the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 27 Jul 2026
A global assessment of warm-phase convective invigoration from aerosol-cloud interactions. Part 2: Observations
Abstract. The hypothesis of warm-phase convective invigoration from aerosol-cloud interactions posits that aerosol pollution increases the cloud-droplet number concentration in the liquid layer of convective clouds, causing rising air parcels to consume supersaturation more readily. This leads to smaller supersaturations, larger buoyancies, and faster updrafts. We consider this mechanism to be substantial if it increases the updraft speed w by 25 % or more when the cloud-droplet number concentration increases by a factor of five. A theoretical prerequisite for substantial invigoration is that supersaturation must exceed ~0.9 % if w = 3 m s-1 and ~5 % if w = 7 m s-1. This work searches for evidence of these high supersaturations in 12 globally distributed aircraft campaigns that collectively measure shallow and deep convection over ocean, coasts, and land. The quasi-steady approximation is applied to estimate supersaturation in liquid elements of convective updrafts with ~100 m horizontal resolution. Across the 13,342 total samples, 463 samples have supersaturation between 1 % and 5 %, seven samples have supersaturation between 5 % and 9 %, and no samples have supersaturation exceeding 9 %. Cloud susceptibility to warm-phase invigoration is quantified as the fraction of updrafts with supersaturation exceeding the threshold for substantial invigoration. 62 % of oceanic updrafts are susceptible when 1 ≤ w < 2 m s-1, but only 1 % are susceptible when w > 5 m s-1. 14 % of coastal and continental updrafts are susceptible when 1 ≤ w < 2 m s-1, but only 0.05 % are susceptible when w > 5 m s-1. These results imply that precipitating shallow cumuli over the ocean are uniquely susceptible to warm-phase convective invigoration.
- Preprint
(8352 KB) - Metadata XML
- BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2026-3986', Anonymous Referee #1, 09 Sep 2026
-
RC2: 'Comment on egusphere-2026-3986', Anonymous Referee #2, 11 Sep 2026
The manuscript investigates aerosol-cloud interactions, particularly focussing on warm-phase convective invigoration. I was able to smoothly follow the assumptions underlying the analysis and assess how they propagate through the methodology to the main conclusions of the study.
Overall, I appreciate the manuscript’s careful writing, clear structure, and scientific rigour. Notably, the authors demonstrated meticulous attention to detail in harmonising observations from multiple field campaigns. They adopted a conservative approach to data selection, excluding flight segments and conditions that could invalidate the assumptions underlying supersaturation calculations. They invest a lot of effort in identifying potential uncertainties, clearly state the methodology’s assumptions, and discuss their associated limitations.
I generally support publication, but I believe two additional points should be addressed to further demonstrate the robustness of the main findings against potential measurement and analysis biases.
1- It’s very important to quantify and propagate measurement uncertainties into the reported invigoration statistics. This will provide a more comprehensive understanding of the uncertainties associated with the findings.
The manuscript presents a transparent and reasonably rigorous discussion of the sources of uncertainty in the parameters used to calculate supersaturation and their associated error propagation. However, the resulting uncertainty of approximately 20–60% in supersaturation is substantial and could potentially impact some of the reported conclusions. While the authors have partially addressed this issue by adopting relatively broad supersaturation categories, it would be beneficial to quantify directly how measurement uncertainties propagate into the final statistical results rather than solely relying on the width of the supersaturation bins.
Therefore, I strongly recommend performing a Monte Carlo-type uncertainty analysis. For instance, the relevant input parameters for each field campaign could be repeatedly jittered according to their estimated measurement uncertainties, leading to recalculations of supersaturation and the derived statistics for each realisation. The perturbation distributions should ideally reflect the known uncertainty characteristics of each measurement; if only bounded uncertainty estimates are available, an appropriately justified uniform distribution could be used.
Repeating this analysis over a sufficiently large ensemble would enable the authors to quantify the robustness of the reported relationships to measurement uncertainty. Instead of presenting a single curve for quantities such as f, the manuscript could, for example, display the ensemble median alongside a shaded confidence/uncertainty interval. Moreover, this approach would provide uncertainty estimates for higher-level conclusions, including the fraction of clouds classified as susceptible to substantial invigoration for each cloud type, as well as other aggregate statistics reported throughout the manuscript.
2. Potential contamination by cloud-edge or cloud-free samples, particularly for shallow cumulus.
The analysis in Fig. 8 illustrates the importance of high spatiotemporal-resolution measurements for this type of study. Even within a cumulus congestus penetration, a cloud-free or cloud-edge parcel can enter the analysed time series despite the application of multiple filtering criteria. This highlights the difficulty of completely excluding such samples and raises the question of whether the problem could be more pronounced for shallow cumulus, where individual cloud penetrations are generally more limited in spatial and temporal extent.
Compared to deeper cumulus or congestus clouds, shallow cumuli generally have shorter in-cloud sampling periods and potentially a larger relative contribution from cloud-edge regions. Consequently, contamination by cloud-edge or cloud-free samples may become particularly important for this cloud category. This issue deserves careful consideration, as one of the notable conclusions of the manuscript is that precipitating shallow cumuli over the ocean are usually susceptible to substantial warm-phase invigoration.
I am unsure which cloud microphysical measurements are consistently available across all the field campaigns included in the study. However, many aircraft datasets include measurements from instruments such as the CDP and/or FCDP, and some configurations provide particle-by-particle (PbP) data at very high temporal resolution. Where available, these measurements could provide a powerful and largely independent criterion for identifying genuine in-cloud sampling and removing samples affected by cloud edges or intermittent cloud-free air.
I recommend that the authors investigate whether high-resolution cloud-droplet measurements can be used as an additional cloud-presence/cloud-edge filter, particularly for the shallow-cumulus subset. At the very least, the authors should quantify the sensitivity of the shallow-cumulus results to a more restrictive definition of in-cloud sampling. If sufficiently high-resolution measurements are not available consistently across all flights or campaigns, this sensitivity analysis could be performed for a representative subset of shallow-cumulus flights for which such measurements (e.g. PbP) are available. Even a limited analysis of this kind would provide a useful demonstration of the extent to which potential cloud-edge or cloud-free contamination may influence the reported results and would help establish the robustness of the conclusions drawn for shallow cumulus clouds.
Overall assessment
This is a carefully conducted study that effectively utilises a challenging multi-campaign observational dataset. The conservative data-selection strategy, transparent treatment of assumptions, and detailed discussion of limitations are important strengths of the manuscript. My two comments above are primarily concerned with establishing the robustness of the conclusions against (1) propagated measurement uncertainty and (2) possible cloud-edge/cloud-free contamination, particularly for shallow cumulus. Addressing these points would, in my view, provide stronger quantitative support for the manuscript’s central conclusions and further increase confidence in the reported evidence for warm-phase convective invigoration.
Citation: https://doi.org/10.5194/egusphere-2026-3986-RC2
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 151 | 71 | 19 | 241 | 15 | 15 |
- HTML: 151
- PDF: 71
- XML: 19
- Total: 241
- BibTeX: 15
- EndNote: 15
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Combining observations from twelve aircraft campaigns is a major achievement, and the effort to harmonize and screen these measurements is impressive. Therefore, the observational compilation and supersaturation analysis are valuable and merit publication. However, the quantitative interpretation in terms of convective invigoration requires a major revision. The thresholds, susceptibility fractions, and cloud-regime conclusions should be restricted explicitly to the idealized analytical model, and the broader claims concerning shallow versus deep convective invigoration should be withdrawn or substantially qualified.
I have several groups of comments:
My principal concern is the quality and relevance of the metrics used to quantify invigoration. Your analytical metric is mathematically coherent for a narrowly defined quantity: the fractional change in terminal updraft speed predicted for an idealized thermal after an imposed fivefold increase in droplet number. However, I do not think it is a general metric of convective invigoration, due to the following reasons:
The authors define substantial invigoration as a 25% increase in updraft speed. The required increase in kinetic energy per unit mass is:
ΔKE/m = ½(w_polluted² − w_clean²)
= ½[(1.25 w_clean)² − w_clean²]
= 0.28125 w_clean².
Consequently, a 15 m s⁻¹ updraft must receive 25 times more additional kinetic energy per unit mass than a 3 m s⁻¹ updraft to satisfy the same criterion. A weak updraft can therefore be classified as substantially invigorated after receiving relatively little additional energy, whereas a strong updraft can receive much greater aerosol-induced buoyancy work without achieving a 25% increase in speed.
Thus, the result that weak updrafts are more susceptible follows partly from the selected fractional-velocity metric. More precisely, weak updrafts are more likely to experience a large fractional change in terminal speed. This does not mean that they necessarily experience the largest energetic invigoration.
Your calculation compares two terminal velocities after instantaneously increasing the droplet number by a factor of five while holding the liquid-water content, normalized spectral shape, thermal size, and environment fixed. It does not follow the vertical evolution of activation, condensation, condensate loading, collision–coalescence, precipitation removal, drag, or cloud survival.
These processes determine how aerosol-induced energy is expressed. Invigoration can appear as deeper or longer-lived ascent, increased work against drag, delayed precipitation, greater active updraft area, or enhanced convective mass flux without producing a large local increase in updraft speed.
The terminal response also assumes that sufficient vertical distance is available for the polluted thermal to approach its new equilibrium velocity. Your time-dependent calculation indicates that attaining most of this response may require ascent through approximately twice the thermal diameter. For the assumed radius range of 200–1000 m, this corresponds approximately to 0.8–4 km. This distance should be compared with the actual vertical depth available after the imposed droplet perturbation. Some shallow clouds may not have sufficient depth to realize the predicted terminal response, whereas deep warm-base clouds may have several kilometers over which the additional buoyancy can act. This issue is directly relevant to the conclusion that shallow clouds are the most susceptible.
I therefore suggest distinguishing explicitly among:
The fraction f is not the observed fraction of updrafts that became invigorated. It is the fraction of retained one-second samples for which the steady-terminal-thermal model predicts Δw/w ≥ 0.25 under an imposed fivefold increase in droplet number.
I suggest calling f a “model-defined fractional-speed susceptibility index” and defining it as:
“The fraction of analyzed quasi-steady cloud-interior samples for which the steady-terminal-thermal model predicts at least a 25% increase in terminal updraft speed under an imposed fivefold increase in droplet number.”
The reported 2.5–97.5% range should not be interpreted as a complete confidence interval. It reflects only the assumed distributions of thermal radius and drag coefficient. It does not include measurement uncertainty, autocorrelation among consecutive samples, campaign-to-campaign variability, uncertainty in the fivefold droplet perturbation, or structural uncertainty in the terminal-thermal model.
The suitability of the sampled clouds for developing high supersaturation needs to be evaluated more explicitly. High supersaturation relevant to aerosol invigoration is expected most readily when several conditions coincide: a very clean initial aerosol background, a warm cloud base, a sufficiently deep liquid-cloud layer, and enough collision–coalescence and precipitation removal to reduce the surface area of the pre-existing droplets.
According to Figure 1, the campaigns sampling congestus and deep convection were not located in the very clean remote tropical-ocean regime. The continental and coastal deep-cloud campaigns occurred in regions with appreciable climatological sulfate. The maritime congestus and deep-cloud campaigns near the Philippines, the Gulf of Mexico, and the Caribbean also do not capture the cleanest aerosol conditions that dominate the open oceans, as shown in the figure.
Conversely, the campaigns in the cleanest geographical environments primarily sampled shallow clouds. CSET and EUREC⁴A sampled shallow trade cumuli, RICO sampled precipitating shallow cumuli, and SOCRATES sampled shallow clouds in cold-sector Southern Ocean conditions. These clouds do not provide the combination of a very clean aerosol background, a warm cloud base, and a deep warm layer required to test high supersaturation in deep tropical convection.
Therefore, the dataset appears to lack the cloud regime most relevant to the disputed mechanism: deep, warm-based convection developing in an extremely clean tropical ocean environment. The absence of high supersaturation in the sampled deep clouds cannot be extrapolated confidently to a regime that was not represented.
Figure 1 also shows a long-term climatological surface sulfate concentration, not the aerosol population actually entering each sampled cloud. It cannot establish the CCN or UAP concentration during an aircraft penetration.
Moreover, the “ocean” composite combines shallow trade cumuli, precipitating shallow cumuli, congestus, deep convection, and cold-sector Southern Ocean clouds. These regimes have very different cloud-base temperatures, warm-cloud depths, aerosol populations, and precipitation histories. A susceptibility fraction calculated from their combined one-second samples does not directly demonstrate that “precipitating shallow cumuli over the ocean are uniquely susceptible.” That conclusion requires separate results by cloud type and precipitation state.
The role of ultrafine aerosol particles is essential to both the physical mechanism and the validity of the quasi-steady supersaturation estimate.
In an initially very clean cloud, collision–coalescence and precipitation can reduce the number and total surface area of the original droplets. Supersaturation may then rise sufficiently for UAP that did not activate at cloud base to undergo secondary activation aloft. The newly activated droplets increase the total condensational surface area, consume supersaturation, release additional latent heat, and can increase buoyancy. This is the central warm-phase pathway proposed for UAP-induced invigoration.
The quasi-steady approximation assumes that droplet number and total droplet diameter concentration remain approximately constant over the supersaturation-adjustment timescale. It is therefore not valid during secondary activation. In a high-S cloud, the method is valid only after activation ceases, either because all UAP capable of activating at that supersaturation have been exhausted or because supersaturation has fallen below their activation thresholds.
Requiring the aircraft to be more than 250 m above the estimated lifting condensation level does not demonstrate that this condition has been satisfied. UAP can activate kilometers above cloud base. The study provides no measurements of interstitial UAP or CCN and therefore cannot establish that the relevant aerosol reservoir had been exhausted in the retained high-S samples.
This creates an important interpretive problem. If UAP are still activating, the quasi-steady method is not valid. Excluding observations within 250 m of cloud base is a reasonable and necessary step because primary droplet activation there violates the quasi-steady assumption. However, this filter addresses cloud-base activation only. It does not exclude secondary activation of UAP higher in the cloud, where supersaturation can increase after collision–coalescence and precipitation have reduced the surface area of the pre-existing droplets.
Ongoing secondary activation violates the quasi-steady supersaturation assumption because droplet number and total droplet diameter concentration are changing. It also violates the terminal-velocity assumption because the associated condensation and buoyancy perturbation initiate acceleration. Even after activation has ceased and supersaturation has returned to quasi-steady conditions, the thermal may still be adjusting dynamically toward a new terminal velocity. Thus, locally weak acceleration does not necessarily demonstrate that the effects of a recent activation event have fully disappeared.
Both assumptions may again become approximately valid only after the relevant UAP have been exhausted or supersaturation has fallen below their activation thresholds, and after the subsequent dynamical adjustment has been completed. In that case, however, the observed cloud element may already represent an aerosol-modified state. Applying a further fivefold increase in droplet number estimates its susceptibility to an additional hypothetical perturbation, not necessarily the response from the original clean-cloud state. Because the study does not measure the interstitial UAP population or document where secondary activation begins and ends, excluding the cloud-base activation layer alone is insufficient to establish these conditions.
The equal-LWC assumption does not invalidate the observed supersaturation distributions, but it does invalidate the use of the calculated Δw as a quantitative estimate of the net aerosol effect on updraft speed in a developing precipitating cloud. By forcing condensate loading to be identical in the clean and polluted thermals, the model removes the aerosol-induced differences in collision–coalescence, rain formation, condensate retention, and precipitation unloading that contribute directly to buoyancy and acceleration. This is not simply an omitted uncertainty; it suppresses a central component of the dynamical response by construction.
The consequence is that the calculated Scrit is neither a necessary nor a sufficient physical threshold for substantial invigoration in real clouds. A sample with S > Scrit may fail to achieve the predicted acceleration if the polluted thermal retains additional condensate and experiences greater loading. Conversely, a sample with S < Scrit may still undergo a dynamically important aerosol response through altered precipitation, unloading, cloud survival, or cumulative buoyancy work. The direction of the error cannot be corrected with a single factor because it depends on the evolving precipitation and loading history.
This directly invalidates three principal interpretations in the study:
The inconsistency is amplified by the different timescales in the calculation. Equal LWC is imposed at the instant droplet number is increased, whereas the predicted terminal response may require 0.8–4 km of subsequent ascent. Over this distance, the clean and polluted droplet spectra, rainwater production, sedimentation, and retained condensate cannot reasonably be assumed to remain equal. The study therefore combines an instantaneous equal-LWC perturbation with a long-distance terminal response while excluding the microphysical evolution that occurs between them.
Accordingly, the supersaturation observations remain valuable, and the analytical calculation may still be presented as an idealized condensational-heating-only sensitivity test. However, the derived thresholds, susceptibility fractions, and cloud-regime conclusions must be explicitly restricted to that hypothetical equal-LWC system and cannot be presented as quantitative constraints on aerosol convective invigoration in natural clouds.
Recommendations
½(w_polluted² − w_clean²),
alongside Δw/w. This would reveal whether the apparent preference for weak updrafts persists when the response is expressed energetically rather than as a fractional velocity change.