EProACH: ML Eutrophication Prognosticator for Prediction of Complex HAB Patterns

Overview

Water shortages are a global issue, now impacting North America, where cyanobacterial HABs pose the greatest threat to inland water quality. Each year the US alone incurs over $50 million in loss due to HAB clean-up, drinking water supply, tourism, and fisheries closures. While few machine learning (ML) models exist, they capture and predict cyanobacteria proliferation locally using a binary classification system. The study presents EProACH, a robust model, that quantifies HAB proliferation on a continuum through a single data point input and acts as a standardized tool for Early Warning Systems across the US. Over 5 months, 240 water samples (500 ml each) collected from 30 sites across NC, VA, and MD were measured for algal weight, NO3, PO4, and temperature. Using Jupyter Labs, Pandas, and SciKit Learn, ML classification and regression models were built. Libraries of ML were used to identify the most accurate ML, with R2 > 0.75. The coefficients were consistent across states and could be combined into a US model. Coefficients from the regression model were then used to build a Monte Carlo (MC) simulation that could run over 100,000 scenarios to accurately predict the impact of a percentage change of any of the variables on HAB proliferation. EProACH demonstrated precision in forecasting HAB occurrences on specific dates, as corroborated by USGS and NASA, affirming its effectiveness. The results showed that EProACH, an ML Regression model with MC simulations, can be readily used for applying USGS and NASA Data to quantify and predict HAB risk on a continuum, making early mitigation efforts across the US possible.

Competition history

• AJAS 2025 Category not listed

Related projects

Closest projects by meaning, across every fair and year in the corpus.

Browse more like this

• All Earth and Environmental Sciences projects
• AJAS Earth and Environmental Sciences projects
• Every AJAS 2025 project