Climate warming is altering thermal structure in lakes worldwide, with increased duration and intensity of stratification driving changes in nutrient cycling. Individual lake responses to warming may depend on baseline nutrient concentrations, and predicting lake-specific changes requires understanding drivers and ecological responses that operate at multiple interconnected spatial and temporal scales. Ecologists are increasingly using numerical simulation models to understand complex ecological feedbacks and predict how future global changes will affect ecosystems. Because it is computationally intensive to model multiple climate scenarios for lakes, we developed a distributed computing platform called GRAPLEr, which allows users to run thousands to millions of lake model simulations within the R environment using cloud computing and cyberinfrastructure resources. Simulations are distributed across hundreds of processing nodes that have been aggregated into an overlay virtual network, dramatically reducing computation time. Here, we used the GRAPLEr platform to model nutrient dynamics of two lakes of contrasting trophic state, eutrophic Lake Mendota (Wisconsin, USA) and oligotrophic Lake Sunapee (New Hampshire, USA), under a range of downscaled climate warming scenarios (+0 to +7°C based on RCP 8.5 for 2099) over an 11-year period, with the goal of comparing sensitivity to warming based on lake trophic state.
Results/Conclusions
Our temperature warming simulations revealed substantial year-to-year and between-lake variability in the responses of different water quality metrics in Lake Mendota and Lake Sunapee. Under a +7°C warming scenario, both lakes experienced extended periods of hypolimnetic hypoxia (DO<2 mg/L) and anoxia (DO~0 mg/L) compared to a baseline (+0°C) scenario, with a higher degree of change from baseline in Mendota than Sunapee. While surface concentrations of total nitrogen were consistently decreased under the +7°C warming scenario in both lakes, surface total phosphorus concentrations were consistently increased, though the amplitude of seasonal variability in phosphorus concentrations was dampened, particularly in oligotrophic Lake Sunapee. These differential responses of nitrogen and phosphorus may result in altered nutrient limitation for phytoplankton in these lakes. Our observed differences in nitrogen and phosphorus concentrations between a eutrophic and oligotrophic lake in response to warming highlight the power of simulation modeling and tools, such as the GRAPLEr, for capturing both intra- and inter-annual variability across a range of lakes, which in turn allow us to more effectively predict ecological responses to climate change.