ENSEMBLE SIMULATION AND FORECASTING OF SOUTH ASIAN MONSOON by Siraj U1 Islam M.Phil. (Computational Physics), Quaid-e-Azam University, Pakistan, 2006 M.Sc. (Physics), Quaid-e-Azam University, Pakistan, 2003 B.Sc., Punjab University, Pakistan, 2000 DISSERTATION SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY IN NATURAL RESOURCES AND ENVIRONMENTAL STUDIES UNIVERSITY OF NORTHERN BRITISH COLUMBIA April 2015 © Siraj U1 Islam, 2015 UMI Number: 3663190 All rights reserved INFORMATION TO ALL USERS The quality of this reproduction is dependent upon the quality of the copy submitted. In the unlikely event that the author did not send a complete manuscript and there are missing pages, these will be noted. Also, if material had to be removed, a note will indicate the deletion. Di!ss0?t&Ciori P iiblist’Mlg UMI 3663190 Published by ProQuest LLC 2015. Copyright in the Dissertation held by the Author. Microform Edition © ProQuest LLC. All rights reserved. This work is protected against unauthorized copying under Title 17, United States Code. ProQuest LLC 789 East Eisenhower Parkway P.O. Box 1346 Ann Arbor, Ml 48106-1346 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon Abstract This research thesis first examines the ability o f Community Atmosphere Model (CAM) and Community Climate System Model (CCSM) in simulating the South Asian Monsoon (SAM) summer precipitation in a framework o f ensemble. On this basis, the climatic relevant singular vectors (CSVs) perturbation theory is applied to investigate the optimal error growth o f SAM seasonal forecast due to the uncertainties in the Pacific and Indian Oceans. Then, the ensemble prediction o f SAM constructed by CSVs is evaluated, and further compared with one traditional ensemble method. It is found that CAM4 adequately simulated monsoon precipitation, and considerably reduced systematic errors that occurred in its predecessors, although it tends to overestimate monsoon precipitation when compared with observations. In terms of monsoon interannual variability and its teleconnection with sea surface temperature (SST), CAM4 showed modest skill. In the CCSM4 coupled simulations, several aspects of the monsoon simulation are improved, including the cross-variability o f simulated precipitation and SST. A significant improvement is seen in the spatial distribution o f monsoon mean climatology where a tooheavy monsoon precipitation, which occurred in CAM4, is rectified. A detailed investigation of precipitation reduction, using sensitivity experiments, showed that the large systematic cold SST errors in the northern Indian Ocean reduces monsoon precipitation and delays the monsoon onset by weakening local evaporation. The CSV analysis using CAM4 revealed that the SST uncertainties in Indian Ocean can result in much larger error growth o f SAM seasonal forecast than those in the equatorial ii Ph.D. Dissertation: University of Northern British Columbia Pacific Ocean. It is seen that the CSVs error growth rate changes significantly depending on the initial states whereas the CSVs patterns are insensitive to the initial conditions. The CAM4 comparison with CCSM4 coupled model indicated that the CSVs patterns from CAM4 are similar to those from CCSM4 while the error growth rate is lower in CAM4 than in CCSM4. CAM4 ensemble hindcasts, constructed using CSVs method and Time Lag Ensemble (TLE) method, for the period from 2000-2009, showed that the ensemble mean prediction by CSVs has a better skill than both TLE and control run prediction, indicating the merit o f CSV for SAM ensemble forecast. Overall, this thesis research has theoretical significance in understanding the uncertainties of SAM seasonal predictions and practical importance in developing a new ensemble method for SAM seasonal predictions. S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon Table of Contents A bstract................................................................................................................................. ii Table o f C ontents.................................................................................................................. iv List of T ables......................................................................................................................... vii List of Figures....................................................................................................................... viii Glossary.................................................................................................................................. xvii Acknowledgements.............................................................................................................. xix C hapter 1: In tro d u ctio n ..................................................................................................... 1 1.1 Asian M onsoon......................................................................................................... 1 1.2 Monsoon Simulations............................................................................................... 3 1.3 Monsoon Seasonal Forecasts................................................................................... 5 1.3.1 Ensemble Strategies....................................................................................... 7 1.4 Objectives and O utline............................................................................................. 8 C hapter 2: Asian monsoon simulations by Community Climate Models CAM4 and C C S M 4 ..................................................................................................... 12 2.1 Introduction.............................................................................................................. 13 2.2 Models, Experiments and Validations................................................................... 15 2.2.1 M odels............................................................................................................ 15 2.2.2 Experimental D esign..................................................................................... 17 2.2.3 V alidation....................................................................................................... 18 2.3 Simulations by CAM4 and C A M 5.......................................................................... 19 2.3.1 Climatological Mean and Seasonal C y cle.................................................... 19 2.3.2 Monsoon Interannual Variability................................................................. 26 2.3.2.1 Asian Monsoon Indices........................................................................ 26 2.3.2.2 Composite A nalysis............................................................................. 28 2.3.3 Teleconnection o f SAM with ENSO and IOD in CAM Sim ulations 30 iv Ph.D. Dissertation: University of Northern British Columbia 2.4 Coupled Simulation using CSM4............................................................................ 33 2.4.1 Mean Climatology of C C SM 4....................................................................... 33 2.4.2 Teleconnection o f SAM with ENSO in Coupled Simulation...................... 36 2.5 The Contribution o f Air-Sea Interaction to SAM Simulation............................... 39 2.6 Summary and Conclusions...................................................................................... 42 C hapter 3: F urther Investigation of ENSO-Monsoon relationship in climate models: A new validation stra te g y ............................................................... 68 3.1 Introduction.............................................................................................................. 69 3.2 Models Experiments and M ethodology................................................................. 72 3.3 Results and Discussion............................................................................................. 76 3.3.1 Composite A nalysis....................................................................................... 77 3.3.1.1 DJF-only Com posite.............................................................................. 78 3.3.1.2 JJAS-only Com posite............................................................................ 79 3.3.1.3 DJF&JJAS Com posite........................................................................... 80 3.3.2 Temporal Evolution of SST and SAM R ainfall.......................................... 82 3.3.3 Response o f Atmospheric Circulations........................................................ 84 3.3.4 Response o f Humidity and Land-Sea Thermal C ontrast............................ 85 3.3.5 Idealized Experiments................................................................................... 88 3.4 Summary and Conclusions...................................................................................... 89 C hapter 4: Optim al erro r growth of South Asian Monsoon Forecast Associated with the uncertainties in the Sea Surface T e m p e ra tu re .......................... 104 4.1 Introduction.............................................................................................................. 105 4.2 Method and M odels.................................................................................................. 108 4.2.1 Estimation o f Singular Vectors....................................................................... 108 4.2.2 Calculation o f R and Application to the SAM Forecast System ................ 110 4.2.3 Experimental S etups...................................................................................... Ill 4.2.4 Target V ariable............................................................................................... 114 4.3 The Optimal Error due to the SST Uncertainties in Indian and Pacific Oceans .. 114 v S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon 4.3.1 Results during the 2000-2009 Period.......................................................... 115 4.3.2 Results during the 1980-2009 Period........................................................... 118 4.3.3 The Results from the Coupled Model C C SM 4........................................... 120 4.4 Sensitivity Experiments o f CSV to some Param eters........................................... 121 4.4.1 Convergence o f CSV with Number of E O F s.............................................. 121 4.4.2 Convergence o f CSV with Number o f Ensemble M em bers....................... 122 4.4.3 Validity o f C S V .............................................................................................. 122 4.5 SAM Ensemble Forecast using C S V s..................................................................... 123 4.6 Summary and Conclusion....................................................................................... 125 Chapter 5: Summary and Discussion.............................................................................. 143 5.1 Thesis Sum m ary....................................................................................................... 143 5.2 Discussion................................................................................................................ 147 References............................................................................................................................ 150 vi Ph.D. Dissertation: University of Northern British Columbia List of Tables Table 2.1: List o f experiments used in CAM4, CAM5 and CCSM4 validation. 47 Table 2.2: Details o f frequently used Asian monsoon indices. The correlation 49 coefficient o f simulated indices with observations is shown in last two columns. Table 4.1: List o f experiments used for CSV analysis including detail o f perturbation 127 domain, number of perturbation EOFs, number o f ensembles and forecast years. vii S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon List of Figures Figure 1.1: Seasonal (winter and summer) mean observed climatology o f SAM precipitation (mm/day) and 850 mb wind vectors (m/s). Figure 2.1: Seasonal mean (December-February; DJF) precipitation differences (biases) and root means square error (RMSE) from observation (CMAP) for (a, c) CAM4 and (b, d) CAM5. Units are in mm/day. Figure 2.2: Same as Figure 2.1 but for the summer season (June-September; JJAS). Figure 2.3: Seasonal mean (June-September; JJAS) precipitation and 850 mb winds from: (a) CAM4, (b) CAM5 and (c) observations (CMAP/NCEP). Precipitation (shaded) in mm/day and 850 mb winds (vectors) in m/s. Figure 2.4: Time-latitude evolution o f precipitation averaged over (50°-120°E) for: (a) CAM4, (b) CAM5 and (c) observations (CMAP). Precipitation annual cycle area averaged over SAM region is shown in (d). Units are in mm/day. Figure 2.5: Seasonal mean (June-September; JJAS) precipitation differences from observations (CMAP) for: (a) CAM5_BAM (CAMS with Bulk Aerosol Model (BAM) scheme), (b) CAM5_BAM_CAMRT (CAM5 with BAM and CAMRT (radiation) Schemes), (c) CAM5 BAM HB (CAM5 with BAM and HB (Holtslag-Boville, boundary layer turbulence) schemes) and (d) C A M 5 B A M C A M R T H B (CAM5 with BAM, HB and CAMRT schemes). Units are in mm/day. Ph.D. Dissertation: University of Northern British Columbia Figure 2.6: Time series o f Indian summer rainfall (ISR) index for (a) CMAP, (b) 55 CAM4 and (c) CAM5. Details of this index are given in Table 2.1. The time series are normalized and thus unitless. Figure 2.7: Time series o f (a) Webster-Yang (WY) index, (b) Indian Monsoon index 56 (IMI), (c) Western North Pacific monsoon (WNPM) index, and (d) South Asian monsoon (SAMj) index o f observation (CMAP/NCEP, black line), CAM4 (blue line) and CAM5 (red line). Details of all the indices are given in Table 2.1. The time series are normalized and thus unitless. Figure 2.8: Difference between model (CAM4 and CAM5) and observed (NCEP) 57 JJAS mean velocity potential (10'6/s) and corresponding divergent winds (m/s) at (a), (b) 850 and (c), (d) 200 mb. Figure 2.9: June-September (JJAS) anomaly precipitation composites o f weak (1982, 58 1984, 1986, 1987, 1989 and 2002) and strong (1980, 1981, 1983, 1988, 1994, 1996, 1998 and 2007) monsoon years for (a, b) observation (CMAP), (c, d) CAM4 and (e, f) CAM5. Units are in mm/day. Figure 2.10: The linear regression o f observed June-September (JJAS) Nifio3.4 (- 59 5°S-5°N, 120°-170°W) and IOD ((-10°S-10°N, 50°-70°E) - (-10°S-0°, 90°-110°E)) SST indices with June-September (JJAS) observed and simulated precipitation and 850 mb winds, (a, b) observation (CMAP/NCEP), (c, d) CAM4 and (e, f) CAM5. Figure 2.11: Lag-lead correlation o f monthly mean precipitation with (a) Nino3.4 (- 60 ix S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon 5°S-5°N, 120°-170°W) and (b) IOD ((-10°S-10°N, 50°-70°E) - (-10°S-0, 90°110°E)) indices. Area averaged (0°-40°N, 55°-100°E) time series (ISR) o f observed CMAP (solid black line) precipitation, observed All-India Precipitation (AIR - dashed black line) time series and simulated (CAM4, solid blue line and CAM5, solid red line) time series are correlated with observed Nino3.4 and IOD SST indices. The month with negative (positive) sign indicate that SST leads (lags) the ISR with maximum lead o f 12 months (1 year). Month 0 and 12 indicates June whereas month 4 and 8 correspond to February and October (minus sign for previous month). Correlations are calculated using a 5 month sliding window. Figure 2.12: June-September (JJAS) mean differences between simulation and 61 climatology (CMA) for: (a) CAM CLIM (CAM4 climatology run) and (b) CCSM4. Shading corresponds to the precipitation difference in mm/day. Figure 2.13: Same as Figure 2.3 but for CCSM4. Precipitation (shaded) in mm/day 62 and 850 mb winds (vectors) in m/s. Figure 2.14: The variance power spectrum for (a) observed SST (HadSST) and (b) 63 CCSM4. Nino3.4 (-5°-5°N, 120°-170°W) SST index is in a solid red line and IOD ((10°S-10°N, 50°-70°E) - (-10°S-0°, 90°-l 10°E)) index is in a solid blue line. Figure 2.15: The linear regression (Top) o f CCSM4 June-September (JJAS). Nino3.4 (-5°S-5°N, 64 120°-170°W) SST index with June-September (JJAS) simulated precipitation and 850 mb winds. Lag-lead correlation (Bottom) o f CCSM4 monthly mean precipitation with CCSM4 Nino3.4 (-5°S-5°N, 120°-170°W) SST index. Area x Ph.D. Dissertation: University of Northern British Columbia averaged (0°-40°N, 55°-100°E) time series (ISR) o f observed CMAP (solid black line) precipitation and observed All-India Precipitation (AIR - dashed black line) time series are correlated with observed Nino3.4 SST index. Correlations are calculated using a 5 month sliding window. Figure 2.16: Seasonal SST differences between CCSM4 and observation (HadSST) 65 for a thirty year mean climatology during: (a) December-February (DJF), (b) MarchMay (MAM), (c) June-August JJA and (d) September-November SON. The shading corresponds to model SST bias in °C (red for warm SST and blue for cold SST). Figure 2.17: June-September (JJAS) differences o f (a) CAM4 POP (simulation 66 forced with the predicted SST of CCSM4) from observation (CMAP/NCEP) and (b) from CCSM4. Shading corresponds to the difference in mm/day whereas vector represents 850 mb winds in m/s. Figure 2.18: Annual cycle o f observed SST (HadSST, solid black line) and simulated 67 SST (CCSM4, blue dotted line and CCSM4_TR, red dotted line) over (a) Arabian Sea (40°-80°E, 7°-30°N) and (b) Bay o f Bengal (80°-100°E, 7°-30°N). (c) JuneSeptember (JJAS) mean precipitation difference o f CAM4_AS_BoB (CAM4 climatology SST run with SST modification in both Arabian Sea and Bay o f Bengal) from CAM4 CLIM (CAM4 climatology SST run), (d) Same as (c) but for CAM4 AS run (with SST modification in Arabian Sea only). The shading corresponds to the difference in mm/day. Figure 3.1: Lag lead correlation o f monthly mean SAM precipitation with (a) Nino3.4 92 xi S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon (-5°S-5°N, 120°-170°W) and (b) northern Indian Ocean (NIO, 5°-25°N, 60°-100°E) indices. Area averaged (5°-25°N, 60°-100°E) time series (ISM) o f observed CMAP (solid black line) precipitation, CAM4 (solid blue line) and CCSM4 (solid red line) time series are correlated with observed Nino3.4 and northern Indian Ocean SST indices except for CCSM4 case where CCSM4 simulated SST are used. Month with a negative (positive) sign indicate that SST leads (lags) the ISR with maximum lead of 12 months (1 year). Month 0 and 12 indicates June whereas month 4 and 8 correspond to February and October (minus sign for previous month). Correlations are calculated using a 5 monthly sliding window. Figure 3.2: Bimonthly composite anomalies o f observed SST (HadSST) for DJ, FM, AM, JJ, and AS (top to bottom). The 1st, 2nd and 3rd columns represent DJF-only. JJAS only and DJF&JJAS types respectively (see text for detail). Shading corresponds to the values in C. Figure 3.3: Bimonthly composite anomalies o f precipitation for DJ, FM, AM, JJ, and AS in DJF-only type (top to bottom). The 1st, 2nd and 3rd columns represent observed (CMAP), CAM4 and CCSM4 respectively. Shading corresponds to the values in mm/day while the contour shows their significance. Figure 3.4: Same as Figure 3.3 but for the JJAS only type. Figure 3.5: Same as Figure 3.3 but for the DJF&JJAS type. Figure 3.6: Same as Figure 3.2 but for CCSM4 simulated SST. Ph.D. Dissertation: University of Northern British Columbia Figure 3.7: Composite JJAS rainfall anomalies (mm/day) for the DJF-only type, 98 JJAS-only type and DJF&JJAS type based on observed (CMAP) and simulated (CAM4 and CCSM4) precipitation. Figure 3.8: Area averaged composite o f normalized 3-month running mean anomalies 99 o f Nino3.4 SST (green curves), northern Indian Ocean SST (NIO, red curves), CAM4 or CCSM4 precipitation (blue curves) and CMAP precipitation (IMR, black curves) in (a) DJF-only type, (b) JJAS-only type, and (c) DJF&JJAS type. SST curves in left column are from observation while right column SST curves are from CCSM4 simulation. Figure 3.9: Observed (top), CAM4 (center) and CCSM4 (bottom) JJAS mean 100 composite anomalies o f velocity potential and the corresponding divergent winds at 850 mb height (left column) and at 200 mb height (right column) for the JJAS-only and DJF&JJAS types. Shading corresponds to the velocity potential in 106 s'1 and vectors in m/s. Figure 3.10: Composites of normalized 3-month running mean anomalies o f surface 101 (2 m) air humidity over the region o f 0°-10°N and 50°-80°E (solid curves) and 500200 mb thickness difference (10 m) between area 20°-40°N, 50°-100°E and 08°20°N, 50°-100°E in (a) DJF-only type, (b) JJAS-only type, and (c) DJF&JJAS type. Left column is for CAM4 and observation while right column shows CCSM4 simulated curves. Figure 3.11: Composite JJAS rainfall anomalies (mm/day) and 850 mb winds for the 102 xiii S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon DJF&JJAS type based on C A M 4 E N S O I O and C A M E N S O experiments. See text for detail o f the experiments. Figure 3.12: JJAS mean composite anomalies o f velocity potential and the 103 corresponding divergent winds at 850 mb height (top left) and at 200 mb (top right) for DJF&JJAS type based on CAM4 ENSO IO and CAM ENSO experiments, (bottom) Composite o f normalized 3-month running mean anomalies for C A M 4 E N S O I O and CAM4 ENS O experiments for DJF&JJAS type. Figure 4.1: First three leading correlation EOFs patterns o f the June SST over (a) 128 Indian Ocean and (b) Pacific Ocean. These patterns are calculated from monthly observed SST for time period 1980 to 2009, and are added as perturbation in the instantaneous initial states o f persistent SST. Figure 4.2: SST anomalies for May o f each forecast year. Each anomaly pattern is 129 calculated based on the mean SST climatology of 1980-2009 and is kept persistent throughout the forecast. Units are in °C. Figure 4.3: Mean climatology o f June, July, August and September for Outgoing 130 Longwave Radiation (OLR). NCEP OLR is shown in left column and CAM4 control run is in right column. Mean is calculated over the period 1980 to 2009. The shading corresponds to OLR values in W/m2. Figure 4.4: The optimal leading Climatology relevant Singular Vectors (CSVs) o f SST and corresponding OLR final patterns (FPs) of, optimized for 4 months (lead time 3) over the Indian Ocean domain. Each individual CSV and FP is shown starting from 131 Ph.D. Dissertation: University of Northern British Columbia year 2000 to 2009. The SST units are °C and OLR is in W/m2 Figure 4.5: The first leading EOF patterns o f the CSVs (left) and FPs (right) 132 computed over 10 individual CSV patterns for all the 4 lead times. CSVs are extracted over Indian Ocean. The SST units are in °C and OLR is in W/m2. Figure 4.6: Same as Figure 4.5 but for CSVs calculated over Pacific Ocean and 133 corresponding FPs over SAM region. The SST units are in °C and OLR is in W/m2. Figure 4.7: Same as Figure 4.6 but for CSV calculated over combined Indian-Pacific 134 Ocean. Only FPs are shown here. Figure 4.8: Annual variation o f singular values averaged over all the lead times for (a) 135 Indian Ocean (b) Pacific Ocean and (c) 1030 experiments, (d) The lead time variation o f singular values over all the three cases (see text for detail). Figure 4.9: Same as Figure 4.5 but for CSVs and FPs calculated over Indian Ocean 136 using 1030 experiment. 30 years CSVs and FPs are used for EOF analysis. The initial conditions for 1030 experiment are generated using CAM4 prescribed SST run. Figure 4.10: The spatial correlation for (a) Indian Ocean, (b) Pacific Ocean and (c) 137 1030 experiments. It is computed between the first EOF mode (obtained by all CSVs) and each individual CSV pattern, same as in Tang and Deng (2011). Figure 4.11: (Spatial) The optimal leading SST CSVs and OLR FPs for lead time 0, 138 1, 2 and 3 obtained using CCSM4 model. The domain o f perturbation is Indian Ocean. The SST units are °C and OLR is in W/m2. (Line) Lead time variation o f CCSM4 xv S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon singular values over Indian Ocean. Figure 4.12: Spatial patterns of the leading CSV (left column) based on the increasing 139 number o f EOFs used in the perturbation. 10 leading SST correlation EOFs are used by increasing EOFs from 3 to 10 in each CSV and FPs calculation. The optimization time is 4 months. Right column shows associated FPs. The SST units are in °C and OLR is in W/m2. Figure 4.13: The convergence o f the leading singular values with increasing number 140 o f ensemble members, optimized for 4 month time interval. Each color line represent individual year. In each year, 20 singular values are calculated by increasing the ensemble size from 1 to 20 in the CSV method. Figure 4.14: Leading EOF o f OLR FPs, at lead time 3, (a) calculated using CSV as 141 perturbation and (b) estimated by applying linear propagator R to CSV i.e. by linear approximation. For (a), the difference o f perturbed and control forecast is calculated each year and leading EOF is extracted using all the differences. In (b), EOF pattern is calculated using estimated FPs from linear approximation for all the 10 years (20002009). Figure 4.15: Root mean square error (RMSE) o f the SAM ensemble forecast 142 generated using CSV perturbation, (a) RMSE calculated over 10 years (2000-2009 using observed initial conditions) whereas in (b) RMSE is for 30 years forecast (19802009 using CAM4 control initial conditions). Error bar is drawn using bootstrap method. xvi Ph.D. Dissertation: University of Northern British Columbia Glossary AGCM: Atmospheric General Circulation Model APCC: Asian Pacific Economic Cooperation (APEC) Climate Center AIR: All-India Rainfall CGCM : Coupled General Circulation Model CAM: Atmospheric General Circulation Model CliPAS: Climate Prediction and its Application to Society CCSM: Community Climate System Model CSV: Climatology Relevant Singular Vector CLM : Community Land Model CICE: Community Ice Code C PC : C limate Predict ion Center CMAP: CPC Merged Analysis o f Precipitation CFCAS: Canadian Foundation for Climate and Atmospheric Sciences DEM ETER: Development of a European Multimodel Ensemble System for Seasonal to Interannual Prediction D JF: December, January and February FP: Final Pattern EOF: Empirical Orthogonal Function EAM: East Asian Monsoon ENSO: El Nino/La Nina and the Southern Oscillation HPC: High Performance Computing Lab IM : Indian Monsoon S Islam: Ensemble Simulation and Forecasting o f South Asian Monsoon IOD: Indian Ocean Dipole ISR: Indian Summer Rainfall ITCZ: Intertropical Convergence Zone IPCC: Intergovernmental Panel on Climate Change GCM: General Circulation Model JJA S: June, July, August and September POP: Parallel Ocean Program PROVOST: Prediction o f climate variations on seasonal to interannual timescales MAM: Modal aerosol scheme NSERC: Natural Sciences and Engineering Research Council NCEP: National Centers for Environmental Prediction NRES: Natural Resource and Environmental Science NCAR: National Center for Atmospheric Research NCL: NCAR Command Language RMSE: Root mean square error SAM: South Asian Monsoon SAMi: South Asian Monsoon index SST: Sea Surface Temperature SVD: Singular Vector Decomposition Westgrid: Western Canada Research Grid WNPM: Western North Pacific Monsoon WY: Webster-Yang Ph.D. Dissertation: University of Northern British Columbia Acknowledgements I want to express immense gratitude to my supervisor Dr. Youmin Tang and co-supervisor Dr. Peter Jackson for their professional guidance and encouragement throughout this research work. Their keen interest, healthy discussion and enlightening criticism enabled me to complete my dissertation. Without their kind and valuable help, it would have almost been impossible for me to accomplish my research goals. I am deeply grateful to my committee members, Dr. Bill Merryfield, Dr. Erik Jensen and Dr. Jianbin Li for their valuable suggestions and support during my thesis research. I wish to say special thanks to the faculty o f Natural Resource and Environmental Science (NRES) program for providing an excellent research environment in the department. I am also grateful to the anonymous reviewers for their comments and suggestions to improve my research manuscripts. I extend my thanks to UNBC High Performance Computing (HPC) lab and Western Canada Research Grid (Westgrid) for providing supercomputing facilities for my research work. I am thankful to National Center for Atmospheric Research (NCAR) for providing models, input datasets and technical assistance. Thanks to NCAR Command Language (NCL) developers for helping me to unlock the power o f data processing and graphical analysis. Heartiest thanks to my lab members Dejian Yang, Manoj Nambiar, Waqar Younas, Yanjie Cheng and Zewang Deng for their helpful comments, suggestions, and discussions. Thanks are due to my friends in the Prince George community for being terrific fellows through the years. They never failed to show concern when things got stressful. S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon Finally I wish to record my deepest obligation to my parents, brothers, sister (late), and my wife (Ambreen) for their prayers, encouragement and love which always keep me on the right track. The debt I owe to my family has no measures. This research work is supported by Canadian Foundation fo r Climate and Atmospheric Sciences (CFCAS) under research grant GR-7027 and Natural Sciences and Engineering Research Council (NSERC) Discovery grant to Dr. Youmin Tang. xx Chapter 1 Introduction 1.1 Asian Monsoon The monsoon is one of the most dynamic and prominent phenomena o f the climate system and has a large effect on weather and climate anomalies at both local and global scales. The monsoon system is caused by the seasonal reversal o f winds due to differential heating between land and ocean, and results in seasonally changing heavy precipitation patterns. The dominant monsoon systems around the globe are the Asian, Australian, African and American monsoons (Webster et al. 1998). Among these various monsoon systems, the Asian summer monsoon comprising the East Asian Monsoon (EAM) and the South Asian Monsoon (SAM) (Lau and Li 1984) receives the heaviest seasonal precipitation during the summer and has a major impact on global atmospheric circulations. The SAM region includes parts o f the Arabian Sea, the Indian subcontinent and the Bay o f Bengal (Goswami et al. 1999). The Indian subcontinent is characterized by complex topographical features, such as the Himalayas (in the north and north east) and the Western Ghats (along the western coast o f India). Both of these extensive mountain ranges strongly influence the SAM circulations. Figure 1.1 shows the seasonal mean precipitation and 850 mb winds climatology over South Asian region. In winter (December, January and February, DJF), surface winds originating from South Asia sweep across the equator to meet the south Indian Ocean trades winds in the southern hemisphere. The precipitation in winter is therefore mostly confined in the equatorial Indian Ocean while the Indian subcontinent exhibits dry conditions. During the summer season in the months o f June, July, August and September (JJAS), the winds blow 1 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon north from the Indian Ocean south o f the Indian subcontinent and produce heavy rains contributing around 75% o f SAM annual precipitation (Hastenrath and Polzin, 2004). The seasonal winds reversal in the SAM region is associated with the surface temperature contrast between the Indian subcontinent and the Indian Ocean, caused by the different responses of land and sea to solar heating during April and May. The onset o f the summer monsoon involves the formation o f a low pressure region over the SAM region that is called the monsoon trough. The monsoon trough is part o f the Inter-tropical Convergence Zone (ITCZ) between the wind patterns o f the southern and northern hemispheres. The SAM precipitation has a very strong temporal and spatial variation due to the interaction between regional topography, atmospheric circulation and slowly varying sea surface temperature (SST) anomalies. Many studies have shown that the significant SAM variability is linked to El Nino/La Nina and the Southern Oscillation (ENSO), a coupled atmosphereocean interaction in the tropical Pacific Ocean (e.g. Sikka 1980; Rasmusson and Carpenter 1983 and Shukla 1987). ENSO influences the SAM regions either through teleconnections or ocean current anomalies, or both. In addition to ENSO, SST in the Indian Ocean also modulates SAM variability by means o f the Indian Ocean Dipole (IOD; Saji et al. 1999; Ashok et al. 2001). Other large scale boundary conditions such as soil moisture (Webster et al. 1998) and snow cover (Bamzai and Shukla 1999; Kripalani and Kulkarni 1999) also modify the SAM dynamics, intensity and onset. Along with large-scale influences, it is seen that any variation in SAM is partially controlled by the atmospheric internal dynamics and random fluctuations (Krishnamurthy and Shukla 2000; Goswami and Ajaya 2001; Saha et al. 2011). The combination o f these interactions result in significant interannual or intraseasonal variation o f moisture availability over the SAM region either in the form o f extreme dry or 2 Ph.D. Dissertation: University o f Northern British Columbia wet conditions or delay of SAM onset. The relative role o f these interactions makes SAM a very complex dynamical system. Being a major convective system in the northern hemisphere, the fluctuations in SAM intensity affects the lives o f many people due to the agrarian basis o f their society. The fluctuations are often associated with delay in the onset o f SAM, floods, droughts, and other climatic extremes (Malik et al. 2010) and significantly disturb the water budget, agricultural practice and economy o f the region causing enormous socio-economic impacts. In view o f the dependence o f water availability and agriculture on monsoon precipitation, study o f monsoon dynamics and its forecasting becomes an issue o f immense importance. A detailed understanding of the role o f different external forcing and other mechanisms that modulate SAM precipitation is essential for skillful SAM prediction as well as for societal and economical needs o f the South Asian region. An accurate seasonal forecast can help planners mitigate the adverse impact in case o f monsoon failure or amplification. Indeed, many efforts have been made to explain the process, variability and teleconnection o f monsoon, but a lull understanding of the phenomenon and its predictability is still being developed. 1.2 Monsoon Simulations The past few decades have witnessed steady improvements in the skill o f global climate models (GCMs) in simulating and forecasting climatic systems. These improvements have been largely driven by improved numerical schemes, increased resolution, advanced computational power and enhanced quality o f input observational data. The overall framework o f the dynamical simulations and predictions is refined based on comprehensive model tests and evaluation o f their outputs on various time scales. The assessment o f a 3 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon GCM’s simulation determines the fidelity with which models simulate different processes and predict climate. In the case of the SAM, conceptual models or fully coupled climate models are widely used either to understand the mechanisms involved in SAM development or to improve its prediction. Using these models, many studies have investigated the SAM dynamics focusing on different boundary condition such as SST (Webster et al. 1998; Ju and Slingo 1995; Wang et al. 2003; Loschnigg and Webster 2000; Lau and Nath 2000), snow (Shen et al. 1998; Becker et al. 2000; Jhun and Lee 2004; Wang et al. 2008), soil moisture (Webster et al. 1998; Walker and Rowntree, 1977; Shukla and Mintz 1982; Xue et al. 2006) and orography (Rodwell and Hoskins 1995; Wu et al. 2007). These studies, along with many others, have been beneficial in identifying the essential mechanisms and processes for SAM, and in improving its predictions. While the current versions o f GCMs can provide credible quantitative estimates o f climate and can reasonably well simulate the mean climate anomalies o f many variables (Randall et al. 2007, Liang et al. 2009, Lee et al. 2010 and many more), they still show significant discrepancies in simulating relatively local-scale systems such as monsoons (Kang et al. 2004; Wang et al. 2004; Covey et al. 2003; Meehl et al. 2005). The SAM variability at seasonal and interannual time scales and its relationship with local and remote SST anomalies are not well reproduced in Atmospheric General Circulation Models (AGCMs) due to large systematic errors (Kang and Shukla 2006). Although the Coupled General Circulation Model’s (CGCMs) mean state could generally lead to a realistic simulation o f monsoon teleconnection (Lau and Nath 2000; Turner et al. 2005), a comparison between uncoupled and coupled model simulations suggested that the SST biases o f coupled models 4 Ph.D. Dissertation: University of Northern British Columbia partially offset the benefits o f an active air-sea coupling (Cherchi and Navarra 2007, Islam et al. 2013). It is also seen that climate model bias in the mean state and in the seasonal cycle could degrade the seasonal and interannual predictability skill (Gadgil and Sajani 1998; Sperber et al. 2001). It is therefore important to identify systematic errors in monsoon simulations if the models are to be used for prediction. One o f the key motivations o f this thesis is therefore to study the systematic errors in uncoupled and coupled models for their simulations o f the monsoon system. 1.3 Monsoon Seasonal Forecasts The tremendous advances in understanding and modeling climate variability and predictability have made seasonal prediction an active area o f research. As a result, seasonal climate prediction and the related uncertainties using multiple climate models has become operational (e.g. Lee et al. 2009; Palmer et al. 2004). In current dynamical models, the skill of seasonal predictions for some large scale climate systems is gradually approaching the practical limit o f predictability that arise due to the nonlinear scale interactions inherent in the atmosphere (Lorenz 1963) coupled with an incomplete knowledge o f the initial state (i.e. chaos) (Kang and Shukla, 2006). These nonlinear interactions impose a theoretical limit o f predictability which can be defined as the performance o f a perfect model with perfect initial conditions. The theoretical (inherent) and practical predictability limits have been estimated in many studies to investigate the potentials for further improvement in dynamical seasonal predictions (e.g. Younas and Tang, 2013). Chamey and Shukla (1981) initiated research on SAM seasonal forecasting by reporting its potential predictability due to the slowly varying boundary conditions such as SST. They 5 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon documented the potential to predict SAM beyond the limit o f deterministic predictability based on its close interaction with tropical SST. Therefore the practical value o f SAM prediction depends on the accurate simulation o f forcing’s signal (mainly from tropical SST) and its strength compared to the strength o f unpredictable variability inherent in the atmosphere. Substantial efforts have been devoted to the study o f monsoon predictability either by dynamical models (e.g. Rajeevan et al. 2011; DelSole and Shukla, 2012) or by statistical models (e.g. Delsole and Shukla, 2002). While many different methodologies are used to improve the monsoon forecast skill, the overall skill o f dynamical SAM forecast is considerably lower than the model forecast skill over the tropical central Pacific region (Kang et al. 2002; Wang et al. 2004). Different initiatives and projects such as “Asia-Pacific Economic Cooperation Climate Center/Climate Prediction and its Application to Society (APCC/CliPAS, Wang et al. 2009)”, “Development o f a European Multimodel Ensemble System for Seasonal to Interannual Prediction (DEMETER, Palmer et al. 2004)” and “Prediction o f climate variations on seasonal to interannual timescales (PROVOST, DoblasReyes et al. 2000)”, focusing seasonal dynamical forecasting, have been performed using both AGCMs (Tier-2) and CGCMs (Tier-1) models. The results from these projects have shown that the monsoon prediction skill mainly depends on simulation o f accurate ENSO signal in models. The skill o f dynamical model is limited mainly by inherent nonlinear characteristics of the atmosphere and the inaccurate simulation o f ENSO signal (Kang et al. 2004). This means that the predictability o f the SAM depends on the external forcing and the relative contributions from the internal dynamics o f the monsoon system. 6 Ph.D. Dissertation: University of Northern British Columbia 1.3.1 Ensemble Strategies Practical forecast system requires a careful distinction between model errors and uncertainties in initial conditions (states). Both o f these are not realistically distinguishable due to the reason that the estimation o f the initial conditions involves a forecast model and thus model errors affects initial condition errors (Leutbecher and Palmer, 2008). Even if the models become perfect, the error due to the uncertainty involved in its initial conditions can degrade the forecast skill. The uncertainties present in initial states may grow due to the chaotic nature o f the evolution equations o f the dynamical system, thereby significantly decreasing the forecast skill. One practical approach to sample these uncertainties is to perturb the initial conditions and run the model with each set o f perturbed initial conditions. The resulting group o f model simulations is known as an ensemble which improves the forecasting skill by reducing the nonlinear error growth and averaging out unpredictable components. Many studies have shown that ensemble prediction can decrease the prediction error that originates from the uncertainty in initial conditions (Molteni and Palmer (1993) and Buizza et al. 1998; Yoo and Kang 2005). Generating an ensemble in a forecast system can improve the forecast skill by diminishing the initial uncertainties. The crucial issue is to design a reliable ensemble forecast strategy that should include the major uncertainties o f forecast initial states. Many operational centers still face the difficulty that the ensemble perturbations from a single model have limited error growth at early forecast lead times as compared to the amplitude o f the mean error (Palmer et al. 2004; Saha et al. 2006, Kug et al. 2011). This implies that their initial perturbations used to generate ensemble may not be optimal. 7 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon Thus, it is necessary to treat the uncertainty o f the initial condition in an optimal way so that error growth in the forecast system is realistic. It has been a challenge to construct optimal ensemble prediction at seasonal time scales for uncoupled or coupled GCM, which is another motivation for this thesis research. 1.4 Objectives and Outline Correct representation o f the monsoon system in climate models and its skillful seasonal prediction are important for different socioeconomic sectors o f South Asia such as water resources, agriculture and other climate-sensitive sectors. This thesis therefore addresses the SAM simulations and ensemble forecasts using state o f the art climate models. The long standing objective o f this thesis research is to (i) investigate models’ capability in predicting SAM variability and in characterizing its teleconnections to SST in Indian and Pacific Oceans and (ii) to improve SAM seasonal forecast skill by sampling initial uncertainties including the analysis o f the factors and mechanisms that yield these uncertainties. The specific objectives o f this research are: i) Evaluate models for their performance in reproducing SAM precipitation with an emphasis on the mean climatology, seasonal and interannual variability in an AGCM and a CGCM. ii) Analyze SAM variability and its relationship with SST anomalies o f Indian and Pacific Oceans in the AGCM and CGCM, including a detailed comparison o f SAM response to the prescribed SST in AGCM and to the simulated SST in CGCM. The impact o f SST bias on the SAM simulation. 8 Ph.D. Dissertation: University of Northern British Columbia iii) Implement and analyze the optimum perturbation growth o f SAM seasonal prediction in AGCM and CGCM. Discuss the main characteristics o f the leading optimum perturbation patterns, corresponding final patterns, and perturbation growth rate over the SAM region. Investigate the sensitivity o f optimum perturbation and optimal growth to various parameters. iv) Use the optimum perturbation to generate ensemble SAM forecasts. Compare the forecast skill with that o f control forecast and traditional ensemble forecast. Investigate the practicality o f the optimum ensemble method in maximizing the benefit o f dynamical forecasts. To accomplish the above objectives, this thesis is organised into five chapters. Except Chapter 1, all chapters independently address one or more objectives described above. Chapter 2 is devoted to the detailed investigation of climate models performance in simulating SAM. Both coupled and uncoupled models are validated by means o f different analysis and sensitivity experiments. Chapter 3 extends the analysis presented in Chapter 2, by further investigating the teleconnections o f SAM to SST o f the equatorial Pacific Ocean. The philosophy, design and implementation o f optimum perturbation theory are discussed in Chapter 4 followed by the construction o f ensemble SAM forecast. Chapter 5 summarizes and concludes the overall findings o f the thesis. This research has theoretical components and practical significance to seasonal forecast research. The results from this thesis may leverage existing activities and resources to achieve better and skillful SAM seasonal forecasting capability. Indeed, a good perturbation method for ensemble generation can effectively improve sub-seasonal to seasonal forecast o f SAM 9 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon which could have profound impacts on agricultural planning, water resource management and other socio-economic activities o f the region. 10 Ph.D. Dissertation: University of Northern British Columbia Rainfall and 850mb Winds 60E (Winter, DJF) 90E 120E 150E (Summer, JJAS) 60E 2 4 90E 6 8 10 120E 12 14 16 150E 18 20 Figure 1.1: Seasonal (winter and summer) mean observed climatology o f SAM precipitation (mm/day) and 850 mb wind vectors (m/s). 11 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon Chapter 2 Asian monsoon simulations by Community Climate Models CAM4 and CCSM4 Islam S, Y Tang and P Jackson (2013), Asian monsoon simulations by Community Climate Models CAM4 and CCSM4, Climate Dynamics, DOI 10.1007/s00382-013-1752-6 Published version is available at: http://link.springer.com/article/10.1007%2FsQ0382-013-1752-6 This paper is reformatted to serve as Chapter 2 in the thesis. 12 Ph.D. Dissertation: University of Northern British Columbia 2.1 Introduction Climate models have been significantly improved in simulating the mean global climate (Randall et al. 2007) and in predicting climate anomalies at the seasonal time scale (Liang et al. 2009, Lee et al. 2010; Kang et al. 2002; Kang and Shukla 2006; Wang et al. 2004). These models are fairly good at simulating the average atmospheric state and large scale patterns, but poorer at simulating relatively small and local atmospheric systems such as the monsoon. In some o f the studies such as Kang et al. 2004 and Wang et al. 2004, it has been seen that even when forced with observed SSTs, GCM performance over the SAM region is not satisfactory and presents large systematic biases. Even in CGCMs, which are believed to simulate the most realistic physical processes, there are notable biases in simulation o f the mean climate and its variability (Covey et al. 2003; Meehl et al. 2005). These discrepancies include a Pacific cold bias, a double ITCZ, and a westward shift o f ENSO variability (AchutaRao and Sperber 2006; Covey et al. 2000; Joseph and Nigam 2006). Gimeno et al. (2010) have shown that the northern Indian Ocean, particularly the Arabian Sea, is an important moisture source for SAM and any changes in Indian Ocean SSTs affect monsoon precipitation by altering the amount o f moisture available for transport towards South Asia. Furthermore, due to the strong air-sea coupling over the Indian Ocean, any variation in the strength o f the SAM precipitation influences the SST variation which significantly complicates the detection o f monsoon variability related to other changes in the lower boundary o f the atmosphere in the coupled model. It has been a challenging issue to correctly simulate the monsoon variability at seasonal and interannual time scales (Annamalai et al. 2007; Dai 2006; Kripalani et al. 2007; Lin 2007; Waliser et al. 2007) and the relationship between SST anomalies and SAM precipitation 13 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon variability (Annamalai and Liu 2005; Meehl and Arblaster 2002; Shukla and Paolino 1983; Rasmusson and Carpenter 1983). The link between the SAM precipitation and ENSO has been well documented in both observations and modeling. For example, it has been reported that the warm phase (El Nino) is associated with weakening o f the Indian monsoon and an overall reduction in SAM summer precipitation, while the cold phase (La Nina) is associated with the strengthening of the Indian monsoon and an enhancement o f SAM summer precipitation (Kanamitsu and Krishnamurti, 1978; Krishnamurti et al. 1989; Palmer et al. 1992; Pant and Parthasarathy, 1981; Rasmusson and Carpenter, 1983; Shukla and Paolino, 1983; Shukla and Mooley, 1987; Sikka, 1999). Meehl et al. 2012, described SAM as a fully coupled air-sea-land system which can be better reproduced by air-sea coupled models. Many other studies also reported that coupled models perform better than atmosphere-only GCMs in simulating the SAM (e.g. Kumar et al. 2005 and Wang et al. 2005). As discussed above, an intensive research effort has been made to improve simulation o f monsoon systems by climate models and significant progress has been made in recent years. Among these models, community climate models developed at the US National Center for Atmospheric Research (NCAR) have played an important role in monsoon research due to their complete physical dynamics and easy implementation. In this chapter we evaluate NCAR climate models, i.e., the Community Atmosphere Model version 4 and version 5 (CAM4 and CAM5) and the Community Climate System Model version 4 (CCSM4). The important factors affecting monsoon-SST relationships, such as air-sea coupling and SST bias, need to be studied in detail to determine the strengths and weaknesses o f these models. A systematic evaluation is also important if these models are to be used for seasonal prediction or climate change studies. In this chapter, we therefore explore in detail the 14 Ph.D. Dissertation: University of Northern British Columbia strengths and limitations of CAM4, CAM5 and CCSM4 in simulating SAM precipitation with an emphasis on the mean climate, seasonal and interannual variability and the relationship between SAM and SST (local and remote). Our focus is placed on: 1) the SAM interannual variability when simulations are forced with observed SST; and 2) the role o f airsea coupling and the impact o f SST bias in simulating the SAM. For the latter (2), using sensitivity experiments, we also examine the effect o f northern Indian Ocean SST bias on SAM. This chapter is organized as follows. Section 2.2 describes models, data and experiments. Section 2.3 investigates the mean climatology, annual cycle and evaluates simulated monsoon interannual variability in CAM4 and CAM5. Section 2.4 highlights and compares the coupled simulations o f CAM4 (CCSM4) in terms o f the mean climatology and SAM-SST relationship as well as the effect and importance o f air-sea coupling over the SAM region. To address the effect o f CCSM4 SST bias on the SAM precipitation, section 2.5 explores results o f sensitivity experiments followed by summary and conclusions in section 2.6. 2.2 Models, Experiments and Validations 2.2.1 Models Simulations are performed using the CAM4, CAM5 and CCSM4 models. These models are the latest in a succession of AGCMs and CGCMs that have been made widely available to the scientific community from NCAR. For the convenience o f the reader we briefly introduce each model, but refer interested readers to the cited references for full details o f each model. The models description discussed below is mostly adopted from the cited reference. 15 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon CAM4 (Neale et al. 2010a) is developed from CAM3 (Collins et al. 2006a) with modifications to the deep convection (Neale et al. 2008) and the polar cloud fraction in extremely cold conditions parameterization schemes (Vavrus et al. 2008). It uses an updated convection parameterization scheme (Neale et al. 2008; Richter and Rasch 2008). This model can be used with three different dynamic schemes (an Eulerian spectral scheme, a semiLagrangian scheme and a finite volume scheme) along with different resolution settings. CAM5 (Neale et al. 2010b) is modified significantly compared to CAM4, with a range o f improvements in the representation o f physical processes. It includes a new shallow convection scheme (Park and Bretheron, 2009), a stratiform cloud microphysical scheme (Morrison and Gettelman 2008), an updated radiation scheme (Lacono et al. 2008) and 3mode modal aerosol scheme (MAM3) (Liu et al. 2011). The CCSM4 (Gent et al. 2011) coupled model descended from its predecessors, CCSM3 (Collins et al. 2006b) and CCSM2 (Kiehl and Gent 2004). It contains a new coupler that exchanges fluxes and state information among all the embedded models. These embedded models are the CAM4 atmospheric model, the Community Land Model (CLM4), the Los Alamos Parallel Ocean Program ocean model version 2.2 (POP 2.2) (Smith and Gent 2002) and the Community Ice Code version 4 (CICE4) sea ice model (Hunke and Lipscomb, 2008). The CLM4 model operates on the same grids as the CAM4 model whereas CICE4 uses the same horizontal grid as POP 2.2, which has a displaced dipole grid (Smith and Kortas, 1995). 16 Ph.D. Dissertation: University of Northern British Columbia 2.2.2 Experimental Design A series o f experiments are performed to achieve three goals: i) exploring and comparing the ability o f CAM4, CAM5 and CCSM4 in simulating the SAM; ii) evaluating the contribution o f air-sea coupling to the simulation and iii) investigating the effect o f SST bias on SAM precipitation. These experiments can be generally categorized as below: 1) Control runs: Thirty-two years (1978 to 2008) o f uncoupled simulations are performed using the CAM4 and CAM5 atmospheric models forced with observed prescribed SST (HadSST, Reynolds et al. 2002) and sea ice data. Both models share the same 1.9° x 2.5° horizontal resolution using the finite volume dynamical core with 26 (in CAM4) and 30 (in CAM5 1 ) vertical levels using a hybrid terrain-following coordinate system. Higher resolution simulation of CAM4 and CAM5 models are also performed using 0.9° x 1.25° horizontal grids. 2) Climatology run: CAM4 is also run forced with the climatological (based on the observations from 1982-2001) seasonal cycle o f SST and sea ice for thirty years. This is referred to as CAM4 CLIM. 3) Coupled run: In the case o f the CCSM4 coupled experiment, a 100-year coupled integration is performed using present day climatological forcing. The output o f the last 30 years o f this coupled simulation is used2 . In this simulation, the horizontal resolution o f 1.9° x 2.5° and finite volume grids in both the atmospheric and land model is used, whereas the 1 CAM5 is run for different set of schemes. Standard run (control) of CAM5 has 30 vertical levels with all the default setting. The remaining set of CAM5 runs will be denoted with their particular name throughout the text. 2 Subject to the computational conditions, the spin-up run was carried out 70 years, which basically allows atmospheric states to reach equilibrium. 17 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon ocean and ice models share the same 1° x 1° resolution with a displaced pole grid. To facilitate comparison, observational datasets are interpolated to the resolution o f model grids. 4) Sensitivity runs: Idealized experiments using different boundary forcings are performed with CAM4 and CAM5. Details o f these experiments are given in the relevant sections. Even a realistic model always contains random components and uncertainties such as those in boundary forcing or in initial conditions. To alleviate the inpact o f these random components and obtain a deterministic response o f the model behavior to forcing (such as SST), an ensemble strategy is used for the above experiments except for the coupled run. For the control run, ensembles are constructed through perturbing the initial conditions, which allows us to separate the “ SST-forced” (or external) response (Rowell et al. 1995). The perturbation of the initial conditions is performed by using the initial conditions lagged in time. For the climatology run, the same method is used to construct the ensembles. A detailed summary o f all the experiments and the ensemble runs is given in Table 2.1. All simulation results from the control and climatology runs used for validation and presented in the next sections are the ensemble mean, unless otherwise indicated. 2.2.3 Validation The following validation steps are used to examine the performance o f the models in simulating the SAM monsoon: i) the simulated SAM precipitation and winds by the CAM4 and CAM5 control runs are compared against the observed counterparts in terms o f climatology, interannual variability, and the relationship to SST. ii) The climatological means from the CAM4 runs are compared against those from CCSM4 to explore the effects o f air- 18 Ph.D. Dissertation: University of Northern British Columbia sea coupling on the SAM simulation, iii) To explore the impact o f SST bias on monsoon simulations, CAM4 forced with a modified SST climatology that contain SST bias, is compared with that forced with the observed SST climatology (CAM4_CLIM). The metrics and methods used to evaluate the simulations include mean bias, Root Means Square Error (RMSE), variance, correlation and regression analysis. In all the uncoupled simulations, the first year o f the integration output is discarded as a spin-up time, which is considered sufficient for atmospheric-only simulations. In the case o f coupled runs, the first seventy years are discarded as the ocean model needs more time for equilibration. Observed precipitation data from the Climate Prediction Center (CPC) Merged Analysis o f Precipitation (CMAP; Xie and Arkin 1997), on a 2.5° x 2.5° grid, is used for validation o f precipitation. The All-India Rainfall time series (AIR; Parthasarathy et al. 1995), which is a combination o f 306 uniformly distributed station measurements, is also used in the analysis. National Centers for Environmental Prediction (NCEP; Kistler et al. 2001) reanalysis data, also on a 2.5° x 2.5° grid, are used to validate winds. Observed SST (HadSST) data (Reynolds et al. 2002) are used for coupled model SST validation. 2.3 Simulations by CAM4 and CAM5 We first examine ensemble mean simulations o f CAM4 and CAM5 forced with prescribed observed SST and sea ice data. Before focusing on the SAM, we evaluate both models over the tropical region. 2.3.1 Climatological Mean and Seasonal Cycle The distribution o f precipitation bias and root mean square error (RMSE) is shown in Figures 2.1 and 2.2 for the winter (December-February, DJF) and summer (June-September, JJAS) S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon seasons. The whole tropical region is shown to assess overall model differences. Bias and RMSE are calculated by comparing model output data with observations (CMAP) for the time period 1979-2008. The model bias for winter (DJF) precipitation simulations are shown in Figures 2. la and 2.lb for CAM4 and CAM5 respectively and their corresponding RMSE are shown in Figures 2.1c and 2.Id. Significant large-scale spatial biases (Figures 2.1a and 2.1b) over the equatorial Indian Ocean and western Pacific Ocean, along with many regional biases, are found in the CAM4 and CAM5 simulations. The magnitude of these biases is higher in CAM4 whereas CAM5, to a large extent, significantly rectifies them in its simulation. Major improvements in the CAM5 simulation occur over the South African land areas where the precipitation overestimation seen in the CAM4 simulation is diminished. The RMSE patterns (Figure 2.1c and 2.Id) further highlight the improved winter precipitation in the CAM5 simulation. In summer, when the precipitation activity over the northern hemisphere is enhanced, both models show biases in the form o f excessive precipitation over the western Indian Ocean, central China, Himalayas, and in the subtropical Pacific Ocean (Figures 2.2a and 2.2b). Over the eastern Indian Ocean, China Sea, central parts o f Africa and in the west and east Pacific Ocean, the models underestimate precipitation. The spatial patterns o f bias from both models are similar, but the magnitude o f biases is higher in the CAM4 simulation whereas CAM5 is able to reduce many regional biases. Compared to the land, biases are significantly higher over the ocean. High magnitudes o f RMSE (Figure 2.2c and 2.2d) are seen over the northern Indian Ocean (Arabian Sea and Bay of Bengal) and the complex topography o f the Himalayas. Similar to the winter, the summer RMSE o f the CAM5 simulation is less than for the CAM4 simulation. 20 Ph.D. Dissertation: University of Northern British Columbia In general it is seen that, while both CAM4 and CAM5 are able to capture many observed features, they have regional biases somewhat similar to those in previous versions o f these models (i.e., CAM3, reported in Meehl et al, 2006). In fact, these precipitation biases especially over the Indian and Pacific Oceans are probably an intrinsic error o f the atmospheric model itself, as seen in Lin (2007) and our simulations. Compared to CAM4, the CAM5 simulation is improved with less regional bias. The magnitudes o f tropical two meter air temperature biases (not shown) in both CAM4 and CAM5 are small except in areas with complex topography such as the Himalayan region which is true for many climate models (IPCC 2007). Both models showed warm biases over most o f the tropical domain. Larger errors are in regions o f sharp elevation changes which may result simply from the mismatches between the models’ smoothed topography and the actual topography. As this chapter mainly focuses on the Asian region, the rest o f the analyses for CAM4 and CAM5 include only the Asian domain, and particularly discuss the South Asian region (summer season only). Figure 2.3 shows seasonal mean summer precipitation and 850 mb winds for (a) CAM4, (b) CAM5 and (c) observations (CMAP/NCEP). In the observations, there are two precipitation maxima, with heavier precipitation around the northern Indian Ocean and a weaker precipitation maximum along the equatorial Indian Ocean. This is an important characteristic o f the SAM precipitation. Although both o f these maxima are captured in the models, significant large-scale biases such as excessive precipitation over the Arabian Sea and diminished precipitation in the central and the eastern Indian Ocean extending into the Bay of Bengal is seen. Simulations also show reduced precipitation along the coast o f Bangladesh and excessive rain over the Western Ghats o f India. CAM4 simulates 21 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon excessive precipitation in the eastern Arabian Sea and in the Bay o f Bengal, with the maximum center around the Bay o f Bengal shifted to the west o f the observed maximum center. This is also true for the CAM5 simulation but the spatial magnitude o f the precipitation is reduced bringing its climatology close to the observation. This same conclusion regarding the CAM4 simulation is found in the recent study by Meehl et al. (2012). Apart from the SAM region, the East Asian monsoon system, covering both subtropics and mid-latitudes, is well captured in both models. CAM5 shows spatial patterns similar to those observed whereas the CAM4 simulation is drier than observations over the South China Sea. All simulations show very good correlation and RMSE skill for the East Asian monsoon, compared with the SAM. In the observed 850 mb winds pattern (Figure 2.3c), the most important features are the monsoon westerlies, the northward movement o f the low pressure area from the Bay o f Bengal and the low level jet stream passing across the equator onto the Indian subcontinent. The strengthening o f westerly 850 mb winds during the summer monsoon seasons can be seen in both models (Figure 2.3a and 2.3b) over the 10°N to 2°5N latitude belt extending eastward from the western Arabian Sea through India and Bay o f Bengal. The Bay o f Bengal is considered as the moisture source o f heavy precipitation events over the central South Asian region (Malik et al. 2010) and precipitation over this central region is mainly caused by the northward movement o f low pressure areas from the Bay o f Bengal (Lai et al. 1995). This interpretation is seen in both CAM4 and CAM5 simulations showing strong winds flowing from the Bay o f Bengal to the north over central South Asia. In general, CAM4 and CAM5 are able to simulate the winds circulation at 850 mb (such as the equatorial monsoon 22 Ph.D. Dissertation: University of Northern British Columbia flow and lower level jet stream) realistically, even though there are biases in the strength o f monsoon westerlies over the Indian region. The simulation o f the seasonal migration o f the ITCZ is a challenging issue in GCMs. Many studies (such as Hack et al 1998 and Wu et al 2003) reported that most GCMs are unable to reproduce the seasonal migration of the ITCZ precipitation. Gadgil and Sajini (1998) found that the atmospheric models which can simulate the northward migration o f the ITCZ can also simulate the interannual variation o f the Indian monsoon reasonably well, whereas in models with poor SAM simulation, the ITCZ remains over the equatorial oceans in all seasons. Over the SAM region, the seasonal migration o f the ITCZ from the equatorial region in winter to the heated continent in summer is the most important feature o f seasonal variation. We therefore briefly analyze the seasonal migration o f ITCZ in our model simulations by analyzing the mean January and July surface winds (not shown). It is found that this planetary scale feature o f the general circulation is well captured by both CAM4 and CAM5. Also the location and strength o f both the westerly jets over the northern Indian region during January and the tropical jets during July (early Monsoon) are fairly well reproduced in simulations which indicate that both models realistically capture the large shift o f the ITCZ from January to July. We have also performed simulations o f CAM4 and CAM5 at higher resolution (0.9° x 1.25°) to analyze the effect o f better resolved topography (which is an important aspect for the simulation of precipitation). We found (not shown here) that increasing the resolution improved the simulation over areas o f complex terrain such as the Western Ghats and Himalayas in the SAM region. The Western Ghats capture much o f the rain on the Arabian Sea-facing side, while the other side o f these mountains (to the east in southeastern India) 23 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon remains dry in the summer season. This is a localized effect and can only be seen in the higher resolution simulation. Also in the higher resolution run, heavy precipitation on the coastal mountain slopes o f Myanmar, across the Bay o f Bengal, is well simulated but with the same overestimation in the amount as seen in the lower resolution run. Also the excessive precipitation over the Tibetan Plateau in CAM4 (also the case in CAM5) is reduced in its higher resolution simulation. Both higher resolution simulations are somewhat closer to observations for the complex terrain regions o f the SAM. The seasonal evolution of SAM precipitation is examine in time-latitude diagrams averaged over the SAM longitudes (50° to 120°E) for observations, CAM4 and CAM5 (Figure 2.4). In the observations (Figure 2.4c), a well-defined seasonal precipitation pattern that varies with latitude and moves significantly northward from 5°N in winter to 20°N in summer is visible. In models (Figure 2.4a and 2.4b), the northward shift o f precipitation starting in winter and reaching a maximum in summer is well captured but there are considerable systematic errors such as the simulated summer precipitation northward extent reaching too far north with a second maximum around 30°N. In the CAM4 and CAM5 simulations, the precipitation reaches a maximum at 15°N and exhibits an essentially realistic seasonal migration, but the simulations have heavier than observed precipitation over the Western Ghats. The overestimation is higher in CAM4 than in CAM5 along with more penetration o f precipitation towards the north. In CAM5, the maximum contour o f precipitation is somewhat reduced and is comparable to observations. The annual cycle o f precipitation averaged over the latitude and longitude o f the SAM region is simulated quite well in models, with a well-defined seasonal cycle as shown in Figure 2.4d. Several characteristics o f the annual cycle of SAM precipitation, such as the rapid onset between May and June, the 24 Ph.D. Dissertation: University of Northern British Columbia sustained high precipitation from June to August and the slow withdrawal during September to October, are well simulated by both models. As noted earlier in the discussion o f spatial patterns, both models produce realistic seasonal variation o f precipitation in the SAM region, but with considerable overestimation. The annual cycle highlights this overestimation more clearly by showing excessive precipitation in the months July, August and September as well as in December, January and February. From the above discussion o f mean climatology simulations, it is seen that CAM5 has significantly improved simulations compared to CAM4. We further investigate this improvement by performing different sets o f CAM5 simulations using the same radiation (CAMRT), aerosol (BAM) and boundary layer (HB, Holtslag-Boville, 1993) schemes which are used in the CAM4 default configuration. This switching o f new schemes (in CAM5) with old ones (in CAM4) allowed us to explore the effect o f each individual scheme in improving CAM5 simulations over the SAM. We named each individual experiment o f CAM5 as CAM5 BAM, CAM5 BAM CAMRT, CAM5 BAM HB and CAM5_BAM_CAMRT_HB (see Table 2.1 for details). Figure 2.5 represents the summer precipitation difference o f CAM5 BAM, CAM5 BAM CAMRT, CAM5 BAM HB and C A M 5 B A M C A M R T H B from observation. Comparing these differences, with the default CAM4 and CAM5 runs (Figure 2.2a and 2.2b), reveals that the implementation o f new boundary layer schemes (UW moist turbulence) in CAM5 has the greatest effect (in our case) on decreasing the overestimation seen in the CAM4 simulation. As reported by Park and Bretheron, (2009), the new UW moist turbulence scheme improved the cloud top boundary layers in the CAM model. They used the CAM3.5 version to test this scheme and found significant reduction in model bias. In our case, although the new RRTMG radiation scheme and the foil 25 S Islam: Ensemble Simulation and Forecasting o f South Asian Monsoon representation o f aerosol indirect effects do not contribute as much to the improved SAM simulation, virtually every atmospheric process (revised/replaced) and its physical representation in the new version makes an improvement in the simulation (the individual discussion o f all these new features o f CAM5 is beyond the scope o f this study). 2.3.2 Monsoon Interannual Variability In this section, the monsoon variability is examined by focusing on the simulation o f monsoon indices (precipitation and circulation) for both CAM4 and CAM5 models. The strong and weak monsoon composite analysis is also discussed to further explore the simulation’s interannual variability. It has been well recognized that the interannual variability o f many climatological variables on earth can stem from ENSO, which is the strongest interannual variability o f the earth's climate system. Thus, the link between simulated SAM precipitation and ENSO is also explored using lag correlation. This is also performed to explore the relationship between the Indian Ocean Dipole (IOD, Saji et. al, 1999) and SAM precipitation. 2.3.2.1 Asian Monsoon Indices We examine the models simulations o f monsoon interannual variability with several commonly used monsoon indices, including Indian Summer Rainfall (ISR), the W ebsterYang monsoon index (WY index) (Webster and Yang 1992), the Southeast Asian Monsoon (SEAM) index or Western North Pacific Monsoon (WNPM) index (Wang and Fan 1999), Indian Monsoon (IM) index (Wang et al. 2001) and South Asian Monsoon Index (SAMi, here the subscript i denotes “index” to differentiate it from SAM) (Gowani et al. 1999). The definitions o f these indices are given in Table 2.2. 26 Ph.D. Dissertation: University of Northern British Columbia Figures 2.6 and 2.7 show the simulated and observed interannual variability o f ISR, WY, IMI, WNPM, and SAMi monsoon indices normalized with their respective standard deviation. All these indices are circulation indices except ISR which represents SAM precipitation. To analyze model simulations for individual strong and weak monsoons, the ISR index is separated from the circulation indices (Figure 2.7) and is presented in Figure 2.6. In Figure 2.6a, the observed ISR index representing the strength (strong and weak monsoon) and interannual variation o f SAM precipitation, is shown. Strong monsoon years such as 1980, 1988 and 2007 are differentiable in the observations, whereas in CAM4 and CAM5 (Figure 2.6b and 2.6c) only the year 1988 has the same sign. Both 1980 and 2007 are characterized as weak monsoon years in both models, opposite to the observations. Similarly, the observed weak monsoon years 1984, 1986 and 2002 are simulated as strong monsoon years in model. This means that both CAM4 and CAM5 failed to capture the interannual variability o f the SAM, except in some years. We find that there are large errors in the simulation o f some extreme seasons which lead to the overall poor skill. Considering SAM extreme precipitation, for those associated with ENSO, both models simulate at least the sign of SAM accurately. For example, the La Nina o f year 1988 is successfully simulated by both models while CAM4 failed to spatially capture the El Nino year o f 1987. This analysis suggests that the low skill in simulation o f monsoon interannual variation arises mainly from a poor simulation ENSO-Monsoon teleconnections rather than the lack o f air-sea interaction (coupling). Since the SAM has remote and local SST teleconnections, the poor simulation of SAM interannual variability in these atmospheric models (forced with observed SST) can be explained by analyzing the ENSO-Monsoon teleconnection and regression o f Nino SST on SAM precipitation. We will further discuss this issue in the next section. 27 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon In case o f circulation indices (Figure 2.7), it is seen that both CAM4 and CAM5 show considerable skill in simulating the interannual variation o f the WY and WNPM indices with significant correlation coefficients (see Table 2.2). For the IM and SAMi indices, correlation coefficients are insignificant, meaning that these indices are not well simulated in these models. As the WNPM index represents the East Asian summer monsoon, the significantly higher correlation o f this index means that East Asian monsoon circulations are better simulated in both models compared to the SAM. Analyses o f monsoon spatial patterns also support this result. This may be due to the fact that the East Asian monsoon has a stronger response to ENSO than the Indian monsoon. To get better insight into the simulated circulation in CAM4 and CAM5 simulations, summer mean velocity potential and divergent wind anomalies are calculated for both models and observations (NCEP). These show upper-level convergence and lower-level divergence over the equatorial central Pacific, and upper-level divergence and lower-level convergence over the SAM region (not shown here). In Figure 2.8, we analyze the difference between model and observation (NCEP) o f the summer mean velocity potential and corresponding divergent winds at 850 mb and 200 mb. Although some o f the circulation indices have higher correlation for CAM4 than CAM5, the overall spatial patterns from the CAM5 simulations are better. At both atmospheric levels, over the Pacific and Indian regions, CAM5 have much better skill (less difference) compared to CAM4. 2.3.2.2 Composite Analysis Modeling extreme events is one o f the most challenging issues and validating model extreme event simulations is therefore important to assess their performance. In this subsection, we 28 Ph.D. Dissertation: University of Northern British Columbia focus on several particular years which were recorded as strong and weak monsoon years over the SAM region (as seen in Figure 2.6). Strong and weak monsoon years are characterized on the basis o f significant weak or strong summer precipitation over the monsoon region. Years with anomalies o f summer mean precipitation greater than 0.5 standard deviation above the mean are categorized as strong monsoon years (1980, 1981, 1983, 1988, 1994, 1996, 1998 and 2007) and those with mean precipitation less than -0.5 standard deviation below the mean are categorized as weak monsoons (1982, 1984, 1986, 1987, 1989 and 2002). Composites o f weak and strong monsoon years from observations (CMAP), CAM4 and CAM5 are shown in Figure 2.9. The observed weak and strong monsoon composites have a large-scale structure with anomalies o f the same sign over many parts within the SAM region. The observed weak (Figure 2.9a) monsoon composite has negative anomalies over the whole Indian region whereas in the strong monsoon (Figure 2.9b) composite there are positive anomalies over the Indian land area, Bay o f Bengal, and the maritime continent. The models composites (Figures 2.9c, 2.9d, 2.9e and 2.9f) show that the simulations failed to reproduce the observed anomaly patterns with some areas having significant differences. For its weak monsoon composite, CAM4 and CAM5 have a much different pattern with more intense and large scale positive anomalies (which is opposite from the observed composite patterns) over the Western Ghats and Bay of Bengal. CAM4 and CAM5 to some extent, reproduce the strong monsoon composite over southern India. In short, both models show poor skill in differentiating the strong and weak monsoon years. This is also seen in the models ISR index (Figure 2.6) as both showed poor skill in simulating the interannual variability o f the SAM region. This is probably due to the simulated overestimation o f precipitation, as well as strong internal dynamics (noise) in the models. An examination o f the interannual variability o f CAM4 and CAM5 monsoon simulations show S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon that, although the observed SST tends to enhance the variability, the internal dynamics also produce considerable interannual variability in these simulations. It is interesting to note that the large variance in control runs is dominated by a few events such as 1983, 1988 and 2007. For many years, the interannual variability produced by internal dynamics is often larger than that in the control runs, suggesting that the interannual variability generated purely by internal dynamics is comparable to that forced by the slowly varying SST boundary forcing in many cases. This is probably the reason why these models perform poorly in differentiating strong and weak monsoon seasons. 2.3.3 Teleconnection of SAM with ENSO and IOD in CAM Simulations In this section, we explore how well CAM4 and CAM5 capture SAM-ENSO and SAM-IOD relationships. We perform regression analysis to analyze spatial patterns o f these relationships whereas lag-lead correlation is used for temporal analysis. Figure 2.10 shows the linear regression o f Niiio3.4 (-5°S-5°N, 120°-170°W) and IOD ((-10°S-10°N, 50°70°E) - (-10°S-0°, 90°-110°E)) SST indices with JJAS precipitation and 850 mb winds for observations and both models. Unless stated otherwise, all regression maps show the covariance o f the normalized Nino3.4 and IOD indices. The regression o f precipitation and 850 mb winds onto the Nino3.4 index is an important key to understand the behavior o f CAM4 and CAM5 in simulating realistic ENSO properties as these fields are direct indicators o f the connection between the ocean and the atmosphere. In Figure 2.10a, the regression o f the observed Nino3.4 index with precipitation and 850 mb winds is shown. The observations reveal enhanced precipitation over the Bay o f Bengal, accompanied by a westerly winds anomaly and decreased precipitation over most o f the Indian region. In the models’ results, (Figure 2.10c and 2.10e) quite realistic patterns are seen over the equatorial 30 Ph.D. Dissertation: University of Northern British Columbia Indian Ocean whereas over the Western Ghats and central India, both CAM4 and CAM5 have opposite response compared to observations. In the IOD regression pattern, (Figure 2.10d and 2.1 Of), spatial modes are well reproduced over the Indian Ocean (enhanced precipitation over the western equatorial Indian Ocean and decreased precipitation over the Eastern Indian Ocean) whereas over the Indian subcontinent the models have the opposite sign similar to the ENSO regression. The analysis above show that precipitation response to local and remote SST in the models' simulations is not preserved. To further investigate this issue, the lagged correlation o f Nino3.4 and IOD indices with the simulated ISR index is shown in Figure 2.11. The area averaged (0°—40°N, 55°-100°E) time series o f observed CMAP precipitation (ISR, solid black line), observed All-India Precipitation index (AIR, dashed black line) and simulated (CAM4, solid blue line and CAM5, solid red line) time series are correlated with observed Niiio3.4 and IOD SST indices. The months with negative (positive) sign indicate that SST leads (lags) the ISR with maximum lead o f 12 months (1 year). Months 0 and 12 indicate June whereas months 4 and 8 correspond to February and October (minus sign for previous months). Correlations are calculated using a 5-month sliding window. The observed positive correlations occur when the SST leads the SAM precipitation and negative correlations occur when SST lags the SAM precipitation. A negative correlation is seen for SST from the same summer to the following winter, showing a weak (strong) SAM in El Nino (La Nina) developing years. The highest negative correlations occurs when the monsoon leads Nino 3.4 SSTs which suggests, as reported in Kirtman and Shukla 2000, that monsoons can provide favorable conditions for triggering or enhancing El Nino or La Nina events in the Pacific during the following winter. The observed ISR index has significant lagged and lead 31 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon correlations with SST, with the highest value o f around 0.5 when monsoon precipitation is slightly led by SST. This suggests a cross-interaction between monsoon and ENSO, namely, ENSO impacts on SAM precipitation and meanwhile the monsoon variability may affect the ENSO evolution, intensity, and periodicity. The mutual influence between ENSO and SAM has been widely reported in other studies as well (Chung and Nigam 1999; Kitoh et al. 1999; Meehl and Arblaster 1998; Wang et al. 2004). It is noted that the difference in correlation magnitude between Figure 2.11a and those in other studies (Kirtman and Shukla 2000; Yasunari 1990) may be accounted for a dramatic change o f the ENSO-SAM relationship in the late 1970s, since Figure 2.1 la is obtained only using the data after 1979. In the CAM4 and CAM5 simulations, this relationship is poorly captured, particularly when SST lags monsoon. CAM4 shows a somewhat comparable result when SST leads monsoon. Both of these models failed to maintain the monsoon and ENSO relationship because, as discussed in the composite analysis, the internal dynamics o f these models can overwhelm the Pacific SST influence on monsoon precipitation. In Figure 2.11b, the lag-lead correlation o f SAM with IOD is computed. A positive correlation during late spring and the simultaneous summer is seen. The correlation changes to negative in the following fall, suggesting a negative feedback o f SAM on the IO. For the models, the simultaneous response is not clear but when IOD leads monsoon, both models show a comparable response to the observations. Overall the above analysis shows that the interannual skill o f CAM4 and CAM5 in simulating SAM is poor. Both models failed to differentiate strong and weak monsoon which is indirectly linked to their poor reproduction o f the ENSO-monsoon relationship. The correct representation o f the ENSO-SAM relationship in models is crucial, since it is the basis for seasonal forecasts o f SAM using climate models. The poor skill o f CAM4/CAM5 interannual 32 Ph.D. Dissertation: University of Northern British Columbia variability, even when forced by observed SST, is attributed to the models’ poor skill in simulating the SST-precipitation relationship over the Indian and Pacific oceans and a lack o f atmosphere-ocean coupling that has been reported as critical for skillful simulation o f the monsoon (Wang et. al. 2005). Since this lack o f atmosphere-ocean coupling in GCMs is one o f the possibilities for their poor skill, we focus our analysis on the coupled CCSM4 model in the next section and compare its simulation with CAM4 and observations. 2.4 Coupled Simulation using CCSM4 We investigate the mean climatology and relationship between SAM precipitation and SST in the CCSM4 coupled model, which will allow insight into the role o f coupling on the simulation o f SAM precipitation. As previously discussed, CCSM4 uses CAM4 as its atmospheric model and POP2.2 as its ocean model. Along with observations, we will also contrast CCSM4 with the CAM4 results presented in previous sections. Here we will mainly use the CCSM4 climatology run whereas for some sensitivity experiments, data from a CCSM4 transient run (CCSM4 TR, downloaded from NCAR) forced with observed forcing o f greenhouse gases) is also used. 2.4.1 Mean Climatology of CCSM4 We first analyze, as for the atmospheric simulations, the coupled model over the tropical region including both Indian and Pacific Oceans. Figure 2.12b shows the JJAS mean precipitation difference (in mm/day) between CCSM4 and CMAP observation. The difference between the CAM4 climatology run (CAM4 CLIM) and observations is also shown in Figure 2.12a. CCSM4 shows significant differences, particularly over the Indian and Pacific Ocean. Specifically, the coupled simulation underestimates precipitation over the 33 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon western and eastern equatorial Pacific and over the eastern Indian Ocean, and it overestimates precipitation over the western Indian Ocean and central equatorial Pacific Ocean. These differences, shown in Figure 2.12, result from the ocean component o f CCSM4. The SST bias from the ocean model influences the precipitation directly, making it different from the observed precipitation climatology. Comparing CCSM4 with CAM4 CLI M shows that CCSM4 precipitation biases are at a broad scale (especially over oceans). In CCSM4, the negative precipitation bias increases over the equatorial area in the Pacific Ocean. This is probably due to the feedbacks o f air-sea coupling in the coupled model that amplifies the bias in the atmospheric and oceanic components. Same as reported in Gent et al., (2011), CCSM4 has the double ITCZ bias, characterized by heavy simulated precipitation over most o f tropical Pacific and the equatorial Indian Ocean, and light precipitation in the west and central Pacific between 15° and 30° south. The double ITCZ problem was also present in previous versions of the CCSM model. Lin (2007) found that most o f the current coupled models and uncoupled models have this double ITCZ problem to some extent. Focusing on the SAM region reveals that while the CAM4 simulation overestimates precipitation over much o f this region, the coupled simulation is more realistic and comparable to observations although the amounts are still overestimated somewhat. This is similar to results from the CCSM3 model reported in Meehl et al, 2006. The CCSM4 simulation has reduced bias over the Arabian Peninsula and the western coast o f India (Figure 2.12b), which is a direct consequence o f the thermodynamic air-sea interactions in the Arabian Sea, Bay o f Bengal, and South China Sea, which are absent in uncoupled simulations. Meehl et al. 2012 also reported that the CCSM4 simulation over the SAM region is much better than the CAM4 simulation. It is also reported in Wang et al. (2004) that the implementation o f air-sea coupling could improve the model simulation o f monsoon 34 Ph.D. Dissertation: University of Northern British Columbia precipitation and circulation in the Asian monsoon. The absence o f air-sea coupling in CAM4 results in continuous heating o f the atmosphere by the prescribed SST (warm SST) which increases the evaporation, resulting in increased precipitation in its simulations. This amplifies the SAM variability and therefore CAM4 simulations overestimate precipitation in the SAM region. We will further discuss air-sea coupling and role o f SST bias in the next section. In general, CCSM4 shows a large reduction in precipitation and less bias over the SAM region particularly over Arabian Sea and west equatorial Indian Ocean as compared with the CAM4 simulation. The reduction o f SAM precipitation in the coupled model can also be seen in the seasonal cycle (not shown) averaged over the region. Seasonal mean summer precipitation and 850 mb winds pattern from CCSM4 simulation, shown in Figure 2.13, is more realistic and comparable to the observations as compared to CAM4 and CAM5 simulation (Figure 2.3). The extreme high precipitation area over the northern West Indian Ocean is diminished in the coupled simulation making it more comparable to observations. However, CCSM4 also removed the heavy observed precipitation over the Bay o f Bengal making its simulation different than observation. This is a significant shortcoming o f the coupled simulation. Also, in CCSM4 runs, the precipitation is more concentrated in the western Indian Ocean, which was also seen in CCSM3 (Meehl et al, 2006). Considering these spatial patterns only, the overall mean climatology o f CCSM4 seems to be more realistic and much better than from the uncoupled simulation, providing evidence that coupled air-sea interaction is necessary for climate models. 35 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon 2.4.2 Teleconnection of SAM with ENSO in Coupled Simulation A correct ENSO-monsoon relationship is one o f the prerequisites needed for a coupled model to produce reliable simulations o f the monsoon. Here we document the ability o f CCSM4 to reproduce the observed lag-lead relationships between the SAM and ENSO, and to understand how systematic errors may affect the simulation o f this relationship. We first discuss the simulation of Nino3.4 and IOD indices in the coupled model. Figure 2.14 shows the variance spectrums of monthly Nino3.4 and IOD index for (a) observation and (b) CCSM4. In the observations, the broad peak shows a probable frequency o f ENSO events of 3 to 6 years (0.24/year) with the maximum variance o f about 20. In the CCSM4 simulation, the same frequency o f 3 to 6 years is seen with maximum peak at 4 years although the CCSM4 Nino3.4 variability is significantly larger than the observed variability. In the previous version o f CCSM i.e. CCSM3, the ENSO frequency (about 2 years) was reported as very poor by Collins et al. (2006). This shows that the new version o f CCSM has significant improvements in its dynamics and can therefore simulate the ENSO properties more closely to observation. For the case o f the IOD spectrum, being irregular in its oscillatory period, there is no well-defined peak in the observed frequency whereas in CCSM4, the peak o f the IOD spectrum remains in phase with that o f the Nino3.4 index showing a frequency o f 3 to 6 years. Also these results show that the IOD index varies significantly with ENSO in CCSM4. The observed IOD variance is very small whereas the model variability is significantly larger showing a similar pattern as the Nino3.4 index. Figure 2.15 (top) shows the linear regression o f the CCSM4 simulated JJAS Nino3.4 (-5°S5°N, 120°-170°W) SST index with JJAS precipitation and 850 mb winds. Comparison of this regression pattern with Figure 2.10 reveals the improved spatial Nino3.4 regression 36 Ph.D. Dissertation: University of Northern British Columbia pattern o f the CCSM4 simulation, which is due to the improved and coherent atmospheric response in coupling. This can be also seen in Figure 2.15 (bottom), representing lag-lead correlation o f CCSM4 simulated SAM precipitation with simulated Nino3.4 SST index. The observations show positive correlation when precipitation lags ENSO and strong negative correlations when precipitation leads ENSO. The CCSM4 can partially capture the observed variation o f correlation timing but with quite different magnitudes. Comparing CCSM4 laglead correlation with CAM4 (Figure 2.11) shows significant improvement in the coupled simulation relationship. This is probably due to an improved or consistent SST simulation and its interaction with atmosphere, which is absent in atmosphere-only simulations. The response o f ENSO to SAM precipitation (i.e. SST lagging the monsoon) is realistic in CCSM4, better than in CAM4. The lag-lead correlation for ENSO suggests that in the CCSM4, SST has significant biases over the central equatorial Pacific and Indian Ocean. The fact that CCSM4 reveals a better connection between ocean and atmosphere in its simulation, although the Nino3.4 amplitude is considerably larger than the observed one, supports the low sensitivity to SST in the CAM4 model. Apparently, the atmospheric component (i.e. CAM4 in CCSM4) shows a realistic response to ENSO variability only when being forced by a strong SST signal which is confirmed by CAM4 sensitivity experiments (not shown or discussed further here). Since SST and precipitation are strongly coupled in the tropics, an unrealistic simulation of SST distribution should lead to an unrealistic SAM-ENSO relationship. To assess this we analyzed the average mean seasonal SST differences between CCSM4 and the observed climatology over the Indian and Pacific oceans. This also helps further investigate the SAM improvements in coupled simulations. In the observed mean SST climatology (not shown 37 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon here) the most important feature is the warm pool region over the west Pacific with SSTs more than 28°C and a cold SST tongue along the east Pacific associated with easterly trade winds along with a strong east-west SST gradient across the equatorial Pacific. A north-south irregularity is present in the eastern Pacific where warm water is located north o f the equator and cold water is present along the west coast o f South America. This north-south irregularity is important for the formation o f the annual cycle (Xie 1994) and ENSO (Cane and Zebiak 1986). CCSM4 SST patterns, showed almost the same climatology but with significant differences in magnitude. Figure 2.16 shows the difference between the simulated CCSM4 SST climatology and the observed SST climatology for DJF, MAM, JJA and SON seasons. In CCSM4, the model simulates warm SSTs over most o f the Pacific and Indian Oceans in all four seasons whereas over the eastern equatorial Pacific and northern East Indian Ocean, the model has cold SST biases in spring and summer. The cold SST bias in the equatorial Pacific significantly reduces the temperature o f the warm pool whereas the warm biases near the coast of Peru reduce the meridional SST gradient. This may be the cause o f the double ITCZ in the CCSM4 simulations, as seen in Figure 2.12b. Comparing the SST bias in summer with the summer precipitation bias o f CCSM4 (Figure 2.12) reveals that the negative precipitation anomalies over much o f the Pacific Ocean are largely associated with the cold SST bias in the same region. This most likely originates from errors in the atmospheric model. As reported in Li and Hogen 1999, and Manganello and Huang 2008, deficiencies in the simulated SSTs are likely responsible for some o f the unrealistic ENSO properties in the coupled model. Although CCSM4 simulated summer monsoon precipitation in South Asia is considerably improved, the SST bias over the Pacific Ocean (ENSO region) can still lead to a spurious response o f the SAM to ENSO in CCSM4. 38 Ph.D. Dissertation: University of Northern British Columbia 2.5 The Contribution of Air-Sea Interaction to SAM Simulation In previous sections, we investigated the SAM simulations using CAM4 and CCSM4. Their differences were due to two factors: 1) the role o f coupling in CCSM4 which is absent in CAM4; 2) the SST consistency in CCSM4. To better isolate the role o f coupling in simulating the SAM, we designed another set o f experiments referred as to CAM4 POP, in which the predicted SST by CCSM4 forces CAM4. These experiments allow us to investigate the air sea coupling in CCSM4 and its effect on SAM precipitation. An ensemble strategy is used as discussed in section 2.2, and therefore the following discussions are from the ensemble mean analysis. The difference between the CAM4 POP and CCSM4 (Figure 2.17b) reveals that the precipitation and westerly winds in the CAM4_POP simulation are enhanced over the northern Indian Ocean including the Arabian Sea and Bay o f Bengal. These differences show that SAM monsoon precipitation is amplified in the absence o f the air-sea coupling. A considerable increase is also seen in the variance (not shown here), revealing that the monsoon variability is amplified by about half as compared with the CCSM4. The absence o f the air-sea coupling keeps SSTs warm in the Indian Ocean (as discussed previously), which increases the local evaporation and precipitation. This suggests that the air-sea coupling works to stabilize the monsoon and hence suppress the variability, which is the case in CCSM4 simulations. Figure 2.17a shows the difference between observations (CMAP/NCEP) and CAM4_POP mean precipitation and winds. The spatial pattern in this case is almost same as seen in Figure 2.12b for the SAM region. This means that the reduction in excessive precipitation seen in the CCSM4 simulation is not only due to the airsea interaction but also due to the more consistent SST simulation in the coupled model 39 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon integration. In other words, the more consistent SST simulated in CCSM4 (different from observation in magnitude) is responsible for decreasing the overestimation o f precipitation. To further elaborate this point we performed another set o f experiments and analyzed the CCSM4 SST in the Indian Ocean and its effect on SAM monsoon. For this set of experiments, we used the SST data from the CCSM4 transient run (C C SM TR ). The CCSM4 TR precipitation is similar to precipitation in the CCSM4 run, but the SST bias is more significant in CCSM4_TR simulation. The Indian Ocean SST bias, seen in Figure 2.16, prompted us to find its influence on the SAM. Comparison o f CCSM4 and CAM4 runs suggests that the coupled model cold SST biases significantly reduce monsoon precipitation as seen in Figure 2.17. The role o f these biases is for CCSM4 to remove the overestimation (reduction in the monsoon) seen in the CAM4 simulation. Gimeno et al. (2010) have shown the northern Indian Ocean to be an important moisture source for Indian monsoon precipitation and therefore understanding the monsoon dependence on Indian Ocean SST, and correctly representing this in climate models, is important in order to realistically predict monsoon fluctuations. In Figure 2.18a and 2.18b, the annual cycle o f SST over the northern Indian Ocean for two separate regions, the Arabian Sea and Bay o f Bengal, is shown for observation and for both the CCSM4 and CCSM4 TR coupled model runs. In CCSM4, the magnitude o f the cold SST bias in the Arabian Sea is small, diminishing after the summer season, whereas CCSM4 TR shows a large cold bias throughout the year, with a slight decrease in its magnitude during the fall. Over the Bay o f Bengal, instead o f a cold SST bias, both models show a warm SST bias, with a larger bias in the CCSM4 climatology run. This means, from January to July, the cold 40 Ph.D. Dissertation: University of Northern British Columbia bias becomes more confined to the Arabian Sea in these models, while a warming appears in the Bay o f Bengal. To further analyze the role o f this Indian Ocean bias, we designed two different sensitivity experiments with CAM4 using climatological SST repeated every year but with the addition of the CCSM4 northern Indian Ocean SST bias into the SST climatology. In the first experiment, we added the annual cycle o f CCSM4 TR SST bias (bounded in the Ocean region 7°-30°N, 40°-100°E) covering the whole northern Indian Ocean to the annual cycle o f climatological SST (named as CAM4_AS_BoB). To remove any discontinuity at the boundary o f the modified region, the bias is added in a tapered manner over the region, being highest at the center and approaching zero at its boundaries. In the second set o f the experiments, the bias o f CCSM4_TR SST is confined only to the Arabian Sea covering 7°30°N, 40°-80°E (named as CAM4_AS). Both o f these experiments are compared against the CAM4 climatology run (CAM4 CLIM) forced with the climatological SST cycle (see Table 2.1 for details). To account for uncertainties, each simulation is performed three times starting with three different initial conditions. As mentioned previously, CAM4 is not very sensitive to the small changes in its boundary conditions which means that only a strong anomaly added to its boundary condition will produce a significantly altered simulation. We therefore use the CCSM4 TR simulated SST in this case which has larger magnitude o f SST bias, especially in the Arabian Sea. In Figure 2.18c and 2.18d, precipitation differences of both of these experimental setups from the CAM4 CLIM run are shown. In CAM4_AS_BoB, a significant reduction in the precipitation is seen over the whole Indian subcontinent, Arabian Sea and Bay o f Bengal. Also the increased precipitation is seen over the West Indian Ocean around 10°N. In the second experiment (CAM4 AS) where we used 41 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon the SST bias only over the Arabian Sea, the reduced precipitation over the Bay o f Bengal diminished while the Arabian Sea continued to be affected by the bias. Both these experiments support our conclusion that the reduction o f CCSM4 SAM precipitation is mainly due to the SST bias in northern Indian Ocean. The weakened (reduced precipitation as compared to CAM4) monsoon seen in the CCSM4 run is therefore mainly caused by systematic cold SST biases o f northern Indian Ocean particularly in the Arabian Sea. This cold bias keeps the ocean cool, and thus reduces evaporation, which results into the reduction o f SAM precipitation. This issue needs more attention in the coupled model to improve monsoon simulation. CCSM4, while being revised with improved physics and dynamics, needs improvement in its oceanic component. 2.6 Summary and Conclusions The SAM precipitation is an important climatic feature due to its profound influence on droughts and floods over Asia, along with its influence on the global general circulation. Improved and accurate simulation o f the SAM system is therefore crucial to predict decadal and seasonal climate as well as projecting long-term climate change in the region. Also it is necessary to assess whether climate models can realistically simulate monsoon systems before using them for such predictions. This chapter discussed selected features o f the SAM precipitation in the simulations o f NCAR's versions o f the uncoupled (CAM) and coupled (CCSM) climate models. Simulations using CAM4, CAM5 and CCSM4 models are performed and compared against observations to identify improvements and discrepancies in these newly updated models. Along with the simulated mean climatology o f the region, the interannual variability and SAM-ENSO/IOD teleconnections are evaluated using lag-lead 42 Ph.D. Dissertation: University of Northern British Columbia correlation and regression analysis. The improvements due to air-sea interactions and impact o f SST biases from the CCSM4 are assessed in coupled model simulations. It is found, in the comparison o f atmosphere-only simulations, that the improvements in CAM4 and CAM5 dynamics and convection parameterizations have eliminated many regional differences (especially for CAM5). Many improvements in these simulations, compared to previous versions, are seen in both models. The detailed structure o f spatial patterns and the seasonal cycle o f monsoon precipitation are well reproduced in both CAM4 and CAM5. The annual cycle o f average precipitation is well simulated along with its major characteristics such as the rapid monsoon onset between May and June, the high precipitation during June-August and slow withdrawal during September-October. The large northward shift o f the ITCZ from January to July is also well simulated by both CAM4 and CAM5 and its location and strength are fairly well reproduced. Increasing the resolution shows a pronounced improvement in precipitation simulation with a reduction in many regional biases, especially over regions o f complex terrain. The CAM5 new dynamics and physics showed improved simulation results over the SAM region. Sensitivity experiments using CAM5 showed that the implementation o f new boundary layer schemes (UW moist turbulence) in CAM5 contributes to decreasing the CAM4 simulation overestimation. Analysis showed that both CAM4 and CAM5 poorly simulate the ENSO-monsoon teleconnection. These models partially captured the monsoon interannual variability with inconsistencies in oscillatory period and amplitude. It is also found that the simulation o f East Asian summer monsoon is much better than the simulation o f the SAM in both CAM4 and CAM5. The better simulation o f the WY and WNPM monsoon indices and poor simulation o f the IM, SAMi and ISR monsoon indices in both CAM4 and CAM5 also supported this 43 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon conclusion. Both the models are able to simulate the winds circulation such as equatorial monsoon flow and lower level jet stream very well. Both models simulate excessive precipitation over the western Indian Ocean and subtropical Pacific Ocean whereas decreased precipitation is simulated over the eastern Indian Ocean, China Sea and South America. Over the SAM region their simulations show significant large-scale biases such as excessive precipitation over the Arabian Sea and over the Western Ghats o f India, and reduced precipitation over the eastern Indian Ocean extending into the Bay o f Bengal. The CCSM4 simulated SAM precipitation is considerably improved compared with CAM4 with the reduction of many biases particularly over the Arabian Peninsula and the western coast o f India. The results showed that the air-sea coupling has significantly improved the monsoon simulation. Along with these improvements, interrupted northward progression and delayed onset o f the monsoon over the SAM region is seen. The CCSM4 underestimated the precipitation over the equatorial area in the Pacific Ocean. Also CCSM4 still has the double ITCZ problem that was also present in the previous versions o f the CCSM model (CCSM3). CCSM4 showed a systematic cold bias in the simulation o f SSTs over the tropical Pacific Ocean and hence showed problems in simulating the observed SST-precipitation relationship. Analysis over the whole tropical region revealed that biases in CAM4 and CCSM4 are somewhat similar to those in previous versions o f these models. The frequency o f ENSO in CCSM4 is found to be more realistic than was simulated in its previous version (CCSM3). The SAM-ENSO teleconnection in the CCSM4 climatology run is partially captured. Significant cold biases over the equatorial Pacific Ocean are found in CCSM4, particularly in winter and early summer. It is seen that the air-sea coupling can improve the simulation o f precipitation. Forcing CAM4 with coupled model SST clarified the 44 Ph.D. Dissertation: University of Northern British Columbia impact o f the air-sea coupling in the interannual variability o f the SAM precipitation. The local air-sea coupling over the SAM region acts to modulate the activity o f the SAM summer monsoon as well as the remote SST forcing. The SST continuously warms over the SAM region as the feedback from the atmosphere to the ocean does not exist in the CAM4 model forced with SST from the coupled model. The SST warming contributes to increased evaporation, which results in the monsoon destabilization over the SAM region. Another impact, the absence of the air-sea coupling, enhances heavier precipitation in the SAM. It is found that, along with air-sea interaction, SST bias in the CCSM4 model plays an important role in simulation o f SAM precipitation variability and magnitude. Using CAM4 sensitivity experiments, the influence o f the coupled model SST bias in the northern Indian Ocean on SAM precipitation is investigated. It is found that the reduction o f SAM precipitation in the coupled simulation, as compared to the uncoupled simulation, is mainly due to cold SST bias in the Arabian Sea. The strengths and limitations in simulating Asian summer monsoon in CAM4, CAM5 and CCSM4, depend mainly on how well they simulate the mean state o f atmosphere, its variability, the internal dynamics o f monsoon systems and ocean-atmosphere interactions. Although these recent model versions have many improvements and are able to capture the observed features o f SAM precipitation, many biases are still present. This chapter shows that while the NCAR systems models can serve as tools in simulating and understanding Asian monsoon climate systems, they still have simulation errors that need further consideration. Along with the improvements in the model physics and resolution, understanding o f the coupled physical processes in conjunction with the complex topography over the SAM region is crucial. As the ocean dynamics also play an important role in Indian 45 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon Ocean SST, further studies are needed to clarify its relative importance compared with the role o f air-sea interaction in SST cooling during the SAM. It is necessary for the coupled model to simulate realistic SST variation to improve the SST climatology, which can then improve the SAM precipitation teleconnection in CCSM4. In general, modeling monsoon fluctuations mainly depends on understanding the fundamental processes that affect local climate, good parameterization and representation o f these processes and the methods used for numerical implementation o f these processes. 46 Ph.D. Dissertation: University of Northern British Columbia Table 2.1: List o f experiments used in CAM4, CAM5 and CCSM4 validation. Experiment Name CAM4 Model Used Time Period CAM4 19782008 CAM5 CAM5 19782008 CAM4_CLIM CAM4 30 years CCSM4 CCSM4 100 years CA M 5B A M CAM5 19782008 C A M 5B A M C A M R T CAM5 19782008 Description / Boundary Conditions (BC) 1.9° x2.5° horizontal resolution and 26 vertical levels. Prescribed observed SST data as BC 1.9°x2.5° horizontal resolution and 30 vertical levels. Prescribed observed SST data as BC 1.9°x2.5° horizontal resolution and 26 vertical levels. Climatology SST data repeated each year 1.9°x2.5 “ horizontal resolution and 26 vertical levels for CAM4 and l°x l° horizontal resolution with 60 vertical levels for POP2.2. Present day climatology forcing 1.9°x2.5° horizontal resolution and 30 vertical levels with Bulk Aerosol Model (BAM) scheme. Prescribed observed SST data as BC 1.9°x2.5° horizontal resolution and 30 vertical levels with BAM and CAMRT (radiation) schemes. Ensemble size 10 10 05 01 01 01 47 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon Prescribed observed SST data as BC 1.9°x2.5° horizontal resolution and 30 vertical levels with BAM and HB 1978CAM5 (Holtslag-Boville, CAM5BAMHB 01 2008 boundary layer turbulence) schemes. Prescribed observed SST data as BC 1.9°x2.5° horizontal resolution and 30 vertical levels with 1978BAM, HB, and C A M 5 B A M C A M R T H B CAM5 01 2008 CAMRT schemes. Prescribed observed SST data as BC Same as CAM4 but 1971with predicted SST CAM4 CAM4 P OP 03 2000 from CCSM4 climatology run Same as CAM4 CLIM 1978but with climatology CAM4 CAM4_AS_BoB 03 2000 SST data modified in the AS and BoB region Same as CAM4_CLIM 1978but with climatology C A M4 A S CAM4 03 SST data modified in 2000 the AS region only CLIM = climatology, B O Boundary Conditions, AS = Arabian Sea, BoB = Bay of Bengal 48 Ph.D. Dissertation: University of Northern British Columbia Table 2.2: Details o f frequently used Asian monsoon indices. The correlation coefficient o f simulated indices with observations is shown in last two columns. Name of the Index Indian Type of Index Precipitation Definition Reference PREC (5°- 40°N, 60°- Summer 100°E) Rainfall (ISR) averaged JJAS precipitation Correlation CAM4 CAMS -0.13 -0.31 0.38 0.45 0.66 0.60 -0.037 -0.11 -0.26 -0.13 over the domain Webster- Circulation U850 - U200 (0°-20°N, Webster Yang 40°- 110°E) and Yang monsoon vertical shear o f zonal 1992 (WY) winds between 850 mb and 200 mb levels Western North Circulation U850 (5°-15°N, 90°- Wang and Pacific 130°E) - U850 (22.5°- Fan 1999 Monsoon 32.5°N, 110°-140°E) (WNPM) difference o f 850 mb zonal winds Indian Circulation Monsoon (IM) U850 (5°-15°N, 40°-80°E) Wang et al. - U850 (20°-30°N, 70°- 2001 90°E) difference o f 850 mb zonal winds South Asian Circulation V850 - V200 (10°-30°N, Goswami et Monsoon 70°- 110°E) al. 1999 Index (SAMj) vertical shear o f meridonal winds between 850 mb and 200 mb levels 49 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon Precipitation (D JF Bias) CAM4 0 30E pr 60E 90E 120E 150E jp p 180 150W 120W CAM5 90W 60W 30W 0 Precipitation (D JF Bias) 20N - 30E 60E 90E 120E 150E 180 150W 120W 90W 60W 30W Precipitation (DJF RMSE) 40N -1 - 1 90W 60W ■ ^ . ... 20N - 20S 0 30E 60E 90E 120E 150E 180 CAM5 150W 120W 30W 0 Precipitation (DJF RMSE) 40N 20N 20S 0 30E 60E 90E 120E 150E 180 150W 120W 90W 60W SOW 0 Figure 2.1: Seasonal mean (December-February; DJF) precipitation differences (biases) and root means square error (RMSE) from observation (CMAP) for (a, c) CAM4 and (b, d) CAM5. Units are in mm/day. 50 Ph.D. Dissertation: University of Northern British Columbia Precipitation (JJA S Bias) CAM4 I 0 30E 60E 90E 120E 150E 180 150W 120W CAM5 90W 60W 30W 0 Precipitation (JJA S Bias) 40N 30E 60E 90E 120E 150E 180 150W 120W 90W 60W 30W Precipitation (JJA S RMSE) CAM4 20N - 0 30E 60E 90E 120E 150E 180 150W 120W CAM5 90W 60W 30W Precipitation (JJA S RMSE) 40N K ■T — 30E 60E 90E 120E 150E 180 - T- 150W 120W ....... .T." T~ 90W 60W 30W Figure 2.2: Same as Figure 2.1 but for the summer season (June-September; JJAS). 0 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon CAM4 SOE («JJAS Mean) 90E 12% CAM5 150E 60E CMAP / NCEP (OBS) 60E 2 (JJAS Mean) 90E 120E 150E (JJAS Mean) 90E 6 120E 10 150E 14 Figure 2.3: Seasonal mean (June-September; JJAS) precipitation and 850 mb winds from: (a) CAM4, (b) CAM5 and (c) observations (CMAP/NCEP). Precipitation (shaded) in mm/day and 850 mb winds (vectors) in m/s. 52 Ph.D. Dissertation: University of Northern British Columbia CAM4 i I I I I (mm/d) CAM5 1 i 1__ 1___I___I___L 40N - -40N - 20N - -20N i i i (mm/d) ..................... % I I I I I I I I 1 J F M A M J J A S O N D J F M A M J J A S O N D ICOWTOUB FBOM 2 T O 12 B r i l CMAP (OBS) j i i i i i i i ICOWTOU* F ftp W I 2 T 0 12 BV 21 Annual Cycle (mm/d) i L j i i i i i i i i L. 40N 6.0 5.0 - 20N 3.0 CAM5 2.0 CAM* I I I I I I I I I I I I J F M A M J J A S O N D ICOWTOUB FROM 2 T 0 12 g V 2 l 2 4 6 8 10 12 F ig u re 2.4: Time-latitude evolution o f precipitation averaged over (50°-120°E) for: (a) CAM4, (b) CAM5 and (c) observations (CMAP). Precipitation annual cycle area averaged over SAM region is shown in (d). Units are in mm/day. 53 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon CAM5_BAM -OBS 60E (JJAS difference) 90E 120E CAM5_BAM_HB - OBS 60E CAM5_BAM_CAMRT - OBS !50E (JJAS difference) 90E 120E -10 -8 -6 60E 150E -4 -2 90E CAM5_BAM_CAMRT_HB - OBS 0 60E 2 4 90E 6 8 (JJAS difference) 120E 1S0E (JJAS difference) 120E 150E 10 Figure 2.5: Seasonal mean (June-September; JJAS) precipitation differences from observations (CMAP) for: (a) CAM5 BAM (CAM5 with Bulk Aerosol Model (BAM) scheme), (b) CAM5 BAM CAMRT (CAM5 with BAM and CAMRT (radiation) Schemes), (c) CAM5_BAM_HB (CAM5 with BAM and HB (Holtslag-Boville, boundary layer turbulence) schemes) and (d) C A M 5 B A M C A M R T H B (CAM5 with BAM, HB and CAMRT schemes). Units are in mm/day. 54 Ph.D. Dissertation: University of Northern British Columbia 2.0 oo o -20 -3.0 1980 1965 1990 1995 2000 2005 1980 1985 1990 1995 2000 2005 1980 1985 1990 1995 2000 2005 3.0 3 0.0 O .10 - 2.0 Figure 2.6: Time series o f Indian summer rainfall (ISR) index for (a) CMAP, (b) CAM4 and (c) CAM5. Details o f this index are given in Table 2.1. The time series are normalized and thus unitless. 55 S Islam: Ensemble Simulation and Forecasting o f South Asian Monsoon 3.0 •CAM5 r s 20 1-0 I .£ o.o £ - 1.0 - 2.0 -3.0 1980 1985 1990 1995 2000 2005 1980 1985 1990 1995 2000 2005 1980 1985 1990 1995 2000 2005 1985 1990 1995 2000 2005 3.0 2.0 .E 0.0 § -1.0 2.0 -3.0 - 3.0 g 2.0 2 1.0 2 0.0 |-1.0 5 -2.0 -3.0 3.0 x 2.0 -i[3 f “ 10 i \ 0.0 » -2.0 - i A -3.0 4 —1 1980 Figure 2.7: Time series o f (a) Webster-Yang (WY) index, (b) Indian Monsoon index (IMI), (c) Western North Pacific monsoon (WNPM) index, and (d) South Asian monsoon (SAM;) index o f observation (CMAP/NCEP, black line), CAM4 (blue line) and CAM5 (red line). Details o f all the indices are given in Table 2.1. The time series are normalized and thus unitless. 56 Ph.D. Dissertation: University of Northern British Columbia (850mb, m/s) CAM4 cntrl 60E 90E 120E 150E 180 150W 120W CAM4 cntrl 60E 90E 120E 150E 180 90W 60E (200mb, m/s) CAM5 cntrl 150W 120W -8 -8 -4 90W -2 (850mb, m/s) CAM5 cntrl 0 2 90E 120E 1S0E 180 150W 120W 90W (200mb, m/s) 60E 90E 120E 4 6 8 150E 180 1S0W 120W 90W Figure 2.8: Difference between model (CAM4 and CAM5) and observed (NCEP) JJAS ft “} mean velocity potential ( 1 0 ' m /s) and corresponding divergent winds (m/s) at (a), (b) 850 and (c), (d) 200 mb. 57 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon CMAP Weak Composite (mm/d) CMAP Strong Composite 40N - 40N 20N - - 20N - (mm/d) f i r — 0- 60E 90E CAM4 Weak Composite 120E (mm/d) CAM4 Strong Composite 40N - - 40N - 20N - 20N -► (mm/d) 0t 1------ 1------ r 60E 90E CAM5 Weak Composite 40N - 120E (mm/d) 60E 90E CAMS Strong Composite 120E (mm/d) HJ 20N - 20N % 120E r r m Figure 2.9: June-September (JJAS) anomaly precipitation composites o f weak (1982, 1984, 1986, 1987, 1989 and 2002) and strong (1980, 1981, 1983, 1988, 1994, 1996, 1998 and 2007) monsoon years for (a, b) observation (CMAP), (c, d) CAM4 and (e, f) CAM5. Units are in mm/day. 58 Ph.D. Dissertation: University of Northern British Columbia CMAP / NCEP (Obs) 60E (JJAS Regression Nino3.4) 90E CAM4 60E 90E CAMS 80E 120E 150E (JJAS Regression Nino3.4) I20E 150E (JJAS Regression Nino3.4) 90E 120E 150E CMAP / NCEP (Obs) 60E (JJAS Regression IOD) 90E CAM4 60E 90E CAM5 60E 120E 150E (JJAS Regression IOD) 120E 150E (JJAS Regression IOD) 90E »20E 150E -2 -1.6-1.2-0.8-0.4 0 0.4 0.8 1.2 1.6 2 Figure 2.10: The linear regression o f observed June-September (JJAS) Nino3.4 (-5°S-5°N, 120°-170°W) and IOD ((-10°S-10°N, 50°-70°E) - (-10°S-0°, 90°-l 10°E)) SST indices with June-September (JJAS) observed and simulated precipitation and 850 mb winds, (a, b) observation (CMAP/NCEP), (c, d) CAM4 and (e, f) CAM5. 59 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon SAM and NIN03.4 0.60 - SAM and IOD 0.60 - ■ < ;CAM4 -■------- >AIR .■......... ; CMAP- 0.30 0.30 - | 0.00 f 0.00 o -0.30 -0.30 - -0.60 - -12 8 SST Lead ■4 0 4 8 SST Lag 12 -12 -8 -4 0 SST Lead 4 12 8 SST Lag Figure 2.11: Lag-lead correlation o f monthly mean precipitation with (a) Nino3.4 (-5°S-5°N, 120°-170°W) and (b) IOD ((-10°S-10°N, 50°-70°E) - (-10°S-0, 90°-110°E)) indices. Area averaged (0°-40°N, 55°-100°E) time series (ISR) o f observed CMAP (solid black line) precipitation, observed All-India Precipitation (AIR - dashed black line) time series and simulated (CAM4, solid blue line and CAM5, solid red line) time series are correlated with observed Niiio3.4 and IOD SST indices. The month with negative (positive) sign indicate that SST leads (lags) the ISR with maximum lead o f 12 months (1 year). Month 0 and 12 indicates June whereas month 4 and 8 correspond to February and October (minus sign for previous month). Correlations are calculated using a 5 month sliding window. 60 Ph.D. Dissertation: University of Northern British Columbia CAM4_CLIM - OBS 60E Precipitation 180 90E 150W 120W 90W Precipitation CCSM4 - OBS 60E -8 -7 90E -6 -5 120E -4 -3 150E -2 -1 180 0 1 150W 2 3 120W 4 5 90W 6 7 8 Figure 2.12: June-September (JJAS) mean differences between simulation and climatology (CMA) for: (a) CAM CLIM (CAM4 climatology run) and (b) CCSM4. Shading corresponds to the precipitation difference in mm/day. 61 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon CCSM 4 (JJA S M ean) 60E 90E 2 4 6 120E 8 10 12 14 150E 16 Figure 2.13: Same as Figure 2.3 but for CCSM4. Precipitation (shaded) in mm/day and 850 mb winds (vectors) in m/s. 62 Ph.D. Dissertation: University of Northern British Columbia O B S S p ectu rm V a ria n c e 40 - C C SM 4 S p ectu rm 40 - •IOD •NIN03M 8 30 - 30 - 20 - > - - 10 - 10 - 0.00 20 - 0.02 0 .0 4 0.06 0.08 F req u en cy (cycles/m onth) 0.10 0.0 0 0 .0 2 0 .0 4 0 .0 6 0 .0 8 0.1 0 F re q u en c y (cycles/m onth) Figure 2.14: The variance power spectrum for (a) observed SST (HadSST) and (b) CCSM4. Nino3.4 (-5°-5°N, 120°-170°W) SST index is in a solid red line and IOD ((-10°S-10°N, 50°-70°E) - (-10°S-0°, 90°-l 10°E)) index is in a solid blue line. 63 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon CCSM4 (WAS Regression Nino3.4) 60E 90E -2 -1.6 -1 2 -0.8 -0.4 120E 0 0.4 0.8 1.2 150E 1.6 2 SAM an d N IN 03.4 CCSM4 0.60 - ■AIR 0.30 ■CMAP 0.00 -0.80 - -12 -8 S S T Lead -4 0 4 8 12 S S T Lag Figure 2.15: The linear regression (Top) o f CCSM4 June-September (JJAS). Nino3.4 (-5°S5°N, 120°-170°W) SST index with June-September (JJAS) simulated precipitation and 850 mb winds. Lag-lead correlation (Bottom) o f CCSM4 monthly mean precipitation with CCSM4 Nino3.4 (-5°S-5°N, 120°-170°W) SST index. Area averaged (0°-40°N, 55°-100°E) time series (ISR) of observed CMAP (solid black line) precipitation and observed All-India Precipitation (AIR - dashed black line) time series are correlated with observed Nino3.4 SST index. Correlations are calculated using a 5 month sliding window. 64 Ph.D. Dissertation: University of Northern British Columbia 40N S S T (C ) D JF C C S M 4 20N 20S 60E 40N 90E 120E 150E 180 150W 120W 90W S S T (C ) MAM C C S M 4 20N 20S 60E 40N 90E 120E 150E 180 150W 120W JJA CC SM 4 90W S S T (C ) M r 20N 20S 60E 40N 90E 120E 150E 180 150W 120W SO N CCSM 4 20N v 90W S S T (C ) ,£fr : * i & 20S 60E - 4 - 3 90E 120E 150E - 2 - 1 0 180 150W 120W 1 2 3 90W 4 Figure 2.16: Seasonal SST differences between CCSM4 and observation (HadSST) for a thirty year mean climatology during: (a) December-February (DJF), (b) March-May (MAM), (c) June-August JJA and (d) September-November SON. The shading corresponds to model SST bias in °C (red for warm SST and blue for cold SST). 65 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon (J J A S D ifferen ce) C A M 4_PO P - O B S 40N 1--------- 1......... l ' 'V 1 \ ’ [3 20N - 20S 60E 90E 120E CA M 4 P O P - C C S M 4 40N 150E (J J A S D ifferen ce) CF 20N - I ---♦ 20S 60E -8 -6 90E -4 -2 120E 150E 8 Figure 2.17: June-September (JJAS) differences o f (a) CAM4 POP (simulation forced with the predicted SST o f CCSM4) from observation (CMAP/NCEP) and (b) from CCSM4. Shading corresponds to the difference in mm/day whereas vector represents 850 mb winds in m/s. 66 Ph.D. Dissertation: University of Northern British Columbia Arabian Sea (AS) Bay of Bengal (BoB) 30.0 30.0 29.0 - 29.0 - 28.0 28.0 - 27.0 - 27.0 CO CO 26.0 - 26.0 - 25.0 - CCSM4 TR 25.0 - 24.0 - CCSM4 24.0 - ■OBS 23.0 J F M A M J J A S O CAM4 POP AS BOB N 23.0 O J 40N F M A M J J A N D CAM4_POP_AS m ton - 60E -6 -5 -4 -3 -2 -t 0 1 2 3 4 5 6 ME -6-5-4-3-2-1 0 1 2 3 4 5 6 Figure 2.18: Annual cycle o f observed SST (HadSST, solid black line) and simulated SST (CCSM4, blue dotted line and CCSM4 TR, red dotted line) over (a) Arabian Sea (40°-80°E, 7°-30°N) and (b) Bay o f Bengal (80°-100°E, 7°-30°N). (c) June-September (JJAS) mean precipitation difference of CAM4_AS_BoB (CAM4 climatology SST run with SST modification in both Arabian Sea and Bay o f Bengal) from CAM4 CLIM (CAM4 climatology SST run), (d) Same as (c) but for CAM4 AS run (with SST modification in Arabian Sea only). The shading corresponds to the difference in mm/day. 67 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon Chapter 3 Further Investigation of ENSO-Monsoon relationship in climate models: A new validation strategy Islam S, Y Tang and P Jackson (2015), Further Investigation o f ENSO-Monsoon relationship in climate models: A new validation strategy, in review process, to be submitted to Atmosphere-Ocean Journal This manuscript is reformatted to serve as Chapter 3 in the thesis. 68 Ph.D. Dissertation: University of Northern British Columbia 3.1 Introduction The relation between ENSO and SAM summer precipitation is one o f the earliest observed teleconnections in global climate studies (Walker 1924). As discussed in Chapter 2, SAM tends to experience a below (above) normal monsoon during an El Nino (La Nina) or the tropical eastern Pacific Ocean warm (cold) event (Sikka 1980; Ropelewski and Halpert 1987). In the last few decades, many efforts have been made to understand the ENSO-monsoon relationship including its physical mechanisms (Mooley and Parthasarathy 1984; Shukla and Paolinol983; Webster and Yang 1992; Ju and Slingo 1995 and many others). The ENSOSAM relationship can be explained by the variation o f the Walker circulation (Shukla and Wallace 1983; Palmer et al. 1992; Soman and Slingo 1997). For example, during El Nino, the ascending branch of the Walker circulation moves eastward due to the abnormal warming in the central and eastern Pacific and accounts for below normal precipitation over the SAM region. Apart from the remote impact o f the Pacific Ocean, either through atmospheric circulation or oceanic currents, on the SAM, the Indian Ocean can also influence the SAM variability (Boschat et al. 2012). The Indian Ocean affects SAM variability in many different ways particularly through the IOD positive and negative modes. Its complex linkage with SST in Pacific Ocean can also modulate the SAM variability which can be seen in the frequent co­ occurrence o f IOD and ENSO events. Several studies have investigated the effect o f Indian Ocean in modulating the ENSO-Indian monsoon teleconnection (Webster et al. 1999; Ashok et al. 2001) and have shown that the role o f air-sea interaction in the Indian Ocean related to the monsoon dynamics is crucial for the ENSO-monsoon teleconnection. 69 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon Many studies have investigated the link and feedback between the Indian Ocean SST and ENSO in observation or in models (Saji et al. 1999; Webster et al. 1999). It has been found that both Indian and Pacific Oceans SST anomalies affect SAM in different ways depending on their phases, amplitudes and interactions. For example, Schott (2009) found that the impacts of both the Pacific and Indian Oceans on SAM are complicatedly linked due to the frequent co-occurrence o f ENSO and IOD events. Krishnamurthy and Shukla (2008) have also shown that ENSO and the Indian Ocean variability may act together or against each other in certain years. As ENSO is a key source o f SAM variability, a realistic ENSO-Monsoon relationship is one of the prerequisites of a climate model for reliable monsoon simulations and predictions. Before using climate models for SAM prediction, the systematic errors in simulating the ENSO-Monsoon relationship must be addressed, since a poorly represented tele-connection consequently limits the prospects for seasonal monsoon prediction. Studies such as Sperber and Palmer (1996) and Wang et al. (2004) investigated many AGCMs, forced with prescribed SST, and showed that the poor interannual variations o f SAM precipitation are mainly due to an incorrect SST-precipitation teleconnection. It is reported that the models with better representation o f the ENSO-SAM teleconnection result in the best simulation of the mean climate, revealing the link between model climatology and interannual variability. Despite the fact that some coupled models can produce a realistic simulation for many variables, the monsoon variability and its teleconnections to ENSO is not properly simulated in model due to poor representation o f the air-sea interactions and SST bias (Bollasina and Nigam 2009). Many coupled CGCMs, where the sea surface is not prescribed but evolves 70 Ph.D. Dissertation: University of Northern British Columbia naturally, feature systematic biases in their mean climate, particularly in the SST field and fail to realistically capture negative ENSO-SAM correlation (e.g. Achuthavarier et al. 2012). In a recent study, Wu et al. (2012) (WU12 here after) used observations to identify three different types o f ENSO influences on SAM precipitation variability, and to explore their possible physical mechanisms. They showed that the anomalous SAM precipitation can be induced by different influences o f east equatorial Pacific SSTs, including the indirect influence o f the SST anomalies o f the preceding winter (DJF-only), the direct influence o f the SST anomalies o f the concurrent summer (JJAS-only) and the combined influence o f both the preceding winter and concurrent summer SST anomalies (DJF&JJAS). The DJFonly type reflects the years with development o f ENSO in the east equatorial Pacific region, the JJAS-only type are the years when ENSO decays from its mature phase and DJF&JJAS type are the years when ENSO changes its sign from positive (negative) to negative (positive). In all o f these different SST influences, the SAM exhibits above or below normal rainfall, suggesting its strong response to changes in east equatorial Pacific. Motivated by the WU12 analysis, which provides the basis for this study, we examine the three different ENSO-SAM relationships in atmospheric (CAM4) and coupled (CCSM4) model simulations using the WU12 methodology. In Chapter 2, both climate models are validated for their mean monsoon climatology and interannual variability including the brief analysis o f their ENSO-SAM teleconnection. In this chapter, we extend the ENSO-SAM teleconnection analyses by characterizing the models’ performance in the different ENSOSAM types i.e. DJF-only, JJAS-only and DJF&JJAS. 71 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon The primary objective o f this work is to investigate which type o f ENSO-SAM teleconnection is better represented in simulations with and without air-sea coupling, and how these teleconnections are affected by SST bias (in the coupled model). We emphasize the processes connecting ENSO to the SAM in simulations o f each ENSO-SAM type by highlighting each model’s systematic biases. We also investigate the role o f the Indian Ocean particularly the northern Indian Ocean to explore how it contributes toward anomalous SAM rainfall. This is explored by isolating northern Indian Ocean from ENSO by means o f CAM4 idealized experiments by specifying prescribed SST anomalies in the tropical Pacific Ocean and climatology SST in the Indian Ocean. It is expected that this kind o f models' validation will be helpful in their continuous evolutions toward better simulations which indirectly leads to a better model forecasting capability, particularly over the SAM region. This chapter is organized into three sections. Section 3.2 describes the methodology used for making composites, the simulation data and the idealized experiments. Section 3.3 investigates the different ENSO-SAM relationships in CAM4 and CCSM4 simulations. Section 3.4 highlights the results o f idealized experiments and concludes the discussion. 3.2 Models, Experiments and Methodology The detailed description o f CAM4 and CCSM4 models is already documented in Chapter 2, section 2.2. A brief description will be provided here. The 32-year (1979 to 2009) CAM4 control ensemble simulations, forced with observed prescribed SST and sea ice data (HadSST, Reynolds et al. 2002), are conducted at 1.9° x 2.5° horizontal resolution grid using the finite volume dynamical core with 26 vertical levels and a hybrid terrain-following coordinate system. To account for the uncertainty in simulations, 72 Ph.D. Dissertation: University of Northern British Columbia an ensemble strategy is used for the CAM4 experiments by using 6 hourly lagged initial conditions. This allows us to separate the “ SST-forced” (or external) response. Ten ensemble members are constructed using the time lag ensemble strategy and the ensemble mean is used in all the analyses of CAM4 control simulations. In CCSM4, the last 50 years from a 100-year simulation, forced with present day climatological boundary conditions, is used for coupled simulation composites. The first fifty years are discarded as the equilibration time for CCSM4 model. The horizontal resolution o f 1.9° x 2.5° and finite volume grids are used for CCSM4 atmospheric and land models, whereas the ocean and ice models share the same 1° x 1° resolution with a displaced pole grid. Two different 32-year (1979 to 2009) sensitivity experiments are performed by CAM4 for two ideal scenarios, namely CAM4_ENSO_IO and CAM 4ENSO. In the CAM4 ENSO IO experiment, the observed SSTs (HadSST) are used to force the CAM4 in both the Indian and Pacific Oceans whereas the climatology SST is used elsewhere. In the CAM4 ENSO experiment, the observed SSTs are only used for the Pacific Ocean while keeping the climatological SST elsewhere. To reduce the uncertainties, ten ensemble simulations are performed for both idealized experiments using the same time lag ensemble strategy as used for the CAM4 control simulation. Observed precipitation data from the Climate Prediction Center (CPC) Merged Analysis o f Precipitation (CMAP; Xie and Arkin 1997) are used for comparison. National Centers for Environmental Prediction (NCEP; Kistler et al. 2001) reanalysis data are used to validate 73 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon winds. Observed SST (HadSST) data (Reynolds et al. 2002) are used for calculating observed SST indices and to validate coupled model SST. Following the WU12 methodology, for constructing composites, the years exhibiting different types o f ENSO-SAM impacts are selected based on the year-to-year variations of SAM and Nino3.4 SST indices in the preceding winter and concurrent summer. The SAM intensity is measured by its area averaged precipitation over 5°-25°N and 60°-100°E. The anomalous SAM is characterized based on the criteria that the magnitude o f its rainfall anomalies exceeds 0.5 standard deviation above the mean in summer during an ENSO year. The area averaged anomalies of SST are calculated over -5S°-5°N and 120°-170°W for Nino3.4 (ENSO) and 5°-25°N and 60°-100°E for northern Indian Ocean. The ENSO years are selected based on the criteria that five consecutive 3-month running mean SST anomalies in Nino3.4 region exceed 0.5°C. Based on the anomalous SAM rainfall criteria discussed above, different years are selected in observation and simulation. The bimonthly composite of selected years are made for each type o f ENSO-SAM teleconnection using the WU12 methodology, as outlined below. Throughout the text, DJF and JJAS months correspond to winter and summer season and MAM and SON reflects spring and fall. When required, the physical interpretation o f the composites based on observation data is mostly adopted from W U12. DJF-only: A positive (negative) preceding DJF Nino3.4 SST anomaly is followed by a positive (negative) SAM anomaly with the requirement that the concurrent JJAS Nino3.4 SST anomaly is not opposite o f the SAM anomaly. This is to distinguish it from the DJF&JJAS type mentioned below. In this type, the Nino3.4 SST anomalies decrease or 74 Ph.D. Dissertation: University of Northern British Columbia remain unchanged from preceding DJF to concurrent JJAS, indicating that the ENSO either decays or persists. JJAS-only: A positive (negative) concurrent JJAS Nino3.4 SST anomaly is accompanied by a negative (positive) SAM anomaly with the condition that the preceding DJF Nino3.4 SST anomaly is not o f the same sign as the SAM anomaly. That results in Nino3.4 SST anomalies that increase or remain unchanged from preceding DJF to concurrent JJAS, indicating that the ENSO either develops or persists. D JF& JJA S: The Niiio3.4 SST anomaly switches from positive (negative) in preceding DJF to negative (positive) in concurrent JJAS and the SAM anomaly is positive (negative). In this type, the ENSO switches its phase during the half year period. In this study, the anomalous SAM years in observations (CMAP) are the same as those in WU12 i.e. 1983, 1984, 1985, 1989, 1999 and 2001 for the DJF-only type, 1982, 1996, 2002 and 2004 for the JJAS-only type and 1988, 1998 and 2007 for the DJF&JJAS type. In CAM4 simulations, which are forced with monthly observed SST, we consider the same years as anomalous SAM years although the magnitude o f the SAM index does not meet our criterion at least o f 0.5 standard deviation in some years. For CCSM4 model simulation, both anomalous SAM and ENSO years are identified based on the criteria described above, since the coupled model does not include any observed information in its simulation. The investigation o f SAM and Nino3.4 indices in coupled simulations revealed that there are 6 DJF-only type, 5 JJAS-only type and 5 DJF&JJAS type years in CCSM4. 75 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon 3.3 Results and Discussion We first focus our attention on the observed and simulated lag-lead correlation o f the SAM index with Nino3.4 and northern Indian Ocean indices. Figure 3.1 shows the SAM relationships with ENSO and northern Indian Ocean in observation and simulations. Observed CMAP (solid black line) and CAM4 (solid blue line) SAM rainfall indices are correlated with observed Nino3.4 and northern Indian Ocean SST indices whereas CCSM4 (solid red line) SAM rainfall indices are correlated with the simulated Nino3.4 and northern Indian Ocean SST indices. As discussed in Chapter 2, the observations show that strong negative correlations can be found between the SAM summer rainfall and the SSTA from the concurrent summer to the following winter, representing a weak (strong) SAM in El Nino (La Nina) developing years. In the CAM4 model, the negative correlation relationship is quite weak reflecting the lack o f correct teleconnection o f SSTA to SAM summer rainfall. Actually the CAM4 simulations reflect an opposite relationship between ENSO and SAM during a boreal summer, with very small correlation coefficients. It also shows there is no relationship between SSTA in the late spring and the SAM summer rainfall, which is common in many models that have difficulties in capturing summer precipitation, due in part to monsoon “ spring prediction barrier” (Webster and Yang 1992). In case o f the CCSM4 model, the negative correlation is captured comparatively better than CAM4 but still is quite different from observations. No significant conclusion can be made regarding the spring predictability barrier in CCSM4. In Figure 3.1b, the lag-lead correlation o f SAM summer rainfall with northern Indian Ocean is computed using observed data. A weak positive correlation can be found between the SAM summer rainfall and the SSTA during late spring and summer whereas negative correlation 76 Ph.D. Dissertation: University of Northern British Columbia can be found between the SAM and SSTA from the concurrent summer to the following winter. The negative correlation o f SAM summer rainfall to the SSTA o f the following winter suggests a negative feedback o f SAM on the northern Indian Ocean. The negative feedback features enhanced (reduced) surface winds and evaporation inducing negative (positive) SST anomalies (Wu and Kirtman, 2004). The negative feedback o f SAM is not reproduced in CAM4 and CCSM4 as they only show positive correlation. The CAM4 shows more positive values of correlation as compared to CCSM4 due to lack o f an air-sea interaction in the Indian Ocean. In the CAM4 model, such a negative feedback is suppressed because the SST cannot respond to atmospheric changes. The lag-lead result suggests that neither model can represent the ENSO-SAM interaction correctly. However, this lag lead correlation is based on all the years o f model simulations. How the model simulates different types o f ENSO-SAM teleconnection (as mentioned in the introduction) and which type o f relationship is better simulated by the model can lead to a more reliable validation o f a model. We therefore further investigate the ENSO-SAM relationship based on its different types, and examine the performance o f both models for each type separately based on the analysis of composites. 3.3.1 Composite Analysis We follow the procedure discussed in WU12 in constructing the composite. For each ENSOSAM4 type, composites are constructed by averaging the years mentioned in section 3.2. To simplify the analysis and spatial representation, the composite o f each ENSO-SAM4 type is constructed to reflect positive anomalies o f SAM summer rainfall i.e. SAM anomalies are reversed (by multiplication o f -1) in the years when it is negative and grouped together with 77 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon the years when it is positive. Although WU12 discussed the physical mechanism o f different types of ENSO-SAM relationships using observations, it is important to extend this analysis to model data and to explore the models ability in characterizing each type o f ENSO-SAM relationship. 3.3.1.1 DJF-only Composite Two month composite anomalies o f observed SST for the DJF-only type are shown in Figure 3.2 (1st column) for Dec-Jan (DJ), Feb-Mar (FM), Apr-May (AM), Jun-Jul (JJ), and Aug-Sep (AS). This composite shows the yearly evaluation o f SST from the preceding winter to the concurrent summer. The evolution o f ENSO decay is characterized by a gradual decrease in SST from its mature phase in the preceding DJ to a weakened phase in the concurrent AS in the equatorial Pacific. Opposite to the ENSO decay phase, the positive SST anomalies in northern Indian Ocean are developed from the preceding DJ to concurrent JJ, and then decreased afterward. The accompanying bimonthly rainfall variation can be seen in Figure 3.3 (column 1) depicting the decay o f anomalous rainfall over the equatorial Pacific consistent with the decay o f ENSO in east equatorial Pacific while the rainfall anomalies over the SAM region grow significantly in summer. Over the SAM region, rainfall anomalies behave in a dipolar form over the Indian Ocean with different signs between the east and west in the preceding DJ which change from north to south in FM to AM. Above normal rainfall develops over the Arabian Sea in JJ spreading over the Indian subcontinent and the Bay o f Bengal in AS. The composite bimonthly rainfall variation o f CAM4 is shown in Figure 3.3 (column 2). CAM4 reproduces the rainfall transition over the east Pacific reasonably well whereas 78 Ph.D. Dissertation: University of Northern British Columbia significant disagreement can be seen over the western Pacific and Maritime continent particular in DJ to AM. The model response over the Maritime continent is quite weak. The overall simulation in DJF-only type is reliable as the large scale features are similar to the observed rainfall anomaly decay over the tropical Pacific and the increase o f rainfall over the Indian subcontinent in late monsoon (AS). CCSM4 bimonthly composites o f rainfall are shown in Figure 3.3 (column 3) whereas its simulated SST composites are presented in Figure 3.6 (column 1). In comparison with observations and CAM4, the CCSM4 rainfall composite exhibits large disagreements over different regions. The double ITCZ is simulated over the tropical Pacific Ocean along with significant differences in rainfall anomalies over the Indian Ocean in the preceding winter to early spring. The double ITCZ issue was also seen in the analysis o f Chapter 2 (Islam et al. 2013). The large differences in CCSM4 rainfall composites are mainly due to the systematic errors in CCSM4 SST simulation (Figure 3.6 (column 1)) degrading SST atmospheric response to rainfall. The ENSO response in CCSM4, to anomalous SAM, is quite strong during its mature phase in winter. CCSM4 exhibits poor skill in spring (FM-AM) which is probably due to the spring barrier which is seen in many coupled models. Over all the anomalous SAM in CCSM4 is quite weak in early summer and comparable to observation in late summer. The CAM4 and CCSM4 comparison for this type (in the ENSO decay phase) reveal that the CAM4 model is better than CCSM4 in reproducing the anomalous rainfall, although both have significant biases. 3.3.1.2 JJAS-only Composite Figure 3.2 (column 2) represents the bimonthly SST composite for the JJAS only type. The 79 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon developing phase of ENSO can be seen as the negative SST anomalies increase in the east equatorial Pacific from AM to DJ. The SST anomalies in the northern Indian Ocean, being weak in spring, change to strongly negative in late summer (AS) and DJ. The corresponding observed rainfall anomalies are shown in Figure 3.4 (column 1), with negative anomalies over the equatorial Pacific typical o f those in La Nina years. These negative equatorial anomalies propagate to the eastern Pacific from JJ to DJ whereas positive rainfall anomalies increase over the Maritime continent from JJ to ON. Positive rainfall anomalies spread over the Arabian Sea, the Indian subcontinent, and the Bay o f Bengal in early summer (JJ). In early fall (ON), negative rainfall anomalies begins over the tropical Indian Ocean and mature in the DJ. In CAM4, although the equatorial rainfall is well captured, significant disagreements in JJ rainfall anomalies over the northern Indian Ocean and western Indian subcontinent can be seen in Figure 3.4 (column 2). The CAM4 response is quite dry compared to observations over the SAM region. In the CCSM4 model, Figure 3.4 (column 3), significant difference in rainfall patterns mainly over whole tropical belt o f Indian and Pacific Oceans are seen. These are mainly induced by the simulated SST bias. Over the northern Indian Ocean and Indian subcontinent, the CCSM4 rainfall anomalies are better captured compare to those in the CAM4 simulations. 3.3.1.3 DJF&JJAS Composite In the DJF&JJAS composite, ENSO switches from positive in winter to negative in summer (Figure 3.2 (column 3)). In the whole year evolution, the tropical Indian Ocean shows consistent positive SST anomalies. The accompanying rainfall anomalies (Figure 3.5 80 Ph.D. Dissertation: University of Northern British Columbia (column 1)) over the tropical Pacific follow the switch o f east equatorial Pacific SST anomalies by being positive in winter and negative in late summer (AS). CAM4 and CCSM4 simulations for this type are presented in Figure 3.5 (columns 2 and 3 respectively). CAM4 better reproduces the above normal rainfall anomalies over the northern Indian Ocean and Maritime continent in JJ and AS, whereas the CCSM4 model reproduces quite poor rainfall response over the same region. This finding shows that this type o f the ENSO-SAM relationship is not properly captured in the CCSM4 simulation as compared to the uncoupled CAM4 model. The above analyses are based on bi-monthly composites that reveal the yearly evolution o f SST and rainfall. To analyze the SAM average JJAS (summer) response, we investigate the average composite rainfall anomalies for June, July, August and September months in all the three types. A comparison o f JJAS means is shown in Figure 3.7 for the DJF-only, JJAS-only and DJF&JJAS types based on CMAP, CAM4 and CCSM4 rainfall. This analysis can conclude our discussion by showing that in all three types, CAM4 simulations are better for DJF-only and DJF&JJAS type whereas CCSM4 shows good simulation only in JJAS-only type, particularly over the SAM region. This means that only the concurrent response o f SST is well represented in CCSM4. Comparison o f the JJAS mean composites, o f Figure 3.7 with Figures 3.2, 3.3 and 3.4 indicates that in observations, the late summer (AS) rainfall anomalies contribute more towards the total mean SAM summer rainfall in DJF-only and DJF&JJAS types, whereas in the JJAS-only type, early summer (JJ) anomalies largely contribute to total JJAS mean rainfall. In these models, such contributions are not clear and therefore cannot be 81 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon differentiated due to their systematic errors over the SAM region. 3.3.2 Temporal Evolution of SST and SAM Rainfall We now explore the temporal evolution o f the ENSO-SAM relationship using area averaged indices. The SST index is calculated over the Niiio3.4 and northern Indian Ocean region while the SAM index is calculated by averaging the observed or simulated rainfall over the SAM region. Figure 3.8 shows the normalized 3-month running mean anomalies o f Nino3.4 SST, northern Indian Ocean SST and SAM rainfall for all the types for both observed (CMAP and HadSST) and simulated (CAM4 and CCSM4) rainfalls and SST indices. As discussed in WU12, in the observations (Figure 3.8a, left column), positive Nino3.4 SST anomalies in the preceding winter are followed by positive northern Indian Ocean SST anomalies during spring through early summer in the DJF-only type, indicating the effect o f preceding east equatorial Pacific SST anomalies on SAM through northern Indian Ocean SST changes. The SAM rainfall anomalies begin increasing in June and reach a maximum in late August and September. The delay o f the SAM rainfall anomaly to the northern Indian Ocean SST suggests the positive feedback o f the northern Indian Ocean warming to the SAM region in the form o f more evaporation. The warm SST anomalies in the northern Indian Ocean in summer lead to lower level convergence and anomalous rainfall in summer (Chang etal. 2011). The CAM4 rainfall over the SAM region follows the observed pattern except smaller in magnitude in the summer. CCSM4 SST indices, shown in Figure 3.8a (right column) basically follow the observed SST variation, whereas its rainfall patterns are different. In CCSM4, the warming in northern Indian Ocean region is not strong, probably resulting in 82 Ph.D. Dissertation: University of Northern British Columbia less impact on the SAM rainfall. In the JJAS-only type (Figure 3.8b, left column), the northern Indian Ocean SST anomalies are negative in summer, leading to less favorable conditions for the SAM rainfall. In this type, negative east equatorial Pacific SST anomalies may directly induce an anomalous SAM via atmospheric circulation changes (WU12). These circulations modulate and amplify the SAM rainfall by increasing the convergence over the SAM area (WU12). In the CAM4 simulations, due to the absence of local air-sea coupling over the Indian Ocean, the atmospheric circulation cannot induce changes in the northern Indian Ocean SST which could be the reason that the CAM4 model is not properly reproducing the anomalous rainfall in this type. Air-sea coupling is important for atmospheric circulations caused by the largescale ocean temperature gradient. This can be seen in the CCSM4 (Figure 3.8b, right column), where anomalous rainfall over the SAM region is better simulated than in CAM4 in summer. Another reason for better CCSM4 simulation is that the northern Indian Ocean SST warm anomaly in summer significantly contributes toward increasing the SAM rainfall. This makes it difficult to decide if the better representation o f JJAS-only type in CCSM4 is purely due to the air-sea coupling or if it is due to the warm anomaly in CCSM4 northern Indian Ocean SST. In the DJF&JJAS type, shown in Figure 3.8c (left column), positive northern Indian Ocean SST anomalies persist from the preceding winter to summer. In this type, the ENSO switches from a warming phase in winter to a cooling phase in summer, suggesting that the persistence o f positive northern Indian Ocean SSTA is mainly due to the preceding winter positive east equatorial Pacific SST anomalies that cause the anomalous SAM. As in the JJAS-only type, the concurrent negative east equatorial Pacific SST anomalies may contribute to an 83 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon anomalous SAM via a direct atmospheric circulation change (WU12). Thus, both the preceding (via northern Indian Ocean) and concurrent (via atmospheric circulation) east equatorial Pacific SST anomalies result in an anomalous SAM rainfall (WU12). The CAM4 simulation reproduces this type remarkably well as compared to the DJF-only and JJAS-only types. This could be due to the persistent warm northern Indian Ocean SST anomalies, which are reflected in CAM4 simulations. In CCSM4 simulations, although the Niiio3.4 pattern is a partial match for the observations, the rainfall response is quite the opposite. The likely source o f CCSM4 failure to simulate the anomalous SAM summer rainfall is probably due to a decrease o f persistence positive SST anomalies in the northern Indian Ocean in spring (3.8c, right column), causing less moisture evaporation. Indeed the CCSM4 ocean response is not consistent from preceding winter to concurrent summer in this type. 3.3.3 Response of Atmospheric Circulations As discussed above in the JJAS-only and DJF&JJAS types, the anomalous SAM rainfall strongly modulates via the atmospheric circulation changes over the SAM region. We explore this in Figure 3.9 by plotting the JJAS mean composite anomalies o f velocity potential and divergent winds at 850 mb and 200 mb heights for the JJAS-only and DJF&JJAS type. Observed winds are from NCEP reanalysis data sets whereas simulated winds are from CAM4 and CCSM4. In the JJAS-only type, the anomalous SAM rainfall is induced by a direct SST forcing in the Pacific Ocean through large scale circulation changes. This can be seen in the observations for the upper level convergence and lower level divergence over the equatorial central Pacific and upper level divergence and lower level convergence over the Maritime continent, 84 Ph.D. Dissertation: University of Northern British Columbia Australia and SAM regions (Figure 3.9, row 1). In CAM4, while the low level convergence is captured over the Maritime continent, the SAM convergence zone is not reproduced which causes the poor anomalous SAM rainfall simulation in the JJAS-only type. CCSM4 shows strong convergence over the SAM region and this intensifies the anomalous SAM. In the DJF&JJAS type (Figure 3.9, row 2) as discussed in WU12, the preceding positive east equatorial Pacific SST anomalies induce positive northern Indian Ocean SST anomalies through atmospheric circulation changes. Negative SST anomalies in the east equatorial Pacific produce lower level divergence and upper level convergence over the tropical Pacific. This leads to upper level divergence and convergence at the surface causing moisture to intensify over the SAM region. CAM4 convergence and divergence zones are comparable to observations whereas CCSM4 circulations are weak in these types and cannot reproduce the observed circulation patterns. 3.3.4 Response of Humidity and Land-Sea Thermal Contrast It has been shown that land-sea heating contrast is a fundamental mechanism powering the summer monsoon circulation (Webster 1987; Young 1987). The relationship between landsea thermal contrast and SAM has been investigated in many studies (Fu and Fletcher 1985; Meehl 1994a; Li and Yanai, 1996) where they found that the intensity o f the SAM is related to the strength o f the temperature gradient due to the land-sea thermal contrast. Increasing the temperature gradient increases the intensity o f the SAM and vice versa. In this section, we compare the surface moisture and thermal contrast form observation and models in each ENSO-SAM type. Following the methodology o f WU12, the surface air humidity is averaged over the western Indian Ocean (0°-10°N, 50°-80°E), as the increased moisture in this region 85 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon may be transported to the SAM region by mean southwesterly winds and thus may contribute to a stronger SAM rainfall. The thermal contrast is represented using a 500-200-hPa thickness difference between the zones from 20°- 40°N, 50°-100°E and from 0°-20°N, 50°100°E. In Figure 3.10 (left column), negative thickness difference anomalies persist from winter to summer for the DJF-only type. Positive anomalies o f surface air humidity are present in early spring and summer in western Indian Ocean and cause an increase in the moisture availability for the SAM region. The air humidity increases due to the positive SST anomalies during late spring and early summer (as in Figure 3.9). Although the CAM4 atmospheric response to thickness difference is quite consistent with observations, its moisture content peaks in late winter instead o f summer. This means that the anomalous SAM summer rainfall is underestimated in CAM4 while its thermal contrast is negative over the SAM region. The positive moisture availability is found to have a reliable performance for the DJF-only type. In CCSM4 (Figure 3.10, right column), the thermal contrast difference remains negative throughout the year similar to observations, but a significant underestimation can be seen in moisture availability in spring and early summer (as compared to 3.10, left column). This demonstrates the lack o f its ability to simulate anomalous SAM (as seen in spatial patterns previously). In the JJAS-only type, the small positive anomalies of thickness difference in summer lag the rainfall anomalies (similar to that seen in WU12), indicating that the enhanced monsoon rainfall is due to thermal contrast. Surface air humidity anomalies are small corresponding to Indian Ocean negative SST anomalies (Figure 3.1) and contribute less to anomalous SAM than the SSTA does. In CAM4, the thickness difference becomes positive in late summer 86 Ph.D. Dissertation: University of Northern British Columbia instead o f early summer, as seen in observations. The moisture content anomaly is near zero in early summer and becomes negative afterward. This is a probable source for the poor simulation o f anomalous SAM rainfall. In CCSM4, the thickness difference is negative in spring and the accompanying moisture peaks at the same time. In summer the thickness difference becomes positive with a reduction in moisture. As seen in the spatial patterns (Figure 3.4, last column), the CCSM4 shows strong anomalous rainfall over SAM region in summer. This means that the CCSM4 atmospheric circulations have a larger contribution in modulating the intensity o f anomalous rainfall than the thermal contrast does. In the DJF&JJAS type, negative thickness difference anomalies appear in winter and spring and become positive after midsummer. Air humidity anomalies are quite large and positive in western Indian Ocean from winter to early summer. Positive SST anomalies during late spring and early summer, as shown in Fig. 3.8, induced an increase in surface air humidity and thus favor a stronger SAM rainfall (WU12). The thermal contrast makes a limited contribution to the SAM late summer anomalies. In CAM4, the moisture content and thickness difference are in-phase with observation and produce good simulations. In the late summer and early fall, the moisture persists due to the lower values o f thickness difference whereas the observed moisture decreases sharply at that time. In CCSM4, the thickness difference and moisture content are almost in out of phase with maximum and minimum amplitude appearing in the early spring and in the summer, respectively. This is very different from the observations and CAM4. 87 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon 3.3.4 Idealized Experiments We now explore how the Indian Ocean controls the ENSO-SAM relationship with the help o f CAM4 idealized experiments. As the role o f the Indian Ocean in amplifying the anomalous SAM rainfall is significant in DJF&JJAS years, we focus only on this particular type. In the analyses discussed above, we have seen that the CAM4 performance is much better in DJF&JJAS type as compared to the DJF-only and JJAS-only type. This is another reason to conduct the idealized experiments for the DJF&JJAS type. We performed two different ensemble experiments, namely C A M 4 E N S O I O and CAM4 ENSO, for the period from 1979 to 2009. In CAM4_ENSO_IO experiments, observed monthly SSTs are specified in both the Indian and Pacific Oceans whereas in CAM4 ENSO experiments, the observed SSTs are only prescribed over the Pacific Ocean. The annually repeating climatology SSTs, blended with observed SST, are used elsewhere in both experiments. The two experiments are designed to explore the role that the different oceans, including the Indian, Pacific, or other play in the anomalous SAM rainfall. In Figure 3.11, composites o f JJAS rainfall anomalies and 850 mb winds for the DJF&JJAS type are presented for CAM4 ENSO IO and CAM4 ENSO experiments. The results for CAM4 ENSO IO are similar to the CAM4 control simulation, shown in Figure 3.7, with some small scale refinements. This implies that the anomalous SAM is mainly influenced by Indian and Pacific Oceans only. In the CAM4 ENSO case, the absence o f the Indian Ocean influence can be characterized by a significant removal o f rainfall over the tropical Indian Ocean and Indian subcontinent. The persistence o f positive northern Indian Ocean SST anomalies induced by the preceding winter east equatorial Pacific SST anomalies is absent in 88 Ph.D. Dissertation: University of Northern British Columbia this experiment, resulting in reduced anomalous rainfall. The weak contribution from concurrent negative east equatorial Pacific SST anomalies, via direct atmospheric circulation changes, can be seen in Figure 3.12 (spatial plots). Although the negative SST anomalies in the east equatorial Pacific induce lower level divergence and upper level convergence over the tropical Pacific, their effect on the SAM region is unrealistically confined to the Maritime continent only, further reducing the surface moisture and rainfall there. This can be also seen in the temporal evolution (Figure 3.12 (bottom plot)) o f rainfall for both experiments. The rainfall in the CA M4E NSO experiment is suppressed during the summer as compared to the C A M 4 E N S O I O rainfall. These results support the physical mechanisms o f the DJF&JJAS type discussed in WU12. Also the importance o f the Indian Ocean in modulating (amplifying or suppressing) the rainfall in ENSO-SAM teleconnection can be seen in this analysis. 3.4 Summary and Conclusions This chapter presented a detailed investigation of the ENSO-SAM relationships in the CAM4 and CCSM4 models simulations. This investigation seeks to identify the strengths and weaknesses o f these models in preserving different types o f ENSO-SAM relationships. This analysis was performed following the recent work by Wu et al. (2012) where they identified three different types o f ENSO influences on the SAM from analysis o f observations. We have extended their analysis to our simulation results, and further explored the physical understanding o f particular types o f ENSO influences on SAM using sensitivity experiments. Composite analysis was performed for both model simulations and observations including spatial composite map and temporal variability o f the area average. The CAM4 simulation in the DJF-only type is better than CCSM4, and can reproduce the realistic anomalous rainfall 89 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon reasonably well over the east Pacific along with some biases over the western Pacific and Maritime continent that are particularly significant in the preceding winter and the early summer. It can also capture the realistic evolution and propagation o f rainfall anomalies from their decay over the tropical Pacific in the preceding winter to their development phase over the Indian subcontinent in the late monsoon season. In contrast, the CCSM4 shows large systematic errors including significant rainfall anomaly biases in the Indian Ocean from the preceding winter to the early spring, probably due to the errors in the SST simulation. The CAM4 model performed better than CCSM4 in DJF-only years. For the JJAS-only type where the concurrent eastern Pacific SST anomalies directly induce anomalous SAM through atmospheric circulation, the CAM4 simulations show disagreement in SAM rainfall anomalies between the observations and the CCSM4 simulations. It is found that ENSO induced warming in northern Indian Ocean via atmospheric circulation is absent in CAM4 simulations in the JJAS-only type probably due to the lack o f air-sea coupling. The CCSM4 shows strong convergence over the SAM region, intensifying the anomalous SAM. The CCSM4 results are more realistic than CAM4 in comparison with observations. It is found that the atmospheric circulations o f velocity potential and the corresponding divergent winds at 850 mb and 200 mb heights in CCSM4 contribute more than the thermal contrast in modulating the intensity o f anomalous rainfall. For the third type o f DJF&JJAS, the impact o f ENSO on the SAM is characterized by a switch o f the ENSO phase during the cycle from the warm phase in preceding winter to cold phase in the summer, amplifying the SAM rainfall. In this type, CAM4 better reproduces the realistic anomalous rainfall than the CCSM4 which has very weak simulated rainfall anomalies. The convergence and divergence zones in the CAM4 model are comparable to the 90 Ph.D. Dissertation: University of Northern British Columbia observations whereas the CCSM4 has very weak simulations in these circulations. In comparing the other two simulation types, the CAM4 better models the ENSO-SAM relationship in the DJF&JJAS type, including the realistic simulation in moisture content and thickness difference. In contrast, the CCSM4 simulates the moisture content and thickness difference out of phase with realistic evolution. The CAM4 sensitivity experiment highlights the role o f the Indian Ocean in controlling the DJF&JJAS type ENSO-SAM teleconnections. It is found that in the absence o f Indian Ocean SST, the anomalous SAM summer rainfall is suppressed in the DJF&JJAS type, suggesting that the important modulation by the Indian Ocean SST probably occurs through the preceding winter east equatorial Pacific SST forcing and the atmospheric circulations. The results presented in this chapter suggest that model biases can significantly affect the ENSO-SAM relationship. Although the air-sea coupling is important for better SAM simulation and its relationship with ENSO, the bias in the ocean modeling can significantly degrade the SAM relationship. The fidelity o f the CCSM4 model is most likely not at a level where the local air-sea feedbacks could be clearly shown to be an important component o f the forecast system that improves the seasonal predictability o f SAM. This study also suggests that the atmospheric model could be used as an alternative tool for monsoon predictions until the coupling biases are better resolved. 91 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon CCSM 4 0.8 - — — CAM4 CMAP 0.4 0.0 -0.4 - 0.8 SST Lead Conlrol SST Lag NtO 0.8 0.4 0.0 -0.4 - 0.8 Figure 3.1: Lag lead correlation o f monthly mean SAM precipitation with (a) Nino3.4 (-5°S5°N, 120°-170°W) and (b) northern Indian Ocean (NIO, 5°-25°N, 60°-100°E) indices. Area averaged (5°-25°N, 60°-100°E) time series (ISM) o f observed CMAP (solid black line) precipitation, CAM4 (solid blue line) and CCSM4 (solid red line) time series are correlated with observed Niiio3.4 and northern Indian Ocean SST indices except for CCSM4 case where CCSM4 simulated SST are used. Month with a negative (positive) sign indicate that SST leads (lags) the ISR with maximum lead o f 12 months (1 year). Month 0 and 12 indicates June whereas month 4 and 8 correspond to February and October (minus sign for previous month). Correlations are calculated using a 5 monthly sliding window. 92 S S T O J C o fflp o a ta S S T OJ C om p ocB * S S T F M C o m p o c il* S S T J J C o m p o s it t SS T JJC om pgM a S S T J J C M npoM * ME ME PJCow rpoate -1 - 0.6 0 .2 0.2 0.6 - 0.6 - 0.2 0.2 0 .6 - 0.2 0.2 0.6 Figure 3.2: Bimonthly composite anomalies of observed SST (HadSST) for DJ, FM, AM, JJ, and AS (top to bottom). The 1st, 2nd and 3rd columns represent DJF-only. JJAS only and DJF&JJAS types respectively (see text for detail). Shading corresponds to the values in C. 93 S Islam: Ensemble Simulation and Forecasting o f South Asian Monsoon CMAP DJ CompoaW (OJF-ooty) (mnVd) — < ta o c tso e too 9 ^ ~ a n d rrV s) C C S M 4 FM C o m p o s it e (m rrvd ) (m n tfd a y a n d m f r ) C C S M 4 AM C o m p o a to (m rrvd ) CAM 4 j j Com poaA e (m nW day a n d m t* ) C C SM 4 JJ Com poaN a C A M 4 A S C o n a so a r ta (ron tfd ay a n d m f t ) G A M 4 F M C o m p o a it a KM s r ............................■ 130* 190* ■ 4 CMAP A M C om pca*a 00* C o ta o w ta o w (n w tv d a y a n d n v > ) C M A PFM C om pow f OOF 'SOW (m m M a y a n d rofr 120* CM AP JJ C om poaN a m M AP A S C o m p o o tte <00 1 199W 120W a n d rrVo ) (m m tte y a n d m * } C C SM 4 A S Com poaA o (m n v d 30N & 1 2 3 4 Figure 3.3: Bimonthly composite anomalies o f precipitation for DJ, FM, AM, JJ, and AS in DJF-only type (top to bottom). The 1st, 2nd and 3rd columns represent observed (CMAP), CAM4 and CCSM4 respectively. Shading corresponds to the values in mm/day while the contour shows their significance. 94 Ph.D. Dissertation: University o f Northern British Columbia CMAP AM Com port* (JMS-onfy) (m nV day a n d n V » ) C A M 4 c n e t AM C o m p o r t * { J J A S -c n ly ) { m m tt a y a n d n V s ) C C S M 4 AM C o m p o a A e im n v d a y a n d n v * ) eoe me C C S M 4 JJ G orrn oen e 'SOW 12DW (m n J d a y a n d m / a ) (mmA$ o 90S ME CM AP JJ C o m p o rt* (m n w a y and m /e ) C A M 4 j j C o m p o r te M A P A S C o m p o s it e (m m A Jay a n d n V s } C A M 4 A S C o m p o rt* 1K> i* * e tec BOW 1*0 taow i? o w SOE IW 190W C C S M 4 A S C o m p oaA e 120V (nrnti) C C SM 4 O N C o m p os* * (m n vd ) IWW CM AP O N C o m p o rt* iw w (m m /d a y a n d rrv* i { m m t t e y a n d try t ) C M A P O J C o m p o s it e (m m A Jay a n d "**» (m n W te y a n d m f t ) m ’s o c iit f i -4 -3-2 - 1 0 1 2 3 4 -4-3 -2 - 1 0 1 2 3 4 - 4 - 3 - 2 - 1 0 F ig u r e 3 .4 : S am e as F ig u re 3.3 b u t fo r th e JJA S o n ly type. 95 1 2 3 4 S Islam: Ensemble Simulation and Forecasting o f South Asian Monsoon CMAP OJ Comport* (DJF&JJAS) BOC JO t 120E I C M A P FM C o m p o r t * (m m tta y a n d m /» ) C A M 4 c r tr i D J C o m p o r t * (D J F & J J A S ) (m m A M y a n d m/% ) 1S0W 120W ( m n V d a y a n d m /» C A M * FM C o m p o r t* (m m * d * y a n d m/% ) C C SM 4 OJ C o m p o rt* (mm/d) C C S M 4 FM C o m p o r t e (m rv d ) MN - fB L Ik . •Or 120E 150E 100 C M A P A M C o mp o r t * ME 190W 1 (m m /d a y a n d m / * ) tzOE loOE 100 C M A P JJ C o m p o rt* 1B0W» SCC SQC 10QE '50C <80 (SOW 1 C A M 4 AM C o m p o r t* (m m /d a y a n d m/% C C S M 4 AM C o m p o r t * (m m /d ) C AM 4 JJ C o m p o rt* (m w r t y a n d m / * ) C C SM 4 JJ C om p os** (m m /d ) tSOW ( m m t ia y a n d m /* X I20E M A P A S C o m p o rt* 1S0E 100 «90W - 4 - 3 - 2 - 1 0 1 2 3 4 190E 1W * (m n V d a y r td m /s) - 4 - 3 - 2 - 1 0 isoW 100 l?OW (m m /d a y * r d m/% ) C A AM A S C o m p o r t * 1 2 3 4 0OC 90C 10OE ’ 50E 100 1 C C SM 4 A S C o m p o rt* - 3 - 2 - 1 0 (m m /d ] 1 2 3 F ig u r e 3 .5 : Sam e as F igure 3.3 but fo r the D JF & JJA S type. 96 Ph.D. Dissertation: University o f Northern British Columbia S S T DJ C o m p o s* * S S T AM C o m p o s* * S S T DJ C o m p o s* * eoc *o€ S S T FM C o m p o s * * S S T j j c o m p o rt* S S T FM C o m p o s * * S S T A S C o m p o rt* S S T AM C o m p o r t s S S T O N C om p os** SS T JJ C o m p o rts m fC S C * ■oe S S T J J C o mp o s * * ia o e is o e tao S S T D J C o m p o rts A -1 - 0.6 - 0.2 0.2 0.6 1 -1 - 0.6 - 0.2 0.2 0.6 1 1 - 0.6 F ig u r e 3 .6 : S am e as F igure 3.2 b u t fo r C C S M 4 sim u lated SST. 97 - 0.2 0.2 0.6 1 S Islam: Ensemble Simulation and Forecasting o f South Asian Monsoon CMAP JJAS Composite (DJF-only) (mm/day and m /s) (mnVday and m /s ) CAM4 cntri JJA S Composite (DJF-only) CCSM4 JJA S Com posite (DJF-only) 3 • 606 906 1206 1506 180 CMAP (JJAS-only) 606 906 1206 1506 180 CMAP (DJF&JJAS) _1------ 606 120W 150W 120W (mnVday and m /s ) 1------ 906 -4 150W (mm/day and m / s ) 60 6 ty A 906 1206 1506 180 CAM4 (JJAS-only) 606 150W 120W (mnVday and m /s ) 906 1206 1506 180 CAM4 (DJF&JJAS) 150W 120W (mm/day and m /s ) l_ 1206 -3-2 -H - 1 0 1506 1 180 2 3 150W 4 120W 606 906 -4 1206 -3 -2 1506 -1 0 1 180 2 3 150W 4 120W 606 906 1206 1506 180 150W CCSM4 (JJAS-only) 606 906 1206 1506 180 150W CCSM4 (DJF&JJAS) _ j __________ , t 606 90 6 -4 120W m/s 123W m/s 1206 -3 -2 -1 0 1506 1 180 2 3 150W 4 Figure 3.7: Composite JJAS rainfall anomalies (mm/day) for the DJF-only type, JJAS-only type and DJF&JJAS type based on observed (CMAP) and simulated (CAM4 and CCSM4) precipitation. 98 129W 3 months running mean DJFonly 3 months running mean s.o DJF only 30 — NINO 2.0 2.0 1.0 - 0.0 0.0 8(8 1.0 2.0 -3.0 -3.0 o O o JJAS only JJAS only 3.0 3.0 2.0 2.0 0.0 * 5 - - 2.0 -2.0 -3.0 -3.0 3.0 2.0 2.0 - I 0.0 0.0 1.0 - - 1.0 - 2 ,0 3.0 -3,0 O Figure 3.8: Area averaged composite o f normalized 3-month running mean anomalies o f Nino3.4 SST (green curves), northern Indian Ocean SST (NIO, red curves), CAM4 or CCSM4 precipitation (blue curves) and CMAP precipitation (IMR, black curves) in (a) DJFonly type, (b) JJAS-only type, and (c) DJF&JJAS type. SST curves in left column are from observation while right column SST curves are from CCSM4 simulation. 99 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon NCEP (JJAS-only c m ) 0 306 ME ME 1M6 ISOE Y«0 l» W 1S0W MW S tt NE Ni '3 0 t 190E 0 30C 180 1S0W 1MW flOW tOW SOW 0 0 30E -1.5 CAM4 (JJAS-oNy to e ME 1HE 1ME so t toe ME *1 -0.5 180 150W 130W fOW tOW SOW 130E 1S0E 1«0 180W 120W 90W 80W S0W 2 -1.5 CCSM4 (JJASoNy c m ) SOE toe ME 1306 1S0E CC8M4 (DJF&JJAS c m ) 0 1M 1S0W 120W MW -1 ME tso e (SOOmb. mm) 190C 1M 'SOW YMW to w ixoe tM E 180 I SOW 120W ! NCEP (DJF&JJAS c m ) 0 0 0 1 ME 1.5 3QE IOC 0 0 0.5 ME 1 I0E MC <200m b, m m i 130E ISOE ME 1.5 180 'SOW 1MW MW OOW SOW 0 SOI 80E ME SOW 0 (200mb, mm) 180E ISOE CCSM4 (JJAS-only c m ) 0 tOW 1M 1MW 120W 80W MW SOW 0 2 <850mb, m/t) (850mb, m/») I0W SOW 0 (200m b, rrVs) 2 CAM4 (DJF&JJAS c m ) 0 -0.5 0.5 ME CAM4 (JJAS-only cate) («50mb,mfe) CAM4 (DJF&JJAS c m ) 0 (850mb, mfe) (850mb, nVt) •on-ha^TTtrrt 306 MW 0 -2 0 NCEP (JJAS-only c a m ) SOW NCEP (DJF&JJAS c m ) 0 (SSOmb, mIt) (2 0 0 m b , m /s) ISOE ISOE 1M iMW 120W MW CCSM4 (DJF&JJAS c m ) MW SOW 0 (200mb,m/t) CAT15 0 Mt MC >01 120t I M t W 1W * I20W W ■ a -2 -1 -0.5 -1.5 I 0 0 0.5 IOC 1 H i m 1.5 '20k 1500 1M 190W I20W m «M MW 0 2 Figure 3.9: Observed (top), CAM4 (center) and CCSM4 (bottom) JJAS mean composite anomalies o f velocity potential and the corresponding divergent winds at 850 mb height (left column) and at 200 mb height (right column) for the JJAS-only and DJF&JJAS types. Shading corresponds to the velocity potential in 106 s'1 and vectors in m/s. 100 Ph.D. Dissertation: University of Northern British Columbia 3 months running mean 3 months running mean DJF only 30 3.0 WO (Q2m, CAM4) WO (Q2m, c t» ) cam4) 2.0 — (02m, CCSM4) 2.0 - • I MKl Mo k, o t e ) 1.0 00 m 2.0 20 -3.0 30 o O JJAS only JJAS only 3.0 30 2.0 2.0 0.0 IB E oo •1 0 1 1.0 • -20 -30 -30 DJF&JJAS 3.0 3.0 2.0 2.0 00 0.0 1.0 -20 -2.0 -3.0 3.0 o O o Figure 3.10: Composites o f normalized 3-month running mean anomalies o f surface (2 m) air humidity over the region o f 0°-10°N and 50°-80°E (solid curves) and 500-200-hPa thickness difference (10 m) between area 20°-40°N, 50°-100°E and 08°-20°N, 50°-100°E in (a) DJF-only type, (b) JJAS-only type, and (c) DJF&JJAS type. Left column is for CAM4 and observation while right column shows CCSM4 simulated curves. 101 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon CAM4_ENSO_IO (DJF&JJAS) (mm/day and m /s) 30N 0 30S 60E 90E 120E 150E - 4 - 3 - 2 - 1 0 1 180 2 3 150W 120W 4 CAM4_ENSO (DJF&JJAS) (mm/day and m /s ) 30N - o - 30S 60E 90E 150E 120E -4 -3 -2 180 150W 120W -1 Figure 3.11: Composite JJAS rainfall anomalies (mm/day) and 850 mb winds for the DJF&JJAS type based on C A M 4 E N S O I O and CAM ENSO experiments. See text for detail o f the experiments. 102 Ph.D. Dissertation: University of Northern British Columbia CAM4_ENSO_IO (DJF&JJAS) (850mb, m/s) CAM4_ENSO_tO (200mb, m/i) CAM4_ENSO (DJF&JJAS) (850mb, m/s) CAM4ENSO (200mb, m/s) - jP-S^epFx V I*Ii , Eq. (1) can be written as X ( t ) + A X ( t ) = F [ X ( t' ) + ] (2) Subtracting Eq. (1) from Eq. (2), we have = (3) where the linear operator R in Eq. (3) is the first-order derivative o f F with respect to X ft (at the time o f 1 ). It is often called the propagator o f Eq. (1) and gives the time evolution o f the dynamical system by representing perturbation growth matrices. The singular vectors of the system, which are the perturbations that amplify maximally over the time period ( t —t 1), are the eigenvectors (E) of R 7R with the largest real part (e.g. Buizza and Palmer 1995), where R T is the transpose of ^ . Thus the SV can be obtained by two methods: the eigenvector analysis o f the R 7R matrix or singular value decomposition (SVD) analysis o f R. Mathematically it can be shown that: R (t,t')E ,= ^Sx ^ 109 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon Where ^ represents first singular value and ^ is first SV. Hence can be derived by applying the propagator R(t,t') to the initial pattern ^ as shown by (4). is called the final pattern. Here 5, is the left vector o f SVD of R . The calculation o f R is discussed below. 4.2.2 Calculation of R and Application to the SAM Forecast System For calculating R o f the dynamical system (1), a perturbation variable (denoted by T ) at the initial time and the target variable (denoted by A ) used to measure forecast errors, are selected. By definition, the leading SV o f A indicates what kind o f uncertainty in Tp can lead to the fastest error growth for predicting A . The detailed procedure for obtaining R is similar to that described in Kleeman et al. (2003) and is summarized as: 1. An ensemble of 10 forecasts with lead time o f 4 months (June, July, August and T September, named as JJAS) is constructed by randomly perturbing the initial p field \j/ (t) with 10 “very small” random patterns. The ensemble mean of A is denoted by oV . 2. Each o f the leading three correlation EOF modes e' o f 1*2,3) ^ a multiplication factor o f 0.1 to ensure linearity) to the initial condition described in step (1) and a new ensemble o f 10 forecasts is produced. The corresponding ensemble vp ( f ) mean o f A is denoted by ' . 3. A reduced-state space matrix version r‘J o f the propagator R is then obtained taking the difference o f both the ensembles. Mathematically it can be represented as _____ __ __ 3 Re, = A¥ ( /) = y¥ l (/) - xVa(0 = Y Jr„e , + residual H (5) 110 Ph.D. Dissertation: University of Northern British Columbia The residual in equation (5) is generally very small and can be ignored (Kleeman et al. 2003). The climatically relevant singular vectors are thus obtained by SVD analysis o f R as aforementioned. These singular vectors are then projected back to real T space using the EOF basis vector expansion. In implementing this method, it is important to note that ensemble experiments are carried out to filter the weather noise by averaging in order to extract fast error growth due to climatically relevant instabilities. Another advantage o f the CSV method is that the choice o f the analysis domain and optimization lead time can be made after the ensembles have been completed which allows one to explore the sensitivity to different choices without further model experiments. 4.2.3 Experimental Setups The models used in our seasonal forecast system are CAM4 and CCSM4. The ability o f CAM4 and CCSM4 to simulate the summer monsoon has been explored in Chapter 2 and 3. It is found that CAM4 overestimates the monsoon rainfall over most o f the SAM region when compared to observation and CCSM4. The overall simulations o f CAM4 reasonably captured the monsoon mean climatology (Islam et al. 2013), allowing its use to study predictability. As discussed above, to implement the CSV method, three leading modes o f the correlation EOF are used to perturb the initial conditions. Figure 4.1a and 4.1b show the first three correlation EOF modes for SST over the Indian and Pacific Oceans. The major reason for SV analysis over the two Ocean domains is to explore the individual effect o f the Indian and 111 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon Pacific Ocean on the error growth of SAM, respectively. As reported in many studies, both o f these Oceans can significantly affect SAM and play a key role in SAM variability (e.g. Meehl and Arblaster 2002; Annamalai and Liu 2005; Wang et al. 2009; Chowdary et al. 2010; Kosaka et al. 2012). Leading EOFs are calculated for the month o f June using the observed SST data (Hadley Centre SST, Rayner et al. 2003) for the period 1980-2009. In case o f the Indian Ocean (Figure 4.1a), 1st and 2nd leading EOF mode account for 39% and 12% o f the total variance. EOF1 shows the Indian Ocean Basin (IOB) Mode and EOF2 depicts the east-west Indian Ocean dipole (Saji et al. 1999, IOD) mode. The former is closely linked to ENSO, whereas the latter occurs with and without ENSO (Pokhrel et al. 2012). The two modes differ not only in spatial structure, but also in their seasonal dependency. In Figure 4.1b, the 1st leading EOF over the Pacific domain represents the ENSO mode. The leading mode shows 30.7% explained variance over the Pacific Ocean. The spatial pattern associated with the warm phase o f ENSO consists o f positive SST anomalies across the eastern equatorial Pacific Ocean and weaker negative anomalies over the western tropical Pacific. In our CSV implementation, each o f these EOFs is used as a perturbation added to the initial SST state. Using the procedure mentioned in section 4.2.2, for each initial condition, eq. (5) can be evaluated by averaging 10 different ensemble members. The initial conditions used in experiments are from NCAR’s Data Assimilation Research Tool (DART) system (Anderson et al. 2009). DART employs an Ensemble Kalman Filter (Houtekamer et al. 2005) which nudges the underlying models toward a state that is more consistent with information from a set o f observations. Overall, these initial conditions are quite close to the National Center for Environmental Prediction (NCEP) reanalysis data sets, but are more consistent with the 112 Ph.D. Dissertation: University of Northern British Columbia CAM4 model. In coupled model experiments (i.e. CCSM4), the atmospheric (CAM4) and land (CLM4) counterparts are initialized using the same initial conditions as used in CAM4 initialization. To generate initial condition for ocean (POP2.2) and sea ice (CICE4) models, a multi years ocean control hindcast experiment is performed using the CORE II (Coordinated Ocean-ice Reference Experiments - Phase II, Griffies et al. 2012) input data set as boundary conditions. The CSV implementation for CCSM4 model is the same as discussed for the CAM4 model. The three leading EOF patterns in this case are calculated using the CCSM4 simulated SST instead of observations. In most of our analysis, we mainly focus on CAM4 simulations (uncoupled) and its analysis. The results from the coupled model are only used for the purpose o f comparison between uncoupled and coupled CSVs. The experiments are performed for the time period o f 2000 to 2009 using DART initial conditions. Starting on June 1st o f each year, 4-month (lead time) integration is made by initializing both atmospheric and land models (embedded in CAM4) at the same time. In the CAM4 forecast, boundary forcing such as SST (Hadley Centre SST, Rayner et al. 2003) and the May sea ice anomaly (one month prior to the forecast initialization time) is kept persistent over the entire forecast period, i.e. June, July, August and September. Figure 4.2 displays May SST anomalies from 2000 to 2009. It can be seen that they vary each year, which can influence the growth o f forecast error when perturbations are applied. To get a more robust statistical analysis, another set o f ensemble forecasts is performed for the 1980 to 2009 time period using persistent SST anomalies as boundary forcing. The initial conditions in this case are from CAM4 control run forced with observed SST. The CSV method is implemented in the same way as in the simulations using DART initial conditions. 11 3 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon A summary o f all the experiments using different domains, models, time periods, number o f EOFs and ensemble size is given in Table 1. 4.2.4 Target Variable In this study, the forecast variable is chosen to be Outgoing Longwave Radiation (OLR) and optimum interval (i.e. lead time) is 4 months. The use of OLR, instead o f precipitation, is because it is comparatively better predicted than actual precipitation. On the other hand, it has been widely used to estimate SAM precipitation often as an index related to convective activity as lower OLR values are associated with higher cloud top heights in well-developed clouds (Schmetz; Liu, 1988 Kousky and Kayano, 1994; Moron, 1995). In summary, we use SST as perturbation variable ( Tp ) at the time / = 0 and investigate OLR as target variable ( A ) at t = t ' . As discussed before, the main reason to select SST as perturbation variable, over the Indian and Pacific Oceans, is due to the fact that SAM variability, at the seasonal time scale, is strongly influenced by the SST in central Pacific and Indian Ocean. 4.3 The Optimal Error due to the SST Uncertainties in Indian and Pacific Oceans The precipitation over the SAM region in CAM4 simulation is validated in Islam et al. (2013) and it is found that CAM4 can capture the main climatological features o f the SAM summer precipitation. Figure 4.3 shows the 30-year (1980-2009) mean OLR climatology from NCEP reanalysis data (left column) and CAM4 control simulation (right column). 114 Ph.D. Dissertation: University of Northern British Columbia Generally, negative (positive) values o f OLR indicate more (less) convection and hence more (less) cloud coverage. In observations, for all four monsoon months, there is a widespread area o f heavy convection around the northern Indian Ocean and Indian subcontinent reflecting heavier rainfall. In contrast, most o f the equatorial Indian Ocean has higher OLR values showing less convection. In CAM4, these observed OLR features are well captured except in some regions such as suppressed rainfall over the Southern China Sea and excessive rainfall over the western Arabian Sea. Overall, the OLR simulated by CAM4 compares well with the observed OLR climatology. 4.3.1 Results during the 2000 - 2009 Period We now examine how error growth varies with different initial conditions and lead time under different scenarios. In our CSV analysis, the first singular value is significantly larger than the remaining singular values. We therefore only discuss the leading CSV and its optimum growth i.e. the final pattern (FP hereafter) throughout the text. Over the Indian Ocean, the leading CSVs and corresponding FPs, optimized for time interval o f 4 months (lead o f 3 months), are shown in Figure 4.4 for each individual forecast during 2000 to 2009. In case o f leading optimized perturbation (i.e. CSV), a distinct north-south dipole pattern can be identified in most o f the years (while ignoring the arbitrary sign o f CSVs). In the years having the dipole-like CSV pattern, the corresponding FP shows error growth mainly over most of the Indian Ocean particularly over the north Indian Ocean and the central Indian subcontinent. Although most of the CSVs in individual years are similar, the FPs are quite different in each year. Large variation o f OLR FPs within different years could be due to the low-frequency processes causing asymmetry between the responses o f OLR to the SST. This could be also due to the different initial SST state anomalies which may significantly change 115 S Islam: Ensemble Simulation and Forecasting o f South Asian Monsoon the CSV growth. Indeed these SST anomalies, as the background state, play a crucial role in modifying the OLR spatial variation. This can be seen in some observed summer OLR anomalies over the SAM region for different years. For example, Charabi and Abdul-Wahab (2009) found that the July anomaly o f OLR in ENSO years such as 2000 and 2002 are significantly different compared to other years. They have seen that the geographical variations o f OLR anomalies are positive in the west Indian Ocean and negative in the East Indian Ocean in 2000 and 2002 whereas the patterns for years such as 1997 (strong El Nino year) and 1994 (strong IOD year) are the opposite. To extract the general characteristic o f these CSVs and FPs, we perform the EOF analysis for the 10 individual years for each lead time. The first leading EOF mode o f CSVs (left column) and FPs (right column) is shown for all lead times in Figure 4.5. As can been seen, the CSV patterns now more prominently resemble the dipole structure, which is not dependent on the lead time if the arbitrary sign is ignored. The FP has differences for different lead times, due to the impact o f initial stated on perturbation growth as discussed before. During the 4-month optimization interval, the error growth gets concentrated over the northern Indian Ocean and Indian subcontinent. Therefore the maximum perturbation growth is over those areas which are usually used to measure SAM intensity and variability i.e. the northern Indian Ocean and Indian subcontinent. We now discuss the Pacific Ocean case, where CSVs are extracted over the ENSO region and the prediction target is still the OLR o f SAM region. The results show that the CSVs are not sensitive to the initial conditions as in the Indian Ocean case, although the FPs vary significantly (not shown here). Figure 4.6 shows the leading EOFs o f CSVs and FPs for all initial conditions at different lead times. A prominent structure in CSVs is the equatorial 116 Ph.D. Dissertation: University of Northern British Columbia ENSO-like mode, representing a large-scale structure with major weighting in the equatorial eastern Pacific and less weighting in the western Pacific. The CSVs in this case are similar to SVs analyzed in many intermediate models when targeting ENSO forecasts (e.g. Chen et al. 1997). This resemblance indicates the link between the error growth o f ENSO prediction and that o f SAM, suggesting ENSO as a major source o f SAM predictability. The FPs (Figure 4.6) in the Pacific Ocean case have close similarities to the FPs calculated in the Indian Ocean case (Figure 4.5) except in some regions. The above CSV analysis considers the individual impact of the Indian and Pacific Oceans on the error growth o f SAM prediction where the Indian and Pacific Oceans are perturbed separately. We will next explore the total impact o f the two Oceans on the error growth of SAM prediction. For this purpose, we analyze CSVs for the Indian-Pacific Ocean. In this case, the spatial structure o f CSVs over the Indian and Pacific Oceans are the same as seen in the individual oceans case (not shown here). Figure 4.7 shows the leading EOF computed using 10 individual FPs from all years for each lead time. The interesting outcome from the joint analysis o f the two oceans is the damped error growth over most o f the Indian subcontinent and the Indian Ocean. The SST uncertainties from two different oceans partially cancel their growth with increasing lead time. This means that domain size choice is crucial, as uncertainties over a large region can suppress its growth and contribute less in error optimization. This was also confirmed by running an ensemble forecast using the CSVs over the whole Indian and Pacific Oceans which shows lower skill compared to the forecast perturbed with CSVs over Indian Ocean only. We will discuss the SAM ensemble forecast later. 117 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon The above discussion o f error growth can be further investigated by analyzing singular values o f CSVs over the Indian Ocean and over the Pacific Ocean. These values measure the fastest growth rate of perturbation during an optimum interval. The annual variation o f singular values is shown in Figure 4.8 for the Indian (Figure 4.8a) and Pacific (Figure 4.8b) Oceans. The singular values o f all the lead times are averaged each year to show an overall error growth. In case o f the Indian Ocean only, the average singular values are higher than those o f the Pacific Ocean. In Figure 4.8d, the lead time dependence o f singular values is shown by averaging all the individual years. The gradual increase o f singular values from lead times o f 0 to lead times of 3 months is more prominent in the Indian Ocean case (black line) compared to the Pacific Ocean case (red line). 4.3.2 Results during the 1980 - 2009 Period Due to the availability o f DART initial conditions, based on real observation, we only used 10 years (2000-2009) of independent experiments in section 4.3.1. For more rigorous statistics, in this section we extend the CSV experiment to the period 1980 to 2009 (30 individual SAM summer forecasts, 1030 hereafter), however the initial conditions in this case are not from real observation. All settings in this experiment are the same as those o f 10yr previous analyses except the initial conditions, which were produced from the control simulation of CAM4 forced by observed SST. As mentioned in the singular values discussion in section 4.3.1, the Indian Ocean has a more significant impact on the SAM seasonal prediction than the Pacific Ocean. We therefore emphasize results only over the Indian Ocean in further analyses. For simplicity, we only focus on analysis at the seasonal time scale (i.e. 3-month lead time). 118 Ph.D. Dissertation: University of Northern British Columbia The leading EOF mode obtained using the 30 individual CSVs and FPs is shown in Figure 4.9 at a lead time of 3 months. The spatial structures o f CSV found in 1030 experiment are very similar to CSVs found in the 10-year analysis (Figure 4.5) confirming the robustness o f CSVs obtained in the analysis o f section 4.3.1. The FP has some regional differences along with significant year to year variation (not shown here) mainly due to the impact o f initial state superimposed by CSV as discussed in section 4.3.1. The annual variation o f 1030 singular values (Figure 4.8c) and its lead time variation (Figure 4.8d, blue line) are quite consistent in magnitude with the 10-year CSV values. It has also been found in previous studies that the SVs are insensitive to initial conditions in many models (e.g. Chen et al. 1997; Xue et al. 1997a; Zhou et al. 2007; Cheng et al. 2009). To explore the sensitivity o f CSVs to initial conditions, we computed the spatial correlation between the EOFc and each individual CSV pattern, as in Tang and Deng (2011), where the EOFc is the first EOF mode obtained by all CSVs. If the EOFc and the CSV* are denoted by the normalized one-dimensional vectors eo fc and csv,, respectively, the spatial correlation Rsp is calculated as below, 1 NX, A'O -1 G=1 ( 6) where N G is the number o f total model grids over the model domain and i = 1,2,.... 10. The spatial correlation values are calculated for Indian Ocean and Pacific Ocean cases for each individual year and are shown in Figure 4.10 (a and b). For most years, the value o f the spatial correlation coefficient is quite high, for both Indian Ocean (Figure 4.10a) and Pacific 119 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon Ocean (Figure 4.10b) cases. As mentioned in Cai et al. 2003, the positive phase o f the same pattern should be regarded as equivalent to the negative phase, at least from a linear system perspective. In addition, the pattern sign can be arbitrary from the view o f EOF analysis. Thus, the preliminary features o f CSVs, as represented by the first EOF mode (equivalent to the average), are insensitive to initial conditions. This conclusion is further confirmed by using CSVs from the 1030 experiment by increasing the individual years to 30 i.e. i' = l,2,.....30 as shown in Figure 4.10c. Higher values o f individual correlations indicate independence of the CSVs from initial conditions. 4.3.3 The Results from the Coupled Model CCSM4 The CSVs and FPs discussed above are based on atmospheric model simulation with persistent SST boundary conditions. How do the CSVs change in a coupled model? As argued in the introduction, the air-sea interaction plays a crucial role in SAM simulation and prediction. To explore this issue, we calculate the CSVs for CCSM4. The leading CSVs and FPs o f CCSM4 over the Indian Ocean for all lead times are shown in Figure 4.11. We only show one particular year here, i.e. 2000, as the CSVs for other years are similar. Interestingly, the spatial structure o f the CCSM4 CSVs is similar to CAM4 CSVs, but the growth rate shown in Figure 4.11 has differences with the CAM4 counterpart in Figure 4.8. The growth rate of CCSM4 varies more significantly with lead times than that o f CAM4, so that the former is larger than the latter for longer lead time beyond 2 months. This suggests that the air-sea coupling could act as an amplifier by positive feedbacks to strengthen the perturbation growth for a long lead time. Due to the difference in the growth rate, the CCSM4 show stronger anomalies than CAM4 in their FPs, as evidenced in Figure 4.11 and Figure 4.8. 120 Ph.D. Dissertation: University of Northern British Columbia The similarities o f CSVs between CAM4 and CCSM4 suggest that one may be able to use the CAM4 to construct an ensemble prediction for the CCSM4, which is more computationally efficient and has potential in operational forecast. 4.4 Sensitivity Experiments of CSV to some Parameters Before using CSVs to construct ensemble forecasts, it is necessary to evaluate CSV robustness to the various choices made in the implementation o f the methodology. In this section we therefore discuss the robustness o f CSVs to the number o f EOFs used as initial perturbations, and the number o f ensemble members used in calculating means over the Indian Ocean with help o f sensitivity experiments. 4.4.1 Convergence of CSV with Number of EOFs As discussed in previous sections, the first three EOF modes are used to generate the CSVs. In the sensitivity experiment, we explore the CSVs with the number o f EOFs used for initial perturbations. For simplicity, we randomly choose an initial condition o f an individual year for the sensitivity analysis. Figure 4.12 shows CSVs computed using different numbers o f EOFs. The lead time for CSV is 4 months. As we increase the number o f EOF perturbations from 3 to 10 EOFs, the CSV structure does not change significantly. The corresponding FP gets optimized even with a relatively small number o f EOFs i.e. 3 to 5 EOFs. This suggests that the dominant features o f the CSVs can be captured by the leading 3 EOF modes. Indeed, increasing the number o f EOF perturbations means including more noise in the initial conditions which can, beyond a certain limit, start degrading the optimization o f perturbation. This analysis suggests that our choice o f using the three leading SST EOFs in construction o f CSVs is appropriate. We repeated the same procedure for other years with similar results. 121 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon 4.4.2 Convergence of CSV with Number of Ensemble Members Figure 4.13 shows how the first singular value changes for different numbers o f ensemble members, when using three EOFs as perturbations. If only 10 ensemble members are used, singular values converge to its minimum values for lead time 3 months and for all individual years (colored lines). Beyond 10 to 15 ensemble members, the singular values remain quite similar. This means that the true mean can be estimated with the use o f only 10 ensemble members. A similar convergence is seen for CSV spatial patterns suggesting the use o f 10 ensemble members being reliable for their calculation (not shown here). 4.4.3 Validity of CSV The CSVs and FPs, shown in Figure 4.4, are obtained by a linear approximation approach for the nonlinear model CAM4. It is therefore important to inspect whether the CSVs can resembled results that are directly obtained by the nonlinear model. This validation can be conducted by integrating the CAM4 twice, one from the initial condition superimposed by the CSVs (scaled by using a multiplication factor o f 0.1), and the other from the initial condition only. The difference between two integrations, denoted as D n o n , indicates the perturbation growth o f the CSVs obtained directly by the nonlinear model, which can be compared with the FPs and serves as the purpose o f the validation. We conduct this experiment for each year during the period from 2000 to 2009. Figure 4.14a shows the leading EOF o f D non over the entire period at 3 months lead time. The leading EOF pattern of the FP directly from the CSV method, i.e. applying the propagator R to the CSVs, is also shown in Figure 4.14b for comparison. It can be seen that the leading EOF o f D non (Figure 4.14a) is very similar to the FP from the linear approximation (Figure 4.14b) although there 122 Ph.D. Dissertation: University of Northern British Columbia are some discrepancies. Despite these discrepancies, the overall features very resemble each other, indicating that the estimates o f the CSVs are robust and correct. 4.5 SAM Ensemble Forecast using CSVs In the previous sections, we mainly focused on the general characteristics and validation o f CSV and the FP. We now apply the CSVs to generate ensemble predictions. It is expected that such an ensemble prediction should be optimal and more skillful than single predictions and non-optimal ensemble predictions. For comparison, we also use the Time Lag Ensemble (TLE) method to construct ensemble predictions. The TLE is achieved using different initial conditions with the lag o f six hours to initialize six forecasts, starting at 1a June o f each year. We construct ensemble predictions by the CSVs extracted over the Indian Oceans. Each individual CSV is multiplied by random noise and is superimposed onto the initial conditions to generate an ensemble forecast. The ensemble size is 20, by using different random numbers. In practice, we use 10 random numbers multiplied by a positive CSV and 10 random number multiplied by a negative CSV, since the sign o f CSVs are arbitrary. The ensemble mean o f positive and negative ensembles is labeled as CSVp and CSVn in discussions below, respectively. The ensemble prediction is run for the period from 2000 to 2009 and from 1980 to 2009, respectively. The RMSE values from the period from 2000 to 2009 are shown in Figure 4.15a, for a single control run prediction without perturbation (black line), the ensemble o f CSVp and CSVn (blue and red line), and ensemble o f TLE (green line). For each forecast, an error bar is drawn using the bootstrap method (Efron et al. 1993) for its significance. The bootstrap 123 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon method is obtained as follows, i) At a given forecast lead month, the observation and forecast are paired together, based on same target month. In this way, the forecast and observation sampled are paired together for the whole period (2000-2009). ii) The 95% o f the paired sample is then chosen randomly and used for RMSE calculation, iii) The step (ii) is repeated 1000 times to obtain 1000 RMSE. iv) The standard deviation o f these 1000 correlation coefficients at each lead month is drawn as an error bar. As shown in Figure 4.15a, the ensemble mean RMSE o f CSVp and CSVn forecasts are significantly better than the Control and TLE forecasts at lead times beyond 2 months. The probable reason for the significant difference in the last two months is that the CSV used to construct ensemble prediction is based on the analysis o f an optimal interval o f 4 months. In other words, it is due to the fact that the CSVs used for perturbation are optimized for a four month period (i.e. from June to September each year). To investigate the significance on a larger sample size, we also perform 30-year forecasts (CSV30p) from 1980 to 2009 using the CSVs extracted from the 1030 experiment. This forecast is compared against 1030 control forecast and RMSE is shown in Figure 4.15b. The bootstrap methodology is repeated for 1030 forecasts, the same as for the 10-year forecasts. The improvement in RMSE for lead time 3 and 4 is significant in this case which increase our confidence in the reliability o f RMSE results found for 10-year forecasts (Figure 4.15a). 124 Ph.D. Dissertation: University of Northern British Columbia 4.6 Summary and Conclusion In this chapter we applied a recently developed technique o f climatologically relevant Singular Vector (CSV) to examine the error growth o f OLR prediction over South Asian Monsoon (SAM) region, at monthly time scales using both CAM4 and the CCSM4 model. Different perturbation domains and optimal growth intervals are chosen to identify the perturbation structures most favorable for the error growth o f SAM prediction. The CSVs are computed using both uncoupled and coupled models to investigate the role o f coupling in the error growth o f SAM prediction. The robustness of CSVs is demonstrated through the analysis o f sensitivity experiments. The CSVs over Indian Ocean are extracted to generate SAM ensemble forecasts. It is seen that the CSV resembles a dipole-like structure over the Indian Oceans and the ENSO-like pattern over the Pacific Ocean. The magnitude o f error growth (singular value) is different over the Indian Ocean and over the Pacific Ocean. When the CSVs are extracted over the Indian Ocean, their growth rates are found to be more consistent with the increase o f lead time and generally larger than the counterparts over the Pacific Ocean. Different parameters such as the number o f the EOF modes used for initial perturbation and the number o f ensemble members are tested to evaluate CSV robustness over the Indian Ocean. Ensemble forecasts constructed using negative and positive CSVs over the Indian Ocean are compared with forecasts using the Time Lag Ensemble (TLE) method and the single control forecast for the period o f 10 years from 2000 to 2009. It is seen that the ensemble forecast generated by CSV perturbations has a more reliable ensemble mean compared to both the TLE mean and the control forecast, and its RMSE is found significantly better than TLE and 125 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon control forecast at lead times beyond 2 months. To explore the consistency and robustness o f the hindcast results, we also extended the hindcast to the period o f 30 years from 1980-2009, and obtained similar conclusions, namely that, the ensemble prediction by the CSV is better than that by TLE and the control forecast, further confirming the merits of the CSV method for SAM prediction. While this study facilitates the investigation o f the SAM optimal error growth by using the CSV method, some cautions should be mentioned. First, the main model used in this study is an atmospheric general model (CAM4), which lacks some key dynamics related to the air-sea coupling. Limited validation is conducted using the coupled model (CCSM4) to conduct the analysis o f error dynamics using the CSV method. Nevertheless this work seems to be the first to explore the optimal error growth o f SAM seasonal prediction, and the results reported here offer valuable insight to SAM predictability and have practical significance for ensemble prediction. These findings have implications for the SAM seasonal forecast in both the construction o f the ensemble forecast system and the detection o f key oceanic areas that impact SAM forecast, where the uncertainties can be reduced by adaptive observations and data assimilation. 126 Ph.D. Dissertation: University of Northern British Columbia Table 4.1: List o f experiments used for CSV analysis including detail o f perturbation domain, number o f perturbation EOFs, number o f ensembles and forecast years. Exp. Name Description Perturbation Domain Time Span Initial Conditions (IC) No of EOFs used Ensemble Members for each EOF 1030 30 years perturb run using EOFs as perturbations Indian Ocean 19802009 Control run IC created with prescribed SST 3 10 IOIO 10 years perturb run using EOFs as perturbations Indian Ocean 20002009 Observed IC from NCAR’s DART system 3 and 10 10 and 20 PO10 10 years perturb run using EOFs as perturbations Pacific Ocean 20002009 Observed IC from NCAR’s DART system 3 10 IOCCSM4 4 years coupled run using EOFs as perturbations Indian Ocean 2000, 2004, 2008 and 2009 Observed IC from NCAR’s DART system + 100 Year control run SST 3 10 TLE Time Lag Ensemble forecast “ 20002009 6 hours lag observed IC from NCAR’s DART system “ 6 CSVp Ensemble forecast using +ve CSV as perturbations Indian Ocean 20002009 Observed IC from NCAR’s DART system 10 CSVn Ensemble forecast using ve CSV as perturbations Indian Ocean 20002009 Observed IC from NCAR’s DART system 10 CSV30p Ensemble forecast using +ve CSV as perturbations Indian Ocean 19802009 Control run IC created with prescribed SST 10 Control Forecast Ensemble forecast without CSV perturbation 20002009 and 19802009 Observed IC from NCAR’s DART system 10 All experiments start from Jun 1st of every year with lead time of 4 months (June, July, August and September, JJAS). In CAM4 experiments, persistence SST is used as boundary conditions (BC) whereas in CCSM4, SST is used from the multiyear year control run of ocean model. 127 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon (a) -0.5386 -0 2 3 4 2 0.1171 0 4.3Z 98 -0.3303 0.1 -0.3602 -0.1442 0 07 H 02678 (b) 30.7% 20N 20S -1.5183 -0.6072 0.3039 1.215 13.1% 20N 1.7765 -0.7106 0.3553 1.4212 11.4% 20 N -0.8622 -0.345 0.1722 0.6894 Figure 4.1: First three leading correlation EOFs patterns o f the June SST over (a) Indian Ocean and (b) Pacific Ocean. These patterns are calculated from monthly observed SST for time period 1980 to 2009, and are added as perturbation in the instantaneous initial states o f persistent SST. 128 Ph.D. Dissertation: University of Northern British Columbia 8ST 2000 Mar Anomalies 906 906 (C) 1206 Me 306 906 '206 SST 2003 May Anomalies SST 2002 May Anomalies 60C SST 2001 May Anomalies 120c SST 2004 May Anomalies 906 906 1 SST 2005 May Anomalies 0 2 G 3k^ SST 2007 May Anomalies 906 906 1206 SST 2008 May Anomata* 906 906 1 SST 2009 May Anomalies Figure 4.2: SST anomalies for May o f each forecast year. Each anomaly pattern is calculated based on the mean SST climatology o f 1980-2009 and is kept persistent throughout the forecast. Units are in °C. 129 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon N C E P (J U N ) C A M 4 (J U N ) iw soi m CAM4(JUL) IIOE 30C NCEP (JUU W /m2 ! 10 k ME NCEP (AUQ) ME C A M 4 (A U G ) ISOI N C E P (S E P ) 3K eoe 11QE C A M 4 (S E P ) 190 230 270 tSOE W /m 2 310 Figure 4.3: Mean climatology o f June, July, August and September for Outgoing Longwave Radiation (OLR). NCEP OLR is shown in left column and CAM4 control run is in right column. Mean is calculated over the period 1980 to 2009. The shading corresponds to OLR values in W/m2. 130 Ph.D. Dissertation: University of Northern British Columbia C8VU m m a t ta t CSV U a* a t a t ir n P T T ia a t to t a t CCVUI m m a t a t m a t ta t PFT LS t a c a t «• at in* ta t CSV U CSVU m m m 1m m m a t Figure 4.4: The optimal leading Climatology relevant Singular Vectors (CSVs) o f SST and corresponding OLR final patterns (FPs) o f optimized for 4 months (lead time 3) over the Indian Ocean domain. Each individual CSV and FP is shown starting from year 2000 to 2009. The SST units are °C and OLR is in W/m2 131 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon C S V * FPT, P V A R . S S T , TVAft - F lU T U adO EO FtC SV wXDJO 77.1% U ttdO EO Fl FPT LMd 1 EOF1 810% LMd 1 EOF 1 iM d 2 EOF1 68 9% Lead2EOFi 20.8% ION - ION - ♦ A c s - Laad 3 EOFt L*w) 3 EOF 1 25.8% Figure 4.5: The first leading EOF patterns o f the CSVs (left) and FPs (right) computed over 10 individual CSV patterns for all the 4 lead times. CSVs are extracted over Indian Ocean. The SST units are in °C and OLR is in W/m2. 13 2 Ph.D. Dissertation: University of Northern British Columbia C S V ft ^ P T , PV A A • S $ t . TVAR . H U T 77.®% UftdOEOFI C8V«cfl>_PO iM d 0 EOF1 FPT 23.0% SON tJQW 100 i 1 EGF1 IOW 93.0% SOB LMd 1 EOFt 33.0% SOB 9K LM0 2EOF1 LMd 3 EOF1 939% BOE tXOC 'H E i M d 2 EOFt 32.1% LMd 3 EOF! 24.0% Figure 4.6: Same as Figure 4.5 but for CSVs calculated over Pacific Ocean and corresponding FPs over SAM region. The SST units are in °C and OLR is in W/m2. 133 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon CSV &FPT, PVAft - SST. TVAA - FLUT L«ad 0 EOF1 FPT t— ... 306 24 5% ;.... «* 906 1------130E Load 1 EOF1 306 «OE *906 41.6% 906 1206 1S06 Figure 4.7: Same as Figure 4.6 but for CSV calculated over combined Indian-Pacific Ocean. Only FPs are shown here. 134 Ph.D. Dissertation: University of Northern British Columbia All Leads Average Singular Value, 10/10 —I I I I I 1_____ 1_____ 1_____ I_____ I_____ L l i i a J I I > (/> 2000 PO/IO I I 2000 2002 2010 1030/1030 i i i i I i i i i I i i i i I i i. J * to A AA A 0 ----- 1---1---1---1---j— 1980 AA AA * I I I I I I I 1 I I A . * A A A 4 A -------- j--------,-------- 1-------- 1-------- ,-------- j— 4 ---j---,---1---,---,---1---,--- 1---1---r 1985 1990 1995 2000 2005 2010 0.5 1.0 1.5 2.0 2.5 3.0 SST/FLUT 8 IF* $ (O 6 — 0.0 Figure 4.8: Annual variation o f singular values averaged over all the lead times for (a) Indian Ocean (b) Pacific Ocean and (c) 1030 experiments, (d) The lead time variation o f singular values over all the three cases (see text for detail). 135 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon C SV t FPT, PVAR - SST TVAR - FUrT Lead 0 EOF1 CSVeotDJP 56.8% Figure 4.9: Same as Figure 4.5 but for CSVs and FPs calculated over Indian Ocean using 1030 experiment. 30 years CSVs and FPs are used for EOF analysis. The initial conditions for 1030 experiment are generated using CAM4 prescribed SST run. 136 Ph.D. Dissertation: University of Northern British Columbia 10 Years Spatial Correlation POs/IO 10 Years Spatial Correlation IQ/IO 1.0 hi Corr. 0.5 - 0.0 - 0.0 - -0.5 - -0.5 - 2000 2002 2004 2006 2000 2008 2002 2004 2006 2006 30 Years Spatial Correlation IO/IO oo 0.5 - -0.5 - 1980 1985 1990 1995 2000 2005 Figure 4.10: The spatial correlation for (a) Indian Ocean, (b) Pacific Ocean and (c) 1030 experiments. It is computed between the first EOF mode (obtained by all CSVs) and each individual CSV pattern, same as in Tang and Deng (2011). 137 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon CC8M4 CSV LO, P w . SST. Tvar« FLUT CCSM4 FPT LO. Pvar • SST, Tv» - FLUT OS 01 306 1Z06 tSOE 30E CSV LI 120E 190E IBDE CSVL2 I9 0 e FPT 12 19K taoe taoe FPTL3 taoe 190c CSVL3 aoc see CCSM4 16 14 12 > 10 e 3 6 ? 4 v_ 30 CVJCMCMCMCMCMCMCMCMCM rt 3 O) c CO 20 10 0 3 6 9 12 15 18 Ensemble members Figure 4.13: The convergence o f the leading singular values with increasing number o f ensemble members, optimized for 4 month time interval. Each color line represent individual year. In each year, 20 singular values are calculated by increasing the ensemble size from 1 to 20 in the CSV method. 140 Ph.D. Dissertation: University of Northern British Columbia CSV Validation, EOF at L3 26.9% 30N S" vJ 30S - 30E -24 60E -16 90E -8 0 120E 8 18 R*CSV=FP, EOF at L3 30E 60E 150E 24 25.6% 90E 120E 150E Figure 4.14: Leading EOF o f OLR FPs, at lead time 3, (a) calculated using CSV as perturbation and (b) estimated by applying linear propagator R to CSV i.e. by linear approximation. For (a), the difference o f perturbed and control forecast is calculated each year and leading EOF is extracted using all the differences. In (b), EOF pattern is calculated using estimated FPs from linear approximation for all the 10 years (2000-2009). 141 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon (a) 3 5 1--------- 1----------1----------■----------■----------1----------■--------- 1---------- 1--------- 1---------- 1--------- r CNTRL CSVp CSVn LAG ! __ I__I_____ Lead 0 r--- ,--- 1--- 1__ Lead 1 Lead 2 L Lead 3 Lead 2 Lead 3 Months (b) 5 4 2 1 Lead 0 Lead 1 5 Months Figure 4.15: Root mean square error (RMSE) of the SAM ensemble forecast generated using CSV perturbation, (a) RMSE calculated over 10 years (2000-2009 using observed initial conditions) whereas in (b) RMSE is for 30 years forecast (1980-2009 using CAM4 control initial conditions). Error bar is drawn using bootstrap method. 14 2 Ph.D. Dissertation: University of Northern British Columbia Chapter 5 Summary and Discussion 5.1 Thesis Summary This thesis investigates some important issues related to South Asian Monsoon (SAM) simulation and its seasonal predictability using comprehensive climate models. A number o f findings and conclusions are obtained through detailed analysis o f model experiments and simulation outputs, which are summarized below. The different features o f the SAM precipitation in the simulations o f NCAR's uncoupled (CAM4/5) and coupled (CCSM4) climate models are discussed in Chapter 2. The comparison o f simulations is performed against observations to identify strengths and discrepancies in these models. The mean climatology o f the SAM, its interannual variability and air-sea interactions are evaluated using different measures. The impact o f SST biases on the SAM simulation and the importance o f air-sea coupling are also explored in this chapter. The results reveal numerous key aspects o f CAM4 and CCSM4 in simulating SAM precipitation. Compared to the versions previous to CAM4 and CCSM4, improvements in eliminating many regional biases are seen in both models. The spatial pattern and the seasonal cycle of monsoon precipitation is fairly well reproduced in the simulations along with the monsoon onset between May and June, the high precipitation during June-August and slow withdrawal during September-October. The large northward shift o f the ITCZ from January to July is also well simulated. A pronounced improvement in precipitation simulation, such as the regional bias reduction over complex terrain, is seen when the CAM4 143 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon resolution is increased. The improved SAM monsoon is seen in CCSM4 simulations with the reduction o f many biases particularly over the Arabian Peninsula and the western coast o f India. The period o f ENSO is found to be more realistic in CCSM4 than in the previous version (e.g. CCSM3). Along with improvements, the systematic errors o f both models are also identified. Inconsistencies in the oscillatory period and amplitude o f monsoon interannual variability are seen in CAM4 simulations. The bias in the CAM4 seasonal mean climatology results in excessive precipitation over the Arabian Sea and over the Western Ghats o f India and reduced precipitation over the eastern Indian Ocean extending into the Bay o f Bengal. In CAM4, the East Asian summer monsoon precipitation is simulated better than the SAM precipitation. The interrupted northward progression and delayed onset o f the monsoon over the SAM region is seen in the CCSM4 simulation including the double ITCZ problem that was also present in the previous versions of the CCSM model (CCSM3). A systematic cold SST bias over the tropical Pacific Ocean is revealed in simulations. Significant cold biases over the equatorial Pacific Ocean are found in CCSM4, particularly in winter and early summer. The importance o f air-sea coupling is investigated by forcing CAM4 with coupled model SST which revealed that the local air-sea coupling over the SAM region acts to modulate the SAM summer rainfall activity. Further, it is seen that along with air-sea interaction, SST bias in the CCSM4 model plays an important role in modulating the variability and magnitude o f SAM precipitation. The influence o f the coupled model SST bias in the northern Indian 144 Ph.D. Dissertation: University of Northern British Columbia Ocean on SAM precipitation is investigated, with a conclusion that the reduction in excessive SAM rainfall in CCSM4 simulation is mainly due to a cold SST bias in the Arabian Sea. Chapter 3 focuses on a detailed investigation o f ENSO-SAM relationships in CAM4 and CCSM4, attempting to identify the strength and discrepancies o f these models in preserving different types o f ENSO-SAM relationships. The analyses in this chapter are performed on the basis of a recent classification scheme o f the ENSO-SAM relationships by Wu et al, (2012). They have identified three different types o f ENSO influences on the SAM using observations (DJF-only, JJAS-only and DJF&JJAS). We have extended their analysis to our simulation results, and further explored the physical understanding o f particular types o f ENSO influence on SAM using sensitivity experiments. Using the composite analysis for model simulations and observations, the performance o f CAM4 and CCSM4 is evaluated for the ENSO-SAM relationship. It is seen that the CAM4 simulation in DJF-only type is better than CCSM4, and can reproduce a realistic anomalous rainfall over the east Pacific reasonably well along with some biases over the western Pacific Ocean and Maritime continent. In contrast, the CCSM4 showed large systematic errors including significant rainfall anomaly biases in the Indian Ocean from the preceding winter to the early spring, probably due to the errors in SST simulation. In DJF-only years, the CAM4 performed better than CCSM4. For JJAS-only type, the CAM4 showed significant disagreement for SAM rainfall anomalies as compared to observations and CCSM4 simulation. It is found that ENSO induced warming in northern Indian Ocean via atmospheric circulation is absent in CAM4 simulation o f the JJAS-only type, probably due to the lack o f air-sea coupling. The CCSM4 simulation has 145 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon shown strong convergence over the SAM region, intensifying the anomalous SAM. The CCSM4 results are found to be more realistic than CAM4 as compared with observation. In the DJF&JJAS type, realistic anomalous rainfall is reproduced by CAM4 as compared to CCSM4 which has a very weak simulation o f rainfall anomalies. Compared to the other two types, the CAM4 has the better simulation o f the ENSO-SAM relationship in the DJF&JJAS type, including a more realistic simulation in moisture content and thickness difference. The role of the Indian Ocean in controlling the DJF&JJAS type ENSO-SAM teleconnection is further investigated by CAM4 sensitivity experiments. It is found that in the absence of Indian Ocean SST anomalies, the anomalous SAM summer rainfall is suppressed in the DJF&JJAS type, suggesting the importance o f Indian Ocean SST in modulating the east equatorial Pacific SST and SAM interaction. In Chapter 4, both the CAM4 and CCSM4 models are used to implement a recently developed technique of climatologically relevant Singular Vector (CSV) to examine the error growth o f OLR prediction over SAM region. Both the Indian and Pacific Oceans are used as perturbation domains and different optimal growth intervals are chosen to identify the perturbation structures most favorable for the error growth o f SAM prediction. The CSVs and corresponding final patterns (FP) are calculated for both uncoupled and coupled simulations. The robustness o f CSVs is confirmed by sensitivity experiments. Different parameters such as the number o f the EOF modes used for initial perturbation and the number o f ensemble members are tested in these sensitivity experiments. The CSVs over the Indian Ocean are used to perform SAM ensemble forecasts. 146 Ph.D. Dissertation: University of Northern British Columbia It is seen that the magnitude o f error growth (singular value) is different over the Indian Ocean and over the Pacific Ocean. The growth rates over the Indian Ocean are found to be more consistent with the increase o f lead time and generally larger than the growth rates over the Pacific Ocean. The CSVs are found to be insensitive to initial conditions whereas FPs are much different from case to case. The optimized CSVs over the Indian Ocean are further used to construct a SAM ensemble forecast. The CSV-based ensemble forecast is compared with that by Time Lag Ensemble (TLE) and the single control forecast for the period from 2000-2009. A more realistic ensemble mean is obtained in the CSV-based ensemble forecast as compared to TLE and the control forecast. There is also a significantly lower RMSE in the CSV ensemble forecast than in other methods at lead time beyond 2 months. The hindcast is also extended to the period of 30 years, from 1980-2009, further confirming the merits o f the CSV ensemble method for SAM prediction. 5.2 Discussion The simulation o f South Asian summer monsoon by models depends mainly on how realistically they simulate the mean state and variability o f the SAM, and how well they represent the monsoon internal dynamics and air-sea interactions. In general, modeling SAM fluctuations depends mainly on understanding the fundamental processes o f SAM, model parameterization schemes and the methods used for numerical implementation o f these processes. The NCAR climate modeling system, including CAM4 and CCSM4, has been widely used to improve our capability in simulating and predicting climate, and made significant progress in 147 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon various aspects. However biases in these models can adversely affect the SAM simulation, particularly in the region where the ENSO-monsoon relationship is important. Thus, further efforts are required in model development, including improving model physics and model resolution as well as coupling processes, particularly over the complex SAM topography. The reliability o f current coupled models, due to their SST biases, is not at a level where the local air-sea feedbacks can be used to improve the seasonal predictability o f SAM. In case o f the seasonal prediction, the skill o f comprehensive climate models may be increased up to the theoretical limit of predictability inherent in the nonlinear and stochastic processes o f climate system. Since there are uncertainties in initial and boundary conditions, a single forecast is of limited value. Instead, an ensemble o f forecasts is necessary to capture the most likely range o f forecast states that are expected to arise given the uncertainties in the initial and boundary conditions. The methods used to sample the initial uncertainty are crucial and can be achieved by using the advanced ensemble construction methods such as the one discussed in this thesis. The challenging issue is to design and implement a reliable ensemble forecast strategy that should include the major uncertainties o f forecast initial states. The work reported in this thesis is subject to some cautions. For example, the perturbation growth method used in this study can be significantly influenced by the model used in its implementation. The main implementation model is an atmospheric general circulation model (CAM4), which lacks coupled air-sea dynamics and has systematic biases in SAM simulation. The small sample size used in perturbation theory analysis may limit the generality and robustness o f results. Nevertheless, this work has theoretical significance and practical importance in SAM simulation and prediction. For example, the perturbation 148 Ph.D. Dissertation: University of Northern British Columbia growth theory presented in this research is to our knowledge the first to explore the optimal error growth of SAM seasonal prediction. This research offers a valuable insight on SAM predictability and proposes a practical method in constructing optimal SAM ensemble predictions. 149 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon References AchutaRao KM and Sperber KR (2006), ENSO simulation in coupled ocean-atmosphere models: are the current models better? Climate D ynam ics, 27:1-15. doi: 10.1007/s00382-006-0119-7 Acharya N, Kar SC, Mohanty UC, Kulkami MA and Dash SK (2011), Performance o f GCMs for seasonal prediction over India - a case study for 2009 monsoon. Theoretical A pplied Climatology, 105:505-520. Achuthavarier D, Krishnamurthy V, Kirtman BP and Huang B (2012), Role o f the Indian Ocean in the ENSO-Indian summer monsoon teleconnection in the NCEP climate forecast system. Journal o f Climate, 25(7):2490-2508. Annamalai H and Liu P (2005), Response o f the Asian Summer Monsoon to changes in El Nino properties. Q uarterly Journal o f the R oyal M eteorological Society, 131:805-831. Annamalai H, Hamilton K and Sperber KR (2007), South Asian summer monsoon and its relationship with ENSO in the IPCC AR4 simulations. Journal o f Climate, 20:10711092. doi:10.1175/JCLI4035.1 Anderson J, Hoar T, Raeder K, Liu H, Collins N, Tom N, and Avellano A (2009), The data assimilation research testbed: A community facility. Bulletin o f the Am erican M eteorological Society, 90:1283-1296. 150 Ph.D. Dissertation: University of Northern British Columbia Ashok K, Guan Z, and Yamagata T (2001), Impact o f the Indian Ocean Dipole on the relationship between the Indian monsoon rainfall and ENSO. G eophysical Research Letters, 28:4499-4502. Bamzai AS and Shukla J (1999), Relation between Eurasian snow cover, snow depth and the Indian summer monsoon: an observational study. Journal o f Clim ate, 12:3117-3132. Becker BD, Slingo JM, Ferranti L, and Molteni F (2000), Seasonal predictability o f the Indian Summer Monsoon: What role do land surface conditions play? M ausam , 52:4762. Boschat G, Terray P and Masson S (2012), Robustness o f SST teleconnections and precursory patterns associated with the Indian summer monsoon. Clim ate D ynam ics, 38(11-12):2143-2165. Bollasina M and Nigam S (2009), Indian Ocean SST, evaporation, and precipitation during the South Asian summer monsoon in IPCC AR4 coupled simulations. Climate D ynam ics, 33:1017-1032. doi:10.1007/s00382-008-0477-4 Buizza R and TN Palmer (1995), The singular vector structure o f the atmosphere global circulation. Journal o f Atm ospheric Science, 52:1434-1456. Buizza, R, and TN Palmer (1998), Inpact o f ensemble size on ensemble prediction. M onthly Weather Review, 126:2503-2518. Cane MA, Zebiak SE and Dolan SC (1986), Experimental forecasts o f El Nino. Nature, 321:827-832 151 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon Charabi Y, and Abdul-Wahab SA (2009), Synoptic aspects o f the summer monsoon o f southern Oman and its global teleconnections, Journal o f G eophysical R esearch, 114:D07107. doi: 10.1029/2008JD010234. Chang EC, Yeh SW, Hong SY, Wu R (2011), The role o f air-sea interaction over the Indian Ocean in the in-phase transition from the Indian summer monsoon to the Australian boreal winter monsoon, Journal o f G eophysical Research, 116:D 01107. Chen YQ, Battisti DS, Palmer TN, Barsugli J, Sarachik ES (1997), A study o f the predictability o f Tropical Pacific SST in a coupled atmosphere-ocean model using singular vector analysis: the role o f the annual cycle and the ENSO cycle. M onthly Weather Review, 125:831-845. Chen D, Cane MA, Kaplan A, Zebiak SE and Huang D (2004), Predictability o f El Nino over the past 148 years. Nature, 428:733-736. Cheng Y, Tang Y, Zhou X, Jackson PL, Chen D (2010), Further analysis o f singular vector and ENSO predictability in the Lamont model - Part I: singular vector and the control factors, Climate D ynam ics, 35:807-826. doi:10.1007/s00382-009-0595-7. Cherchi A and Navarra A (2007), Sensitivity of the Asian summer monsoon to the horizontal resolution: differences between AMIP-type and coupled model experiments. Climate Dynam ics, 28:273-290. doi:10.1007/s00382-006-0183-z Chamey JG and Shukla J (1981), Predictability o f monsoons, in Monsoon Dynamics, edited by J. Lighthill and R. P. Pearce, Cam bridge University Press, 99-109. 15 2 Ph.D. Dissertation: University of Northern British Columbia Chowdary J, Xie SP, Lee JY, Kosaka Y and Wang B (2010), Predictability o f summer Northwest Pacific climate in eleven coupled model hindcasts: local and remote forcing, Journal o f G eophysical R esearch , 115:D22121, doi:10.1029/2010JD014595 Chung C and Nigam S (1999), Asian summer monsoon-ENSO feedback on the Cane-Zebiak model ENSO. Journal o f C lim ate , 12:2787-2807 Collins WD, Rasch PJ, Boville BA, Hack JJ, McCaa JR, Williamson DL, Briegleb BP, Bitz CM, Lin S and Zhang M, (2006a), The Formulation and Atmospheric Simulation o f the Community Atmosphere Model Version 3 (CAM3). Journal o f Climate, 19:21442161. doi: 10.1175/JCLI3760.1 Collins WD, Bitz CM, Blackmon ML, Bonan GB, Bretherton CS, Carton JA, Chang P, Doney SC, Hack JJ, Henderson TB, Kiehl JT, Large WG, McKenna DS, Santer BD and Smith RD, (2006b), The Community Climate System Model Version 3 (CCSM3). Journal o f Climate, 19:2122-2143. doi: 10.1175/JCLI3761.1 Covey C, AchutaRao KM, Lambert SJ and Taylor KE (2000), Intercomparison o f present and future climates simulated by coupled ocean-atmosphere GCMs. PCMDI report no 66. Program for climate model diagnosis and intercomparison, Lawrence Livermore National Laboratory. University o f California, Liverm ore Covey C, AchutaRao KM, Cubasch U, Jones P, Lambert SJ, Mann ME, Phillips TJ and Taylor KE (2003), An overview o f results from the coupled model intercomparison project (CMIP). G lobal and Planetary Change, 37:103-133. doi:10.1016/S09218181(02) 00193-5 153 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon Dai A (2006), Precipitation characteristics in eighteen coupled climate models. Journal o f Climate, 19:4605-4630. doi: 10.1175/JCLI3884.1 DelSole T and Shukla J (2012), Climate models produce skillful predictions o f Indian summer monsoon rainfall. G eophysical Research Letter, 39:L09-703, doi: 10.1029/2012GL051 279. Delsole T Shukla J (2002), Linear prediction o f Indian monsoon rainfall. Journal o f Climate, 15:3645-3658. Doblas-Reyes FJ, Deque M and Piedelievre JP (2000), Multi-model spread and probabilistic seasonal forecasts in PROVOST. Q uarterly Journal o f the R oyal M eteorological Society, 126:2069-2087. Drbohlav HL and Krishnamurthy V (2010), Spatial Structure, Forecast Errors, and Predictability of the South Asian Monsoon in CFS Monthly Retrospective Forecasts. Journal o f Climate, 23:4750-4769. doi: http://dx.doi.Org/10.1175/2010JCLI2356.l Du H, Doblas-Reyes FJ, Garcia-Serrano J, Guemas V, Souf-flet Y and Wouters B (2012), Sensitivity o f decadal predictions to the initial atmospheric and oceanic perturbations, Climate D ynam ics, 7(8):2013-2023. Efron B and Tibshirani RJ (1993), An Introduction to the Bootstrap. Chapm an and H all/C RC, New York, USA, ISBN: 0412042312. Fu C and Fletcher JO (1985), The relationship between Tibet-tropical Ocean thermal contrast and interannual variability o f Indian monsoon rainfall. Journal o f Clim ate a n d A pplied M eteorology, 24:841-847. 154 Ph.D. Dissertation: University of Northern British Columbia Fu X, Wang B, Bao Q, Liu P and Lee JY (2009), Impacts o f initial conditions on monsoon intraseasonal forecasting. G eophysical Research Letters, 36:L08801. Fu X, Wang B, Lee JY, Wang W and Gao L (2011), Sensitivity o f dynamical intraseasonal prediction skills to different initial conditions. M onthly W eather Review, 139:25722592. Gadgil S, Rajeevan M and Nanjundiah R (2005), Monsoon prediction Why yet another failure? Current Science, 88:1389-1400. Gadgil S and Sajani S (1998), Monsoon precipitation in the AMIP runs. Clim ate D ynam ics, 14:659-689. Gent P, Danabasoglu G, Donner L, Holland M, Hunke E, Jayne S, Lawrence D, Neale R, Rasch P, Vertenstein M, Worley P, Yang ZL and Zhang M (2011), The Community Climate System Model Version 4. Journal o f Climate, 24:4973-4991. doi: 10.1175/2011JCLI4083.1 Gimeno L, Drumond A, Nieto R, Trigo RM, Stohl A (2010), On the origin o f continental precipitation. G eophysical Research Letter, 37:L13804. doi: 10.1029/2010GL043712 Goswami BN, Krishnamurthy V and Annamalai H (1999), A broad-scale circulation index for the interannual variability o f the Indian summer monsoon. Q uarterly Journal o f the R oyal M eteorological Society, 125:611-633. Goswami BN, Ajaya RS (2001), Intra-seasonal oscillations and inter-annual variability o f the Indian summer monsoon. Journal o f Climate, 14:1180-1198. 155 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon Griffies SM, Winton M, Samuels B, Danabasoglu G, Yeager S, Marsland S, Drange H and Bentsen M (2012), Datasets and protocol for the CLIVAR WGOMD Coordinated Ocean-sea ice Reference Experiments (COREs), WCRP Report, No. 21/2012, pp. 21 Hack JJ, Kiehl JT and Hurrell JW (1998), The hydrologic and thermodynamic characteristics of the NCAR CCM3. Journal o f Climate, 11:1179-1206. Hastenrath S and Polzin D (2004), Dynamics o f the surface wind field over the equatorial Indian Ocean. Quarterly Journal o f the Royal M eteorological Society, 130: 503-517. Hawkins ED, Sutton R (2010), Estimating climatically relevant singular vectors for decadal predictions of the Atlantic Ocean, Journal of Climate, 24:109-123. doi: 10.1175/2010JCLI3579.1 Holtslag A, Boville B (1993), Local versus nonlocal boundary-layer diffusion in a global climate model. Journal o f Climate, 6:1825. Houtekamer PL, Mitchell HL, Pellerin G, Buehner M, Charron M, Spacek L and Hansen B (2005), Atmospheric Data Assimilation with an Ensemble Kalman Filter: Results with Real Observations. M onthly W eather Review, 133:604-620. doi: 10.1175/MWR-2864.1 Hunke EC and Lipscomb WH (2008), CICE: the Los Alamos sea ice model user’s manual, version 4. L os Alam os N ational Laboratory Technical Report, LA-CC-06-012. Iacono MJ, Delamere J, Mlawer E, Shephard M, Clough S and Collins W (2008), Radiative forcing by long-lived greenhouse gases: Calculations with the AER radiative transfer models. Journal o f G eophysical Research, 113:D13103. doi: 10.1029/2008JD009944. 156 Ph.D, Dissertation: University of Northern British Columbia IPCC (2007), Climate Change 2007, The Physical Science Basis. Contribution o f Working Group I to the Fourth Assessment Report o f the Intergovernmental Panel on Climate Change [Solomon, S., D. Qin, M. Manning, Z. Chen, M. Marquis, K.B. Averyt, M.Tignor and H.L. Miller (eds.)]. Cam bridge University Press, Cambridge, United Kingdom and New York, NY, USA. Islam S, Tang Y and Jackson P (2013), Asian monsoon simulations by Community Climate Models CAM4 and CCSM4, Climate Dynamics, 41(9-10):2617-2642. doi: 10.1007/s00382-013-1752-6 Jin EK, Kinter JL, Wang B, Park CK, Kang IS, Kirtman BP, Kug JS, Kumar A, Luo JJ, Schemm J, Shukla J and Yamagata T (2008), Current status o f ENSO prediction skill in coupled ocean-atmosphere models, Climate D ynam ics, 31(6):647-664. doi: 10.1007/s003 82-008-0397-3. Joseph R and Nigam S (2006), ENSO evolution and teleconnections in IPCC’s twentiethcentury climate simulations: realistic representation? Journal o f Climate, 19:43604377. doi: 10.1175/JCLI3846.1 Jhun JG and Lee EJ (2004), A new East Asian winter monsoon index and associated characteristics o f winter monsoon. Journal o f Climate, 17:711-726. Ju J and Slingo J (1995), The Asian summer monsoon and ENSO. Q uarterly Journal o f the R oyal M eteorological Society, 121:1133-1168. Kalnay E and Toth Z (2003), Bred Vectors o f the Zebiak-Cane Model and Their Application to ENSO Predictions. Journal o f Climate, 16:40-56. 157 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon Kanamitsu M and Krishnamurti TN (1978), Northern summer tropical circulation during drought and normal rainfall months. M onthly Weather Review, 106:331-347. Kang IS, Jin K, Wang B, Lau KM, Shukla J, Krishnamurthy V, Schubert SD, Wailser DE, Stem WF, Kitoh A, Meehl GA, Kanamitsu M, Galin VY, Satyan V, Park CK and Liu Y (2002), Intercomparison o f the climatological variations o f Asian summer monsoon precipitation simulated by lOGCMs. Climate D ynam ics, 19:383-395. Kang IS, Lee JY and Park CK (2004), Potential predictability o f summer mean precipitation in a dynamical seasonal prediction system with systematic error correction. Journal o f Climate, 17:834-844. Kang IS and Shukla J (2006), Dynamic seasonal prediction and predictability (Chap. 15). The Asian Monsoon, Springer Praxis, Chicherster, 585-612. Kiehl JT, and Gent PR (2004), The Community Climate System Model, version 2. Journal o f Climate, 17:3666-3682. Kirtman BP and Shukla J (2000), Influence o f the Indian summer monsoon on ENSO. Q uarterly Journal o f the R oyal M eteorological Society, 126:213-239. Kistler R, Kalnay E, Collins W, Saha S, White G, Woollen J, Kalnay E, Chelliah M, Ebisuzaki W, Kanamitsu M, Kousky V, Dool H, Jenne R and Fiorino M (2001), The NCEP-NCAR 50-Year Reanalysis: Monthly Means CD-ROM and Documentation. Bulletin o f the Am erican M eteorological Society, 82:247-267. doi: 10.1175/1520- 0477(2001 )082<0247:TNNYRM>2.3.CO;2 158 Ph.D. Dissertation: University of Northern British Columbia Kitoh A and Arakawa O (1999), On overestimation o f tropical precipitation by an atmospheric GCM with prescribed SST. G eophysical Research Letter, 26:2965-2968. Kleeman R, Tang Y and Moore AM (2003), The calculation o f climatically relevant singular vectors in the presence o f weather noise as applied to the ENSO problem. Journal o f Atm ospheric Science, 60:2856-2868. Kosaka Y, Chowdary JS, Xie SP, Min YM and Lee JY (2012), Limitations o f seasonal predictability for summer climate over East Asia and the Northwestern Pacific. Journal o f Climate, 25:7574-7589. Kousky VE and Kayano MT (1994), Principal modes o f outgoing longwave radiation and 250 mb circulation for the South American sector. Journal o f Climate, 7:1131-1143. Kripalani RH and Kulkami A (1999), Climatology and variability o f historical Soviet snow depth data: some new perspectives in snow-Indian monsoon teleconnections. Climate D ynam ics, 15:475-489. Kripalani RH, Oh JH, Kulkami A, Sabade SS and Chaudhari HS (2007), South Asian summer monsoon precipitation variability: Coupled climate model simulations and projections under IPCC AR4. Theoretical A pplied Climatology, 90:133-159. doi: 10.1007/s00704-006-0282-0 Krishnamurti TN, Bedi HS, Subramaniam M (1989), The summer monsoon o f 1987. Journal o f Climate, 2:321-340. Krishnamurthy V and Shukla J (2000), Intraseasonal and interannual variability o f rainfall over India. Journal o f Climate, 13:4366-4377. S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon Krishnamurthy V and Shukla J (2008), Seasonal persistence and propagation o f intraseasonal patterns over the Indian monsoon region. Climate D ynam ics, 30, 353-369. Kumar K, Hoerling M and Rajagopalan B (2005), Advancing dynamical predictions o f the Indian monsoon rainfall. Geophys Res Lett 32:L08704. doi: 10.1029/2004GL021979 Kulkami MA, Acharya N, Kar SC, Mohanty UC, Tippett MK, Robertson AW, Luo JJ and Yamagata T (2011), Probabilistic Prediction o f Indian Summer Monsoon Rainfall Using Global Climate Models, Theoretical a n d A pplied Clim atology, 107(3-4):441450. doi: 10.1007/s00704-011 -0493-x Kug JS, Ham YG, Lee EJ and Kang IS (2011), Empirical singular vector method for ensemble El Nin~o-Southem Oscillation prediction with a coupled general circulation model. Journal o f Geophysical Research, 116:C08029. Liang J, Yang S, Hu ZZ, Huang B, Kumar A, Zhang Z (2009), Predictable patterns o f Asian and Indo-Pacific summer precipitation in the NCEP CFS. Climate D ynam ics, 32:9891001 . Lacono M, Delamere J, Mlawer E, Shephard M, Clough S and Collins W (2008), Radiative forcing by long-lived greenhouse gases: Calculations with the AER radiative transfer models. Journal o f G eophysical R esearch, 113D, 13103. Lau NC and Nath MJ (2000), Impact o f ENSO on the variability o f the Asian-Australian monsoons as simulated in GCM experiments. Journal o f Climate, 13:4287-4309. Lau KM and Li MT (1984), The monsoon o f east Asia and its global associations - A survey. B ulletin o f the Am erican M eteorological Society, 65:114-125. 160 Ph.D. Dissertation: University of Northern British Columbia Lai M, Bengtsson L, Cubash U, Esch M, and Schlese U (1995), Synoptic scale disturbances of the Indian summer monsoon as simulated in a high resolution climate model. Climate Research, 5:243-258. Lee JY, Wang B, Kang IS, Shukla, Kumar A, Kug JS, Schemm JKE, Luo JJ, Yamagata T, Fu X, Alves O, Stem B, Rosati T and Park CK (2010), How are seasonal prediction skills related to models’ performance on mean state and annual cycle? Clim ate D ynam ics, doi: 10.1007/s00382-010-0857-4 Lee WJ and Coauthors (2009), APEC Climate Center for climate information services, A P C C F inal Report (www.apcc21.org/eng/research/pub/repo/japcc040602_lst.jsp) Leith CE (1974), Theoretical skill o f Montecarlo forecasts. M onthly W eather Review, 102: 409-418. Leutbecher M and Palmer TN (2008), Ensemble forecasting. Journal o f Com putational Physics, 227(7):3515-3539. Li C and Yanai M (1996), The onset and interannual variability o f the Asian summer monsoon in relation to land-sea thermal contrast, Journal o f Climate, 9:358-375. Li T and Hogan TF (1999), The Role o f the Annual-Mean Climate on Seasonal and Interannual Variability o f the Tropical Pacific in a Coupled GCM. Journal o f Climate, 12:780-792. Liang J, Yang S, Hu ZZ, Huang B, Kumar A, Zhang Z (2009), Predictable patterns o f the Asian and Indo-Pacific summer precipitation in the NCEP CFS. Clim ate D ynam ics, 32:989-1001. 161 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon Lin JL (2007), The double-ITCZ problem in IPCC AR4 coupled GCMs: ocean-atmosphere feedback analysis. Journal o f C lim ate, 20:4497-4525. doi:10.1175/JCLI4272.1 Liu X, Easter RC, Ghan SJ, Zaveri R, Rasch P, Shi X, Lamarque JF, Gettelman A, Morrison H, Vitt F, Conley A, Park S, Neale R, Hannay C, Ekman AML, Hess P, Mahowald N, Collins W, Iacono MJ, Bretherton CS, Flanner MG and Mitchell D (2011), Toward a minimal representation of aerosol direct and indirect effects: model description and evaluation, Geoscientijic M odel D evelopm ent D iscussion, 4:3485-3598. doi: 10.5194/gmdd-4-348 5-2011 Lorenz EN (1963), Deterministic Nonperiodic Flow, Journal o f the A tm ospheric Sciences, 20(2):130-141. Lorenz EN (1965), A study o f the predictability o f a 28-variable atmospheric model. Tellus, 17:321-333. Loschnigg JG and Webster PJ (2000), A coupled ocean-atmospheric system o f SST regulation for the Indian Ocean. Journal o f Climate, 13:3342-3360. Malik N, Marwan N and Kurths J (2010), Spatial structures and directionalities in monsoonal precipitation over south Asia. N onlinear Process Geophysics, 17:371-381. Manganello, Julia V, Huang and Bohua (2008), The influence o f systematic errors in the Southeast Pacific on ENSO variability and prediction in a coupled GCM. Climate D ynam ics, DOI: 10.1007/s00382-008-0407-5 Meehl GA (1994), Influence o f the land surface in the Asian summer monsoon: Internal conditions versus internal feedback, Journal o f Climate, 7:1033-1049. 162 Ph.D. Dissertation: University of Northern British Columbia Meehl GA and Arblaster JM (1998), The Asian-Australian monsoon and El Nino-Southern Oscillation in the NCAR climate system model. Journal o f Climate, 11:1356-1385 Meehl GA and Arblaster JM (2002), Indian monsoon GCM sensitivity experiments testing tropospheric biennial oscillation transition conditions. Journal o f Climate, 15:923-944. Meehl GA, Covey C, McAvaney B, Latif M and StoufFer RJ (2005), Overview o f the coupled model intercomparison project. Bulletin o f the Am erican M eteorological Society, 86:89-93. doi: 10.1175/BAMS-86-1-89 Meehl, GA, Washington WM, Santer BD, Collins WD, Arblaster JM, Hu A, Lawrence DM, Teng H, Buja LE and Strand WG (2006), Climate change projections for the twentyfirst century and climate change commitment in the CCSM3. Journal o f Climate, 19:2597-2616. Meehl GA, Arblaster JM, Caron JM, Annamalai H, Jochum M, Chakraborty A and Murtugudde R (2012), Monsoon Regimes and Processes in CCSM4. Part I: The Asian-Australian Monsoon. Journal o f Climate, 25(8):2583-2608. Moore AM and Kleeman R (1996), The dynamics o f error growth and predictability in a coupled model o f ENSO. Quarterly Journal o f the R oyal M eteorological Society, 122:1405-1446. Mooley DA and B. Parthasarathy (1984), Fluctuations in All-India summer monsoon rainfall during 1871-1978. Clim atic Change, 6:287-301. 163 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon Molteni F and Palmer TN (1993), Predictability and finite time instability o f the northern winter circulation. Q uarterly Journal o f the R oyal M eteorological Society, 119:269298. Morrison H and Gettelman A (2008), A new two-moment bulk 688 stratiform cloud micro physics scheme in the NCAR Community Atmosphere Model (CAM3), Part I: Description and numerical tests. Journal o f Climate, 21:3642-3659. Moron V (1995), Variability o f the African convection centre as viewed by outgoing longwave radiation records and relationships with sea surface temperature patterns. International Journal o f Climatology, 15:25-34. Mureau R, Molteni F and Palmer TN (1993), Ensemble prediction using dynamically conditioned perturbations. Quarterly Journal o f the R oyal M eteorological Society, 119:299-323. Neale RB, Jochum M and Richter JH (2008), The impact o f convection on ENSO: From a delayed oscillator to a series o f events. Journal o f Climate, 21:5904-5924. Neale RB, Richter JH, Andrew JC, Park S, Lauritzen PH, Gettelman A, Williamson DL, Rasch PJ, Vavrus SJ, Taylor MA, Collins WD, Zhang M, Lin S (2010a), Description o f the NCAR Community Atmosphere Model (CAM 4.0). N C AR Technical Note, NCAR/TN-485+STR, National Center for Atmospheric Research, Boulder, Colorado. Neale RB and Coauthors (2010b), Description o f the NCAR Community Atmosphere Model (CAM 5.0). N C AR Technical Note, NCAR/TN-486+STR, National Center for Atmospheric Research, Boulder, Colorado. 164 Ph.D. Dissertation: University of Northern British Columbia Nigam S (1994), On the dynamical basis for the Asian summer monsoon rainfall-El Nino relationship. Journal o f Climate, 7:1750-1771. Palmer TN, Molteni F, Mureau R, Buizza R, Chapelet P and Tribbia J (1993), Ensemble prediction. Proceedings o f the ECMWF Seminar on Validation o f models over Europe: vol. I, EC M W F, Shinfield Park, Reading, RG2 9AX, UK. Palmer TN, Brankovic C, Viterbo P, Miller MJ (1992), Modelling interannual variations o f summer monsoons. Journal o f Climate, 5:399-417. Palmer TN, Alessandri A, Andersen U, Cantelaube P, Davey M, Delecluse P, Deque M, Diez E, Doblas-Reyes FJ, Feddersen H, Graham R, Gualdi S, Gueremy JF, Hagedom R, Hoshen M, Keenlyside N, Latif M, Lazar A, Maisonnave E, Marletto V, Morse AP, Orfila B, Rogel P, Terres JM and Thomson MC (2004), Development o f a European multimodel ensemble system for seasonal-to-interannual prediction (DEMETER). B ulletin o f the Am erican M eteorological Society, 85:853-872. Park S and Bretherton CS (2009), The University o f Washington shallow convection and moist turbulence schemes and their impact on climate simulations with the Community Atmosphere Model. Journal o f Climate, 22:3449-3469. Parthasarathy B, Munot AA and Kothawale DR (1995), All India monthly and seasonal rainfall series: 1871-1993. Theoretical A pplied Clim atology, 49:217-224. doi: 10.1007/BF00867461 165 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon Pokhrel S, Chaudhari HS, Saha SK, Dhakata A, Yadav RK, Salenke K, Mahaptra S and Rao SA (2012), ENSO, IOD and Indian summer monsoon in the NCEP climate forecast system. Clim ate D ynam ics, doi: 10.1007/s00382-012-l349-5 Randall DA, Wood RA, Bony S, Colman R, Fichefet T, Fyfe J, Kattsov V, Pitman A, Shukla J, Srinivasan J, StoufFer RJ, Sumi A and Taylor KE (2007), Climate Models and Their Evaluation. In: Climate Change 2007: The Physical Science Basis. Contribution o f Working Group I to the Fourth Assessment Report o f the Intergovernmental Panel on Climate Change. Cam bridge University Press, Cambridge, United Kingdom and New York, NY, USA Rajeevan M, Unnikrishnan CK and Preethi B (2011), Evaluation o f the ENSEMBLES multi­ model seasonal forecasts o f Indian summer monsoon variability. Clim ate D ynam ics, doi:10.1007/s00 382-011-1061-x. Rasmusson EM and Carpenter TH (1983), The relationship between eastern equatorial Pacific sea surface temperature and rainfall over India and Sri Lanka. M onthly Weather Review, 111:517-528. Rayner NA, Parker DE, Horton EB, Folland CK, Alexander LV, Rowell DP, Kent EC and Kaplan A (2003), Global analyses o f sea surface temperature, sea ice, and night marine air temperature since the late nineteenth century. Journal o f G eophysical Research, 108:4407. doi: 10.1029/2002JD002670 Reynolds RW, Rayner NA, Smith TM, Stokes DC and Wang W (2002), An improved in situ and satellite SST analysis for climate. Journal o f Climate, 15:1609-1625. 166 Ph.D. Dissertation: University of Northern British Columbia Richter J and Rasch P (2008), Effects o f convective momentum transport on the atmospheric circulation in the Community Atmosphere Model, version 3. Journal o f Climate, 21:1487-1499. Ropelewski CF and Halpert MS (1987), Global and regional scale precipitation patterns associated with the El Nin'o/Southem Oscillation, M onthly Weather Review, 115:1606-1626. Rowell DP, Folland CK, Maskell K and Ward MN (1995), Variability o f summer rainfall over tropical north Africa (1906-92): Observations and modelling. Q uarterly Journal o f the R oyal M eteorological Society, 121:669-674. Rodwell MJ and BJ Hoskins (1995), A Model o f the Asian Summer Monsoon. Part II: CrossEquatorial low and PV Behavior. Journal o f Atm ospheric Science, 52:1341-1356. Saji NH, Goswami BN, Vinayachandran PN and Yamagata T (1999), A dipole mode in the tropical Indian Ocean. Nature, 401:360-363. Saha SK, Haider S, Krishna Kumar K and Goswami BN (2011), Pre-onset land surface processes and ‘internal’ interannual variabilities o f the Indian summer monsoon. Climate D ynam ics, 36:2077-2089. Saha S, Nadiga S, Thiaw C, Wang J, Wang W, Zhang Q, Van-den Dool HM, Pan HL, Moorthi S, Behringer D, Stokes D, Pena M, Lord S, White G, Ebisuzaki W, Peng P and Xie P (2006), The NCEP Climate Forecast System. Journal o f Climate, 19:34833517. doi: 10.1175/JCLI3812.1 167 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon Schmetz J and Liu Q (1988), Outgoing longwave radiation and its diurnal variation at regional scales derived from Meteosat. Journal of G eophysical Research, 93(D9): 111 92-11204. Schott FA, Xie SP, McCreary J (2009), Indian Ocean circulation and climate variability. Review s o f Geophysics, 47:RG1002. doi:10.1029/2007RG000245 Sikka DR (1999), Monsoon drought in India. Joint CO LA/CARE Technical R eport No. 2. Center for Ocean-Land-Atmosphere Studies and Center for the Application of Research on the Environment, 93pp Sikka DR (1980), Some aspects o f the large scale fluctuations o f summer monsoon rainfall over India in relation to fluctuations in the planetary and regional scale circulation parameters. Proceedings o f the Indian A cadem y o f Sciences - Earth and Planetary Sciences, 89(2): 179-195. Sikka DR and Gadgil S (1980), On the maximum cloud zone and the ITCZ over Indian longitudes during the southwest monsoon, M onthly Weather Review, 108:1840-1853. Shukla J (1981), Dynamical predictability o f monthly means. Journal o f A tm osphere Science, 38:2547-2572. Shukla J and Mintz Y (1982), Influence of Land-Surface Evapotranspiration on the Earth's Climate. Science, 215:1498-1501. Shukla J and Wallace JM (1983), Numerical simulation o f the atmospheric response to equatorial Pacific sea surface temperature anomalies. Journal o f A tm ospheric Science, 40:1613-1630. 168 Ph.D. Dissertation: University of Northern British Columbia Shukla J and Paolino DA (1983), The Southern Oscillation and long-range forecasting o f the summer monsoon rainfall over India. M onthly Weather Review, 111:1830-1837. Shukla J and Moo ley DA (1987), Empirical prediction o f the summer monsoon rainfall over India. M onthly Weather Review, 115: 695-703. Shukla J (1987), Interannual variability o f monsoons. Monsoons, Ed. J.S. Fein and P.L. Stephens, John Wiley a nd Sons, pp 399-463. Sperber KR, Brankovic C, Deque M, Frederiksen CS, Graham, R., Kitoh A, Kobayashi C, Palmer T, Puri K, Tennant W and Volodin E (2001), Dynamical Seasonal Predictability o f the Asian Summer Monsoon. M onthly W eather Review, 129:22262248. doi: 10.1175/1520-0493(2001 )129<2226:DSPOTA>2.0.CO;2 Sperber, KR, and Palmer TN (1996), Interannual tropical rainfall variability in general circulation model simulations associated with the Atmospheric Model Intercomparison Project. Journal o f Climate, 9:2727-2750. Singh A, Acharya N, Mohanty UC, Robertson AW and Mishra G (2012), On the predictability o f Indian Summer Monsoon Rainfall in general circulation model at different lead time, D ynam ics o f Atm ospheres and Oceans, 58:108-127. Slingo JM and Annamalai H (2000), 1997: The El Nino o f the century and the response o f the Indian summer monsoon. M onthly Weather Review, 128:1778-1797. Smith RD and Gent P (2002), Reference manual for the Parallel Ocean Program (POP). Los A lam os U nclassified Report, LA-UR-02-2484. 169 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon Smith RD and S Kortas (1995), Curvilinear coordinates for global ocean models. L os A lam os U nclassified Report, LA-UR-95-1146 Sohn SJ, Tam CY, Ashok K and Ahn JB (2012), Quantifying the reliability o f precipitation datasets for monitoring large-scale East Asian precipitation variations, International Journal o f Climatology, doi: 10.1002/joc.2380 Soman, MK and Slingo J (1997), Sensitivity o f the Asian summer monsoon to aspects o f seasurface-temperature anomalies in the tropical Pacific Ocean. Q uarterly Journal o f the R oyal M eteorological Society, 123:309-336. Tang Y, Kleeman R and Moore A (2005), On the reliability o f ENSO dynamical predictions. Journal o f Atm ospheric Science, 62(6): 1770-1791. Tang Y, Kleeman R and Miller S (2006), ENSO predictability o f a fully coupled GCM model using singular vector analysis. Journal o f Climate, 19:3361-3377. Tang Y and Deng Z (2011), ENSO bred vectors and predictability in a hybrid coupled model from 1881-2000, Journal o f Climate, 24( 1):298-314. Trenberth KE, Jones PD, Ambenje P, Bojariu R, Easterling D, Klein TA, Parker D, Rahimzadeh F, Renwick JA, Rusticucci M, Soden B, Zhai P (2007), Observations: Surface and Atmospheric Climate Change. In: Solomon S, Qin D, Manning M, Chen Z, Marquis M, Averyt KB, Tignor M, Miller HL (ed) Climate Change 2007: The Physical Science Basis. Contribution o f Working Group I to the Fourth Assessment Report o f the Intergovernmental Panel on Climate Change. Cam bridge University Press, Cambridge, United Kingdom and New York, NY, USA 170 Ph.D. Dissertation: University of Northern British Columbia Toth Z and Kalnay E (1993), Ensemble forecasting and NMC: the generation of perturbations. B ulletin o f the Am erican M eteorological Society, 74:2317-2330. Turner AG, Inness PM, Slingo JM (2005), The role o f the basic state in the ENSO-monsoon relationship and implications for predictability. Q uarterly Journal o f the R oyal M eteorological Society, 131: 781-804 Vavrus S and Waliser D (2008), An improved parameterization for simulating Arctic cloud amount in the CCSM3 climate model. Journal o f Climate, 21:5673-5687. Waliser D, Seo KW, Schubert S and Njoku E (2007), Global water cycle agreement in the climate models assessed in the IPCC AR4. G eophysical Research Letters, 34:L16705. doi: 10.1029/2007GL030675 Walker GT (1924), Correlation in seasonal variations o f weather, IV, A further study o f world weather. M onthly Weather Review, 24:275-332. Walker J and Rowntree PR (1977), The effect o f soil moisture on circulation and rainfall in a tropical model. Q uarterly Journal o f the R oyal M eteorological Society, 103:29-46. Wang B and Fan Z (1999), Choice o f South Asian summer monsoon indices. B ulletin o f the Am erican M eteorological Society, 80:629-638. Wang B, Wu R and Lau KM (2001), Interannual variability o f the Asian summer monsoon: Contrasts between the Indian and the western North Pacific-East Asian monsoons. Journal o f Climate, 15:4073-4090. 171 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon Wang B, Wu R and Li T (2003), Atmosphere warm ocean interaction and its impacts on the Asian-Australian monsoon variation. Journal o f Climate, 16,1195-1211. Wang B, Kang IS and Li JY (2004), Ensemble simulation o f Asian-Australian monsoon variability by 11 AGCMs. Journal o f Climate, 17:803-818. Wang B, Ding QH, Fu XH, Kang IS, Jin K, Shukla J and Doblas-Reyes F (2005), Fundamental challenge in simulation and prediction o f summer monsoon rainfall. Geophysical Research Letters, 32:L15711. Wang B, Bao Q, Hoskins B, Wu G and Liu Y (2008), Tibetan Plateau warming and precipitation change in East Asia. G eophysical Research Letters, 35:L14702. doi: 10.1029/2008GL034330. Wang B, Lee JY, Kang IS, Shukla J, Park CK, Kumar A, Schemm J, Cocke S, Kug JS, Luo JJ, Zhou T, Fu X, Yun WT, Alves O, Jin EK, Kinter J, Kirtman B, Krishnamurti T, Lau NC, Lau W, Liu P, Pegion P, Rosati T, Schubert S, Stem W, Suarez M and Yamagata T (2009), Advance and prospectus o f seasonal prediction: assessment o f the APCC/CliPAS 14-model ensemble retrospective seasonal prediction (1980-2004). C lim ate Dynam ics, 33:93-117. DOI 10.1007/s00382-008-0460-0 Webster PJ (1987), The elementary monsoon. In: Fein JS, Stephens PL (eds) Monsoons. John Wiley, New York, NY, pp 3-32. Webster PJ and Yang S (1992), Monsoon and ENSO: selectively interactive systems. Quarterly Journal o f the R oyal M eteorological Society, 118:877-926. 172 Ph.D. Dissertation: University of Northern British Columbia Webster PJ, Moore A, Loschnigg J and Leban M (1999), Coupled dynamics in the Indian Ocean during 1997-1998. Nature, 401:356-360. Webster, PJ, VO M, Palmer TN, Shukla J, Tomas RA, Yanai M and Yasunari T (1998), Monsoons: processes, predictability and the prospects for prediction. Journal o f G eophysical Research, 103:14451 -14510. Wu X, Liang X and Zhang GJ (2003), Seasonal migration o f ITCZ precipitation across the equator: Why can’t GCMs simulate it? G eophysical Research Letters, 30( 15): 18241828. Wu R and Kirtman B (2004), Inpacts o f the Indian Ocean on the Indian Summer M onsoonENSO Relationship. Journal o f Climate, 17, 3037-3054. Wu G, Liu Y, Wang T, Wan R, Liu X, Li W, Wang Z, Zhang Q, Duan A and Liang X (2007), The Influence o f Mechanical and Thermal Forcing by the Tibetan Plateau on Asian Climate. Journal o f H ydrom eteorology, 8:770-789. Wu R, Chen JL and Chen W (2012), Different types o f ENSO influences on the Indian summer monsoon variability. Journal o f Climate, 25:903-920. Xie SP (1994), On the Genesis o f the Equatorial Annual Cycle. Journal o f Climate, 7:20082013. Xie P and Arkin PA (1997), Global Precipitation: A 17-Year Monthly Analysis Based on Gauge Observations, Satellite Estimates, and Numerical Model Outputs. Bulletin o f the Am erican M eteorological Society, 78:2539-2558. 173 S Islam: Ensemble Simulation and Forecasting of South Asian Monsoon Xue Y, Cane MA and Zebiak SE (1997a), Predictability o f a coupled model o f ENSO using singular vector analysis. Parti: optimal growth in seasonal background and ENSO cycles. M onthly W eather Review, 125:2043-2056. Xue Y, Sales F, Li WP, Mechoso CR, Nobre CA and Juang HM (2006), Role o f land surface processes in South American monsoon development. Journal o f Climate, 19:741-762. Xue Y, Cane MA and Zebiak SE (1997b), Predictability of a coupled model o f ENSO using singular vector analysis. Partll: optimal growth and forecast skill. M onthly W eather Review, 125:2057-2073. Yasunari T (1990), Impact o f Indian monsoon on the coupled atmosphere/ocean systems in the tropical Pacific. M eteorology and Atm ospheric Physics, 44:29-41. Younas W and Tang Y (2013), PNA Predictability at Various Time Scales. Journal o f Climate, 26:9090-9114. doi: 10.1175/JCLI-D-12-00609.1 Yoo JH and Kang IS (2005), Theoretical examination o f a multi-model composite for seasonal prediction. G eophysical Research Letters, 32:L18707. Zhou X, Tang Y and Deng Z (2007), The impact of nonlinear atmosphere on the fastest error growth of ENSO prediction. Clim ate D ynam ics, 30:519-531. DOI: 10.1007/s00382007-0302-5. 174