Modeling Fire Hotspots in Kalimantan, Indonesia Using Nested 3-Copula Regression Based on Precipitation and Dry Days During Different ENSO Phases

Authors: TW Mas’oed, S Nurdiati, A Sopaheluwakan, MK Najib, A Salsabila.  

Abstract: The persistent forest and land fires in Indonesia have piqued the interest of numerous social groups. Due to the huge number of losses caused by fires, fire prediction is an important part of fire prevention. This study focused on Kalimantan, which is one of the major contributors to fires in Indonesia. This article focuses on modeling the relationship between total precipitation, the number of dry days, and hotspots in Kalimantan, Indonesia for each ENSO phase using a nested 3-copula approach. Using the selected copula structure, the number of hotspots was estimated using nested 3-copula regression with two predictors. Copula regression offers more robustness to outliers and non-normality in the data compared to traditional regression techniques. The results reveal that the regression model based on ENSO phases has an RMSE of 1204 hotspots per month and can explain up to 70% of the variance in hotspots. These results outperform models without ENSO phases, highlighting the importance of ENSO phases in simulating hotspots in Kalimantan. From the regression plane, the ENSO phase has a small impact on the hotspots at low levels. When it comes to high or intense hotspots, the ENSO phase is very important. El Nino is the most dangerous phase for extreme hotspots, while La Nina is the least dangerous. The findings of this study can help researchers better understand the influence and dependence of local and global climate conditions on hotspots in Kalimantan, which can be evolved into an early warning model for forest fires in Indonesia in the future.

Keywords: Copula regression, Dry spells, Hierarchical copula, High-dimension, Multivariate copula, Uncertainty, Wildfire

link: http://dx.doi.org/10.21163/GT_2024.192.21

Dipublikasikan pada Geographia Technica, Vol 19 (2): 264-281