A 35-Year Daily Precipitation Analysis for Hydroclimate Applications
Pingping Xie1) and Hai-Tien Lee2)
1) NOAA Climate Prediction Center 2) ESSIC, University of Maryland at College Park
Many precipitation data sets have been produced in the past two decades based on gauge measurements, satellite observations and their combinations. Their applications in hydroclimate applications, however, are restricted by a number of shortcomings in long-term homogeneity, quantitative accuracy, and time/space resolution. The objective of this work is to construct a homogenous analysis of daily precipitation on a 0.25olat/lon grid over the global land for a 35-year period from 1979 to the present through combining information from satellite observed outgoing longwave radiation (OLR), as well as daily and monthly gauge analyses.
First, CPC unified daily gauge analysis is calibrated against the GPCC monthly gauge data set to correct the under-estimates in the daily gauge analysis caused by the negatively biased daily station reports. The resulting adjusted daily gauge analysis presents improved quantitative accuracy over regions with reasonable gauge coverage.
A new technique is then developed to derive precipitation estimates from the OLR data observed by HIRS aboard NOAA polar orbiting satellites. To this end, probability density function (PDF) tables are created for the HIRS OLR and matched against those for the collocated CMORPH satellite precipitation estimates using data for a 15-year period from 1998 to the present. An observed HIRS OLR value is converted to precipitation through the matched OLR and precipitation PDF tables. The PDF tables are constructed for each pentad period and for each land grid box of 1olat/lon to account for the seasonal variations and regional dependence of the OLR-precipitation relationship. The technique is applied to generate precipitation estimates on a 0.25olat/lon grid over the globe for the entire HIRS OLR data period from 1979.
Finally, the adjusted daily gauge analysis and the HIRS OLR-based precipitation estimates are blended through the optimal interpolation (OI) technique. The OLR based precipitation estimates are used as the first guess, while the gauge analysis is utilized as the observation to update the first guess. The gauge-OLR blended analysis of daily precipitation analysis is dominated by the gauge analysis over regions of dense station networks, while the OLR-based estimates play an important role over regions of sparse surface observations (e.g. equatorial Africa and Amazon).
A suite of comprehensive procedures are implemented to examine the quantitative accuracy and homogeneity of the 35-year daily precipitation data set. Detailed results will be reported at the AMS meeting.