Spatiotemporal modelling and monitoring of harmful algal blooms using IoT in Lake Victoria Basin Kenya

Abstract The escalating global concern of harmful algal blooms (HABs) in aquatic environments, poses significant challenges in the context of sustainable development goals (SDGs). The impacts of HABs extend beyond increased waterborne diseases, causing high aquaculture mortality, diminished aesthetic appeal, elevated drinking water treatment costs, and negative effects on tourism and gross domestic product (GDP). Traditional HAB assessment methods involving field sampling and laboratory analysis are inefficient due to their labor intensiveness and cost. This study introduces a novel approach, combining near real-time satellite remote sensing with a low-cost in-situ internet of things (IoT) system that adds a layer of real-time data collection and timely reporting. The method utilizes the ocean color algorithm and the mono-window lake surface air temperature (LSAT) algorithm, enabling widespread detection and mapping of cyanobacteria-induced HABs with a 30 m spatial resolution landsat imagery. The research involved deployment of immobile in-situ IoT system for continuous near real-time LSAT monitoring at select locations which were as well Kenya marine and fisheries research institute (KMFRI) HAB sampling sites that were prone to early HAB occurrence. From 2015 to 2020, the study used Landsat 8 data to monitor chlorophyll-a (Chl-a) concentrations as proxies for HABs. The Ocean Colour 2 algorithm, combined with landsat 8 thermal infrared (TIR) sensor data, estimated LSAT during blooms. Results are validated against datasets from Copernicus’ Sentinel-3 and national aeronautics and space administration’s (NASA) moderate resolution imaging spectroradiometer (MODIS) missions. Findings showed significant increases in Chl-a values (31 to 57.1 Mg/M 3 ) and LSAT (35.1 to 36.6 °C) during blooms, while unaffected areas had lower Chl-a concentration values ranging from (−1.2 to 16.4 mg/m 3 ) and corresponding lower LSAT values of (16.9 to 28.7 °C). Validation using sentinel-3 ocean and land colour instrument (OLCI) and MODIS data confirmed the algorithm’s reliability with coefficients of determination (R 2 ) ranging from 0.837 to 0.899 for sentinel 3-OLCI and 0.667 to 0.821 for MODIS datasets. An automated IoT powered in-situ system for LSAT monitoring reported abnormally higher average temperature rises (to the above scores, against the normal 25.4 °C), for potential real-time HAB alerts.