Multivariate Fire Risk Models Using Sample-Size Reduction and Copula Regression in Kalimantan, Indonesia

The copula-based joint distribution can construct a fire risk model to improve forest fires' early warning system, especially in Kalimantan. In this study, we model and analyze the copula-based joint distribution between climate conditions and hotspots. We used several climate conditions, such as total precipitation, dry spells, and El Nino-Southern Oscillation (ENSO). We used copula functions with sample size reduction to construct the joint distributions and the copula regression model to estimate the fire size. The results show that the probability of extreme hotspots number during normal ENSO conditions is very rare and almost near zero during La Nina. Other than that, extreme hotspot event (more severe than in 2019) during El Nino is more sensitive to total precipitation than dry spells based on the conditional survival function. However, the copula regression model found that the model used dry spells as a climate condition better than total precipitation. In this model, the 95% confidence interval of the expected hotspots can cover all actual hotspots data.