AUTHOR=Tso Chak-Hau Michael , Magee Eugene , Huxley David , Eastman Michael , Fry Matthew 

TITLE=River reach-level machine learning estimation of nutrient concentrations in Great Britain

JOURNAL=Frontiers in Water

VOLUME=Volume 5 - 2023

YEAR=2023

URL=https://www.frontiersin.org/journals/water/articles/10.3389/frwa.2023.1244024

DOI=10.3389/frwa.2023.1244024

ISSN=2624-9375

ABSTRACT=Nitrogen (N) and phosphorus (P) are essential nutrients necessary for plant growth and to support life in aquatic ecosystems. However, excessive N and P can lead to algal blooms that deplete oxygen and lead to fish death and the release of toxins that are harmful to humans. Predictions of N and P in rivers are typically calculated at station or grid (>1km) scale; therefore, it is difficult to visualize the evolution of water quality as water travels downstream. Using a high-resolution reach-scale river network and associating each reach with land cover fractions and catchment descriptors, we trained random forest models against aggregated data (2010-2020) from the Environmental Agency Open Water Quality Data Archive for 2343 stations to predict long-term nitrate and orthophosphate concentrations at each river reach in Great Britain (GB). We separate the model-training and predictions for different seasons to investigate the potential difference in feature importance. Our model predicted concentrations with an average testing coefficient of determination (R 2 ) of 0.71 for nitrate and 0.58 for orthophosphate using five-fold cross-validation. Our model shows slightly better performance for higher Strahler stream order, highlighting the challenges for making predictions in small streams. Our results reveal that arable and horticultural land use is the strongest and most reliable predictor for nitrate, while floodplain extents and standard percentage runoff are stronger predictors for orthophosphate. Nationally, higher orthophosphate concentrations are observed in urbanized areas. This study demonstrates the joint use of a river network model and machine learning can readily help provide a river-network understanding on the spatial distribution of water quality levels.