the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 30 Sep 2026
Satellite-based constraints on temperature-dependent aerosol–cloud feedback over a clean boreal forest and a rural continental site
Abstract. Temperature-dependent biogenic aerosol formation has been proposed as a negative climate feedback mechanism. However, the indirect cloud-albedo radiative feedback remains difficult to quantify from observations because in situ observations from clouds are sparse and satellite-derived cloud properties are sensitive to retrieval assumptions, cloud regime, and meteorology. In our attempt to tackle this issue, we combine multiyear field observations with carefully sampled satellite retrievals for two diverse locations: a boreal forest site, Hyytiälä in Finland, and a continental site, Southern Great Plains in USA. We focus on low-level liquid clouds and evaluate the impact of sampling criteria (general cloud-quality filters, a local quality filter, and an adiabatic-consistency criterion) and different fitting methods on the estimation of the temperature-dependent radiative effects of biogenic aerosols. We found that if the adiabaticity requirement is not used, the indirect radiative effects are likely underestimated because non-ideal cloud scenes dilute the observed aerosol-cloud response. Based on carefully selected data sets, we estimate that the clear-sky direct radiative feedback caused by biogenic aerosols is approximately -0.6 Wm−2 °C−1 in a clean environment and about half of that (-0.3 Wm−2 °C−1) in a rural region. The indirect radiative feedback of these aerosols is in the same range with the direct radiative feedback (-0.6 Wm−2 °C−1) in the clean region but it is almost three times stronger (-0.8 Wm−2 °C−1) in the rural region. However, the quantification of the temperature-dependent aerosol–cloud feedback is more straightforward in clean environments than in rural ones, where other emissions can drive the aerosol load and number.
Competing interests: At least one of the (co-)authors is a member of the editorial board of Atmospheric Chemistry and Physics.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.- Preprint
(6512 KB) - Metadata XML
-
Supplement
(3719 KB) - BibTeX
- EndNote
Status: open (until 11 Nov 2026)
- RC1: 'Comment on egusphere-2026-4750', Anonymous Referee #3, 19 Oct 2026 reply
-
RC2: 'Comment on egusphere-2026-4750', Anonymous Referee #1, 23 Oct 2026
reply
The manuscript investigates temperature-dependent aerosol radiative feedbacks at two sites, Hyytiälä (clean boreal forest) and SGP (rural continental), using in situ measurements, satellite retrievals, and cloud parcel model simulations. The authors assess how cloud sampling criteria and regression methods affect aerosol-cloud susceptibility and radiative feedback estimates. The manuscript is generally well organized and well written. I think its main contribution lies in the systematic evaluation of cloud sampling criteria and their influence on the estimated aerosol-cloud relationships. I only have mostly minor comments before publication.
- L146. The adiabatic-consistency filter of CER3.7 ≥ CER2.1 is a useful screening approach, but the difference between these retrievals is also sensitive to retrieval noise, cloud heterogeneity, and drizzle. Therefore, this criterion does not necessarily identify adiabatic clouds. I suggest clarifying this limitation where the filtered scenes are interpreted as having an adiabatic structure.
- Because CER also enters the CDNC calculation, selecting pixels based on differences between CER channels may also affect the fitted CDNC–N100 relationship. The increase in susceptibility after filtering could partly reflect differences in the sampled cloud population, in addition to improved retrieval quality. It would be useful to compare the characteristics of retained and rejected scenes and, if feasible, compare the fitted slopes using a common set of overpasses.
- L359. Only overpasses with at least 30 valid MODIS pixels are retained, while the reported pixel retention fractions are based on two example overpasses. Please report the numbers of pixels, overpasses, and independent days retained at each screening stage for the full dataset.
- The manuscript refers to non-overlapping confidence intervals when comparing the susceptibility slopes and describes the increase in indirect feedback at Hyytiälä as statistically significant. Because the adiabatic-filtered sample is a subset of the broader sample, I suggest testing the difference between the estimates directly rather than inferring significance from separate confidence intervals.
- L305. The indirect radiative feedback includes liquid-cloud fraction, while the main comparison uses the clear sky direct feedback. For consistency, please also compare the indirect feedback with the reported all-sky direct estimates. This would change the stated relative magnitudes, particularly at SGP.
- The observed temperature dependence is interpreted as evidence for a biogenic aerosol contribution. If available, aerosol composition measurements could provide additional support for the proposed biogenic contribution. Otherwise, I suggest describing the estimates more cautiously as temperature-associated aerosol radiative responses and tempering the attribution in the conclusions.
- The cloud parcel model uses fixed thermodynamic conditions and an updraft velocity of 0.3 m/s. However, this velocity is justified using positive Doppler-lidar velocities at Hyytiälä and vertical velocity standard deviations at SGP. Please clarify this choice and consider testing the sensitivity of the modeled susceptibility to a reasonable range of updraft velocities based on Fig. S3, particularly when interpreting the model-observation agreement.
- Please provide additional justification for excluding measurements with total particle number concentrations above 8000 cm/3, as this threshold may remove new particle formation events that are relevant to the temperature-dependent aerosol response.
- Eq. 1. Please check the units of γ.
Citation: https://doi.org/10.5194/egusphere-2026-4750-RC2 -
RC3: 'Comment on egusphere-2026-4750', Anonymous Referee #4, 25 Oct 2026
reply
The manuscript estimates temperature-dependent direct aerosol and indirect cloud-albedo feedbacks at two contrasting sites, extending the earlier boreal framework to SGP. Cloud selection, regression comparisons, and parcel modeling support and test those estimates. The combined observational and modeling approach is valuable, but the interpretation and methodological choices below require clarification. I recommend a major revision.
1. Abstract and Introduction
The motivation is clear, but the abstract presents biogenic attribution and systematic underestimation without the effective-radius filter more conclusively than the analysis supports. Distinguish observed temperature associations, cloud-selection effects, and inferred climate feedbacks. Explain explicitly what advances beyond the earlier Hyytiälä study and what the two-site comparison can establish.
2. Section 2.1: Aerosol observations, CCN proxy and screening
Justify using N100 rather than N70, particularly in light of Virtanen et al. (2025), already cited in the manuscript. Their collocated in situ measurements yield strong Nd-N70 relationships, with susceptibilities of 0.89 ± 0.02 at Puijo, 0.84 ± 0.02 at Pallas and 0.82 ± 0.04 at Zeppelin (95% confidence bounds; their Fig. 2). These substantially exceed the present estimates, including after cloud filtering.
A comparison using N70 would help establish whether the cutoff contributes to this discrepancy or affects the comparison between Hyytiälä and SGP. However, Virtanen et al. report only a minor sensitivity to their proxy's lower size limit; the discrepancy should therefore not be attributed to the cutoff alone. Discuss differences in cloud regime, activation conditions, fitting methodology and, particularly, collocation: their aerosol and droplet measurements sample the cloud directly, whereas the present study relates surface aerosol to satellite-retrieved cloud properties. Surface-cloud decoupling could weaken the latter relationship.
Report the fractions rejected by the aerosol concentration limits and lognormal-fit criterion at each site, including any preferential removal of particular aerosol conditions. Validate fitted N100 against direct integration using agreement and bias statistics, rather than statistical significance alone.
3. Sections 2.2 and 3.2: Cloud selection and mixing
The physical meaning of the effective-radius criterion requires clarification. Adiabatic re does not imply adiabatic LWC. Under inhomogeneous mixing, droplets can evaporate completely, reducing LWC and droplet number, while surviving droplets retain adiabatic sizes. This behavior is typical of convective clouds and preserves the droplet-size signature needed to retrieve Nd and its aerosol susceptibility.
The stronger relationship after applying re,3.7 ≥ re,2.1 may therefore reflect preferential selection of convective clouds or updraft regions where this signature is preserved, in addition to improved retrieval quality. Discuss this possible explanation and avoid describing the selected cases as having demonstrated adiabatic LWC.
Report the fractions of pixels, overpasses, and aerosol-cloud matches retained across the full dataset at both sites, not just for the illustrative cases, to establish which cloud population the resulting susceptibility represents.
4. Section 2.4: Representative updraft velocity
The representative updraft should account for both the nonlinear dependence of Nd on w and the greater cloud volume supplied by stronger updrafts. A simple average of positive velocities - or a velocity standard deviation - is therefore not automatically the appropriate input for droplet activation.
Calculate the cloud-volume-weighted activation response over the observed updraft distribution, or determine an effective velocity that reproduces that response. Assess how this treatment changes both simulated Nd and the Nd-N100 susceptibility. It could improve the modeled susceptibility and its agreement with observations, providing a more physically justified comparison than prescribing 0.3 m s-1 at both sites.
5. Sections 3.1 and 3.3: Temperature dependence at SGP
Section 3.1 reports a strong positive NO2-adjusted N100-temperature relationship at SGP, whereas lines 419-425 interpret the absence of a clear temperature dependence in the cloud-matched sample as indicating that temperature-dependent emissions do not drive N100. Clarify whether this contrast results from different sampling, NO2 adjustment, or other factors. The absence of a visible relationship in the smaller cloud-matched sample alone does not support that conclusion.
For Figures 2-3, specify how the summer means are calculated, whether aerosol and temperature averages use identical valid times, and whether the SGP markers are NO2-adjusted.
6. Section 3.3: Potential explanations for differences between sites
The discussion should consider plausible physical and observational explanations for the apparent site differences, while recognizing the overlapping susceptibility estimates.
Surface-cloud decoupling could weaken the Nd-surface-N100 relationship because surface aerosol poorly represents cloud-base CCN. This mechanism has been demonstrated at SGP (Su et al., 2024). Spatial separation and the ±2 h collocation window could further reduce representativeness. Differences in cloud regime, updraft distributions, and entrainment-mixing could also affect susceptibility.
Aerosol source contributions may help explain differences in the temperature response. SGP has a larger anthropogenic contribution than Hyytiälä, whereas biogenic precursor emissions generally respond more strongly to temperature than many anthropogenic emissions. A less temperature-sensitive anthropogenic background could therefore dilute the fractional temperature response of the total aerosol population at SGP. Atmospheric processing and anthropogenic-biogenic interactions complicate this expectation, so it should be discussed as a possible explanation rather than assumed.
Differences in aerosol size distributions and hygroscopicity may additionally make N100 an unequal proxy for the activating population at the two sites.
These are possible explanations warranting discussion, not mechanisms demonstrated by the present analysis.
Minor corrections
Correct logarithmic units throughout, including line 401 and the axes of Figures 3 and 6. Logarithms are dimensionless, and temperature slopes of ln(N100) have units K-1, not cm-3 K-1. Formally, the logarithm should be written as ln[N100/(1 cm-3)].
Correct the Table 1 caption: direct feedback uses the AOD-temperature coefficient, not the N100-temperature coefficient.
References cited in this review
Virtanen, A. et al. (2025). High sensitivity of cloud formation to aerosol changes. Nature Geoscience, 18, 289-295. doi:10.1038/s41561-025-01662-y.
Su, T., Li, Z., Henao, N. R., Luan, Q. and Yu, F. (2024). Constraining effects of aerosol-cloud interaction by accounting for coupling between cloud and land surface. Science Advances, 10, eadl5044. doi:10.1126/sciadv.adl5044.
Citation: https://doi.org/10.5194/egusphere-2026-4750-RC3
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 337 | 153 | 48 | 538 | 78 | 41 | 35 |
- HTML: 337
- PDF: 153
- XML: 48
- Total: 538
- Supplement: 78
- BibTeX: 41
- EndNote: 35
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
This paper analyses the temperature-dependent aerosol-radiative feedbacks over two contrasting sites: Hyytiälä, characterized by a relatively clean boreal forest environment, and a rural continental environment of the Southern Great Plains, USA. The authors combined ground-based aerosol measurements with satellite-derived cloud properties to quantify the temperature dependence of aerosol concentrations and aerosol–cloud interactions, and associated radiative feedbacks. Particular attention is given to how the selection of clouds satisfying an adiabatic-consistency criterion influences the inferred aerosol–cloud susceptibility and, consequently, the estimated cloud-albedo radiative feedback.
The paper is well-written, and the overall presentation is good. My only major concerns are around methodology and the uncertainties in the estimated aerosol-radiative feedbacks, as mentioned below.
Major Comments
The Doppler-lidar observations used to prescribe the vertical velocity in the parcel model would be useful to show, particularly the distribution of observed vertical velocities. Also, a prescribed vertical velocity may not sufficiently represent the observed variability in aerosol-cloud interactions, especially at SGP, where stronger updrafts may occur. Furthermore, the differences in using positive Doppler-lidar velocities at Hyytiälä and the standard deviation of vertical velocity at SGP should be explained.
I encourage the authors to simulate the parcel model experiment at different updrafts, e.g., 0.1, 0.3, 0.5 m/s. This would establish whether the model-observation agreement is robust or depends on the selected velocity. Also, I wonder whether the model reproduces the observed CDNC magnitude, in addition to aerosol-cloud susceptibility.
For the adiabatic cloudy conditions, the authors used the difference between effective radii at 3.7 and 2.1 microns. However, can the authors clarify whether the stronger aerosol–cloud susceptibility obtained after applying this criterion reflects the exclusion of clouds inconsistent with adiabatic conditions or partly differences in the cloud populations retained before and after filtering? A brief comparison of the cloud properties between these samples would help support the interpretation.
The temperature dependence of N100 is an important component of the reported radiative feedbacks. However, aerosol conditions can be greatly influenced by meteorological conditions, transport, and wet scavenging. I suggest that the authors clarify or comment on the extent to which these factors could influence the reported aerosol-temperature relationship. This could be particularly crucial for SGP, where multiple aerosol sources are available. A discussion of these uncertainties would help establish the robustness of the inferred radiative feedbacks.
The estimated aerosol-cloud radiative feedbacks are estimated from the observed aerosol-cloud susceptibility and the temperature dependence of aerosol concentrations. However, these calculations account mainly for the changes in cloud albedo associated with CDNC, without considering potential changes in LWC, cloud fraction (or precipitation). I recommend that the authors clarify how these assumptions limit the interpretation of the estimated feedbacks. It can also be useful to mention which sources of uncertainty are included in the reported confidence intervals (CI)?
At SGP, the aerosol-cloud feedback has a wide CI, hence a possibility of considerable uncertainty in its magnitude. Also, the results appear to be sensitive to adiabatic-consistency criteria within the sampling domain. Could the authors discuss how these uncertainties affect the robustness of the inferred feedback at this site? Clarifying the number of observations retained after the different filtering steps would also help assess the statistical reliability of the results.
Minor comments
Abstract: “biogenic aerosols” may imply stronger source attribution than the manuscript observed. Distinguish the measured aerosol-temperature relationships from their proposed biogenic contribution
Sect. 2.2: Please ensure that the terms “adiabatic”, “near-adiabatic” and “adiabatically consistent” are used consistently, because the retrieval criterion may not establish adiabaticity directly.
Sect. 2.4: A brief explanation of the differences in Doppler-lidar measurements between the sites would be helpful and will motivate the physical reasoning behind selecting 0.3 m/s updraft velocity.
Conclusions: Overall well-written. Distinguish the estimated cloud-albedo feedback from the broader aerosol-cloud feedback. Please also consider adding a paragraph or so describing sources of uncertainties that may affect the presented analysis.