————————————————–

3.3 Coarse-resolution surface soil moisture product

In 2012, the ESA CCI project for SSM was established to fulfill global long-term SSM monitoring by merging multiple available active and passive microwave-based SSM products (Wagner et al., 2012). That same year, the first SSM product from the ESA CCI (v0.1) was publicly released. By involving new sensors and improving the merging scheme, the subsequent SSM dataset has been updated over an extended spatiotemporal coverage. The daily SSM product provides a consistent SSM record from 1978 to the present. The latest version (v04.4) of the ESA CCI SSM product at depths of 0.5–5 cm was used in this study, with a spatial resolution of 0.25 degree (https://www.esa-soilmoisture-cci.org). The SSM data were interpolated and resampled to 25 km × 25 km regular grids (Figure 2).

3.4 MODIS products

The Moderate Resolution Imaging Spectroradiometer (MODIS) is a key instrument onboard the Terra and Aqua satellites. Fine-resolution ancillary variables LST, NDVI, BSA and LC information were collected from the Version 6 products of Aqua MODIS (https://lpdaac.usgs.gov/). The daily LST and 16-day NDVI were provided by MYD11A1 and MYD13A2 at 1-km resolution, while the 16-day albedo and annual LC were provided by MCD43A3 and MCD12Q1 at 500-m resolution. The BSA data were calculated from shortwave radiation of MCD43A3, which uses a linear combination of the black-sky and white-sky albedo data, with weights of 0.34 for the former and of 0.66 for the latter. All MODIS products were reprojected consistently with the ESA CCI product. Missing values were filled using the aforementioned spatiotemporal prediction method to ensure complete coverage. The LST and NDVI data were resampled and aggregated into 1 km × 1 km and 25 km × 25 km regular grids. The average aggregations of BSA and modal aggregations of LC were achieved at both fine and coarse grids.

3.5 DEM products

The DEM at 90-m resolution provided by the NASA Shuttle Radar Topographic Mission (SRTM) within the study area was employed. The Void Filled DEM product was downloaded fromhttps://www.usgs.gov/centers/eros. The DEM data were resampled into 1 km × 1 km and 25 km × 25 km regular grids by using average aggregations. The basic terrain factors at 1 km and 25 km, including aspect and slope, were calculated from the DEM information.