Poster Session #2: UC Ballroom

Examining Airborne Infrared Fire Detection Data in the Context of Fire Severity

Presentation Type

Poster

Faculty Mentor’s Full Name

Carl Seielstad

Faculty Mentor’s Department

Department of Forest Management: National Center for Landscape Fire Analysis

Abstract / Artist's Statement

Application of the LANDSAT-based delta-Normalized Burn Ratio (dNBR) for identifying unburned islands within fire perimeters is a subject of interest to ecologists examining vegetation refuges in burned landscapes. However, questions remain as to whether fire actually visited some these refuges but did not trigger change responses in the dNBR. This research integrated forty-one infrared fire detection acquisitions to examine the coincidence of thermal infrared heat detections and dNBR severity classes in the 274 km2 Selway-Bitterroot East Complex of fires (Clearwater National Forest, ID) which burned in 2012. Delta-NBR was imputed using conventional methods from LANDSAT-5 and -8 datasets. Thermal infrared fire detection data were collected by the USDA Forest Service’s National Infrared Operations (NIROPS) Phoenix System on 35 days in August and September and detection points were integrated and attributed by date and time to produce a single fire detection point layer spanning the duration of the incident. Although Phoenix data are not calibrated and have variable fire detectability thresholds depending on the amount of fire activity and background temperature, they provide a conservative map of heat locations that does not suffer from errors of commission (e.g, the heat detected is real) and is useful for identifying locations of heat at distinct points in time. For this study, IR detections were compared with dNBR polygons in a GIS. Ten percent of the total fire detections occurred in ‘unburned’ areas. The proximity of these detections to severity class edges was examined to account for potential spatial incongruities between the datasets as well as areas of sensor saturation (bloom). Further, detections were summarized by severity classes to identify possible relationships between detection frequency and burn severity. These comparisons will serve as an initial framework documenting the efficacy of using LANDSAT-derived dNBR to locate unburned areas within large wildfires.

Category

Physical Sciences

This document is currently not available here.

Share

COinS
 
Apr 11th, 3:00 PM Apr 11th, 4:00 PM

Examining Airborne Infrared Fire Detection Data in the Context of Fire Severity

Application of the LANDSAT-based delta-Normalized Burn Ratio (dNBR) for identifying unburned islands within fire perimeters is a subject of interest to ecologists examining vegetation refuges in burned landscapes. However, questions remain as to whether fire actually visited some these refuges but did not trigger change responses in the dNBR. This research integrated forty-one infrared fire detection acquisitions to examine the coincidence of thermal infrared heat detections and dNBR severity classes in the 274 km2 Selway-Bitterroot East Complex of fires (Clearwater National Forest, ID) which burned in 2012. Delta-NBR was imputed using conventional methods from LANDSAT-5 and -8 datasets. Thermal infrared fire detection data were collected by the USDA Forest Service’s National Infrared Operations (NIROPS) Phoenix System on 35 days in August and September and detection points were integrated and attributed by date and time to produce a single fire detection point layer spanning the duration of the incident. Although Phoenix data are not calibrated and have variable fire detectability thresholds depending on the amount of fire activity and background temperature, they provide a conservative map of heat locations that does not suffer from errors of commission (e.g, the heat detected is real) and is useful for identifying locations of heat at distinct points in time. For this study, IR detections were compared with dNBR polygons in a GIS. Ten percent of the total fire detections occurred in ‘unburned’ areas. The proximity of these detections to severity class edges was examined to account for potential spatial incongruities between the datasets as well as areas of sensor saturation (bloom). Further, detections were summarized by severity classes to identify possible relationships between detection frequency and burn severity. These comparisons will serve as an initial framework documenting the efficacy of using LANDSAT-derived dNBR to locate unburned areas within large wildfires.