Understanding fire in the Australian landscape

/

Earth observation for fire risk assessment and monitoring

Hosted by Dr Andrew Edwards, Darwin Centre for Bushfire Research, Charles Darwin University.

This special webinar was developed to explore the very topical issue of fire in the Australian landscape. It looks at how earth observation and remote sensing can be used to understand fire severity and extent, and EO and remote sensing's important role in fire management.

Presenters:

  • Dr Marta Yebra, Australian National University

    • Remote Sensing Contributions to Bushfire Science and Management.

  • Phil McKenna, The University of Queensland

    • Planet, Drone and dNDVI/dNBR Fire Severity Mapping.

  • Dr Leo Hardtke, Joint Remote Sensing Research Program, Qld Department of Environment and Science & The University of Queensland

    • Operational Fire Extent Mapping with Sentinel-2.

Downloads:

The purpose of this session was to inform the community of the datasets available for fire prediction, monitoring and mapping what they are currently used for, and provide an update on the latest research to improve them.

In Australia, remote sensing has been used with other information for some time to map, monitor and model conditions related to fire occurrence:

  • Pre-fire: fuel load and flammability − and from this, determining fire danger

  • During fire − to monitor light, smoke, temperature and determine hotspots and soft containment lines

  • Post-fire − to measure, map and monitor burned leaves, ashes (carbon) and biomass reduction to determine burn: area, emissions, severity and regeneration potential.

Local, State and Commonwealth governments and volunteer groups all have different levels of responsibility for pre-fire management, fire control and post fire clean up and assessment. They all require different information to conduct their work at each of the three stages of fire prediction, monitoring and post-fire management.

A selection of resources for Australian fires (pre, during and post fire) is provided below:

Fire fuel load, curing and flammability assessments:

Bushfire monitoring – near real time:

Current and post fire scar mapping and burn assessment (most states will cover post fire extent and severity mapping):

Additional resources:

Build your own map!

Our three presenters tackled one part each of the three stages of fire mapping:

Pre-fire

Dr Marta Yebra’s work in fuel condition monitoring is helping land and fire managers plan and schedule prescribed burns, and through mapping areas with very low moisture thresholds, understand potential fire flashpoints.

Their research with the bushfire and Naturals CRC is on improving the spatial assessment of fire risk and flammability across the landscape, both strongly determined by the amount of vegetation load, structure and moisture content. Fire services and land managers need this information to undertake hazard reduction burns, to prepare for the fire season, and to anticipate the difficulty of suppressing bushfires throughout the fire season. In this context, Dr Yebra presented the Australian Flammability Monitoring System based on an algorithm to retrieve daily information on vegetation moisture content at a continental scale using the physical phenomenon of radiative transfer (e.g. the propagation of radiation through a medium is affected by absorption, emission, and scattering processes) and remote sensing data. She also presented examples of how fire emergencies services are using LiDAR-derived maps on forest structure and load for prescribed burning planning and planning fire response. Finally, she briefly presented different algorithms to derived fire severity and burn extend. One of them, based on an automated process to infer change detection, was developed in collaboration with Geoscience Australia.

During Fire

Phill McKenna’s work is mapping fire severity using Planet and drone data.

(There is a difference between fire intensity and fire severity. Intensity is the energy released, while severity relates to organic matter loss. Fire intensity is the amount of fuel that has been combusted x rate of spread x a fuel constant – but you need to have a fairly accurate understanding of how fast the fire is moving.)

Phill has assessed biomass, litter fuel loads and fire line intensity at more than 100 transects. His work has shown that:

  • Planet Dove (Classic) data can be used to accurately map fire severity and recovery

  • the scale of the fire determines which sensor is best

  • it’s important to have ground validation field points

  • the process of converting imagery into maps can be automated using R/Python/Google EE script

  • severity rankings are ecosystem specific.

Post-fire

Dr Leo Hardtke has developed and operationalised a method to map fire extent from Sentinel 2 imagery across Qld. This method is also being used in the NT. Leo’s method is novel in that it used the Sentinel fractional cover layer (a physical base - proportion of bare soil, green vegetation and dry vegetation for each pixel and satellite reflectance information). He has developed an algorithm that compares a reference image with the last image that comes in from the satellite (it can also handle Landsat data). It is a three step process that also copes with false positives and produces a discrimination between burned and unburned areas. Validated across 145,000 points/93% precision. This method has been found better than NBR overall, but NBR is still better in tall forested systems. Limitations include when there are long periods without cloud free images, or smoke, which may mean a gap until you get the full extent of a fire.

The Sentinel 2 archive is fully processed – 245 tiles from 2016/04 to 2019/11 – more than 63,700 images. He is trialling mosaics now – which will be manually corrected for known errors soon and then available through TERN and NAFI.

Using the same automated process on fire severity data is a possibility, but good training data is needed.

Postscript:

This material was put together and delivered prior to the now catastrophic spread of Australian fires. It is important to understand that there is considerable work still to be done in this space if we are to continue to predict future events, prepare for them in advance and monitor their spread. Even with the best current resources, high temporal resolution satellite data cannot be collected nationally and at fine enough spatial scale to capture the spread of recent fires in real time.

Multiple industries, communities, research and education facilities are working on this problem.

EOA Inc. was set up as a national coordination entity to help bring them together, advocate impartially for them and provide advice to the Australian Space Agency and other Commonwealth agencies on these issues.