Evaluation of Fuzzy Regression Efficiency for Reconstructing Missing Annual Precipitation Data in Karoon Basin

Continuous and disperse blanks in most hydrological data (e. g. rainfall data) often occur due to data loss، elimination of incorrect data and the malfunctioning of measuring instruments; these then need to be estimated and/or evaluated for subsequent analysis. There are various methods available for estimating and regenerating these data، the accuracy of which depends very much on the specific conditions of the station، so that one specific method may suit a particular station. Generally، data from four adjacent stations are used for regenerating the missing data at a particular station. In this research، fuzzy regression efficiency is employed for reconstructing yearly rainfall data in Karoon basin. The results are compared with methods such as normal ratio، graphical، simple linear regression and multivariate linear regression. Reconstruction groups were formed using the clustering method in minitab software. Twenty-five stations، similar in their duration of data collection، were selected from among stations in the northern Karoon basin and these were classified into 5 clusters. Following data elimination by cross validation، their value was estimated using the above mentioned methods. Then، using the root mean square of errors (RMSE)، the priority was evaluated for each method. The results of yearly data regeneration indicate that fuzzy regression yielded more accurate estimates in 12 out of the 25 stations studied، or in 3 clusters out of the 5 classified group، making it the most appropriate method for regenerating data for the whole of Karoon basin.