View in EDS

A Hybrid Framework for Forest Fire Detection and Severity Prediction using Sequential Deep Learning on Multitemporal Satellite Imagery.

Saved in:
Bibliographic Details
Title: A Hybrid Framework for Forest Fire Detection and Severity Prediction using Sequential Deep Learning on Multitemporal Satellite Imagery.
Authors: Bhosale, Rohini, Railkar, Poonam
Source: EPJ Web of Conferences; 11/20/2025, Vol. 341, p1-12, 12p
Subject Terms: FOREST fire management, DEEP learning, SUSTAINABLE development, MACHINE learning, LONG short-term memory, REMOTE-sensing images, CHANGE-point problems, FIRE risk assessment
Abstract: The incident, intensity, detection and prediction strategies of forest fires are increasing day by day which is having a significant impact on infrastructure and the global economy around the world, and therefore affecting the Sustainable Development Goals. The aim of this research study is to detect and predict forest fires based on multi-temporal images captured by satellites. It was observed that the majority of fires occur during the pre-monsoon period, especially during the month of March. Out of all areas surveyed, the current and anticipated high-risk areas were marked in the regions with the largest concentration of protected zones. It is vital to control the underground bio-mass burning in the forests at lower elevations to minimize the chances of fire in the peak season. The study underscores the necessity for a well-defined framework, to predict, identify, and prioritize fire-prone zones. Additionally, a deep learning-based hybrid approach using change detection, Long Short-Term Memory (LSTM) and attention mechanism on pre-processed satellite images is proposed for the early detection of forest fires. Change detection is used for the comparison of multiple raster datasets, typically collected for one area at different times, to determine the type, magnitude, and location of change. It is used to track forest fires, access forest wildfire impacts, detect burned areas, and for reducing damage and cost. Long Short-Term Memory Networks or LSTM in deep learning, is a sequential neural network that allows information to persist. It is used to analyse temporal dependencies in the change-detected regions. The attention mechanism is a technique used in machine learning and natural language processing to increase model accuracy by focusing on relevant data by assigning higher weights to important parameters, which makes the model better fit the current data. The trained model demonstrates high accuracy, surpassing traditional methods, and aids in early warning and decision-making for fire management authorities. This combination of remote sensing and deep learning offers a robust system for accurate forest fire detection and prediction, essential for mitigating the impact of forest fires on ecosystems and communities. [ABSTRACT FROM AUTHOR]
Copyright of EPJ Web of Conferences is the property of EDP Sciences and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database: Complementary Index
Description
Abstract:The incident, intensity, detection and prediction strategies of forest fires are increasing day by day which is having a significant impact on infrastructure and the global economy around the world, and therefore affecting the Sustainable Development Goals. The aim of this research study is to detect and predict forest fires based on multi-temporal images captured by satellites. It was observed that the majority of fires occur during the pre-monsoon period, especially during the month of March. Out of all areas surveyed, the current and anticipated high-risk areas were marked in the regions with the largest concentration of protected zones. It is vital to control the underground bio-mass burning in the forests at lower elevations to minimize the chances of fire in the peak season. The study underscores the necessity for a well-defined framework, to predict, identify, and prioritize fire-prone zones. Additionally, a deep learning-based hybrid approach using change detection, Long Short-Term Memory (LSTM) and attention mechanism on pre-processed satellite images is proposed for the early detection of forest fires. Change detection is used for the comparison of multiple raster datasets, typically collected for one area at different times, to determine the type, magnitude, and location of change. It is used to track forest fires, access forest wildfire impacts, detect burned areas, and for reducing damage and cost. Long Short-Term Memory Networks or LSTM in deep learning, is a sequential neural network that allows information to persist. It is used to analyse temporal dependencies in the change-detected regions. The attention mechanism is a technique used in machine learning and natural language processing to increase model accuracy by focusing on relevant data by assigning higher weights to important parameters, which makes the model better fit the current data. The trained model demonstrates high accuracy, surpassing traditional methods, and aids in early warning and decision-making for fire management authorities. This combination of remote sensing and deep learning offers a robust system for accurate forest fire detection and prediction, essential for mitigating the impact of forest fires on ecosystems and communities. [ABSTRACT FROM AUTHOR]
ISSN:21016275
DOI:10.1051/epjconf/202534101057