Modelling Australia's Fire Seasonality

There are now enough years of high quality MODerate-resolution Imaging Spectroradiometer (MODIS) hotspot data to attempt a review Australia's fire seasons. MODIS is carried on both the Aqua and Terra spacecraft. Standard algorithms assess the spectral signature of each pixel and the pixel's neighbourhood to tag some as indicating fire. The databases of these are now readily available for analysis. Hotspots are point objects with some positional uncertainty due to viewing angle, terrain and orbital instability. They are fully attributed with data about collection time and assessed intensity. Aggregation of these for the purposes of identifying patterns is not straightforward. Australia's fire industry still largely relies on fire-season maps produced decades ago, long before modern remote sensing technology. Changes in land use, social structures and even climate can be expected to have potentially altered the seasonality pattern. A seasonality pattern derived from hotspot data was produced on the following basis. Firstly, month was the temporal unit for aggregation. Secondly, all fire hotspots were used, including both wildfire and higher intensity fuel reduction burns. While this produces issues, it is suitable for a range of key applications. Thirdly, aggregation was done by means of a 1 degrees grid. Finally, all hotspots from July 2002 to June 2013 were included. Spatial software was used to produce, for each grid cell, a "wind rose" type diagram, with twelve spokes radiating out from a central core. The length of each spoke was proportional to the relative frequency of hotspots in that grid-cell, with months arranged like on a clock-face with January being month one. It was found that to a large degree there were extensive, coherent groups of these roses, referred to as zones. In all 29 zones have been identified. Many have a clear unimodal distribution, while some are clearly bimodal. Six were classified more by a lack of a clear modality, in contrast to their neighbouring zones. These were termed "aseasonal". The results were also used to produce a national hotspot frequency map. This identified some areas where the existing climatology may be insufficient to have confidence of a stable zonation. Dynamically extending the climatology may resolve this in future years. It was found that the seasonality is very different from that current in use. This may be due to the conflation of deliberate and wildfire hotspots. It is not currently practical to separate all wildfire hotspots, but those due to major wildfires will be identified as part of on-going research.