Weed remote sensing: a case study in ‘finding a needle in a haystack’ | Mark Hamilton
Tracks
| Thursday, August 7, 2025 |
| 11:00 AM - 11:20 AM |
| Miconia Room |
Speaker
Mr Mark Hamilton
Project Officer (Weeds)
NSW National Parks and Wildlife Service
Weed remote sensing: a case study in ‘finding a needle in a haystack’ | Mark Hamilton
Abstract
Distribution and abundance data for weed species is limited owing to the cost and difficulty in collecting such data. Data paucity can hamper planning, control and monitoring efforts. Remote (aerial) imagery capture, paired with automated detection and mapping, is a burgeoning field in weed management. Aerially mapping weeds often involves a camera, drone and detection algorithm. These technologies are constantly improving, and possibilities are expanding, however they are not currently within the means of all weed management program budgets.
Depending on the weed and its detection difficulty, highly specialised drones and cameras may be needed. More importantly, an effective weed detection algorithm is usually specific only to that species, which is challenging given hundreds of weeds are managed in NSW. Detection can be manual if the weed is highly detectable and the area of interest is small, such as a land manager digitising weed presence on an imagery orthomosaic. More commonly, the weed is not easily distinguishable, and the area of interest is much larger than can be feasibly mapped manually. A detection algorithm can offer immense cost savings, as automated image processing allows rapid mapping of large areas.
Here we present a case study of remote detection of a small, orange-flowered herb in remote Kosciuszko National Park, covering the technology trialled, successes and failures, relative costs and the building of a detection pipeline. Orange hawkweed (Pilosella aurantiaca) is Prohibited Matter and under eradication in NSW. Delimitation of this weed is critical to eradication success, with aerial imagery and automated detection using algorithms being key. Being a well-resourced state eradication target, in attempts to accelerate eradication a suite of novel image capture, sensors, image resolutions and detection methods were trialled over the last 10 years. This case study provides valuable insights and potential cost saving mechanisms to assist those contemplating the use of weed remote sensing technologies.
Depending on the weed and its detection difficulty, highly specialised drones and cameras may be needed. More importantly, an effective weed detection algorithm is usually specific only to that species, which is challenging given hundreds of weeds are managed in NSW. Detection can be manual if the weed is highly detectable and the area of interest is small, such as a land manager digitising weed presence on an imagery orthomosaic. More commonly, the weed is not easily distinguishable, and the area of interest is much larger than can be feasibly mapped manually. A detection algorithm can offer immense cost savings, as automated image processing allows rapid mapping of large areas.
Here we present a case study of remote detection of a small, orange-flowered herb in remote Kosciuszko National Park, covering the technology trialled, successes and failures, relative costs and the building of a detection pipeline. Orange hawkweed (Pilosella aurantiaca) is Prohibited Matter and under eradication in NSW. Delimitation of this weed is critical to eradication success, with aerial imagery and automated detection using algorithms being key. Being a well-resourced state eradication target, in attempts to accelerate eradication a suite of novel image capture, sensors, image resolutions and detection methods were trialled over the last 10 years. This case study provides valuable insights and potential cost saving mechanisms to assist those contemplating the use of weed remote sensing technologies.
Biography
Mark Hamilton is a weed ecologist at the NSW National Parks and Wildlife Service. He has experience in planning, monitoring, implementing, undertaking research, detecting/mapping, data handling and analysis, and evaluating diverse environmental weed management programs. Most recently he has coordinated threatened plant conservation programs where weeds are the major threat, developed aerial remote detection options for several weeds, and prepared weed monitoring protocols.
