Patterns and Drivers of Riverine Nitrate Concentrations in the Contiguous United States
Abstract
Nitrate is one of the most widespread and persistent pollutants of our time. Nitrate contamination will exacerbate in the future, as a doubling of urban land area is expected by 2060 in the United States, and croplands have been expanding at a rate of over one million acres per year. Forecasting capabilities are essential for water management under future changing climate and land use conditions, yet our forecasting capabilities have lagged behind due to sparse and discontinuous datasets, in tandem with entangled influences of climate, land use, and geology at the continental scale. Here we collated nitrate data from 2061 rivers in the contiguous United States (CONUS) along with 32 site characteristic indexes, and developed four machine learning models to reconstruct spatial patterns. The training yielded similar performance of the four models, from which one of the models (Boosted Regression Tree) was analyzed to identify the most influential drivers of riverine nitrate concentrations. The analysis revealed that five (out of 32) indexes (drivers) can explain about 69% of spatial variations in nitrate concentrations. The five drivers are nitrogen application rates Nrate and urban area Aurban% (human drivers), mean annual precipitation and temperature (climate drivers), and sand percent Sand% (soil property driver). Nitrate concentrations in undeveloped sites are primarily modulated by climate and soil property; they decrease with increasing Sand% and mean discharge, the water surplus after the evaporative and transpiration water demand. Nitrate concentrations in human-impacted lands increase with Nrate and Aurban% until reaching their "saturation limit" values around 10,000 kg/km2/yr and around 25%, respectively. The results allude to a conceptual model that highlights distinct impacts of different drivers: while human activities predominate nitrogen addition to the land and water, climate and soil properties modulate its removal via export and transformation, the balance of which determine nitrate concentrations. We have also shown that the machine learning model can be used to forecast nitrate concentrations in places without data, a crucial capability as we expect changing climate and growing agriculture and urbanization.
- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2022
- Bibcode:
- 2022AGUFM.H13G..05S