Abdulmajed Bostani; Sharareh Pourebrahim; Afshin Danehkar
Abstract
Recognizing and mapping the sensitivity of forests to fires is crucial for the preservation of ecosystems and biodiversity. This study, utilizing the time-series capability of Landsat 8 satellite imagery and developing an efficient model within the Google Earth Engine (GEE) platform, managed to map the ...
Read More
Recognizing and mapping the sensitivity of forests to fires is crucial for the preservation of ecosystems and biodiversity. This study, utilizing the time-series capability of Landsat 8 satellite imagery and developing an efficient model within the Google Earth Engine (GEE) platform, managed to map the sensitivity of Kurdistan province forests to fires over the the past decade, from 2013 to 2023, in two study areas located in the Marivan, Sarvabad, and Baneh counties. It provided valuable information for land use management and effective resource allocation to prevent and mitigate the impacts of forest fires in the Kurdistan region. The Normalized Burn Ratio (NBR) index was applied to pre- and post-fire season images to detect forest fires. To enhance classification results, areas such as vegetation, residential zones, and water bodies were highlighted as non-fire regions. The Random Forest (RF) model within the GEE platform was employed to achieve the highest classification accuracy. Appropriate training samples were derived from the highlighted results, and image classification using the RF model with 50 decision trees was performed on the GEE platform.To ensure the reliability of the selected training samples, the fire mapping results were compared with point-based fire data from the Kurdistan Province Natural Resources Department. The classification results for the two forest study areas- Marivan and Sarvabad regions in 2016, 2018, and 2020, and the Baneh region in 2018-demonstrated an overall accuracy of 99% and a Kappa coefficient of 0.97. The findings of this study underscore the capability of Landsat 8 imagery in mapping forest fire susceptibility and confirm the acceptable accuracy of the RF model in this context.
