Multitemporal Validation of IMERG Precipitation Estimates over South Korea

The Integrated Multi-satellite Retrievals for GPM (IMERG) is a quasi-global gridded precipitation product that unifies measurements from a network of satellites in the GPM constellation. IMERG covers up to +/- 60^olatitudes with a high resolution of 0.1^oevery half hour. To accommodate the various user requirements for data latency and accuracy, the IMERG is generated in the form of near-real-time data (IMERG Early and Late Runs) and post-real-time research data (IMERG Final Run).Multitemporal validation of IMERG for a specific region is essential to the algorithm improvement and product applications. Substantial efforts have been devoted for the validation of IMERG precipitation estimates against ground observations such as radars and gauges, in different areas of the world. This study validates the Version-05B IMERG Early, Late and Final Runs using ground-based Korean Quantitative Precipitation Estimation (QPE). The Korean QPE data, developed by Korea Meteorological Administration (KMA) from a Real-time ADjusted Radar-AWS (Automatic Weather Station) Rainrate (RAD-RAR) system utilizing eleven radars over the Republic of Korea, are at a 1-hour temporal resolution and 1-km by 1-km spatial resolution.This analysis is performed over the area (124.5^oE-130.5^oE, 32.5^oN-39^oN)at multiple time scales, ranging from annual to diurnal, over a commonly available period from March 2014 to October 2017. The results demonstrate the reasonably good ability of Version-05B IMERG products in estimating precipitation over South Korea's complex topography that consists mainly of hills and mountains, as well as large coastal plains. IMERG Final precipitation estimates are in the best overall agreement with the ground data, followed by IMERG Late and IMERG Early Run estimates, in terms of some metrics such as correlation and probability of detection. In terms of relative error, however, IMERG Early and Late Runs outperform the Final Run. It seems that monthly gauge-adjustments cause the substantial overestimation of precipitation rates, particularly over the eastern mountain area. The total precipitation estimate error can be decomposed into three independent components: hit error, missed precipitation, and false precipitation. These errors cancel one other, which results in a much smaller total error. The precipitation diurnal cycles are fairly well reproduced by IMERG over the land and ocean areas, but with underestimated amplitude and delayed peak time relative to Korean QPE.

Detailed validation results are available at https://wallops-prf.gsfc.nasa.gov/KoreanQPE.V05/index.html