| 29 Jul 2026
Measurement Report: Vertical New Particle Formation Profile observed at 356 m Shenzhen Meteorological Tower
Abstract. Vertical measurements of New Particle Formation (NPF) are essential yet scarce for understanding the role of ultrafine particles in lower boundary layer atmospheric processes. This study presents comprehensive vertical observations conducted at the 356 m Shenzhen Meteorological Tower, utilizing mobility particle size spectrometers (SMPS and Nano-SMPS) at five heights (5–350 m) during two intensive observation periods (IOPs) in 2023. Results reveal distinct vertical dependencies in NPF characteristics. The NPF occurrence frequency was higher at ground level, while the particle growth rate (GR) was significantly enhanced aloft, increasing from 9±4 nm h−1 at the surface to 14±4 nm h−1 at 320 m. The vertical structure of NPF was classified into four types, which were closely related to the condensation sink (CS) higher at ground 0.04 s−1 than aloft 0.02 s−1. Turbulence promoted near- surface nucleation with peak growth rates up to ~13 nm h−1 (average 9±4 nm h−1 at 5 m) and enhanced growth in the middle layers with GR values of 7–8 nm h−1 at 100–200 m (e.g., 8 nm h−1 at 100 m and 200 m, and 5–8 nm h−1 at 150 m), while growth rate at 350 m was mostly within 4 nm h−1. Furthermore, the upper tower environment contributed more effectively to the conversion of new particles into cloud condensation nuclei, as supported by the significantly higher GR at 320 m (14±4 nm h−1) compared to the surface (9±4 nm h−1), indicating that particles aloft can grow into CCN active sizes more rapidly despite weaker turbulence at the highest levels. This work provides critical insights into aerosol vertical transport process and CCN production in the urban boundary layer.
Review of egusphere-2026-4039
Measurement Report: Vertical New Particle Formation Profile observed at 356 m Shenzhen Meteorological Tower
Summary of the manuscript
The paper reports vertically resolved measurements of new particle formation (NPF) at the 356 m Shenzhen Meteorological Tower (SZMGT), a coastal urban site in the Pearl River Delta. Two intensive observation periods (IOPs) in 2023 are used: IOP1 (stated as 1 March to 16 October) with mobility particle sizers at ground level and on the tower, together with Vocus PTR VOC measurements and trace gases; and IOP2 (14 to 26 October) with five sizing instruments at 5, 100, 150, 200 and 320 m, plus turbulence measurements from a sonic anemometer at 320 m and a Doppler wind lidar at the base of the tower.
The main reported results are: (i) NPF occurred more frequently at ground level (30 events, 26 %) than aloft (21 events, 18 %); (ii) events are classified into four vertical types (A: ground and tower simultaneously; B: ground only; C: tower only; D: decreasing mode diameter), which are linked to differences in the condensation sink; (iii) VOCs (isoprene, ethanol, acetone) and derived OOM production rates show pronounced vertical and diurnal structure, with maximum OOM production around midday in the 50 to 300 m layer; (iv) turbulence (TKE, dissipation rate) is linked to the particle growth rate, with near-surface accumulation and nucleation, enhanced growth at 100 to 200 m, and a weak response at the top of the tower; and (v) the authors conclude that the upper tower environment is more effective at converting new particles into CCN.
The dataset itself is valuable and rare. Very few sites worldwide can deliver simultaneous multi-height PNSD, VOC and turbulence measurements in a coastal megacity, and this is the reason the work deserves to be published in some form. The concerns below therefore address the analysis, the internal consistency and the presentation, not the value of the measurements.
Recommendation
Reject.
Overall, this article needs much further work, a better discussion of the facts and the evidence supporting them, and considerably more engagement with the existing literature. In its present form several of the central conclusions are contradicted by the authors' own figures, and a reader cannot verify any of the headline numbers. The condensation sink values quoted in the text differ from Figure 5 by a factor of three to four and reverse the direction of the key result; the growth rate given in the abstract and conclusion contradicts Section 4.1 and Figure 9; the turbulence quantities carry incorrect units; and the CCN conclusion rests on no CCN, hygroscopicity or survival-probability calculation at all. The manuscript also has no Discussion section, presents its Results almost entirely without references, and does not compare its findings with the vertical-NPF literature it cites in its own Introduction. Several analyses promised in the Introduction (event duration, formation rate, the sulfuric acid proxy) never appear.
These are not points that can be settled within a revision cycle: the numbers underlying the main conclusions have to be recomputed from the raw data, the cross-height comparison has to be redone over a common size range because the five levels used instruments with five different size ranges, and a Discussion section has to be written from scratch. I therefore recommend rejection. I would like to stress that this judgement concerns the analysis and the presentation, not the measurements. The dataset is rare and valuable, and I would be glad to see it resubmitted as a new manuscript once the analysis has been rebuilt on a sound footing. The comments below are written with that resubmission in mind rather than as a list of corrections to the present version.
General comments
There are many spacing errors throughout the manuscript, for example missing spaces after commas and full stops, and missing spaces between numbers and their units. I do not think it is useful to list the individual positions; the whole manuscript needs a careful pass.
In addition, please make unit notation consistent and SI-conform throughout (nm h-1 rather than nm/h, s-1, m2 s-2, m2 s-3, molec cm-3 s-1); at present these are mixed even within single sentences. Finally, the abbreviation of the site is used inconsistently as SZMGT, SZMG and SZMT; please pick one.
The manuscript mainly uses Chinese references, where in many places other references would fit better. For example, in lines 65 to 67, where tower studies in Beijing, the Amazon and the Southern Great Plains are introduced, several well-fitting ATTO papers are missing. The same applies to the European long-term vertical and mountain-site work, and to the Hyytiälä and Leipzig urban datasets. Broadening the reference base would also let the authors benchmark their own numbers, which is currently not done anywhere.
Generally, the results and discussion part needs references. Many described facts are stated without any reference. Related to this: the manuscript goes directly from the Results to the Conclusion with no Discussion section at all. Even for a Measurement Report, the results must be placed in context. Please add a Discussion comparing the occurrence frequencies and growth rates with other vertical NPF studies and with ground-based studies.
Several passages, in particular the paragraph in lines 358 to 374, contain a large number of assertions without proof. These need either to be demonstrated with the data and figures, or supported by references. The conclusion in particular builds very strong theories on quite little data, without those results having been established in the Results section. Concrete examples of claims that are asserted but never shown: the GR-CS correlation being "stronger aloft" (line 421), for which no correlation analysis appears anywhere in the manuscript; the attribution of Type A to turbulent vertical transport and Type D to subsidence (lines 285 to 290), for which no wind, mixing-layer-height or vertical-velocity data are shown; and the entire turbulence-NPF mechanism of Section 4.1, which is argued from Figure 8 even though Figure 8 contains no turbulence data at all.
The figure captions are generally very sparse. In my opinion a figure caption should be sufficient on its own to understand the figure. Please improve all captions accordingly. At minimum each caption should state the height(s) shown, the measurement period and the time convention (local time or UTC), the instrument the data come from, the quantity plotted with its units, and, for statistical plots, the number of events or days in each category.
Some of the graphics use a rainbow color map. It is well known and well established in science never to use rainbow colors, for a number of reasons. Please check all graphics for color-blind readiness. Please also regenerate every figure at higher resolution. In the current PDF, Figures 3, 4, 6, 7, 8 and 10 have axis labels, tick labels and color bars that cannot be read at normal magnification, which makes independent evaluation of the results impossible. Perceptually uniform maps such as viridis, cividis or the crameri suite would solve both problems at once.
The values stated in the text for the condensation sink do not match Figure 5 at all, and in one case the statement is the opposite of what the figure shows. This is very critical, because it contradicts several statements elsewhere in the manuscript. It is hard for me to believe how this happened.
Two further inconsistencies of the same kind need to be resolved:
The uppermost measurement level is given as 320 m in Section 2.2.2 and in the abstract, and as 350 m in Figure 9 and in Section 4.1 (lines 20, 26, 364, 372, 391, 420, 425). These appear to be used interchangeably. Please state unambiguously which platform was instrumented and use one height consistently, because the two are attached to contradictory GR values (see general comment 7).
Figure 8 shows that the five levels of IOP2 used instruments with five different size ranges: 10 to 700 nm (320 m), 7 to 65 nm (200 m), 13 to 500 nm (150 m), 13 to 700 nm (100 m) and 10 to 532 nm (5 m). This is a fundamental problem for every quantitative comparison in the paper. Levels with a 13 nm lower cut-off cannot fully detect the nucleation mode, so NPF event identification at 100 and 150 m is not comparable with the levels that reach 7 to 10 nm, and the finding that NPF occurrence is higher at the ground may be partly or wholly an instrumental artefact. The 200 m instrument stops at 65 nm and cannot track growth beyond that size, yet its GR is compared with levels reaching 700 nm. Since a lognormal-fit GR is strongly sensitive to the size range fitted, I recommend re-deriving all NPF statistics, CS and GR values over a common size range present at all levels (for example 13 to 65 nm), and reporting the original values only as a sensitivity test.
For a tower deployment with gradient sampling systems and high-altitude sampling ports (lines 124 to 126), diffusional losses in the inlet lines are a first-order effect below 20 nm and can easily reach 30 to 70 % at 10 nm depending on line length and flow. No penetration efficiency, line length, flow rate or residence time is reported anywhere. Because the core claims concern nucleation-mode particles at different heights, which presumably had different line lengths, this is essential information. Please report line lengths and flows for each level and either apply and document a diffusion loss correction or quantify the resulting uncertainty on the number concentrations and GR.
There is no CCN measurement, no hygroscopicity measurement used in the analysis, no growth-to-CCN-size calculation and no coagulation scavenging estimate. The claim that the upper tower is more effective at converting new particles into CCN (lines 25 to 27, 427 to 428) rests entirely on the disputed GR comparison. Either add further analysis of CCN or the statement should be removed from the abstract and conclusion.
The Introduction (lines 86 to 93) commits to quantifying occurrence frequency, duration, nucleation rate and growth rate. In the Results, the event duration is defined in detail in Section 2.3.5 but never reported for a single event or height. Either use these or remove them from the Methods and the Introduction.
Lines 309 to 313 state that low levels are influenced by ground emissions and higher levels by atmospheric transport. During daytime the ABL is well mixed anyway, with little variation across these heights. At night the layers can decouple, with complex effects. But what about chimneys? What about emissions from thermal industry, which create buoyancy? This definitely needs further analysis and discussion. The same applies to lines 392 to 395: the ABL normally reaches 1 to 2 km above ground, so the 356 m tower is well inside the ABL, and convection inside the ABL is generally quite homogeneous, so the air is well mixed during the day. This is the central weakness of the mechanistic argument and it should be addressed head-on rather than in passing. Concretely: no boundary layer height time series is presented anywhere, although "boundary layer development" is invoked as the explanation for GR at almost every height (lines 379 to 395), and the lidar should be able to provide the mixing layer height directly. Adding it would let the authors show explicitly when the tower is inside a well-mixed layer, when it straddles the nocturnal stable layer and the residual layer, and whether the observed vertical gradients survive that stratification. For a coastal megacity, the sea breeze circulation is also likely to be a first-order control on which air mass is sampled at 5 m versus 320 m, and a wind sector analysis would help. The highway 800 m away and the airport 10 km away are noted at lines 109 to 113 but never revisited.
Generally, please give the manufacturer of every instrument used. This is especially important in a Measurement Report. Maybe add a full instrument table listing, for each IOP and each height: platform height, instrument make and model, size range, time resolution, inlet length and flow, and data coverage in days. This single table would resolve a large part of the confusion in Section 2.2, and it is the kind of information a Measurement Report exists to provide. A second summary table giving, per height, the number of events, mean and median GR, CS, event duration and nucleation-mode number would similarly replace much of the anecdotal narrative in Section 4.1.
Specific comments
L16. "five heights (5-350 m)" applies to IOP2 only; please make the abstract describe each IOP accurately.
L19. "at the surface" is actually 5 m. Please state the measurement height rather than "surface". The same applies at lines 26 to 27 and 420, and the ground versus 5 m distinction should be made consistently throughout, since "ground" and "surface" are also used for the separate ground-based site in IOP1.
L19-20, L26-27. The value of 14 +/- 4 nm h-1 is unsupported and contradicts Section 4.1. See general comment 7.
L21. The CS values contradict Figure 5. See general comment 7.
L55. Carnerero et al. (2018) is cited but does not appear in the reference list.
L65-67. Many well-fitting ATTO papers are missing here. See general comment 2.
L74 and L100. The word "tower" is missing: "the tallest meteorological in Asia". Note also that line 74 says "among the tallest" while line 100 says "the tallest"; please use one claim consistently.
L113-114. There seems to be a leftover half-sentence from the writing phase. In this context it would also be useful to give the size of the urban area.
L117-118. This sentence fits better in the paragraph before.
L119. "...monitored continuously by a Meteorological by a Vaisala..." seems strange. As written, a wind sensor is said to measure the particle number size distribution. Please rewrite and state which instrument covered 1.8 to 500 nm and at which height. Note also that a 1.8 nm lower limit is not achievable with the stated model 3776 CPC, whose nominal d50 is about 2.5 nm; please clarify.
L120. BNU is not explained.
L124. Was the PSL calibration really done every 5 minutes? I assume the 5 minutes refers to the scan cycle; please rewrite. In addition, a sizing uncertainty of +/- 1 % is optimistic for a nano-DMA; please state how it was determined and give the sizing uncertainty at 10 nm specifically.
L126-127. For the Vocus PTR, please state at which heights it sampled, with what time resolution, how the gradient sampling was sequenced and what calibration was applied. None of this is given, yet Section 3.2 and the whole OOM calculation depend on these data.
L131. "serving as the reference instrument" - what serves as the reference instrument? And what does THU mean? The intercomparison as described is also far too brief to establish comparability between the levels.
L134. What is a Bi SMPS? please give manufacturer and model for every instrument, and replace descriptions such as "conventional particle size range" (100 m) with actual size ranges.
Figure 2. Why are the levels 200 m and 150 m named without instruments?
L152. PNSD is the particle NUMBER size distribution.
L190. "particle swarm" sounds strange in my opinion.
L195. How long are 12 data points? At 5 min resolution this is one hour, but it should be stated explicitly.
L204-205. For the wind lidar, please state the lowest usable range gate. Figure 10 shows data from 100 m, which is plausible, but the blind zone should be given.
L206-207. "...and data with a single wind speed profile missing rate exceeding 20 %..." - this sentence is not clear to me.
L210. "The above quality control process was repeated three times" - why three, and what changed between the iterations?
L250-253. "indicating that the ground atmosphere is more conducive to the occurrence of NPF" is a strong causal claim from a 26 % versus 18 % difference. please either test it properly or soften the wording.
L255. "It is not difficult to see that" and similar rhetorical constructions should be removed throughout.
L256-258. In my opinion this phrase does not make sense. It would benefit from a clearer formulation. What are "micro-particles", and why are they concentrated around the tower?
Figure 3. Please state the heights of the two panels. The description is very sparse, and it might help to add meteorological data as context.
L267. The first sentence needs rephrasing, as it can be misunderstood. The type definitions also need to be precise: what does "tower" mean here, at which height, and what happens on days when only some tower levels show NPF?
NPF types. How were the four types of NPF chosen? What were the thresholds, or was the classification made optically?
Figure 4. The axis labels are impossible to read, even in the PDF. Rainbow color is used again. The colors also appear burned out in parts, so the color scale needs recalculation. The description is sparse.
Figure 5. See general comment 7, the values in the text and the figure do not match, and in the case of Type N the direction is reversed. In addition, the caption says "mean CS level" but the plot shows medians and quartiles; the number of days per category should be given; and the claim of "significant differences" at line 272 requires an actual statistical test. It should also be stated whether Figure 5 shows ground CS, tower CS or a mixture, and over what time window, since the physically meaningful quantity is normally the pre-event morning CS.
Figure 6. The graphic seems to use some kind of interpolation between the levels; this has to be removed. Rainbow color again. The legends and labels are too small to read. It needs a better caption. The title also reads "ethvl alcohol"; it should be "ethanol". And since only a single day (7 April 2023) is shown, please state whether it is representative and add a campaign-averaged version.
Figure 7. The panel labels a to d are missing. Panel d (lower right) has two identical line types. The graphic is of very poor quality and looks AI-generated. I cannot read the labels. All four panels need readable axis labels with units, and the legend times (08:00 to 16:00) should state whether these are single-day or campaign-mean profiles.
L327-343. This paragraph partly does not represent the data shown in the graphics, and it discusses facts which have not been proven, especially lines 341 to 343. Specifically, the OOM generation rate at 16:00 was higher than stated (see the graphic), which contradicts lines 329 to 330
L358-374. This paragraph contains a lot of assertions without proof. It needs either proof with data and graphics, or references. The specific problem is that the mechanism is argued from Figure 8, which shows only PNSD and contains no turbulence data at all; the turbulence data appear only in Figure 10, for a different subset of days and only for 100 to 300 m.
L388-389. This sentence makes no sense to me.
L392-395. See general comment 13: the tower is well inside the ABL, which is generally well mixed during the day, so the proposed attenuation-with-height argument needs to be reconsidered.
Figure 8. It needs a better caption. The y-axis labels and titles are missing. Rainbow color again. From which time period are the data, and is the time axis UTC or local time?
Figure 9. The x axis is labelled "GR(m/h)"; it should be nm h-1. The y axis is a discontinuous linear height axis that makes the 350 m bars look detached from the rest. A horizontal bar chart is also not the natural representation of a vertical profile; consider a profile plot with error bars, or a boxplot per height.
L403. Figure 13 does not exist. Presumably Figure 10 is meant.
L416-428 (Conclusion). The conclusion builds very strong theories on quite little data, without those results having been proven in the Results section. It needs a full rewrite once the numerical inconsistencies are resolved, and it should state the limitations of the study and what the results imply for models, which the Introduction promises at lines 96 to 97 but the Conclusion never delivers.
Technical and editorial