the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 08 Apr 2024
CHclim25 – a spatially and temporally very high-resolution climatic dataset for Switzerland
Abstract. CHclim25 is a climatic dataset with a 25 m resolution for Switzerland that includes daily, monthly and yearly layers for temperature, precipitation, relative sunshine duration, growing degree-days, potential evapotranspiration, bioclimatic variables and aridity. The dataset is downscaled from a daily 1 km resolution dataset from the Swiss federal agency for meteorology using local regressions with an elevation model to better account for local topography and complex local climatic phenomena. Climatic layers are provided for individual years, 1981–2010 baseline period and future periods 2020–2049, 2045–2074, and 2070–209. Future layers incorporate three regional/global circulation models and three representative concentration pathways. We compare our predictions with values observed at independent weather stations and show that errors are minimal in comparison to the original dataset at 1 km resolution, and that the dataset is more accurate than available climatic global datasets at 30’ resolution, especially at high elevation. CHclim25 improves the temporal and spatial accuracy of climatic data available for Switzerland and enables new studies at very high resolution in ecology and environmental sciences.
- Preprint
(1425 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on essd-2024-79', Anonymous Referee #1, 07 May 2024
The manuscript by Olivier Broennimann et al. presents CHclim25, a climatic dataset with a 25m resolution for Switzerland.
In general, I believe that the manuscript needs at the very least a great deal of rewriting and rethinking. I feel that there is a lack of awareness of the complexity of the subject matter and the dataset presented is created far too simplistically for the advertised use. I recommend rejection because there is not enough time for the review to change the manuscript accordingly and recommend resubmitting it after careful redrafting.
General comments
What I lack is a critical sense of the work done and the dataset created. Its qualities are largely praised but attention is not paid to the many problems that one must be aware of when using such a dataset. First and foremost is the 25-meter resolution. In itself, one could create datasets with a resolution of 1 millimeter or even less, but this does not mean that the effective resolution of the dataset is really that. The fact that the existing datasets mostly are in the km resolutions is precisely because the lack of information (and most of the time it does not even correspond to the effective resolution). Adding a local regression or using bilinear interpolation is a “poor” downscaling, which can certainly have a sound purpose in certain cases, but this needs to be explained very precisely to avoid user thinking to really have a gridded dataset with 25m of resolution and being able to resolve at this scale. In the manuscript I see a huge discrepancy between the introduction and the discussion, where publications are cited that explain several issues depending on the variable (for example inversions, cool air pools) as well as “microclimatic” aspects and then the dataset with a very simple downscaling.
A detailed analysis of the performance of the dataset is missing. In the manuscript, there is only an analysis (moreover monthly, thus losing important information on a daily level) with respect to stations. As the author acknowledges, the comparison with the "NBCN" stations has the problem that they are already present in the source dataset. For this reason, it is not particularly useful information for characterizing the new dataset. The other stations used are also rather special (either at high altitudes or in cities, where an effect of urbanization on temperatures is to be expected in part). Besides being a far too limited evaluation, almost nothing is explained about it. Even the discussion, where one would expect a detailed look at what has been found, ends up in a simple list and results of other publications (L169-189), which would make more sense in the introduction or if there were a link here to the method or the results found.
Some statements are not supported by concrete analyses and results, as “CHclim25 dataset represents a significant advancement in the availability of high-resolution climatic data in Switzerland” or “The downscaling process employed, using local regressions with elevation models, enhances the dataset's ability to capture complex local climatic phenomena, especially in regions with rugged terrain”. With such downscaling, where would the ability to capture complex local climatic phenomena come from? How can microclimates that "strongly influence ecological patterns and dynamics" be represented?
Reading the last sentence, a person interested in the dataset really thinks they have fields that allow them to analyses on the 'micro' scale, which is incorrect apart from the consideration of the height for temperature and horizontal position for precipitation. But this is a far from for example detecting cold pools in a depression (an example for “microclimate”). With this dataset it is not possible to address "the limitations of macro-climatic data".
Specific comments
- L58: please use the same abbreviation in text and plots (Tave vs. TaveD, Prec vs PrecD,…)
- The dataset used for downscaling comes from MeteoSwiss. In the manuscript it is sometimes written Meteoswiss, sometimes MeteoSwiss and Swiss federal agency for meteorology. Better check the correct affiliation (I think it should be Federal Office of Meteorology and Climatology MeteoSwiss or MeteoSwiss)
- Note the big difference for the relative sunshine duration between the monthly average simply by averaging the percentages and the true monthly value by passing through the correct value of the maximum possible sunshine duration on a certain month.
- L220 typo: also instead of aslo
Citation: https://doi.org/10.5194/essd-2024-79-RC1 -
AC1: 'Reply on RC1', Olivier Broennimann, 25 Jun 2024
First, we would like to thank reviewer 1 for the valuable comments on the paper and the dataset, which will contribute to significantly improve the manuscript, and would like to apologize for not properly explaining the context of this dataset, which caused some misunderstanding on its intended use. As we explain in our general comment below, we fully agree that the dataset should not be considered for precise analyses at the final scale, i.e. to paraphrase the reviewer “a person interested in the dataset should not think they have fields that allow them to run analyses on the micro scale” This will be made crystal clear in the revision. Instead, this dataset was mainly developed to provide a finer scale set of bioclimatic maps at relatively high resolution (25 m) for applications in spatial ecology and biogeography in mountainous landscapes. Accordingly, we propose to change the title to: ” CHclim25 – a spatially and temporally high-resolution bioclimatic dataset for ecological applications in Switzerland”. Hereafter we detail how the problems mentioned by the reviewer will be addressed in the revised version.
About the unjustified choice of a 25-m resolution, we agree that it was not sufficiently described and supported in the current version of the manuscript, and we agree that the resolution of the downscaling procedure is not directly linked to the precision of the dataset. The 25-m resolution chosen for the dataset corresponds to the resolution of most gridded topographic and land cover datasets available for Switzerland. Historically it stems from the use of the national Swiss DEM that has been available at this resolution for decades. 25 m is therefore commonly used as a standard resolution for modelling studies and conservation planning in Switzerland (see examples in our general comment below). For instance, the recently published SWECO25 dataset uses this resolution as a standard for many cross thematic environmental GIS layers spanning from climate, vegetation, land use and land cover information across Switzerland. Given the size of the country, it is a good trade-off between spatial accuracy, resolution of input sources, and size of output databases (Külling et al. 2024, Scientific Data). Note that the national ecological infrastructure for Switzerland is currently being developed at this resolution (partly based on our dataset) by the federal and cantonal authorities (i.e. https://valpar.ch/index_en.php?page=home_en). We are willing to add these explanations in the text. In the revised manuscript, we will recenter the introduction and discussion on the importance of using 25-m resolution maps in ecological and biogeographical applications such as modelling species distributions and spatial conservation planning. We will better explain why climatic data at 1 km such as Worldclim and Chelsa are not informative enough, especially to depict temperature and precipitation patterns for use in ecological research and applications in mountain landscapes, and why obtaining a bioclimatic dataset a 25 m resolution fills a strong need by researchers and practitioners (see again the already numerous usages - i.e. citations and downloads - of this dataset in ecological research in Switzerland).
Next, we acknowledge that analysis of the performance of the dataset could be improved and better described. We started this data paper with the idea of being succinct, but we agree more explanations, and potentially more analyses, were needed. As the reviewer rightly says, NBCN stations are used to calibrate the downscaling procedure and comparisons between observed and modeled climatic values at these stations are thus not useful as an independent evaluation. This dependent dataset was used to illustrate the magnitude of errors present in the independent datasets; this was probably not stated clearly enough. The independent datasets cover respectively low elevation sites (close to cities, but cities are everywhere at low elevation in Switzerland) and high elevation environments. Figures 2 and 3 illustrated quite clearly in our opinion how errors in CHclim25 are minimal at low elevation and increase at higher elevation but remains largely inferior compared to Chelsa and Worlclim2 datasets. We do agree however that if the figures are clear, the result and the discussion sections could have been more elaborated and will be vastly improved in the revised version. Note that we have now identified another independent dataset of weather stations for agricultural areas, which could be gathered and tested to complement the current validation. If deemed necessary by the editor, we would be happy to complement the analyses by including these. Finally, it is correct that the evaluation of the dataset at daily temporal resolution was not provided. We indeed decided not to provide this evaluation on purpose because we focused on the monthly and yearly maps that are mainly used in ecological applications, but we will provide this additional evaluation in the revised version of the manuscript (see also our response to reviewer 2).
Finally, we would like to address the critic implying that we overstated our dataset's ability to capture complex local climatic phenomena, and that our dataset does not provide information about microclimate. Our dataset was clearly not meant to be an accurate microclimatic dataset, but rather a finer-grain representation of the official Swiss climate maps at 1 km resolution. We therefore apologize again if this was not appropriately explained and if the reviewer felt we overstated some qualities of our dataset, and especially about its ability to detect micro-climatic features such as cold pools: we fully agree that it cannot represent the accurate microclimatic variations. Our downscaling approach is based as input layers on the official MeteoSwiss gridded product at 1 km that already includes spatial interpolations at a daily temporal scale and deals with nonlinearities in the vertical thermal structure such that basin-scale inversions such as valley-scale cold pools and foehn should be detected (Frei et al. 2014). The detection of such basin-scale climatic features is thus achieved in the MeteoSwiss product at the 1 km resolution scale, not in our downscaling approach. We thus agree that the term “microclimate” is overstated here, and we will rewrite the concerned sections of the manuscript accordingly. In remains however that our downscaling approach further improves the quality of this daily dataset featuring basin-scale climatic features by incorporating more detailed topographic effects based on the adiabatic relationship of temperature with elevation at 25 m. The superiority of our dataset over classic macroscale datasets such as Chelsa and Worldclim thus stems from the combination of both approaches, daily nonlinearity interpolations at 1 km, further strengthened with local regressions. Showing that our dataset is closer to measured climatic values than other datasets indeed supports, in our opinion, the idea that it better captures local climatic phenomena, but again this rationale was probably not sufficiently explained, and the message was therefore misleading. We believe that the manuscript can easily be revised in this regard by (1) better specifying the intended use of the dataset (i.e. for bioclimatic-based ecological applications), and (2) by improving the explanation of the methodology behind the dataset.
Finally, please note that the fact that we state that microclimates strongly influence ecological patterns is not a result of this study but rather stems from previous ecological research. We mentioned this to stress the importance of developing such fine-resolution bioclimatic maps for ecological applications in Switzerland, and we will make sure to cite these studies appropriately in the revised manuscript.
Thanks again for all your valuable inputs.
Citation: https://doi.org/10.5194/essd-2024-79-AC1
-
RC2: 'Comment on essd-2024-79', Anonymous Referee #2, 14 May 2024
The authors describe a 25 m high-resolution dataset of various climate variables for Switzerland. The dataset is based on a downscaling of 1 km data of MeteoSwiss. In addition, it provides data for future periods based on a few climate models and three RCPs.
The downscaled data seems to show good skill in the comparison to reference datasets, and could be useful for high resolution applications e.g., in ecolology. However, the validation exercise is quite limited and only considers monthly mean data, while the daily data is not validated at all. Why this limitation to monthly data only? In my opinion, also the daily data needs to be evaluated if the dataset is "sold" as such.
Regarding the methodological aspects, there is no motivation provided for using simple bilinear interpolation for the non-temperature variables. How justified is such a simple approach given the large spatial heterogeneity of precipitation in Switzerland, and also given the daily time scale that is used here?
The presentation of the dataset is appropriate for a technical data paper but could have been done more carefully. There a quite a few inconsistencies throughout the text e.g., naming of institutions, units etc. Therefore, the paper may need some rewriting to make it more concise.
Navigation through the provided data on zenodo was difficult due to unclear file naming conventions (see specific comments below on the data organisation and file name conventions). I propose to provide a proper description of the naming conventions either in the paper or on zenodo to guide the user.
As such, the paper and the provided data needs major revisions before publication.
Specific comments:
Line 12/13: Official name of MeteoSwiss is “Federal Office of Meteorology and Climatology”.
Line 15: “209” should be “2099”.
Line 40: Again, the official name is “Federal Office of Meteorology and Climatology”.
Line 43: There is no footnote 6 in my document. There must likely be a proper reference document from MeteoSwiss describing the dataset.
Line 48: The recommended reference period has been updated to 1991-2020 by WMO. I think the dataset would be more valuable if you could align it with this new reference period.
Line 49/50: What do you mean with compatible with the 5th assessment report, in what sense? Also, there is already the 6th assessment report of IPCC out now. Why not making you dataset “compatible” with the latest assessment report?
Line 69: I could not find the argument method “nbg” in the resample function of the raster package. I guess you mean nearest neighbour “ngb”?
Line 82: “periode” should be “period”.
Line 89: What means “D” here in the spatial resolution?
Line 90: Propose to use “Changes” instead of “Anomalies”.
Line 97: Please use consistently either "°" or "degree" throughout the manuscript.
Line 106: “R-package dismo” instead of “package dismo”.
Line 113: “variables” instead of “variable”.
Line 119: “time series” instead of “climatologies”.
Line 127: Change “for monthly averages for the 1981-2010 period” to “for long-term monthly averages of the 1981-2010 period” to clarify the procedure.
Line 128-132: Please rephrase to increase readability and clarity.
Line 135, Figure caption: Please note in the caption that y-axes have differing ranges between reference datasets.
Line 149/150: “Interestingly, errors on 30 year monthly averages for the high elevation sites are lower than errors for separate years, …” Wouldn't one expect this behaviour since the representation of the mean yearly cycle is much easier to achieve than the year-to-year variability?
Line 173: “organism-centered” instead of “organism-cantered”Data organisation and file name conventions:
- What is the meaning of the 10, 18, 50 and 98 km in the file naming? (e.g., in Tave_monthly_1981_2010_average.zip)
- What is the meaning of the suffix “climatologies” in these cases? What's the difference to the monthly averages, which I also interpret as climatologies?
- Also, there are partly .rData files instead of .tif files in the repository. Why? (e.g, TmaxY_1989.rData and others)These points are not described in the paper or the description on zenodo and need clarification. I propose to provide a proper description of naming conventions either in the paper or on zenodo to guide the user.
Table 1:
- Why not at least providing the daily layers as climatologies (i.e., 1981-2010 average, 2020-2049 average etc.)?
- Individual years of yearly, monthly and daily layers are only available for the historical period, is this correct? Please indicate this in the table.Citation: https://doi.org/10.5194/essd-2024-79-RC2 -
AC2: 'Reply on RC2', Olivier Broennimann, 25 Jun 2024
We thank reviewer 2 for the useful and valuable comments, and for acknowledging that “the downscaled data seems to show good skill in the comparison to reference datasets, and could be useful for high resolution applications e.g., in ecology”, but that “the validation exercise is quite limited and only considers monthly mean data”. We acknowledge that the main applications of this bioclimatic dataset could be in ecology and biogeography (or similar “non-climatic” fields), which we will also better specify (see our response to reviewer 1) and fully agree also that the validation of the dataset can be further improved, and the comments provided will be very helpful in this regard. It is correct that the evaluation of the dataset at daily temporal resolution was not provided. As also responded to reviewer 1, we decided on purpose not to provide it, because we wanted to focus on the monthly and yearly maps that are mainly used in ecological applications, but we understand the reviewer’s concern and will now provide this information in the revised version of the manuscript.
We would like to further address the concern about the lack of rationale, in the manuscript, for using simple bilinear interpolations for the non-temperature (i.e. precipitation) variables. Theoretical support for downscaling temperature based on the relationship between temperature and elevation derives from the adiabatic process which is well described in thermodynamics. Precipitations on the contrary are driven by more complex factors and less related to elevation (at least at local scale), but of which regional features are already largely included in the 1-km Swiss climatic maps. In a previous version of the dataset, we indeed downscaled precipitation from 1 km to 25 m using local regressions with elevation and difference in elevation, but the results were not conclusive. We thus decided to use instead a simple bilinear interpolation, mostly to provide precipitation layers at the same resolution as temperature layers. Note that at daily temporal scale, the amount of precipitation is quite well delineated at 1 km resolution in the MeteoSwiss dataset, with interpolations derived from 520 rain-gauge stations (RhiresD, MeteoSwiss Grid-Data Products). We will add this missing information to the revised manuscript.
Finally, we acknowledge the difficulties of navigating the files on Zenodo due to the unclear file naming conventions. In the “1981_2010_average” folders of temperature variables, files without “climatologies” in the name are calculated by averaging daily layers at 25 m (and thus integrate >900 layers (i.e. 30 days x 30 years), each of them based on an individual daily downscaling procedures). These are the core layers of the dataset. The ones with “climatologies” in the name are calculated from a single downscaling procedure based on the average of the daily layers at 1 km (way less computationally intensive, but the impact of daily local phenomena such as cold air pools are likely blurred). But the ease of calculation allowed us to test different moving windows in the downscaling procedure with increasing diameters of 10, 18, 50, and 98 km (note that in figure 3 the radius instead of the diameter of the window is indicated). These “climatologies” layers are used here mostly for testing. In the new version of the manuscript, we will make sure that the description of layers and naming conventions are clear and consistent.
We also checked all the minor comments carefully. Thank you for spotting all the inconsistencies throughout the text. These are mostly clarification issues, which we will correct in the revised manuscript and should substantially improve the presentation of the dataset.
Thanks again for all your valuable inputs.
Citation: https://doi.org/10.5194/essd-2024-79-AC2
-
AC2: 'Reply on RC2', Olivier Broennimann, 25 Jun 2024
-
AC3: 'Comment on essd-2024-79', Olivier Broennimann, 25 Jun 2024
We read with interest the comments of the reviewers and we believe that they provide very important points for the improvement of the manuscript. Most importantly, we realize that the scope and intended use of the dataset was unclear, and that a redrafting of the introduction and discussion will be necessary to convey the correct message to the readers. We thus propose a new title “CHclim25 – a spatially and temporally high-resolution bioclimatic dataset for ecological applications in Switzerland” that should better reflect the new orientation of the manuscript. Accordingly, we would like to stress that the dataset we propose is not intended to provide absolute values at a 25 m resolution for climatic or meteorological applications, but rather to provide bioclimatic layers that can improve ecological modeling and biogeographic research, and ultimately conservation applications in Switzerland (e.g. Augustijnen et al. 2022, Eeckman et al. 2022, Mazel et al. 2022, Adde et al. 2023, Black et al. 2023, Ortiz-Rodriguez et al. 2023, Kulling et al.2024a, b, Verdon et al. 2024, ...).
The availability of climatic information at such a resolution that is compatible with the scale at which ecological applications are conducted is crucial. So far, the only dataset available at such a fine resolution (25 m) for the whole Switzerland was a bioclimatic dataset developed by the Institute of Forest, Snow and Landscape (WSL) which provided 30-year average maps at 25 m resolution. However, this dataset covered the period 1961-1991 and was thus vastly outdated. Furthermore, it did not include predictions for the future under various climate change scenarios. This was the incentive to create this new bioclimatic dataset. By downscaling most recent and national daily MeteoSwiss climatic data from 1 km to 25 m with daily local regressions that incorporate the adiabatic effect of elevation on temperature, we thus substantially improve the fine-scale mapping of bioclimatic data that are critical to the fields mentioned above. As mentioned by reviewer 1, the absolute values at 25 m might not be perfect, but errors remains fairly low, and the dataset is anyway a great improvement from only having access to information at 1 km resolution, which in a rugged environment such as the Swiss mountains encompasses far too much climatic variability (i.e. elevation gradient of several hundreds of meters can occur in a 1 km pixel) to inform usefully ecological processes.
If granted by the editor, we will thus redraft the manuscript to cover these aspects. We further agree with the two reviewers, that the validation of the dataset could be improved and better described, with a validation of predictions not only at monthly and yearly temporal scales, but also at daily scale. We could also constitute a third independent evaluation dataset from meteorological stations in agricultural areas. In the new version of the manuscript, we will make sure that the description of layers and naming conventions are clear and consistent to facilitation navigation on the Zenodo repository.
We hope that the proposed changes will convince the editor that our revised manuscript will meet the quality requirements of ESSD.
References:
H. Augustijnen, T. Patsiou, K. Lucek, Secondary contact rather than coexistence—Erebia butterflies in the Alps, Evolution, Volume 76, Issue 11, 1 November 2022, Pages 2669–2686, https://doi.org/10.1111/evo.14615
Adde, A., Rey, P.-L., Brun, P., Külling, N., Fopp, F., Altermatt, F., Broennimann, O., Lehmann, A., Petitpierre, B., Zimmermann, N.E., Pellissier, L. and Guisan, A. (2023), N-SDM: a high-performance computing pipeline for Nested Species Distribution Modelling. Ecography, 2023: e06540. https://doi.org/10.1111/ecog.06540
J. Eeckman, J.-M.Fallot et O. Broennimann, « Air temperature estimation at very high resolution in mountainous areas », Dynamiques environnementales, 49-50 | 2022, 85-97.
Mazel, F., Malard, L., Niculita-Hirzel, H., Yashiro, E., Mod, H.K., Mitchell, E.A.D., Singer, D., Buri, A., Pinto, E., Guex, N., Lara, E. and Guisan, A. (2022), Soil protist function varies with elevation in the Swiss Alps. Environ Microbiol, 24: 1689-1702. https://doi.org/10.1111/1462-2920.15686
B. Black, M.J. van Strien, A. Adde, A. Grêt-Regamey, Re-considering the status quo: Improving calibration of land use change models through validation of transition potential predictions, Environmental Modelling & Software, Volume 159,2023, https://doi.org/10.1016/j.envsoft.2022.105574.
Ortiz-Rodríguez DO, Guisan A, Van Strien MJ (2023) Sensitivity of habitat network models to changes in maximum dispersal distance. PLoS ONE 18(11): e0293966. https://doi.org/10.1371/journal.pone.0293966
Külling, N., Adde, A., Fopp, F. et al. SWECO25: a cross-thematic raster database for ecological research in Switzerland. Sci Data 11, 21 (2024). https://doi.org/10.1038/s41597-023-02899-1
Verdon, V., Malard, L., Collart, F., Adde, A., Yashiro, E., Lara Pandi, E., Mod, H., Singer, D., Niculita-Hirzel, H., Guex, N. and Guisan, A. (2024), Can we accurately predict the distribution of soil microorganism presence and relative abundance?. Ecography e07086. https://doi.org/10.1111/ecog.07086
Citation: https://doi.org/10.5194/essd-2024-79-AC3
Status: closed
-
RC1: 'Comment on essd-2024-79', Anonymous Referee #1, 07 May 2024
The manuscript by Olivier Broennimann et al. presents CHclim25, a climatic dataset with a 25m resolution for Switzerland.
In general, I believe that the manuscript needs at the very least a great deal of rewriting and rethinking. I feel that there is a lack of awareness of the complexity of the subject matter and the dataset presented is created far too simplistically for the advertised use. I recommend rejection because there is not enough time for the review to change the manuscript accordingly and recommend resubmitting it after careful redrafting.
General comments
What I lack is a critical sense of the work done and the dataset created. Its qualities are largely praised but attention is not paid to the many problems that one must be aware of when using such a dataset. First and foremost is the 25-meter resolution. In itself, one could create datasets with a resolution of 1 millimeter or even less, but this does not mean that the effective resolution of the dataset is really that. The fact that the existing datasets mostly are in the km resolutions is precisely because the lack of information (and most of the time it does not even correspond to the effective resolution). Adding a local regression or using bilinear interpolation is a “poor” downscaling, which can certainly have a sound purpose in certain cases, but this needs to be explained very precisely to avoid user thinking to really have a gridded dataset with 25m of resolution and being able to resolve at this scale. In the manuscript I see a huge discrepancy between the introduction and the discussion, where publications are cited that explain several issues depending on the variable (for example inversions, cool air pools) as well as “microclimatic” aspects and then the dataset with a very simple downscaling.
A detailed analysis of the performance of the dataset is missing. In the manuscript, there is only an analysis (moreover monthly, thus losing important information on a daily level) with respect to stations. As the author acknowledges, the comparison with the "NBCN" stations has the problem that they are already present in the source dataset. For this reason, it is not particularly useful information for characterizing the new dataset. The other stations used are also rather special (either at high altitudes or in cities, where an effect of urbanization on temperatures is to be expected in part). Besides being a far too limited evaluation, almost nothing is explained about it. Even the discussion, where one would expect a detailed look at what has been found, ends up in a simple list and results of other publications (L169-189), which would make more sense in the introduction or if there were a link here to the method or the results found.
Some statements are not supported by concrete analyses and results, as “CHclim25 dataset represents a significant advancement in the availability of high-resolution climatic data in Switzerland” or “The downscaling process employed, using local regressions with elevation models, enhances the dataset's ability to capture complex local climatic phenomena, especially in regions with rugged terrain”. With such downscaling, where would the ability to capture complex local climatic phenomena come from? How can microclimates that "strongly influence ecological patterns and dynamics" be represented?
Reading the last sentence, a person interested in the dataset really thinks they have fields that allow them to analyses on the 'micro' scale, which is incorrect apart from the consideration of the height for temperature and horizontal position for precipitation. But this is a far from for example detecting cold pools in a depression (an example for “microclimate”). With this dataset it is not possible to address "the limitations of macro-climatic data".
Specific comments
- L58: please use the same abbreviation in text and plots (Tave vs. TaveD, Prec vs PrecD,…)
- The dataset used for downscaling comes from MeteoSwiss. In the manuscript it is sometimes written Meteoswiss, sometimes MeteoSwiss and Swiss federal agency for meteorology. Better check the correct affiliation (I think it should be Federal Office of Meteorology and Climatology MeteoSwiss or MeteoSwiss)
- Note the big difference for the relative sunshine duration between the monthly average simply by averaging the percentages and the true monthly value by passing through the correct value of the maximum possible sunshine duration on a certain month.
- L220 typo: also instead of aslo
Citation: https://doi.org/10.5194/essd-2024-79-RC1 -
AC1: 'Reply on RC1', Olivier Broennimann, 25 Jun 2024
First, we would like to thank reviewer 1 for the valuable comments on the paper and the dataset, which will contribute to significantly improve the manuscript, and would like to apologize for not properly explaining the context of this dataset, which caused some misunderstanding on its intended use. As we explain in our general comment below, we fully agree that the dataset should not be considered for precise analyses at the final scale, i.e. to paraphrase the reviewer “a person interested in the dataset should not think they have fields that allow them to run analyses on the micro scale” This will be made crystal clear in the revision. Instead, this dataset was mainly developed to provide a finer scale set of bioclimatic maps at relatively high resolution (25 m) for applications in spatial ecology and biogeography in mountainous landscapes. Accordingly, we propose to change the title to: ” CHclim25 – a spatially and temporally high-resolution bioclimatic dataset for ecological applications in Switzerland”. Hereafter we detail how the problems mentioned by the reviewer will be addressed in the revised version.
About the unjustified choice of a 25-m resolution, we agree that it was not sufficiently described and supported in the current version of the manuscript, and we agree that the resolution of the downscaling procedure is not directly linked to the precision of the dataset. The 25-m resolution chosen for the dataset corresponds to the resolution of most gridded topographic and land cover datasets available for Switzerland. Historically it stems from the use of the national Swiss DEM that has been available at this resolution for decades. 25 m is therefore commonly used as a standard resolution for modelling studies and conservation planning in Switzerland (see examples in our general comment below). For instance, the recently published SWECO25 dataset uses this resolution as a standard for many cross thematic environmental GIS layers spanning from climate, vegetation, land use and land cover information across Switzerland. Given the size of the country, it is a good trade-off between spatial accuracy, resolution of input sources, and size of output databases (Külling et al. 2024, Scientific Data). Note that the national ecological infrastructure for Switzerland is currently being developed at this resolution (partly based on our dataset) by the federal and cantonal authorities (i.e. https://valpar.ch/index_en.php?page=home_en). We are willing to add these explanations in the text. In the revised manuscript, we will recenter the introduction and discussion on the importance of using 25-m resolution maps in ecological and biogeographical applications such as modelling species distributions and spatial conservation planning. We will better explain why climatic data at 1 km such as Worldclim and Chelsa are not informative enough, especially to depict temperature and precipitation patterns for use in ecological research and applications in mountain landscapes, and why obtaining a bioclimatic dataset a 25 m resolution fills a strong need by researchers and practitioners (see again the already numerous usages - i.e. citations and downloads - of this dataset in ecological research in Switzerland).
Next, we acknowledge that analysis of the performance of the dataset could be improved and better described. We started this data paper with the idea of being succinct, but we agree more explanations, and potentially more analyses, were needed. As the reviewer rightly says, NBCN stations are used to calibrate the downscaling procedure and comparisons between observed and modeled climatic values at these stations are thus not useful as an independent evaluation. This dependent dataset was used to illustrate the magnitude of errors present in the independent datasets; this was probably not stated clearly enough. The independent datasets cover respectively low elevation sites (close to cities, but cities are everywhere at low elevation in Switzerland) and high elevation environments. Figures 2 and 3 illustrated quite clearly in our opinion how errors in CHclim25 are minimal at low elevation and increase at higher elevation but remains largely inferior compared to Chelsa and Worlclim2 datasets. We do agree however that if the figures are clear, the result and the discussion sections could have been more elaborated and will be vastly improved in the revised version. Note that we have now identified another independent dataset of weather stations for agricultural areas, which could be gathered and tested to complement the current validation. If deemed necessary by the editor, we would be happy to complement the analyses by including these. Finally, it is correct that the evaluation of the dataset at daily temporal resolution was not provided. We indeed decided not to provide this evaluation on purpose because we focused on the monthly and yearly maps that are mainly used in ecological applications, but we will provide this additional evaluation in the revised version of the manuscript (see also our response to reviewer 2).
Finally, we would like to address the critic implying that we overstated our dataset's ability to capture complex local climatic phenomena, and that our dataset does not provide information about microclimate. Our dataset was clearly not meant to be an accurate microclimatic dataset, but rather a finer-grain representation of the official Swiss climate maps at 1 km resolution. We therefore apologize again if this was not appropriately explained and if the reviewer felt we overstated some qualities of our dataset, and especially about its ability to detect micro-climatic features such as cold pools: we fully agree that it cannot represent the accurate microclimatic variations. Our downscaling approach is based as input layers on the official MeteoSwiss gridded product at 1 km that already includes spatial interpolations at a daily temporal scale and deals with nonlinearities in the vertical thermal structure such that basin-scale inversions such as valley-scale cold pools and foehn should be detected (Frei et al. 2014). The detection of such basin-scale climatic features is thus achieved in the MeteoSwiss product at the 1 km resolution scale, not in our downscaling approach. We thus agree that the term “microclimate” is overstated here, and we will rewrite the concerned sections of the manuscript accordingly. In remains however that our downscaling approach further improves the quality of this daily dataset featuring basin-scale climatic features by incorporating more detailed topographic effects based on the adiabatic relationship of temperature with elevation at 25 m. The superiority of our dataset over classic macroscale datasets such as Chelsa and Worldclim thus stems from the combination of both approaches, daily nonlinearity interpolations at 1 km, further strengthened with local regressions. Showing that our dataset is closer to measured climatic values than other datasets indeed supports, in our opinion, the idea that it better captures local climatic phenomena, but again this rationale was probably not sufficiently explained, and the message was therefore misleading. We believe that the manuscript can easily be revised in this regard by (1) better specifying the intended use of the dataset (i.e. for bioclimatic-based ecological applications), and (2) by improving the explanation of the methodology behind the dataset.
Finally, please note that the fact that we state that microclimates strongly influence ecological patterns is not a result of this study but rather stems from previous ecological research. We mentioned this to stress the importance of developing such fine-resolution bioclimatic maps for ecological applications in Switzerland, and we will make sure to cite these studies appropriately in the revised manuscript.
Thanks again for all your valuable inputs.
Citation: https://doi.org/10.5194/essd-2024-79-AC1
-
RC2: 'Comment on essd-2024-79', Anonymous Referee #2, 14 May 2024
The authors describe a 25 m high-resolution dataset of various climate variables for Switzerland. The dataset is based on a downscaling of 1 km data of MeteoSwiss. In addition, it provides data for future periods based on a few climate models and three RCPs.
The downscaled data seems to show good skill in the comparison to reference datasets, and could be useful for high resolution applications e.g., in ecolology. However, the validation exercise is quite limited and only considers monthly mean data, while the daily data is not validated at all. Why this limitation to monthly data only? In my opinion, also the daily data needs to be evaluated if the dataset is "sold" as such.
Regarding the methodological aspects, there is no motivation provided for using simple bilinear interpolation for the non-temperature variables. How justified is such a simple approach given the large spatial heterogeneity of precipitation in Switzerland, and also given the daily time scale that is used here?
The presentation of the dataset is appropriate for a technical data paper but could have been done more carefully. There a quite a few inconsistencies throughout the text e.g., naming of institutions, units etc. Therefore, the paper may need some rewriting to make it more concise.
Navigation through the provided data on zenodo was difficult due to unclear file naming conventions (see specific comments below on the data organisation and file name conventions). I propose to provide a proper description of the naming conventions either in the paper or on zenodo to guide the user.
As such, the paper and the provided data needs major revisions before publication.
Specific comments:
Line 12/13: Official name of MeteoSwiss is “Federal Office of Meteorology and Climatology”.
Line 15: “209” should be “2099”.
Line 40: Again, the official name is “Federal Office of Meteorology and Climatology”.
Line 43: There is no footnote 6 in my document. There must likely be a proper reference document from MeteoSwiss describing the dataset.
Line 48: The recommended reference period has been updated to 1991-2020 by WMO. I think the dataset would be more valuable if you could align it with this new reference period.
Line 49/50: What do you mean with compatible with the 5th assessment report, in what sense? Also, there is already the 6th assessment report of IPCC out now. Why not making you dataset “compatible” with the latest assessment report?
Line 69: I could not find the argument method “nbg” in the resample function of the raster package. I guess you mean nearest neighbour “ngb”?
Line 82: “periode” should be “period”.
Line 89: What means “D” here in the spatial resolution?
Line 90: Propose to use “Changes” instead of “Anomalies”.
Line 97: Please use consistently either "°" or "degree" throughout the manuscript.
Line 106: “R-package dismo” instead of “package dismo”.
Line 113: “variables” instead of “variable”.
Line 119: “time series” instead of “climatologies”.
Line 127: Change “for monthly averages for the 1981-2010 period” to “for long-term monthly averages of the 1981-2010 period” to clarify the procedure.
Line 128-132: Please rephrase to increase readability and clarity.
Line 135, Figure caption: Please note in the caption that y-axes have differing ranges between reference datasets.
Line 149/150: “Interestingly, errors on 30 year monthly averages for the high elevation sites are lower than errors for separate years, …” Wouldn't one expect this behaviour since the representation of the mean yearly cycle is much easier to achieve than the year-to-year variability?
Line 173: “organism-centered” instead of “organism-cantered”Data organisation and file name conventions:
- What is the meaning of the 10, 18, 50 and 98 km in the file naming? (e.g., in Tave_monthly_1981_2010_average.zip)
- What is the meaning of the suffix “climatologies” in these cases? What's the difference to the monthly averages, which I also interpret as climatologies?
- Also, there are partly .rData files instead of .tif files in the repository. Why? (e.g, TmaxY_1989.rData and others)These points are not described in the paper or the description on zenodo and need clarification. I propose to provide a proper description of naming conventions either in the paper or on zenodo to guide the user.
Table 1:
- Why not at least providing the daily layers as climatologies (i.e., 1981-2010 average, 2020-2049 average etc.)?
- Individual years of yearly, monthly and daily layers are only available for the historical period, is this correct? Please indicate this in the table.Citation: https://doi.org/10.5194/essd-2024-79-RC2 -
AC2: 'Reply on RC2', Olivier Broennimann, 25 Jun 2024
We thank reviewer 2 for the useful and valuable comments, and for acknowledging that “the downscaled data seems to show good skill in the comparison to reference datasets, and could be useful for high resolution applications e.g., in ecology”, but that “the validation exercise is quite limited and only considers monthly mean data”. We acknowledge that the main applications of this bioclimatic dataset could be in ecology and biogeography (or similar “non-climatic” fields), which we will also better specify (see our response to reviewer 1) and fully agree also that the validation of the dataset can be further improved, and the comments provided will be very helpful in this regard. It is correct that the evaluation of the dataset at daily temporal resolution was not provided. As also responded to reviewer 1, we decided on purpose not to provide it, because we wanted to focus on the monthly and yearly maps that are mainly used in ecological applications, but we understand the reviewer’s concern and will now provide this information in the revised version of the manuscript.
We would like to further address the concern about the lack of rationale, in the manuscript, for using simple bilinear interpolations for the non-temperature (i.e. precipitation) variables. Theoretical support for downscaling temperature based on the relationship between temperature and elevation derives from the adiabatic process which is well described in thermodynamics. Precipitations on the contrary are driven by more complex factors and less related to elevation (at least at local scale), but of which regional features are already largely included in the 1-km Swiss climatic maps. In a previous version of the dataset, we indeed downscaled precipitation from 1 km to 25 m using local regressions with elevation and difference in elevation, but the results were not conclusive. We thus decided to use instead a simple bilinear interpolation, mostly to provide precipitation layers at the same resolution as temperature layers. Note that at daily temporal scale, the amount of precipitation is quite well delineated at 1 km resolution in the MeteoSwiss dataset, with interpolations derived from 520 rain-gauge stations (RhiresD, MeteoSwiss Grid-Data Products). We will add this missing information to the revised manuscript.
Finally, we acknowledge the difficulties of navigating the files on Zenodo due to the unclear file naming conventions. In the “1981_2010_average” folders of temperature variables, files without “climatologies” in the name are calculated by averaging daily layers at 25 m (and thus integrate >900 layers (i.e. 30 days x 30 years), each of them based on an individual daily downscaling procedures). These are the core layers of the dataset. The ones with “climatologies” in the name are calculated from a single downscaling procedure based on the average of the daily layers at 1 km (way less computationally intensive, but the impact of daily local phenomena such as cold air pools are likely blurred). But the ease of calculation allowed us to test different moving windows in the downscaling procedure with increasing diameters of 10, 18, 50, and 98 km (note that in figure 3 the radius instead of the diameter of the window is indicated). These “climatologies” layers are used here mostly for testing. In the new version of the manuscript, we will make sure that the description of layers and naming conventions are clear and consistent.
We also checked all the minor comments carefully. Thank you for spotting all the inconsistencies throughout the text. These are mostly clarification issues, which we will correct in the revised manuscript and should substantially improve the presentation of the dataset.
Thanks again for all your valuable inputs.
Citation: https://doi.org/10.5194/essd-2024-79-AC2
-
AC2: 'Reply on RC2', Olivier Broennimann, 25 Jun 2024
-
AC3: 'Comment on essd-2024-79', Olivier Broennimann, 25 Jun 2024
We read with interest the comments of the reviewers and we believe that they provide very important points for the improvement of the manuscript. Most importantly, we realize that the scope and intended use of the dataset was unclear, and that a redrafting of the introduction and discussion will be necessary to convey the correct message to the readers. We thus propose a new title “CHclim25 – a spatially and temporally high-resolution bioclimatic dataset for ecological applications in Switzerland” that should better reflect the new orientation of the manuscript. Accordingly, we would like to stress that the dataset we propose is not intended to provide absolute values at a 25 m resolution for climatic or meteorological applications, but rather to provide bioclimatic layers that can improve ecological modeling and biogeographic research, and ultimately conservation applications in Switzerland (e.g. Augustijnen et al. 2022, Eeckman et al. 2022, Mazel et al. 2022, Adde et al. 2023, Black et al. 2023, Ortiz-Rodriguez et al. 2023, Kulling et al.2024a, b, Verdon et al. 2024, ...).
The availability of climatic information at such a resolution that is compatible with the scale at which ecological applications are conducted is crucial. So far, the only dataset available at such a fine resolution (25 m) for the whole Switzerland was a bioclimatic dataset developed by the Institute of Forest, Snow and Landscape (WSL) which provided 30-year average maps at 25 m resolution. However, this dataset covered the period 1961-1991 and was thus vastly outdated. Furthermore, it did not include predictions for the future under various climate change scenarios. This was the incentive to create this new bioclimatic dataset. By downscaling most recent and national daily MeteoSwiss climatic data from 1 km to 25 m with daily local regressions that incorporate the adiabatic effect of elevation on temperature, we thus substantially improve the fine-scale mapping of bioclimatic data that are critical to the fields mentioned above. As mentioned by reviewer 1, the absolute values at 25 m might not be perfect, but errors remains fairly low, and the dataset is anyway a great improvement from only having access to information at 1 km resolution, which in a rugged environment such as the Swiss mountains encompasses far too much climatic variability (i.e. elevation gradient of several hundreds of meters can occur in a 1 km pixel) to inform usefully ecological processes.
If granted by the editor, we will thus redraft the manuscript to cover these aspects. We further agree with the two reviewers, that the validation of the dataset could be improved and better described, with a validation of predictions not only at monthly and yearly temporal scales, but also at daily scale. We could also constitute a third independent evaluation dataset from meteorological stations in agricultural areas. In the new version of the manuscript, we will make sure that the description of layers and naming conventions are clear and consistent to facilitation navigation on the Zenodo repository.
We hope that the proposed changes will convince the editor that our revised manuscript will meet the quality requirements of ESSD.
References:
H. Augustijnen, T. Patsiou, K. Lucek, Secondary contact rather than coexistence—Erebia butterflies in the Alps, Evolution, Volume 76, Issue 11, 1 November 2022, Pages 2669–2686, https://doi.org/10.1111/evo.14615
Adde, A., Rey, P.-L., Brun, P., Külling, N., Fopp, F., Altermatt, F., Broennimann, O., Lehmann, A., Petitpierre, B., Zimmermann, N.E., Pellissier, L. and Guisan, A. (2023), N-SDM: a high-performance computing pipeline for Nested Species Distribution Modelling. Ecography, 2023: e06540. https://doi.org/10.1111/ecog.06540
J. Eeckman, J.-M.Fallot et O. Broennimann, « Air temperature estimation at very high resolution in mountainous areas », Dynamiques environnementales, 49-50 | 2022, 85-97.
Mazel, F., Malard, L., Niculita-Hirzel, H., Yashiro, E., Mod, H.K., Mitchell, E.A.D., Singer, D., Buri, A., Pinto, E., Guex, N., Lara, E. and Guisan, A. (2022), Soil protist function varies with elevation in the Swiss Alps. Environ Microbiol, 24: 1689-1702. https://doi.org/10.1111/1462-2920.15686
B. Black, M.J. van Strien, A. Adde, A. Grêt-Regamey, Re-considering the status quo: Improving calibration of land use change models through validation of transition potential predictions, Environmental Modelling & Software, Volume 159,2023, https://doi.org/10.1016/j.envsoft.2022.105574.
Ortiz-Rodríguez DO, Guisan A, Van Strien MJ (2023) Sensitivity of habitat network models to changes in maximum dispersal distance. PLoS ONE 18(11): e0293966. https://doi.org/10.1371/journal.pone.0293966
Külling, N., Adde, A., Fopp, F. et al. SWECO25: a cross-thematic raster database for ecological research in Switzerland. Sci Data 11, 21 (2024). https://doi.org/10.1038/s41597-023-02899-1
Verdon, V., Malard, L., Collart, F., Adde, A., Yashiro, E., Lara Pandi, E., Mod, H., Singer, D., Niculita-Hirzel, H., Guex, N. and Guisan, A. (2024), Can we accurately predict the distribution of soil microorganism presence and relative abundance?. Ecography e07086. https://doi.org/10.1111/ecog.07086
Citation: https://doi.org/10.5194/essd-2024-79-AC3
Data sets
CHclim25 - a spatially and temporally very high resolution climatic dataset for Switzerland Olivier Broennimann https://zenodo.org/communities/chclim25
Model code and software
CHclim25 - R scripts Olivier Broennimann https://doi.org/10.5281/zenodo.7898914
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 765 | 157 | 47 | 969 | 39 | 38 |
- HTML: 765
- PDF: 157
- XML: 47
- Total: 969
- BibTeX: 39
- EndNote: 38
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
