The Development of Low-Cost Remote Sensing Technology to Monitor Invasive Alien Plant Species: The Case of Japanese Knotweed

Abstract

Invasive Alien Plant Species (IAPS) are becoming more resilient to global warming. Their effects alter biodiversity causing harm to native organisms and plants. Their early detection followed by a quick response are more helpful and cost-effective than controlling their spread. Freely available satellite remote sensing data such as Landsat-8 and Sentinel-2 can provide multispectral and historical imagery with a spatial resolution between 10 to 30m. However, such low spatial resolution imagery cannot effectively identify and segment the several IAPS like Japanese Knotweed (Fallopia japonica) especially at the beginning of its growth, and on large-scale areas (small areas). Aerial photography, LiDAR and Radar technologies are expensive and also lack the temporal resolution required for regular monitoring of IAPS. It is also difficult to attain the 1 to 2 cm spatial accuracies required for segmenting the different IAPS from their neighbourhood plant species. A review of contemporary research for the monitoring of IAPS shows considerable variation in the application of UAV remote sensing due to their adaptivity and cost-effectiveness for different topographies. Despite this, existing methodologies using UAV do not offer a direct technique for acquiring images in a wide range of RGB and grayscale bandwidths that can attain the spectral propensities for different IAPS identification. Healthy vegetation absorbs most of the visible light (for photosynthesis) and reflects a large portion of the near-infrared (non-visible) light. Consequently, detecting and classifying the different IAPS vegetation requires an interpretation of imagery at other wavelengths. Farmers and landowners also monitor IAPS by foot, but this is time consuming and sometimes some areas are inaccessible or safe to traverse. This research develops a new low-cost and effective technique by integrating different low-cost multispectral sensors onto drones, precise IAPS recognition and delineation irrespective of IAPS size or age, using a wide bandwidth of 440nm – 900nm. Four different low-cost sensors were mounted on the drone and flown at different heights. The derived spectral resolution obtained was 440nm to 900nm while the spatial resolution was less than 1cm. This improved and increased accuracy was able to detect Japanese Knotweed and the other IAPS at their early stage of development using object-oriented classification methods. In conclusion, the greater cost-effectiveness, the high spectral, spatial, and temporal resolution achieved with the developed low-cost UAV platform with the range of sensors tested prove to be more effective to monitor IAPS than the traditional remote sensing technologies such as Satellite, hyperspectral, Aerial photography, and LiDAR.

Date of Award	2024
Original language	English
Supervisor	David Kidner (Supervisor), Iain Shewring (Supervisor) & Nathan Thomas (Supervisor)