EGU2020-18274
https://doi.org/10.5194/egusphere-egu2020-18274
EGU General Assembly 2020
© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

Detection of precipitation and snow cover trends in the the European Alps over the last century using model and observational data

Julien Beaumet1, Martin Menegoz1, Hubert Gallée1, Vincent Vionnet2, Xavier Fettweis3, Samuel Morin4, Juliette Blanchet1, Nicolas Jourdain1, Bruno Wilhelm1, and Sandrine Anquetin1
Julien Beaumet et al.
  • 1CNRS, Université Grenoble Alpes, Institut de Géosciences de l'Environnement (IGE), 38000 Grenoble, France (julien.beaumet@univ-grenoble-alpes.fr)
  • 2Environmental Numerical Research Prediction, Environment and Climate Change Canada, Dorval, QC, Canada (vincent.vionnet@canada.ca)
  • 3F.R.S.-FNRS, Laboratory of Climatology, Department of Geography, University of Liège, 4000 Liège, Belgium (xavier.fettweis@uliege.be)
  • 4Univ. Grenoble Alpes, Université de Toulouse, Météo-France, CNRS, CNRM, Centre d'Études de la Neige, 38000 Grenoble, France (samuel.morin@meteo.fr)

The European Alps are particularly sensitive to climate change. Compared to temperature, changes in precipitation are more challenging to detect and attribute to ongoing anthropic climate change mainly as a result of large inter-annual variability, lack of reliable measurements at high elevations and opposite signals depending on the season or the elevation considered. However, changes in precipitation and snow cover have significant socio-environmental impact mostly trough water resource availability. These changes are investigated within the framework of the Trajectories initiative (). The variability and changes in precipitation and snow cover in the European Alps has been simulated with the MAR regional climate model at a 7 km horizontal resolution driven by ERA20C (1902-2010) and ERA5 (1979-2018) reanalyses.

For precipitation, MAR outputs were compared with EURO-4M, SAFRAN, SPAZM and E-OBS reanalyses as well as in-situ observations. The model was shown to reproduce correctly seasonal and inter-annual variability. The spatial biases of the model have the same order of magnitude as the differences between the three observational data sets. Model experiment has been used to detect precipitation changes over the last century. An increase in winter precipitation is simulated over the North-western part of the Alps at high altitudes (>1500m). Significant decreases in summer precipitation were found in many low elevation areas, especially the Po Plain while no significant trends where found at high elevations. Because of large internal variability, precipitation changes are significant (pvalue<0.05) only when considering their evolution over long period, typically 60-100 years in both model and observations.

Snow depth and water equivalent (SWE) in the French Alps simulated with MAR have been compared to the SAFRAN-Crocus reanalyses and to in-situ observations. MAR was found to simulate a realistic distribution of SWE as function of the elevation in the French Alpine massifs, although it underestimates SWE at low elevations in the Pre-Alps. Snow cover over the whole European Alps is evaluated using MODIS satellite data. Finally, trends in snow cover and snow depth are highlighted as well as their relationships with the precipitation and temperature changes over the last century.

How to cite: Beaumet, J., Menegoz, M., Gallée, H., Vionnet, V., Fettweis, X., Morin, S., Blanchet, J., Jourdain, N., Wilhelm, B., and Anquetin, S.: Detection of precipitation and snow cover trends in the the European Alps over the last century using model and observational data, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-18274, https://doi.org/10.5194/egusphere-egu2020-18274, 2020

Comments on the presentation

AC: Author Comment | CC: Community Comment | Report abuse

Presentation version 1 – uploaded on 01 May 2020
  • CC1: Comment on EGU2020-18274, Matthieu Lafaysse, 04 May 2020

    Thanks Julien for your results and great presentation.

    Your are talking about « spurious trends » in SAFRAN. I acknowledge they may seem lower than expected, I can't really explain. However, I would just say that there are different from MAR-ERA20C ones. Indeed, who knows the truth ? All data have their sources of errors, even ERA reanalyses, even observations. I also notice that S. Scherrer and S. Kotlarski in this session also obtain lower trends in observation-based products than in reanalyses products, it is interesting because their observation dataset was homogeneized (not the case for SAFRAN).

    Anyway, both the guess and the observed data assimilated in SAFRAN are very heterogeneous over time and it is demonstrated that it has a significant impact on local trends as published by Vidal et al., 2010 . To minimize this effect, it could a bit be more robust to do these comparisons at a larger scale (the whole Alps) rather than at this very local scale.

    On another topic, how do you demonstrate that the snow albedo feedback is responsible for the elevation-dependent warming ? Is the cause-consequence relationship really obvious from your scatter plot ?

    Thank you

    Matthieu

    • AC1: Reply to CC1, Julien Beaumet, 04 May 2020

      Hi Matthieu,

      Thanks for remarks and interest. I reply here below :

      Your are talking about « spurious trends » in SAFRAN. I acknowledge they may seem lower than expected, I can't really explain. However, I would just say that there are different from MAR-ERA20C ones. Indeed, who knows the truth ? All data have their sources of errors, even ERA reanalyses, even observations. I also notice that S. Scherrer and S. Kotlarski in this session also obtain lower trends in observation-based products than in reanalyses products, it is interesting because their observation dataset was homogeneized (not the case for SAFRAN).

      First, I want to precise I used so far "old" SAFRAN version, and I am planning to move to S2M very soon, possibly this will change the results. I am in contact with S. Morin about this, I can add you in the loop if you will. Then I said "spurious" not because they are different and smaller than in MAR-ERA20C the values changes rapidly in sometimes even from sign between two 300m alt bands... I don't get this when I look at SAFRAN trends in precipitation, snow etc...I looked at S. Scherrer and S. Kotlarski works as well, I posted a comment on their display, I plotted the trend in MAR for this surprising winter cooling period between 1988 and 2012, MAR seems to do a fairly good job although it underestimates a bit the cooling as well but seems a bit better than reanalyses (ERA5 eg).

      Anyway, both the guess and the observed data assimilated in SAFRAN are very heterogeneous over time and it is demonstrated that it has a significant impact on local trends as published by Vidal et al., 2010 . To minimize this effect, it could a bit be more robust to do these comparisons at a larger scale (the whole Alps) rather than at this very local scale.

      All in all, you are right saying than upscalling will gives more robusts results.

      On another topic, how do you demonstrate that the snow albedo feedback is responsible for the elevation-dependent warming ? Is the cause-consequence relationship really obvious from your scatter plot ?

      All I can say, is that R2 for scatter between Tmean trends and SWnet (due to snow related albedo changes) trends are quite convincing at least for spring and summer. Considering for instance spring, where elevations of maximum warming coincides very well with elevations of maximum snow loss and increased SWnet, it is hard to imagine another mechanism at play. Allthough looking at the same thing for tmax and tmin rather than tmean as suggested in the chat could be a bit more convincing. I am also planning at looking other possible mechanism : temperature feedback (LWU,LWD,LWnet), latent/sensible heat fluxes and may-be soil moisture.

      Regards,

      Julien

       

       

  • CC2: Comment on EGU2020-18274, Edoardo Raparelli, 08 May 2020

    Hi Julien!

    Thanks for the questions you made about my presentation (https://meetingorganizer.copernicus.org/EGU2020/EGU2020-19408.html).

    I'm replying here because I uploaded another version of the presentation (typo corrections) and I can't reply to your comment anymore.

    1. How do you do grid point measurement to model comparison for snow height (e.g slide 14) ? Do you take nearest grid point ?
      • To compare WRF simulated values to observations I make a bilinear interpolation of the simulated values on the coordinates of the observations. I do this for snow depth, air temperature, etc. To compare Alpine3D simulated snow depth to the observed snow depth I make a bilinear interpolation on the coordinates of the observations of the WRF simulated values which I use to force Alpin3D. This is done specifying some points of interest in the Alpine3D simulation. Thus Alpine3D generates a gridded output for each cell of domain, but also a more detailed output for the specified points of interest.
    2. What is the elevation difference between model and station elevation and do you take it into account ? 
      • The elevation difference between AWS and model topograpfy is few hundred meters for the snow stations. For sure it has an impact but I'm not correcting the simulated values to try to compensate the elevation difference because it was beyond the scope of the study. However it would be interesting to make a plot of the elevation differences for all the considered AWS.
    3. Also I wanted to know, is there similar observations data you use for the Apennine available for the Italian Alps ?
      • Yes, I used the data of the Italian Civil Protection database. The data come from different regional organizations (e.g. ARPA) and converge in that database. You could ask to these organizations or to the civil protection.

    I read your presentation and it's really interesting. Your ERA5 simulations cover entirely the Central Apennines and I would be really curious to see if you could confirm your findings over the Apennines using local observations. 

    Can you tell me which is the timestep of your 7km ERA5-driven simulation output?

    Thanks,

    Edoardo

    • AC2: Reply to CC2, Julien Beaumet, 11 May 2020

      Hi Eduardo,

      Many thanks for your reply.

      I save mostly the outputs of MAR simulations (including the one driven by ERA5) at daily time steps.

      Regards,

      Julien