Analyzing rainfall trends of various sites of south and south-western areas of Pakistan

Abstract

Abstract. Frequency analysis of rainfall is important in many aspects including agriculture water management, preemptive measures against natural disasters, etc. This study identified best fitted distribution(s) to predict the rainfall trends of south and south-western areas of Pakistan namely Balochistan and Sindh. Secondary data is obtained from Pakistan Meteorological Department for the period 1980 to 2015. Firstly, we have checked the basic assumptions such as randomness and identical distribution of the observed data series by using Run and Mann Whitney tests. Further, three parameter distributions named as Generalized extreme value, Weibull and Frechet three parameters distributions, and compared with two parameters Weibull, Frechet and Log-logistic distributions. The parameters of the three and two parameters candidate distributions were obtained by using maximum likelihood and probability weighted moment methods.The performances of the candidate distributions were checked out by using Kolmogorov Smirnove test at 5 percent level of significance. Moreover, probability density functions graphs were used to indorse the behavior and the theoretical framework of the rainfall distributions of selected sites. Furthermore, quantiles were estimated for future prediction. The resultant estimates depicted that two parameters distributions and three parameters distributions were the most appropriate for annual maximum rainfall series.
Keywords:Generalized extreme value distribution; Frechet distribution; Weibull distribution; Log-logistic distribution; Maximum likelihood method; Probability weighted moment method.