Statistical Frequency Analysis of Extreme Rainfall and Development of Intensity–Duration–Frequency Curves Using Chirps Satellite Data

Abstract

Urban flash flooding is a major challenge in cities because of its climate and its frequent urbanizing of land, being converted to impervious surfaces supporting surface runoff and drainage overflow. This study analyzed extreme rainfall data of the campus, developing an Intensity-Duration-Frequency (IDF) curve using statistical frequency analysis, using high-resolution 25-year CHIRPS satellite data, because ground data wasn’t available. Non-parametric trend tests were conducted to understand the data stability for extreme value analysis. The Mann-Kendall’s test and Cox-Stuart tests confirmed that the series computed in Hydrognomon exhibits stationarity over time. The Pettitts Test indicated a sudden change in 2021, followed by a reduced peak anomaly over the next three years. To select the best fitting model, six probability distribution models were compared resulting to Log-Pearson Type III being chosen under the Bayesian Information Criterion (BIC) because of its mathematical parsimony and its unique ability to handle highly skewed data without overfitting. This model was used to estimate the rainfall depth of 10-year, 25-year, 50-year and 100-year projecting a depth of 110.24mm and 143.42mm by year 10 and 50 respectively. These depths were converted to 24-hour rainfall duration and used to develop the IDF curve using the Sherman’s equation. The resulting equations provided calibrated values engineers can use to design better drains, and stormwater structures on campus.