Figures

Figure 1. Sagehen Creek Basin, located in the northeastern Sierra Nevada. The area of interest (black outline) shows the study domain given available lidar and modeled snow density. a. The elevation profile spans ~1800 to 2800 meters, increasing to the southwest. b. Northness shows steep slopes to the southwest. c. The thinning project summary shows proposed thinning boundaries overlaying a canopy height model (CHM) in meters, which align well with the fraction of vegetation (fVEG) difference between the pre- and post-thinning flights (d).
Figure 2. SNOTEL SWE values compared with flight dates (vertical dashed lines) used in this study for water year (a) 2008, (b) 2016, and (c) 2022. Orange = Independence Lake SNOTEL Site (541), Black = Independence Camp SNOTEL Site (539), Blue = Independence Creek SNOTEL Site (540).
Figure 3. (a) Example image displaying the lidar snow-on classification with an example subsection of a lidar grid cell (tan = ground, green = vegetation, blue = snow surface). The bottom color bar aligns with the correct canopy cover class as classified from the height strata (dark blue = open, light blue = short canopy, dark green = tall canopy, light green = understory). The black zones indicate that the area above would be removed from the analysis, either because there are points between 1.5-3 m or because the snow is not >0.3 m above the vegetation surface (note: not to scale and point cloud colors do not correspond with classification colors). b. Vegetation classification output. The refined classification includes short and understory classes, allowing us to include shorter vegetation types if the snow depth is ≥ 0.3 m above the vegetation. c. Snow-depth classification example for the 04/17/2016 flight.
Figure 4. Spatial distribution of pre-disturbance forest structure metrics across the area of interest, colored by northness, where negative values indicate south-facing slopes. Box plots show quartiles with the horizontal bar representing the median value. Higher elevations tend have denser canopy on northern-facing slopes.
Figure 5. Spatial distribution of SWE across the domain. We see greater SWE at higher elevations and more northern-facing slopes.
Figure 6. Median DSnowAc (shown by horizontal lines within each box plot) is always below zero, indicating greater accumulation in open areas. The spatial distribution of DSnowAc across the domain is varied with a more pronounced difference at lower elevations on more northern-facing slopes in 2016 and more muted effects in 2008 and 2022.
Figure 7. DSnowAc against fVEG for the a. NCALM flight from 02/10/2008. b. ASO flight from 03/26/2016 c. NCALM flight from 03/21/2022. Lidar points are colored by density with darker blue = greater point density. Black points show the Varhola et al. (2010) data.
Figure 8. Normalized DSnowAb against fVEG for both early (a) and late (b) season ablation show a weak positive relationship, with the differences between forested and open sites increasing (becoming more positive) with increasing vegetation.
Figure 9. Annotated partial plots showing the response area determination. Higher pre-disturbance fVEG areas are identified as high response areas because thinning would lead to a pronounced increase (positive response) in DSnowAc whereas lower pre-disturbance Openness (ln[gap diameter/mean tree height]) areas are identified as high response because thinning would lead to a pronounced decrease (negative response) in DSnowAc.
Figure 10. Predicted response classes based on the Random Forest results. Shading shows areas that are predicted to experience a low (light purple), moderate, or high (dark purple) response to treatment. a. The spatial distribution of these maps on top of actual treatments, shown both as planned treatment outline (dotted black line) and the delta fVEG raster. b. Snow accumulation dynamics across all three flights and predicted response classes.
Figure 11. An expanded response analysis using the 2014 lidar flight. The LANDFIRE EVT map (a. and c.) shows that the domain is dominated by Ponderosa Pine Forests, which are also primarily where areas are impacted by fire. Moderate response zones account for the majority of the domain, and high response areas are concentrated between Douglas/Grand/White Fir and Ponderosa Pine Forests (c.). Treatment and burned areas overlain on an expanded response map show that low response areas overlap with burned areas (b.) and taking the area of pixels per each date range, the proportion of low response areas increases as the treatment dates become more recent (e.g. 2007-2014). Planned treatment areas (2014-2024) show an increase in high response areas.