Leveraging Catchment Attributes to Explain Patterns of Concentration-Discharge Relationships Across the Contiguous United States

Identifying catchment attributes that influence solute production and transport can aid in understanding solute loading to surface waters and predicting solute export regimes. We analyzed the new CAMELS-Chem dataset, which pairs water chemistry data with the existing CAMELS (Catchment Attributes and Meteorology for Large-sample Studies) dataset to identify links between catchment attributes and solute export behavior across the contiguous US. We focused on 12 geogenic, exogenous, and biologically associated constituents across 44-249 catchments.

To evaluate solute export behavior patterns, we quantified concentration-discharge (C-Q) metrics (slope, intercept, ratio of coefficients of variation) for each constituent using Bayesian inference to estimate segmented regression parameters and threshold positions. We then used several approaches to link catchment attributes to these metrics. Specifically, random forest models and Spearman correlations were used to select attributes that best predicted C-Q metrics. We then used hierarchical clustering to group similar catchments based on attribute values and examined how C-Q metrics differed across catchment clusters.

C-Q relationships were dominated by linear responses (79-100% of models for each constituent lacked thresholds). Interestingly, a large majority of the C-Q relationships that displayed threshold behaviors were located in the central US. Geogenic constituents (e.g., magnesium and sodium) exhibited a weakly dilutionary (slope < 0) to constant (slope not different than 0) response with increasing streamflow. Conversely, dilutionary behaviors were rare for more biogeochemically reactive constituents (e.g., aluminum and dissolved organic carbon) that demonstrated constant or enriched (slope > 0) concentration with increasing streamflow. Further, across geographic clusters, geogenic constituents displayed minimal differences in C-Q metrics while biogeochemically reactive constituents exhibited larger differences. In summary, relationships between catchment attributes and C-Q patterns were useful in explaining variability in constituent export behavior, but these relationships depended on the solute source and reactivity and geographic location.