Non-stationary generalised extreme value modeling of extreme rainfall in Makassar, Indonesia

Abstract

Extreme rainfall modeling is crucial for flood-risk assessment, drainage design, and urban planning in tropical monsoon regions. This study analyzed annual maximum rainfall in Makassar, Indonesia, using stationary and non-stationary generalized extreme value (GEV) models. Daily NASA POWER rainfall estimates from 1 January 1981 to 30 September 2024 were aggregated into 44 annual maxima and analyzed using maximum likelihood estimation. Candidate non-stationary models allowed the location and/or scale parameters to vary with time, while the shape parameter remained constant. Model selection used the Akaike Information Criterion and likelihood ratio tests, and return-level uncertainty was quantified using parametric bootstrap confidence intervals. The Mann-Kendall test indicated no significant monotonic trend. The best-fitting model had a quadratic time-varying location parameter and constant scale and shape parameters. A sensitivity analysis excluding the partial 2024 block retained the same selected model. For 2025-2045, the fitted model produced conditional return levels that were higher in later projection years, but these estimates do not confirm a long-term monotonic increase. The model-based 100-year return level was 212.86 mm in 2025 and 285.18 mm in 2045, with wide 95% confidence intervals indicating substantial tail uncertainty. The findings are robust to exclusion of the partial 2024 block but remain conditional on the quadratic temporal structure and the use of gridded precipitation data.

Access to Document

Other files and links

• Link to publication in Scopus
• Open Access Version Available

Fingerprint