Abstract Title: Enhancing the Understanding of Riverine Phosphorus Dynamics in the Contiguous United States: A Remote Sensing and Machine Learning Approach

Abstract Submitted to: HYDROLOGY

Abstract Text:

Approximately 42% of rivers in the contiguous United States (CONUS) are affected by total phosphorus (TP), threatening ecosystems and public health on a national scale. However, our current understanding of riverine TP, especially its variations in space and time, is limited. Traditional monitoring of TP requires manual water sampling and lab analysis, limiting the sampling frequency and the resulting data availability. Consequently, only about 45% of the U.S. rivers have been monitored, which inhibits our ability to study riverine TP dynamics and understand its spatial and temporal variations. To address this issue, we collected reflectance data from Landsat satellites for rivers across CONUS and matched them with in situ TP measurements to create a national scale matchup dataset from 1984 to 2021 and develop an optimal machine learning (ML) model to estimate riverine TP. We then applied this ML model to all U.S. rivers that are observable by Landsat satellites to quantify TP variations in the past four decades. Finally, we assessed the relationships between TP variations and natural and anthropogenic factors that potentially impact TP. This work will expand our knowledge of riverine TP dynamics and the driving mechanisms.