Identification and Tracking of Wildfires and Wildfire-Induced Smoke Plumes During FIREX-AQ 2019 using Orbital and Suborbital Instruments.

Abstract:

The detection and tracking of objects, like smoke plumes, within a single instrument’s data has long required the development of instrument-specific retrieval algorithms. Such development is labor intensive and requires domain-specific parameters and instrument-specific calibration metrics, alongside manual effort to track retrieved objects across multiple scenes. With a special emphasis on supporting NASA’s Fire Influence on Regional to Global Environments Experiment - Air Quality (FIREX-AQ) campaign, we present our progress in the development of a systematic deep learning framework to track wildfires and wildfire-induced smoke plumes in datasets from multiple sensors as well as multi-sensor fused datasets. The tracked plumes complement measurements made during FIREX-AQ to assess the formation, properties, and transport of smoke plumes, as well as their impacts on downwind air quality. By exploiting the synergy between L1B observations from multiple airborne instruments, geostationary, and polar orbiting satellites, we demonstrate our capability to use data from multiple sensors to track and more precisely characterize wildfires and smoke plumes than can be done for wildfires and smoke plumes observed by a single instrument.