A Deep Learning Model Using Geostationary Satellite Data for Forest Fire Detection with Reduced Detection Latency

May 19, 2026

Upload by Anda

Although remote sensing of active fires has been widely studied, early detection remains less addressed, and contextual threshold methods face generalization limits. The study introduces a deep learning forest-fire detection algorithm designed to reduce detection latency using 10-minute interval Himawari-8 Advanced Himawari Imager geostationary satellite data. Random forest and convolutional neural network models are compared, with accuracy and latency benefits tied to temporal/spatial feature contributions, data augmentation, and spatial pattern learning.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages18

Reading Time45 min

Introduction

Satellite-based active fire monitoring

Problem of early detection and generalization

FAQs

What is the main goal of the proposed study?

To develop a deep learning forest fire detection approach that reduces detection latency using high temporal resolution geostationary satellite data.

Which satellite data and time interval are used?

The method uses Himawari-8 Advanced Himawari Imager observations with a 10-minute interval to support early detection.

How do temporal and spatial information affect performance?

Temporal factors primarily reduce detection latency, while spatial factors primarily reduce false alarms; the best results come from combining both types of information.

May 19, 2026

Upload by Anda