Evaluation of Remote Sensing-derived Active Fire Perimeter Delineation Methods for Operational Purposes
Date:
Abstract: Forest fires significantly impact forested areas across the globe including Canadian ones. While forest fires are often perceived negatively due to their damaging effects on ecosystems, biodiversity, and human property, they also have several advantages, such as controlling diseases and insects, altering wildlife habitat patterns, and exposing minerals in the soil. Considering these impacts, it is crucial to conduct more research on accurately and promptly monitoring forest fires to aid in minimizing losses while maximizing benefits. The objective of this study was to address the necessity of early forest fire monitoring by evaluating the effectiveness of buffer, concave hull, and convex hull algorithms, as well as their combinations, in estimating active forest fire perimeters. A comprehensive comparison was conducted to identify the most effective method for delineating forest fire boundaries. This study used VIIRS 375 m and MODIS 1000 m active fire datasets for 30 selected forest fires in the Alberta and Northwest Territories region from 2015 to 2021. The study results highlighted that the VIIRS-derived forest fire perimeters exhibited superior performance, yielding lower Commission Errors (CEs) and Omission Errors (OEs) compared to MODIS. Specifically, the α shape Concave Hull algorithm demonstrated higher precision with the lowest CEs and OEs at 24.60% and 22.23%, particularly in sparse or irregularly distributed datasets. In contrast, the round shape Buffer Algorithm performed better in detecting errors in dense or regularly distributed datasets with CEs and OEs at 24.56% and 24.89%, although it tended to overestimate. Additionally, it was found that combination methods generally achieve higher matching percentages with referenced areas but also depicted higher CEs. These findings underscore the importance of enhancing rapid response, resource allocation, and evacuation planning in forest fire management. To our known knowledge, this study may be the first to employ multiple algorithms in both individual and synergistic approaches applicable for both Near Real-Time (NRT) and Ultra Real-Time (URT) active fire data, providing a critical foundation for future studies aimed at improving the accuracy and timeliness of forest fire perimeter assessments.
Conference page: https://wildlandfirecanada.com/