Estimation of mean monthly, annual rainfall in Andhra Pradesh using geostatistical analysis

Rainfall is a hydrological phenomenon that varies in magnitude in space as well as in time and requires suitable tools to predict mean values in space and time. Estimation of rainfall data is necessary in many natural resources and water resource studies. There are several methods to estimate rainfall among which interpolation is very useful approach. In this research, geostatistical interpolation methods are used to estimate monthly (June-December), seasonal (South-West and North-East Monsoon seasons) and annual rainfalls in Andhra Pradesh, India. Monthly rainfall data from a network of 23 meteorological stations for the period 19702003 has been in the study. The main objectives of this work are: (1) to analyze and model the spatial variability of rainfall, (2) to interpolate kriging maps for different months as well as seasons, (3) to analyze and model the structural cross correlation of rainfall with elevation for different seasons, (4) to investigate whether co-kriging would improve the accuracy of rainfall estimates by including elevation as a secondary variable, and (5) to compare prediction errors and prediction variances with those of kriging and cokriging methods for different seasons. Rainfall surfaces have been predicted using ordinary kriging method for these analyses. Cokriging analysis has been done to improve the accuracy of prediction, by including the elevation as a co-variate. It has not resulted in significant improvement in the prediction. It was observed that the rainfall data is skewed and Box-cox transformation has been used for converting the skewed data to nomal. It is observed that the trend is present in all the cases, and is constant for November, North-East monsoon. The first order polynomial fits well for June, August, Sept, October, December, Annual period and South -West monsoon. The second order polynomial fits best for July. It has been observed that the directional effects are predominant in October, November, South-West Monsoon and Annual rainfall. Spherical model fits well for June, July, November, South-West and North-East monsoons, where as the Gaussian model fits well for August, September, October, December and annual rainfalls. Nugget effect is zero for June, November, and North-East monsoon. The cross-validation error statistics of OCK presented in terms of coefficient of determination (R-2), kriged root mean square error (KRMSE), and kriged average error (KAE) are within the acceptable limits (KAE close to zero, R-2 close to one, and KRMSE from 0.98 to 1.341). The exploratory data analysis, variogram model fitting, and generation of prediction map through kriging were accomplished by using ESRI'S ArcGIS and geostatistical analyst extension.