Improving the spatial and temporal monitoring of cyanotoxins in Iowa lakes using a multiscale and multi-modal monitoring approach.

Cyanobacterial harmful algal blooms (CyanoHABs) are pervasive and negatively impact lake water quality, resulting in economic losses and public health risks through exposure to cyanotoxins. Therefore, it is critical to better monitor and understand the complexity of CyanoHABs, but current methods do not fully describe the spatial and temporal variability of bloom events. In this work, we developed a framework for a multiscale and multi-modal monitoring approach for CyanoHABs combining drone-based near-range remote sensing with analytical measurements of microcystin cyanotoxins and chlorophyll-a. We analyzed weekly beach monitoring samples from 37 lakes geographically distributed across the state of Iowa (USA) over a 15-week period in the summer of 2019 to quantify ELISA (bioassay), 12 microcystin congeners (LC-MS/MS), and chlorophyll-a. We developed a novel microcystin congener-normalized equivalent toxin metric to compare CyanoHAB impacted waters; this microcystin-LR normalized sum-of-congeners approach yields lower predicted toxicity than parallel ELISA results suggesting ELISA is conservative for assessment. A significant linear relationship existed between chlorophyll-a and microcystin for lakes throughout Iowa (R = 0.39, p < 0.001); lakes with low watershed:lake area ratio and long residence times exhibited a stronger correlation. We then developed a novel geometry-based image processing approach to allow for stitching over-water drone images, a previous barrier in photogrammetry. We applied our mutli-modal framework to a case study on Green Valley Lake to assess initial viability and predicted microcystin concentrations within 33%. We concluded that multispectral imaging is possible but may presently be insufficient for predicting microcystin concentrations due to limitations in the spectral capabilities of the multispectral camera, but technologies are quickly advancing, and lightweight hyperspectral imaging could soon become feasible for investigating spatial bloom variability on lakes.