Ocean Heat Model Reframes Greenhouse Gas Cost Estimates

Integrated assessment models often turn climate damages into a single social cost of carbon, with a discount-rate choice doing much of the work. This paper argues that the setup is too narrow: damages should be linked first to physical heat accumulation, then mapped into economic loss under explicit uncertainty. The authors introduce OPTiMEM, an Ocean-Heat-Content Physics and Time Macro Economic Model, and pair it with a “heat conjecture” that treats weather damages as stochastically proportional to ocean heat increase.

Core Contribution

The central move is to replace a temperature-centered damage curve with an ocean-heat-centered one. The paper’s rationale is physical: the ocean stores about 88.0% of the planetary heat energy relevant to climate and weather, while the atmosphere stores about 0.9%. Temperature responds faster than total heat content and can misrepresent long-lived climate energy, especially when greenhouse gas emissions rise, fall, or stop before the climate system reaches equilibrium.

That change shifts the output from a point estimate to a surface. The authors argue that the social cost of CO₂, CH₄, N₂O, and halogenated gases cannot be represented as single numbers. Instead, they present social-cost estimates across greenhouse gas scenarios, time horizons, and real-dollar discount rates. Figure 8 is the clearest visual summary of that claim: each gas has an isosurface over discount rate and number of years, with marker tabs for a 1.57% carbon-bond mean discount rate and a 0.483% DICE-derived discount rate.

Technical Approach

OPTiMEM starts from fossil-fuel extraction and emissions assumptions, then propagates those into atmospheric greenhouse gas concentrations, Earth energy imbalance, and ocean heat content. The model uses 18 primary scenarios to span uncertainty and a smaller DICE-family scenario set to test implicit assumptions in DICE-style modeling. CO₂ persistence is fitted to Archer’s long-tail atmospheric lifetime estimates with high, central, and low remainder curves; methane includes permafrost emissions and conversion to CO₂; N₂O is driven by population and gross world product; halogenated gases enter through radiative forcing.

Figure 3 shows the model boundary. Land and cryosphere are marked but not modeled as heat reservoirs; the ocean heat box is the validated store. That is a strength for transparency and a constraint on scope. The model deliberately builds around the best available global heat reservoir data rather than trying to represent every Earth-system component.

The economic side applies the heat conjecture to NOAA total weather damages. The authors fit an exponential curve to observed weather damages, scale forward damage curves with modeled OHC, and add tail-risk estimates using Chebyshev-style multipliers. They also propose long-term “carbon bonds” as a practical way to observe real discounting through market rates rather than settling the prescriptive-versus-descriptive discount-rate dispute in theory.

Results and Analysis

The most concrete empirical anchor is the fit to observed weather damages. For the risk-curve construction, the paper reports an exponential fit over the NOAA weather-damage span with x-axis scale 0.3626 and $R^2 = 0.9563$. Those fitted curves are then extended well beyond the data window, which the paper itself flags as a validity concern for projections past roughly 2080.

The tail-risk calculation is more useful than the point estimates because it exposes time aggregation. A 0.1% annual weather-damage risk sounds remote, but Table 1 converts it to 9.52% over 100 years, 25.93% over 300 years, 39.36% over 500 years, and 63.23% over 1000 years. At 1% annual risk, the corresponding probabilities are 63.40% over 100 years and 95.10% over 300 years. Figure 6 then places these outlier weather-damage curves against projected U.S. GDP; the authors note that the 1-in-1000 annual curve appears close to the 2025 GDP region around 2200–2350 CE.

The paper’s sharpest comparison is with DICE-family assumptions. Under the scenario implementing DICE carbon and growth assumptions, the authors report that warming crosses +18°C by 2210 CE and +110°C by 2300 CE. Their interpretation is that such outcomes are incompatible with the comparatively benign conclusions produced by those models. That is a strong criticism, but it depends heavily on accepting OPTiMEM’s emissions, heat, and damage-chain assumptions.

Evidence and Caveats

The evidence is best read as a transparent model-based challenge to standard SCC framing, not as a settled replacement. The paper validates parts of the physical chain against greenhouse gas, temperature, and ocean heat datasets, uses NOAA damage data for the economic fit, and makes its uncertainty dimensions visible. At the same time, the damages extrapolation runs centuries beyond its empirical base, land and cryosphere heat storage are excluded, black-swan events are outside the scenario set, and the heat conjecture remains a conjecture rather than a causal identification of damages from ocean heat. Policymakers, actuaries, infrastructure planners, and climate-economics modelers would benefit most from the risk-surface framing, provided they treat the numerical surfaces as scenario diagnostics rather than forecasts.