Predicting river ecosystem metabolism across large environmental gradients: Drivers and temporal dependencies in the Iberian Peninsula

Abstract

River ecosystem metabolism plays a significant role in the global carbon cycle. However, the limited spatial or temporal scale of most river metabolism studies hinders our ability to draw general patterns, identify common drivers, and make reliable global predictions. We developed Random Forest models for predicting daily metabolism rates using a large database of more than 100 river reaches across the Iberian Peninsula covering a large environmental gradient. As potential drivers, we included static variables (e.g., catchment area, distance to the sea), anthropogenic factors (e.g., land uses), and short-term dynamic variables (e.g., light, water temperature, discharge) averaged over different periods (from 0 to 40d) to explore the role of shorter vs. longer-term environmental control on daily river metabolism rates. Both daily gross primary production and ecosystem respiration rates responded more strongly to average environmental conditions over the previous 40d than to daily values. The 40-d average random forest models explained up to 77% of gross primary production and 82% of ecosystem respiration variance. The most important drivers of GPP were stage (depth), distance to the sea, and light, while the main predictors of ER were stage and GPP. Dynamic variables were generally the most important drivers of daily metabolic rates, although static ones such as distance to the sea also played a role. Our results indicate that temporal patterns in river metabolism are influenced by a combination of environmental conditions integrated over several weeks, seasonal timing, and to a lesser extent, topology.