Drone technologies help combat invasive species for coastal Texas conservation program
At a glance
GHD Digital helped the Coastal Bend Bays & Estuaries Program (CBBEP), a non-profit conservation organisation, enhance data collection and identify invasive plant species’ location and distribution. Using drone technology and image classification methods, we determined the location and area calculation of three invasive plants - salt cedar (Tamarix ramosissima), Johnson grass (Phyllostachysaurea), and lead tree (Leucaena leucocephala) for the Sunset Lake Park and Violet Andrews Park invasive species eradication plan.
The challenge
The CBBEP is dedicated to protecting and restoring the health and productivity of the bays and estuaries in the Texas Coastal Bend, while supporting economic growth and public use of the bays. With the help of numerous partners, the CBBEP has restored thousands of acres of marsh habitat, funded dozens of projects designed to improve water quality, and installed infrastructure to enhance public access opportunities. Recently, CBBEP partnered with the City of Portland to identify invasive plant species at Sunset Lake and Violet Andrews parks and develop plans to eradicate the invasive plants.
In the last twenty years the City has experienced a significant increase in invasive species, leading to challenges with the growth of salt cedar, Johnson grass, lead tree, and other invasive plants in the Sunset Lake and Violet Andrews parks. The city and CBBEP needed to quantify invasive plants so they could better develop strategies for eradication.
Failure to eradicate the invasive species would prevent the native plants from growing and would negatively impact the natural avian habitats of wildlife.
Our response
GHD Digital’s drones and geomatics team provided CBBEP with a targeted approach to better understand the location and distribution of invasive species at Sunset Lake, a 450-acre coastal wetland area and public park. Digital solutions would expediate their data collection and enable them to quantify the area and distribution of each invasive species while significantly reducing the field time of staff biologists.
Using Geographic Information Systems (GIS), digital technologies and QGIS software, we provided the necessary data to streamline decision-making and enabled the mapping of invasive species.
Our certified drone operator flew a multi-rotor unmanned aircraft over the site and captured thousands of aerial images. Additionally, eight ground control points (GCP) were surveyed using ground positioning systems to obtain accurate coordinates and elevations of known locations. The imagery and GCPs were imported into software to generate one high-resolution, georeferenced image of the site – a process called photogrammetry.
CBBEP staff biologists collected data at 20 locations at Sunset Lake where there were known invasive plant species. These data points were used to locate the plants in the final drone imagery. Using machine learning tools, we used the 20 field data points – as well as over 600 points from the imagery - to train a model to detect and delineate each of the plant species based on specific parameters, including color and growth patterns. Water, ground, and native plant species were also classified as part of the model training.
A raster (i.e. a classified image that they could put into their GIS applications), and a Google Earth map was created delineated by colour where the invasive species were located and total acreage of each. We identified that salt cedar made up 39.5 acres, lead tree made up 9.29 acres and Johnson grass made up 2.87 acres of the park.
The impact
CBBEP revisited the site to verify the data’s accuracy, and the results were impressive for salt cedar and lead tree, with accuracy rates of 86 percent and 83 percent, respectively. The accuracy for Johnson grass was significantly lower at only 16 percent, due to its indistinct physical characteristics. To enhance the mapping accuracy for Johnson grass during the next round phase, we would incorporate additional sensors such as multispectral or hyperspectral.
The use of drone technologies and machine learning to codify each invasive species provided a roadmap for the CBBEP with the data needed to calculate the total acreage of affected areas requiring remediation efforts. This provided them with accurate information to use in their submission to vegetation management subcontractors and to receive project funding.
Traditionally, data collection would be conducted by biologists manually collecting data in the field, a very time-intensive process that is very difficult to implement over large areas. Leveraging digital technologies enabled data collection in just one day. The reduction in field time has allowed the staff to focus on other important projects and reduced the risks associated with field surveys. The data also provides the vegetation management subcontractors with valuable intelligence for planning and executing the eradication efforts.
This methodology has the potential to transform the way we detect and map invasive plant species – which is vital to reducing the impacts of invasives to native plant species and overall biodiversity.
Connect with us to learn more about leveraging drone technologies to streamline data collection and improve environmental outcomes.