Algal Bloom Detection and Water Monitoring Using IoT Connectivity

This paper presents an internet-of-things system for algal bloom detection. It relies on microcontrollers, low-cost sensors, the cloud, machine learning, and domain knowledge. An Arduino MKR WiFi 1010 measures pH, temperature, turbidity, and phosphorus (via an RGB sensor and reagent), and sends telemetry over MQTT to a ThingsBoard dashboard. The objective of this work is to detect the presence of harmful algal blooms using guideline-based thresholds, the purpose of which is to support environmental water-monitoring in lakes and rivers, using low-cost systems. In this paper, multiple machine learning algorithms and processes were used to analyze multiple water-quality markers, predict the presence of a harmful algal bloom, and display warning information and diagnostics in a cloud-based dashboard. Because no labelled bloom data were available, chlorophyll-a was used as a surrogate training target, removing the need for a costly chlorophyll sensor in field deployment. The random forest and gradient boosting models trained in this fashion predicted all three risk levels, with random forest reaching 78% accuracy and a macro F1-score of 0.68. This proof of concept is evidence that local water quality monitoring can be completed with low-cost hardware. Overall, this work highlights the absence of, and subsequent need for, low-cost water sampling and data collection in the freshwater bodies of British Columbia (BC).