The BRIGHT/AHI algorithm, funded as part of the Australian federal government’s Smartsat CRC Bushfire initiative, was developed by a team of researchers from the former Natural Hazards CRC and Natural Hazards Research Australia.
It uses images from the Himawari-8 satellite of the Japanese Meteorological Agency to provide automated, near-continuous, and near-real-time surveillance of potential fire activity across Australia. Previously, polar-orbiting earth observation satellites could only provide one or two images per day, limiting the detection of potential hotspots. However, with the addition of geostationary satellites, new opportunities have emerged for persistent continental monitoring of fires, improving the ability to observe fire in the landscape from space.
Fire detection algorithms typically identify hotspots by comparing the temperature of a candidate pixel to a reference value. Traditional fire detection methods compare the candidate pixel to its surrounding pixels, using fixed threshold values to trigger a detection.
However, this technique can be problematic, especially when fires occur in complex landscapes or when cloud and/or smoke obscures or partially obscures pixel values. In a country as large and diverse in fire regimes as Australia, these challenges are further compounded.
The BRIGHT/AHI algorithm adopts a novel approach to detect fire activity from earth observation data. By leveraging the AHI Himawari-8 geostationary imager that captures images of Australia every ten minutes, day and night, the algorithm calculates bioregion-specific control values and associated thresholds based on a pixel’s temperature history. This history covers a 30-day period, and the values are optimised to minimise errors of omission in near-real-time. Upon receiving the satellite imagery, the algorithm processes the entire continent of Australia in less than 45 seconds, providing timely and accurate bushfire information. The algorithm’s detections have been compared to those of existing polar-orbiting, satellite-based fire detection products (Chatzopoulos-Vouzoglanis et al., 2022).
The release of the new BRIGHT/AHI hotspot product to Australian fire management agencies and the public is timely, given that two polar-orbiting fire detection products, namely the Moderate Resolution Imaging Spectrometer (MODIS) and the Visible Infrared Imaging Radiometer Suite (VIIRS), are approaching the end of their operational lifecycles.
The MODIS instruments were launched in 1999 (Terra satellite) and 2002 (Aqua satellite), and both have exceeded their designed operational lifetimes of six years. While the Suomi NPP VIIRS maintains the continuity of earth observations, this was recently compromised when the satellite went into safe mode in July 2022 and only recently came back online.
The BRIGHT/AHI algorithm, together with geostationary satellites, represents a new generation of near-real-time earth observation products that are revolutionising our ability to understand fire activity, regimes, impacts, and recovery in a systematic and replicable way for Australia.
Recent developments and testing have led to the inclusion of Fire Radiative Power (FRP) estimate alongside hotspots, with associated products of fire severity and combustion completeness being developed by the RMIT team. However, the challenge of persistent surveillance and tracking of wildfires across Australia is complex and requires a suite of technologies and techniques.
While no single technology or solution can provide us with robust answers to every question, BRIGHT/AHI contributes significantly to this ongoing challenge through the provision of hotspots for Australia via the Geoscience Australia Digital Earth Australia Hotspots website.
The Australian federal government’s Smartsat CRC now funds this research via its Real-time Fire Analytics research programme.
The aim of this project is to develop a comprehensive satellite system that combines geostationary, polar-orbiting, and aerial-based sensors to enable real-time attribution of fires across landscapes. The project will design and implement a data and platform ecosystem that allows for the autonomous detection, processing, and delivery of real-time fire information to end users.
The project comprises two key elements:
- The design and implementation of a data and platform ecosystem that can enable real-time fire surveillance using geostationary, polar-orbiting, and aerial sensors. This “system of systems” approach will ensure that the best available information is always provided.
- The project aims to develop autonomous AI algorithms for real-time fire surveillance and attribution, such as burn severity and FRP.
