Abstract
Atmospheric aerosols and their impact on cloud properties remain the largest uncertainty in the human forcing of the climate system. By increasing the concentration of cloud droplets ($N_d$), aerosols reduce droplet size and increase the reflectivity of clouds (a negative radiative forcing). Central to this climate impact is the susceptibility of cloud droplet number to aerosol ($\beta$), the diversity of which explains much of the variation in radiative forcing in global climate models. This has made measuring $\beta$ a key target for developing observational constraints of the aerosol forcing. While the aerosol burden of the clean, pre-industrial atmosphere has been demonstrated as a key uncertainty for the aerosol forcing, here we show that the behaviour of clouds under these clean conditions is of equal importance for understanding the spread in radiative forcing estimates between models and observations. This means that the uncertainty in the aerosol impact on clouds is, counterintuitively, driven by situations with little aerosol. Discarding clean conditions produces a close agreement between different model and observational estimates of the cloud response to aerosol, but does not provide a strong constraint on the radiative forcing from aerosol-cloud interactions. This makes constraining aerosol behaviour in clean conditions a key goal for future observational studies.
Introduction: The Aerosol-Cloud Uncertainty
Climate scientists have long struggled to quantify how much aerosols—such as soot or sulfates—cool the planet by making clouds more reflective. This cooling effect, known as Radiative Forcing from aerosol-cloud interactions (RFaci), is one of the most significant uncertainties in current climate projections. When aerosols are added to the atmosphere, clouds develop more droplets ($N_d$), making them smaller, more numerous, and more reflective.
The core of this uncertainty lies in a parameter called susceptibility ($\beta$), which measures how sensitive cloud droplet numbers are to changes in aerosol concentration. A high $\beta$ implies that even small amounts of pollution can significantly cool the planet, while a low $\beta$ suggests a negligible effect. Currently, global climate models (GCMs) and satellite observations provide a wide range of values for $\beta$, complicating efforts to predict future climate sensitivity.
Methodology: Measuring Susceptibility
To address this discrepancy, researchers compared satellite-derived observations from MODIS with a suite of Global Climate Models (GCMs) from the AeroCom project. The study introduced a critical distinction between $\beta$ (calculated across all conditions) and $\beta_{hi}$ (calculated only for "polluted" conditions with a high Aerosol Index).
The analysis utilized:
- Regression Analysis: Linear regression on joint histograms of aerosol proxies and cloud droplet numbers to determine the slope ($\beta$).
- Parcel Modeling: Simulations of different updraught speeds and aerosol types (e.g., sea salt vs. sulfates) to understand the underlying physics of droplet formation.
- Error Analysis: Investigation of "regression dilution," where measurement errors in clean conditions (low signal-to-noise ratio) artificially deflate the calculated $\beta$ values.
Results: The Clean Air Divergence
The study reveals a counterintuitive truth: the massive uncertainty in RFaci estimates is driven not by polluted skies, but by how clouds behave in clean, low-aerosol conditions.
Surprisingly, when focusing solely on polluted air ($\beta_{hi}$), almost all models and satellite observations converge on a susceptibility value between 0.5 and 0.6. The disagreement occurs almost entirely in clean air, where satellite data shows a "shallowing" effect (low $\beta$) while models often maintain high sensitivity.
This divergence is largely a signal-to-noise problem. In clean air, the aerosol signal is weak compared to the "noise" of the Earth's surface reflectance. This leads to higher relative errors in satellite retrievals, which biases the $\beta$ calculation toward zero—a phenomenon known as regression dilution.
Implications for Radiative Forcing
Because human-caused cooling is calculated by comparing today's atmosphere to the "clean" pre-industrial baseline, the way models handle these clean conditions dictates the final forcing estimate. If a model assumes high sensitivity in clean air, it predicts a much stronger cooling effect from historical pollution.
Key findings include:
- Uncertainty Reduction: $\beta$ explains approximately 50% of the uncertainty in RFaci estimates.
- Halving the Error: Constraining cloud behavior in clean air to the same level of accuracy as polluted air could reduce the uncertainty in aerosol-cloud forcing by half.
- New Constraints: By accounting for the clean-air bias, the study suggests a likely range for RFaci of -0.45 to -1.2 $W m^{-2}$.
Conclusion and Future Directions
The primary limitation remains the sensitivity of current satellite instruments, which struggle to distinguish small aerosol signals from surface background noise in pristine environments like the remote Southern Ocean.
Future work should focus on:
- Advanced Instrumentation: Utilizing multi-angle polarimeters (e.g., PACE mission) to better separate aerosol signals from surface reflectance.
- In-situ Integration: Combining satellite data with sensitive ground-based and aircraft measurements in pristine environments.
- Refining Model Physics: Improving GCM representations of "aerosol-limited" regimes where droplet formation is highly sensitive to individual particles.
Summary updated