Eradicating the red imported fire ant with remote sensing

Twitter
Posted on August 8, 2013

Recently we reported on the issue of red imported fire ants in Brisbane – a pernicious pest, first detected in Queensland in 2001, that affects agricultural crops, native species and human health.

Although the number of fire ant nests in south east Queensland is decreasing, their geographic range has increased.

Neither we nor Biosecurity Queensland regard the battle to eradicate fire ants as having been lost. On the contrary, in this article, we discuss what may be the best hope of eradicating the fire ant: a technology known as remote sensing surveillance.

Meeting the challenge

Aerial survey camera. © The State of Queensland (Department of Agriculture, Fisheries and Forestry) 2010–2013

The infestation is now large (by our estimates, in December 2011 a bounding rectangle had an area of 342,000ha) and there is a crucial need to survey the perimeter of the known infested area to detect and destroy any new nests thereby stopping further spread.

Spotting nests from the air may be the best hope of eradicating fire ants. © The State of Queensland (Department of Agriculture, Fisheries and Forestry) 2010–2013

Spotting nests from the air may be the best hope of eradicating fire ants. © The State of Queensland (Department of Agriculture, Fisheries and Forestry) 2010–2013

Ground-based surveillance is costly and labour intensive. The best hope therefore lies, we believe, in new remote-sensing technologies using airborne cameras, which cover large areas in a fraction of the time taken by conventional surveillance methods.

The cameras are mounted beneath a helicopter, which flies over the target area at a height of 150m. Images are captured in three separate frequency ranges: visible, near infrared and thermal (see images below). These are then processed in parallel to identify objects that may be fire ant nests.

Nests have a distinctive size and shape, but more importantly, in suitable weather conditions are about 10C hotter than the surrounding ground. They’re not called fire ants for nothing!

Image analysis is automated using clever detection algorithms developed at the University of Sydney. This is an essential part of the process because the amount of imagery captured is vast.

Remote sensing image in the visible range (fire ant mounds detected are circled). © The State of Queensland (Department of Agriculture, Fisheries and Forestry) 2010–2013
Remote sensing image in the near infrared range. © The State of Queensland (Department of Agriculture, Fisheries and Forestry) 2010–2013
Remote sensing image in the thermal range. © The State of Queensland (Department of Agriculture, Fisheries and Forestry) 2010–2013

The strategy

The remote-sensing technology makes possible a new strategy for eradicating the ant, which begins with determining the extent of the infestation. Ground-based surveillance is relatively expensive and would take too long to survey the area required to fully delimit the invasion.

Under the original strategy, a core area was treated and a 5km buffer around this was surveyed. Engagement of the public and industry was used to detect infestations beyond this.

However, in the mid-2000s there was undetected movement of fire ants to a landfill west of the known infested area and beyond the operational area. Our reconstruction of the invasion history suggests this occurred before 2004.

This infestation was not reported until 2007. In that time it spread from this new epicentre and this accounts for much of the east-west expansion that has occurred since fire ants were first detected in 2001.

Estimates of nest density 2000-2011. The most dense areas are red and the least dense yellow. The black arrow in the 2004 map indicates the beginnings of the uncontrolled expansion centred on Amberley. Credit: Jonathan Keith

Remote sensing allows a much larger area to be searched each year, making it feasible to patrol large boundary regions outside the known infested region, and thus work from the outside in, progressively shrinking the infestation.

The estimated cost for remote sensing surveillance is just over A$72 per hectare. This will save over A$24M per year compared to ground based surveillance.

Chance of success

With remote sensing it is possible to capture up to 750ha of imagery per day per camera. However, the technology requires fine weather to be effective and can only be used during the cooler months, when soil temperatures are low enough to provide contrast with the higher temperatures generated by nests.

Based on Bureau of Meteorology records, the necessary weather conditions occur about 100 days a year. Operational use of remote sensing began in May 2012. There are two cameras currently in operation, surveying around 1,500ha per day. The program is currently aiming to search a maximum of 100,000ha per year.

The estimated footprint of the infested area is less than 350,000ha. Remote sensing is not intended to monitor the entire infested region, but is prioritised at the edges of the infestation where expansion could potentially occur.

Fire Ant Operational Map, showing zones in which various levels of monitoring and intervention are in place. © The State of Queensland (Department of Agriculture, Fisheries and Forestry) 2010–2013

The Program has used our model’s estimates of locations possibly infested with fire ants in combination with knowledge of known infested regions to select areas that will be surveyed with remote sensing.

The Program has now captured imagery of more than 85,000ha of the outer edges of the suspected area of infestation. Field staff conducting follow-up surveillance on points identified by the detection algorithm found two fire ant mounds within this area to date. Given the high sensitivity of remote sensing, this supports the Program’s belief that there is very little fire ant activity in these areas. This surveillance work is continuing.

In addition, the public have also reported the presence of several fire ant mounds after receiving notification that remote sensing surveillance was occurring in their area, but before imagery could be analysed.

This notification prompted these residents to check their yards and report back to the Department of Agriculture, Fisheries and Forestry (DAFF). The detection algorithm found a number of these mounds during subsequent analysis.

The effectiveness of remote sensing technology depends on two critical numbers: the sensitivity and the false positive rate.

Red imported fire ants. Credit: Wikimedia Commons

Sensitivity is the proportion of nests in searched areas that are detected. A false positive is an object that is not a fire ant nest but is identified as one by the detection algorithm, and the false positive rate is the number of such objects detected per unit area.

Sensitivity is important because if too many nests go undetected, the infested area will continue to expand. The false positive rate is important because each detection must be followed up, either to investigate whether the object really is a fire ant nest or to treat it. Too high a false positive rate will ineffectively use program resources.

In trial work conducted in 2012, the algorithm had an overall detection rate of 75% and the current false positive rate is less than 5 points per hectare. The Program is continuing its assessment of the sensitivity and is also working on reducing the false positive rate through continual refining of the detection algorithms.

The future

In ongoing work, we are conducting computer simulations of remote sensing in an effort to determine the ranges of sensitivity, false positive rate, and other parameters, that will lead to successful eradication. We hope to work with Biosecurity Queensland and the detection algorithm developers in a process of continual testing and improvement to ensure the parameters needed for success are achieved.

The program has a goal to delimit the infested area by mid 2015. The national program will then make a decision on whether to keep aiming for full eradication, which we believe could be done by 2022.

The success of remote sensing surveillance will be critical to achieving this target.

Source: The Conversation, story by Jonathan Keith, Craig Jennings and Daniel Spring