*3.1. Preliminary Assessment of MODIS ATI Surface Soil Moisture*

To decide and select the best approach for the MODIS ATI SSM before applying the SWI model, different MODIS ATI SSM estimations were tested against the in situ SSM across REMEDHUS (Table 1). In general, the correlations obtained from the MODIS ATI SSM (R ~0.32 to 0.70) were lower than those obtained from other SSM products, such as SMOS, SMAP, and CCI SSM, confirmed in previous validation studies [1,45,59,60]. However, the errors were very similar (cRSMD ~0.04 m3/m3).

Analyzing the different albedo approaches, no significant differences were found between using the *α*shortwave or *α*visible in terms of correlation (maximum R ~0.66), errors (minimum RMSD ~0.040 m3/m3 and minimum cRMSD ~0.030 m3/m3), bias, and coverage. The capabilities of both albedo approaches to characterize the full range of vegetation were comparable, as shown in other studies [61]. Therefore, because the *α*shortwave computation required reflectance data from a higher number of MODIS bands, *α*visible was preferred.

Considering the different methods to compute the diurnal temperature range, it was clearly seen that the Δ*LST*4values displayed correlation and bias similar to those of the remaining approximations although with lower errors (RMSD ~0.035 to 0.133 m3/m3 and cRMSD ~0.019 to 0.085 m3/m3). However, higher coverages were obtained when using Δ*LST*Aqua (N ~35 %), Δ*LST*Terra (N ~40 %), and Δ*LST*Aqua/Terra (N ~50 %). These results mean that the availability of the MODIS ATI SSM was mainly limited by the availability of the LST observations. Previous research used similar

approximations for Δ*LST* [57] or complex methods to interpolate the LST data [12,13,62,63]. In this study, the Δ*LST*Aqua/Terra was chosen for the estimation of the diurnal temperature range.

Regarding the different dynamic ranges used as reference, the Soil Hydrologic Database (SHD) afforded statistics similar to those from the CCI approach. This fact, together with the higher spatial resolution of SHD compared to CCI (1 vs. 25 km), supported the selection of the SHD for the dynamic range.



All values are significant (*p*-value < 0.05).
