مقايسه روش هاي زمين آمار در برآورد توزيع مكاني بارش سالانه در مناطق خشك و نيمه خشك جنوب شرقي ايران

Application of Geostatistical Methods for Estimation of Annual Spatial Rainfall in Arid and Semiarid Regions of South East of Iran

Estimation of rainfall is indispensable in many natural resources and agricultural studies. There are several methods of estimating rainfall among which interpolation is helpful in this regard. In most cases, a method is selected and estimation is then made with unknown precision. In this study, 3 methods were used to estimate annual rainfall. These methods were: Thin Plate Smoothing Splines (TPSS), with and without covariable, Weighted Moving Average (WMA) and Kriging (ordinary, Cokriging, Log kriging). The data in 167 stations from south east of Iran with 22 years of record were used in this study. Elevation was used as covariable and its correlation with rainfall evaluated. Because of size of the area and dispersing stations, the correlation coefficient was low and unacceptable for all stations. Therefore, the study area was divided into subcatchments, based on watershed catchments. In cases where few stations existed in the catchment or regression coefficient (between elevation and rainfall) for two stations was the same, those subcatchments were combined. Finally, 3 units of subcatchment were obtained. Regression coefficient between rainfall and elevation was low in one of the subcatchments (R=0.12) while in the other two subcatchments was higher (0.7 and 0.8). The results show that regression coefficient (between rainfall and elevation) must be grater than 0.6 to improve interpolation accuracy when using elevation as covariable. The results also show that TPSS method with elevation as covariable is the most precise method in estimating annual rainfall (MAE=13-38 mm). Kriging (MAE= 29-50 mm) and WMA methods (26-57 mm) stood in second and third order, respectively.