Comparison of Monthly GPCPV3 Precipitation Estimates with PACRAIN Atoll Gauge Observations

The Global Precipitation Climatology Project (GPCP) V3 Monthly and Daily satellite-based precipitation products follow on to the highly successful GPCPV2 products. GPCPV3 takes advantage of, and builds upon, the availability of more modern input data sources to provide high-quality estimates at higher spatial resolution. The goal is to produce a consistent, homogenous record of precipitation (e.g., a Climate Data Record) appropriate for long-term precipitation studies.

While validation of GPCPV3 precipitation estimates over land is relatively straightforward due to the plethora of in-situ observations such as rain gauges and radars, validation over open ocean is much more challenging, but equally important as ocean/land precipitation patterns are inextricably linked. In this study, we compare and assess the quality of the GPCPV3 monthly product against the Pacific Rainfall Database (PACRAIN) monthly atoll gauge observations for the period 1983 – 2022. Although the PACRAIN database contains observations from islands with varying geography, we have limited the comparison to low-lying atolls to minimize land effects for a more representative comparison with open ocean. Due to the nature of gauge observations, strict quality-control measures are implemented to ensure the integrity of the PACRAIN observations. For example, we require that each individual atoll station report for each day of the month to be included in the monthly data set, and we exclude reported accumulations longer than 24 hours (“weekend effect”) to avoid potential issues such as evaporation. Following quality control, 72 atoll stations from the Western and Central tropical Pacific are included in this analysis. It is important to note that no PACRAIN atoll stations are included in the GPCC gauge analyses that are incorporated into GPCPV3, so GPCPV3 is independent of the PACRAIN data. Furthermore, we apply a blended monthly Legates-Fuchs wind-loss correction to the GPCC gauge analysis prior to merging with the satellite-only estimates in an effort to correct for gauge undercatch. However, no wind-loss correction is currently applied to the PACRAIN atoll gauges.

Preliminary analysis shows that GPCPV3 estimates are biased low by -10.6% over the period, compared with -16.6% for GPCPV2.3. Seasonally, GPCPV3 biases range from -9.4% in MAM to -12.0% in JJA. Seasonal and annual scatterplots comparing GPCPV3 estimates and PACRAIN atoll observations will be presented as well as statistics such as bias, percent bias, correlation, and RSME. GPCPV3 performance relative to GPCPV2.3 will also be examined.

The results presented here will provide the framework for continued assessment of GPCPV3 as well as future comparison of the GPCPV3 daily product with the PACRAIN atoll observations.