Characterizing Soil Heterogeneity at the Watershed Scale

Mountainous headwater watersheds have an outsized impact on water resources across the world. Snowmelt from headwater watersheds supports a significant fraction of regional water supplies. Climate change - temperature increases and early snowmelt - could have a significant impact on soil moisture, soil biogeochemistry, ecosystem and water resources/quality. Soil - mixture of various mineral particles, organic matters and organisms- plays a critical role in supporting the ecosystem, and regulating water and nutrient cycling. The soil heterogeneity is yet extremely difficult to characterize over the watershed scale, since direct measurements are limited to sparse core samples. Recently, there have been many attempts to map soil based on remote sensing (RS) data as proxies. However, the RS applications are often limited in mountainous areas due to complex terrains as well as lack of training data.

The goal of this study is to characterize the soil variability, based on over 500 samples across the East River watershed. The soil properties include soil texture (e.g., clay/soil content), soil carbon content, soil density and other geochemical properties (such as pH). Statistical and machine learning methods are applied to (1) quantify the variability across scales (i.e., plot-scale, watershed-scale), (2) compute the correlations between soil properties and spatially extensive RS data, geophysical data and other spatial data layers, and (3) estimate soil properties over space as a function of RS data.