Carbon dioxide (CO2) concentrations in lakes vary strongly over time. This variability is rarely captured by environmental monitoring, but is crucial for accurately assessing the magnitude of lake CO2 emissions. However, it is unknown to what extent temporal variability needs to be captured to understand important drivers of lake carbon cycling such as climate and land management. We used environmental monitoring data of Swedish forest lakes collected in autumn (n=439) and throughout the whole open-water season (n=22) from a wet and a dry year to assess temporal variability in effects of climate and forestry on CO2 concentrations across lakes. Effects differed depending on the season and year sampled. According to cross-lake comparisons based on autumn data, CO2 concentrations increased with annual mean air temperature (dry year) or catchment forest productivity (wet year) but were not related to colored dissolved organic matter (CDOM) concentrations. In contrast, open-water season averaged CO2 concentrations were similar across temperature and productivity gradients but increased with CDOM. These contradictions resulted from scale mismatches in input data, lead to weak explanatory power (R2=9-32%) and were consistent across published data from 79 temperate, boreal and arctic lakes. In a global survey of 144 published studies, we identified a trade-off between temporal and spatial coverage of CO2 sampling. This trade-off clearly determines which conclusions are drawn from landscape-scale CO2 assessments. Accurate evaluations of the effects of climate and land management require spatially and temporally representative data that can be provided by emerging sensor technologies and forms of collaborative sampling.
Lakes interact with the atmosphere by exchanging greenhouse gases such as carbon dioxide that have implications for the global climate. In order to predict future carbon dioxide exchanges, the effects of climate and land management need to be determined. Here, we show that the apparent effects of climate and forestry on carbon dioxide concentrations in temperate, boreal and arctic lakes are inconsistent and depend on whether lakes are sampled during autumn or throughout the year, and whether it was a wet or dry year. To accurately quantify effects, hundreds of lakes would need to be sampled at least six times per year. Published literature suggests that this is unrealistic. This implies a need for new automated measurement technologies and collaborative team work to provide representative data and allow accurate predictions of the effects of climate and land management.