ASM Science Journal: Modelling Extreme Rainfall Using Adjusted Sandwich Estimator

Rainfall patterns can vary from place to place, and treating each location separately may lead to inaccurate results. By considering how nearby locations are connected, this study gives a clearer picture of how often heavy rain happens.

The Generalised Extreme Value (GEV) distribution is widely used to model the frequency of extreme rainfall events. However, applying the GEV distribution independently to spatial data does not capture the dependencies that often exist between different sites. This limitation can lead to inaccurate statistical assumptions when using a marginal approach. To address this, adjustments can be made using a sandwich estimator.

In this study, the author first applied the conventional marginal fitting of the GEV distribution to extreme rainfall data, and then corrected the standard errors to account for inter-site dependence. Additionally, the author employed penalised maximum likelihood estimation to improve the accuracy of parameter estimates.

A case study was conducted using annual maximum rainfall data from several stations in Western Sabah. The findings indicate that the variances were larger than the standard errors produced by marginal estimation, highlighting the influence of inter-site dependence on extreme rainfall modelling.

Visit the ASM Science Journal to read the full article.