Global patterns and key drivers of stream nitrogen concentration: A machine learning approach.

Area of Science:

• Environmental Science
• Hydrology
• Data Science

Background:

• Anthropogenic nitrogen loading threatens freshwater ecosystems and human health.
• Understanding nitrogen dynamics and flow patterns is crucial for effective management.
• Global-scale spatiotemporal analysis of riverine nitrogen is needed.

Purpose of the Study:

• To develop a machine learning model for predicting global riverine nitrogen concentrations.
• To identify key drivers of nitrogen dynamics at a global scale.
• To map nitrogen pollution hotspots and assess trends.

Main Methods:

• Utilized a random forest machine learning approach.
• Regressed monthly nitrogen concentrations onto 17 predictors at 0.5-degree resolution (1990-2013).
• Validated model with independent data and predicted concentrations for 520 major river basins.

Main Results:

• Identified hotspots of high median nitrogen concentrations in 2013, including major rivers in the US, Pakistan, India, China, and Europe.
• Detected the highest rates of nitrogen concentration increase in eastern China, Canada, Pakistan, and Southeast Asia.
• Determined that temporality (month of year, cumulative count) is the most influential predictor, followed by hydroclimatic, agricultural, and topographic factors.

Conclusions:

• The developed machine learning model effectively captures spatiotemporal variability in riverine nitrogen.
• The study provides a global assessment of nitrogen pollution, highlighting critical regions and trends.
• Findings can inform the development of targeted strategies for mitigating freshwater nitrogen pollution.