Publications

Kumari, S; Mamgain, S; Roy, A; Prince, HC; Ahlawat, A (2024). Earth Observation Based Characterization of Environmental Conditions for Forest Fire Risk in Western Himalayan Ecosystems Using Machine Learning Approach. JOURNAL OF THE INDIAN SOCIETY OF REMOTE SENSING.

Abstract
The increased frequency and intensity of forest fires have necessitated the characterization and evaluation of factors influencing fire occurrences in the Western Himalaya, which is essential for managing and mitigating forest fire disasters. Environmental conditions leading up to fire episodes such as pre-fire conditions, anomalies in weather phenomenon and seasonality in biophysical parameters, have been identified to be intricately linked to forest fire occurrences. Information on the factors influencing forest fire, retrieved from satellite data in Google Earth Engine (GEE) and analysed using R Studio through machine learning algorithms (Random Forest and Maximum Entropy), was used to estimate spatial forest fire risk maps. Results showed a significant influence of seasonality of biophysical parameters in determining forest fire season. Furthermore, local conditions and fuel availability shows strong association with forest fire risk. For instance, Uttarakhand falls under moderate to very high forest fire probable risk category compared to Himachal Pradesh and Jammu and Kashmir. Statistically, Random Forest method showed better performance compared to Maximum entropy method, as indicated by receiver operating characteristics (ROC) curve. The study provides a clear understanding of quantitative influence of environmental and local factors responsible for forest fire risk and can be used for forest fire management and mitigation.

DOI:
10.1007/s12524-024-02002-0

ISSN:
0974-3006