Current version: v0207 (Jan.1979 to Dec.2002), available from the web site and ftp site cited above.

Phil Arkin provides a great technical overview of the CMAP data set below, in the Expert User Guidance section of this page.

The CPC Merged Analysis of Precipitation ("CMAP") is a technique which produces pentad and monthly analyses of global precipitation in which observations from raingauges are merged with precipitation estimates from several satellite-based algorithms (infrared and microwave). (Provided by CDC.)

Taken from Pingping XIe's announcement of CMAP v0207 from NCEP ftp site: The CMAP data set may contain an artificial downward trend for the period after 1996. Care must be taken when using the data set for accurate quantitative applications.

Questions concerning the method used to merge the preciptation estimates should be directed to Pingping Xie.

A number of observations exist from which the spatial and temporal distribution of precipitation can be inferred. For the study of seasonal to decadal variability in precipitation, or the use of such data in model validation, it is helpful to construct a time series of global gridded fields (analyses) of time-averaged precipitation for the globe. Given the shortcomings of the various available observations, this requires a sensible method for combining inputs of widely varying character and quality. CMAP is one such effort, comprising a family of products; another, fairly similar but not identical, product suite is that generated by the Global Precipitation Climatology Project and generally referred to as GPCP.

The CMAP products include time series of monthly and pentad (5-day) mean precipitation for 2.5 degree grid areas for the globe for the period from January 1979 to (presently) March 2001. Since observations are not available for some areas, particularly the polar caps, a supplementary version incorporating model forecasts of precipitation from the NCEP/NCAR reanalysis is also available. The inputs used and some of their characteristics are:

1. An analyzed field derived from approximately gauge observations, using inverse distance weighting and directional shadowing (Xie and Arkin, 1996a).

2. Estimates derived from a variety of satellite observations, including geostationary and polar orbiting infrared and passive microwave.

3. Except in the "observation only" data sets, forecasts of precipitation from the NCEP/NCAR reanalysis.

The gauge-based analysis is best when a number of gauges are available in each grid areal; this is the case only in limited regions. The estimates based on infrared observations are essentially based on variations in cold cloudiness, and are best suited for identifying changes in deep convective precipitation. They have greatest accuracy in the tropics and warm season mid-latitudes, and are least accurate in colder conditions. Two types of passive microwave-based estimates are used: scattering and emission. Scattering estimates are best suited to detecting deep convection, while emission estimates are sensitive to liquid rain over the oceans. All microwave-based estimates are significantly limited by sparse sampling. A review, moderately out of date but still useful, of such techniques is given in Arkin and Ardanuy (1989).

The analysis technique used comprises two phases. In the first, all estimates except the gauge-based analysis are combined using a maximum liklihood estimate with weights derived by comparison to the gauge analysis (over land) or in a simplified form based on earlier investigations (over oceans). This reduces the overall random error, but leaves the systematic error. We then use the gauge analysis over land, when sufficient gauge observations are available, or atoll-based gauge observations over the tropical oceans, to determine the absolute value of the analysis, while preserving the gradients derived earlier. The algorithm is described in Xie and Arkin (1996b) and preliminary results in Xie and Arkin (1997).

Crucial points to be kept in mind by potential users:

1. The quality of the analysis is highly dependent on the quality and amount of input data used. Areas with sparse or no gauges, or areas where the satellite estimates have large errors or poor sampling, are likely to exhibit larger errors.

2. All of the satellite estimates used are significantly flawed. The infrared-based estimates depend upon an empirical relationship between cloudiness and precipitation that is poorly known, and which surely varies in space and time. The microwave-based observations are derived from sparse sampling, and also depend, although to a lesser degree, on empirical calibrations.

3. In general, the data set is best suited to identifying and quantifying in a relative manner the spatial and temporal variability of precipitation in the tropics. Variability in mid-latitudes is characterized less well, but still usefully. Great care is required in using the data at latitudes poleward of 60 degrees.

4. The absolute values given are generally less worthy of confidence than the variability. Global averages appear to be accurate to within 5-10%, but individual grid area values probably have much greater uncertainties.

5. The data set is not likely to be useful for analysis of trends in global or large-scale precipitation. The merging of varying data sources relies to some extent on a presumption of stationary, probably incorrect, that makes residual trends untrustworthy.

6. The differences between CMAP and GPCP products, except where clearly associated with algorithm or data differences, are likely to be our best estimate of uncertainty in the analysis.

Phil Arkin
October 2001

Arkin, P.A. and P. Ardanuy, 1989: Estimating climatic-scale precipitation from space: A review. J. Climate, 2, 1229-1238.

Xie, P. and P. A. Arkin, 1996a: Gauge-based monthly analysis of global land precipitation from 1971 to 1994. J. Geophys. Res., 101, 19023-19034.

Xie, P. and P. A. Arkin, 1996b: Analyses of global monthly precipitation using gauge observations, satellite estimates and numerical model predictions. J. Climate, 9, 840-858.

Xie, P. and P. A. Arkin, 1997: Global precipitation: A 17-year monthly analysis based on gauge observations, satellite estimates and numerical model outputs. Bull. Amer. Met. Soc., 78, 2539-2558.

Click on the links below to view data coverage maps for a particular time period. Percentage of non-missing data per time period is plotted. Coverage is consistent throughout the period of record.

(1/1979-3/2001)

Updated: 10/15/03
Maintained by asphilli@ucar.edu