Comparative Study of Machine Learning Methods for Mapping Forest Fire Areas Using Sentinel-1B and 2A Imagery

Aug 5, 2026

Upload by Liam

The study investigates recognition and mapping of burning and burnt areas for a large-scale forest fire in Xintian County, China, occurring in October 2022. Three machine-learning algorithms—SVM, Random Forest, and Neural Network—are compared for adaptability under pre-fire, during-fire, and post-fire scenarios using Sentinel-1B and Sentinel-2A imagery. Preprocessing combines remote-sensing features to identify fire-related land-cover types. Results show SVM performs best pre- and post-fire, while RF achieves top accuracy during fire.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages19

Reading Time48 min

AccessFree to download

Introduction

Forest fires and remote sensing needs

Methods

Data preprocessing and feature construction

Machine learning models and experimental design

Results

Pre-fire period performance (SVM)

During-fire period performance (RF)

Post-fire period performance (SVM)

Discussion and implications

Algorithm selection across scenarios

Technical support for mapping damage from fires

TYPE Original Research PUBLISHED 04 December 2024 DOI 10.3389/frsen.2024.1446641 OPEN ACCESS EDITED BY Adrián Regos, Forest Technology Center of Catalonia (CTFC), Spain REVIEWED BY Huiran Gao, Ministry of Emergency Management, China Xintao Liu, Hong Kong Polytechnic University, Hong Kong SAR, China Bilal Muhammad, Beijing Forestry University, China *CORRESPONDENCE Xinbao Chen, [xchen@hnust.edu.cn](xchen@hnust.edu.cn)[ ](xchen@hnust.edu.cn)Yaohui Zhang, [eminem@mail.hnust.edu.cn](eminem@mail.hnust.edu.cn)[ ](eminem@mail.hnust.edu.cn)Zecheng Zhao, [1901080108@mail.hnust.edu.cn](1901080108@mail.hnust.edu.cn) RECEIVED 10 June 2024 ACCEPTED 18 November 2024 PUBLISHED 04 December 2024 CITATION Chen X, Zhang Y, Wang S, Zhao Z, Liu C and Wen J (2024) Comparative study of machine learning methods for mapping forest ﬁre areas using Sentinel-1B and 2A imagery. Front. Remote Sens. 5:1446641 . doi: 10.3389/frsen.2024.1446641 COPYRIGHT © 2024 Chen, Zhang, Wang, Zhao, Liu and Wen. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY) . The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms. Comparative study of machine learning methods for mapping forest ﬁre areas using Sentinel-1Band 2A imagery Xinbao Chen 1,2*, Yaohui Zhang 2*, Shan Wang 3, Zecheng Zhao 2*, Chang Liu 2 and Junjun Wen 3 1Sanya Institute of Hunan University of Science and Technology, Sanya, China, 2School of Earth Sciences and Spatial Information Engineering, Hunan University of Sciences and Technology, Xiangtan, China, 3Hunan Institute of Geological Disaster Investigation and Monitoring, Changsha, Hunan, China The study focuses on identifying ﬁreburning and burnt areas in a large-scale forest ﬁre that occurred in Xintian County, China, in October 2022 . To investigate the adaptability of machine learning methods in various scenarios for mapping forestﬁre areas, this study presents a comparative study on the recognition and mapping accuracy of three machine learning algorithms, namely, Support Vector Machine (SVM), Random Forest (RF), and Neural Network (NN), based on Sentinel-1B and 2A imagery. Initially, three sets of pre-ﬁre, during-ﬁre, and post-ﬁre remote sensing data were preprocessed. Various feature parameters from Sentinel-1B and 2A imagery were combined to identify ﬁrerelated land cover types. The experimental results revealed that: (i) During the pre-ﬁre period, the SVM method demonstrated superior accuracy compared to the other two methods. The combination of spectral and Normalized Difference Vegetation Index (NDVI) features achieved an optimal accuracy for identifying forest areas with an overall accuracy (OA) of 93 .52% . (ii) In the during-ﬁre period, RF method exhibited higher accuracy compared to the other two methods with peak ﬁre identiﬁcation accuracy reached by combining spectral and Normalized Burn Ratio (NBR) index features atan OA of 95 .43% . (iii) In the post-ﬁre period, SVM demonstrated superior accuracy compared to other methods. The highest accuracy of 94 .97% was achieved when combining spectral and radar features from Sentinel-1B imagery, highlighting the effectiveness of using spectral and radar backward scattering coefﬁcients as feature parameters to enhance forest ﬁre recognition accuracy for burnt areas. These ﬁndings suggest that appropriate machine learning algorithms should be employed under different conditions to obtain more precise identiﬁcation of forest ﬁre areas. This study provides technical support and empirical evidence for extracting and mapping forest ﬁre areas while assessing damage caused by ﬁres. KEYWORDS comparative study, fores

FAQs

What fire event and study area are used in the research?

The study targets a large-scale forest fire in Xintian County, China, in October 2022, focusing on identifying burning and burnt areas.

Which machine learning algorithms are compared for forest fire area mapping?

Support Vector Machine (SVM), Random Forest (RF), and Neural Network (NN) are compared for mapping accuracy under different time periods.

How do the best-performing methods differ across pre-fire, during-fire, and post-fire periods?

SVM shows superior accuracy in the pre-fire and post-fire periods, while RF achieves higher accuracy during the during-fire period.

Aug 5, 2026

Upload by Liam