Recent global performance of the Climate Hazards group Infrared Precipitation (CHIRP) with Stations (CHIRPS)

• Published in: Journal of hydrology (Amsterdam) 2020-12, Vol.591, p.125284, Article 125284
• Main Authors: Shen, Zhehui, Yong, Bin, Gourley, Jonathan J., Qi, Weiqing, Lu, Dekai, Liu, Jiufu, Ren, Liliang, Hong, Yang, Zhang, Jianyun
• Format: Article
• Language: English
• Subjects: CHIRP; CHIRPS; Error analysis; Global evaluation; Satellite precipitation
• ISSN: 0022-1694; 1879-2707
• DOI: 10.1016/j.jhydrol.2020.125284

The Climate Hazards group Infrared Precipitation (CHIRP) and with Stations (CHIRPS) datasets are two new quasi-global (50oS–50oN), high-resolution (0.05° × 0.05°), long-term (1981-present) precipitation estimates based on infrared Cold Cloud Duration (CCD) observations. This study investigates, for the first time, the global performance of CHIRP and CHIRPS against the gauge-based GPCC (Global Precipitation Climatology Centre) data at monthly scale using 36 complete years of data record (1981–2016). Global assessment results indicate that both CHIRP and CHIRPS have negative biases (−5.93% for CHIRP and −2.01% for CHIRPS) before 2000, while this systematic underestimation was effectively removed after 2000. Global analyses also show that the gauge-adjusted CHIRPS estimates generally represent a substantial improvement over CHIRP due to gauge-based bias correction. With respect to regional statistics, temporal analysis and intensity distribution, the gauge-adjusted CHIRPS estimates agree well with GPCC and outperforms CHIRP over most regions, such as the United States, Europe, Africa, Australia and South America. However, southeast China is an exception. Over this region, CHIRPS has a systematic overestimation of 5.55% against GPCC during 2000–2016, especially for spring and summer months, while such positive biases were not found for the pure satellite-derived CHIRP. Possible causes for the discrepancy between these two satellite products over the global and regional scales were further discussed and analyzed. The results reported here will both provide the algorithm developers of CHIRP and CHIRPS with some valuable information and offer the hydrometeorological users a better understanding of their error characteristics and potential limits for various hydrological applications from the global perspective.