Dynamics of Forest Fires in P.N.N. El Tuparro: Application of Remote Sensing and Machine Learning in Environmental Monitoring

Abstract. Forest fires are among the leading causes of sudden vegetation cover loss, originating from natural phenomena or human-induced activities. These events are significant ecological disturbances that can severely alter ecosystem structure, degrade soil properties, and disrupt natural regeneration processes. In Colombia's El Tuparro National Natural Park (P.N.N. El Tuparro), fire dynamics are strongly influenced by climatic variability and land-use practices. This study addresses the need for precise monitoring tools to assess the impact of wildfires in protected areas. We used multi-temporal Landsat imagery (2013–2023) and supervised classification with Support Vector Machines (SVM) to detect and quantify burned areas. A set of spectral indices (NDVI, NBR, SAVI, NDWI) was calculated to evaluate fire severity and vegetation recovery. Results reveal a strong relationship between fire events and reductions in vegetation vigor and biomass, with grassland ecosystems showing the highest resilience. Anomaly detection using the Isolation Forest algorithm identified extreme fire-related events and climatic anomalies. These findings emphasize the importance of continuous vegetation monitoring and the development of mitigation strategies to reduce wildfire impacts in vulnerable ecosystems.