Abstract This study investigates the spatial variability of forest fire intensity, burn indices, ecosystem productivity, and Greenhouse Gas (GHG) emissions in Pakistan from 2001 to 2023. Using satellite-derived burn indices such as SAVI, LST, NMDI, LSWI, NBR, and MSAVI2, the study examines the relationship between forest fires and net primary productivity (NPP) across diverse ecological regions. The analysis reveals that northern Pakistan, particularly Khyber Pakhtunkhwa and Gilgit-Baltistan, experiences high fire intensity, resulting in significant reductions in NPP and increased emissions of COx, NOx, and CH₄. Central and southern Pakistan, including the arid regions of Balochistan and Sindh, exhibit lower fire intensity but remain vulnerable due to climate-driven dry conditions. The study also applies the ΔNPP/ΔBurn approach to evaluate how changes in burn indices correspond to shifts in NPP, revealing that small increases in fire intensity can lead to substantial ecosystem productivity loss. Additionally, a comparative analysis of Random Forest (RF) and XGBoost machine learning models for fire prediction found RF to be the more accurate model, achieving 88.0% accuracy and a 93.8% AUC score. These findings underscore the importance of developing region-specific fire management strategies to mitigate the ecological and environmental impacts of wildfires. The study highlights the critical need for improved fire prediction, early warning systems, and long-term monitoring of post-fire ecosystem recovery. By drawing comparisons with global research, this study contributes to understanding the broader implications of forest fires on carbon dynamics and ecosystem productivity, providing valuable insights for future fire management policies in Pakistan.
Shahzad et al. (Fri,) studied this question.