Thesis (M.Eng.) - Asian Institute of Technology, 1993

A thesis submitted in partial fulfillment of the requirements for the degree of Master of Engineering.

Analysis of the hydrological data of the Karnali river basin, Nepal has been carried out by kriging. A theoretical outline of the geostatistical principles and the kriging system has been presented in a concise form. Based on the principles of kriging a methodology has been developed for selecting the best locations for a given number ofrain gauges planned to be added in a network with an objective to improve the reliability of the data. Use has been made of the software Geo-EAS for structure identification in the form of variogram analysis and for carrying out point and block kriging. The kriging technique has been utilized to interpolate precipitation depths at ungauged sites, to estimate the areal average of precipitation and most importantly for optimal network design. The results are compared with other interpolation methods in practice. Kriging, principally the 'best linear unbiased estimator' (BLUE) is found to be superior to other traditional approaches in the sense that kriging provides a quantitative measure of the accuracy of estimation in the form of 'kriging estimation variance' where as the other interpolation techniques involve subjective bias andfail to quantify the accuracy of estimation. It has been demonstrated that kriging can be of valuable use in identifying the optimal locations for a set of additional rain gauges using Kriging Standard Deviation as an indicator. The results obtained from kriging are realistic to the extent the stationarity assumptions are true. Therefore, if sufficient sample data points are available a large basin should be divided into smaller homoge.neous regions where the assumption of stationary mean can be considered valid. Difficulty involved in such a regionalization approach is delineation of the homogeneous zones and identification of the true structure due to lack of sufficient sample data points.