Mapping temperature using kriging with external drift: Theory and an example from scotland



Land suitability mapping requires information about climate, topography, and soils in relation to crop performance. Climatic data are available from recording stations which present point samples and so interpolation is needed to generate estimates for non-sampled points. The annual temperature cycle must be interpolated differently for different land suitability projects. A common theme, however, is spatial interpolation and here we investigate the application of kriging with external drift to mapping January mean temperatures in Scotland. The spatial structure of the mean temperature is analysed using variograms computed in different directions. From these we see that January temperature is second-order stationary in the north-south direction. Hence the variogram exists in that direction and is taken to represent the underlying variogram. This variogram is modelled and used in universal kriging to produce point-kriged estimates on a 5-km square grid. These estimates do not adequately show the variation in temperature between stations and so the correlation with elevation was exploited in universal kriging with elevation as external drift. This method gives a kriged estimate for temperature that reproduces the correlation with elevation at the climate stations.