Comparing atmospheric boundary layer heights from vertical profiling scanning lidars to ERA5 and WRF

Mulet-Benzo, Cristina; Black, Andrew; Gottschall, Julia

doi:10.5194/wes-2025-155

Preprints

https://doi.org/10.5194/wes-2025-155

Preprints

20 Oct 2025

| 20 Oct 2025

Status: this preprint is currently under review for the journal WES.

Comparing atmospheric boundary layer heights from vertical profiling scanning lidars to ERA5 and WRF

Cristina Mulet-Benzo, Andrew Black, and Julia Gottschall

Abstract. The expansion and creation of new wind farms in recent years brings up challenges to manage both inter- and intra- wind farm wake effects. Wake blockage impact heavily relies on atmospheric conditions for determining how long and how intense the wake propagation is. One of these key atmospheric parameters is the height of the atmospheric boundary layer (ABL), which determines the height of the atmosphere most impacted by surface layer wind speed and temperature regimes. Generally lower ABL is united with more stable conditions, and thus greater the wake propagation. Offshore wind farms experience more frequent stable conditions compared to onshore farms, thus boundary layer conditions must be well parameterized to accurately model wake blockage effects. Scanning lidars present a viable solution for boundary layer height determination. In this study, their measurements are compared against ERA5 and WRF ABL model outputs. The lidar acts as a reference for boundary layer conditions, categorizing the ABL for both the mixing (convectively driven) and residual (stably driven) layer heights. Two campaigns, both using WindCube Scan devices, were assessed: one located completely offshore and another on a coastline. The results demonstrate an overestimation of the boundary layer height from both ERA5 and WRF from the offshore site of around 400 m. The coastal site yielded mixed results when comparing the ABL model height to the mixing and residual layer heights derived from the lidar, with a model overestimation compared to the mixing height of 300 m and for the residual height of 750 m. A sensitivity study demonstrates the bias of both models correlates to the ABL diurnal cycle and to temperature flux misrepresentation in the model.

Received: 15 Aug 2025 – Discussion started: 20 Oct 2025

Competing interests: At least one of the co-authors is a member of the editorial board of Wind Energy Science.

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.

Download & links

Cristina Mulet-Benzo, Andrew Black, and Julia Gottschall

Status: open (until 17 Nov 2025)

Post a comment Subscribe to comment alert

RC1:
'Comment on wes-2025-155', Anonymous Referee #1, 28 Oct 2025 reply
Review of “Comparing atmospheric boundary layer heights…” by Mulet-Benzo et al.
The manuscript describes comparisons of boundary layer heights calculated from lidar measurements with those from two models for two sites over water. The datasets should be quite useful, and a well-formed comparison could be informative for model development. Unfortunately the analysis suffers from poor understanding of the basic concepts of the boundary layer. The paper needs extensive revision to be publishable. Because of the conceptual difficulties, I cannot provide many detailed comments to this version. Another round of review is probably necessary.
General comments:
The authors seem to rely on a conceptual model involving a diurnal cycle (as in fig. 1 for example). This carries through to at least one of the algorithms used to process the lidar data (the ToD algorithm). The diurnal cycle generally does not exist over the water. If a diurnal cycle is present at all, it may be severely damped or even reversed, depending on the proximity to the upwind coast or to SST variations. The Introduction and text need to be rewritten to reflect a correct understanding of boundary layer behavior over water.

The most important issue in observing or comparing boundary layer heights is a clear understanding of what exactly is being observed or modeled. The paper reflects an understanding that the aerosol layer is not the layer being currently mixed by turbulence, an important point. The description of the lidar algorithms is good, although the residual layer and ToD algorithms are clearly not applicable to these cases. However, the description of the modeled heights is unclear. The ECMWF model assimilates radiosonde profiles, but the ABL height it outputs at any given grid cell is not directly derived from that data. Questions that need to be answered include:
Is the modeled height that produced by each of the models in its output file, or is it recalculated in a consistent way?

In light of the answer to a, what algorithm is used within or outside the model?

What version of the models was used? In the case of WRF, exactly what physics options and namelist variables were used?

A recent paper by Caicedo et al. provides detailed analyses of measurements, algorithms, and models of BL heights. Although the data are taken over land, much of the work is relevant to this paper. See Caicedo, V., Sedlar, J., Riihimaki, L. D., Angevine, W., Turner, D.D., Zhang, D., Lantz, K. Enhanced Boundary Layer Height Detection Using Ceilometer, Surface Meteorology, And Radiation Products With A Random Forest Ensemble Method. JGR Atmosphere (in press). If the preprint is not yet available online, it can be gotten from the first author, caicedo@colorado.edu.

Metrics: RMSE includes bias, so in cases with large biases, as here, the RMSE is confusing. We need to be able to distinguish between systematic error (bias) and random error. It’s much clearer to use simple standard deviation, also known as centered RMSE.

After a brief demonstration that the residual and ToD methods are not applicable, they should not be mentioned further. This will simplify the presentation.

Looking for an SST bias is important and useful. However, the reader should be cautioned that a point measurement of SST may not be precisely comparable to the grid cell average from a model. Furthermore, different SST measurement methods (particularly depth) may produce quite different temperatures, none of which is exactly that the model means by SST. SST is almost as conceptually complicated as BL height. So large biases may be important, but small ones are probably not.

Specific comments:
Line 175: Convective conditions may commonly exist during the night over water. This is an example of the lack of basic understanding mentioned in General Comment 1 above.

Line 193: Some literature has shown considerable problems with the Ri method, see for example LeMone et al. 2012 DOI: 10.1175/MWR-D-12-00106.1 That method can produce unreasonably high heights under near-neutral conditions, which are very common over water.

Table 5: Model grid spacing is not resolution. The horizontal and vertical grid spacings should be labeled as such. This also occurs in the text.

Line 373: I don’t understand this sentence. Both WRF and ECMWF are using (different) PBL schemes. See General Comment 2 above.

Reply
Citation: https://doi.org/10.5194/wes-2025-155-RC1
RC2:
'Comment on wes-2025-155', Anonymous Referee #2, 05 Nov 2025 reply
This manuscript presents a study of the atmospheric boundary layer height (ABLH) based on modelled (ERA5 and WRF) and observation (Doppler lidar) data. The evaluation of the estimation of the ABLH by global reanalysis and mesoscale models is an interesting scientific topic, and relevant to wind energy research, since as it is also mentioned in the manuscript the ABLH has an impact on the flow that interacts with wind turbines in wind farms. For the evaluation Doppler lidar data are used acquired at two different locations. Please find below my general, specific and minor comments regarding the manuscript.
General comments
The evaluation of the ABLH derived by ERA5 and WRF is performed based on Doppler lidar observations, which are used as a reference. However, the capability of the used Doppler lidars to identify the ABLH is missing from the manuscript. Therefore, it is difficult to assess whether the derived ABLH from the models is wrong or if the lidar-based estimations are erroneous. This is acknowledged by the authors in the Conclusions section, where they write: “the absence of an absolute reference for the ABLH, as well as multiple physical definitions on how to derive it, leads to the uncertainty investigated in this study”.

In the abstract it is stated that the two lidars were located one “completely offshore and another on a coastline”. However, in the rest of the manuscript it is stated that the ABLH estimations were performed over two offshore sites. I think that this point should be clarified in the manuscript to avoid confusion.

In the analysis the WRF and lidar datasets were averaged over 1-h periods for the comparison with ERA5 data. However, aren’t the ERA5 data describe an instantaneous wind field? Is it valid to compare a 1h mean lidar data with ERA5 data? Furthermore, what is the distance between the ERA5 and WRF grid and the location of the two Doppler lidars?

In the site of the FLOW campaign a ceilometer was installed next to the Doppler Lidar. Have the authors checked what is the comparison of the ABL height from these two remote sensing instruments?

Overall, I find the introduction a bit long and not coherent. I suggest editing the introduction and focussing on the background that is relevant to the study presented in this manuscript.

Specific comments
Many of my specific comments concern sentences of the manuscript that are not clear.
Line 24: It is written: ”This parameter is directly linked to atmospheric stability …” Which is the parameter that this sentence refers to?

Line 25: Please add a reference to support that “offshore conditions are mostly characterized by stably stratified conditions”.

Line 25: I think that the statement that the “stable conditions augment wake propagation” is slightly misleading. During stable atmospheric conditions the velocity deficit in the wake propagates over longer distances in comparison for example with unstable atmospheric conditions, but this don’t mean the wake propagations is augmented. I suggest reformulating this sentence.

Line 30: Here it is needed further information about what are the oscillations.

Lines 68 – 69: The strength of the return signal of a Doppler lidar can provide an insight of the aerosol concentration but cannot measure directly the aerosol concentration. This sentence must be clarified.

Lines 99 – 100: Which is “this difference”?

Line 116: What is meant with the “unique challenge for determination from measurements” and why is the reference of Puccioni et al. 2024 used here?

Lines 119 – 120: It is written that “… to obtain a return signal and perfectly follow the wind speed direction”. I guess it is meant here that the aerosol flow with the same speed and direction of the wind. I suggest clarifying this sentence and removing the word “perfectly” since it implies that Doppler lidars perform perfect measurements.

Line 135: When it is written “only the ABL scan was used for this analysis”, is it meant the 2-minute fixed vertical scan?

Line 139: The reference Patel et al. 2025 is incomplete.

Page 7: Figures 3 and 4. It is not clear what is the purpose of these two figures. They illustrate instruments (Fig3a and Fig.4b) and scanning patterns (Fig.4a). I suggest either discussing more about the figures in the manuscript or removing them.

Line 156: The radial wind speed corresponds to the projection of the wind vector on the line-of-sight of the emitted laser beam.

Line 160: What is meant when the authors write that “aerosols behave as passive tracers” in the residual layer? Furthermore, under certain weather conditions there could be transfer of aerosols in the free troposphere. What would be the impact of those events in the estimation of the ABLH?

Line 165: What is the definition of the “confidence index”?

Line 181: What is meant with the “Averaging both ABLH retrieval types”?

Page 10: Figure 6. Which was the sample dataset acquired and for which tool’s validation was it used? Also, please add units at both the x-axis and at the colour legend.

Lines 209 – 210. Which provides an indication that the stable conditions are not well simulated?

Line 259: The comparison of which of the two modelled ABLH shows an overestimation?

Page 16, Figure 9. The distribution of the ABLH at the GLOBE site is characterized by a dominate peak at 100 m. Could this finding be attributed to erroneous interpretation of the Doppler lidar data? Furthermore, which ABL values are used in the x-axis, based on the Doppler lidar or based on the models?

Lines 294 – 296: Why is the presence of wakes considered to be a problem here? Aren’t the observations used to estimate the ABL higher than the height of the wind turbines’ rotor?

Line 298. Which sensor and at which height is used to estimate the MOL and what is meant with the “MOL bias”? How is it explained the most common atmospheric stability stratification is unstable in the GLOBE site?

Line 299 – 300. Figure 13 shows a histogram of amount of data per month and per wind direction. How is the statement that “Figure 13 shows slightly more stable distributions” supported?

Lines 306 – 307: The ToD average in the FLOW site has a strong diurnal variation. However, this is not true for the GLOBE site, where the bias of the ToD average is visibly dependent on the hour of the day.

Lines 330 – 331: First, the increase of the ABLH reported by Abraham et al. 2024 was observed under stable atmospheric conditions. Most of the cases at the GLOBE site, based on the Fig. 12(c), correspond to unstable stratification. Second, hypothetically the presence of wakes would increase the ABLH, which means that the models should report lower ABLH values. However, the results from the GLOBE site show the opposite.

Lines 334 – 336. In connection to one of my general comments. What is the distance between the lidar and the grid of the models in the case of the FLOW site. Could the observed differences be attributed different locations?

Lines 440. In the conclusions it is stated that “this study could further improve the validity of lidar boundary layer height retrievals…” by comparison of different methods. Is it meant that the results of this study could be improved if different data processing methods were used?

Minor comments
I think that minor corrections are required in many parts of the manuscript, therefore a thorough review of the text in the manuscript is required. Please find very few suggestions for corrections below.
Line 6: I think that the word “associated” is more appropriate than the word “united”.

Line 11: I suggest deleting the word “completely” before “offshore”.

Lines 69 – 70: The word “scanning” is written three times in the same sentence.

Line 209: Please provide a description of the acronym MYNN.

Lines 316 – 317. Please correct the “…to a more accurately evaluate …”

Reply
Citation: https://doi.org/10.5194/wes-2025-155-RC2
CC1:
'Comment on wes-2025-155', Nicolai Gayle Nygaard, 07 Nov 2025 reply
Why is a distinct diurnal cycle expected offshore. The illustration in figure 1 is what is commonly seen in textbooks for the onshore boundary layer development through a diurnal cycle. Onshore the daytime convection is driven radiative heating near the surface. In the marine boundary layer, it was my impression that the clear diurnal cycle was absent because the SST changes on a much longer time scale due to the higher heat capacity of water. See also Figure 18.14 in Stull, Practical Meteorology where he illustrates a seasonal difference in the cycle. In the profiles I have seen from WRF there is a high occurrence of profiles without a capping inversion, where the boundary layer I stable all the way to the free atmosphere.

It would be nice if you added plots of the ABLH versus, hour, month, and stability to the paper. The plot of the sensitivity of the bias to these parameters is nice, but the first thing is to understand how the behaviour actually is from observations and models and to contrast that with what is expected.

Figure 1. Can the lidar determine the height of the capping inversion? I guess that you are arguing that this is the height of the top of the residual layer and that this is determined from the drop-off in CNR. The latter is related to the aerosol concentration. My interpretation is that these aerosols are remaining aloft in the residual layer after being lifted there by daytime convection. But is it possible that aerosols are later washed out of the residual layer by precipitation, and would the residual layer height than by underestimated by the lidar?

Since the lidar measures both the mixing layer height and the residual layer height, you should consider up front in the paper which of these (if any) the model estimates of the ABLH is meant to resemble. Clearly it can’t be both.

Please provide more details of the lidar ABLH algorithm. Is it the same algorithm used in Abraham et al (2024)? ABLH from model data is quite sensitive to the choices made in the algorithm. I suspect the same holds true for the lidar algorithm.

Has the lidar ABLH been validated against radiosonde measurements. I consider that the gold standard, since the soundings provide profiles of the potential temperature.

Is the lidar algorithm for the residual layer height accounting for the drop-off in CNR that would occur even if the aerosol concentration was constant with height?

The lidars only measure the ABL periodically, doing periodic VAD scans in between, while the model outputs are at hourly resolution. You could comment on the rate at which the ABLH is expected to change and relate that to the measurement and model sampling rate.

Figure 6. What is the unit on the x axis? It would be simple to indicate the hours if it meant to represent 24 hours

Figure 6. Is the sample dataset from an onshore or an offshore measurement campaign?

Figure 6. The data in this figure seems to have 3-4 periods with a residual layer (residual layer height >> mixing layer height). This is in contrast with Figure 1.

Page 9. The formula for the bulk Richardson number is what is often presented in the literature. However, the expression used in ERA5 for the BLH diagnostic is slightly different. See equation (3.92) in this documentation: https://www.ecmwf.int/sites/default/files/2023-06/Part-IV-Physical-Processes.pdf. Or at least it is more explicit about how the different terms are defined.

In section 2.2.3 you should specify how the ABLH is determined in WRF, or at least if it is a model output and not something that you estimate afterwards based on other model outputs. Some of this discussion is in the Discussion section, but it would be easier to follow the paper if it appeared in this section.

Table 5 states that the WRF simulation includes the Fitch WF parameterization. Do you know if this has any influence on the ABLH determined from the model? If not it would be interesting to compare with an otherwise identical simulation without the turbines.

You refer to the paper by Abraham that discusses how wind farms increase the boundary layer height locally. As I understand this process it is due to the flow deflection above the wind farm which is a consequence of the turbine drag forces. As the horizontal wind speed is decreased the vertical speed must increase due to the conservation of mass. In the coastal transition the wind speed also decreases as the flow transitions from lower to higher roughness. Do you also expect a vertical displacement of the top of the boundary layer in that case_

Table 7. Do you mean RMOL(= 1/L, the inverse Obukhov length) and not MOL?

Table 7. What are the limits of the different stability regimes based on? A reference would be useful here.

Figure 8. The mixing layer height seems to drop by more than 1000 m in 30 seconds (from 15:51:45 to 15:52:15). Is that realistic?

Figure 9. What is the shortest range of the lidar and how does that impact the lowest detectable ABLH?

Figure 9. Do you have a sense why the GloBE data have a sharp peak < 100 m while the FLOW dataset has a much more subtle peak at low ABLHs? Is there a correlation to the months included in the two datasets? They almost cover complementary seasons.

Figure 9. Is the peak in the FLOW data at 2000 m physical or an artefact?

Line 296 states that the models do not include wakes, but Table 5 specifies that the WRF simulation has the Fitch WFP

Figure 12. Contrast the observed stability distribution with the statement on line 25 that offshore is mostly characterised by stably stratified conditions. Did you mean stable stratification in the free atmosphere above the boundary layer?

Line 319. The diurnal cycle sensitivity “of the bias”

Line 358. Wind speed bins should be wind direction sectors?

Line 347. Positive SS should be positive SST bias?

Have you tried looking at the bias as a function of the temperature difference between air and the sea surface?

Discussion. I would move at least part of the PBL scheme and how this influence the ABLH from WRF to section 2.2.3

Understanding the sensitivity of the WRF ABLH to the choice of PBL scheme would be very nice

Line 409-412. The fact the lidar measurements of horizontal wind speed are trusted for wake measurements and WRA does not by itself imply that lidar ABLH estimates can be trusted. That can only be achieved by validation against other trusted sources of ABLH measurements like radiosondes. The two papers from the AWAKEN experiment you cite provide some evidence and I am sure there are others

You could also include a reference comparing ERA5 (or WRF) directly with ABLHs determined from radiosonde. Like Xi et al, https://doi.org/10.1029/2024JD040779

Reply
Disclaimer: this community comment is written by an individual and does not necessarily reflect the opinion of their employer.

Citation: https://doi.org/10.5194/wes-2025-155-CC1

Cristina Mulet-Benzo, Andrew Black, and Julia Gottschall

Viewed

Total article views: 224 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
181	26	17	224	13	9

HTML: 181
PDF: 26
XML: 17
Total: 224
BibTeX: 13
EndNote: 9

Views and downloads (calculated since 20 Oct 2025)

Month	HTML	PDF	XML	Total
Oct 2025	119	15	11	145
Nov 2025	62	11	6	79

Cumulative views and downloads (calculated since 20 Oct 2025)

Month	HTML	PDF	XML	Total
Oct 2025	119	15	11	145
Nov 2025	62	11	6	79

Viewed (geographical distribution)

Total article views: 223 (including HTML, PDF, and XML) Thereof 223 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 15 Nov 2025

Short summary

Wind turbine wake modeling requires accurate representation of atmospheric conditions, especially the atmospheric boundary layer height (ABLH). Scanning lidars provide a viable solution for ABLH determination. This study assesses ABLH data from coastal and offshore scanning lidars, comparing it to ABLH derived from models typically used for wake modelling. The results demonstrate a general overestimation of the models compared to the lidar, and a bias sensitivity to the diurnal cycle.


Total:	0
HTML:	0
PDF:	0
XML:	0