Fire spread models are used by Australian fire agencies to forecast the speed, direction and intensity of vegetation fires across the fire season. Whilst there are a number of variables in grassland fire spread models, biomass (fuel weight) and moisture content in the fuel are important factors. The Country Fire Authority of Victoria (CFA) has led research into assessing the annual drying of grasslands, a process known as Curing. Curing data is used for support decision making for Fire Restrictions, Total Fire Bans, fire operations and public warnings. CFA have established a system for monitoring Curing using processed satellite imagery and a network of Grassland Curing Observers (people). CFA recently received a National Emergency Management Project (NEMP) Grant to trial the program throughout south-eastern Australia. Any technical advances to improve the accuracy and resilience of the Curing data program are encouraged by CFA. Two particular challenges have been identified as intrusion of reflectance from woody vegetation and green up of grasslands following summer rains.
At an October 2014 meeting at the University of Melbourne, interested parties from CFA, CSIRO and RISER discussed opportunities for research into grassland fire model inputs using a range of sensing devices including existing static ground based sensors and un-manned aerial vehicles (UAV). A project to capture a multi-date colour, infra-red and Normalised Difference Vegetation Index (NDVI) image collection from a UAV platform in conjunction with ground observations was agreed.
From a potential 10-12 sites where grassland curing observers were already in place, a site at Moorooduc, on the Mornington Peninsula, was selected. The Moorooduc site, approximating 400m × 400m, has a mixture of pastures, crops and land uses. Given that the spatial resolution of the UAV imagery is 0.35cm, there is plenty of variety in the data and imagery for researchers to investigate.
By April, 2015, five image collections have been compiled along with accompanying field observations which documented changing pasture onditions and land use interventions (e.g. cutting for hay). Figures 1 and 2 shows the optical and NDVI images of the Moorooduc site, respectively, taken on 28th December, 2014. Geo-referenced photo’s were taken to match each field observation (see Figure 3).
As part of the Curing research program, CFA and CSIRO have established several ground-based fixed sensors in a paddock at Scoresby, in Melbourne’s eastern suburbs. CFA and RISER expanded the UAV program and collected four image data sets. This data is supplemented with some physical sampling of grasses. A data library was created and includes raw UAV imagery, processed vegetation indices, site photo’s, ground-based sensor data, and human observations in an easily digestible geo-database for CFA and CSIRO researchers.
CFA is extending their grassland curing estimation system to interstate, where there are some issues with lightly treed grassland landscape (right). The proportion of exposed soil is high in lightly treed areas. Satellite based grass curing monitoring methods tend to underestimate re-growth after summer rainfall events and fail to detect late season green up. Hence the RISER team and CFA started the second phase of the UpRISER sub-project to investigate if UAV imagery can provide more accurate grassland curing estimations. A second series of images and field data were collected in the summer of 2015-2016, this time in the Mallee region of Victoria.
Mallee has a lightly treed grassland landscape. Hence the data enable us to understand the performance of UAV imagery in the future interstate sites. The figures below illustrate one of the optical images taken in Mallee and the NDVI in the same area.
Applications of the UAV imageries
The high-resoluation UAV imageries reveal considerable spatial details of the distribution of grassland curing. We will analyze the spatial characteristics of the data to understand the finer-grained changes of grassland curing. The time series of the imageries is valuable to investigate how grassland curing rate varies spatially. We may be able to find out reasonable interpretations of the variation from the correlation between the variation and other spatial information. Finally, it is important to understand how the finer-grained variation of grassland curing will affect the bushfire simulation models. These discoveries will be combined to evaluate whether UAV imagery can provide useful information about grassland curing for bushfire management, which cannot be obtained from satellite imagery or ground observers.
- Mark Garvey. “Using next generation spatial technologies to advance knowledge of grassland curing”, Locate16