Water Quality Prediction Dataset - Georgia Multi-Site Water Monitoring

Overview

The Water Quality Prediction dataset contains daily water quality measurements from 36 monitoring sites across Georgia, USA, designed for spatio-temporal forecasting of water quality parameters.

Data Collection

• 705 daily instances collected from two major water systems: one centered on Atlanta and the other on the eastern coast of Georgia.
• Data funded by the National Science Foundation for spatio-temporal water quality prediction research.
• The dataset explicitly captures spatial dependency among different geographic locations.

Variables

• Input features (11 indices): Specific conductance (max, min, mean in microsiemens/cm at 25°C), dissolved oxygen (max, min, mean in mg/L), temperature (max, min, mean in °C), and pH measurements (max, min).
• Target variable: pH median value (unfiltered, field, standard units) for next-day prediction.

Use Cases

• Water quality forecasting for environmental monitoring and public health protection.
• Spatio-temporal regression modeling capturing geographic dependencies among monitoring sites.
• Developing early warning systems for pH anomalies that may indicate pollution or ecosystem stress.