Development of High Resolution Multi-Layer Soil Moisture Information

Abstract:

Soil moisture is an important land surface variable responsible for land-atmosphere interactions. Operational soil moisture missions such as Soil Moisture Active Passive (SMAP) successfully retrieve surface soil moisture (0-5 cm) with reasonable accuracy. However, there is a greater need for us to understand the soil moisture variability in the deeper layers. To this effect, SMAP soil moisture is assimilated into land surface models to estimate rootzone soil moisture over a profile depth of 0-100 cm (9 km resolution). Multi-layer soil moisture information at high resolution is necessary for the agriculture sector (irrigation water application) and also towards modeling extreme events, among other areas of research.

In this work, we develop a high-resolution multi-layer soil moisture information over the Contiguous United States (CONUS) region. The CONUS has a dense in-situ soil moisture station network comprising 695 stations, which monitor soil moisture at multiple layers. We use this data in a machine learning framework along with Soil Moisture Active Passive (SMAP) Level 4 surface (0-5 cm) and rootzone (0-100 cm) soil moisture, geomorphological, topographical, climate, and vegetation data to estimate soil moisture across 5 layers – 5, 10, 20, 50, and 100 cm depths at 1 km resolution.

Model testing resulted in ubRMSE range of 0.0294-0.0432 m³/m³, R range of 0.9058-0.9605, and bias of 0.0002-0.0011 m³/m³ across the five layers. Better accuracy is achieved in the regions of low to moderate vegetation, and flatter topography. Interestingly, the importance of climate variables reduced in the models as the depth increases. The models are noticed to have given more importance to the soil textural information in the deeper layers. This result is in alignment with several studies in the literature, which indicated the strong dependency of textural properties with the deeper later soil moisture. The newly developed soil moisture product is validated at 86 locations. Results suggest that the new soil moisture product successfully captures the temporal dynamics across the five layers with reasonable accuracy. There is also an agreement between the spatial patterns of surface and rootzone (aggregated) soil moisture of SMAP Level 4 product and newly developed product.