Improving spatial and temporal monitoring of harmful algal blooms in aquatic ecosystems using unpiloted aerial vehicle based remote sensing

Cyanobacterial Harmful Algal Blooms (CyanoHABs) are increasing in frequency and intensity in freshwater systems worldwide, particularly in Midwest lakes, where agricultural nutrient runoff is high. Cyanobacteria can produce toxins, posing environmental, public health, and economic threats. Multiple monitoring strategies have been developed, but they fail to fully describe water quality variations in space and time. There is a critical need for the advancement of monitoring techniques that can accurately measure the extent and fluctuations of CyanoHABs. Cameras mounted on small unpiloted drones offer low costs, flexible deployment schedules, and the ability to capture highly detailed images of water surfaces. Drone cameras often need to collect many overlapping images during monitoring, which are then stitched together to generate spatial maps displaying the extent and intensity of CyanoHABs. However, current geographic image-processing software is unable to stitch over-water drone imagery due to highly variable surface reflections. This work presents a novel geometric approach for aligning and stitching over-water images. Additionally, methods were developed to improve data quality, such as the removal of reflections on the water surface. Grab samples and aerial images were used to demonstrate the utility of a drone-based remote sensing approach for monitoring chlorophyll concentrations, a common indicator of CyanoHAB biomass, in Lake Darling, Iowa.