Document Type : Research Paper
Authors
1
Assistant Professor, Research Division of Forests and Rangeland, Khorramabad Agricultural and Natural Resources Research and Education Center, Agricultural Research Education and Extension Organization (AREEO), Khorramabad, Iran
2
Associate Professor, Department of Nature Engineering, Faculty of Natural Resources and Earth Sciences, Shahrekord University, Iran
3
Associate Professor, Research Division of of Forests and Rangeland, Khorramabad Agricultural and Natural Resources Research and Education Center, Agricultural Research Education and Extension Organization (AREEO), Khorramabad, Iran.
10.22092/ijfrpr.2024.365082.1618
Abstract
Background and objectives: The evaluation of disturbances using satellite imagery is a crucial sub-field in natural resource science, serving as a tool for monitoring changes such as fire and livestock grazing in forest and pasture ecosystems. Grazing intensity and fires are significant disturbances in arid and semi-arid rangelands. Remote sensing for burned area mapping has been extensively studied. Spectral indices are widely used to monitor vegetation cover changes, particularly post-fire, and to generate burned area maps. Given the vastness and inaccessibility of Iran's mountain rangelands, multi-spectral remote sensing data was employed to identify burned areas. This research aimed to determine the most effective auxiliary data to enhance classification accuracy for identifying and delineating burned semi-steppe rangelands, and to estimate burned pasture extent for post-fire management using Landsat-8 imagery.
Methodology: Burned rangeland sites were selected based on information from the Chaharmahal and Bakhtiari Province Department of Natural Resources Protection and local expert input. Old fire occurrence and grazing intensity at selected sites were determined. Twenty-seven burned sites, representing old fire occurrences and high to medium grazing intensity, were delineated using GPS polygons. Maximum likelihood classification (MLC) in Idrisi TerrSet software was used to identify burned pasture areas. Auxiliary data, including pan-sharpened raw bands, Tasseled Cap transformation components, digital elevation model (DEM) derivatives, principal component analysis (PCA) components, and the Normalized Burn Ratio-Thermal (NBRT) index, were evaluated to improve classification accuracy. Overall and Kappa accuracies were assessed using error matrices, and Friedman's rank test was used to compare the effectiveness of different data combinations.
Results: The combination of DEM derivatives and the NBRT index significantly improved the classification accuracy of burned areas with varying grazing intensities, achieving an overall accuracy of 66% and a Kappa accuracy of 63%. The NBRT index, based on near-infrared (NIR), short-wave infrared (SWIR1 and SWIR2), and thermal infrared (Thermal1) bands, effectively distinguished burned semi-arid rangelands with different fire ages and grazing intensities. The high sensitivity of these bands to post-fire vegetation changes contributes to the NBRT index's effectiveness.
Conclusion: Multi-spectral remote sensing data, particularly Landsat-8 imagery, combined with auxiliary data layers, effectively identifies and delineates burned semi-steppe pastures. MLC demonstrated capability despite spectral similarities between regenerated vegetation and surrounding areas. The NBRT index's effectiveness stems from the high sensitivity of its constituent bands (NIR, SWIR1, SWIR2, and Thermal1) to post-fire vegetation changes, enabling accurate identification of burned semi-steppe grasslands.
Keywords
Main Subjects
)