Statistical modeling of weekly rainfall : a case study in Colombo City in Sri Lank

Abstract

The modeling of the weekly rainfall percentile is imperative for better understanding of rainfall patterns in any region. This study focuses on selecting the most appropriate probability distributions for weekly rainfall and use those to make reliable rainfall percentile with the 95% confidence intervals. Daily rainfall data of 56 years (1960-2015) during the South West Monsoon in Colombo City were used for this analysis. The three parameter Weibull distribution has been found most probable distribution for most of weekly rainfall totals. Weibull, two parameter Exponential, Exponential and Lognormal distributions were well fitted distributions for remaining totals. Based on the 95% confidence intervals of percentiles, the weeks 18-23, and 38-39 during SWM showed not only high rainfall, but also high rainfall variation results which caused high possibility to form extreme rainfall events. Heavy rainfall with great variation during the period of 30th April to 10th June and 17-30th of September was further confirmed by the result of running total of weekly rainfall.