The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences
Publications Copernicus
Articles | Volume XLI-B8
22 Jun 2016
 | 22 Jun 2016


B. Hally, L. Wallace, K. Reinke, and S. Jones

Keywords: Himawari, fire detection, thermal imaging, temporal analysis, Kalman filtering

Abstract. Wildfire detection and attribution is an issue of importance due to the socio-economic impact of fires in Australia. Early detection of fires allows emergency response agencies to make informed decisions in order to minimise loss of life and protect strategic resources in threatened areas. Until recently, the ability of land management authorities to accurately assess fire through satellite observations of Australia was limited to those made by polar orbiting satellites. The launch of the Japan Meteorological Agency (JMA) Himawari-8 satellite, with the 16-band Advanced Himawari Imager (AHI-8) onboard, in October 2014 presents a significant opportunity to improve the timeliness of satellite fire detection across Australia. The near real-time availability of images, at a ten minute frequency, may also provide contextual information (background temperature) leading to improvements in the assessment of fire characteristics. This paper investigates the application of the high frequency observation data supplied by this sensor for fire detection and attribution. As AHI-8 is a new sensor we have performed an analysis of the noise characteristics of the two spectral bands used for fire attribution across various land use types which occur in Australia. Using this information we have adapted existing algorithms, based upon least squares error minimisation and Kalman filtering, which utilise high frequency observations of surface temperature to detect and attribute fire. The fire detection and attribution information provided by these algorithms is then compared to existing satellite based fire products as well as in-situ information provided by land management agencies. These comparisons were made Australia-wide for an entire fire season - including many significant fire events (wildfires and prescribed burns). Preliminary detection results suggest that these methods for fire detection perform comparably to existing fire products and fire incident reporting from relevant fire authorities but with the advantage of being near-real time. Issues remain for detection due to cloud and smoke obscuration, along with validation of the attribution of fire characteristics using these algorithms.