Dear reviewer,

We would like to thank you for your time, effort and support! We have revised the manuscript according to all comments which has resulted, for example, in the removal of species richness analysis and the clarification of the plant functional type analysis. Please find below our point-by-point responses to your comments indicated in bold text. 

In addition to these changes, we revised the figures for blindness color issues and in the Materials and Methods, the reported seed densities and the soil sampling protocol as we noticed some shortcomings in these sections.

Best regards,

Justine Trémeau on behalf of all authors

General comments

This paper contributes some valuable case study data to an important area of research and management. The authors combination of GHG, drought and PFTs is interesting and fruitful approach. I particularly enjoyed the drought dataset and discussion, which I though was especially authoritative and novel. The figures look really good. The paper is well referenced, though add Marshall et al. 2023 Ecological Solutions and Evidence. The conclusions are clear and useful. In general the authors are careful not to overstate their findings and are clear about which results are robust and which are more circumstantial, except when it comes to plant biodiversity. Only non-grass species were identified in grassland systems, which is definitely a weakness. Thus I am uncomfortable with the species richness conclusions. I suggest to drop species richness entirely from the paper and refer only to plant functional types (which are well studied). Or, you could drop the Poaceae sp. taxon from the analyses and make the species richness models explicitly forb richness, rather than species richness. And in the title, perhaps ‘biodiversity’ would be more accurate as ‘plant functional types’. I am also missing a brief justification for why intensive study sites were combined with satellite sites. Why not study more sites intensively?

Thank you very much for the kind appreciation! We do understand your finding about the species richness, we discarded this analysis from the manuscript and adjusted the related section accordingly. The title was changed according to your comment.

We agree that it would be optimal to study a high number of sites in detail. However, observing CO2 exchange in a temporally representative manner to determine annual carbon balance is laborious and thus expensive. Therefore, we had two sets of sites: 1) intensive sites with good temporal coverage of CO2 exchange for the carbon balance and 2) satellite sites with better spatial coverage to answer the other research questions. However, as the methods partly overlap, we combined the datasets in the research question 4). We further clarified this at the end of the revised Introduction and in the Materials and Methods.

Specific comments and technical corrections

Summary

Line 20 – make clear that this is two sites not four

We clarified this in the revised manuscript.

Introduction

Line 47 – what is species cortege?

We changed the term to species community.

Line 58 – mention some typical species (in the UK often Lolium perenne for a hardwearing sports turf, or Agrostis stolonifera/Festuca rubra for a fine-leaved display lawn)

Around Helsinki Metropolitan Area, it seems to be combinations in different proportions of Poa pratensis and Lolium perenne, followed by Agrostis capillaris were the most common grass species in mesic sites. Dactylis glomerata, Elymus repens and Phleum pratense were also detectable onsite, although their presence was scarcer. In the most xeric sites, Poa species got gradually substituted in favour of Fescue spp, especially for Festuca rubra, which tend to grow in association with Lolium perenne. We clarified it in the revised manuscript.

Line 65 – Marshall et al 2023 also highly relevant: https://besjournals-onlinelibrary-wiley-com.ezp.lib.cam.ac.uk/doi/full/10.1002/2688-8319.12243

True. We added the reference in the revised manuscript.

Methods

line 90 – mention the criteria used to match lawns and meadows e.g. distance, …?

We indeed matched the plots with distance. This is now clarified this in the revised manuscript.

line 93 - in the UK the first two species are non-native invasive garden escapes. Is this the case in Finland? If so, mention that.

For sure, Lupinus polyphyllus is indeed a non-native invasive garden escapes; however, the others are neither alien nor invasive, they are archaeophytes (i.e. introduced in ancient times by humans). In the revised MS, we included a note after Lupinus that it is a non-native invasive species.

Line 96 – mowing robot! Cool

Yes!

Line 120 – mention species. Weak if the paper includes a plant biodiversity perspective.

There is no further details, and it is now in line with the revised manuscript, which does not include plant species biodiversity perspective anymore.

Methods – I am missing a (brief) comment on why intensive and extensive study approaches were chosen. why 2 x intensive and 8 x extensive, and how sites were allocated to the two study approaches?

Thank you for this comment which made us improve the description! As stated above, it is not feasibly to study labour-intensive aspects with high temporal and spatial coverage and therefore, we combined two approaches to have them both. As shortly explained above, originally, the intensive site study was designed for the one and the satellite site study designed for another project. The first one focuses on carbon sequestration, so the GPP was included in the study, and different types of grasslands, where one of the meadows is old, were studied. The second project focuses on conversion from a lawn to a meadow, so all meadows were transformed at the same time following the same procedure end of 2020 (so young meadows) and there, only TER (no GPP), in addition to CH4 and N2O fluxes were studied Moreover, for the case of the extensive sites, their locations were chosen from those made available by the housing associations and their number was larger to fulfill the public interest the transformation initiative produced. This is why we make the distinction between the different sites. The usage of these two sets of sites is now motivated and it improved the description throughout the M&M section.

Section 2.2 – this is not my area of expertise so I can’t comment.

Line 235 – why this difference in approach? A mistake/for convenience? Does it matter for the results?

The sampling actually took place during different years as KMP lawn was studied already during a previous project and the other intensive sites were included later. Unfortunately, the timing of the sampling was missing earlier but we included the sampling dates in the revised manuscript. Truly, spatial coverage in KMP lawn was wider due to a higher number of sampling points but the number is still sufficient and thus the difference should not hinder the comparison of the sites. We also want to stress that this data describes the sites as background and is not used to answer any research questions.

Line 263 – combining all grass species to one species is a weakness of the paper, especially as the study sites are grass-dominated grassland systems, and biodiversity is mentioned in the title. It contrasts with the thorough flux, soil and drought methods. Suggest to reframe the paper from ‘biodiversity’ (for which you don’t have data as grass species were not identified except in broad non-taxonomically defined groups) and towards ‘plant functional types’, for which you have good data.

As described earlier, the species richness perspective was removed from the reviewed manuscript, and the revised manuscript now focuses only on the plant functional types as suggested by the reviewer. We also added a new table to show the cover estimates of different functional types and a short comparison of plant functional type’s distributions between lawns and meadows.

Line 331- corrections for multiple tests?

We have now added to the analysis a Kenward-Roger correction (Kenward et al., 1997), because of the small size of the dataset.

Results

Figure 2 looks good

Thank you!

It would be easier to follow the results if you used the type of lawn/meadow (following fig 2) instead of the abbreviated site names. i.e. line 373, KMP meadow -> mesic meadow, VKI meadow -> dry meadow.

We changed the names in the results section as suggested.

Discussion

Line 480 – or even net positive

We added the aspect of net positive here.

Line 482 – “a very fertile and a very poor meadow” . What is meant here? Poor in what? Earlier you call the sites ‘mesic meadow’ i.e. middling, neither especially fertile nor nutrient poor, and ‘dry meadow’ i.e. lacking moisture . nutrient availability is related to soil moisture, but they are distinct concepts

One is a mesic, mesotrophic meadow and the other one is a dry sandy (->xeric) nutrient-poor meadow. This is now clarified in the manuscript.

Line 500 good point

Thank you!

Line 511 – yes, I think time since establishment would be important.

Yes, we do agree. It has been clarified that in next studies, it would be valuable to compare meadows with varying fertility but comparable age.

Line 515 – more quadrats would have helped to represent better the variation present at the site?

Yes, probably, but again we cannot say anything about the inhomogeneity, which would probably be the same with more quadrats.

Line 576 and line 585 – no link between plant diversity and C and N cyles in your study could be because you didn’t identify grass species?

It is possible, but the section about “plant richness” is now removed from the revised manuscript and the discussion has been updated accordingly.

Conclusions

Line 614 – Meadows do increase biodiversity -> meadows have higher species richness and PFTs

It is changed as suggested in the revised manuscript.

Line 618 – new sentence .Nonetheless…

It is changed as suggested.

Line 620 – I am uncomfortable with this conclusion given your approach to plant identification. You could drop species richness entirely from the paper and use only PFTs. Or, you could drop the Poaceae sp. Taxon from the analyses and make the species richness models explicitly forb richness, rather than species richness. And then line 620 should read “Regarding the plant diversity, the FORB richness does not seem to affect significantly 620 the C and N cycles. Some PFTs are correlated to specific C and N status”.

We understand your point of view and we chose to discard this richness section and focus on the plant functional types proportion as mentioned above. It is corrected the discussions on that accordingly.

Dear reviewer,

We would like to thank you for your time, effort and support! We have revised the manuscript according to all comments which has resulted, for example, in unified terminology, in the removal of species richness analysis, as well as in the clarification and further discussion about the JSBACH modelling. Please find below our point-by-point responses to your comments indicated in bold text. 

In addition to these changes, we revised the figures for blindness color issues and in the Materials and Methods, the reported seed densities and the soil sampling protocol as we noticed some shortcomings in these sections.

Best regards,

Justine Trémeau on behalf of all authors

Trémeau et al. provide an interesting study about urban green places, their biodiversity, and their ecosystem functions. They have a novel study setting and new results, and the language in the manuscript is mostly clear. I personally would have split this manuscript in two (e.g., one focusing on flux magnitudes and modeling, and another one on the key ecological and soil drivers) to be able to analyze all the core topics in enough detail. For example, JSBACH seems to perform quite badly (especially for the non-irrigated lawn and dry meadow) and it would have been useful to explore this in detail. However, as this manuscript is now close to its final stage, I do not think such a split makes sense, and have provided my major (or moderate) and minor feedback considering the current structure below.

Major issues:

1. Biodiversity data

The terminology related to the biodiversity has to be edited. I think you study species richness and the cover of certain plant functional types, and would use these terms throughout the manuscript. It would also be good to report the number of species that you observed and discuss those in relation to earlier studies.

According to the comments from both reviewers, the species richness analysis is now removed from the manuscript, and where it is expected, the term “diversity” has been changed to “plant functional types”. Therefore, we did not include the number of species either. Instead, we added a new table to show the cover estimates of different functional types and a short comparison of plant functional type’s distributions between lawns and meadows.

Line 18: Here and throughout – I don’t think you are studying plant functional diversity. I generally consider plant functional diversity as diversity metrics derived from trait measurements (see e.g.  https://www.frontiersin.org/articles/10.3389/fpls.2022.993051/full). You are considering species richness metrics here. Be careful with the wording throughout the manuscript.

As mentioned above, we removed the species richness analysis and focused on plant functional types. We have carefully gone through the manuscript to be consistent with the terminology.

1. Cumulative fluxes are highly uncertain due to the poor performance of JSBACH

I am not sure if the utilization of JSBACH here makes a lot of sense after seeing the very poor performance of the model. Understanding why the models perform so badly would be important. Also, why was JSBACH chosen as a first step? Is it the best model to predict fluxes in northern grasslands?

JSBACH is the land model of the earth system model ICON and is thus widely used especially in Europe. There is not a dedicated model for boreal, urban grasslands as those differ greatly from agricultural ecosystems, especially by management and cultivated species and therefore, JSBACH can be assumed to describe these studied habitats as well as any other process-based ecosystem model.

In some cases, the observed emissions (TER) were high compared to the input of carbon from the standing vegetation. It seems evident that the studied soil was not stabilized during the measurements. Such mismatch is a result of urban construction where growing media with unknown properties was brought to the site, and it is not a result of long-term carbon input by the standing vegetation. This is most evident in the young sites (less than 15 years old) i.e., the non-irrigated lawn and the dry meadow, highlighting the variable and unpredictable characteristics of the urban soils that are commonly brought on the site from somewhere else or the land use has changed and therefore, the quality and quantity of organic matter is not connected to the history of the site.

Here, the simulations included a long spin-up period, i.e. in the simulations the soil carbon pools are a result of the carbon input from the vegetation over a long time. The simulations therefore represent carbon fluxes typical for these habitats over longer time periods. We decided to choose this approach as we wanted to be more general and show the longer-term effects of these vegetation types than just two cases where the ecosystems have not stabilized yet. Thus, instead of just fitting the model to the observations and giving the carbon balance of these four sites with unconnected soil underneath, we wanted to give a more general overview of the different habitats in the long term. Therefore, the model was originally initialized to reach a steady state with each vegetation type and the simulated soil carbon pool represents the result of long-term carbon input of the habitat itself. We realized now that we had not sufficiently motivated the use of the model and the approach in the earlier version of the manuscript. In the revised manuscript, we have now clarified why we have used the approach.

It is also possible to better fit the model fluxes to the observations by adjusting the soil carbon pools, as shown in Figure 1 (below). This, however, does not necessarily result in fluxes that would represent the average fluxes of these habitats over extended periods. Some outliers can still be seen in Figure 1, mostly data points with large ranges. Furthermore, the soil respiration in the model likely reacts slowly to short hot periods, as the flux is driven by a two-week running mean of the temperature. In the revised manuscript, we added annual balances also by this approach i.e., forcing the model simulation to the observations. In addition, the new figure (Fig. 1, attached pdf document) was added in the supplementary material and discussion about the effect of this decision in the Discussion section.

Do you come to the same conclusions about the general sink strength when you compare the pure measurements?

The measurements were only conducted every two weeks and there are uncertainties according to the day and the time of the day. Thus, it is not possible to give annual fluxes like with JSBACH. However, it seems evident from the data that the annual balance would have been higher based on the TER observations that were more often higher than the model simulation. In the revised manuscript, as mentioned earlier, we now also provide annual estimates that are estimated with a model version that is forced closer to the observations as already described earlier.

420: Perhaps add some kind of a model performance metric (R2, RMSE) here to demonstrate that dry meadow estimates, for example, are highly uncertain.

We added some metrics (R2) for discussing the model performance regarding the annual cycle and the changes in the carbon fluxes during a dry period.

503: This comparison seems quite uncertain here after seeing the major issues with the model. I would put more emphasis on comparing the measured growing season fluxes (and SOC, SON, species richness, and plant communities in general) with earlier studies.

As described earlier, we modified the discussion at this part to include the purpose of the modelling exercise and the possible uncertainty considering the annual balances of especially TER and NEE of these sites.

1. Methods section needs to be clearer

I suggest the authors carefully revise the methods section and make sure that it is chronological. Right now the sections are sometimes a bit hard to follow because they can start with a description of sample collection, then go to laboratory analysis, then talk about field description and study design again. I would always start with study design, then move on to field sampling, and then laboratory analysis.

Thank you for the feedback! We have carefully reviewed and edited the manuscript and re-organized the material and method section in a better way. A “study design” section was added after the site description, in which we describe the design (the transects and four quadrats along the transects), the flux measurements and soil sampling sections were also slightly reordered.

135: Somewhere here, specify how many collars you had and how many times they were visited.

The number of collars has been clarified in the new “study design” section and a sentence was added to specify how many times each site, at least one collar at a site, was visited (including also the number of visits, when we were able to measure the light response).

133: Maybe it is mentioned here somewhere, but if you were taking dark measurements at a collar, how many replicates did you take?

At each site, we measured four different plots/collars only once. This was clarified in the methods.

138: How many measurements did you take to derive the light response? Did you take measurements at several light intensities? (I now see that this is mentioned on line 155 – perhaps make sure the text is in a logical location always; on line 155 also describe the PAR values for the three different light levels)

By default, five measurements were collected at each collar at different light intensities from full ambient sun light (transparent chamber) to no light at all (opaque cover), with three intermediate measurements done at 50%, 25% and 12.5% of full ambient sun light. If some of these needed to be discarded due to poor quality and possible technical errors, at least three measurements were needed to derive the light response curve. A sentence was added to clarify the minimum of measurements needed for the LR curve, and the different light intensities are now explicitly written.

162: I am a bit confused now, earlier it was said that it was with a dark chamber. I also find it a bit challenging to follow which gases were measured and why there were differences. It might help to show these various measurement details with a timeline or a table format.

We use two different terminologies to describe the same chamber. We chose “opaque” as a consensus. We appreciate the table idea; we added one to the supplementary material.

167: Should the number and distribution of plots be presented already earlier with the study designs? Perhaps with a description of how many times they were visited in total?

We clarified the description by adding a section right after the site description section, in which we described the number and distribution of the plots at each site. We also added two sentences to describe how many times the sites were visited in total (TER), and the number of times when we recorded LR.

205: Earlier it was said that ca. 10 % of PAR variability was accepted?

The 10% of PAR variability was used on the field, and then the 150 µmol m−2 s−1 threshold was used by the Python script. Thus, there were two control steps at two different phases. However, after double checking, it appears that the first filtering is more restrictive than the second and no measurements were deleted from the second filtering. We chose to remove the mention of this second phase of filtering.  

230: Did I understand correctly that no soil temperature or moisture data were collected during the measurements? This seems problematic to me since these are two key drivers of GHG fluxes. After seeing Fig. 2 I think you measured these but I am not sure

As stated in the material and method section, soil temperature and soil moisture were all measured at the same time as the fluxes were recorded. To make it clearer, the description has been moved from the flux measurement protocol subsection to soil measurements section.

248: I would state the number of samples, study designs, and collection time early in the paragraph instead of late in the paragraph.

We added a section “study design” much earlier in the material and method, and added a sentence at the beginning of the paragraph to clarify that it is the start of the satellite site paragraph (after the paragraph describing the method for the intensive site).

273: Why not the SMEAR station that was used earlier?

The data from the Finnish Meteorological Institute (FMI) is produced for the operational weather forecast and likely has a more rigorous quality control. The stations are situated on the same campus area and show very similar data. The FMI observations were gapfilled using the SMEARII data.

276: How were the LAI data used in the model? Did it have to be gapfilled?

The LAI was not gapfilled for the simulations. The model does not use the observed LAI, except for the maximum value of LAI. The JSBACH has its own phenology model that calculates the growth and shedding rates based on environmental conditions (mainly temperature and soil moisture).

283: Do I read this correctly – the main property that differentiates the lawns and meadows in the model is the satellite-derived LAI (because the meteorological data were the same) and the flux measurements?

The model doesn’t use the satellite derived LAI as input. Only the maximum LAI, used by the phenology model, was estimated from the observations. This is now clarified in the revised Materials and Methods section. The model calculates the growth and shedding rates from the environmental conditions. The model also calculates the soil moisture, which depends on the precipitation, temperature, radiation, vegetation and soil properties. The fluxes are not used by the model either, all fluxes are calculated by the model. It is true that the LAI has a big impact on the fluxes etc., however, as the model derives the LAI several factors will affect the simulated fluxes.

295: Can you explain how the model was using the GHG flux measurements? Were the parameters somehow altered based on that information?

Only the GPP was used to adjust the model parameters. The soil carbon pools, that the model uses for calculating the heterotrophic soil respiration, were accumulated by running a long spin-up period (thousands of years). The simulated total ecosystem respiration will therefore not always reproduce observed fluxes. For example in cases where the vegetation has been changed recently, which may often be the case in urban areas. In the revised manuscript, we clarify how the simulations were performed.

295: Did you simulate for the peak winters too? Fig. 3 only shows visualizations that lack winter months. How did the model perform during the winter?

The model was run for full years, also including the winter months. We don’t have measurements for the winter months and therefore no direct comparison was made. The only flux that is active during the winter is the heterotrophic soil respiration, which decreases slowly when the soil temperature decreases, and reaches a minimum in late winter. We added a note in the caption of Figure 3 stating "The model simulates the whole year but for clarity, Jan-March and Dec are not visible as there were no measurements during those months".

301: reference to FAO

It was added to the reviewed manuscript.

310: components - resistance and recovery - we s…

It is corrected in the reviewed manuscript.

329: SOC and SON not defined yet. Also, how do you know that you do not have inorganic C? Did you test for carbonates?

Thank you, we defined SOC and SON in the revised manuscript. We assumed that the share of inorganic C and N is very low and added a sentence about that in the revised manuscript: “Total soil C and N contents were assumed to be only organic components without carbonates i.e., soil organic carbon (SOC) and soil organic nitrogen (SON).”

335: What is plant functional category? Not defined yet. This should be described in detail in the methods section.

Now, it is described in the plant inventory section.

1. Some potential issues with the statistics need to be addressed

333: What is location? The coordinates? The name of the study design (intensive, spatial)?

Thank you for noting this! “Location” factor describes the site of the lawn/meadow pairs. We changed the formulation to make it clearer in the revised manuscript.

335: Did you check the correlation of the predictor variables? It is generally not recommended to put variables that correlate >0.7 with each other to a multivariate linear model. Predictor collinearity can for example reverse the slopes and cause many other issues in the model.

It had not been done so far, but now it is corrected: correlation factor between Sedges and Horsetails was 0.73, so Sedges factor was removed from the analysis since sedges covered only 0.5% of one of our sites vs. horsetails were found in two sites and covered 30% of one of them. The correlation table now figures in the supplementary materials.

339: What do you mean by testing any regression?

The original wording was wrong, by "running" would have been better. Nevertheless, it was completely removed from the sentence for a better clarification.

440: Did you control for soil moisture, soil temperature and air temperature variations between the two groups here? Did you do the measurements at the control and experiment during the same day and same conditions? How robust is this comparison?

Soil temperature and soil moisture were measured at the same time as the fluxes were measured. There were no statistical differences between the two treatments (Fig. S7). Each individual lawn/meadow pair was always measured consecutively during the same day to achieve as similar conditions as possible.

450: But the R^2 is still high for some variables? Why is that? Also, is it possible that the variables would be non-linearly related but you did not test that? I would consider exploring the non-linearity here too. Also, how much variability did you have for species richness metrics (and how does that compare with earlier studies)? If there is not much variability in the first place, it cannot really explain the fluxes either.

This richness section has been deleted from the manuscript as mentioned above, because the grass identification was missing in the study.

Minor

19: Probably good to write these as full words when first mentioned (carbon dioxide,…). Same with GHG, C, and N.

They are now mentioned as full word.

34: Slightly confusing to refer to “these intensive study sites” and “six other pairs” here when they have not been introduced in the abstract

We made it clearer.

25: Can you give some numbers for the sink strength (e.g. just for the growing season)?

It is now added to the revised abstract.

25: slightly confusing way to use the word “whereas” here. It usually is tightly linked to the previous sentence (e.g., lawns are clear sinks, whereas meadows are sources) – now these two sentences do not seem to be linked at all

This is now corrected.

29: I would write a stronger concluding sentence

Thank you, it makes sense. We changed the concluding sentence as follows: “Lastly, the proportion of flowering plants (forbs, including legumes) was found to be higher in meadows than in lawns. Even though other aspects would need to be investigated, these results can be taken into account in urban planning when considering biodiversity and carbon-smartness.”

31: You need more references throughout the paragraph. I would also consider including the urban perspective here all ready

We added more references throughout the paragraph and included the urban perspective already earlier in the introduction.

55-58: References needed

We added key references here as well.

64: i.e. meadows? Make sure that meadows are clearly defined too

It is now defined clearly.

70: Nice and clear research questions!

Thank you!

77: Are the recently created meadows not urban because that word is not used here but it is used in the other vegetation types?

No, they are all urban, but we just omitted how crucial it could be to have it. We clarified the nomenclature throughout the manuscript.

125: It would be more interesting to show the distribution of urban green spaces here instead of roads.

We agree and we also thought to show urban green spaces first, but there is no urban green spaces layer available for the whole Helsinki Metropolitan Area, but just for some cities. Therefore, we don’t have enough material to improve the map further. The main idea of the map is to show how the sites are spread around the sites and not for example the green corridors which are not in the focus of this manuscript.

130: Add uncertainty estimates or standard deviations? Also, what is the type of vegetation column, it was not described in the text and seems to be a mixture of genus, moisture and mixture definitions?

We added standard deviation for the values, that were not just from a single measurement. The vegetation type column is here to give an idea of the vegetation found at each site. We changed it to "Vegetation’s characteristics" for a better understanding.

365: The uncertainties in the plot are standard deviations across years? So should the standard deviations be the same for the two years for the same time period?

The wording in the caption was apparently confusing, but the standard deviation was calculated for each day of measurements with the four replicates measured at a site. It is now rephrased in the caption.

479: Can you give a clearer and stronger concluding sentence here?

We arranged a stronger concluding sentence as recommended, as follows: “Furthermore, as expected, the forbs proportion was found to be higher in meadows than in lawns and seems to be negatively connected to N2O fluxes and C/N ratio, and positively to SOC.”

485: Are you talking about annual fluxes here now?

It is the case, and it has been specified in the revised version.

555: not just water use efficiency, but also light and nutrient use efficiency?

It is now added into the revised manuscript.

613: Is this statement true? On some years the average CH4 and N2O fluxes were positive.

This is true, in some years the fluxes were positive, but when comparing lawns to meadows they are similar, so the positive fluxes are not due to the transformation it-self. This part was slightly rephrased to make it clearer.

617: What do you mean by larger scale?

It was there to mention a higher number of sites, which is now clarified.

626: What about the analysis codes, will you share them too?

Gasmet and LI-COR codes will be shared in the new Code availability section at the end of the manuscript.