Assessment of time of emergence of anthropogenic deoxygenation and warming: insights from a CESM simulation from 850 to 2100 CE

Marine deoxygenation and anthropogenic ocean warming are observed and projected to intensify in the future. These changes potentially impact the functions and services of marine ecosystems. A key question is whether marine ecosystems are already or will soon be exposed to environmental conditions not experienced during the last millennium. Using a forced simulation with the Community Earth System Model (CESM) over the period 850 to 2100, we find that anthropogenic deoxygenation and warming in the thermocline exceeded natural variability in, respectively, 60 % and 90 % of total ocean area. Control simulations are typically used to estimate the pre-industrial variability level. However, the natural variability of oxygen (O2) and temperature (T) inferred from the last millennium period is systematically larger than the internal variability simulated in the corresponding control simulation. This renders natural variability from control simulations to be biased towards low estimates. Here, natural variability is assessed from the last millennium period (850–1800 CE), thus considering the response to forcing from explosive volcanic eruptions, solar irradiance and greenhouse gases in addition to internal, chaotic variability. Results suggest that in the tropical thermocline, where biological and solubility-driven O2 changes counteract each other, anthropogenic changes in apparent oxygen utilisation (AOU) and in O2 solubility (O2,sol) are detectable earlier than O2 changes. Both natural variability and change in AOU are predominantly driven by variations in circulation with a smaller role for productivity. By the end of the 21st century, ventilation becomes more vigorous in the tropical thermocline, whereas ideal age in deep waters increases by more than 200 years relative to the pre-industrial period. Different methodological choices are compared and the time for an anthropogenic signal to emerge (ToE) is earlier in many thermocline regions when using variability from a shorter period, from the control simulation or estimates from the industrial period instead of the variability from the last millennium. Our results highlight that published methods may lead to deviations in ToE estimates, calling for a careful quantification of variability. They also highlight that realised anthropogenic change exceeds natural variations in many regions.