Toward a self-calibrated and independent SM2RAIN rainfall product

Rainfall monitoring is fundamental in many hydrological applications such as flood and landslide forecasting and water resources management. In-situ measurements are the traditional data source of rainfall, but the worldwide declining number of stations, their low spatial representativeness and the data access problem limit their use. Satellite products are being widely used as an alternative data source. Among them, SM2RAIN-based products, which exploit the inversion of the water balance equation to derive rainfall from soil moisture observations, have shown relatively good skills for hydrological applications. However, the need of calibrating the SM2RAIN parameter values against a reference represents one important limitation, particularly over data scarce regions.In this study, we explore the possibility to self-calibrate SM2RAIN and thus to obtain rainfall estimates from the Advanced SCATterometer (ASCAT) soil moisture independently from any reference rainfall dataset. Four parametric relationships relating SM2RAIN parameter values to static descriptors (average rainfall, topography, soil moisture noise) are developed. To develop such relationships, a sample of 1009 points uniformly distributed over the areas covered by rain gauges in Australia, India, Italy and the United States is selected. A global validation of the methodology is conducted by comparing the performances of the parameterized product with the classical product in which the parameter values are estimated by calibration against a reference rainfall dataset. The Final Run of the Integrated Multi-Satellite Retrievals for Global Precipitation Measurement (IMERG) precipitation dataset is used for performance assessment, together with the triple collocation techniques by using the gauge-based Global Precipitation Climatology Center (GPCC) product and the Late Run of IMERG.The aim of the analysis is to obtain an uncalibrated SM2RAIN methodology to retrieve rainfall whose performance are similar to those obtained with calibration. The results at 1009 points show that the performances of the parameterized SM2RAIN product are in line with those of the calibrated one, with an increased capability in the detection of intense rainfall events and an acceptable reduction of the performance according to both Pearson Correlation and Root Mean Square Error indexes. The application of triple collocation confirms these findings on a global scale, showing that the SM2RAIN product outperforms both GPCC and IMERG - Late run estimations in areas characterized by low density of rain gauges and good quality of ASCAT soil moisture retrievals (i.e., Africa and South America.