Predicting socioeconomic drivers of deforestation in Murree forest region, Pakistan using statistical inference and machine learning approach

Rapid deforestation in Pakistan poses severe environmental challenges, threatening the fragile ecosystem, altering the hydrological cycle, intensifying the risk of flash flooding and contributing to climate change. The socioeconomic drivers of deforestation in Pakistan, particularly in mixed state and community forest landscapes, remain poorly understood. This study investigates the socioeconomic drivers of deforestation in the Murree Forest Region, Pakistan using statistical inference and machine learning models: Support Vector Machine (SVM), Random Forest (RF) and Logistic Regression. Descriptive statistics revealed that a total of 84.3% of respondents were landholders in Murree, with Business (43.8%) and Government employment (32.7%) as primary sources of income. Approximately 72.5% of respondents collect 1-25 kg of fuelwood daily, while 2.4% collect >90 kg of fuelwood per day. Logistic regression identified fuelwood collection, inadequate education, forest fire, proximity to forest, and illegal logging as perceived major drivers of deforestation (p < 0.001). Moderation analysis indicated that education did not change the perceived effect of fuelwood on deforestation. Machine learning models demonstrated strong predictive performance, with SVM achieving the best performance. Feature importance analysis consistently identified illegal logging, forest fire, livelihood dependence, distance from forest, and poor governance as the most influential predictors. These findings highlight that the hybrid analytical approach enhanced the accuracy of identifying the perceived drivers of deforestation. This approach may be used to inform community and government-centered interventions aimed to mitigate deforestation by addressing perceptions regarding forest use and deforestation.