*4.3. Graphical Assessment of Spatial Distribution of Downscaled Soil Moisture*

After the downscaled SM was validated using in situ sites, we proceeded to conduct a spatial analysis to evaluate the effectiveness of the downscaling approaches. The downscaled SM from the XGBoost model was visually compared with high-resolution MODIS EVI and MODIS ET products. The Enhanced Vegetation Index (EVI) is an indicator used

to assess and monitor the health and growth status of vegetation [55]. When SM is low, vegetation may be constrained by water availability, leading to slowed or stressed plant growth. This may manifest as lower EVI values. Conversely, when SM is high, plants may have ample water supply, promoting growth and resulting in higher greenness and elevated EVI. Evapotranspiration (ET) refers to the sum of evaporation from the land surface and transpiration from plants [56]. When SM is high, there is ample water supply in the soil, and plant roots can absorb sufficient water for transpiration, thereby promoting the ET process. Higher SM typically results in higher ET. Conversely, when SM is low, the water supply in the soil decreases and plants face water limitations, leading to reduced plant transpiration. Lower SM typically results in lower ET. Therefore, examining the variations in EVI and ET within the study area can indirectly reflect changes in SM.

The following sections compare the relationships among the downscaled SM, EVI, and ET for four periods: 9–14, 10–16, 11–17, and 12–19. We processed the downscaled SM from the XGBoost model using simple Kriging interpolation, and then conducted a spatial analysis with MODIS EVI and MODIS ET products.

As seen in Figure 10, the EVI values in the central and eastern of the study area are relatively high, while those in the northwest and southwest are lower. This is related to the vegetation cover in the study area and is consistent with the geographical characteristics of the study area described in Section 2.1. Compared with the downscaled SM and EVI at the same time, we can observe that areas in the study region with higher SM also have higher EVI values, such as the sides of the Central Valley in the middle and the Appalachian Mountains in the east. Conversely, areas with lower SM also have lower EVI values, such as in the western regions of Oklahoma and Salt Lake City. It can be proved that there is a correlation between SM and EVI. As seen in Figure 10(a-1), on September 14, the SM values in the Homochitto National Forest, the Sabine National Forest in the south-central study area, and the southeastern region are relatively high. Comparing this with Figure 10(a-3) at the same time, we can see that the ET values in these areas are also high. The same pattern can be observed when comparing Figure 10(b-1) with Figure 10(b-3), Figure 10(c-1) with Figure 10(c-3), and Figure 10(d-1) with Figure 10(d-3). SM, as one of the main sources of water for ET, may lead to higher ET in areas with higher SM. However, it is worth noting that there are some discrepancies in the spatial distribution of downscaled SM, EVI, and ET in some areas in the south-central study area (areas within the red box in Figure 10). Because the downscaling model is established at a 36 km grid scale, some extreme values are smoothed during the spatial aggregation process. As a result, the training samples selected in the model construction process are all smooth data, with fewer extreme values. This is not unique to our study, as all existing downscaling methods necessitate calibration with coarse-resolution data initially, making the aggregation of high-resolution predictors inevitable [26]. The result is that the downscaled SM has some mistakes.

Overall, the spatial distribution and temporal variation of the downscaled SM product generated in this paper are relatively consistent with EVI and ET, both of which have a certain correlation with SM. Therefore, this indirectly verifies the accuracy of the downscaled SM for the retrieval of SM in the study area.

**Figure 10.** Distributions of the downscaled CYGNSS SM, MODIS EVI, and MODIS ET on 14 September 2019, 16 October 2019, 17 November 2019, and 19 December 2019.
