FY 2024

Site-specific multiple stressor assessments based on high frequency surface observations and an Earth system model

Olson, E.M.B., J.G. John, J.P. Dunne, C. Stock, E.J. Drenkard, and A.J. Sutton

Earth Space Sci., 11(7), e2023EA003357, doi: 10.1029/2023EA003357, View open access article at AGU/Wiley (external link) (2024)

Global Earth system models are often enlisted to assess the impacts of climate variability and change on marine ecosystems. In this study, we compare high frequency (daily) outputs of potential ecosystem stressors, such as sea surface temperature and surface pH, and associated variables from an Earth system model (GFDL ESM4.1) with high frequency time series from a global network of moorings to directly assess the capacity of the model to resolve local biogeochemical variability on time scales from daily to interannual. Our analysis indicates variability in surface temperature is most consistent between ESM4.1 and observations, with a Pearson correlation coefficient of 0.93 and bias of 0.40°C, followed by variability in surface salinity. Physical variability is reproduced with greater accuracy than biogeochemical variability, and variability on seasonal and longer time scales is more consistent between the model and observations than higher frequency variability. At the same time, the well-resolved seasonal and longer timescale variability is a reasonably good predictor, in many cases, of the likelihood of extreme events. Despite limited model representation of high frequency variability, model and observation-based assessments of the fraction of days experiencing surface T-pH and T-Ωarag multistressor conditions show reasonable agreement, depending on the stressor combination and threshold definition. We also identify circumstances in which some errors could be reduced by accounting for model biases.

Plain Language Summary. Ocean ecosystems are under stress from changing temperature and acidity due to the human-driven increase in global atmospheric carbon dioxide. Global Earth system models (ESMs) are used to study the effects of climate variability and change on marine ecosystems. However, computing power and storage constraints limit the level of detail represented by these simulations. Some short timescale variability present in the real world is missing from ESM output. Despite this inconsistency, we show that at an array of sites where daily observation data from ocean moorings is available, models accurately capture observed spatial patterns in estimates of the amount of time the locations experience combined temperature and acidification stress. We also demonstrate circumstances in which some model errors can be reduced through bias correction.