slide1 n.
Download
Skip this Video
Loading SlideShow in 5 Seconds..
Institut für Küstenforschung PowerPoint Presentation
Download Presentation
Institut für Küstenforschung

Loading in 2 Seconds...

play fullscreen
1 / 40
margo

Institut für Küstenforschung - PowerPoint PPT Presentation

109 Views
Download Presentation
Institut für Küstenforschung
An Image/Link below is provided (as is) to download presentation

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.

- - - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - - -
Presentation Transcript

  1. Institut für Küstenforschung I f K Issues in regional atmospheric modelling: large scale control and divergence in phase space Hans von StorchInst. Coastal ResearchGKSS Research CenterGeesthacht, Germany CGD/CU-PAOS Seminar Boulder, Colorado, 3 Dec. 2001

  2. Institut für Küstenforschung I f K • Validation – the „Big Brother“ experiment of Denis and Laprise • Boundary value problem or information recovery problem? – spectral nudging • The problem of regional noise – indeterminacy

  3. GCM GCM RCM Denis and Laprise: BBE Validation – the „Big Brother“ experiment of Denis and Laprise Recovering regional scale detail with a RCM. Coarse resolution Jump in resolution at the lateral boundary: 1:6 Denis, B., R. Laprise, D. Caya and J. Cote, 2001: Downscaling ability of one-way nested regional climate models: The Big Brother Experiment. Climate Dyn. (in press)

  4. Denis and Laprise: BBE Specific humidity at 700 hPa T = 4.0 days Control “J6”- Experiment

  5. Denis and Laprise: BBE Specific humidity at 700 hPa T = 8.0 days Control “J6”- Experiment

  6. Denis and Laprise: BBE BB J6 C = 88% Temporal standard deviation : precipitation rate Contour intervals : 5 mm day-1

  7. Denis and Laprise: BBE BB C = 90% G = 98% J6 Temporal standard devation of fine-scale features : precipitation rate Contour intervals : 5 mm day-1

  8. Big Brother Experiment … demonstrates that • regional atmospheric model recovers small scale structures as a response to internal dynamics and small scale physiographic details, • jump up to 12:1 is acceptable (at least in the BBE set-up). Thus, RCMs do what they are constructed for.

  9. Institut für Küstenforschung I f K Boundary value problem or information recovery problem? – spectral nudging von Storch, H., H. Langenberg and F. Feser, 2000: A spectral nudging technique for dynamical downscaling purposes. Mon. Wea. Rev. 128: 3664-3673 Feser, F., R. Weisse and H. von Storch, 2001: Multidecadal atmospheric modelling for Europe yields multi-purpose data. EOS 82, 305+310

  10. Institut für Küstenforschung I f K global model „Energy“ Insufficiently resolved Well resolved Spatial scales

  11. Institut für Küstenforschung I f K regional model „Energy“ Insufficiently resolved Well resolved Spatial scales Added value

  12. Institut für Küstenforschung I f K Data drivenmodeling ...

  13. Institut für Küstenforschung I f K Usually, a regional model is forced only in a „sponge zone“ along the lateral boundaries. („standard“) We use „large-scale nudging“ instead, i.e., additionally to the lateral forcing the large-scale (spectrally filtered) analysed state is imposed in the interior as well. d*t = (filtered) large-scale NCEP re-analysis

  14. Institut für Küstenforschung I f K The regional atmospheric model REMO is forced with 6-hourly NCEP re-analyses of global weather.

  15. Institut für Küstenforschung I f K Similarity of zonal wind at 850 hPa between simulations and NCEP re-analyses large scales medium scales standard formulation large-scale nudging

  16. Institut für Küstenforschung I f K

  17. Institut für Küstenforschung I f K Correlation between gridded precip analysis (MAP) and REMO (left) and NCEP estimates (right) (N. Groll, 2001, unpublished)

  18. Institut für Küstenforschung I f K Precip stats at 10 Spanish stations Percentage of wet and dry days in one RCM grid box (~50 x 50 km²),four RCM grid box average (~200 x 200 km²),ECMWF operational analysis grid box (~200 x 200 km²)

  19. Zonal component of the 10m-wind Patterns of the first 2 EOFs (winter 1969-1979) Station data / Model data 5.526.94 1.551.10 List Hallig Hooge Büsum 5.576.75 0.750.29 5.52 3.77 0.01 -0.31 5.446.08 -1.39-0.64 Norderney Cuxhaven 4.494.38 -0.58 -0.43 Bremerhaven 3.893.40 -0.67-0.77 Feser, 2001, (unpublished)

  20. Institut für Küstenforschung I f K

  21. Institut für Küstenforschung I f K

  22. Institut für Küstenforschung I f K • Conclusions • Regional atmospheric modelling is not a boundary value problem but a problem of efficiently combining empirial knowledge and theoretical insight. • Regional atmospheric modelling aims at modelling regional scales while satisfying large-scale constraints. • Spectral nudging is one method to deal with the problem.

  23. Institut für Küstenforschung I f K The problem of regional noise – indeterminacy Weisse, R., H. Heyen and H. von Storch, 2000: Sensitivity of a regional atmospheric model to a sea state dependent roughness and the need of ensemble calculations. Mon. Wea. Rev. 128: 3631-3642

  24. Institut für Küstenforschung I f K The Rinke & Dethloff study on regional modelling of the Arctic atmosphere Ensemble standard deviation 500 hPa height [m²/s²] Rinke, A., and K. Dethloff, 2000: On the sensitivity of a regional Arctic climate model to initial and boundaryconditions. Clim. Res. 14, 101-113.

  25. Institut für Küstenforschung I f K Thus, the development in the interior of the limited domain is only partially controlled by the lateral boundary conditions. Instead, the nonlinear chaotic processes acting on all spatial scales have a marked impact on the development. Small disturbances, be they in the initial conditions, lateral boundary conditions, or in the parameterizations introduce the potential of divergent evolution at any time. The stronger the influence of the large-scale state, the smaller the potential for divergence.

  26. Institut für Küstenforschung I f K Not only in global GCMs but also in regional GCMs variations unrelated to external causes (noise) are formed. The assessment of a paired model experiment, in which the effect of a treatment is studied, needs the discrimination between the effect of the treatment (signal) and noise.

  27. Institut für Küstenforschung I f K Example: The case of the relevance of the sea state on the atmospheric variability Hypothesis: The dynamical state of the ocean waves (specifically the shape of the spectra, or age) affect in a physically significant way the state of the overlying atmosphere. Growing (young) waves suck momentum from the wind field, thereby damping the formation of storms.

  28. Institut für Küstenforschung I f K Experimental design: Regional atmospheric model (HIRLAM) covering the North Atlantic. Control: roughness of sea surface parameterized by the Charnock formula. Anomaly: roughness of sea surface determined from wave spectra simulated interactively with wave model WAM. In each configuration one full year was simulated (conventional setup.)

  29. Institut für Küstenforschung I f K HIRLAM computation domain, covering the North Atlantic storm track, where wind-wave interaction is maximum.

  30. Institut für Küstenforschung I f K 1 year simulation (January – December 1993), SLP Area average of rms difference between control (Charnock) and experiment (interactive WAM model)

  31. Institut für Küstenforschung I f K control (Charnock) experiment (WAM) difference SLP in hPa 15. January  14. January  13. January January episode with large differences

  32. Institut für Küstenforschung I f K Additionally, another 20 months were simulated with HIRLAM. For each configuration, control (Charnock) and anomaly (WAM model coupled), 5 Januaries and 5 Junes were simulated. They differed only with respect to the initial state, which was taken from the year-long simulation one day apart (e.g. 2, 3, 4, 5 and 6 January). Thus for the basic experiment, two ensembles of 6 „control“ and „anomaly“ members each were available to assess the internal variability (noise) and the systematic difference (signal).

  33. Institut für Küstenforschung I f K SLP January Area averaged rms of the six control simulations, relative to their joint spatial average (solid)and of the six anomaly simulations relative to their joint spatial average (dashed). Note that the rms is calculated for each time separately – the noise is not stationary but time dependent.

  34. Institut für Küstenforschung I f K #3 - #1 #6 - #1 #6-#3 13. Jan  14. Jan  15. Jan Differences between members of the „control ensemble“

  35. Institut für Küstenforschung I f K B A Rms of members of the anomaly ensemble (interactive WAM model) compared to control ensemble variations.For both ensembles, the rms is calculated relative to the control average. The blue band is the estimated 95% „confidence“ interval of rms of the control ensemble. 95% of all states consistent with the control should be within the band. A is a situation with an insignificant difference, B a situation with a significant difference.

  36. Institut für Küstenforschung I f K A: Large differences and large noise, thus inconclusive result. Ensemble mean differences in SLP [hPa] Points with significant t-statistics are in blue. 15. Jan, 0 UTC Six anomaly (interactive WAM; solid) and six control simulations (Charnock; dashed) of 500 hPa height [gpm]

  37. Institut für Küstenforschung I f K B: Small differences but statistically significant. Evidence for physically insignificant treatment. Ensemble mean differences in SLP [hPa] Points with significant t-statistics are in blue. 29. Jan, 0 UTC Six anomaly (interactive WAM; solid) and six control simulations (Charnock; dashed) of 500 hPa height [gpm]

  38. Institut für Küstenforschung I f K Effect of spectral nudging to suppress divergence Weisse and Feser, unpublished SLP Standard ensemble Spectral nudging ensemble standard obs wind speed Spectral nudging

  39. Institut für Küstenforschung I f K • Conclusions • Also in regional climate models internal variability is formed; only part of the variability is related to varying boundary forcing. • Numerical experiments with RCMs need to discriminate between noise and signal, like in global GCM experiments. • The noise in RCMs is not stationary so that its statistics can hardly be extracted from extended simulations; instead sufficiently large ensembles are needed.

  40. Institut für Küstenforschung I f K Report of the "Joint WGCM/WGNE ad hoc Panel on Regional Climate Modelling“: Atmospheric regional climate models (RCMs): A multiple purpose tool? Richard Jones (Hadley Centre, England), Ben Kirtman (Center for Ocean-Land Studies - COLA, USA), René Laprise, (Convenor; Université du Québec à Montréal, Canada), Hans von Storch (GKSS Research Centre, Germany), Werner Wergen (Deutscher Wetterdienst - DWD, Germany) • Recommendations • Obviously, all models suffer from various defects. In fact, trivially, numerical models are a reduced image of a considerably more complex reality. In this sense, all models are wrong and can be made more realistic in very many different ways. Therefore the process of improving models should be guided by the needs of the specific applications. • The reduction of errors in the driving GCMs should remain a priority for climate modellers. • The assessment of RCM climate simulations continues to be hampered by the lack of high-resolution observed gridded climate data over many regions of the globe. Regional data re-analysis projects using observations from national archives should be encouraged.