Choosing the best fit probability distribution in rainfall design analysis for Pulau Pinang, Malaysia

Design rainfall is frequently utilized in the planning and design of urban infrastructure, including culverts and urban drainage systems. The choice of an appropriate probability distribution that sufficiently fits the observed rainfall data is one of the key elements in designing rainfall estimation. The current study compared five probability distribution methods which include Gumbel, log-Normal, Normal, Pearson III, and log-Pearson III probability distribution method (PDMs) to analyze the rainfall pattern of the study area. The goodness of fit of the PDMs was also tested by determining the root mean square error (RMSE) scores of PDMs results. Three spatial distribution modelling methods which include inverse distance weighting (IDW), universal kriging (UK), and ordinary kriging (OK) were used to spatially distribute and compare the resulting rainfall pattern from the five PDMs. The findings of the study revealed that the log-Pearson III probability distribution method (PDM) was very outstanding among the five PDMs used in analyzing the rainfall pattern design of the study area. The LPII also showed the lowest root mean square error (RMSE) value. Furthermore, the IDW presented a clearer and more accurate representation of the rainfall design and average annual rainfall of the study area by showing the method with the least RMSE value in comparison to the OK and UK which slightly underestimated and overestimated their prediction results. The result of this study can be applied to effective early warning systems for future rainfall prediction and accurate design of water resource structures for a sustainable environment.