Smartphones and sky scans for better fire mapping

This article first appeared in the Summer 2015/2016 edition of Fire Australia magazine. By Nathan Maddock. 

In the age of smartphones, there is just about an app for everything. And why should land management be any different? Research in the Disaster landscape attribution project at the Bushfire and Natural Hazards CRC has developed an app, currently in beta version, to help land managers quickly and more accurately assess fuel loads before and after prescribed burns.

“We know that when we visually assess (bushfire) fuel, estimates can differ depending on the person who assesses it,” said researcher Dr Karin Reinke from RMIT University.

“We wanted something that was rapid, easy to use and more accurate.”

Dr Reinke is working with a team of researchers at RMIT, co-led by Professor Simon Jones. There are two parts to the study: the first is developing the beta app to assess fuels in the landscape pre and post-fire; the second is investigating active fire mapping and detection using satellites.

Terrestrial LiDAR, which involves lasers and light, has been utilised in the project to map bushfire fuels. While effective, it is an expensive process that requires experts to run the technology. Enter the beta smartphone app; Fuels 3D.

“We want to attribute fire landscapes in terms of the amount, structure and connectivity of the fuel present, and to understand the impact of fire and how fuel characteristics may change, and what remains as residual afterwards,” explained Professor Jones.

“Using a computer vision technique, called Structure From Motion (SFM), a camera phone can be utilised, or any similar commercial-grade camera, to take a series of photos to create a point cloud.”

A point cloud represents the external surface of an object. SFM allows for the extraction of information about the structure, which results in a 3D model of the fuels. A number of different technologies were reviewed by the team to assess which would work best.

“We validated these technologies against dry weight samples of the fuel,” explained Dr Reinke.

 “After we collected the imagery, the fuel was harvested to mineral earth, oven dried and weighed to compare it back to our point cloud estimates from the different technologies evaluated.

“We found good agreement between the terrestrial LiDAR and the Structure From Motion approaches, and good agreement back to the harvested samples.

“But terrestrial laser scanning typically requires experts to implement it, and has much longer data collection and set up times. Structure From Motion data acquisition is really quick,” said Dr Reinke.

The team has developed the beta version of Fuels 3D for Android platforms using SFM technology, through the know-how of researcher Dr Luke Wallace.

The app is being developed in conjunction with a sampling protocol to complement existing fuel hazard assessment practices, allowing land managers to take photos of prescribed burn areas, both before and after a burn, to understand the impact of the burn on fuel loads.

“Fuels 3D should be easier and quicker than visual estimation,” said Dr Reinke.

A workshop was held in Melbourne in early December, where the research team explained Fuels 3D to land managers from Victoria’s Department of Environment, Land, Water and Planning and Country Fire Authority, and South Australia’s Department of Environment, Water and Natural Resources. While Fuels 3D has already been undergoing trials between the research team and end-user partners, large scale validation is needed. It is hoped the app will not only standardise estimation of fuel loads, but make a real difference to how land managers assess the efficiency of prescribed burns.

This would be a big advance, according to Simeon Telfer, an end-user and fire manager at the Department of Environment, Water and Natural Resources SA.

“Not only would it be an advantage for our prescribed burning program, in understanding more accurately how fuel is distributed and the fire severity of our prescribed burns, but it will also help with a greater understanding of what the change in fuel means for a bushfire,” outlined Mr Telfer.

“In the end we are trying to reduce risk through lowering bushfire intensity, but at the moment we are making some estimates about the percentage of fuel burnt based on visual assessments. A greater understanding of what fuel reduction means for risk reduction will be fantastic.

“DEWNR is pretty excited to be involved in this project and have the ability to get some good measures around reduction of fuel loads and seeing how useful the outcomes are,” Mr Telfer said.

A fixed fire daytime experiment

Eyes in the sky

The project is also assessing the role satellites play in mapping bushfires, from their boundaries to their hotspots and flare ups.

“We are looking at the detection of fires as the starting point,” explained Dr Reinke.

“One of the focus satellites is TET-1, a German satellite launched in July 2012, which is the first in the German FireBird constellation.”

Up to seven satellites are expected to complete the constellation in coming years, with the German Aerospace Agency, Deutsches Zentrum für Luft und Raumfahrt, a formal partner in the project.

The other focus satellite is Japanese satellite Himawari-8, launched in October 2015. The Bureau of Meteorology, Geoscience Australia and Australia’s emergency services have access to data from Himawari-8.

TET-1 and Himawari-8 are different types of satellites explains Professor Jones, with each having their pros and cons.

“Himawari-8 is a geostationary satellite that is 36,000km above the earth’s surface,” said Professor Jones.

“Himawari-8 provides new imagery every ten minutes, but at a relatively coarse spatial resolution. It is sitting in the same spot in the sky and continually observing an entire hemisphere of the earth, while a polar orbiting satellite, such as TET-1, is much nearer to the earth, between 600 to 1000km away, resulting in about 14 orbits a day.

“For fire managers, this (TET-1) might mean a new image every couple of days, which is not brilliant for monitoring. But the benefit of the TET-1 satellite is that it provides a lot more spatial detail about the configuration of the fire.”

“The key is to firstly, understand what the utility of each of these platforms are for fire management, and secondly, in the long term, come up with a data assimilation approach where you can actually start to say ‘how can we intelligently use all of this information together with all the other technologies available?’

This study will help fire agencies understand which facets of the available satellite technology are useful for Australian conditions, and which are not.

“It is about determining the confidence we can have in these products, where do they perform and at what point does their performance fall off?” explained Dr Reinke.

To assess this, the research team has designed several experiments. An important aspect of this is the work being done by RMIT and CRC PhD student Bryan Hally. Mr Hally has created a virtual computer environment to simulate fires in different landscapes from which equivalent satellite images based on system characteristics, or even hotspot products, can be generated.

This approach helps to negate some of the logistical and timing constraints posed by working with fire and satellites.

“Being on the ground (at a fire) at the instant the orbiting satellite goes over is really hard to achieve, said Professor Jones.

“Because of that we use the simulated fire landscapes. We can then impose on them different viewing angles, different pixel sizes and different sensor bandwidths to try and understand what a fire looks like for all of the different satellites available for fire observation, and ultimately compare these back to the true, original fire landscapes. We can do this hundreds of times so we achieve a statistically valid representation of performance. This approach can be extended to evaluate future satellite sensors.”

This analytical ability will greatly benefit fire and land management agencies around the country.

“Through our simulations we can begin to understand the potential performance of the different satellites,” explained Dr Reinke.

“For example, we might say ‘you know what, this is going to be great for the Mallee, but the minute we hit the tall closed forests, it is going to have limited value until fires get to a size where sufficient energy comes through the canopy for the satellite to detect accurately’.

“The simulated landscapes and imagery can also be used as a test bed to validate new hotspot algorithms and help improve hotspot algorithms for Australian conditions, because many of the existing algorithms are designed to be global products,” said Dr Reinke.

To allow for empirical observations, the team is conducting field experiments under the satellite’s pass, to determine how accurately they calculate. This is a crucial part of the process in validating the algorithms

“It is building up these case studies that we can use to empirically validate algorithm development,” outlined Dr Reinke.

“It is difficult to obtain comprehensive ground truthing data, unless we can time it around prescribed burns. Even then this is challenging. So in addition to prescribed burns we also create our own fire experiments that are spatially fixed, essentially bonfires of different sizes, intensities and configurations.

“With a fixed fire the logistics are relatively easy, and we will do some more fixed fire testing once we are outside of the fire season. But we also want to apply our validation to prescribed burns. We are hoping to work with CFA in the Victorian Otway’s, DEWNR in South Australia and ACT Parks and Wildlife on this.”