Cyanobacteria Hot Spot Detection Integrating Remote Sensing Data with Convolutional and Kolmogorov-Arnold Networks

BrianZambrano

University of Alberta

"Cyanobacteria Hot Spot Detection Integrating Remote Sensing Data with Convolutional and Kolmogorov-Arnold Networks"

Monitoring cyanobacterial blooms promptly and accurately is crucial for public health management and understanding aquatic ecosystem dynamics. Remote sensing, particularly satellite observations, offers a viable approach for continuous monitoring. This study utilizes multispectral images from the Sentinel-2 satellite constellation in conjunction with ERA5-Land data to facilitate broad-scale data collection. We proposed a simple deep convolutional neural network (CNN) architecture to analyze cyanobacteria (CB) concentration dynamics in Pigeon Lake, Canada, over a five-year period. Utilizing the Local Getis-Ord statistic, we identified and analyzed trends in hot and cold spots under the null hypothesis of random distribution. We observed changes in the distribution and median CB concentration in hot spots over time. Additionally, we trained a Kolmogorov-Arnold Network (KAN) to classify segments of the lake shoreline into hot and non-hot spots using the Dynamic World dataset within a 500-meter radius of the lake.
Additional authors:

---