2.2.1. Satellite-Based Precipitation (GPM IMERG)

The GPM mission was launched on 27 February 2014 by NASA and JAXA as an international joint project for frequent measurements of near-global precipitation. Like the antecedent mission TRMM, the GPM is a constellation of multiple satellites, which comprise the GPM Core Observatory carrying combined passive/active sensors, and microwave measurements by partner satellites. The Core Observatory information acts as a reference standard to combine other satellites' microwave precipitation measurements that orbit within the same constellation. The GPM Core Observatory technically consists of two major sensors: GPM Microwave Imager (GMI) and Dual-frequency Precipitation Radar (DPR). The GMI sensor measures precipitation characteristics, such as intensity, type, and size, while DPR observes the internal structure of storms within and under the clouds and measures precipitation in 3D throughout the atmospheric column and provides an insight into the structure of rain drops [7].

The success of the GPM mission depends upon algorithm development that combines both GMI and DPR observations and the partner satellites' measurements. GPM data are available online at three various processing levels, including geo-located and calibrated (Level 1), geophysical data product derived from Level 1 data (Level 2), and a composite of Level 2 data products (Level 3) [35]. The GPM Level-3, namely IMERG, is derived from multiple satellites and available as three different types of daily products consisting of early run (IMERG-Early), late run (IMERG-Late), and final run (IMERG-Final), which are published with 4-h, 12-h, and 2.5-month latencies, respectively, and a monthly product (IMERG-Monthly). These products are accessible at 0.1◦ × 0.1◦ spatial resolution for regions between 60◦ S and 60◦ N globally and at several temporal resolutions, i.e., half-hourly, 3-hourly, daily, and monthly [15]. More detailed information and an algorithm description can be found in [35]. While IMERG-Early provides a quick estimate with only a 3-h latency period, IMERG-Late successively presents better estimates as data from more partner satellites is merged with a 12-h latency period. IMERG-Final estimates are, however, presented after bias adjustment with monthly rain-gauge measurements with a three-month latency. The IMERG-Early data can be used for potential flood or landslide warnings while the IMERG-Late and Final data can be used in agricultural forecasting, drought monitoring, and hydrological modeling [24].

It is noted that the IMERG-Final run algorithm provides two different types of precipitation estimates, including precipitationCal (with rain gauge-adjusted processes) and precipitationUnCal (without rain gauge-adjusted processes). Based on the literature, precipitationCal has indicated better performance relative to ground-based measurements [25]. Therefore, this algorithm together with the IMERG-Early, -Late, and -Monthly products from April 2014 until December 2017 were used in this study.
