Independent precipitation estimates may be combined to produce data sets that emphasize the best estimate in each region. Such estimates are being produced on the monthly scale in Global Precipitation Climatology Project (GPCP; on a 2.5x2.5-deg grid) and Tropical Rainfall Measurement Mission (TRMM; on a 1x1-deg grid). However, feedback from data set users to GPCP revealed significant interest in a finer-scale, long-term product.
An experimental global, daily, 1x1-degree GPCP data set has been developed that combines infrared (IR), microwave, and TOVS estimates, with indirect contribution by gauge analyses. The algorithm for this new data set is related to the monthly schemes cited, with calibration of the IR by the microwave, estimation of root-mean-square random error for each independent estimate, and combination using inverse-error weighting. The major differences are:
1. The IR calibration is done with threshold matching. As before, although more rigorously here, the coefficients are computed only from data that are coincident in space and time, accumulated for a month.
2. The daily estimates are scaled to sum to the monthly GPCP satellite-gauge combination. This ensures consistency between the two data sets and introduces gauge information into the daily estimates. A global daily gauge analysis is not feasible at this time.
The data set covers 1 Jan. 1997 through the present (about 2 months behind real time). A loop of the time series shows impressive continuity of features from day to day and from one data region to another. From the hydrologic perspective, such data are ideally suited to forming composites for user-defined periods/regions. For example, the time history of precipitation falling on a specific drainage basin can be computed. The data also allow power spectrum and Empirical Orthogonal Function decomposition over a wide range of scales. The estimates can be used at the finest scales, for example to drive land surface hydrology models. Even though the data set likely provides the best global data to date at these scales, users should be aware of the relatively large error bars for individual grid values.
The status of the data set will be discussed, including validation efforts and the feasibility of moving to 3-hourly 1x1-degree resolution