Machine learning (ML) has emerged as an effective tool for estimating snow depth and snow water equivalent at unsampled times and locations. Airborne lidar surveys are particularly useful for ML applications: the high-resolution, high-precision snow depth data allow for algorithm training and testing to an extent not previously possible. Here, we train a random forest model to estimate snow depth relative to a nearby Snotel site using physiographic data and other dynamic snowpack information as predictor variables and lidar for the target variable. The model output is daily, 50 m resolution snow depth for basins that have both lidar and Snotel data in Colorado. We evaluated multiple approaches for random forest training: using historic lidar data in a basin (temporal transfer), using lidar data from other basins in a region (spatial transfer), and both together. All scenarios demonstrate success, with RMSE values of 0.38-0.45 m at 50 m resolution, indicating that information from lidar can be transferred to different times and locations within the region. When upscaled to the 1-30 km scales, the model has RMSEs of 0.13-0.15 and biases of 0.01-0.03. The model scenario which includes both temporally and spatially transferred lidar data is the most robust to the number and timing of lidar surveys used in model training. This framework extends the spatial footprint of Snotel and the temporal coverage of lidar by leveraging the strengths of the two datasets, with applications for water resource management and validation of gridded snow products.

Measurements of biophysical tree properties and hydrologic fluxes are necessary for improving models and monitoring the impact of disturbances. Prior research has demonstrated that measurements of tree sway frequency can be used to quantify important ecohydrologic processes, such as drought stress and snow interception, that otherwise require expensive measurement techniques. However, existing instruments used to measure tree sway lack spatial scalability. We investigate whether the virtual vision sensor and multilevel binary thresholding video processing algorithms can be used to accurately extract tree sway frequency at multiple points in a video camera field of view and enable scalable measurements of ecohydrologic processes. Comparing sway frequencies extracted from video and accelerometer data at two sites, we show that for 30-60 s videos, the video processing algorithms can reproduce accelerometer sway frequencies with ±0.03 Hz accuracy. The results suggest that video processing algorithms may be suitable for applications where changes in sway frequency are on the order of tenths of hertz or larger, for example the measurement of snow in trees. Further work is needed to clarify the accuracy of the algorithms when applied to longer videos, which may be required to monitor processes with more subtle changes in sway frequency, such as diurnal changes in tree water content.

Snowpack accumulation in forested watersheds depends on the amount of snow intercepted in the canopy and its partitioning into sublimation, unloading, and melt. A lack of canopy snow measurements limits our ability to evaluate models that simulate canopy processes and predict snowpack and water supply. Here, we tested whether monitoring changes in wind-induced tree sway can enable snow interception detection and estimation of canopy snow water equivalent (SWE). We monitored hourly tree sway across six years based on 12 Hz accelerometer observations on two subalpine conifer trees in Colorado. We developed an approach to distinguish changes in sway frequency due to thermal effects on tree rigidity versus intercepted snow mass. Over 60% of days with canopy snow had a sway signal in the range of possible thermal effects. However, when tree sway decreased outside the range of thermal effects, canopy snow was present 93-95% of the time, as confirmed with classifications of PhenoCam imagery. Using sway tests, we converted significant changes in sway to canopy SWE, which was correlated with total snowstorm amounts from a nearby SNOTEL site (Spearman r=0.72 to 0.80, p<0.001). Greater canopy SWE was associated with storm temperatures between -7 C and 0 C and wind speeds less than 4 m/s. Lower canopy SWE prevailed in storms with lower temperatures and higher wind speeds. We conclude that monitoring tree sway is a viable approach for quantifying canopy SWE, but challenges remain in converting changes in sway to mass and further distinguishing thermal and mass effects on tree sway.