Increasing phosphorus loss despite widespread concentration decline in US rivers

Area of Science:

• Environmental Science
• Water Quality
• Data Science

Background:

• Phosphorus (P) loss from land to water causes pollution and threatens food security.
• Analyzing long-term P trends is difficult due to limited historical data.

Purpose of the Study:

• To reconstruct temporal trends of total phosphorus (TP) using deep learning.
• To analyze TP concentration and flux trends in rivers across the contiguous United States (CONUS).

Main Methods:

• Utilized intensive hydrometeorological data and a multitask long short-term memory (LSTM) deep learning model.
• Trained the model on data from 430 rivers across the CONUS for daily TP reconstruction (1980-2019).

Main Results:

• Reconstructed daily records showed declining TP concentrations in 60% of rivers.
• Urban river TP concentrations declined most, while agricultural rivers saw increases.
• Despite declining concentrations, total P loss (flux) increased by 6.5% per decade nationwide due to rising river discharge.

Conclusions:

• Deep learning effectively reconstructs historical P trends, addressing data gaps.
• Effective P control in urban areas contrasts with challenges in agricultural nonpoint source pollution.
• Increasing river discharge, driven by climate change, significantly complicates efforts to reduce overall P loss.