Frank Stefani, a researcher at the Helmholtz-Zentrum Dresden-Rossendorf (Institute of Fluid Dynamics), has published work examining the relative roles of solar activity (using the geomagnetic aa index as a proxy) and CO₂ in driving global climate changes, particularly sea surface temperatures (SST).

In his most recent paper (published in Atmosphere in 2026), Stefani updates earlier regression analyses comparing the solar aa index and CO₂ concentrations (or emissions) against observed global SST data from HadSST4.2. The aa index measures geomagnetic disturbances influenced by solar wind and activity, serving as a long-term proxy for solar variability.

Key findings from the analysis:

• Up to around 2000, the aa index alone predicts SST variations very well—often better than CO₂ alone, and comparably to models combining both.
• Post-2000 data shows a stronger apparent role for CO₂, but Stefani argues that fixing the solar (aa index) contribution based on robust pre-2000 regressions (around 0.04 K/nT) and then regressing the residual SST against CO₂ narrows the estimated Transient Climate Response (TCR)—the warming from a doubling of CO₂ under transient conditions—to 1.1–1.4 K.
• This TCR range is at the low end of the IPCC’s 2021 estimate (1.2–2.4 K) and aligns closely with other observationally constrained studies (e.g., Lewis & Curry 2018: 0.9–1.7 K; Scafetta 2023: 1.0–1.2 K for HadSST4.2).

Stefani’s approach refines earlier work (e.g., his 2021 paper), where double regressions gave a broader TCR range (0.6–1.6 K), sensitive to including recent data affected by events like strong El Niño periods and low aa values.For projections to 2100, Stefani uses refined aa index forecasts (based on ideas of planetary synchronization of the solar dynamo) and CO₂ scenarios assuming constant emissions (30–50 Gt/year, near current levels) plus a linear sink model. Even at higher sensitivities and unabated CO₂ growth, warming remains modest (~1 K additional by 2100 in pessimistic cases). Lower sensitivities suggest possible near-term cooling or flattening, with temperatures potentially staying below recent highs (e.g., 2024 peaks).

This work is discussed in Andy May’s blog.

Andy May, a retired petrophysicist and climate blogger (andymaypetrophysicist.com), published a detailed review on March 2, 2026, titled “Stefani on the Sun vs. CO2 as climate drivers”.

This post analyzes Frank Stefani’s February 2026 paper in Atmosphere (MDPI): “Solar and Anthropogenic Climate Drivers: An Updated Regression Model and Refined Forecast.”May’s article highlights Stefani’s findings that solar variability—proxied by the geomagnetic aa index (a measure of solar wind and geomagnetic disturbances, recorded since ~1868)—explains global sea surface temperature (SST) changes (HadSST4.2 dataset) remarkably well from ~1850 to ~2000, often outperforming CO₂ alone and matching or exceeding combined solar + CO₂ models.

_____________________________________________________________________________________

Stefani on the Sun vs. CO2 as climate drivers

From Andy May Petrophysicist

By Andy May

Frank Stefani, of the Helmholtz-Zentrum Dresden-Rossendorf, Institute of Fluid Dynamics, has published a very interesting new paper that compares the solar “aa” index and CO₂ emissions to global SST (sea surface temperatures using the HadSST4.2 dataset) and finds a CO₂ sensitivity (TCR or the “Transient Climate Response”) of 1.1 to 1.4K. This is at the low end of the IPCC TCR range of 1.2 to 2.4K (IPCC, 2021, p. 93), but quite close to the values calculated by Lewis and Curry and Nicola Scafetta (Lewis & Curry, 2018), (Scafetta, 2023), and (Lewis, 2023). Scafetta found a plausible range of TCR (versus HadSST4.2) of 1.0K to 1.2K and Lewis & Curry report a range of 0.9K to 1.7K for TCR versus HadCRUT4. The estimates are compared in Table 1.

Table 1. Estimates of TCR, or Transient Climate Response to a doubling of CO₂. Sources: (Stefani, 2026), (Lewis & Curry, 2018), (Scafetta, 2023), and (IPCC, 2021, p. 93).

TCR Estimates
Author	Best Estimate	Range: Low end	Range: High end	Note
Stefani	1.26K	1.1	1.4	2026
Lewis & Curry	1.2K	0.9	1.7	2018
Scafetta	1.1K	1	1.2	2023
IPCC AR6	1.8K	1.2	2.4	p. 93

All the estimates in table 1 are based on regression models of varying complexity and all attempt to account for the influence of solar variability. The Lewis and Curry estimate does not incorporate a solar activity proxy directly but does use the Atlantic Multidecadal Oscillation (AMO) as an indicator of natural climate variability, which is, in large part, solar variability. Stefani uses the aa index of geomagnetic activity, essentially how disturbed Earth’s magnetic field is by the sun. It is measured in nanoteslas (nT).

Nearly every observable form of solar variability is a magnetic phenomenon at its core, including sunspots, flares, and solar wind variability. Thus, the aa index is a good indicator of changes in the sun’s state and output. The aa index as a measure of solar-geomagnetic coupling has been measured consistently since 1868.

Table 1 shows that Stefani’s aa index + CO₂ model compares well to Lewis and Curry’s AMO + CO₂ model and Scafetta’s solar proxies + CO₂ model. Scafetta uses three estimates of total solar irradiance (TSI) in his study, although he acknowledges that variations in TSI may be only ~20% of the sun’s total influence on Earth’s climate. None of these observation-based models support the high-end IPCC TCR estimate of 2.4K per doubling of CO₂ or their best estimate of 1.8K, but all are near the lower end of the IPCC range. The Lewis (2023) correction to the Sherwood (2020) assessment of ECS and TCR relied upon in AR6 is not included in the table. However, after changing Sherwood’s subjective Bayesian assessment of multiple estimates of TCR to an objective Bayesian assessment, Lewis calculated a TCR of 1.37 to 1.4 depending upon the assumptions made. This is still close to the other observation-based objective estimates in Table 1.

Stefani found that the solar aa index can successfully predict HadSST4.2 up to 1990-2000 on its own. After 1990 to 2000 the role of CO₂ in the regression increases significantly. In figure 1, Stefani’s robust aa index weight of 0.04 K/nT (that is the HadSST4.2 anomaly temperature in °C per the aa index in nanoteslas) plus CO₂ with a sensitivity of 1.26K per doubling is shown as a thin dark gray line. It is compared to the HadSST4.2 anomaly (shown as black dots). His projected aa index plus CO₂ at 1.26K per doubling function to 2100 is shown as a red line. The maximum departure in 2023 and 2024 looks startling, but we need to remember that this follows two strong El Niños (2018-19 & 2023-24) and the Hunga Tonga volcanic eruption. In particular the El Niño from June 2023 to May 2024 was one of the strongest El Niños on record. The HadSST4.2 global anomaly has been falling since September 2024. Thus, a portion of the difference between the aa index plus CO₂ function and HadSST4.2 may just be ENSO, the Hunga Tonga eruption, and weather.

Stefani’s optimal model projection to 2100 assumes constant emissions of 30-50 Gt of CO₂ per year (roughly current levels) plus a simple linear carbon sink model and predicts a global SST increase of 0.6°C (1.1°F) compared to the standard HadSST4.2 reference period of 1961-1990. Using pessimistic parameters (high CO₂ emissions, a low sensitivity to the aa-index, and a high sensitivity to CO₂) yields a 2100 temperature increase of ~1K over 1960-1990. This is still a benign result.

A word on Transient Climate Response

The transient climate response is the modeled warming due to increasing the atmospheric CO₂ concentration by 1% each year until it doubles, which would take about 70 years. This is different from the commonly cited value of ECS, which is the equilibrium climate sensitivity, or the final warming due to a sudden doubling of CO₂. ECS is an untestable number, since it would take over 1,000 years for the atmosphere to completely come to equilibrium after CO₂ suddenly doubles. Given the implausible scenario, ECS can never be tested, except in a climate model. Thus, it is not a scientific quantity per Karl Popper (Popper, 1962). TCR on the other hand is very realistic and could be tested given enough time and effort.

Conclusions

The paper is a welcome addition to the growing group of observation-based estimates of climate sensitivity to CO₂. It is appropriate that Stefani chose to create his model around HadSST4.2. SSTs are more stable than air temperatures on land and they respond mainly to changes in insolation, whether due to cloud cover changes or changes in the sun itself. The changes in SST due to the greenhouse effect are smaller for the reasons discussed in my previous post. I recommend the paper, it is interesting, significant, and a good read.

Works Cited

IPCC. (2021). Climate Change 2021: The Physical Science Basis. In V. Masson-Delmotte, P. Zhai, A. Pirani, S. L. Connors, C. Péan, S. Berger, . . . B. Zhou (Ed.)., WG1. Retrieved from https://www.ipcc.ch/report/ar6/wg1/

Lewis, N. (2023, May). Objectively combining climate sensitivity evidence. Climate Dynamics, 60, 3139-3165. https://doi.org/10.1007/s00382-022-06468-x

Lewis, N., & Curry, J. (2018). The Impact of Recent Forcing and Ocean Heat Uptake Data on Estimates of Climate Sensitivity. Journal of Climate, 31, 6051-6071. DOI: https://doi.org/10.1175/JCLI-D-17-0667.1.

Popper, K. R. (1962). Conjectures and Refutations, The Growth of Scientific Knowledge. New York: Basic Books. Retrieved from http://ninthstreetcenter.org/Popper.pdf

Scafetta, N. (2023). Empirical assessment of the role of the Sun in climate change using balanced multi-proxy solar records. Geoscience Frontiers, 14(6). Retrieved from https://www.sciencedirect.com/science/article/pii/S1674987123001172

Sherwood, S. C., Webb, M. J., Annan, J. D., Armour, K. C., J., P. M., Hargreaves, C., . . . Knutti, R. (2020, July 22). An Assessment of Earth’s Climate Sensitivity Using Multiple Lines of Evidence. Reviews of Geophysics, 58. https://doi.org/https://doi.org/10.1029/2019RG000678

Stefani, F. (2026). Solar and Anthropogenic Climate Drivers: An Updated Regression Model and Refined Forecast. Atmosphere, 17(3). https://doi.org/10.3390/atmos17030252