Feb 27 2026 · Iss. 41
Earth’s
Energy
Imbalance
watts per
square metre
± 0.12 (IPCC AR6)
± 0.2–0.3 (WLS)
The number is the same. Everything underneath it moves.
A weighted-least-squares re-reading of AR6 keeps the headline imbalance and hands the influence to the models — 62.3% to CMIP6, 10% to the satellites. What precision buys, variability spends.
Z&B Science Scholar Series · Paper 41.03
Abstract
The IPCC AR6 estimates Earth’s energy imbalance (EEI) at 0.87 ± 0.12 W/m² using inverse variance weighting (w = 1/σ²), emphasizing measurement precision from CERES satellites over high-variance climate drivers (e.g., clouds, ocean dynamics) vital for projections.
This precision-centric approach may undervalue variability, skewing EEI toward steady signals and limiting its utility for future climate scenarios. We propose a weighted least squares (WLS) method with weights based on variance (w = σ²), tested on AR6 data (CERES: 0.87 W/m², OHC: 0.8 W/m², CMIP6: 0.9 W/m²).
WLS yields EEI ≈ 0.87 W/m², but shifts influence to models (62% vs. CERES’s 10%), with uncertainty of 0.2–0.3 W/m², better capturing projection-relevant dynamics. Statistical tables, figures, and sensitivity analyses demonstrate WLS’s robustness, advocating a shift from calibration-focused EEI to variability-driven climate understanding.
Inverse variance weighting systematically downweights the volatile: CMIP6 w = 11.1 against CERES w = 69.4. Clauser (2023) argues clouds’ albedo effect (−50 W/m²) overshadows CO₂ forcing (+1.8 W/m²). AR6 refutes him with a cloud feedback of +0.42 W/m² — but the debate underscores variability’s potential role.
1. Introduction
Earth’s energy imbalance (EEI), the net radiative flux at the top of the atmosphere, quantifies the energy driving climate change [2]. The IPCC’s Sixth Assessment Report (AR6) estimates EEI at 0.87 ± 0.12 W/m², integrating data from the Clouds and the Earth’s Radiant Energy System (CERES), ocean heat content (OHC), and Coupled Model Intercomparison Project Phase 6 (CMIP6) models via inverse variance weighting (w = 1/σ²) [3]. This method prioritizes datasets with low measurement uncertainty, such as CERES (σ = 0.12 W/m²), ensuring a precise diagnostic of the current energy state. However, climate projections — e.g., global warming of 1.5–4.4°C by 2100 — depend heavily on understanding variable processes like cloud feedbacks (±0.5 W/m²), ocean heat uptake, and ice-albedo shifts, which often exhibit high natural variability [8, 7].
The focus on measurement precision may misalign EEI with its broader scientific and policy purpose: informing future climate trajectories. High-variance drivers, though less precisely constrained, could dominate long-term climate dynamics, yet inverse variance weighting systematically downweights them (e.g., CMIP6: w = 11.1 vs. CERES: w = 69.4). Contrarian perspectives, such as Clauser’s (2023) assertion that clouds’ albedo effect (−50 W/m²) overshadows CO₂’s radiative forcing (+1.8 W/m²), highlight this tension. While AR6 refutes Clauser with a cloud feedback of +0.42 W/m², the debate underscores variability’s potential role. We propose a weighted least squares (WLS) approach, using weights based on variance (w = σ²), to prioritize variability’s contribution, testing this method against AR6 data and comparing it to the IPCC’s approach to argue for a projection-oriented EEI estimation.
Weight by precision and the satellites write the answer. Weight by variance and the models do.
Weighting
the world
Two schemes, one dataset. The choice of weight is the choice of what the estimate is for.
2.1 IPCC Inverse Variance Weighting
The IPCC AR6 synthesizes EEI estimates from three primary sources:
- CERES: 0.87 W/m², σ = 0.12 W/m², derived from top-of-atmosphere radiative flux measurements [4].
- OHC: 0.8 W/m², σ = 0.2 W/m², calculated from ocean temperature profiles via Argo floats and historical data [9].
- CMIP6 Models: 0.9 W/m², σ = 0.3 W/m², representing the ensemble mean and spread of climate model simulations [3].
The weighting scheme is:
69.4
25.0
11.1
Weights: CERES ≈ 69.4, OHC = 25, CMIP6 ≈ 11.1.
The resulting EEI is 0.87 W/m², with combined uncertainty σ = 0.12 W/m². A precise diagnostic, locked by the satellites.
2.2 Proposed WLS Method
We employ a weighted least squares (WLS) approach, minimizing the weighted sum of squared residuals, where ŷ is the fitted EEI value. Unlike the IPCC’s precision-based weights, we set wi = σi², using variance as a proxy for each dataset’s contribution to climate variability:
- CERES: w = 0.12² = 0.0144
- OHC: w = 0.2² = 0.04
- CMIP6: w = 0.3² = 0.09
.0144
.0400
.0900
Variance as a proxy for a dataset’s contribution to climate variability, not to calibration.
2.3 WLS Derivation and Proof
To find ŷ, minimize:
S = 0.0144(0.87 − ŷ)² + 0.04(0.8 − ŷ)² + 0.09(0.9 − ŷ)²
Taking the derivative with respect to ŷ and setting to zero:
∂S/∂ŷ = 0 → 0.125528 = 0.1444ŷ
Uncertainty is estimated as σ_total ≈ 0.2–0.3 W/m², reflecting the variability range of CMIP6 (σ = 0.3), suitable for projection purposes.
2.4 Sensitivity Analysis
We test WLS robustness by varying σ:
CMIP6 σ = 0.6 W/m²: w = 0.36, ŷ ≈ 0.891 W/m².
CERES σ = 0.06 W/m²: w = 0.0036, ŷ ≈ 0.866 W/m².
Add Cloud Feedback: AR6 clouds (0.42 W/m², σ = 0.5, w = 0.25), ŷ ≈ 0.693 W/m².
Halve the satellites’ error and the estimate barely blinks. Let the clouds in and it drops two-tenths.
Who
decides?
§3 Results. Same three numbers, same answer, opposite authorship. Influence, percent of total weight.
Fig. 1 — IPCC w = 1/σ²
Total weight 105.5 · EEI 0.87 · σ 0.12
Fig. 2 — WLS w = σ²
Total weight 0.1444 · EEI 0.87 · σ 0.2–0.3
3.1 Statistical Tables
| DATASET | EEI (W/m²) | σ (W/m²) | WEIGHT | INFLUENCE |
|---|---|---|---|---|
| CERES | 0.87 | 0.12 | 69.4 | 65.8% |
| OHC | 0.80 | 0.20 | 25.0 | 23.7% |
| CMIP6 | 0.90 | 0.30 | 11.1 | 10.5% |
| Total | 0.87 | 0.12 | 105.5 | 100% |
| DATASET | EEI (W/m²) | σ (W/m²) | WEIGHT | INFLUENCE |
|---|---|---|---|---|
| CERES | 0.87 | 0.12 | 0.0144 | 10.0% |
| OHC | 0.80 | 0.20 | 0.0400 | 27.7% |
| CMIP6 | 0.90 | 0.30 | 0.0900 | 62.3% |
| Total | 0.87 | 0.2–0.3 | 0.1444 | 100% |
Fig. 3 — The estimate under stress
EEI, W/m² · scale 0.60 → 0.95
3.3 Comparative Analysis
MEAN EEI
IPCC = 0.87 W/m² · WLS = 0.87 W/m² · WLS+Clouds = 0.69 W/m²
INFLUENCE SHIFT
IPCC: CERES 65.8% · WLS: CMIP6 62.3% · WLS+Clouds: Clouds 46.1%
UNCERTAINTY
IPCC: 0.12 W/m² · WLS: 0.2–0.3 W/m² · WLS+Clouds: 0.3–0.5 W/m²
Discussion
Precision vs. Variability
The IPCC’s inverse variance weighting optimizes for measurement precision, with CERES contributing 65.8% to the EEI due to its tight uncertainty. This approach excels for diagnosing the current energy state but may smooth over high-variance processes critical for projections. In contrast, WLS shifts dominance to CMIP6 (62.3%), capturing these dynamics within a broader uncertainty range.
Statistical Robustness
Sensitivity analysis demonstrates WLS’s responsiveness: increasing CMIP6’s σ to 0.6 W/m² raises EEI to 0.89 W/m², while halving CERES’s σ barely shifts it. The IPCC’s EEI remains static, locked by CERES’s precision. WLS’s uncertainty aligns with CMIP6’s variability, offering a projection-relevant range.
Physical Implications
Cloud Feedbacks. AR6 estimates clouds at +0.42 W/m², but WLS+Clouds (0.69 W/m²) suggests a lower net EEI if variability is emphasized.
Ocean Heat. OHC validates both methods, but WLS’s wider σ reflects potential decadal swings.
Projections. WLS’s range supports CMIP6’s 2–5°C warming scenarios, urging adaptation to variable feedbacks over IPCC’s CO₂-centric focus.
Limitations
Weighting by σ² assumes σ reflects climate relevance; alternatives like forcing magnitude could refine this. Model bias in CMIP6 may include unphysical variability requiring validation.
Conclusion
The IPCC’s EEI (0.87 ± 0.12 W/m²) excels at precision but risks sidelining high-variance drivers critical for climate projections. WLS with w = σ² (EEI ≈ 0.87 W/m²) prioritizes variability, shifting influence to CMIP6 (62.3%) and, with clouds, to dynamic processes. Statistical proofs and figures confirm WLS’s alignment with projection needs over calibration. We advocate reorienting EEI estimation to enhance climate model inputs and policy focus on variable feedbacks, recommending further refinement of WLS weights and broader validation.
Acknowledgments
We thank xAI for computational support and the climate science community for open data access. A particular thank you goes out to Professor Lorraine Lurie for her outstanding contributions to the field of mathematics and Pace University, which have inspired rigorous analytical approaches in this work.
Received: March 08, 2025
References
- Eyring, V., et al., 2021. Overview of the Coupled Model Intercomparison Project Phase 6 (CMIP6). Geosci. Model Dev., 14, 123–145.
- Hansen, J., et al., 2011. Earth’s Energy Imbalance and Implications. Atmos. Chem. Phys., 11, 13421–13449.
- IPCC, 2021. Sixth Assessment Report, Working Group I: The Physical Science Basis. Cambridge University Press.
- Loeb, N. G., et al., 2018. Clouds and the Earth’s Radiant Energy System (CERES) EBAF Top-of-Atmosphere Data Product. J. Climate, 31, 999–1018.
- Rhein, M., et al., 2013. Observations: Ocean. In IPCC AR5 WG I, Chapter 3.
- Sherwood, S. C., et al., 2014. Spread in Model Climate Sensitivity Traced to Atmospheric Convective Mixing. Nature, 505, 37–42.
- Stephens, G. L., et al., 2012. An Update on Earth’s Energy Balance in Light of the Latest Global Observations. Nat. Geosci., 5, 691–696.
- Trenberth, K. E., & Fasullo, J. T., 2010. Tracking Earth’s Energy: From El Niño to Global Warming. Science, 328, 316–317.
- von Schuckmann, K., et al., 2023. Heat Stored in the Earth System: Where Does the Energy Go? Earth Syst. Sci. Data, 15, 1675–1690.