In Situ Monitoring and Predictive Modeling of Harmful Algal Blooms with Autonomous Surface Vehicles

Nenyi Kweku Nkensen Dadson, Jean Paul Gonzalez, Daniel P. Arabie, Ryan J. Cedel, Dorin Boldor, Corina BarbalataarXiv 2026 · arXiv

TL;DR

Researchers have developed an autonomous surface vessel (ASV) system designed to tackle the "spatio-temporal headache" of monitoring Harmful Algal Blooms (HABs). By integrating a custom motorized winch for multi-depth sampling with a Random Forest machine learning model, the system can identify bloom conditions with 99.6% precision, offering a scalable and repeatable alternative to manual water quality testing.

What Problem Does This Paper Address?

Harmful Algal Blooms (HABs)—often caused by cyanobacteria—are a major environmental threat, capable of producing toxins that degrade water quality and harm public health. The challenge lies in their dynamic nature: they shift rapidly across large water bodies and vary significantly by depth.

Traditional monitoring relies on manual, point-based sampling, which is too sparse to capture fine-scale variability. While Unmanned Aerial Vehicles (UAVs) offer better spatial coverage, they suffer from short battery lives and cannot measure parameters below the water's surface. This paper introduces an autonomous robotic solution capable of high-resolution, multi-depth, and multi-modal data collection to bridge this gap.

Methods

The authors built their system on the BlueBoat ASV platform, a modular catamaran designed for environmental sensing.

Fig. 2: (a) Baseline BlueBoat ASV platform prior to payload integration (Image credit: Blue Robotics) and (b) after installation of the integrated sensing and actuation subsystems

Hardware Engineering

The core innovation is a custom-designed motorized winch system. To handle the weight and drag of a multi-parameter sonde at depth, the team integrated a NEMA 23 stepper motor with a 5:1 gearbox and a boost converter to increase the supply voltage to 35V, ensuring sufficient torque for reliable deployment and retrieval.

Fig. 3: Integrated winch system installed on the ASV, showing the (a) mechanical assembly and (b) associated electronics enclosure

Sensing Suite

The ASV carries a sophisticated payload:

  • Aqua TROLL 700 Sonde: Measures chlorophyll-a, phycocyanin (BGA-PC), nitrate, turbidity, and temperature.
  • Side-scan Sonar: Provides acoustic imagery of the lake bed texture.
  • Event Camera: Powered by an NVIDIA Jetson Orin Nano to record visual changes on the water surface.

Predictive Modeling

The researchers employed a two-layer machine learning approach. First, a Random Forest model predicts chlorophyll-a concentrations based on environmental variables (temperature, turbidity, specific conductivity, etc.). Second, a binary classification layer uses a threshold of 10 µg/L to label the site as "HAB" or "non-HAB."

Datasets

The system was validated through field deployments in two distinct Louisiana water bodies to ensure environmental diversity:

  1. Lake Pontchartrain: A major estuary prone to HABs due to nutrient influx from the Mississippi River.
  2. False River: An oxbow lake used to expand the spatial diversity of the training data.

The final dataset consisted of 323 data points, which were pre-processed using a Pearson correlation matrix to remove redundant features before training.

Key Insights and Ideas

  • Verticality is Vital: HABs are not just a surface phenomenon. The ability to autonomously sample at multiple depths (0.5m, 1m, and 2m) allows for a 3D characterization of the water column that surface-only sensors miss.
  • Robotic Repeatability: By using ArduRover for autonomous waypoint navigation, the ASV can revisit the exact same coordinates across multiple days or weeks, providing a consistent temporal baseline for bloom development.
  • Multi-modal Context: While water chemistry is the primary indicator, the inclusion of sonar and event cameras provides "contextual environmental information"—such as lake bed texture or surface scum—that could eventually be fused into more robust predictive models.

Key Results

The system demonstrated high reliability in both data collection and prediction:

  • Chlorophyll-a Prediction: The Random Forest model achieved a prediction accuracy of 90% with an $R^2$ value of 0.9952.
  • HAB Classification: The system reached an overall precision of 99.6% in identifying bloom conditions.
  • Operational Efficiency: The ASV successfully executed "lawnmower" coverage patterns, autonomously loitering at waypoints to perform depth-controlled sampling without manual intervention.

Limitations and Future Work

While the results are promising, the authors note several areas for expansion:

  • Dataset Scaling: The current model was trained on 323 points. Larger, more diverse datasets from different climate zones would improve the model's generalizability.
  • Sensor Fusion: Currently, sonar and event camera data provide "context." Future iterations could integrate these data streams directly into the machine learning pipeline to see if visual or acoustic features improve early-warning detection.
  • Long-term Autonomy: Further testing is needed to evaluate the system's performance during multi-day deployments in harsher weather conditions.

Summary updated

Abstract

Harmful Algal Blooms (HABs) pose significant environmental, economic, and public-health challenges in coastal and inland water bodies, creating a need for scalable methods to collect spatially distributed water-quality data. This paper presents an autonomous surface vessel-based sensing system for in-situ water-quality monitoring related to harmful algal bloom detection. The system integrates a BlueBoat Autonomous Surface Vessel (ASV), a motorized winch for depth-controlled sonde deployment, a multiparameter water-quality sonde, side-scan sonar, and an event-camera payload. Autonomous waypoint missions were used to collect measurements across selected sampling regions and multiple depths, enabling both horizontal and vertical characterization of environmental conditions. Field deployments were conducted in Lake Pontchartrain, Louisiana, with additional measurements collected in False River, Louisiana to increase dataset diversity. The collected measurements included algal pigment indicators, nutrient concentration, turbidity, temperature, and other water-quality parameters relevant to bloom formation. In addition, complementary sonar and event-camera data were collected during vehicle operation to provide contextual environmental information. A machine-learning classification approach was also evaluated using the collected water-quality measurements to support HAB and non-HAB classification. The results demonstrate the feasibility of using a modular ASV platform for repeatable, multi-modal water-quality data collection and support future development of autonomous monitoring systems.
BibTeX
@misc{dadson2026situ,
  author = {Nenyi Kweku Nkensen Dadson and Jean Paul Gonzalez and Daniel P. Arabie and Ryan J. Cedel and Dorin Boldor and Corina Barbalata},
  title = {{In Situ Monitoring and Predictive Modeling of Harmful Algal Blooms with Autonomous Surface Vehicles}},
  year = {2026},
  eprint = {2611.18501},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2611.18501},
}
Figure