Assessment of LiDAR snow depth measurements and the spatiotemporal variability of the snowpack in a forested watershed on Vancouver Island

This study uses repeat airborne LiDAR surveys to assess snow distribution within complex forested terrain in Russell Creek, a 33 km2 sub-basin of the Tsitika watershed, on northern Vancouver Island. LiDAR surveys used within this study were acquired in 2022 and 2023, with 4 – 5 flights per year timed with the intent to capture maximum snow depth, to the completion of the melt season. In addition, two separate bare earth models were used to assess error by acquisition date (2017 vs 2023). Over 2000 manual snow depth measurements were taken over the course of this study via a) cardinal plots within six different forest cover types and b) weather station snow courses conducted on gravel roads. Mean difference (MD) of the averaged manual and LiDAR measurements were overall lowest for the weather station snow courses (-29 to 13 cm) and highest in the harvested (juvenile and regenerating) plots (-39 to 134 cm). Analysis of the LiDAR bare earth point returns showed that juvenile and regenerating plots both had a low percentage of pixels with ground returns – average of 24 and 35% respectively - a result of the tall (~ 2 meter) and complex ground vegetation. Error was reduced in the juvenile forest plots (-22 to 29 cm) when LiDAR snow depth was processed with an alternative bare earth model acquired in 2017, in which the average coverage of ground returns was much greater (96%). Total snow storage within Russell Creek was relatively similar (<5% difference) between the two bare earth models, and the higher elevations (>1000 m) of the watershed - comprised of old growth forest and alpine cover types – store the majority (56 – 82%) of the snow. At high elevations (1400 – 1700 m) in the watershed, the snow volumes were much greater (28 – 44%) when processed with the 2023 bare earth model compared to the 2017. An additional component to this study was a paired control-treatment approach to assess the impacts of forest harvest on a) the bare earth model and b) snow distribution. Evaluation of pre- (2020) and post-harvest (2023) snow free models showed a slight (0.15 m) bias within treatment sites due to changes to the ground surface. After bias correction, the rate of increase in snow water equivalence between the pre-and post-harvest snow years in the treatment sites (1.4 – 4.2) doubled the control (0.7 – 1.7). Overall, these results highlighted the importance of the bare earth model to accurately measure snow, especially within complex forested and harvested watersheds.