Scientists have used satellite observations and aerial drone imagery to map the spread of an aggressive non-native weed across coastal dune landscapes, providing essential information that could help reduce the impact of this damaging invasive species on local ecosystems.

The study – published in the IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing – drew on multispectral Earth observation data from Vision-1 and PlanetScope, both of which are part of ESA’s Third Party Missions programme. It also used images from Copernicus Sentinel-2.

Sand dunes in coastal areas are unique habitats that support complex ecosystems of specialised wildlife, including rare birds, insects, and flowers. They also serve as natural barriers that protect inland settlements from flooding and storms.

The environment of the dunes on the coastal plateau of Israel is characterised by the combination of a Mediterranean climate and dry sandy soils, which allows a mix of desert plants and Mediterranean species to exist together in the same area.

However, this habitat is under threat from an invasive plant species, named Heterotheca subaxillaris and commonly known as camphorweed, that can easily outcompete native flora and disrupt local ecological dynamics.

Camphorweed was introduced intentionally in the 1970s from North America to stabilise the dunes and prevent sand moving into roads or fields of crops. It has now spread far beyond its original planting locations to become the dominant plant in large parts of the region.

As part of the study, high resolution drone images were collected over two sites on the Mediterranean coast. A machine learning model was then developed to detect camphorweed using tell-tale visual signatures, such as the appearance of inflorescent yellow petals during the flowering period. The method was reported to identify the plant with an accuracy of 97%.

After the effectiveness of this approach was demonstrated, the researchers used it as a training and validation reference for the mapping of camphorweed from satellite imagery.

In this stage of the analysis, the team drew on very high resolution Vision-1 data, disseminated via ESA’s Third Party Missions programme, and high resolution PlanetScope data, which were provided directly by Planet Labs. A separate machine learning algorithm was developed to identify visual indices for the detection of camphorweed from these data.

Through a comparison of outputs from drone imagery, it was found that Vision-1 was able to correctly identify camphorweed with 96% accuracy, with the lower resolution PlanetScope mission achieving an accuracy of 83%.

In the final step of the project, the outcomes from the Vision-1 analysis were employed to train a probabilistic machine learning model using medium resolution Sentinel-2 data. This was found to have a detection accuracy of 73%.

The results of this study demonstrate how drone imagery can be used as a reference point to map the spread of camphorweed across multiple scales using satellites. This approach has the benefit of capitalising on the respective attributes of several different Earth observation missions.

Vision-1, for instance, was able to provide highly precise detections over small scales due to its very high-resolution observations. Sentinel-2, on the other hand, delivered less precise estimates, but was able to cover broader areas at a high temporal resolution, thanks to the mission’s relatively fast revisit rate.

If it were made available, such multi-scale data could support the early detection and mapping camphorweed, enabling land management teams to plan interventions designed to protect local plant and animal life from this invasive species.

Sharad Kumar Gupta, remote sensing researcher and lead author of the study, concluded, “This study highlights the significance of combining multiple remote sensing techniques to address environmental challenges and calls on users to leverage ESA's capabilities to support such initiatives. From drones to satellites, the technology is turning the tide against invasive species. With early detection and smart mapping, we can protect nature before it’s too late. A satellite sees what the eye misses—now it’s time to act!”

The study was carried out by Sharad Kumar Gupta, along with co-authors Prof. Eyal Ben-Dor and Prof. Marcelo Sternberg of Tel Aviv University, Israel. It was backed by the Israel Ministry of Science and Technology in cooperation with the Italian Ministry of Foreign Affairs and International Cooperation, as part of a dedicated research project.

References:

S. K. Gupta, E. Ben-dor and M. Sternberg, "Unveiling the Invasion: Advancing Ecological Mapping of Heterotheca Subaxillaris Through Integrated Remote Sensing Techniques with Drones and Satellites," in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 17, pp. 7193-7211, 2024, doi: 10.1109/JSTARS.2024.3374232.