An official website of the United States government
Abstract. Earth system models (ESMs) are useful tools forpredicting and understanding past and future aspects of the climate system.However, the biological and physical parameters used in ESMs can have widevariations in their estimates. Even small changes in these parameters canyield unexpected results without a clear explanation of how a particularoutcome was reached. The standard method for estimating ESM sensitivity isto compare spatiotemporal distributions of variables from different runs ofa single ESM. However, a potential pitfall of this method is that ESM outputcould match observational patterns because of compensating errors. Forexample, if a model predicts overly weak upwelling and low nutrientconcentrations, it might compensate for this by allowing phytoplankton tohave a high sensitivity to nutrients. Recently, we demonstrated that neuralnetwork ensembles (NNEs) are capable of extracting relationships betweenpredictor and target variables within ocean biogeochemical models. Beingable to view the relationships between variables, along with spatiotemporaldistributions, allows for a more mechanistically based examination of ESMoutputs. Here, we investigated whether we could apply NNEs to help usdetermine why different ESMs produce different spatiotemporal distributionsof phytoplankton biomass. We tested this using three cases. The first andsecond case used different runs of the same ESM, except that the physicalcirculations differed between them in the first case, while the biologicalequations differed between them in the second. Our results indicated thatthe NNEs were capable of extracting the relationships between variables fordifferent runs of a single ESM, allowing us to distinguish betweendifferences due to changes in circulation (which do not changerelationships) from changes in biogeochemical formulation (which do changerelationships). In the third case, we applied NNEs to two different ESMs.The results of the third case highlighted the capability of NNEs to contrastthe apparent relationships of different ESMs and some of the challenges itpresents. Although applied specifically to the ocean components of an ESM,our study demonstrates that Earth system modelers can use NNEs to separatethe contributions of different components of ESMs. Specifically, this allowsmodelers to compare the apparent relationships across different ESMs andobservational datasets.
more »
« less
This content will become publicly available on June 4, 2027
A machine learning emulator for dissolved oxygen in earth system models
Abstract The current generation of earth system models (ESMs) exhibit substantial biases and inter-model spread in representing ocean deoxygenation and the expansion of tropical oxygen minimum zones (OMZs). Improving dissolved oxygen simulation is challenging as it depends on uncertain physical and biogeochemical processes and their interactions. A machine learning-based emulator, O2EMU, aims to reduce model biases and inter-model spread by replacing ESMs’ biogeochemical parameterizations with the learned relationships between dissolved oxygen and physical variables. The emulator first learns from the historical shipboard and autonomous observations of dissolved oxygen and temperature and salinity from ocean reanalyses. The learned relationships are then applied to the temperature and salinity outputs from an ensemble of ESM historical simulations. The results show significantly improved climatological OMZ structure and long-term oxygen trends including the shoaling and expansion of the tropical Atlantic OMZs. Application of O2EMU to ESM simulations can reduce the inter-model spread by 70%–75% and the root mean square error by 50%–60%, offering a computationally efficient and scalable alternative to standard biogeochemical models, and providing a stepping stone towardshybridbiogeochemical projections that blend mechanistic ESM dynamics with observationally constrained tracer distributions.
more »
« less
- Award ID(s):
- 2123546
- PAR ID:
- 10706417
- Publisher / Repository:
- IOP
- Date Published:
- Journal Name:
- Environmental Research Letters
- Volume:
- 21
- Issue:
- 11
- ISSN:
- 1748-9326
- Page Range / eLocation ID:
- 114011
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Ocean deoxygenation due to anthropogenic warming represents a major threat to marine ecosystems and fisheries. Challenges remain in simulating the modern observed changes in the dissolved oxygen (O2). Here, we present an analysis of upper ocean (0-700m) deoxygenation in recent decades from a suite of the Coupled Model Intercomparison Project phase 6 (CMIP6) ocean biogeochemical simulations. The physics and biogeochemical simulations include both ocean-only (the Ocean Model Intercomparison Project Phase 1 and 2, OMIP1 and OMIP2) and coupled Earth system (CMIP6 Historical) configurations. We examine simulated changes in the O2inventory and ocean heat content (OHC) over the past 5 decades across models. The models simulate spatially divergent evolution of O2trends over the past 5 decades. The trend (multi-model mean and spread) for upper ocean global O2inventory for each of the MIP simulations over the past 5 decades is 0.03 ± 0.39×1014 [mol/decade] for OMIP1, −0.37 ± 0.15×1014[mol/decade] for OMIP2, and −1.06 ± 0.68×1014[mol/decade] for CMIP6 Historical, respectively. The trend in the upper ocean global O2inventory for the latest observations based on the World Ocean Database 2018 is −0.98×1014[mol/decade], in line with the CMIP6 Historical multi-model mean, though this recent observations-based trend estimate is weaker than previously reported trends. A comparison across ocean-only simulations from OMIP1 and OMIP2 suggests that differences in atmospheric forcing such as surface wind explain the simulated divergence across configurations in O2inventory changes. Additionally, a comparison of coupled model simulations from the CMIP6 Historical configuration indicates that differences in background mean states due to differences in spin-up duration and equilibrium states result in substantial differences in the climate change response of O2. Finally, we discuss gaps and uncertainties in both ocean biogeochemical simulations and observations and explore possible future coordinated ocean biogeochemistry simulations to fill in gaps and unravel the mechanisms controlling the O2changes.more » « less
-
Abstract Climate change reduces ocean oxygen levels, posing a serious threat to marine ecosystems and their benefits to society. State‐of‐the‐art Earth System Models (ESMs) project an intensification of global oxygen loss in the future, but poorly constrain its patterns and magnitude, with contradictory oxygen gain or loss projected in tropical oceans. We introduce an oxygen water mass framework—grouping waters with similar oxygen concentrations from lowest to highest levels—and separate oxygen changes into two components: thetransformationof oxygen in water masses by biological, chemical, or physical processes along their pathways in “ventilation‐space,” and theredistributionof these water masses in “geographic‐space.” The redistribution of water masses explains the large projection uncertainties in the tropics. ESMs with more realistic representations of water masses provide tighter constraints on future redistribution than less skilled ESMs, leading to over a third more of tropical area exhibiting consistent oxygen projections (58% vs. 22%), and a 30% reduction in model spread for tropical oxygen projections. These higher‐skilled ESMs also project weaker global deoxygenation than less skilled models (median of −2.9 vs. −4.2 Pmol per °C of surface warming) controlled by an increase in global water residence times, and they project a stronger increase in oxygen minimum zone ventilation by ocean mixing. These tighter constraints on future oxygen changes are critical to anticipate and mitigate impacts for ecosystems and inform management and conservation strategies of marine resources.more » « less
-
Abstract. We introduce and evaluate the regional ocean model MOM6-COBALT-IND12 version 1 coupling the MOM6 ocean dynamics model to the Carbon, Ocean Biogeochemistry and Lower Trophics (COBALT) biogeochemical model at a horizontal resolution of 1/12°. The model covers the northern Indian Ocean (from 8.6° S to the northern continental boundaries), central to the livelihoods and economies of countries that comprise about one-third of the world’s population. We demonstrate that the model effectively captures the key physical and biogeochemical basin-scale features related to seasonal monsoon reversal, interannual Indian Ocean Dipole and multi-decadal variability, as well as intraseasonal and fine-scale variability (e.g., eddies and planetary waves), which are all essential for accurately simulating patterns of coastal upwelling, primary productivity, temperature, salinity, and oxygen levels. Well represented features include the timing and amplitude of the monsoonal blooms triggered by summer coastal upwelling and winter mixing, the strong contrast between the high evaporation/high salinity Arabian Sea and high precipitation/high runoff/low salinity Bay of Bengal, the seasonality of the Great Whirl gyre and coastal Kelvin upwelling/downwelling waves, as well as the physical and biogeochemical patterns associated with intraseasonal and interannual variability. Quantitatively, the model exhibits relatively small biases, as reflected by root mean square error (RMSE) values in key variables: sea surface temperature (0.25–0.3 °C), mixed layer depth (7–8.09 m), sea level anomaly (0.02 m), sea surface salinity (0.53–0.71 psu), vertical chlorophyll (0.03–0.3 mg m−3), subsurface temperature (0.33 °C), and subsurface salinity (0.07 psu). A major model bias (16 µmol kg−1 of oxygen) is the larger oxygen minimum zone simulated in the Bay of Bengal, a common challenge of ocean and Earth system models in this region. This bias was partly mitigated by improving the representation of the export and burial of organic detritus to the deep ocean (e.g., sinking speed, riverine lithogenic material inputs that protect organic material and burial fraction), and water-column denitrification (e.g., nitrate-based respiration at higher oxygen levels) using observational constraints. These results indicate that the regional MOM6-COBALT-IND12 v1.0 model is well suited for physical and biogeochemical studies on timescales ranging from weeks to decades, in addition to supporting marine resource applications and management in the northern Indian Ocean.more » « less
-
Abstract Stable isotope‐based reconstructions of past ocean salinity and hydroclimate depend on accurate, regionally constrained relationships between the stable oxygen isotopic composition of seawater (δ18Osw) and salinity in the surface ocean. An increasing number of δ18Oswobservations suggest greater spatial variability in this relationship than previously considered, highlighting the need to reassess these relationships on a global scale. Here, we use available, paired δ18Oswand salinity data (N = 11,119) to create global interpolations of each variable. We then use a self‐organizing map, a specialized form of machine learning, to define 19 regions with unique δ18Osw‐salinity relationships in the surface (<50 m) ocean. Inclusion of atmospheric moisture‐related variables and oceanic tracer data in additional self‐organizing map experiments indicates global surface δ18Osw‐salinity spatial patterns are strongly forced by the atmosphere, as the SOM spatial output is highly similar despite no overlapping input data. Our approach is a useful update to the previously delimited regions, and highlights the utility of neural network pattern extraction in spatiotemporally sparse data sets.more » « less
