Predicting coastal ocean conditions a decade ahead has long been considered a formidable challenge. Global Earth system models used for decadal forecasting operate at spatial resolutions of roughly 100 to 200 kilometers, far too coarse to represent the complex bathymetry, tidal mixing, and mesoscale circulation that govern conditions on a continental shelf. The Gulf of Maine is particularly tricky because its environment is shaped by a tug-of-war between two very different water masses that enter through the Northeast Channel: cold, fresh, oxygen-rich Labrador Slope Water flowing down from the subpolar North Atlantic, and warm, saline Warm Slope Water associated with greater Gulf Stream influence. These water masses modulate temperature, salinity, nutrients and carbonate chemistry across the entire ecosystem, producing local responses to climate change that can be complex and sometimes counterintuitive.

To overcome the resolution problem, the researchers downscaled global decadal predictions from the SPEAR prediction system into a 1/12-degree regional ocean model, MOM6-COBALT-NWA12, which couples ocean physics, sea ice and the COBALT biogeochemical model. They generated a 10-member ensemble of 10-year retrospective forecasts, or hindcasts, initialized every January from 1965 to 2025, amounting to nearly 6,500 years of simulation. These were complemented by uninitialized historical simulations driven only by external radiative forcing, allowing the team to separate the predictable climate change signal from the predictable internal variability of the ocean. Initial conditions were constrained by a data-assimilating regional simulation gently relaxed toward the EN4 observational dataset in deep waters, while the shallow Gulf of Maine basins themselves were left to evolve freely.

The retrospective forecasts successfully captured the region’s dramatic water mass history. The downscaled simulations reproduce a gradual bottom warming of about 2 degrees Celsius from the 1980s to the mid-1990s, a cooling of similar magnitude thereafter, and intensified warming from the mid-2000s to around 2020. They also track the late-1990s replacement of Warm Slope Water by Labrador Slope Water, the subsequent return of warm water influence from 2004 to 2020, and the marked cooling that began in the early 2020s, when cooler, fresher water was observed entering the deep Gulf through the Northeast Channel. Independent observations, including the Fishing Industry SHared Bottom Oceanographic Timeseries collected with industry partners, corroborate the simulated variability in bottom temperature, salinity, oxygen and carbonate chemistry.

Beneath these physical swings lies a biogeochemical story with a troubling twist. Bottom oxygen anomalies in the Gulf vary inversely with temperature and salinity, and the team’s decomposition shows that roughly 70 percent of the interannual variance in bottom oxygen comes from a non-solubility residual reflecting ventilation, respiration and advection, rather than from temperature-dependent solubility alone. Episodic oxygen maxima in 2004, 2015 and 2019 coincide with deep winter mixing that ventilated the deep basins and with the oxygen content of the inflowing slope waters. Meanwhile, bottom pH and aragonite saturation state, a key measure of how hospitable seawater is to shell-forming organisms, have declined steadily since 1980 as the ocean absorbs rising atmospheric carbon dioxide.

The attribution analysis reveals how the two acidification-relevant quantities diverge. Bottom pH, declining at about 0.013 units per decade, is led by rising dissolved inorganic carbon of roughly 4 micromoles per kilogram per decade, the fingerprint of anthropogenic carbon uptake, with warming contributing a further decline and increasing alkalinity partially buffering the trend. Aragonite saturation state, by contrast, is governed on interannual timescales primarily by alkalinity, which varies with water mass: high-alkalinity Warm Slope Water raises the saturation state and masks acidification, while Labrador Slope Water offers far less buffering. During the warm decade from 2010 to 2020, the influx of Gulf Stream-influenced water actually counteracted the acidification trend in saturation state even as pH fell. The recent shift toward colder, fresher Labrador water reverses that protection, amplifying acidification and producing a sharp drop in aragonite saturation.

The prediction skill analysis shows that different habitat variables draw their predictability from different sources. Bottom temperature forecasts show significant positive anomaly correlations across most of the ten lead years, driven both by the forced warming trend and by predictable water mass variability in the first few years. Bottom pH skill likewise remains high throughout the decade because the carbon uptake trend dominates. Bottom oxygen and aragonite saturation, which depend more strongly on internal variability, retain significant skill for about two years before dropping, although the initialized forecasts still outperform both persistence and uninitialized simulations during that window. The model correctly predicted the timing and magnitude of major transitions, including the rapid 2005 to 2015 warming and the recent shift to cooler, fresher conditions.

Translating ensemble output into decision-ready probabilities, the team used extended logistic regression to produce tercile forecasts of below-normal, normal and above-normal conditions averaged over three-year horizons that match typical stock assessment timelines. All four variables achieved positive ranked probability skill scores relative to persistence, with forecast accuracies well above random benchmarks. The probabilistic system confidently captured the observed mid-2000s cooling and the subsequent warming phase through 2020. For the coming years, the forecasts indicate below-normal bottom temperatures with 88.3 percent probability, above-normal bottom oxygen with 92.0 percent probability, and, with greater than 99 percent probability, below-normal bottom pH and aragonite saturation, meaning the respite from rapid warming will likely come at the cost of accelerating acidification.

That trade-off carries real consequences for the region’s most valuable fisheries. American lobster, scallops and groundfish all depend on bottom habitat, and early life stages of shellfish are consistently more sensitive to acidification than adults, creating recruitment bottlenecks that can drive population declines. Lobster larvae show stage-specific vulnerabilities to combined warming and acidification, and previous projections suggest the entire Gulf of Maine could experience suboptimal aragonite saturation for most of the year by 2050 under high emissions scenarios. The authors caution that the reported skill represents potential skill, verified against a model-based reference in the absence of continuous deep biogeochemical observations, and may be an overestimate. More frequent observations of deep shelf biogeochemistry will be essential to confirm predicted changes and sharpen confidence.

Even with those caveats, the study establishes a foundation for operational decadal prediction of coastal bottom environments. The framework, in which a computationally efficient regional model downscales regularly updated global predictions, is intended to be transitioned toward operational use through NOAA’s Changing Ecosystems and Fisheries Initiative. The researchers emphasize that marine resource management is a wicked problem, and that careful, application-specific analysis of the costs, benefits and tradeoffs of using such forecasts will be needed. But as water mass reorganization continues to reshape the Gulf of Maine, the ability to anticipate bottom temperature, oxygen and acidity years ahead offers fishers, managers and conservationists something they have rarely had: lead time to adapt before the changes arrive.

Subject of Research: Decadal prediction of bottom temperature, oxygen, pH and aragonite saturation state in the Gulf of Maine using a downscaled regional ocean biogeochemical model

Article Title: Decadal biogeochemical predictions for the changing bottom marine environment of the Gulf of Maine

Article References: Koul, V., Ross, A. C., Stock, C., Zhang, L., Wittenberg, A. T., & Delworth, T. (2026). Decadal biogeochemical predictions for the changing bottom marine environment of the Gulf of Maine. Ocean Science, 22(5), 3017-3036. https://doi.org/10.5194/os-22-3017-2026

Image Credits: AI Generated

DOI: 10.5194/os-22-3017-2026

Keywords: Gulf of Maine, decadal prediction, ocean acidification, bottom temperature, dissolved oxygen, aragonite saturation, Labrador Slope Water, Warm Slope Water, MOM6-COBALT, fisheries management, Northeast US Shelf, water mass variability