Climatic Variables Mapping Using Geostatistics in Semi Arid Region: Case Study of Tunisia

Climatic Variables Mapping Using Geostatistics in Semi Arid Region: Case Study of Tunisia

In this paper, two geostatistical algorithms are compared: universal kriging (UK) and kriging with external drift (KED). These algorithms are used to interpolate interannual rainfall observations and interannual potential evapotranspiration (Et0) measured at 22 weather stations in Tunisia covering an area of 164 150 km2. Experimental sample semi variograms are constructed and fitted to estimate levels of interannual rainfall and interannual Et0. Contour maps and corresponding uncertainty maps are then created. Maps of rainfall and Et0 obtained using UK showed a slightly smoother pattern than those obtained using KED. The largest kriging variance values are located in the west and south parts of the study area for both methods with a larger maximum variance for Et0 than for rainfall. Cross validation showed that the RMSE obtained for KED gave better results for rainfall interpolation while the UK gave better result for Et0 interpolation.