Earth System Sensitivity (ESS)

• Earth System Sensitivity (ESS) is the measure of long-term equilibrium warming per unit radiative forcing, accounting for both fast (decadal) and slow (millennial) feedbacks.
• Quantitative estimates derived from SVECM analyses and Bayesian deep-time models yield ESS values ranging from 1.2 to 12 °C per CO₂ doubling, reflecting methodological diversity.
• The inclusion of slow feedbacks in ESS calculations highlights amplified warming beyond short-term ECS estimates, emphasizing robust climate policy and mitigation strategies.

Earth System Sensitivity (ESS) quantifies the long-term equilibrium increase in global mean surface temperature per unit of radiative forcing, accounting for both fast and slow geophysical feedbacks. In contrast to the traditional equilibrium climate sensitivity (ECS), which considers only fast-feedback processes over decadal to centennial timescales, ESS incorporates the millennial-scale effects of ice-sheet dynamics, vegetation changes, carbon-cycle responses, and other slow feedbacks. This parameter is a central metric for understanding past climate transitions and for projecting the ultimate magnitude of warming “locked in” by present-day greenhouse gas concentrations (Nakano et al., 11 Jan 2026, Hansen et al., 2022, Hansen et al., 2012, Wong et al., 2019).

1. Conceptual Foundations and Mathematical Definition

ESS is formally defined as the equilibrium temperature response $\Delta T_{\mathrm{eq}}$ to a specified change in external radiative forcing $\Delta F$ after all fast and slow feedbacks have equilibrated:

$$\mathrm{ESS} = \frac{\Delta T_{\mathrm{eq}}}{\Delta F}$$

For the standard case of CO₂ doubling, ESS is often denoted as $AT2x$, such that:

$AT2x = \Delta T_{\mathrm{eq}}(2\times CO_2)$

and, using the canonical radiative forcing for CO₂ doubling (Myhre et al.), $F_{2\times CO_2} = 5.35\ln 2$ W m⁻², one can write:

$\mathrm{ESS} = \frac{AT2x}{5.35\ln2}$

ESS explicitly includes “slow” processes such as ice-sheet albedo changes and long-lived greenhouse gas adjustments, which are omitted from ECS as classically defined in the Charney Report (Hansen et al., 2022, Hansen et al., 2012, Wong et al., 2019). Because slow feedbacks amplify the warming response, ESS necessarily exceeds ECS except in narrowly defined climatic states near present conditions.

2. Estimation Approaches: Statistical and Process-Based Frameworks

Empirical constraints on ESS derive primarily from paleoclimate archives and integrated process models. Nakano and Nishimura (Nakano et al., 11 Jan 2026) implement a Structural Vector Error Correction Model (SVECM) to Antarctic ice-core records, specifying the following system:

$$\Delta Y_t = \Pi Y_{t-1} + \sum_{i=1}^{3}\Gamma_i \Delta Y_{t-i} + \Psi Z_t + \epsilon_t$$

where $Y_t = [T_t, \ln CO_{2,t}]'$ (temperature anomaly and log-transformed CO₂), $Z_t$ is exogenous orbital forcing (N60J; W/m² at 60°N in June), and the lag structure ($k=3$) is selected by the Akaike Information Criterion. Cointegration analysis confirms a single long-run equilibrium ($r=1$) between temperature, $\ln CO_2$, and orbital forcing.

Long-term carbon cycle models, such as GEOCARBSULFvolc (“GEOCARB”), are calibrated against multi-million-year proxy data with Bayesian methods to yield ESS constraints that integrate tectonic, biogeochemical, and weathering feedbacks (Wong et al., 2019). Posterior parameter distributions are determined via Latin-hypercube sampling and probabilistic acceptance-rejection, fusing observational proxy records (e.g., Foster et al. 2017 for CO₂; Mills et al. 2019 for temperature).

Empirical approaches based on glacial–interglacial cycles use energy-budget closure over transitions such as the Last Glacial Maximum (LGM) to Holocene:

$$\mathrm{ESS}_{LGM-Holocene} = \frac{\Delta T}{F_{GHG} + F_{surfAlbedo}}$$

Composite analyses yield $\Delta T \approx 7^\circ$C, $F_{GHG} \approx 2.25$ W m⁻², $F_{surfAlbedo} \approx 3.5$ W m⁻², resulting in $\mathrm{ESS}\approx1.22$ °C W⁻¹ m² (Hansen et al., 2022).

3. Quantitative Estimates and Constraints

A range of methods yields divergent ESS values, principally due to differences in statistical specification, proxy selection, and process representation.

Source	Methodology	Estimated ESS
Nakano & Nishimura (2026)	SVECM (800 kyr)	$12.0^\circ$C per CO₂ doubling (5.8–18.1 °C, 95% CI) (Nakano et al., 11 Jan 2026)
Wong et al. (2019)	GEOCARB + Bayesian	3.4 °C per CO₂ doubling (2.6–4.7 °C, 5–95%) (Wong et al., 2019)
Hansen et al. (2022)	Glacial cycles	$1.2 \pm 0.3$ °C W⁻¹ m² (≈8–10 °C for current GHG) (Hansen et al., 2022)
Hansen et al. (2012)	Paleoclimate energy budget	$6 \pm 2^\circ$C per 4 W m⁻² (LGM-Holocene scale) (Hansen et al., 2012)

Nakano and Nishimura (Nakano et al., 11 Jan 2026) report that a long-run coefficient $\beta_{CO₂}=17.30$ (z = –3.82, p < 0.01) linking temperature to $\ln CO_2$ in their SVECM framework translates to

$\mathrm{ESS} = \beta_{CO₂} \cdot \ln 2 \approx 17.30 \times 0.6931 \approx 12.0\,^{\circ}C$

per CO₂ doubling for the last 800 kyr of Antarctic climate history.

Wong et al. (Wong et al., 2019) derive a median $AT2x = 3.4$ °C (5–95% range: 2.6–4.7 °C) from a Bayesian calibration to the GEOCARB model, integrating deep-time CO₂ and temperature proxy data (last 420 Myr). Glacial climates are found to exhibit a GLAC amplification factor, with median $AT2x$ doubling to 7.1 °C.

The spread in ESS estimates reflects not only methodological diversity but significant physical uncertainties—particularly in the efficacy of ice-sheet and weathering feedbacks, process parameterizations, and proxy calibration.

4. Feedbacks, Timescales, and State-Dependence

ESS subsumes the amplifying impacts of both fast feedbacks (water vapor, sea ice, clouds, aerosols) and slow feedbacks (ice-sheet albedo, long-lived greenhouse gases, vegetation, continental shelf processes) (Hansen et al., 2022, Hansen et al., 2012). On timescales:

• Fast feedbacks operate over decades to centuries.
• Slow feedbacks equilibrate over centuries to millennia and beyond, with the carbon cycle’s response sometimes taking up to $10^6$ years.

State-dependence emerges in process-based models and is supported by paleoclimate evidence: the magnitude of ESS varies with background climate (e.g., stronger ice-sheet feedbacks during glacial intervals), with global mean sensitivity increasing in warmer states as cloud albedo diminishes and atmospheric water vapor increases (Hansen et al., 2012). The GEOCARB model incorporates a glacial amplification parameter (GLAC), with posterior medians of GLAC = 2.1 (Wong et al., 2019).

Feedback accounting in long-term radiative balance equations is:

$$\Delta T = S_{ESS} \cdot (F_{fast} + F_{slow}) = S_{ESS} \cdot F_{total}$$

In practical paleoclimate calibration, the partitioning of forcing among $F_{GHG}$, $F_{surfAlbedo}$, and other terms is critical for robustly inferring ESS.

5. Statistical Properties, Uncertainties, and Robustness

The principal sources of statistical uncertainty in ESS inference are:

• Proxy calibration errors (e.g., temperature reconstructions, ice-core gas measurements).
• Model structure (e.g., cointegration rank, lag order, structural restrictions in SVECMs).
• Representation of non-CO₂ greenhouse gases and non-radiative forcings.
• Uncertainties in albedo feedback efficacy, especially in the calculation of glacial–interglacial energy budgets (Hansen et al., 2022, Hansen et al., 2012).

Nakano and Nishimura (Nakano et al., 11 Jan 2026) estimate a 95% confidence interval for ESS of 5.8–18.1 °C per CO₂ doubling, based on the standard error of the SVECM long-run coefficient. Their robustness analyses demonstrate that ESS estimates consistently exceed model-based ECS studies even after varying the lag order, exogenous orbital solution, and structural identification assumptions.

Wong et al. (Wong et al., 2019) show that the posterior distribution of $AT2x$ is sharply constrained by including both CO₂ and temperature proxies. Sensitivity experiments reveal that uncertainty in the fraction of weatherable land area ($f_{Aw}/f_A$) is a dominant factor, with implications for the Cretaceous temperature bias in model hindcasts.

6. Climate and Policy Implications

ESS has fundamental implications for equilibrium climate projections and the eventual magnitude of “warming in the pipeline.” The high value ($\sim$12 °C per CO₂ doubling) estimated by Nakano and Nishimura (Nakano et al., 11 Jan 2026) implies that, should elevated CO₂ levels persist, long-term warming would far outstrip short-term ECS-based projections.

Hansen et al. (Hansen et al., 2022) calculate that present-day anthropogenic forcing (~4.6 W m⁻²) commits the climate system to 8–10 °C of eventual equilibrium warming when slow feedbacks are included. Even with current aerosol cooling, committed warming remains ≳8 °C. The discrepancy between ESS and ECS (1.2 ± 0.3 °C W⁻¹ m² vs. 0.6 °C W⁻¹ m²) underscores that slow-feedback amplifications, though often ignored in policy assessments, become dominant over the long horizon.

These projections indicate that centennial- to millennial-scale climatic and sea-level changes will be dramatically larger than those inferred from ECS alone, unless rapid decarbonization and emissions mitigation are pursued. The overwhelming contribution of slow feedbacks over geological timescales presents severe constraints on adaptation-based strategies (Hansen et al., 2012).

7. Future Directions and Open Research Questions

The largest parametric uncertainties in ESS inference concern quantification of weathering feedbacks, efficacy of ice-sheet albedo processes, slow greenhouse gas responses, and proxy record calibration (Wong et al., 2019). Improved proxies for weatherable land area, runoff, and continental relief, as well as integration of hierarchical statistical models for proxy error, are expected to sharpen constraints on ESS. Integration with higher-complexity climate models capable of simulating time-dependent feedback activation will enable assessment of nuanced state-dependence, beyond the piecewise-constant formulation of current carbon-cycle frameworks.

Advancing understanding of slow-feedback timescales and nonlinearities is likewise crucial for robust estimation. Bayesian and data-assimilation techniques in deep-time paleoclimate modeling provide a promising avenue for future constraint (Wong et al., 2019). The practical relevance of ESS for climate policy and risk assessment hinges on rigorous empirical anchoring, highlighting the enduring importance of process-level paleoclimate research and methodological innovation.

---

Key Cited Works

• Nakano & Nishimura (2026), "Structural Cointegration of the Paleoclimate: Estimating Earth System Sensitivity" (Nakano et al., 11 Jan 2026)
• Hansen et al. (2022), "Global warming in the pipeline" (Hansen et al., 2022)
• Hansen et al. (2012), "Climate Sensitivity, Sea Level, and Atmospheric CO2" (Hansen et al., 2012)
• Wong et al. (2019), "A tighter constraint on Earth-system sensitivity from long-term temperature and carbon-cycle observations" (Wong et al., 2019)