Introduction

Mountain snowpacks serve as natural water towers, dictating the timing of peak soil moisture, controlling evapotranspiration dynamics, and sustaining streamflow into the dry summer months . A warming climate suggests a future with increased precipitation variability and a decrease in snowpack accumulation and retention . Climate change has also affected frequency and magnitude of disturbances from fire, insects, and drought in forested areas that are seasonally snow covered. Montane forest ecosystems across the western U.S. are particularly vulnerable to these shifting snow dynamics because they have historically relied on snowmelt to support water demand during the growing season . The Sierra Nevada mountains, for example, have a snow-dominated precipitation regime and supply over half of California’s water needs . In the Sierra Nevada, efforts are underway to introduce low-to-moderate severity fires and selective thinning to create forest conditions that promote ecosystem function resembling historical, pre-fire suppression conditions . A knowledge gap in predicting the effects of forest disturbance on snow retention is limiting forest managers’ ability to contend with current and future water supplies .
Precipitation patterns are typically the primary control on snowpack accumulation, while energy input patterns normally dictate ablation; however, terrain and vegetation characteristics serve as secondary controls (. In most locations, orographic precipitation causes deeper snowpacks at higher elevations with limited redistribution of snow by wind . North-facing slopes in the northern hemisphere receive less radiation, particularly during the winter months when sun angles are lower. Therefore, higher elevation, steeper, and more northern-facing slopes in the Sierra Nevada typically both accumulate and retain more snow .
Predicting snow response to forest canopy (and its removal) remains challenging, despite decades of research, because of complex inter- and counter-acting processes controlling the surface energy and water budget . Snowfall is intercepted in forest canopies, where it sublimates rapidly, but predictions of snow interception are uncertain . In warmer regions with minimal wind redistribution, like the Sierra Nevada, up to 60% of precipitation can be intercepted and sublimated from the forest . Forest canopy also changes wind patterns, reducing snow redistribution and changing turbulent heat fluxes , and alters radiation fluxes by both shading the snowpack from shortwave radiation and emitting longwave radiation (Lundquist et al., 2013; Safa et al., 2021). Shortwave radiation typically dominates energy fluxes available for snow ablation in the Sierra Nevada, but longwave radiation from forests is more important when solar irradiance is low and air temperatures are warmer . The interplay between long and shortwave radiation inputs in the winter and spring - when solar irradiance is relatively low - causes snow to ablate more quickly and disappear earlier under forest canopy in warmer places like the Sierra Nevada . However, the impact of forest canopy loss on snow ablation rates across complex topography is not well understood because it cannot be captured well by point-scale or coarser remote sensing and because we lack consistent methodology using newer fine-scale remote sensing.
Forest effects on snow accumulation and ablation are well documented, yet a cohesive framework for predicting forest disturbance effects on snow is lacking . Early studies in the Sierra Nevada showed that forests with open glades accumulated more snow than dense forest patches (Church, 1933; Anderson, 1956, 1963). Varhola et al. (2010) showed through a meta-analysis that reducing the fraction of area covered by forests (fVEG) promoted accumulation, while ablation showed the opposite trend. Their synthesis of 30 studies had large uncertainties that cannot be attributed to local climate or terrain . The Varhola et al. (2010) model was based on coarse-scale forest cover information; however, recent evidence shows that fine-scale forest structure metrics, like tree-scale forest gaps , forest clumps, edges of different vegetation density, and tree height , are critical to simulating larger scale snow distributions and amounts . The methodological framework used by Varhola et al. (2010), which used an open reference site approach, remains a current predictive paradigm but does not utilize information on fine-scale forest structure. Studies commonly focus on binary canopy classifications (i.e., open/ closed) or coarser scale (10-1000 m2) fractional canopy metrics to investigate forest-snow interactions as opposed to finer scale (1-30m2) metrics that classify vegetation structure and density (e.g., leaf area index, or LAI) as well as gap characteristics (e.g., ratio of vegetation heigh to gap size, or openness) .
Fine-scale forest structure and snow information from airborne light detection and ranging (lidar) data have revolutionized our understanding of snow-forest interactions , but remain underutilized for predicting forest management effects on snowpack. Lidar offers substantial advantages over other remote sensing tools because emitted light pulses can penetrate vegetation canopy and be used to create spatially distributed models of the ground/snow surface, canopy height, and canopy density at a decimeter-scale accuracy over extents greater than 100 km2 . An influx of lidar datasets from initiatives like SnowEX and organizations like the Airborne Snow Observatories, Inc. (ASO) and the National Center for Airborne Laser Mapping (NCALM) present unique opportunities to take advantage of high-resolution spatially distributed data with increasing temporal distribution. However, there are few consistent processing methods to resolve snow depth and snow water equivalent (SWE) from lidar in dense forests due to various complications. For example, the number of returns is reduced due to interactions with dense vegetation, low branches can be confounded with the snow surface, and variable snow density adds spatial heterogeneity (both vertically throughout the snowpack and horizontally over the landscape). Moreover, large-scale processing of dense point clouds is computationally intensive and requires expert knowledge of vertical and lateral biases typical in these datasets . Developing a consistent and open-source, lidar-based tool to estimate the effects of vegetation on snow accumulation and ablation has been the focus of the scientific and forest management community for the last several years . Existing decision support tools show varied results when applied at different scales and no current product effectively utilizes snow depth lidar datasets that are becoming increasingly available .
Taking advantage of a unique set of lidar datasets from the densely forested Sagehen Creek Basin in the central Sierra Nevada, California, we investigate how forest structure metrics interact with terrain to control snowpack accumulation and retention. Ongoing management in this study area includes the introduction of low to moderate severity fires and selective thinning to create more resilient forests . There is promise for increasing hydrological outputs from restoring forests, including an increase in snowmelt magnitude and streamflow . We develop a novel method for processing multitemporal lidar datasets to examine patterns over complex topography and heterogenous forest canopy. We pose the following questions:
How does coarser-scale fractional canopy cover influence snow accumulation and ablation as predicted by the current Varhola et al. (2010) paradigm?
What role do terrain (elevation, aspect, and slope) and finer-scale vegetation structure (openness) play on snow accumulation and ablation?
Beyond addressing these questions, a primary goal of this study is to create a replicable, open-source workflow to process point clouds into novel snow and canopy metrics. Our work serves as an important proof-of-concept for a new lidar-based method to develop local forest information relevant to decision support needed in forest restoration planning.