Results from the GPCP Version 3.2 Products

The Global Precipitation Climatology Project (GPCP) provides long-term precipitation products that emphasize homogeneity, following Climate Data Record (CDR) principles. Following on the operational GPCP Version 2.3 Monthly and Version 1.3 Daily, the new-generation Version 3.2 provides key improvements such as: finer spatial resolution of 0.5°x0.5° (vs. 2.5°x2.5°); wider geosynchronous infrared coverage (58°N-S, vs. 40°N-S); use of the PERSIANN-CDR algorithm (vs. a basic calibrated brightness temperature–precipitation rate scheme); upgraded retrievals from selected passive microwave sensors (GPROF2010v2 algorithm vs. a combination of older algorithms) that calibrate the IR input; revised intercalibrations of TOVS and AIRS data (used at high latitudes); climatologies based on CloudSat, TRMM, and GPM to provide overall calibration by modern satellite estimates; the latest Global Precipitation Climatology Centre (GPCC) precipitation gauge analyses over land areas; regional modifications to the gauge undercatch correction (Fuchs et al. 2001 over northern Eurasia, vs. Legates-Willmott everywhere); and IMERG half-hourly data input to the Daily V3.2 product vs. a locally adjusted GPI-like scheme.

We will show sample analyses that demonstrate aspects of the Version 3.2 precipitation record, such as the global climatology, the time series for global land and ocean total precipitation and snowfall, and the time series of tropical land and ocean daily precipitation rate histograms. Version 3.2 shows improvements in both the Monthly and Daily products compared to the operational Version 2.3 and 1.3 products, respectively. In particular, a suspected artifact in in the Southern Ocean around 60°S that Version 2.3 displays is mitigated. The Daily histograms over ocean for Version 1.3 at the start of 2009 display a shift, which is ascribed to the transition from SSMI to SSMIS as input to GPROF2010v2. Version 3.2 eliminates this shift, but introduces a smaller jump in the daily histograms in June 2014, apparently due to the transition from the combined radar-radiometer product being driven by TRMM instrument data to it being driven by GPM Core Observatory instrument data. In both cases, the mean precipitation values are well-behaved, but the distribution of values (including “extreme” events) is systematically different, with Version 3.2 Daily having smaller differences than the Version 1.3 Daily. In addition, the shift in sampling from 8 per day using a simplified algorithm in V1.3 to 48 per day based on IMERG in V3.2 provides considerably more flexibility to correctly represent extreme daily values. Taken together, these provide more confidence in the V3.2 climatology of extreme values.

Finally, we discuss prospects for the future satellites/sensors and community datasets necessary to continue computation of a consistent CDR product, as well as new analyses of additional input data that might provide potential improvements in the historical record.