*3.2. Regional Evaluation of SM2RAIN-ASCAT Dataset*

Figure 4 shows the spatial distribution of statistical metrics of SM2RAIN-ASCAT. The bias of SM2RAIN-ASCAT (Figure 4a) shows an almost similar trend compared to SM2RAIN-CCI (Figure 3a). However, the magnitude of biases at each RG is significantly reduced when compared to SM2RAIN-CCI. Precipitation is highly underestimated at the north-east sides of the glacial and humid regions of the country, which gradually decreases towards the west of the regions. The reason for high underestimation might be the abundant water availability in the specific region due to dams, barrages, headworks, and well-developed irrigation systems. Besides, snowmelt and high precipitation events that saturate the soil also contribute to an underestimation of precipitation in high elevated regions (i.e., the glacial region and upstream areas of the humid regions). SM2RAIN-ASCAT highly overestimated precipitation in the south-east of the hyper-arid region. The average (median) values of bias in the glacial, humid, arid and hyper-arid regions were −28.76% (−29.25%), −29.86% (−23.31%), −6.18% (−12.68%) and 34.87% (27.66%), respectively.

SM2RAIN-ASCAT depicts an improved performance in comparison with SM2RAIN-CCI considering ubRMSE (Figure 4b). ubRMSE shows the same spatial trend as compared to SM2RAIN-CCI with maximum and minimum values in the humid and hyper-arid regions, respectively. The regional average (median) ubRMSE values were 3.73 mm/day (3.57 mm/day) in glacial, 7.87 mm/day (7.54 mm/day) in humid, 4.83 mm/day (4.32 mm/day) in humid and 2.21 mm/day (2.19 mm/day) in hyper-arid regions. Maximum and minimum ubRMSE values were depicted by HRG12 (9.91 mm/day) and HARG8 (1.21 mm/day).

SM2RAIN-ASCAT shows high forecasting accuracy at the south-east of the arid region while moderate accuracies in extreme south of hyper-arid region (Figure 4c). Lower accuracies are depicted in humid region mostly in the vicinity of hydraulically developed areas. The regional average (median) values of Theil's U coefficient were 0.40 (0.39) in glacial, 0.46 (0.48) in humid, 0.25 (0.19) in arid, and 0.30 (0.28) in hyper-arid regions. The maximum and minimum of Theil's U values were 0.59 (HRG39) and 0.12 (ARG10).

The comparison reveals that SM2RAIN-ASCAT dominates the SM2RAIN-CCI across all climate regions. The average percentage improvements in humid, arid and hyper-arid regions were: Bias (27.01%), ubRMSE (19.61%), Theil's U (9.80%), MSEs (24.55%), MSEr (19.41%), and KGE (5.26%) in the humid region; Bias (5.94%), ubRMSE (20.16%), Theil's U (28.20%), MSEs (13.83%), MSEr (29.20%), and KGE (28.12%) in the arid region; and Bias (6.05%), ubRMSE (25.56%), Theil's U (26.83%), MSEs (8.22%), MSEr (24.14%), and KGE (24.72%) in the hyper-arid region.

Figure 4d,e represents the spatial distribution of systematic and random errors of SM2RAIN-ASCAT over four climate regions of Pakistan. Higher random errors were observed in the humid region, which decreases towards the arid and hyper-arid regions. The regional average (median) random errors were 15.41% (13.81%), 57.12% (59.87%), 22.81% (21.97%), and 6.04% (6.75%) in glacial, humid, arid and hyper-arid regions, respectively. Overall the random error average across Pakistan was 25.35%. On the other hand, systematic errors were larger than a random error with an average of 74.65% across the whole of Pakistan. Maximum systematic errors were depicted in hyper-arid (average: 93.96%, median: 93.25%) and glacial (average: 84.76%, median: 86.20%) followed by arid (average: 77.19%, median: 78.03%) and humid (average: 41.59%, median: 40.13%) regions.

Figure 4f displays the spatial distribution of KGE score compared with RG observations. Maximum KGE is observed in the hyper-arid region showing a high performance of SM2RAIN-ASCAT in the region. Smaller KGE scores are observed in the humid and glacial regions depicting poor performance in those mentioned climate regions. However, the KGE score illustrates moderate performance in the arid region. Average (median) KGE scores in glacial, humid, arid and hyper-arid regions were 0.36 (0.34), 0.38 (0.41), 0.64 (0.64), and 0.72 (0.77), respectively.

**Figure 4.** Spatial distribution of bias (**a**), ubRMSE (**b**), Theil's U coefficient (**c**), systematic error (**d**), random error (**e**), and KGE score (**f**) based on a daily scale across Pakistan from SM2RAIN-ASCAT compared to RGs data for the period of 2007–2015.
