Soil moisture decline across the conterminous United States

Abstract:

Changes in soil moisture (SM) are directly related to food and environmental security around the world. Furthermore, precise information about temporal and spatial patterns of SM is crucial for realistic interpretations of environmental change and policy relevant research. This study shows how data fusion of topography (represented by a digital elevation models and derived terrain attributes) and annual SM (represented by remotely sensed microwave observations from 1978 to 2013) enhance spatial detail and improves the correlation between remotely sensed and ground truth SM observations. On average, topography explains 80% of remotely sensed soil moisture variability using a kernel-based form of regression with a RMSE of 0.026 via cross-validation. Predictions of annual SM were generated across the conterminous United States at 1km pixel size for the 36 years of available data. Previous studies report that SM remote sensing data, derived from microwave observations (~27km pixel size), is representative of the first 2 cm of soil depth. We found that field SM measurements best correlates with our 1 km SM product at 60 cm soil depth (R²= 0.52). Furthermore, by averaging field SM measurements between 25 and 60 cm the correlation improved to R²= 0.62. Our results show a similar negative temporal trend for field SM observations and our predicted SM product at 1km pixel size. For both cases the slope is showing a reduction of -1.87 (-0.64 -3.23) %/year and -0.92 (-0.15, -1.25) %/year, respectively. We found a consistent decline of SM at the national level, and a sharp decay in 2012 and 2013. The temporal variability of SM is partially (~51%) explained by spatial and temporal trends of precipitation and temperature across the United States. These results provide insights of alternative approaches to estimate SM trends across continental-to-global scales.