Exploring confidence and uncertainty in projections of potential marine ecosystem stressors under climate change

Marine ecosystems are increasingly stressed by human-induced climate change affecting their physical and biogeochemical environment. Future projections of potential marine ecosystem stressors – including warming, acidification, nutrient availability and declining oxygen levels–are inherently uncertain, however, complicating assessments of climate change impacts. Here we combine data from a novel 30-member ensemble simulation from the GFDL’s Earth System Model with data from CMIP5 Earth System models that were run under a high- and low-carbon-emissions scenario to assess the different sources of uncertainty (internal, model and scenario) in projections of marine ecosystem drivers.

We show that the uncertainty in century-scale global and regional surface pH projections is dominated by scenario uncertainty, highlighting the critical importance of policy decisions on carbon emissions. In contrast, uncertainty in century-scale sea surface temperature projections in polar regions, oxygen levels in low oxygen waters, and regional nutrient availability is dominated by model uncertainty, underscoring that overcoming deficiencies in scientific understanding and improved process representation in Earth System Models are critical for making more robust predictions. For smaller spatial and temporal scales, uncertainty associated with internal variability also constitutes an important source of uncertainty, suggesting irreducible uncertainty inherent in these projections. We also show that changes in the combined multiple ecosystem drivers emerges from the noise in 44% of the ocean in the next decade and in 57% of the ocean by the end of the century following a high carbon emissions scenario. Changes in pH and sea surface temperature can be reduced substantially and rapidly by the end of 21^st century with aggressive carbon emission mitigation, but only marginally for oxygen and net primary productivity. Implications for downscaling of Earth system model output and for projecting global and regional marine fisheries catch will be discussed.