#regional: Eyes in the Sky: DOC Turns to Satellites and AI to Hunt Invasive Weed in Waikato Wetlands

A pest plant threatening some of Waikato’s most important wetlands is now being hunted from space, with the Department of Conservation combining high-resolution satellite imagery and machine learning to locate golden dodder and potentially transform the way invasive species are detected across Aotearoa.

The Department of Conservation — DOC — is trialling the technology at Whangamarino wetland and Lake Whangape, using satellite images capable of detecting patches of Cuscuta campestris, commonly known as golden dodder.

Whangamarino is an internationally significant Ramsar wetland, but its ecosystems face continuing pressure from invasive plants and animals.

DOC Technical Advisor Susan Emmitt says the trial, undertaken with specialist contractor Fabian Döweler of Bushcraft Analytics, has used machine learning to develop a computer model capable of recognising golden dodder within satellite images.

Seeing a pest plant from space

The level of detail available from modern satellites is remarkable.

The high-resolution imagery used for the project provides detail finer than one square metre per pixel, allowing relatively small patches of vegetation to be detected from orbit and their locations identified on the ground.

But the system is doing more than looking for the colour or shape of a plant.

Plants and other objects reflect light differently, creating what scientists describe as a spectral signature.

The satellite imagery captures wavelengths beyond those visible to the human eye, including near-infrared — NIR — light.

That additional information helps computers distinguish golden dodder from surrounding wetland vegetation.

Machine-learning models can then analyse those patterns and classify areas where the invasive plant is likely to be growing.

Six satellite images used to train technology

DOC commissioned six specific satellite images of Whangamarino from different providers to support development of the system.

Part of the project involved determining which spectral characteristics were most useful for detecting golden dodder and establishing the minimum image quality required to reliably identify it.

As more verified examples are added to the model, the machine-learning system can become increasingly effective at identifying the pest across large images.

That could dramatically reduce the amount of time required to survey difficult landscapes.

From walking wetlands to remote sensing

Traditionally, pest plant surveillance can require people to physically inspect large areas.

That can be slow and resource intensive, particularly in wetlands and other environments that are difficult to access.

Satellite technology provides DOC with what is known as remote sensing — collecting environmental information without needing somebody physically present across every part of the landscape.

It also allows conservation teams to repeatedly examine the same areas and compare what is changing over time.

For golden dodder, that means DOC can assess whether existing control programmes are reducing infestations while simultaneously looking for new outbreaks.

Could be scaled across Aotearoa

The potential applications extend considerably beyond Whangamarino.

Döweler says the success of the trial suggests the technology could potentially be scaled up to search for golden dodder across much larger landscapes.

More importantly, the same methodology could eventually be adapted to identify other invasive plant species.

If different pest plants have sufficiently distinctive spectral signatures, machine-learning models could be trained to recognise them too.

That could give conservation teams an increasingly powerful early-warning system.

Rather than waiting for an infestation to become obvious from the ground, potentially suspicious vegetation could be identified remotely and teams directed to precise locations for investigation and control.

Better information for protecting te taiao

For conservation work, accurate information can determine how effectively limited resources are deployed.

Knowing exactly where an invasive plant is spreading allows teams to target control operations rather than searching enormous areas manually.

DOC says satellite analysis can also reduce potential human error associated with manual surveillance and create datasets that can be shared and developed by other organisations.

That opens possibilities for collaboration between DOC, councils, researchers, landowners, iwi and conservation groups.

For wetlands such as Whangamarino, protecting biodiversity also has wider significance for te taiao, wai and the native species dependent on healthy wetland ecosystems.

AI becomes another conservation tool

Artificial intelligence and machine learning are increasingly associated with offices, automation and digital services.

At Whangamarino, however, the technology is being applied to something much more physical — protecting an internationally important wetland.

Satellites provide the eyes.

Near-infrared imagery reveals details humans cannot ordinarily see.

Machine learning searches for the signature of an invasive species.

And conservation teams can use that information to decide where action is needed on the ground.

DOC will now consider how the technology and methodology could support other conservation programmes.

For a pest plant attempting to hide among thousands of hectares of wetland vegetation, that could mean there are increasingly few places left to hide — even from hundreds of kilometres above Earth.

#Whangamarino #LakeWhangape #GoldenDodder #CuscutaCampestris #DOC #DepartmentOfConservation #Conservation #SatelliteTechnology #SatelliteImagery #ArtificialIntelligence #AI #MachineLearning #RemoteSensing #ConservationTechnology #PestPlants #InvasiveSpecies #Biosecurity #Wetlands #Ramsar #Waikato #TeTaiao #Kaitiakitanga #Biodiversity #NativeSpecies #EnvironmentalProtection #EnvironmentalScience #Technology #Innovation #Matauranga #Aotearoa #EnvironmentNews #ScienceNews #MaoriNews #RegionalNews #WaateaNews #RadioWaatea #WaateaDigital