Tropical Land-Surface Precipitation: Gridded Monthly and Annual Time Series (1950 - 1999)

(Version 1.01)

produced and documented by

Michelle Johnson, Kenji Matsuura, Cort Willmott, and Petra Zimmermann
(with support from NASA's Seasonal to Interannual ESIP)

For additional information concerning this archive, please contact us at:

Center for Climatic Research
Department of Geography
University of Delaware
Newark, DE 19716
(302) 831-2294


Archive (Version 1.01) released May 1, 2003


Five sources of station data were used to produce these gridded monthly total precipitation (P) time series. The sources were: the Global Historical Climatology Network (GHCN version 2) (Peterson and Vose 1997); the National Center for Atmospheric Research (NCAR) daily India data; Sharon Nicholson's archive of African precipitation data (2001); Webber and Webber and Willmott's (1998) South American monthly precipitation station records; and the station climatologies from Legates and Willmott's (1990) archive.

Global Historical Climatology Network: Data for the tropical land surface were extracted from the GHCN version 2 archive, which contains monthly precipitation observations for about 20,600 meteorological stations. The time periods of observation vary from 1840 to the present. For the period 1950-1999, the number of tropical land-surface stations with observations available for a particular month varies from about 600 to 10,700.

Sharon Nicholson: Nicholson's station archive contains monthly P data from 1,338 meteorological stations in Africa. Observations are available for varying numbers of stations from 1840 through 1996. Monthly P records within this archive are primarily from the African mainland and near-shore islands. For the period 1950-1996, the number of stations with observations available for a particular month varies from 791 to 1,263. The year with the most dense station coverage is 1958, while 1996 has the sparsest coverage.

Webber and Willmott: Webber and Willmott's (1998) South American Monthly Precipitation Station Records dataset contains monthly and annual P data for 5,315 stations in South America. Its temporal coverage ranges from 1846 to 1993. For the period 1950-1993, the number of tropical land-surface stations with observations available for a particular month varies from about 240 (in 1993) to 3,450 (in 1972).

NCAR: Whereas the three data sources described above are comprised of monthly observations, the NCAR dataset is made up of daily precipitation totals. It contains 4,041 station records distributed spatially throughout India. The records span varying periods, although no observation is available prior to 1901 or after 1970. Inconsistencies among the daily observations were examined, and unrealistically high daily P values were dropped from the record. Each daily P record also was examined subjectively, with regard to whether it was plausible; if not, a missing record code was inserted in its place. The daily P data then were summed to generate monthly totals, and the units were converted to mm. Monthly P for any month with five or more missing daily values was recorded as missing. Each derived monthly record then was compared graphically with the record of its nearest station neighbor (or with the monthly climatology at the closest grid point from Willmott and Matsuura, 2001) to assess its credibility. For the period 1950-1970, the number of stations with viable observations available for a particular month varies from 3,210 (in 1957) to 2,395 (in 1965).

The station records from these four data sets were merged to create a composite station-record series for the period 1950 through 1999. During this process, station records that had the same geographical coordinates were interleaved or blended to create a single, station time series for that location. If there were two or more station observations for a given month, these observations were averaged to obtain P for that month. When there was only one station observation for a month, it was taken as P for that month. This was done to make use of all available data. Observations from stations which had different geographical coordinates were assumed to belong to different station records, although sometimes parts of nearby station records were extremely similar. The total number of station records available, after merging, varied markedly from about 400 (in 1999) to 13,700 (in 1968). Our merged station archive provides relatively good spatial coverage from 1950 into the early 1990s. After the early 1990s, station-record coverage of the tropical land surface begins to degrade.


Our traditional interpolation algorithm is based on the spherical version of Shepard's distance-weighting method (Shepard, 1968; Willmott et al., 1985). Station observations were spatially interpolated to a 0.5 degree by 0.5 degree of latitude/longitude grid, where the grid nodes were centered on 0.25 degree. The number of nearby stations that influenced a grid-node estimate was increased to an average of 20, from an average of 7 in earlier applications. This resulted in smaller cross-validation errors (see below) and visually more realistic precipitation fields. A more robust neighbor finding algorithm, based on spherical distance, also was used.

Using a climatology available from a relatively dense network of stations also can increase the accuracy of spatially interpolated time series of monthly climate variables. Employing Climatologically Aided Interpolation (CAI) (Willmott and Robeson, 1995), a monthly P at each time-series station can be differenced from a climatologically averaged P for that month which is available at or can be interpolated to the time-series station location. Traditional interpolation then can be performed on the station differences to obtain a gridded difference field. Finally, the gridded difference field can be added to interpolated estimates of the climatology at the same set of grid points.

CAI was used to estimate our monthly total precipitation fields. For the background climatology, two station climatologies were merged. The first was calculated at those of our precipitation time-series stations which had at least five years of observations for each month (within the period 1960-1990). The second was the monthly station P climatology of Legates and Willmott (1990). Only those Legates and Willmott stations which were not collocated with our own 1960-1990 station climatology were included in the background climatology for CAI.


To indicate (roughly) the spatial interpolation errors, station-by-station cross validation was employed (Willmott and Matsuura, 1995). One station is removed from the station network at a time, and then precipitation is interpolated to the removed station location from the surrounding nearby stations. The difference between the interpolated station value and the observed value is a local estimate of interpolation error. After each station cross validation is made, the removed station is put back into the network. To reduce network biases on cross-validation results, absolute values of the errors at the stations were interpolated to the same spatial resolution as the precipitation field.


Monthly and annual total precipitation (for the years 1950 - 1999) interpolated to a 0.5 by 0.5 degree grid. The format of each record is
Field Columns Variable Fortran Format
1 1 - 8 Longitude (decimal degrees)
2 9 - 16 Latitude (decimal degrees)
3-14 17 - 112 Monthly Total Precipitation (mm, Jan - Dec)
15 113 - 120 Annual Total Precipitation (mm)

Cross-validation errors associated with monthly and annual total precipitation interpolated to a 0.5 by 0.5 degree grid. The format of each record is
Field Columns Variable Fortran Format
1 1 - 8 Longitude (decimal degrees)
2 9 - 16 Latitude (decimal degrees)
3-14 17 - 112 Cross-Validation Errors for Monthly Total Precipitation (mm, Jan - Dec)
15 113 - 120 Cross-Validation Errors for Annual Total Precipitation (mm)


Legates, D. R. and C. J. Willmott (1990). Mean Seasonal and Spatial Variability in Gauge-Corrected, Global Precipitation. International Journal of Climatology, 10, 111-127.

Peterson, T.C. and R.S. Vose (1997). An Overview of the Global Historical Climatology Network Temperature Database. Bulletin of the American Meteorological Society, 78(12), 2837-2849.

Shepard, D. (1968). A Two-dimensional Interpolation Function for Irregularly-spaced Data. Proceedings, 1968 ACM National Conference, 517-523.

Willmott, C. J., C. M. Rowe and W. D. Philpot (1985). Small-scale Climate Maps: A Sensitivity Analysis of Some Common Assumptions Associated with Grid-point Interpolation and Contouring. American Cartographer, 12, 5-16.

Willmott, C. J. and K. Matsuura (1995). Smart Interpolation of Annually Averaged Air Temperature in the United States. Journal of Applied Meteorology, 34, 2577-2586.

Willmott, C. J. and S. M. Robeson (1995). Climatologically Aided Interpolation (CAI) of Terrestrial Air Temperature. International Journal of Climatology, 15, 221-229.