Evaluation of soil moisture estimates from satellite based and reanalysis products over two network regions.

Abstract

The soil is an important variable of the hydrological cycle. It plays a key role in the distribution of water and energy fluxes between the surface and atmosphere. Soil moisture data can be used to develop early warning systems for flood and drought monitoring, improve weather and climate forecasting and provide an indication of crop water requirements. Therefore, the regular monitoring of this variable can prove to be beneficial to various management applications. One of the main issues associated with estimating soil moisture is to adequately account for its spatial and temporal variability as it is influenced by factors such as climate, topography, soil properties and land cover. There are different methods available to derive soil moisture estimations such as in-situ, remote sensing and modelling-based approaches. In-situ methods generally produce reliable soil moisture estimates, however, are only suitable for small scale studies. Alternatively, remote sensing and modelled reanalysis methods can provide soil moisture estimates over a large spatial extent, however, they are generally limited by their coarse spatial resolutions and may not be suitable for localised applications. Therefore, the aim of this study was to implement and evaluate a downscaling technique across two regions (South Africa and USA) to ultimately produce finer scale soil moisture and address the scale mismatch between in-situ methods and coarse resolution products. This procedure was facilitated by two data processing platforms, Google Earth Engine (GEE) and R, which showed significant potential for data processing and analysis. Additionally, satellite-based and reanalysis products were also evaluated to determine which of these methods are more suitable for soil moisture estimation. The soil moisture products and the downscaled products were validated against the CRNS instrument, which was particularly chosen for its performance at an intermediate spatial resolution. The SMAP_25 km product performed best at the Two Streams site and was selected for downscaling, whilst the CFSV2 product performed best at the Mead CSP3 and York Benny catchments and was chosen to be downscaled at both these sites. The results from the study indicated that the downscaled products for the Two Streams and Mead CSP3 sites performed better than the original products when compared to the CRNS data. The data acquired for the York Benny site revealed that the downscaled product performed similarly to the CFSV2 product. Therefore, downscaling does not always result in an improved outcome. However, from the results acquired for the Two Streams and Mead CSP3 study sites, it is evident that downscaling shows significant potential in producing better soil moisture estimates, which could be used to improve planning and management operations for various purposes.