Monthly Rainfall and Temperature Surfaces for Africa  1951 - 1995 (CD-ROM) INFORMATION

Production of Climate Surfaces

Climate Surface Information
Data Format; Data Units; File Names; Data Storage; Data Size

Weather Station Locations Information
File Format; File Names; Data Storage; Data Size

Credits and Acknowledgements

CONSTRUCTION OF 3 MINUTE LATITUDE / LONGITUDE MONTHLY CLIMATE SURFACES OVER AFRICA FOR THE PERIOD 1951-1995

Mark New and Mike Hulme
Climatic Research Unit University of East Anglia Norwich NR4 7TJ, UK
1 August 1997

Introduction

This report describes the construction 3 minute latitude / longitude grids of monthly precipitation, mean temperature and diurnal temperature range for continental Africa for the period 1951-1995, as required under contract to Dr. Dave Le Seuer of the Medical Research Council, Durban, South Africa.

Methodology

A major obstacle in the construction of monthly climate surfaces is the paucity of long time-series of station data. This problem is particularly acute developing countries. More readily available are climatological normals, the mean monthly climate at a station for a specified time period (usually 30 years). For example, the Climatic Research Unit (CRU) archives contain more than four times the number of minimum and maximum temperature normals than time series. Our approach has been to separate the time and space components during the construction of the climate surfaces. Using station normals for the period 1961-1990 we constructed mean monthly climate surfaces of precipitation, mean temperature and diurnal temperature range. By exploiting the more extensive network of stations with climate normals, as much information as possible about spatial variability in mean climate was obtained. Thereafter, monthly surfaces of anomalies relative to 1961-1990 were constructed from the more limited network of station time series. Because climatological anomalies are a function of changes in the large scale circulation patterns, they tend to be regional in extent and can be characterised by fewer observations. Finally, the monthly anomaly surface is added to the mean climatology to produce the monthly climate surface.

Data

Climatological Data CRU holds global datasets of station climatological normals and monthly time series of precipitation and mean, maximum and minimum temperature. Data for Africa were extracted from the master datasets. The resultant African subsets of normals comprised 2307 precipitation stations (Figure 1), 1485 mean temperature stations (Figure 2a) and 1431 diurnal temperature range stations (Figure 2b). The latter variable was used instead of maximum and minimum temperature as it reduces the number of variables to be interpolated. The datasets of station time series are considerably smaller than the network of station normals and also show large variation over time (Figures 3-5). The late 1980s and 1990s are particularly poorly sampled; in some cases this is due to political and/or economic factors, but in many cases it is simply a reflection of the time it takes for data to "filter" through to CRU via various official and unofficial contacts. Figure 1. Station network of precipitation normals.

Figure 1. Station network of precipitation.

Figure 2(a). Station network of mean temperature normals.

Figure 2(b). Station network of diurnal temperature range normals.

Elevation Data The procedure used to interpolate the mean climate fields requires elevation as a third predictor variable, along with latitude and longitude, in the form of a digital elevation model (DEM) when the final surfaces are generated. We used the 3 minute resolution African DEM of Hutchinson (1996) for this purpose.

Interpolation

The interpolation of irregular gauge data onto a uniform grid has been the focus of much research and a large number of methods have been proposed, ranging from simple Thiessen and distance weighting methods (Shepard, 1968; Willmott et al., 1985), to geostatistical methods such as kriging (Phillips et al., 1992) or splines (Hutchinson, 1995) and locally-varying regression techniques (e.g. PRISM; Daly et al., 1994). For many climate applications it is important that elevation is included as a covariate or independent variable because the climate variable is dependent on elevation in some manner (Willmott and Matsuura, 1995; Briggs and Cogley, 1996).

Mean climate surfaces We used thin-plate splines to interpolate the mean climate surfaces as a function of latitude, longitude and elevation. The original thin-plate spline fitting technique was described by Wahba (1979), while Hutchinson (1995) provides a theoretical description of their application to surface climate variables such as precipitation. The technique is robust in areas with sparse or irregularly spaced data points. Thin-plate splines are defined by minimising the roughness of the interpolated surface, subject to the data having a predefined residual. This is usually accomplished by determining the amount of data smoothing that is required to minimise the generalised cross validation (GCV). This is calculated by removing each data point in turn and summing, with appropriate weighting, the square of the difference between the omitted point and a surface fitted using all the other points. The interpolation of a large number of data points becomes computationally demanding. In addition, fitting the same spline function to areas with markedly different station densities (e.g. Equatorial Africa vs. North Africa) can result in too much smoothing in data rich areas and too little smoothing in data poor areas. For each variable, therefore, the African continent was divided into a number of geographic tiles over which separate spline functions were fitted (Figure 1-2). The size of the tiles varied primarily according to station density, but also as a function of spatial complexity of the climate variable. Individual tiles were forced to overlap by at least 10 degrees so as to minimise edge effects. The number of stations in a tile varied between about 200 and 500. Output diagnostics from the spline programs were used to identify erroneous data. Most common errors were typographic and locational. Where possible, the errors were corrected; stations which could not be corrected were removed from the dataset. The final fitted spline functions for each tile were applied to the portion of the 3 minute DEM falling within the tile to derive the climate grids for each variable. The tiles were then merged to produce a global field. Finally, the precipitation grid was adjusted so that negative values were set to zero.

Monthly anomaly surfaces Monthly anomalies of precipitation mean temperature were interpolated using thin plate splines as a function of latitude and longitude only because the anomalies are mostly independent of elevation. For precipitation, the larger number of stations required the interpolation to done over two tiles, split at the equator, which were subsequently merged for each year-month. The network of temperature stations was small enough to allow the surfaces to be interpolated as a single tile. The network of maximum and minimum temperature stations was too sparse to permit the interpolation of diurnal temperature range over the entire continent (see Figure 5), as data from only a few African countries are held in the CRU dataset. Using a surface fitting function such as splines was inappropriate because of the dangers of extrapolation to unrealistic values far away from control points. Diurnal temperature range was therefore interpolated using an angular distance weighting procedure similar to that of Willmott (1985), but modified so that grid points further than 500km from any station were assigned a value of zero.

Monthly climate surfaces In the final stage of the construction the monthly anomaly surfaces were combined with their respective mean climate surfaces to produce surface of monthly precipitation, mean temperature and diurnal temperature range. Maximum and minimum temperature surfaces are not included explicitly in the dataset in order to reduce data storage requirements However, these variables can be derived easily from the mean temperature (T) and diurnal temperature range (Dtr) surfaces as follows: This approach for deriving maximum and minimum temperature has the added advantage that the monthly surfaces of mean, maximum and minimum temperature will be spatially consistent. Moreover, in regions where there are no stations with diurnal temperature range (and the monthly diurnal temperature range surface is therefore equal to the mean surface), maximum and minimum temperature will still vary as a function of mean temperature.

Accuracy of Surfaces

Although the surfaces are fitted to a 3 minute (approximately 3km) digital elevation model, it must be stressed that they do not capture the effects of topography on climate at this resolution. Inter-station distances have a range of ten to several hundred km for the mean climate surfaces, and are appreciably larger for the anomaly surfaces. The information contained in the fitted surfaces can only be as good as that provided by input data. The resultant fitted surfaces probably only capture spatio-temporal variability at the 25-50km scale. Variability in the surfaces below this scale is simply a function of the larger scale relationships with latitude, longitude and elevation being applied at a finer resolution.

Conclusions

This document provides a brief description of the methods and data used in the construction of 3 minute latitude and longitude monthly climate surfaces over Africa for the period 1951-1996. Surface climate variables in the dataset include precipitation, mean temperature and diurnal temperature range. Maximum and minimum temperature can be derived from the mean temperature and diurnal temperature range surfaces.

References

Briggs, P. R. and J. G. Cogley, 1996: Topographic bias in mesoscale precipitation networks. J. Climate, 9, 205-218.

Daly, C., R. P. Neilson and D. L. Phillips, 1994: A statistical-topographic model for mapping climatological precipitation over mountainous terrain. J. Appl. Meteor., 33, 140-158.

Hutchinson, M. F., 1995: Interpolating mean rainfall using thin plate smoothing splines. Int. J. Geog. Inf. Sys., 9, 385-403.

Hutchinson, M. F., H. A. Nix, J. P. McMahon and K. D. Ord, cited 1996: The development of a topographic and climate database for Africa. [Available on-line from http://www.ncgia.ucsb.edu/conf/SANTA_FE_CD-ROM/sf_papers/hutchinson_michael_africa/africa.html].

Phillips, D. L., J. Dolph and D. Marks, 1992: A comparison of geostatistical procedures for spatial analysis of precipitation in mountainous terrain. Agric. Forest. Meteor., 58, 119-141.

Shepard, D., 1968: A two-dimensional interpolation function for irregularly spaced data. Twenty-third ACM National Conference, Brabdon Syst. Press, 517-524.

Wahba, G., 1979: How to smooth curves and surfaces with splines and cross-validation. 24th Conference on the Design of Experiments, U.S. Army Research Office, 167-192.

Willmott, C. J. and K. Matsuura, 1995: Smart interpolation of annually averaged air temperature in the United States. Journal of Aplied Meteorology, 34, 2577-2586.

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. Am. Cartogr., 12, 5-16.

Climate data is provided for the African continent as monthly means covering the years 1951 to 1995. CD 1 contains precipitation, CD 2 contains diurnal temperature range and average temperature. Monthly climate surfaces are compressed in self-extracting zip files, and grouped in five-year periods.

Data Format:

Data surfaces are in Idrisi format (*.img). For details run the Idrisi Export Program (d:\export\idexport.exe) and see "File Structures" under Help > Contents.

The Idrisi byte binary format is a stream of bytes; to decode it, it must be treated like a BSQ (Band-sequential) file, with a single band. The integer binary file is the same, but each pixel is 2 bytes, and in the real binary file each pixel is a 4-byte single precision real number. Any GIS software that can import satellite images should be able to import the Idrisi images.

The data is stored in binary code, but can be converted to ASCII code using the CONVERT module in the Idrisi Export Programme (see Section 5.1).

Header information is given in a separate text file - the Idrisi image documentation file (*.doc). Image format (below) is reflected in every image documentation file.

Data type: integer
File type: binary
Columns: 1380
Rows: 1450
Reference system: latlong
Reference units: deg
Unit distance: 1
Min. X: -17.5
Max. X: 51.5
Min. Y: -35
Max. Y: 37.5
Resolution: 0.05 deg

Background value: -9999

Data Units:

Precipitation: millimeters = liters per square meter
Average Temperature: centi-degrees Celsius (tenths of a degree)
Temperature Range: centi-degrees Celsius (tenths of a degree)
        Centi-degrees / 10 = Degrees Celsius with one decimal

To calculate:
Maximum temperature = average temperature + (temperature range / 2)
Minimum temperature = average temperature - (temperature range / 2)

File Names:

File names are as follows: ccccyymm, where c = climate surface, y = year, m = month.

The following abbreviations are used:

Climate Surface:
pptn = monthly average precipitation
tavg = monthly average temperature
tran = monthly diurnal temperature range

Year: 50 = 1950 to 95 = 1995

Month: 01 = January, 02 = February to 12 = December

Example: pptn8603 = Precipitation in March 1986

Data Storage:

The climate surfaces are compressed in five-year periods.
Archive names are as follows: ccccssee, where c = climate surface, s = start year of period, e = end year of period.

Climate Surface:
pptn = monthly average precipitation
tavg = monthly average temperature
tran = monthly diurnal temperature range Start/End

Year: 50 = 1950 to 95 = 1995

Example: tavg7680.exe = average temperature surfaces for 1976 to 1980

Data Size:

Compressed: Full set of 27 self-extracting zip archives = about 750 Mega Bytes (MB).
Uncompressed: Each surface = about 4 MB. Each self-extracting zip archive = about 230 MB.
Full set of 1620 climate data surfaces = about 6.4 Giga Bytes (GB).

Weather station locations are provided to show the spacial spread of weather station data used in the production of individual climate surfaces. CD 1 contains weather stations used for the precipitation surfaces, CD 2 contains those used for temperature range and average temperature data. The files are compressed in self-extracting zip files, and grouped in five-year periods.

File Format:

The weather station locations are given in space-delimited text files in the following format:

Lat Long Elev. CRU Sta.ID Name/Country 11.75 -2.93 270.00 6551600 BOROMO--------BURKINA-FA 32.90 13.20 80.00 6201001 TRIPOLI-------LIBYA-----

Latitude (Lat) and Longitude (Long) are given in decimal degrees.
Elevation (Elev.) is given in metres (m).
CRU Sta.ID are the Climatic Research Unit's Station IDs

File Names:

Weather station file names have the same names as the climate surfaces they belong to (see Section 2.3), but with a .txt extension.

Example: tavg0490.txt contains weather station locations used in the production of the average temperature surface for April 1990.

Data Storage:

Weather station files are compressed in the same five-year periods as the climate surfaces. Archive names are as follows: wwwwssee, where w = weather stations, s = start year of period, e = end year of period.

Weather Stations: wppt = weather stations for precipitation surfaces wtav = weather stations for average temperature surfaces wtra = weather stations for temperature range surfaces Start/End Year of period: 50 = 1950 to 95 = 1995

Example: wppt7680.exe = weather stations for precipitation surfaces 1976 to 1980

Data Size:

Uncompressed: All weather station data = about 85.3 MB.

This climate data was developed by the Climatic Research Unit (CRU). The work was commissioned by the Mapping Malaria Risk in Africa (MARA/ARMA) initiative and funded by the International Development Research Centre (IDRC), Canada. The South African Medical Research Council (MRC) funded the reproduction of the data CD-ROMs, and the MARA/ARMA initiative organized their dissemination.

When referencing the data in publications, the following reference should be used: 'CRU/SAMRC, 1998. African Monthly Climate Data CD-ROM. Climatic Research Unit and the Mapping Malaria Risk in Africa Initiative. Available from the Climatic Research Unit - University of East Anglia, Norwich, NR4 7TJ, United Kingdom.

Copyright belongs to the Climatic Research Unit - University of East Anglia, Norwich, NR4 7TJ, United Kingdom. The data on this CD-ROM may not be reproduced for commercial purposes and may not be passed on to third parties. When the data is used or publicly displayed, the source should be acknowledged.

CD Navigator was created by AndyC Creations.

Back to Climate CD home page.