Comparison of the Global Precipitation Climatology Project (GPCP) Estimates of Precipitation over Snow Covered Surfaces with Gauges

The Global Precipitation Climatology Project (GPCP) produces a global gridded field of monthly mean precipitation at a scale of 2.5˚ x 2.5˚ latitude by longitude. This product merges infrared and microwave estimates with gauges to form the final product of precipitation as described in Huffman et al., 1997. A snapshot of the procedure is to first adjust geostationary infrared estimates, which are used in the region 40˚N-40˚S, to the microwave estimates that are assumed less biased. Over land the blended satellite estimates are then adjusted to a gauge analysis. This gauge analysis is prepared by the Global Precipitation Climatology Centre (GPCC) and consists of about 6,000 gauges worldwide. One major problem in the GPCP analysis is that over snow covered land there is usually not any geostationary IR data (it is out of the useful range for geostationary IR estimates and even if there were data available the technique is not well-suited for cold season precipitation), and it is not possible to retrieve precipitation from the microwave sensors over snow or ice covered surfaces. GPCP has attempted to account for this by using precipitation estimates available from TOVS/HIRS sounding instruments on polar orbiting satellites. Another approach has been to use precipitation estimates based on the Outgoing Longwave Precipitation Index (OPI), which relates anomalies in outgoing longwave radiation to precipitation. The purpose of this work is to evaluate 1) how well GPCP estimates precipitation over snow and ice covered surfaces when compared to an expanded set (~40,000) of gauge data, and to evaluate 2) TOVS/HIRS and OPI retrievals over snow covered surfaces. In order to fully understand and interpret our results we will also compare the 6K stations gauge analysis to the 40K gauge analysis. We are using monthly mean data from a ten-year period from 1988-1997. Preliminary finding indicate: • When comparing the annual average precipitation from 42.5˚N-75˚N for grid boxes containing at least five stations the increased number of gauges in the 40K network exhibits higher values of precipitation. This holds both spatially (from grid box to grid box) and temporally (from month to month) over Europe and North America as well as for monthly area averages. • GPCP estimates (much like the 6K gauge network that influences them) are less than the precipitation values from the 40K gauge network. The differences are generally less than 0.20 mm/day, except areas along the coast where estimates face difficulties. Areas of large precipitation differences appear to be highly correlated with areas with largest difference in number of gauges between the two networks. More is being done to further investigate this. • TOVS/HIRS appear to underestimate precipitation values compared to OPI retrievals and further underestimate compared to GPCP and gauge networks. In some regions, TOVS is more than 1 mm/day under the values from the other estimates. This underestimate is consistent from season-to-season as well as for areal averages.