2.3.1. SM2RAIN-CCI

The European Space Agency Climate Change Initiative (ESA-CCI) released the ESA CCI SM (Soil Moisture) v3.2 datasets in early 2017. SM retrievals from active and passive microwave instruments onboard different satellite platforms were merged to develop the dataset, which provides global daily SM estimates with 0.25◦ spatial resolution from 1978–2015 [52–54]. Accumulated precipitation estimates between 00:00 to 23:59 UTC are obtained from an application of a weighted average-based integration procedure. Quality flag provided within the raw SM observations (i.e., EA CCI SM v3.2) is used to remove the low-quality data and observations distinguished by retrieval issues (e.g., glacial regions, frozen soil, complex topography, and dense vegetation). The SM2RAIN algorithm was calibrated on a pixel-by-pixel basis during the periods of 1998–2001, 2002–2006, and 2007–2013 against Global Precipitation Climatology Centre Full Daily Data (GPCC-FDD) [37]. These different calibration periods are dependent on types of data and sensors that have been utilized in developing the active and passive SM datasets.

The SM2RAIN algorithm, a novel approach, estimates precipitation from SM using the inverted soil water balance equation [9]. The basic assumption of the SM2RAIN algorithm is that surface runoff and evapotranspiration rate are insignificant during the precipitation event [9,32]. The simple form of soil–water balance equation can be illustrated as

$$p(t) = Z^\* \frac{ds(t)}{dt} + a \cdot s(t)^b \tag{1}$$

where *p*(*t*) is precipitation rate [L/T], *Z*\*=*nZ* is soil water capacity [L] in the soil layer with a depth of *Z* and porosity of *n*, *s*(*t*) represents relative soil saturation, *t* is time [T], and the two parameters *a* and *b* represent the nonlinearity between drainage and soil saturation. The parameters *Z*\*, *a*, and *b* are calculated through calibration analysis [37]. The main limitation of the SM2RAIN algorithm is that it is not able to estimate the precipitation when the soil is saturated or nearly saturated, because the SM algorithm is unable to derive SM variation as the SM approaches saturation [30]. A more detailed description of SM2RAIN algorithm can be found in [32]. In the current study, v2.0 of the SM2RAIN-CCI product, available from 1 January 1998 to 31 December 2015 (released in July 2018) with 0.25◦ spatial and daily temporal resolutions, is used.
