General comments: The paper describes a meteorological and hydrological observational record of the Met Office semi-rural field site of Cardington between 2004 and 2024. The paper is well written and structured. I therefore recommend publication after minor modifications.

 

AC: We thank this reviewer for their quick response and efforts in their detailed comments which have led to an improvement in the manuscript.

 

My main concerns are: a) the introduction could be improved (see detailed comments below) and b) the advertisement for modellers to use the data could be improved. E.g. showing some nighttime data with fog – as fog forecast is still a big issue for modellers. This example could also include the CBL evolution after the fog night. c) I understand that the operators of the flux station don’t want to prescribe which corrections on sensible, latent and soil heat flux should be applied. But they are the most appropriate persons to suggest – at least - “accepted” algorithms.

 

AC: New analysis of fog data, and including an extra figure, feels like it would make a long paper even longer. It is also worth stating that fog droplet and aerosol microphysics, although they were intermittently observed at Cardington, were not included in the archived files described in the manuscript because of the difficulty in establishing good QC for this data to make it publishable. Therefore a low visibility/fog analysis from the available datasets must revolve around thermodynamic observations. As an alternative, we have added this paragraph to the introduction and explains reasons for the site in general and focussing on the issue of fog in particular. We hope this covers the "advertising" required to make the paper more attractive to readers.

 

"Cardington was established for both model evaluation (e.g. Price et al., 2018) and improvement of the parametrisations of processes in the Met Office LSM and regional NWP schemes (Boutle et al., 2014). Studies of the stable nighttime boundary layer (Horlacher et al., 2012), the morning and evening transitions (Angevine et al., 2020), and fog/visibility (Haywood et al., 2008) have all been carried out over the past 20 years. The perennial problem of fog forecasting was a recurring research theme at Cardington that has led to improvements in both LES and NWP modelling. Boutle et al. (2018) showed that attention needs to be paid to simulations of weak turbulence and low supersaturations that activate low concentrations of fog droplets. If this activation can be modelled using explicit aerosol data, then the evolution of optically thin, stable fog into thick, unstable fog can be better captured. Osborne & Weedon (2024) showed that correct modelling of the onset of radiation fog requires high-vertical resolution of the near-surface temperature profile, aerosol distribution and removal of fog droplets by canopy occult deposition. Consequently it has yet to be resolved the best way that corrections to sensible-, latent- and ground-heat flux should be modified to allow for fog"

 

The introduction now addresses the energy balance closure and a method to correct the sensible and latent heat fluxes using the Bowen ratio method of Twine et al (2000). This method has been demonstrated to great effect in relatively homogeneous terrain and is now recommended in our paper, in particular when validation of LSM model output is required. We have included a new figure demonstrating the exceptional energy balance closure based on yearly means for this dataset.

Section 3.2.1 has been amended to discuss the processing of the turbulent fluxes (and covariances generally). The raw 10 Hz has a certain level of QC applied, filtering for thresholds, removal of outliers, detrending, and the coordinates of the sonic components are rotated using three matrices, but anything more than this (Webb-Pearman-Leuning (WPL) corrections for the latent heat fluxes, and high-frequency loss corrections for all fluxes) are not applied-- although references are now recommended in the text to be considered by the data user. Because only processed 10-min and 30-min covariances are present in the CEDA archive, i.e. no raw data, then use of the TK3 software (as suggested by the referee) is unsuitable. Therefore simplified analytical approaches must be taken, as now suggested.

 

Detailed comments

1 Introduction

This is a very long paper and not all seems to be so important. It feels like there is a break in the flow (lines 54 – 69, history of the site). I would remove that part.

 

AC: We agree it is a long paper and could benefit from pruning. Lines 54-69 have been removed, which also means that several references have also been removed. The paragraph "The Cardington site was overhauled..." over lines 70-77 has also been removed because it referred to how the "modern" site was established in relation to the old history: it is also no longer required.

 

An introduction should “introduce” the intention/aim of the paper. Here, you mainly focus on the energy balance closure problem – the problem not best/possible solutions. Do you need this in the introduction or could you even move it to section 3? Instead, I missing some information regarding Section 5. I.e., what was the intention for using/installing the non-score instrumentation.

 

AC: The paragraph describing problems with energy budgets has been removed, it wasn't needed anywhere in the paper (but note the new energy budget figure which succinctly handles confidence in the site energy components)-- so that overall the Introduction is significantly shorter.

 

Obviously, there was an aim to install the station: the site was designed to carry out boundary layer, fog and air-surface exchange research to improve the representation of process-based physics within the Met Office Unified Model.

Finally, you should address the reasons why the data set was published and for which model comparisons the dataset might be most relevant. A little bit advertisement could be done.

 

AC: The reasons are now stated clearly in the revised introduction. We agree the introduction needed shortening and being more concise.

 

Table 1: table caption says period 2004–2024 but in the table you already start in 2003.

 

AC: The 2003 line in the table has been removed as it was outside the remit of the paper (Beare et al 2006 ref also removed).

 

RAF should be explained – as abbreviations in general.

 

AC: RAF (Royal Air Force) now not in the paper (paragraph removed)

 

Line 79 – 80: “….. partition the surface energy balance via the fluxes of heat, moisture and

momentum ……”. For me, momentum is not part of the surface energy balance as net radiation, heat, moisture and soil heat flux are.

 

AC: Agree, momentum flux is a controlling factor at the surface, not strictly a component of the energy budget. Removed.

 

2 Site description

Lines 124 – 125: You mention: “An investigation by Grant (1994) showed that the terrain surrounding Cardington can influence the wind field by channelling surface flow along the ridge in a south-west to north-east direction for a stably stratified boundary layer.”

 

The wind rose shows mainly south-west to north-east directions. So, does stably stratified conditions dominate the wind rose? Do you have separate wind roses for day and night or stable and unstable stratifications?

 

AC: No, stable conditions do not dominate the wind rose. Stable flow (i.e. mostly at night) amounts to 35% of the total available at 10m. The wind rose for Cardington resembles equivalent met data from near-by climate sites (https://www.metoffice.gov.uk/research/climate/maps-and-data/regional-climates/index), i.e. in particular the dominance of the wind sector around 240deg is the same. Therefore the effect of the local terrain, although it's measurable according to Grant (1994), does not change the climatological prevailing wind as viewed on a 16-point wind rose.

 

3 Site set-up and data logging

3.1 Site set-up and data logging

Section and Subsection should not have the same title.

 

AC: Agree, section titles now corrected.

 

Line 173: Was the raingauge really positioned at surface level which would be very unusual (splashing water).

 

AC: Text amended to "the rain-gauge was sited on the ground with the inlet rim at 0.45 m above the surrounding grass"

 

3.2 Core dataset instrumentation

Table 2: I would not use RH as in the column “Measurement” and “Variable” at the same time (see Rotronic sensor). Otherwise one column in the table could be removed.

Anyway, the title of the first column does not fit all the time. E.g. “Tri-axis sonic anemometer” is not a measurement but a measuring device instrument.

Also: what is the difference between specific humidity as variable and derived property.

Where do you give the units of the different variables (this concerns also Tables 3 and 4)?

 

AC: Tables 2, 3 and 4 were a bit of a mess. You are correct to make suggestions. "Measurement device" is now the left-hand column, i.e. the technology upon which the instrument depends. There were many corrections in this column across all three tables. "Variable" means the measurement(s) that is(are) directly output from the instrument. Importantly, these tables have been now been moved to the Supplementary material. There is a new Table 3 in the main text (as requested by Referee #2) that summarises the important variables and their uncertainties. We have decided not to add units to this table (or to the moved tables above): all units are given in the supplementary materials relating to the variables in the netCDFs.

 

3.2.1 Sonic anemometers

Could you say something about the ultrasonic measurements at 0.4 m. The height is close to the distance between the transmitter and receiver, i.e. a lot of the turbulence spectrum is probably not resolved.

 

AC: The 0.4 m sonic deployment was carried out to investigate potential collapse of turbulence close to the ground when this shallow layer is decoupled from aloft (at 2 m or 10 m, say). This is now explained in the text. We appreciate that turbulence at 0.4 m is not developed (and hence not resolved). But then, it isn't at 2 m in many conditions. Only at 10 m is turbulence fully resolved.

 

General comment: so, no corrections were applied as described by Mauder’s TK3 software?

 

AC: See response above about covariance corrections. No, corrections like the spectral loss energy using TK3 software are not applied. Some of the QC that is contained within KT3 is applied, however (see above). Simplified versions of spectral high-frequency loss corrections can still be applied by the user. Nevertheless, the overall energy balance is very good (on average 1-10 W m-2 compared to about 450 W m s-2 for net incoming and net outgoing energy.

 

Line 237: Abbreviation PRT should be explained. Is it pressure, relative humidity and temperature?

 

AC: platinum resistance thermometer, now in full where required in the text.

 

3.2.2 Relative humidity

Lines 297 – 299: Have you measured dew point temperature higher than the air temperature? Data flagged or removed? Or indicated as fog detection?

 

AC: data points where [dew point temp > air temp] are flagged as BAD in the archived files.

 

3.2.3 Licor high speed hygrometer

Perhaps I missed it, but did you mention where - in relation to the sonic anemometer (distance and direction) - the Licor is installed?

 

AC: The 10-m sonic anemometer and Licor were co-located on the same mast, offset laterally within 45 cm of each other, so same height, but on a separate 1-m boom sticking out from the 10-m tower. This is now stated in the text.

 

Line 311: Same question as above: how are the covariances calculated - just from the raw data without any corrections?

 

AC: QC is applied to the raw data as with the sensible heat fluxes (see above). But corrections like WPL are not applied. These are left to the discretion of the data user because they typically only amount to a few %. This is now stated more clearly in the text with a reference.

 

3.2.7 Radiation

Line 385: why not just W m-2 instead of (J s⁻¹ m⁻²)?

 

AC: Agree, but units are not needed here. Removed.

 

Line 393: could refer to the WMO report

 

AC: Agree, done.

 

3.2.8 Subsoil sensors

Lines 553 and 564 (and elsewhere): is there any difference between ground and soil heat flux. I would suggest to use the same wording throughout the text when the same variable is meant. Otherwise it could be confusing.

 

AC: Ground and soil heat flux mean the same thing. The paper now uses ground heat flux only throughout.

 

Table 6 seems misplaced in this section

 

AC: It does indeed. We have turned this into Table 2 (now at the start of Section 3) and renumbered the subsequent tables (note some tables now moved to the Supplementary material, but not new Table 2 on the flagging system).

 

4 Land surface model forcing data

A general comment on the flux data (sensible, latent and ground heat flux). I understand that you don’t want to prescribe which possible flux corrections. But for model comparison: those who generate the data are normally most familiar which algorithms should be applied to derive the best quality data (while modellers are not and would be happy if the data producer would do the job). So, why not provide raw and corrected values? Or at least mention a program which could be applied.

 

AC: The issue of correcting the turbulent fluxes has been addressed above i.e. using the Bowen ratio method contained within Twine et al (2000). This is what we recommend in order to be able to validate model data with confidence. This is a straightforward method. The Bowen ratio is assumed to be observed correctly, and so the sensible and latent heat fluxes are both adjusted (which often means increased) equally until an energy balance is achieved. This should be done for every 30-min time interval.

 

Lines 538 – 555: I understand that the observation – modelling comparison is just meant as an example. Nevertheless, it would be interesting to know, whether the fluxes shown in Figure 3 are just the covariances derived from the raw data or corrections have been applied? And if yes, then which?

 

AC: These fluxes are uncorrected. This is now clearly stated in the text. The brief analysis comparing the observations to the model is still valid.

 

The comparison of the sensible heat flux at 0.4 m would be of interest, too. Just for me to know as I mentioned the problem of using a sonic at that height. Or just explain why you didn’t include the data in the comparison.

 

AC: Including the observed 0.4 m heat flux would mean also including a JULES model output flux at 0.4 m, which is not something that can be achieved since the output level (height) is in effect the same as the driving (forcing) level, but such 0.4 m forcing data does not exist (i.e. temp, winds, humidity etc). The only forcing levels possible are those shown i.e. 2, 10, 25 and 50 m. Remember JULES is a land surface model, i.e. there is no atmosphere attached to it, so we cannot produce profiles for a range of heights.

 

I would prefer to see some vertical profiles of the sensible and latent flux (although only two levels are available), e.g. averaged over some hours (around midday). These profiles should also give a hint whether data are trustful or maybe outliers. And if the fluxes are not constant (by ± 10%) in the surface layer some explanation would be helpful.

 

AC: We understand the desire for this, but the purpose of Fig 3 was to illustrate the heights where the observation and JULES output is available. It isn't included to draw scientific conclusions from. Therefore we argue that we keep Fig 3 as it is, especially as the other referee has not requested any changes to this figure. The purpose of the figure is best shown as the time series as it already is rather than changing to mean profiles. As the observed fluxes are uncorrected, drawing scientific meanings from the figure is also a little risky.

 

5 Non-core remote sensing instruments

Table 8 includes several acronyms. I suggest to explain them also in the table caption because the table is already referenced in Section 5 but the acronyms are only explained in section 5.1 and 5.2. What exactly is radial turbulence. Could you be more precise.

 

AC: Abbreviations/acronyms now described in the caption. Radial turbulence is not quite correct, we have changed it to "radial velocity", which is what the Doppler shift provides. Of course, a measure of turbulence can indeed be derived over a sufficient period as explained in the text.

 

One general comment: the detailed descriptions of the remote sensing systems listed in Section 5 are quite different. E.g. in section 5.1 equations are given how the wind vectors (u,v) are calculated. The description of the ceilometer is more general. I would harmonise that. Perhaps a reference to books like Atmospheric Measurements (Foken Editor) could be useful where detailed measurement principles and algorithms of most of the remote sensing systems listed in Sections 5.1 to 5.4 are presented.

 

AC: The detail in section 5.1 for the Doppler lidars are necessary because of the wind and turbulence retrievals possible using the necessary trigonometry. Of course, ceilometers are also lidars but because they only measure backscatter they are much simpler devices and therefore the section on ceilometers is likewise "simple". The subsections on the microwave radiometers has been improved with better technical information on instrument frequencies and measurement uncertainties for brightness temperatures and water vapour and liquid profiles.

 

5.1 Halo Doppler lidars

Line 604: Add signal-to-noise ratio to SNR and replace signal-to-noise ratio by SNR in line 624

 

AC: As the abbreviation "SNR" was used only once in the paper (line 624), it was deemed unnecessary to define it. Therefore we have used signal-to-noise ratio in full twice in the paper.

 

Line 635: what is qe and qn

 

AC: These were meant to be the Greek letter theta, not q ! 

 

Line 649: here you use φ but in the equation in line 647 you use 𝜑. Anyway, I suggest to use either elevation angle 𝜑 or zenith angle α.

 

AC: The equations have been re-formulated for zenith angle (phi) throughout.

 

Line 647: if you number the equations, this should be (5)

AC: Yes, but this equation is actually the fundamental radial velocity equation used in Doppler lidar scans generally. We have moved it to an earlier position and is now Equation 1.

 

Line 650: if you introduce u and v, you should do it already ahead of eq. 3 and 4.

 

AC: Done, in relation to the new Equation 1.

 

Line 655: AT instead of just T.

 

AC: Done.

 

As mentioned above. Perhaps you just refer to the appropriate literature because section 5.1 is the only one which presents equations.

 

AC: We argue that the Doppler lidar scans and the derived wind profiles are important to the dataset. The different types of scan can be confusing. Equations are not deemed necessary for any other part of the archived dataset. Therefore we feel they should remain intact.

 

7 Example of turbulence data⎯ roughness length

The shown example on roughness length and it’s changes over time is interesting. However, I could imagine that some results from CBL and NBL observations in combination with energy balance data would encourage researchers to use the data for model comparisons.

 

AC: Referee #2 suggested an energy budget analysis and this, alongside your comment above, has resulted in Figs 3 & 4 on the seasonal roughness length and yearly averaged energy residuals over the 20-yr period, respectively. The correlation of the increase in both in the second half of the dataset suggests that the changing environment (woodland) for the prevailing fetch affected both the turbulence and hence the perceived energy balance (due to different footprints of the observations that constitute the energy components). 

8.1 File formatting

It’s unusual have only one subsection in a section. I.e., two separate section would be more appropriate.

 

AC: Good spot, but overall we feel no subsections are necessary-- there is only Section 8 now and it has been much reduced because the file information and DOI links are all contained within the landing page on the CEDA website; therefore only one link is required.

 

I had problems to access some sources:

Line 869: “Page not found” for http://www.nationalarchives.gov.uk/doc/open-government870

licence/version/3/I got an error message

AC: corrected to: https://www.nationalarchives.gov.uk/doc/open-government-licence/version/3/

 

Line 939 - 946: “DOI Not Found” for wind radar data https://dx.doi.org/10.5285/eb352545ce1b4476b2580a3e5885c00d/

         

AC: link now not needed as much of section 8 (now section 7) has been removed and replaced with the landing page DOI.

 

Line 878: I suggest to add the time zone – probably UTC.

 

AC: Done.

 

Line 885 and elsewhere: Does timestamps marks the beginning, center or end of the intervals?

 

AC: Timestamps mark the centre (or mid-point, as we call it) of each interval. Now stated twice in the text (when describing the core files and the JULES forcing files).

 

Miscellaneous

Line 33: If it is “well known” you could omit the sentence. Otherwise remove “well known”.

 

AC: Have removed "well known".

 

Line 35: On the other hand, …. . Shouldn’t you better say “additionally, …” because it isn’t in contradiction to something mentioned before.

AC: Agree, we now say "additionally".

 

Line 88: be careful with wording. Do you really mean energy budget or energy balance as discussed before (line 82)? A budget contains a storage term. Try to avoid mixing it up.

AC: Agree, have changed to "balance".

 

Table 1:

The way you list the entries in the table are somewhat inconsistent, e.g. Semicolon, comma, dot, Full stop at the end or no full stop:

In situ; turbulent fluxes

In situ, radiation.

In situ. Fluxes

 

AC: Table 1 now corrected and consistent.

 

Line 494: Normally, references should be listed in chronical order.

 

AC: Agree, now corrected.

Typos

Figure 1c: Wind rose: m s-1 correct to m s-1

Figure 2: the labelling (data completeness) is too small.

 

AC: Agree, but these titles on each of the panels were not really needed anyway; the table caption says what they are. Removed.

 

Line 372: µ instead of u in 0.85 µm.

AC:Done.

 

Line 374: It should be Koschmieder’s law instead of Koschmeider’s law.

AC Done.

 

Line 478: A dot is missing at the end of the sentence, i.e. after for NO2.

AC: Done.

 

Line 665: Depolarisation instead of Depolarization?

AC: Yes, changed.

 

Line 981: remove the second dot.

AC: Done.

 

Table 7: Soil hydraulic conductivity in kg instead of Kg

AC: Good spot ! Done.

 

References

I have not checked the references in detail but some inconsistencies are quite obvious.

“Quart J Royal Meteorol. Soc.” or “Quart. J. Royal Meteorol. Soc.,”.

Sometimes the author’s list stop after three, like Bosveld, F. C., Baas, P., Beljaars, A. C. M. et al. and sometimes all authors are listed.

DOI is missing several times although available, e.g. for El-Madany, T. S., Griessbaum, F., Fratini, G., Juang, J. Y., Chang, S. C., and Klemm, O.: Comparison of sonic anemometer performance under foggy conditions. Agr. For. Meteorol., 173, 63–73, 2013.

Sometimes pages numbers are given as “663–680” and normally “663-680”

Sometimes the words in the title are capitalized, and sometimes they are lowercase.

Sometimes “Boundary-Layer Meteorol” and sometimes “Bound.-Layer Meteor.”

Etc

 

AC: The whole reference list has been overhauled and checked. There were many errors and inconsistencies.

The new Fig 3 showing energy budget residuals.

AC: We thank the reviewer for the comments  on our Cardington data manuscript. There were some high-level suggestions here, which we address below. They have resulted in two new figures (one of the old figures being replaced), one new table, and four instrument tables moving to the Supplementary section. We have acted on most of these suggestions. Where we feel the comment was unfeasible or impractical to act upon, we have stated the reasons. Overall the paper is much improved after responding to your criticisms. The two new figures described below have already been uploaded as individual PDFs in response to Referee RC1. The new "uncertainty" table will be viewable as part of the revised MS when it is uploaded in due course.

*Major comments*

1.    Unclear added value and community relevance relative to existing Cardington data. Cardington has been used extensively in past boundary-layer and land-surface research, and some data streams are already archived and accessible via national data centres. The manuscript needs to much more clearly articulate. What is new in the present data product (variables, time coverage, harmonization, QC, uncertainty quantification, model-ready forcing, etc.) relative to existing archives

AC: We agree that the relevance of this dataset was unclear in the introduction, especially in relation to what had been available before. It is a valid point to raise. There was an older BADC Cardington dataset spanning 2006-2017 [new dataset 2004-2024] but was by no means inclusive in terms of variables and derived quantities. It does not contain any soil data, or turbulent fluxes, for example. QC was relatively poor with little or no metadata, and was in ascii. The size of the old core dataset was 4GB, compared to the new 9GB core dataset. This dataset has now been superseded by the new one described in this paper (CEDA has now removed the old dataset). The total size of the dataset, including the non-core radiosondes, microwave radiometers and lidars amounts to 2TB. None of these specialist instruments were included in the old dataset. The land surface model-ready forcing files are also entirely new and will be of key value for the rapid investigation by modellers. So in summary, the new CEDA archive is a huge improvement in what has gone before, in both quantity and quality, but also accessibility i.e. everything is now NetCDF with comprehensive metadata.

AC: The last paragraph in the Introduction has been expanded to describe how the new dataset is substantially different from the old one.

2.  For ESSD, a central requirement is a rigorous and transparent treatment of data quality, uncertainties, and temporal homogeneity. Instrument calibration procedures and long-term stability are described only qualitatively for many sensors, with limited quantitative uncertainty budgets. Changes in instrumentation, measurement heights, sampling strategies, and site surroundings (e.g. tree growth, urban expansion) are acknowledged but not systematically treated as inhomogeneities. It is unclear how these changes impact time series continuity and how users should handle them.

AC: This is an important point to raise and has made us realise that the turbulence data presented in Section 7 focussing on the roughness length could be reorganised into a time series and shown earlier in the paper, i.e. when addressing the core data in Section 3. The most important aspect of the dataset are the long-term turbulence observations at a range of measurement mast heights. In an effort to link data quality to the changing environment around Cardington (in particular the growing woodland), a new figure (Fig. 3) has been generated based around roughness length (retrieved in neutral conditions). This figure replaces the probability distribution functions in old Fig 6 i.e. the median roughness lengths (+/- 95% confidence intervals) as a function of height. The time series plots in the new figure now become not only an example use of the Cardington data but a quality check as well as showing the evolving conditions upwind of the site for prevailing conditions. A new paragraph has been added in Section 3. Therefore the previous Section 7 has now been deleted which shortens the paper overall. Uncertainties in general are treated in response [3] below.

AC: Related to the new Fig. 3 and the associated discussion is a new Supplementary Section (S5) on major changes to the mast equipment (at 2, 10, 25 m), e.g. deployments, swapping of instruments, changes in calibration. There is a lot of detail in here with many dates (and times of day) and even instrument serial numbers of instruments-- which in themselves do not mean much, but they emphasise changes in the field.AC: 3.  For some key variables (e.g. latent and sensible heat fluxes, soil moisture profiles, groundwater), quantitative uncertainties and bias estimates are only briefly mentioned or not given. Providing variable-by-variable uncertainty estimates, at least in the form of typical ranges or upper bounds, including contributions from calibration, representativeness, and processing assumptions.

AC: We agree that despite listing all the instruments and measurement types in the former Tables 3-6, the handling of errors and uncertainties was poor.

AC: A new table has been created that summarises the uncertainties associated with the principle measurements, i.e. state parameters, turbulence, radiation and soil are now succinctly shown with columns of stated uncertainty (manufacturer's) and real-word (practical) uncertainty. The latter came from laboratory calibrations, evaluation using field data, or using accepted error analyses from the literature. These types of uncertainties are labelled in the table. Because the old Tables 3-6 were essentially based around types of instruments and their manufacturer, these tables are best placed in the Supplementary section and referenced in the main text as appropriate. This has shortened the paper considerably. We believe the more compact summary in the new table that was suggested by the reviewer will make the archive more accessible.

4. Adding a dedicated section on time series homogeneity, listing all major changes (instruments, heights, surrounding land cover) in a table and describing their expected impact on the different variables. Where possible, including simple consistency checks (e.g. energy balance closure statistics, comparison of precipitation with nearby gauges, roughness length estimates vs. independent methods) to underpin the stated data quality.

AC: This comment overlaps with [2] above with regard to impact of land cover and also with [5] below with regard to the energy balance closure. We argue that a new section is not required because these aspects have already been covered by the other responses, i.e. the energy balance, roughness length are both now covered in detail with two new figures. Although comparison with other nearby observations or comparing z0,m to independent methods has been carried out, the two new figures (new Figs 3 and 4) are in effect statements of confidence in the data and summarise a huge quantity of data and address the issue of homogeneity. Associated descriptions naturally fall into different parts of the paper, so creating a new section seems unnecessary.

5.  The paper repeatedly highlights the “near-closure” of the surface energy balance as a key strength of the dataset. However, Energy balance closure is not quantified in a systematic way over the 20-year period (e.g. annual closure ratios, distributions of residuals). A dedicated analysis of energy balance closure statistics (e.g. Rn – G – S vs. H+LE), including seasonal dependence and possible trends. A discussion of how to interpret and possibly correct the fluxes if used for model evaluation, and a clear statement that the dataset provides raw fluxes without closure adjustment.

AC: A discussion on potential corrections to the turbulent fluxes has been addressed to Referee #1 who also raised it several times. Please see our response (i.e. responding to his suggestions to sections 3.2.1, 3.2.3 and 4).

AC: We appreciate the comment on energy closure and feel it's important to address this issue properly in the paper. An analysis of the energy balance in terms of the residual energy has been carried out across all years and the results are shown in a new table. This approach to an annual budget has been attempted previously using Cardington data (in Horlacher et al., 2012), but this was only for years 2006-2007. Therefore a similar analysis has been extended for the 2005-2024 period (whole years). Although ground heat flux can be estimated from the soil temperature profile throughout the period, and indeed was observed with the heat flux plates for about half of the time series, applying corrections and calculating errors in the ground heat flux are deemed to be a study in itself. Therefore like many analyses in the literature, the ground heat flux is assumed to be net zero over the course of a year. This assumption is justified for an unchanging site, i.e. stable environmental conditions such as soil and canopy. Note that although we did attempt to include the ground heat flux (not shown), and the resultant residual energy was a little more erratic with a large dynamic range (i.e. larger apparent error), the overall trend was similar to what we show in the new figure (new Fig 3-- meaning that the change is probably not caused by a long-term shift in soil storage.

AC: The drift in residual energy from -7 W m-2 deficit to +10 W m-2 surplus over the 20 years is very small (+/- 1.9%) compared to the net downwelling and net upwelling energy fluxes (of about 450 W m-2).AC: Breakdown of the components showed that is it explained by changing downward radiative fluxes in both the shortwave and longwave. A further analysis on why this is, e.g. instrumental drift or change in calibration, or a real-world change in cloud over in the last few years (note the largest energy surpluses were in 2022 and 2023) remains open.

AC: A new figure showing the energy residuals and net down/net up is now Fig 3. A new paragraph text explaining the above is now in place in Section 3.2.

6.  ESSD articles should allow users to understand and, in principle, reproduce the processing chain from raw measurements to published data. Add a flowchart or table summarizing the complete processing pipeline for each data class (radiation, turbulence, soil, meteorology), including all filters, thresholds, and corrections. Provide more detail on the turbulence processing settings, including references to codes or processing scripts if available. Clearly distinguish between what is in the “primary” data set (as measured/QC’d) and what is part of the derived forcing (gap-filled, aggregated, etc.), ideally with separate DOIs or at least very clearly separated files and documentation.

AC:  The most involving of all the QC and processing routines is that for the turbulence variables. A full breakdown of the processing that goes into the turbulence calculations (variances, covariances, heat fluxes) is now included in Section 3.2.1 on the sonic anemometers. This is in the form of a numbered 9-point list rather than a figure or flow chart. The processing steps for all other core data variables are, in effect, a simplified version of this as now explained.

7. Adding a compact overview table or “data descriptor” section listing, for each product: time resolution, variables, heights, time span, file naming conventions, QC flag scheme, and example file links.

AC:  We appreciate that adding another overview table might be beneficial to the paper, yet this is deemed unnecessary since the resolutions, heights, period of deployments, QC flagging etc are already covered in Table 2 (now moved earlier in the paper), plus the information in new Table 3, the modified four instrument Tables now moved to the Supplementary material, and to some extent Supplementary S5 which concentrates on sensor changes to the four instrumented masts. The referee is suggesting a compact table but in reality, such a table would be impractically large as there would entries for tens of measurement variables.

AC: We thank the referee for their comments-- in particular highlighting missing technical information and uncertainties for the microwave radiometers and the Halo Doppler lidars. The section on the specialist non-core instruments has therefore been significantly improved.

Major comments

 

1. Section 5 introduces a variety of remote sensing instruments. However, for some instruments, essential information is not sufficiently documented. For instance, it is unclear whether the Doppler lidar operates at 1565 nm or 1.55 μm, and the channel configurations of certain microwave radiometers are not fully provided. This lack of information may limit the usability of the dataset. With complete instrument specifications, users could perform synergistic retrievals of cloud properties or data assimilation using combined observations. Similar issues may arise for other instruments due to missing key information. I therefore recommend either providing the full set of instrument specifications most relevant to users or explicitly referencing detailed technical documentation where this information can be found.

 

AC: Corrected to 1.548 um wavelength throughout. We agree that details were lacking for the microwave radiometers.

Full frequency information is now stated for all three microwave radiometers, with uncertainties in brightness temperature, liquid water path (LWP) and integrated water vapour (IWV) in each case too. Extra references have also been added where appropriate, either from previous Cardington work or in general from the literature.

 

Further investigation into the radiometer details made us realise that the Supplementary table on the Humpro radiometer only showed the timeseries netCDF details, i.e. the profile netCDF variables were missing (although both types of files were mentioned in the Data Availability section of the main text). These have now been added, so that we now have two tables for the Humpro: Table S8.1 and S8.2.

 

2. Climate analysis and data assimilation both require stringent quality control procedures and measurement uncertainties, which are key concerns for users. I note that the core hydrometeorology instrumentation is well documented in terms of quality control. However, this information is missing or incomplete for many other instruments. For some non-core instruments, it may be sufficient to summarize the uncertainty information available in the instrument manuals or to provide appropriate references.

 

AC: This ties in with [1] above: there were missing errors/uncertainties for the microwave radiometers. These have now been stated in each appropriate paragraph in that section. In hindsight, these missing uncertainties (and the missing instrumental details mentioned above) were quite glaring, so thank you for the suggestion. Errors in water and liquid water paths were not simply a matter of quoting a manufacturer's number (as these are often not stated anyway). They in general come from field data using historic radiosonde profiles and machine-learning methods (neural network usually) to assess bias and uncertainty. Some of these radiosonde trials were carried out at Cardington in the early 2000s (such as Price, 2003). Although this radiosonde data is not included in the CEDA archives, the radiosondes used in the trials from the late 90s and early 2000s are included.

 

 

3. Some instruments measure similar variables over overlapping time periods, which provides an opportunity for consistency checks and possible inter-instrument adjustments. Examples include measurements from multiple Doppler radars, wind profilers, and radiosonde observations. While such analyses may be beyond the scope of the current manuscript, it would be helpful to explicitly inform users of these overlaps and to remind them to consider potential inconsistencies when combining data from different instruments.

 

AC: This is an excellent suggestion, and we agree that this constitutes another study entirely-- but we have added text (shown below) to the final paragraph of the Summary) to inform readers of data overlap (with instrument examples) and potential to analyse bias and potential corrections/adjustments.

 

"Another aspect not addressed directly was instruments that measured similar or related variables with overlapping time periods that could allow biases to be quantified and potential instrument corrections to be applied. Examples include (i) use of radiosonde boundary layer wind profiles to compare to the UHF wind profiler and Halo Doppler lidar; (ii) use of radiosonde profiles in further microwave radiometer humidity retrieval trials, (iii) in-situ humidity measurement from multiple sensors (humicaps, chilled-mirror and high-speed optical hygrometers), (iv) use of the two heat flux plates and soil temperature and moisture sensors to assess and correct the ground heat flux in a rigorous manner for its use in energy balance studies."

 

Minor Comments

 

Although the manuscript is generally well written, its overall length is relatively large. Some sections appear to have limited importance. For instance, it is unclear whether Section 8 needs to describe the file storage format in such detail. This information could instead be clearly documented on the data webpage.

 

AC: Yes, we agree. Everything was already in place on the CEDA landing web page, so it is intuitive to follow DOIs and other links from there. The bulk of this section has been removed, with just a small description of CEDA, CF conventions and of course the link to the main page. This has reduced the length of the text.

 

Given the richness of the dataset, providing a time series or catalogue of major weather events observed at the site, if possible, would be highly valuable. This would facilitate targeted analyses of specific weather phenomena and could also help users identify and exclude certain outliers in climate studies.

 

AC: This is a great idea, but one that would unfortunately take an inordinate about of time considering the Cardington group has been disbanded with scant resources now available.

 

Lines 345 and 592: please arrange the supplementary materials in the order in which they are cited in the manuscript.

 

AC: Done !

 

Section 5.1 devotes substantial space to describing the principles of three-dimensional wind retrieval using Doppler lidar. In my view, these principles are largely consistent with commonly used approaches and may not require such detailed explanation. Instead, I would be more interested in the uncertainties associated with the retrieved three-dimensional wind fields. Similarly, the impact of near-field effects on lidar backscatter measurements warrants further attention. It would be beneficial to inform users of the height range where data uncertainty is expected to be relatively high, or to provide the corresponding quality flags. This also applies to Section 5.3.

 

AC: We have now provided full uncertainty and precision information about the Halo Doppler lidar based on the SNR threshold used for the processing. Therefore we quote the vertical velocity and horizontal wind speed and direction errors. The first 3 gates of the Halo are effectively unusable due to beam crossover. This varies between 90 and 108m above the ground depending on the gate length setting. This is also stated in the text now. We can only really state the typical useable height that the Halo can reliably see, which is the boundary layer typically 1-3 km deep which contains sufficient aerosol for a backscatter signal. The SNR filter processing already applied in the data is all the QC that is required in practice.

 

At line 715, the phrase “frequencies between 22.24 and 31.4 GHz” should be made more specific by listing the individual channels and providing their corresponding NEΔT values. For the retrieved temperature and humidity profiles, I also recommend providing retrieval uncertainties profiles, or alternatively referring to relevant technical documentation.

 

AC: See response to major comment [1] above-- much more details on the microwave radiometers and uncertainties are now in the text.

 

Lines 539–540 emphasize that “The configuration of JULES here is the MetUM-JULES Regional Atmosphere and Land configuration as described in Bush et al. (2025).” Does Figure 3 use a different configuration from the other data? If the configuration is the same, this statement could be moved earlier to avoid potential confusion.

 

AC: Our description of driving (or forcing) JULES with the archived netCDF files is one thing, choosing the JULES science configuration is another. There are a huge number of settings and parameters in a configuration; a few examples would be i) depths and number of soil layers, ii) fundamental optical properties of soil and vegetation, iii) plant chemistry, iv) soil hydrology. The configuration of JULES can be chosen from either various past standalone versions, or from various past configurations as defined for coupled model (land + atmosphere) runs, whether that be for regional or global modelling.

That being said, it is straightforward to change certain parameters relating to site conditions. We have used the Bush et al (2025) configuration, albeit with the parameters used as defined in Table 4. This is now made clear in the text.