Know Before You Go: Data-Driven Beach Water Quality Forecasting.

Area of Science:

• Environmental Science
• Data Science
• Public Health

Background:

• Forecasting environmental hazards is crucial for community and ecosystem resilience.
• Environmental data science offers potential for hazard forecasting but requires advanced methodologies for complex datasets.
• Current methods for predicting marine beach water quality are limited in their predictive lead time.

Purpose of the Study:

• To develop and validate a data-driven framework for forecasting bacterial standard exceedances at marine beaches.
• To assess the predictive performance of statistical and machine learning models for water quality forecasting.
• To provide decision support for proactive public health and pollution risk management.

Main Methods:

• Utilized historical water quality data from two California marine beach sites.
• Trained nearly 400 forecast models using diverse statistical and machine learning techniques.
• Compared forecast model performance against persistence and nowcast baseline models using ROC curve analysis.

Main Results:

• Forecast models demonstrated significantly higher area under the ROC curve compared to persistence models, indicating enhanced predictive information.
• Forecast model performance was comparable to baseline nowcast models across all lead times.
• The developed framework successfully forecasts bacterial exceedances with a 3-day lead time.

Conclusions:

• The data-driven forecasting framework provides valuable information beyond historical observations for beach water quality.
• Integrating this framework into beach management can improve public notification and proactive health risk management.
• Environmental data science advancements are key to enhancing decision support for environmental hazard management.