Climate and Land-Use Gradients Drive Spatially Coherent CO2 Driver-Response Relationships Across US Lakes

English Abstract:

Understanding how diverse environmental factors influence partial pressure of CO2 (pCO2) in lakes is challenging because local or regional relationships can rarely be generalized at continental scales. The ability to predict lake pCO2 typically decreases at larger spatial extents and with increasing environmental gradients, suggesting non-linearities in the pCO2 driver-response relationships across large and diverse landscapes. We explored spatial patterns in the response of pCO2 to several biological (chlorophyll a) and chemical (nutrients, dissolved organic carbon and alkalinity) drivers in lakes across the contiguous United States using geographically weighted regressions. Allowing the regression parameters to change spatially more than tripled our ability to predict pCO2 compared to traditional multi-linear regression models. The regression parameters formed spatially contiguous clusters, indicating that there were large-scale differences in pCO2 driver-response relationships and that these clusters were substantially different from widely used, pre-determined regionalisation frameworks. The clusters were mainly differentiated on the basis of climate and land-use. In dry and low development areas, pCO2 was primarily related to alkalinity, whereas high precipitation and intensive land use (e.g., agriculture and urban) areas were negatively related to Chl a and positively related to dissolved organic carbon. These results highlight the need to consider non-linearities in the relationships between lake pCO2 and its drivers across large environmental and spatial gradients, which may improve extrapolations based on smaller-scale studies and will improve our understanding of how the aquatic carbon cycle responds to global change.