Performance assessment of SM2RAIN-NWF using ASCAT soil moisture via supervised land cover-soil-climate classification

The SM2RAIN algorithm developed a simple analytical relationship by inverting the soil-water equation to es-timate rainfall through the knowledge of soil moisture. The recently developed SM2RAIN-NWF algorithm offers an improvement in estimating rainfall by integrating the SM2RAIN algorithm and the net water flux (NWF) model. The Advanced Scatterometer (ASCAT) soil moisture products were used to estimate rainfall and evaluate the reliability of the SM2RAIN-NWF algorithm compared to the SM2RAIN on a national scale. Besides, the impact of Land cover-Soil texture-Climate (LSC) characteristics and the intensity of rainfall (four classes of intensity) on the performance of algorithms were discussed. Five performance metrics, including Correlation Coefficient (R), Kling-Gupta (KGE), Root Mean Square Error (RMSE), False Alarm Ration (FAR), and Probability of Detection (POD) were used to validate the estimated cumulative 5-, 14-, and 30-day rainfall. Furthermore, the effect of evapotranspiration (ET) and drainage terms were investigated in the performance of rainfall estimation through the SM2RAIN-NWF algorithm for the first time on a national scale. Results showed the rainfall estimations through the SM2RAIN-NWF algorithm improved approximately up to 7.5% in each accumulation (e.g. rainfall aggregation intervals (AGGR) 5 to 14 and 14 to 30) based on R and KGE indices. In addition, the SM2RAIN-NWF improved rainfall estimations up to 50% based on the KGE index in the southern half of Iran (arid and semi-arid climate) compared to the SM2RAIN estimates. The comprehensive evaluation and uncertainty analysis of rainfall estimations under the supervised classification of 11 LSC and 4 rainfall classes also showed the calibration of the SM2RAIN-NWF was highly affected by environmental and climatic circumstances. Uncertainty analysis showed the SM2RAIN-NWF algorithm can estimate rainfall more consistently in the five LSC classes namely 1) barren -clay loam-arid-desert, 2) barren-loam-arid-steppe, 3) barren-clay loam-arid-steppe, 4) urban-clay loam-arid -desert, and 5) urban-loam-arid-steppe. Similarly, estimating rainfall in the region with precipitation under 267 mm/year can be retrieved more reliably through the SM2RAIN-NWF algorithm. Results obtained from the ET analysis revealed an insignificant (<4%) effect of this term in improving the performance of the SM2RAIN-NWF algorithm. Also, the NWF equation confirmed the paramount role of the drainage term in enhancing the accuracy of rainfall estimates compared to the SM2RAIN method.

Automated summary

The SM2RAIN algorithm estimates rainfall by inverting the soil-water equation based on soil moisture knowledge. The SM2RAIN-NWF algorithm further improves the estimation by integrating the SM2RAIN algorithm and the net water flux model. The performance of the algorithms is influenced by LSC characteristics and rainfall intensity.