Authors’ response to the 1^st review

First of all, we would like to thank the Referees for their thorough and constructive comments and the positive reception of our manuscript. Their precious work contributes to refinement of the manuscript and hopefully, ultimately to the publication of the revised paper.

Quotation from the review:

“The attention to detail in the current study is impressive, with deep knowledge of the site evident. It is valuable as it includes a more accurate representation of the site, to account for the changes which have occurred there, and it should ultimately be published. A problem, however, with the current manuscript is that various uncertainties in the data are incompletely considered. This is important, as after all the corrections, users will want to know what level of accuracy can be assumed in the final corrected PG. And, following the points made in the conclusions about the need to understand a site fully, can the authors indicate how detailed a knowledge of a site is needed to obtain e.g. 10% or 1% accuracy? Another point is that the method of data selection is not to use the independent fair weather definition, but to select on the magnitude of the data following the long term conventions at the site. This may be a fundamental difficulty. Will the variations in this sampling between years also introduce further uncertainties?”

We understand that the major problems with the paper according to the first referee were the inappropriate handling of various uncertainties and the ineligible method of the fair weather data selection. Indeed, these are flaws in the article that should be revised.

Firstly, we pay more attention to the error propagation throughout the data processing. In the revised form of the paper, we will present in detail how various uncertainties affect our conclusions and how do they propagate through the different operations (e.g., the uncertainty of the measuring instruments, the error of the uni- and multivariate linear fits, etc.). However, we are not able to exactly quantify the total accuracy of the PG (atmospheric electric potential gradient) measurements at the measurement site. We lack the appropriate supplementary measurements (such as the meteorological and air conductivity measurements for instance) that is why we can not account for various other processes (e.g., local turbulent air motions, soil radioactivity, aerosol size distribution and concentration, ion mobility and conductivity, cloud heights, etc.) that could affect the uncertainty of the PG measurements. Please, mind that the principal aim of this paper is to correct for the shielding effect alone.

Secondly, we change the fair weather data selection method to a more appropriate one. Please note that we are confined to the usage of a fair weather data selection method based on the magnitude and statistical distribution of the data as we lack meteorological information that are commonly used to identify fair weather conditions for data selection (as for instance in Harrison&Nicoll, 2018). As there are long-term variation in the data, the application of the same fair weather boundary value to all the years can indeed introduce unwanted uncertanties. The reason to use this method was to preserve consistency with the Märcz&Harrison, 2003 paper. Nevertheless, we will adopt the data selection method as in Lucas et al., 2017 and use a varying upper fair weather boundary value (the lower one remains to be 0 V/m) that is calculated based on the statistical distribution of the data in each year to exclude unwanted year-to-year variations.

Major points

Major point 1:

Review:

“L195. Add the uncertainty to the experimental points. Derive the mean (red line) from the data by allowing for the uncertainty rather than just removing the data from 4th Aug. Through doing this more rigorously it may become clear that the experimental values are not inconsistent with the model. Merely excluding the inconvenient data is unsatisfactory.”

Authors’ response:

L195. Even with taking into consideration the uncertainties, we are afraid that some unwanted effects (e.g., inappropriate handling of the instruments, unconsidered temporary changes in the environmental conditions, etc.) influenced the measurements on 4^th August so that the profiles recorded on that day do not reflect the true shielding effect and thus they would bias the model. Please note that, compared to the measurements of the 3^rd August, the shielding profile should not have changed so much (the difference between the two measurements is only one day). The outlier along line VI. at 16 m from the fence on 4^th August (see Fig. 4/b in the manuscript) is there because the two instruments were put too close to each other and thus they altered the ambient electric field. That is an erroneous value that should not to be taken into consideration during the model setup. The other outlier at 22 m is again too high, we do not see such a sharp jump in the profiles measured on the other three days. Therefore, we are confident in that those values are specifically and incurably biased, unfortunately just in the vicinity of the critical locations so we choose to exclude these two, erroneous values from the investigation and to use the values from 4^th August without them. The reasoning of this decision will be clarified more in the revised manuscript.

Major point 2:

Review:

“L260 (and Table 3). Uncertainties on the annual percentage changes, propagated through, would be useful. Convention for p values is just to give an inequality and one significant figure (e.g. p<0.05, p<0.001).”

Authors’ response:

L260 (and Table 3). Thank you very much. We will implement the suggested changes.

Major point 3:

Review:

“L276 The PG and aerosol number concentration are generally found to be positively correlated. It may be that there is an aerosol size effect here and/or a number concentration effect. More explanation is needed.”

Authors’ response:

L276 Unfortunately, the aerosol particle size is not measured systematically in Swider therefore we can not account for possible variation in the aeorosol size. On long time scales, aerosol concentration and PG are mainly positively correlated at Swider but this relation is not unambiguous. As we do not have information about how the size of the aerosols changed during the years we can not assure that there is not a size effect at play there. We will explain this issue more carefully in the revised paper. Please note, that recently a paper about the long-term PG measurements at Swider has been published (Kubicki et al., 2021) with more extended time series in the aerosol concentration data as well. We will refer to these longer time series in our revised work.

Major point 4:

Review:

“L355. One important implication of this work is that local, site-specific effects may have a large influence on PG measurements and can entirely suppress global signals. How should this be considered in general? The need to see similar variations at displaced sites would seem important. On what timescales could this be expected?”

Authors’ response:

L355. This study focuses on the site-specific electrostatic shielding effect of local, conducting objects. However, there are other local processes that alter the PG. Considering the shielding effect generally, one should carefully evaluate each PG measurement site whether there are conducting objects near the PG instrument and perform similar parallel PG measurements to quantify the shielding effect. According to analytic calculations, thin conducting objects (such as a metallic pole or a fence) can distort the ambient atmospheic electric field up to 5% at adistance of 3 times their height and up to 1% at a distance of 10 times their height (Lees, 1915). If the obejct is more thick (like a forest or a bigger building) the distortion is around 5% at a distance of 5 times the height of the object and 1% at a distance of 33 times the height of the object (Benndorf, 1900, Lees, 1915). These values can help to locate PG instruments at any site. We will refer to them in the revised paper. Nonetheless, in our research, we were able to justify our results by finding similar changes in the PG on long time scales (ca. 5-10 years) recorded at a station (i.e., Swider) in the same region (at ca. 600 km). To find similar variations on shorter time scales and/or at more distant stations is out of the scope of this paper. We will note this more explicitly in the revised text.

Minor points

Minor point 1:

Review:

“L12. Start the sentence differently "In this work it is found…””

Authors’ response:

L12. Thank you, we will revise this part.

Minor point 2:

Review:

“L125. Fig2. The repeated points as stripes are confusing. Average them into a single value, with an uncertainty range. Also, the axes would be easier to read if the units were V/m.”

Authors’ response:

L125. Fig2. This figure shows the data points that were measured by the mobile and stationary field mills. The linear fit was done based on all the points, that is why we do not agree on averaging them. They can resemble stripes on the plot because the field mill can record PG values only in a precision of 10 V/m. On Fig. 2, PG values were averaged in every second (the original data sampling frequency was 2 Hz). Because of the averaging, PG values between every 10 V/m values can appear as well but they are grouped around certain values. Ultimately, we will change the unit to V/m, thank you.

Minor point 3:

Review:

“L143 arose --> arise”

Authors’ response:

L143 Thank you, the word will be corrected.

Minor point 4:

Review:

“L145 (and fig2). The calibrations on different days have different uncertainties. It is worth deriving them and including them on the plot, particular because of the apparently anomalous 4th Aug data in fig 1a.”

Authors’ response:

L145 (and fig2). We will include the uncertainty.

Minor point 5:

Review:

“L154 What is the basis for the 1m uncertainty in annual tree height?”

Authors’ response:

L154 To acquire more accurate tree height curves one should have a detailed knowledge about the soil, climate and ecological environment of the measurement site. As we lack these information, we are confined to determine the tree heights based on national averages. The uncertainty of 1 m is an expectable uncertainty of these national averages as the local conditions may differ from the average. This will be explained in the revised text.

Minor point 6:

Review:

“L235 Table 1. It is likely that the precision given is too great. It should be based on an assessment of the combined uncertainties in the calibration, and the validity of the model.”

Authors’ response:

L235 Table 1. Thank you, you are right. We will revise the precision of the values in the table.

Minor point 7:

Review:

“L285 Conditions local to the site are likely to be the cause. However, it does suggest some doubt about what the absolute value should be at either site.”

Authors’ response:

L285 The Geophysical Observatory of Swider is located near (ca. 15 km away) Warsaw, the capital of Poland and is surrounded by settlements. The anthropogenic pollution is likely to be higher at Swider than at Nagycenk as it is located in a Natural Park and only a smaller city is located near it. Higher pollution decreases the air conductivity thus resulting in higher PG values. Air conductivity at Swider after the perturbed period of atmospheric nuclear weapon tests is around 3-4*10^-15 S/m whereas the average fair weather air conductivity is greater by one order of magnitude (around 1.3*10^-14 S/m) according to Rycroft et al., 2000. Please note that PG measured at different sites can have highly different magnitude. For instance, in a paper where 17 PG stations were compared, the non-disturbed PG median of all the investigated data ranged from 21 V/m to 404 V/m (Nicoll et al., 2019). The high variability of PG at different sites, alongside with the different sensitivity of instruments at NCK and Swider, are likely to be the reason behind the different absolute PG values at the two sites.

Minor point 8:

Review:

“L355. This is too general a statement as written, as it does not consider the timescales that are relevant. What is probably meant here is on long timescales. But even so these are used for comparison with the KSC and Swider data.”

Authors’ response:

L355. Yes, this sentence is indeed too general. We will rewrite this paragraph to focus more on the electrostatic shielding effect and to point out the relevance of making comparisons between relatively close PG measuring sites, wherever this can be done.

Minor point 9:

Review:

“Fig 6. What is the aerosol size? The change in aerosol size with time (as well as number) will also affect the conductivity.”

Authors’ response:

Fig 6. Please, see our answer that was given to the third major point (L276).

Minor point 10:

Review:

“L308. Describe winter as "December, January, February"”

Authors’ response:

L308. Thank you, we will add the description.

Minor point 11:

Review:

“L309. Replace " This behavior is in agreement with the general theory of atmospheric electricity" with "This is frequently found at continental sites".”

Authors’ response:

L309. Thank you, we will rewrite it.

References

Benndorf, H.: Über die Störungen des normalen atmosphärischen Potentialgefälles durch Bodenerhebungen, S.B. Akad. Wiss., Wien, 109, 923–940, 1900.

Harrison, R. G and Nicoll, K. A.: Fair weather criteria for atmospheric electricity measurements, J. Atmos. Sol-Ter. Phy., 179:239-250, 2018.

Kubicki M., Mysłek-Laurikainen B., and Odzimek A.: Nature of Relationships Between Atmospheric Electricity Parameters at Ground Surface and Air Ionization on the Basis of Nuclear Accidents in Power Plants and Weapons Tests, Front. Earth Sci., 9:647913. doi: 10.3389/feart.2021.647913, 2021.

Lees, C. H.: On the Shapes of Equipotential Surfaces in the Air near Long Walls or Buildings and on their Effect on the Measurement of 395 Atmospheric Potential Gradients, Proceedings of the Royal Society A, 440–451, 1915.

Lucas, G. M., Thayer, J. P., and Deierling, W.: Statistical analysis of spatial and temporal variations in atmospheric electric fields from a regional array of field mills, J. Geophys. Res. Atmos., 122, 1158–1174, doi:10.1002/2016JD025944, 2017.

Märcz, F. and Harrison, R. G.: Long-term changes in atmospherical electrical parameters observed at Nagycenk (Hungary) and the UK observatories at Eskdalemuir and Kew, Annales Geophysicae, 21:2193-2200, 2003.

Nicoll, K. A., Harrison, R. G., Barta, V., Bor, J., Brugge, R., Chillingarian, A., Chum, J., Georgoulias, A. K., Guha, A., Kourtidis, K., Kubicki, M., Mareev, E., Matthews, J., Mkrtchyan, H., Odzimek, A., Raulin, J.-P., Robert, D., Silva, H. G., Tacza, J., and Yair, Y.: A global atmospheric electricity monitoring network for climate and geophysical research, J. Atmos. Sol-Terr. Phy., 184, 18–29, 2019.

Rycroft, M. J., Israelsson, S., and Price, C.: The global atmospheric electric circuit, solar activity and climate change, J. Atmos. Sol-Terr. Phy., 62, 1563–1576, 2000.

Authors’ response to the 2^nd review

First of all, we would like to thank the Referees for their thorough and constructive comments and the positive reception of our manuscript. Their precious work contributes to the refinement of the manuscript and hopefully, ultimately to the publication of the revised paper.

Major point 1:

Review:

“Conductivity is a key parameter in the physics of the Global Atmospheric Electric Circuit (GEC) which affects, dfor example, the atmospheric electric field magnitude. In the present study, the conductivity issue is not treated adequately; therefore, I ask the authors to be more specific in dealing with this parameter in their paper.

In line 101 it is stated that “all objects in the model were initially treated as perfect conductors”, which is certainly a gross simplification. First, what does it mean here “perfect conductor” when electrically the wood is a “perfect insulator”? They need to explain and discuss this idealized assumption and if should be applied here. Apparently, it is presumed that the conductivity of the objects (trees and building) is considered equal to that of the conducting ground. This is not well explained. Even the ground conductor concept needs to be discussed and clarified, since its conductivity varies and it may depend on season or is subject to a trend with years.

In validating their model, the authors have been obliged to deal with their assumption on conductivity, because the model led to systematic differences between the predicted PG values and the measurements (Figure 4a). Thus, in line 200 is stated that: “the model with perfectly conducting objects overestimates the shielding effect of the trees”. To account for this discrepancy, the authors adjusted the objects’ dielectric constants to values that fit better their data. However, they do not explain how the dielectric constant is used here to account for the effect of conductivity. Also, they need to explain why they adopt dielectric constants for the objects which are much larger than published values. For example, the authors use wood dielectric constant values of 120 to 130 which differ considerably from those of 25 to 85 reported in relevant publications. This is an issue that needs to be considered. To state it in other words: can the dielectric constant of the objects be used as a “free parameter” to fit the data in order to explain the observations? Are the large dielectric constants used in the present simulations realistic?

I recommend that the conductivity issue is first discussed in the “Model setup” section and explained as how it relates to the dielectric constant in the model. Then the initial calculations should be carried out by using published dielectric constant values, instead of considering the objects to be “perfect conductors”. Once this is done, new model calculations can be done by applying larger effective dielectric constants, which, however, need to be justified as being physically realistic. All this requires a major revision of the paper.”

Authors’ response:

The major problem with the paper according to the second review were the inappropriate handling of the conductivity parameter during the setup and validation of the model. Indeed, we agree with the referee on that treating trees as “perfect conductors” is an inappropriate way and we would like to thank them for drawing our attention to this problematic point. Therefore, we accept the proposed method and will firstly use dielecteric constant values from relevant publications, do the model calculations with them and omit the “perfect conductor” part. Furthermore, more attention will be payed for the conductivity parameter and any discrepancies will be discussed more thoroughly in the revised paper.

However, the model results still imply that trees at NCK have greater (120-130) dielectric constant values than those reported in relevant publications (25-85). Please note that the dielectric constant of living trees is highly dependent on the actual conditions of the trees, especially on their moisture content. As we do not have data about the wetness of the trees we are bound to fit the dielectric constant to the measured shielding profiles. Another reason for this that we can not model the exact, rather complex geometry of overlying branches and the foliage in our 2D model. This introduces an uncertainty in the dielectric constant values so we use an effective dielectric constant which is a bit higher than reported values. Again, this difference originates from the unkown moisture content and the complex geometry which can not be fully incorporated in the model. Realistic dielectric constants of the trees could be derived with this method if the wetness of the trees would be known and the 3D geometry of the tree and the foliage would have been modeled in more detail.

Major point 2:

Review:

“Finally, I wonder why the authors do not consider possible seasonal changes in the objects’ conductivity when discussing the seasonal (winter, summer, and spring) variations. Especially since ground moisture and various degrees of wood wetness, which vary with season, are expected to affect the conductivities, and therefore the model predictions.”

Authors’ response:

The principal aim of this study is to correct for the shielding effect in the long-term annual averages of PG at NCK. The wetness of the trees and ground moisture should indeed have an annual variation but we do not see such an effect in the shielding profiles measured in the winter and summer unambiguously (Fig. 4a-b in the manuscript). Furthermore, the seasonal variation in the PG is present in the uncorrected data as well so it is not introduced by the correction. This phenomenon is worth investigating, however it lies outside of the scope of this paper.

Minor point 1:

Review:

“Text in Page 6 and Figure 2: It is not clear why the mobile sensor fair-weather PG measurements range from 0.0 to 0.7 kV/m and those of the stationary sensor range between 0.0 to 0.275 kV/m.”

Authors’ response:

It is an unexpected behaviour indeed as the two instruments are of the same type (Boltek EFM100 field mill). The two field mills have different sensitivites. Moreover, during the field measurements the orientation of the head of the field mills differed from each other. In case of the stationary field mill, the head was oriented downwards whereas in case of the mobile one, the head was pointed upwards. The different orientation distorts the ambient atmospheric electric field in case of the two instruments differently. The field mill with the upward orientation (the mobile instrument) measures higher PG values as equipotential lines are somewhat denser at the top of the mounting pole. On the other hand, the downard-faced stationary field mill measures smaller PG values as its mounting pole shields the ambient electric field. Please note, however, that we are not interested in the absolute PG values in this study rather in the relative PG, the ratio of the PG measured by the two instruments after cross-calibrating the two field mills. We will describe this part more carefully in the revised paper.

Minor point 2:

Review:

“Is it justified to have 3 and 4 significant figure accuracy for the quantities shown in the various tables? How can you have such accuracy when you deal with measuring a quantity that is highly variable?”

Authors’ response:

Thank you very much for your pertinent remark. This level of accuracy is indeed too much for so highly variable parameters. As this problem was noted by the first Reviewer as well, we addressed it in our response to the first review. We will handle the accuracy problem more carefully and pay more attention to the error propagation and uncertainties throughout the revised paper. For a more detailed answer please see our response to the first review.

Minor point 3:

Review:

“In page 13 the authors rely on Figures 5 and 6 to conclude that the time series of the mean annual PG values at NSK are similar with those at Swider, Poland. However, I note that: (a) there is no Swider data plotted in Figure 5 (!), and (b) the upper panel in Figure 6 shows large differences in magnitude and variability between the electric field annual means at NSK and Swider. How can the authors claim that the two time series are well correlating? From what I see, there seem to be a problem with the fair weather field measurements done at Swider; this is also recognized by the authors but not explained (see lines 281 – 285). The Swider mean fair weather E fields are on the average too large, exceeding in most cases 200 V/m. I suggest the NSK-Swider comparisons to be omitted. ”

Authors’ response:

Figure 5 do not contain the Swider PG data, it is used to demonstrate the similarities between the NCK and Swider data which cannot be seen on Figure 6 because of the different value ranges these two datasets bear. To demonstrate the strong correlation (Pearson’s correlation coefficient is 0.8) between the long-term PG data at NCK and Swider we plotted the annual corrected PG means at NCK on Fig. 5/a and the annual PG means at Swider are presented on Fig. 6 upper panel. These two figures are to used for the comparison. All three long-term trends that were found in the corrected PG time series at NCK appear in the Swider data as well. The Swider data is only shown in Fig. 6 because this figure was adopted from another publication. We put the corrected NCK annual PG time series on Fig. 6 upper panel to demonstrate the difference in magnitude (but not in the variation) between the two datasets. PG values at Swider indeed exceed 200 V/m in most cases, however it does not mean necessarily that there is a problem with PG measurements at Swider. Please note that the Swider Geophysical Observatory is located ca. 15 km away from Warsaw, the capital of Poland and is surrounded by settlements. The anthropogenic pollution is likely to be higher at Swider than at NCK as NCK is located in a Natural Park and only a smaller city is located near it. Higher pollution decreases the air conductivity thus resulting in higher PG values. Air conductivity at Swider after the perturbed period of atmospheric nuclear weapon tests is around 3-4*10^-15 S/m whereas the average fair weather air conductivity is greater by one order of magnitude (around 1.3*10^-14 S/m) according to Rycroft et al., 2000. Please note that PG measured at different sites can have highly different magnitude. For instance, in a paper where 17 PG stations were compared, the non-disturbed PG median of all the investigated data ranged from 21 V/m to 404 V/m (Nicoll et al., 2019). The high variability of PG at different sites, alongside with the different sensitivity of instruments at NCK and Swider, are likely to be the reason behind the different absolute PG values at the two sites. Therefore, we do not agree on omitting the comparison with the Swider PG from the manuscript.

Minor point 4:

Review:

“Finally, omit "Central Europe" from the title, Hungary is enough.”

Authors’ response:

We would like to emphasize the region (Central Europe) in the title as PG data from Swider, Poland (the same region, Central Europe) were used to support that the corrected long term PG time series is correct. One of the conclusions of the submitted study is, that long-term fair weather PG time series measured at NCK are representative at least in a regional scale. That means PG recorded at NCK can mirror variation in the atmospheric electric field at least on a regional scale which do not mean necessarily that the absolute magnitude of the PG at different sites in the region are the same as the PG is highly dependent on local factors. Therefore, we suggest to retain “Central Europe” in the title.

References

Nicoll, K. A., Harrison, R. G., Barta, V., Bor, J., Brugge, R., Chillingarian, A., Chum, J., Georgoulias, A. K., Guha, A., Kourtidis, K., Kubicki, M., Mareev, E., Matthews, J., Mkrtchyan, H., Odzimek, A., Raulin, J.-P., Robert, D., Silva, H. G., Tacza, J., and Yair, Y.: A global atmospheric electricity monitoring network for climate and geophysical research, J. Atmos. Sol-Terr. Phy., 184, 18–29, 2019.

Rycroft, M. J., Israelsson, S., and Price, C.: The global atmospheric electric circuit, solar activity and climate change, J. Atmos. Sol-Terr. Phy., 62, 1563–1576, 2000.