Statistical Analysis of Rainfall Data: A Case Study of Georgetown, Guyana

Abstract

Rainfall is a key component for designing of many engineering projects, including canals, bridges, culverts, and road drainage systems. This research will provide useful information to water resource planners, farmers, and urban engineers for analysis of water availability and for building appropriate engineering structures. In order to calculate the appropriate input value for the design and analysis of engineering structures as well as for crop planning, a thorough statistical study is required. The Botanical Gardens raingauge station in Georgetown was selected for analysis as Georgetown experiences frequent floods. The mean, standard deviation and coefficient of variation, skewness and kurtosis of monthly and annual rainfall for a 30 year period were analyzed. According to the computed data, the rainfall pattern is unpredictable, as demonstrated by the coefficient of variation. The variability in rainfall for May, June and July are (38%, 35% and 37%) respectively are much lower than the other months. An analysis of the average monthly rainfall data shows a bi-modal yearly rainfall pattern. Various plotting position formulae and probability distribution functions were used to analyze the return period of the yearly rainfall. It was determined that the Chegodayev technique provides the best fit distribution for yearly rainfall data using the plotting position method, whilst the Gumbel’s Extreme Distribution yields the highest value for rainfall for various return periods using the probabilistic methods. Chegodayev technique also provides the best fit annual one day maximum rainfall for the plotting position methods The Chegodayev technique projects that rainfall for 5, 10, 50, 100 and 150 year return period are 2668.31 mm, 3024.23 mm, 3850.65 mm 4206.56 mm and 4414.76 mm whilst Gumbel’s Extreme Distribution projects rainfall of 2761.54 mm, 3090.30 mm, 3813.85 mm, 4119.73 mm and 4298.10 mm respectively for the same periods. It should be noted that the result from the plotting position formulae are usually good for small extrapolations as errors increases with the amount of extrapolations.