Freshwater ecosystems around the globe are changing at an unprecedented rate as a result of human activities. These ecosystem responses are complex and driven by feedbacks across multiple temporal and spatial scales, necessitating new approaches for prediction. In response, ecologists are increasingly using numerical simulation models to forecast the effects of future change on lakes and reservoirs. However, it is challenging to predict how different aspects of global change, particularly land use and changing climate, will interact to affect water quality because it is computationally intensive to model thousands of factorial land use and climate scenarios. In response, we developed a unique distributed computing cyberinfrastructure called GRAPLEr, which allows users to submit millions of lake model simulations in the R environment via cloud computing resources. These simulations are parallelized and run separately on a collection of servers located around the world that have been aggregated into an overlay virtual network. We used the GRAPLEr platform to examine the effects of statistically downscaled climate predictions for 2099 and nutrient loading scenarios on water quality in eutrophic Lake Mendota, Wisconsin (USA) over a five-year period.
Results/Conclusions
Our >1 million simulations revealed substantial variability in the responses of different water quality metrics to land use and climate change in Lake Mendota’s catchment. Water temperature and phosphorus (P) loading synergistically interacted to increase maximum total phytoplankton biomass in the warmest and most nutrient-rich scenarios, which represented 7.2oC warmer air temperatures and 25% higher P loads than current conditions. The non-linear response to P and temperature was driven primarily by nitrogen (N)-fixing cyanobacteria; non-N-fixing cyanobacteria exhibited linear responses to changing N, but not P, loads and temperature. Regardless of climate, reductions in P loads from current levels did not substantially decrease in-lake P concentrations or mean total phytoplankton biomass. Our modeling results suggest that the legacy of Lake Mendota’s past eutrophication and accumulated N and P in lake sediments will hinder improvements in future water quality, even if nutrient loads substantially decrease. Given the potential for non-linear increases in bloom-forming cyanobacteria, our study highlights the importance of simulation modeling and tools such as the GRAPLEr to improve our understanding of how different climate and land use drivers will interact to affect water quality.