the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 26 Sep 2024
Too cold, too saturated? Evaluating climate models at the gateway to the Arctic
Abstract. The Arctic wintertime energy and moisture budget are largely controlled by the advection of warm, moist air masses from lower latitudes, cooling and drying of these air masses inside the Arctic and the export of cold, dry air masses. Climate models have substantial difficulties in representing key processes in these air-mass transformations, including turbulence under stable stratification and mixed-phase cloud processes. Here, we use radiosonde profiles of temperature and moisture and surface radiation observations from Ny Ålesund, Svalbard (1993–2014), to assess the properties of air masses being imported into and exported from the central Arctic in CMIP6 climate models. In the free troposphere, models tend to be cold-biased especially for the coldest temperatures. Most models underestimate the frequency of occurence of supersaturation with respect to ice, and a sensitivity experiment suggests that this can be improved by using 2-moment microphysics, i.e. prognostic rather than prescribed ice number concentrations. Cold and dry biases are stronger in air masses being exported from the Arctic than those entering the Arctic. This suggests that previously reported cold biases and excess energy convergence in the Arctic in CMIP6 models are probably due to errors in local thermodynamic processes.
- Preprint
(8401 KB) - Metadata XML
-
Supplement
(121 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2024-2961', Anonymous Referee #1, 12 Nov 2024
Peer review for manuscript: Too cold, too saturated? Evaluating climate models at the gateway to the Arctic
Climate models are widely used tool in studies about atmospheric physics and dynamics and practically the only method to predict how changes on e.g. greenhouse gases affect the future climate. Therefore, it is important evaluate their behavior against the real observed atmospheric conditions. However, these comparisons are not often straightforward to do as many issues associated e.g. representiveness of observations may cause the problems for direct comparisons. In addition, availability of observation may complicate direct comparison especially in remote areas, where observational network is sparse as in the Arctic. Climate models have often difficulties to present polar climate, probably because parametrizations better optimized for lower latitudes. Thus, the manuscript provides interesting and important information about capabilities of climate models to describe polar climate.
The manuscript present detailed comparison between climate models and radiosonde observation from Ny-Ålesund. Even though models are compared against observation from only a single measurement site in the Arctic the method to divide the comparison to two flow regimes, northerly flow and south-westerly flow, gives additional information about the sources of biases in the models. This division allows to separate the contributions of Arctic processes and mid-latitude advections in the model biases.
The study shows that there are biases in climate models in low-tropospheric temperatures and humidity and discuss possible physical causes behind these biases. The discussion about the causes behind insufficient presentation of atmospheric processes in the models focus on cloud processes and resolution of models. The experiments which are made with high-resolution model with two different cloud schemes take the discussion deeper level than only speculations. The experiments suggested that too simple cloud parametrization probably contribute biases in humidity in the climate models. However, causes behind the main result that the models have cold bias especially in the airmasses which come from the central Arctic has only speculated in the manuscript. The reasons for cold bias in the climate models are probably quite complicated, and it could be a topic for a separate study as proposed in manuscript.
Overall, the manuscript is well written conclusion based on evidence of results. Methods are appropriately described allowing readers to understand how study is done. However, there are some small issues in the manuscript which should be considered before publishing. Therefore, I recommend publishing the manuscript in Atmospheric Chemistry and Physics after minor revision.
My main concern is associated with discussion about longwave radiation in the manuscript. The focus of manuscript is temperatures in troposphere, but mostly only longwave radiation at the surface is discussed. This is understandable because observation is available only at the surface, but the effects of cloud on tropospheric cooling could be discussed more as the focus of manuscript is on tropospheric temperatures. Below there is a couple spots in text where you could discuss more about effects of cloud on tropospheric long wave cooling.
Lines 30 – 32 In addition to that clouds reduce long wave radiative cooling at surface, the increase longwave emissions from atmosphere upward and thus affect long wave atmospheric cooling.
Lines 277 – 281 The effects of high emissivity of clouds in CMIP models on tropospheric temperature could be discussed.
Specific Comments:
Line 44. “In real-world mixed-phase clouds, supercooled liquid water can be concentrated in thin (O 10m) layers near cloud top” Could you give a reference for this?
Line 109 – 112. Does this mean that formation of ice particles from water vapor without any nuclei is not included in the model but freezing of ice particle from water droplet without any ice nuclei is included?
Lines 141 -144 and lines 180 -183. Maybe mention explicitly that these results based on radiosonde observation.
Line 184. Maybe we could say that, overall there is not similar dry bias on average in the modes as there is cold bias in temperatures, or how about in terms of dew point temperature how the biases in humidity are compared with the biases in temperature?
Line 205. Figure 7 -> Figure 6
Figure 8. There are no values on y-axis in Figure 8.
Line 223. The CMIP models … or The CMIP model … ?
Figure 9. Do colors mean occurrence? Maybe because of logarithmic scale, but it seems that there are quite many ice particles in 1-moment scheme when relative humidity is low? Do I interpret the figure correctly?
Line 268. Could you explicitly mention what means upper bound?
Line 270. There is not shown land-sea fraction in Figure 2
Lines 270 - 271. Could you explain how surface type affects the downward longwave radiation?
Figures 7, 9, 12, and 13 The meaning of color scale could be mentioned in the captions of figures. I assume that the different tones of gray means frequency of occurrence.
Figure 8, 10 and 11. ”northward flow” and ”southward flow” could be also mentioned in these figures in the title of the left and right panel.
Citation: https://doi.org/10.5194/egusphere-2024-2961-RC1 -
RC2: 'Reply on RC1', Rune Grand Graversen, 13 Nov 2024
Peer review of “Too cold, too saturated? Evaluating climate models at the gateway to the Arctic” by Pithan et al.
This study evaluates climate model against observations in Ny Ålesund on Svalbard, hereby it contributes to a suite of studies evaluating models in the Arctic where observations are sparse. The results show, for example, that models are generally having cold bias at the site in question. However the ensemble of models investigated is small and in many cases these models are variants of each other. Hence it is still an open question whether the conclusions drawn here also are valid for the full CMIP6 ensemble.
At places, the manuscript is not so clear. I here provide some comments and suggestions that the authors can take into account before consideration of publication.
Major comment:
Only a small subset of the CMIP6 models are used in this study. The authors are taking an interesting perspective by studying northerly and southerly flow in Ny Ålesund, to some extent representing Arctic and midlatitude air masses, respectively. Therefore high frequency (6-hour) data are needed, which, as the authors argue, limits selected models to a rather small subset of CMIP6. In order to give an idea of the representativeness of this subset relative to the full CMIP6, I suggest to start the study with evaluating the full CMIP6 based on monthly mean data (which or available for the full CMIP6) and against monthly means of the station data. This can be done at least for the temperature and humidity profiles. We can hereby learn of the overall biases of CMIP6 and the representativeness of the subset investigated further.Minor comments:
L20: “These processes ..” → “It is challenging to represent ..”
L21: “had”->”have”
L22-25: A reference is needed for the influence of large-scale circulation on the Arctic, e.g. Baggett and Lee, J. Atmosph. Sci., 2015 and Graversen and Burtu, Q. J. Royal Met. Sco., 2016.
L26-28: It says stable stratification “shaped by air-mass transformation”, I would rather say “mainly due to atmospheric heat advection from the south and diabatic cooling of the surface”.
L39: Remove space before full stop.
L41-43: Is this regime a curiosity or is it common in the Arctic? Please provide a reference on this aspect.
L60-61: Are the radiosondes automatically launched? Is the system in operation from 1993 to present? Please indicate.
L70: An evaluation of ERA5 in the Arctic is given here: Graham et al., J. Clim., 2019.
Fig. 1: Why not show topography for ICON as well?
L99: remove “s” on “thermodynamics”.
L122: “those”, do you mean “whom”?
L145-149:”Danmarkshavn” (without “e”). The sentence needs a reference.
L149-151: This sentence comes out of the context here and can be moved to the discussion of Fig. 3.
L153-155: This sentence regarding polar amplification of bias seems quite speculative and does not give much meaning to me.
L155-156: When discussing the results in a pan-Arctic context, one should keep in mind that the results are based on a single station only.
L158-161: How can inversions be seen in Fig. 3? The figure seems to show biases and not absolute values. In addition, this sentence seems inconsistent with the appearance of bias inversion in the described conditions for some models, for instance the GISS model. Also remove “the formation of”.
L201: Add “humidity” after “saturation”, since you later refer to a “value”. Add “variability” after “this” to avoid ambiguity.
L206-208: This seems to be inconsistent with at least three models: GISS, MRI-GCM2.0 and Had, which show considerable subsaturation.
Fig. 7: It is not clear what is shown, is it probability-density functions? Units on colour scale are not explained. The line for 700 hPa is almost invisible. Text and numbers (also in other figures) are too small in font size.
Fig. 8: Difficult to distinguish the different lines. ICON in the left frame can't be seen. Perhaps rather present a design with model mean, ICON, and OBS.
L2018-219: I suppose this sentence concerns OBS, but this is not so clear. The mode below 30 % is extremely small, and perhaps not worth mentioning.
L245: Comma after “scheme”.
L267-268: How is it seen that the AWI model exceeds the “upper bound”.
L285-286: MIROC seems not to be shown in Fig. 14.
Fig. 12, same as for Fig. 7.
Section 3.5: Mention that this section is based on only OBS. Humidity could also be investigated here.
L322-323: This statement doesn't seem to be valid for all models.
L340-342: A reference is lacking for supporting these two sentences.Citation: https://doi.org/10.5194/egusphere-2024-2961-RC2 -
AC1: 'Comment on egusphere-2024-2961', Felix Pithan, 03 Dec 2024
We thank both reviewers for their positive assessment of our work and insightful comments and suggestions, to which we respond in detail below:
RC1: […]
My main concern is associated with discussion about longwave radiation in the manuscript. The focus of manuscript is temperatures in troposphere, but mostly only longwave radiation at the surface is discussed. This is understandable because observation is available only at the surface, but the effects of cloud on tropospheric cooling could be discussed more as the focus of manuscript is on tropospheric temperatures. Below there is a couple spots in text where you could discuss more about effects of cloud on tropospheric long wave cooling.
Lines 30 – 32 In addition to that clouds reduce long wave radiative cooling at surface, the increase longwave emissions from atmosphere upward and thus affect long wave atmospheric cooling.
We mention this in the revised manuscript :
Clouds increase the emissivity of the atmosphere and can thereby increase atmospheric radiative cooling and precipitation formation (Pithan and Jung, 2021; Bonan et al., 2024).
Lines 277 – 281 The effects of high emissivity of clouds in CMIP models on tropospheric temperature could be discussed.
Indeed, thanks for pointing this out. We have added the following sentence:
For a given tropospheric temperatures, more emissive clouds lead to warmer surface temperatures in the Arctic, but they also lead to more efficient radiative cooling of tropospheric air, which could contribute to the strong tropospheric cold bias in GISS-E2-1-G.
Specific Comments:
Line 44. “In real-world mixed-phase clouds, supercooled liquid water can be concentrated in thin (O 10m) layers near cloud top” Could you give a reference for this?
We have added a reference to the MPACE experiment (Verlinde et al 2007). Thanks for the suggestion!
Line 109 – 112. Does this mean that formation of ice particles from water vapor without any nuclei is not included in the model but freezing of ice particle from water droplet without any ice nuclei is included?
Yes, the reviewer is right that cloud ice specific mass in the one-moment scheme is not affected by formation of ice particles from water vapor without any nuclei but that freezing of ice particle from water droplet without any ice nuclei is included. We clarify in the manuscript:
The single-moment scheme calculates a diagnostic number of ice particles which depends only on temperature. It does not allow for homogeneous nucleation of ice particles from water vapor without any nuclei but takes into account tendencies of cloud ice specific mass due to homogeneous freezing of cloud droplets and a simple heterogeneous nucleation rate.
Lines 141 -144 and lines 180 -183. Maybe mention explicitly that these results based on radiosonde observation.
The text refers to Figures 2 and 4, which explicitly mention the radiosondes in their captions. We have not changed the manuscript text.
Line 184. Maybe we could say that, overall there is not similar dry bias on average in the modes as there is cold bias in temperatures, or how about in terms of dew point temperature how the biases in humidity are compared with the biases in temperature?
Thanks for the suggestion, we have expanded the text as follows:
Biases in specific humidity (Figure 5) partly parallel those in temperature, but the models do not display a consistent dry bias as could have been expected based in the temperature biases discussed above.
Line 205. Figure 7 -> Figure 6
We have moved the reference to Figure 7 to the following sentence, thanks for pointing out the potential confusion!
Figure 8. There are no values on y-axis in Figure 8.
We have omitted the absolute values of the pdf as they carry to physical meaning beyond the relative scaling that can be inferred from the graph.
Line 223. The CMIP models … or The CMIP model … ?
Corrected, thanks!
Figure 9. Do colors mean occurrence? Maybe because of logarithmic scale, but it seems that there are quite many ice particles in 1-moment scheme when relative humidity is low? Do I interpret the figure correctly?
We have added to the caption that this Figure is showing the bivariate pdf of ice number concentration and RH_ice, thanks for pointing out the ommission! Yes, the frequent high estimate of N_ice at low RH_ice are another consequence of the 1-moment scheme. We do not discuss this in the manuscript, as we focus on the supersaturation regime here.
Line 268. Could you explicitly mention what means upper bound?
Thanks for poining out the lack of clarity! We have revised the statement to say:
AWI-ESM1-1 appears to show values of downward longwave radiation that exceed the near-linear relationship formed by clouds radiating as blackbodies at a given temperature.
Line 270. There is not shown land-sea fraction in Figure 2
Indeed, we have removed the reference to Figure 2. Thanks.
Lines 270 - 271. Could you explain how surface type affects the downward longwave radiation?
We attribute this feature to shallow cold-air outbreaks close to the sea-ice edge, in which warm low-level clouds enhance downward longwave radiation near the surface but not yet 925 hpa temperature. We do not discuss this in detail in the manuscript to avoid detracting from our main point, which is the model evaluation here.
Figures 7, 9, 12, and 13 The meaning of color scale could be mentioned in the captions of figures. I assume that the different tones of gray means frequency of occurrence.
We added to all captions that these show bivariate pdfs. Thanks for pointing out that this was missing in the previous version1
Figure 8, 10 and 11. ”northward flow” and ”southward flow” could be also mentioned in these figures in the title of the left and right panel.
Indeed, we have added this. Thanks!
Rc2 / Rune Grand Graversen
Peer review of “Too cold, too saturated? Evaluating climate models at the gateway to the Arctic” by Pithan et al.
This study evaluates climate model against observations in Ny Ålesund on Svalbard, hereby it contributes to a suite of studies evaluating models in the Arctic where observations are sparse. The results show, for example, that models are generally having cold bias at the site in question. However the ensemble of models investigated is small and in many cases these models are variants of each other. Hence it is still an open question whether the conclusions drawn here also are valid for the full CMIP6 ensemble.
At places, the manuscript is not so clear. I here provide some comments and suggestions that the authors can take into account before consideration of publication.
Major comment:
Only a small subset of the CMIP6 models are used in this study. The authors are taking an interesting perspective by studying northerly and southerly flow in Ny Ålesund, to some extent representing Arctic and midlatitude air masses, respectively. Therefore high frequency (6-hour) data are needed, which, as the authors argue, limits selected models to a rather small subset of CMIP6. In order to give an idea of the representativeness of this subset relative to the full CMIP6, I suggest to start the study with evaluating the full CMIP6 based on monthly mean data (which or available for the full CMIP6) and against monthly means of the station data. This can be done at least for the temperature and humidity profiles. We can hereby learn of the overall biases of CMIP6 and the representativeness of the subset investigated further.We have added the suggested comparison of observed and modelled monthly mean profiles (Figure 3 of the revised manuscript), which is indeed a great way of placing our results into the context of the full CMIP6 ensemble. Thanks for bringing this up!
Minor comments:
L20: “These processes ..” → “It is challenging to represent ..”Modified as suggested.
L21: “had”->”have”Modified as suggested.
L22-25: A reference is needed for the influence of large-scale circulation on the Arctic, e.g. Baggett and Lee, J. Atmosph. Sci., 2015 and Graversen and Burtu, Q. J. Royal Met. Sco., 2016.
References addded.
L26-28: It says stable stratification “shaped by air-mass transformation”, I would rather say “mainly due to atmospheric heat advection from the south and diabatic cooling of the surface”.modified as suggested.
L39: Remove space before full stop.Corrected, thanks.
L41-43: Is this regime a curiosity or is it common in the Arctic? Please provide a reference on this aspect.
We have moved the references to the end of the paragraph to make clear that they cover the entire process chain that is discussed here.
L60-61: Are the radiosondes automatically launched? Is the system in operation from 1993 to present? Please indicate.Radiosondes are launched manually, and we have added this to the revised manuscript.
L70: An evaluation of ERA5 in the Arctic is given here: Graham et al., J. Clim., 2019.Reference added.
Fig. 1: Why not show topography for ICON as well?
We chose not to include ICON topography, as we only use ICON for a dedicated sensitivity experiment, and use data from abroader region that is less affected by the Svalbard topography.
L99: remove “s” on “thermodynamics”.corrected, thanks!
L122: “those”, do you mean “whom”?We have shortened and simplified the sentence to:
This is a short period for studying climatological effects…
L145-149:”Danmarkshavn” (without “e”). The sentence needs a reference.Typo corrected, thanks! We have added a reference to the IGRA dataset.
L149-151: This sentence comes out of the context here and can be moved to the discussion of Fig. 3.
We have rephrased this sentence to improve readability and better connect it to the preceding paragraph, thanks for pointing out that this was unclear!
We therefore do not interpret the deviation of modelled and observed temperatures at lower levels (marked by gray shading in profile plots) as a model bias. These observations might simply not be representative of the scales the models attempt to describe.
L153-155: This sentence regarding polar amplification of bias seems quite speculative and does not give much meaning to me.
We have removed the sentence, which indeed wasn’t very helpful here. Thanks.
L155-156: When discussing the results in a pan-Arctic context, one should keep in mind that the results are based on a single station only.We have added “over Svalbard” to emphasize the local nature of our observations here.
L158-161: How can inversions be seen in Fig. 3? The figure seems to show biases and not absolute values. In addition, this sentence seems inconsistent with the appearance of bias inversion in the described conditions for some models, for instance the GISS model. Also remove “the formation of”.The absence of a temperature inversion ins observations can indeed only be seen in Figure 2 for the observations, and is not directly shown for the model. This paragraph refers to AWI-ESM-1-1-LR, the behaviour of the GISS model is indeed different. We have modified the sentence to better reflect this:The warmest air masses coming from the south do not show any signs of surface decoupling or temperature inversions in either observations (Figure 2) or the model (not shown), such that the specific local conditions of the Fjord that can cause additional mixing do not lead to a mismatch between observed and AWI-ESM-1-1-LR profiles under these conditions
L201: Add “humidity” after “saturation”, since you later refer to a “value”. Add “variability” after “this” to avoid ambiguity.
Modified as suggested.
L206-208: This seems to be inconsistent with at least three models: GISS, MRI-GCM2.0 and Had, which show considerable subsaturation.
Indeed, thanks for pointing this out. We have split the discussion of supersaturation and subsaturation to better reflect the different behaviour of models:
Most models are close to saturation with respect to ice for cold temperatures, and somewhat below saturation with respect to ice at warmer temperatures (Figure 8). In particular at cold temperatures, models tend to lack supersaturated conditions frequently seen in observations. One exception that we will discuss further below is the MRI-ESM2-0 model which more frequently reaches humidities close to saturation with respect to water, and thus supersaturation with respect to ice, at cold temperatures below -15 ◦C. Several models also lack strong subsaturation with respect to ice that is seen in observations.
Fig. 7: It is not clear what is shown, is it probability-density functions? Units on colour scale are not explained. The line for 700 hPa is almost invisible. Text and numbers (also in other figures) are too small in font size.
Yes, this is a pdf, and we have added that to the caption. We have increased font sizes on the Figures. Thanks!
Fig. 8: Difficult to distinguish the different lines. ICON in the left frame can't be seen. Perhaps rather present a design with model mean, ICON, and OBS.We have rescaled the Figure to improve legibility. ICON data are indeed only shown in the right panel, and we have added an explanation to the caption:
ICON data (right) are from a much shorter run than CMIP models and cover the central Arctic ocean (70◦ N to 90◦ N) to obtain useful statistics. They are plotted alongside observations and CMIP model output for air masses originating from the North as they represent air mass properties in that source region.
L2018-219: I suppose this sentence concerns OBS, but this is not so clear. The mode below 30 % is extremely small, and perhaps not worth mentioning.We explicitly state that this refers to observed data and omit the possible intermediate mode for clarity.
In both southwesterly and northerly flows, observed relative humidity with respect to ice is most frequently below 30 or close to 100 %.
L245: Comma after “scheme”.Comma inserted, thanks!
L267-268: How is it seen that the AWI model exceeds the “upper bound”.
We have rephrased this – thanks for pointing out the lack of clarity!
AWI-ESM1-1 appears to show values of downward longwave radiation that exceed the near-linear relationship formed by clouds radiating as blackbodies at a given temperature.
L285-286: MIROC seems not to be shown in Fig. 14.
Indeed, we have removed the erroneous reference from the text and added a sentence to the Figure caption to explain that models with no data at 1000 hPa have been omitted. Thanks!
Fig. 12, same as for Fig. 7.
Corrected.
Section 3.5: Mention that this section is based on only OBS. Humidity could also be investigated here.
We have changed the section title to ‘observed trends’.
L322-323: This statement doesn't seem to be valid for all models.
Corrected as follows:
Relative humidity in models is close to saturation over ice more frequently in models than in observations. Most models lack supersaturation with respect to ice and several models also lack subsaturation at relative humidities with respect to ice around 340 or below 30 %.
L340-342: A reference is lacking for supporting these two sentences.Reference to Dahlke and Maturilli (2017) added to the revised manuscript.
Citation: https://doi.org/10.5194/egusphere-2024-2961-AC1
Data sets
Ny Alesund Radiosonde record Marion Maturilli https://doi.pangaea.de/10.1594/PANGAEA.845373
Model code and software
Code for microphysics sensitivity experiment Ann Kristin Naumann https://doi.org/10.17617/3.OD9NTK
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 164 | 38 | 34 | 236 | 13 | 6 | 5 |
- HTML: 164
- PDF: 38
- XML: 34
- Total: 236
- Supplement: 13
- BibTeX: 6
- EndNote: 5
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1