Modeling the Dynamics of Forest Fires: A Vector Autoregressive Approach Across Three Fire Classifications

Abstract

The problem of forest fires is one that, with each passing year, gets more difficult to mitigate. A significant number of people will be affected by this case, particularly in terms of their health. The need for targeted initiatives must be balanced. Look at the forecasts for the number of forest fires expected to occur in the following period. Cases of forest fires reported to the Ministry of Environment and Forestry are categorized into three distinct categories: high, medium, and low. In addition to future estimates, it is reasonable to anticipate that classifications will also affect one another. The vector autoregressive (VAR) model is a statistical tool that may produce future projections based on three categories of forest fires in a specific period. This information can be utilized to make predictions. The aim of the study was to model 3 classifications of forest fire cases using the Vector Autoregressive (VAR) model. The data utilized is a summary of the number of forest fire cases in Pulang Pisau Regency, Central Kalimantan, categorized as low, medium, and high, from January 2013 to March 2024. During this study, the VAR modelling process was broken down into three primary stages: order identification (the findings that were achieved were VAR(4)), parameter estimation, and diagnostic testing (VAR(4) was declared to fulfil the requirements for the diagnostic test). It is possible to generate a predicted value for the subsequent three times based on these stages, which may be considered when calculating the proper amount of effort to put forward. The accuracy of forest fire case modeling utilizing the VAR(4) model is 70.23%. Moreover, the predictive outcomes for each categorization indicate a rise in medium and low-level forest fires compared to previous data, although the contrary is observed for high-level forest fire incidents.