Metabolism of steppe rivers in Mongolia and the United States: drivers, heterogeneity, methodological bias, and climate warming

Abstract

River metabolism is a central component of global biogeochemical processes and has become a widely used metric of ecosystem function. With increasing data availability, broad-scale models of metabolism are now available towards prediction and improved understanding. Many river metabolism studies do not provide sufficient methodological detail for replication, but extrapolate from numerous local measurements and predict responses to various drivers, including climate. I was therefore interested in 1) how we can make metabolism estimates more accurate, representative, and comparable in methods and reporting? 2) what variables best explained temperate steppe metabolism and how did this vary by region and scale? and 3) how can we expect these rates to change under a warming climate? I use the answers to these questions to improve our understanding, reporting, and representativeness of studies of river metabolism. In evaluating the reporting of open channel river metabolism methods, only 79% of 43 sampled papers published from 2015-2019 mentioned calibration, 44% described sensor placement, and 34% did not describe estimation approaches sufficient for replication. Given that spatial heterogeneity in rivers influences metabolism, and measurement sensitivities vary with sensor model, it is important to have appropriately established replicable protocols and detailed information in reported methods along with a holistic understanding of how river heterogeneity might influence metabolism. I deployed 2-8 sensors at 92 steppe river reaches to characterize site heterogeneity, evaluating how sensor placement and type, deployment length, drift correction, data source, local versus remotely sensed data, and calibration affected metabolism estimates. Estimates of gross primary production (GPP) and ecosystem respiration (ER) were strongly influenced by deployment location within a river reach; GPP and ER rates varied up to 131% and 69% respectively across a river width and up to two orders of magnitude within reach. Dissolved oxygen sensor brands vary widely in precision and accuracy; I found even when operated within stated performance ranges, estimates of GPP and ER could vary by 82% and 198% respectively if not properly calibrated, as determined using field data from a sample site. Inaccuracies from sensor drift over weeklong deployments led to 48% ER overestimation and 2% GPP overestimation comparing uncorrected with corrected field data. With a firmer understanding of methodological and riverine heterogeneity, I could more confidently compare our sites. To explore explanatory structures across scales, I then linked metabolism estimates with reach-to-watershed-scale hydrogeomorphology, vegetation, climate, and anthropogenic impact metrics to evaluate predictors and applicability of traditional ecological frameworks in the Anthropocene. I expected that vegetation and climate related to ecoregion would be more explanatory than human or hydrogeomorphic data. I present the structures with the greatest explanatory power by river type, scale, and location. This required a systematic approach to identify the most explanatory variables, many of which were strongly correlated. I was subsequently interested in using these explanatory mixed models to predict change. Responses of metabolism rates to climate change is critically important to global carbon cycling, so I used the above models to predict changes in GPP and ER under warmer temperatures. I evaluated the downscaling of broad-scale metabolism models using data collected from broad regions of Mongolia and North America. The understudied rivers of the semi-arid steppe of Mongolia are particularly vulnerable to climate change due to high altitude and latitude. This steppe has matching ecoregions with the United States Great Plains, allowing cross-continent investigation of temperature effects on river metabolism. I evaluate how a broad-scale modeling approach applied at the ecoregion level, projecting changes in estimated rates of metabolism under different warming scenarios. Temperature was not the primary explanatory variable, but directly and indirectly influenced modeled rates of metabolism. Our metabolism models did not scale down well. The Grassland Steppe was the most temperature-sensitive ecoregion for both rates on both continents. I offer best practices for more comparable, replicable, representative, and accurate methods in stream metabolism study, and present most explanatory structures of variables by river type, scale, and location. I conclude that macrosystem-scale studies require broader interdisciplinary and multi-scale assessment for prediction and capture of variation in aquatic metabolism, and the observed distribution of spatial patterns of river metabolism is scale-dependent. This suggests that researchers, managers, and policymakers must incorporate factors operating at multiple scales to inform management and climate projections, particularly if interested in modeling the influence of increased temperature on river metabolism.