Developing low-cost remote sensing methods for multiscale habitat mapping of an intertidal seagrass-macroalgae environment

Abstract

Seagrass ecosystems around the UK are in poor condition and continue to decline, in large part due to anthropogenic activities, such as nutrient pollution, which may also lead to macroalgae proliferation that is detrimental to seagrass growth. To better understand declines and support recovery efforts, accurate spatiotemporal monitoring of seagrass habitat health and macroalgae distributions are required. Remote sensing offers the potential to map large or inaccessible areas, cost-effectively, providing coastal managers with promising data for assessment. This PhD thesis evaluates the potential of using remote sensing technologies to map and monitor a complex intertidal seagrass-macroalgae environment in Lindisfarne, Northumberland, UK. A multiscale mapping approach was used to evaluate multiple platforms and sensors, with differing spatial and spectral resolution. Different classification approaches were tested, the monitoring and management implications of each considered. A Maximum Likelihood classifier and multispectral Unoccupied Aerial Vehicle (UAV) imagery successfully mapped seagrass-macroalgae distribution to species level, with an Overall Accuracy (OA) ranging between 84% and 91%. A random forest classifier with airborne hyperspectral imagery and high resolution PlanetScope satellite imagery was able to produce 6-class large-scale habitat maps with OA of over 90%, for each. This was repeatable across multiple images and may enable monitoring of seasonal and interannual changes in seagrass and macroalgae distribution. The benefit of red edge and near infrared bands was highlighted across multiple platforms. These are offered by the low-cost multispectral UAV that is then able to discriminate between vegetation classes, with similar map accuracies to those achieved when reducing hyperspectral imagery spectral bands (23) to 5-8 bands. Large-scale maps can be used to reveal distribution patterns of seagrass and macroalgae as snapshots and over time, elucidating seagrass-macroalgae dynamics, to support coastal managers’ decisionmaking and management. Overall, this PhD provides a comprehensive critical evaluation of optical remote sensing methods for effective monitoring and its operationalisation for use for seagrass ecosystem conservation.