Sensitivity of Spectral Indices on Burned Area Detection using Landsat Time Series in Savannas of Southern Burkina Faso

• Main Authors: Jinxiu Liu, Eduardo Eiji Maeda, Du Wang, Janne Heiskanen
• Format: Article
• Language: English
• Published: MDPI AG 2021-06-01
• Series: Remote Sensing
• Subjects: burned area; spectral indices; Landsat time series; savanna
• Online Access: https://www.mdpi.com/2072-4292/13/13/2492

Accurate and efficient burned area mapping and monitoring are fundamental for environmental applications. Studies using Landsat time series for burned area mapping are increasing and popular. However, the performance of burned area mapping with different spectral indices and Landsat time series has not been evaluated and compared. This study compares eleven spectral indices for burned area detection in the savanna area of southern Burkina Faso using Landsat data ranging from October 2000 to April 2016. The same reference data are adopted to assess the performance of different spectral indices. The results indicate that Burned Area Index (BAI) is the most accurate index in burned area detection using our method based on harmonic model fitting and breakpoint identification. Among those tested, fire-related indices are more accurate than vegetation indices, and Char Soil Index (CSI) performed worst. Furthermore, we evaluate whether combining several different spectral indices can improve the accuracy of burned area detection. According to the results, only minor improvements in accuracy can be attained in the studied environment, and the performance depended on the number of selected spectral indices.