Extreme Rainfall Analysis with Spatial Extreme Value through the Extremal-t Process Husna Miratin Nuroini- Dr. Sutikno, M. Si.- Dr. Purhadi, M. Sc.
Sepuluh Nopember Institute of Technology
Abstract
The topographical conditions in Indonesia consist of coasts, lowlands, highlands, and mountains. This causes the area to have a high diversity of weather and climate, enabling several hydrological phenomena such as extreme rainfall, hurricanes, high temperatures, and storms. Nowadays, global warming has become a widely discussed environmental issue. One of its impacts is climate change, which affects the occurrence of extreme hydrological phenomena and potentially causes floods, disruption of transportation and communication, damage to infrastructure, harm to the agricultural sector, and threat to life. Therefore, this study aims to obtain the best model and return level of extreme rainfall in Ngawi Regency from 1990 to 2022 through Spatial Extreme Value (SEV) using Max-Stable Processes (MSP) method with the extremal-t process. The estimation parameters used are Maximum Likelihood Estimation (MLE) and Maximum Pairwise Likelihood Estimation (MPLE), then solved by Broyden-Fletcher-Goldfarb-Shanno (BFGS) Quasi-Newton numerical iteration methods. Based on the analysis result, the best trend surface model obtained with the average rainfall and variance influenced by latitude. Meanwhile, the shape of rainfall distribution is not affected by longitude or latitude because it is assumed to be isotropic. In addition, the value of return level prediction has better accuracy when used over 3-years period.
