Abstract
Riparian lowlands are important for protecting aquatic ecosystems threatened by contamination. Their ability to attenuate and reduce nutrient rich water provides a useful ecosystem service. Previous field studies have shown that the hydrological conditions within a lowland can affect their reduction capabilities dependent on the dominant flow pathways, i.e., surface runoff, groundwater discharge, and drain flow. For example, the likelihood of nutrient reduction within a riparian lowland dominated by surface runoff is low, conversely if groundwater discharge dominates the likelihood is higher. Hence, knowledge of the flow pathways can be used to provide a qualitative estimate of the reduction capacity in riparian lowlands, information that is vital in assessing catchment scale processes. The objective of this study is to establish a relationship between the dominant riparian flow pathways and lowland features, such as slope and hydrogeology/geology. Previous work had shown the ability of a downscaled high-resolution numerical model to replicate observed annual flow patterns within a riparian lowland. This downscaled numerical model was used to provide quantitative information regarding the riparian flow partitioning. To diversify the dataset, the riparian lowland was segmented to provide flow information at different scales, and topographic and hydraulic properties within the model were perturbed to capture the range of topographic and geological characteristics present at large scale. These data were then used to train a random forest (RF) model, where the target variable was the fraction of overland flow. Applying the RF model to a 12,785.5 km2 large region in Denmark provided a prediction in line with our understanding of the area. This approach can prove useful in enhancing existing nitrate management tools by incorporating variability in lowland nitrate reduction capacity.
Original language | English |
---|---|
Article number | 129536 |
Number of pages | 14 |
Journal | Journal of Hydrology |
Volume | 620 |
Issue number | Part B |
Early online date | 25 Apr 2023 |
DOIs | |
Publication status | Published - May 2023 |
Externally published | Yes |
Keywords
- Flow partitioning
- Local scale
- Machine learning
- Modelling
- Nitrate