Climate Data

            The climate data came from a world climate database compiled by members of the Climatic Research Unit in the School of Environmental Sciences at the University of East Anglia, Norwich, UK.  The data set is comprised of information from the national meteorological agencies, the World Meteorological Organization 1961-1990 Global Standard Normals, and the CRU Global Data sets of Station Time Series.  This data set was used instead of climate station data due to the difficulty in getting climate station data from India.  The UEA database takes data from meteorological weather stations and interpolates those data based on latitude, longitude and elevation by means of the thin-plate spline methodology.  This method creates a database of climate variables for the whole world that is accessible at different spatial resolutions.  The thin-plate spline method gives less weight to weather station data that have large variances or short-term records, whereas it makes the data more robust in areas that have irregular data or few weather stations (New et al., 1998).  The data are only as accurate as the underlying data from the climate station data on which it is based.  In order to check the interpolation methods, internal consistency checks and inter-variable consistency checks were done and the data were compared with other climatologies.  The UEA database is considered better than other climate databases because it has the largest number of station normals, it takes into account nine surface variables, and it uses consistent interpolation methods (New et al., 1998).  The data are set on a grid that is 0.5° latitude by 0.5° longitude, which is about 40 km by 40 km (Cox, 2000, personal communication). 

            The climate database is made from data from individual weather stations.  For the subcontinent of India, a dense geographic cover of meteorological stations submitted information on precipitation, making the precipitation data fairly reliable from the UEA database.  The number of stations that submitted information on temperature and diurnal temperature range is less than the number for precipitation and even fewer stations submitted data on relative humidity (Lister, 2000, personal communication).  The UEA data are more reliable in flat areas than in areas with a complex topography (Cox, 2000, personal communication).  Because the Western Ghats, a mountain range, follow the western coast of Karnataka, the climate data in that region are not as reliable as in the flat regions.

            I used mean temperature and precipitation from the UEA (University of East Anglia) climatology for my study.  Mean temperature (degrees Celsius) is defined as the average of the mean maximum and mean minimum temperatures.  Precipitation (millimeters) measurements were taken directly from the weather stations. 

            There are, however, uncertainties in the data due to the effects caused by wind conditions, the type of gauge used at each station, and the ratio of solid to liquid precipitation.  I did not use relative humidity in my study because the largest errors in vapor pressure are in the Indian subcontinent because of the few stations reporting vapor pressure or relative humidity (New et al., 1998).  Although relative humidity is important in malaria transmission, this data would lead to erroneous conclusions about the predictive abilities of relative humidity in my study.  I did not use wind as a variable in my analysis because it is difficult, due to its variability, to assess the effects of wind on malaria transmission. 

            Mean temperature is one of the most reliably interpolated variables in this dataset, but precipitation is the second least reliably interpolated variable.  Mean temperature was better predicted by the interpolation methods than diurnal temperature range. The main reason for errors in the dataset is from inadequate station networks, and the least reliable areas are ones with poor data coverage or with a lot of topographic relief.  There is a possibility that there were errors in the conversion of variables or errors in station location or through interpolation (New et al., 1998).