Influence of Model Version, Resolution and Driving Data on High Resolution Regional Climate Simulations with CLM C. Meißner, G. Schädler, C. Kottmeier Universität / Forschungszentrum Karlsruhe 7.3.27 LM-User-Seminar 27
Motivation Performance of high resolution climate simulations on regional scales Areas of x km Quantification of the influence of resolution and model version on model results 14km, 7km Version 3.21, Version 3.22 (new cloud microphysical parameterization) Quantification of the influence of lateral and lower boundary conditions on the simulation results Investigation of the influence of lateral boundary conditions by using different driving data Investigation of the influence of the lower boundary condition by changing the initial soil water content 7.3.27 LM-User-Seminar 27
Influence of resolution and model version m.5.5.5 47.5 47 6 7 8 9 1 11 24 22 2 18 16 14 12 1 8 6 4 2 Simulations for the year 21 with GME Analysis Data as driving data every 6 hours 4 simulations with different setups: 7 km with Version3.21 7 km with Version3.22 14 km with Version3.21 14 km with Version3.22 7.3.27 LM-User-Seminar 27
.5.5 47.5 6.5 7 7.5 8 8.5 9 9.5 1 1.5.5.5 Annual total precipitation amount 14km V3.21 V3.21 14km V3.22 V3.22 47.5 6.5 7 7.5 8 8.5 9 9.5 1 1.5 3 3 28 26 24 22 2 18 16 14 12 1 8 6 4 2 3 3 28 26 24 22 2 18 16 14 12 1 8 6 4 2.5.5 16.5 14 12 1 8 6 4 47.5 2 47.5 6.5 7 7.5 8 8.5 9 9.5 1 1.5 6.5 7 7.5 8 8.5 9 9.5 1 1.5 7km V3.21 V3.21 7km V3.22 7km 7km V3.22 - V3.21 - V3.21.5.5 47.5 6.5 7 7.5 8 8.5 9 9.5 1 1.5 Too small precipitation amount with 14 km grid size LM Version 3.22 with 7km grid size gives best results 3 3 28 26 24 22 2 18 3 3 28 26 24 22 2 18 16 14 12 1 8 6 4 2.5.5.5 measurements 47.5 6.5 7 7.5 8 8.5 9 9.5 1 1.5 3 3 28 26 24 22 2 18 16 14 12 1 8 6 4 2 1 4 3 2 1 1 - -1-1 -2-3 -4 - -1 7.3.27 LM-User-Seminar 27
Convective precipitation convective precipitation 14km 3.21 3 28 27 25 24 22 21 19 18 16 1 13 12 1 9 7 6 4 3 1 convective precipitation 14km 3.22 3 28 27 25 24 22 21 19 18 16 1 13 12 1 9 7 6 4 3 1 7 8 9 1 convective precipitation 7km 3.21 3 28 27 25 24 22 21 19 18 16 1 13 12 1 9 7 6 4 3 1 7 8 9 1 convective precipitation 7km 3.22 3 28 27 25 24 22 21 19 18 16 1 13 12 1 9 7 6 4 3 1 7 8 9 1 7 8 9 1 Amount of convective precipitation smaller in simulations with 14 km grid size Less convective precipitation with Version 3.22 7.3.27 LM-User-Seminar 27
Grid scale precipitation gridscale precipitation 14km 3.21 3 28 27 25 24 22 21 19 18 16 1 13 12 1 9 7 6 4 3 1 gridscale precipitation 14km 3.22 3 28 27 25 24 22 21 19 18 16 1 13 12 1 9 7 6 4 3 1 7 8 9 1 gridscale precipitation 7km 3.21 3 28 27 25 24 22 21 19 18 16 1 13 12 1 9 7 6 4 3 1 7 8 9 1 gridscale precipitation 7km 3.22 3 28 27 25 24 22 21 19 18 16 1 13 12 1 9 7 6 4 3 1 7 8 9 1 7 8 9 1 Enhancement of grid scale precipitation in Version 3.22 Larger amount of grid scale precipitation with 7km grid size 7.3.27 LM-User-Seminar 27
Annual mean of 2m-temperature 2m temperature 14km 3.21 2m temperature 14km 3.22 C C 1 1 8 6 8 6 4 4 2 2 7 8 9 1 7 8 9 1 2m temperature 7km 3.21 2m temperature 7km 3.22 C C 1 1 8 6 8 6 4 4 2 2 7 8 9 1 7 8 9 1 Simulations with different grid size and model version are very similar 7.3.27 LM-User-Seminar 27
Influence of resolution and model version Improvement of precipitation forecast using 7 km grid size resolution instead of 14 km New cloud microphysical parameterisation improves precipitation forecast Only small differences in temperature simulation between the different model versions 7.3.27 LM-User-Seminar 27
Quantification of the influence of lateral and lower boundary conditions on the simulation results Normally driving data have coarser resolution than GME Analysis => nesting necessary to get reasonable results for small model area Simulation of big model domain with.44 grid size resolution Uncertainty how strong the influence of boundary data can bee => Simulations with: Variation of driving data: ERA4 Reanalysis data NCEP Reanalysis data Variation of soil initialisation: Taken over from driving data (ERA4, NCEP) Climatological values provided by ETH Zürich 7.3.27 LM-User-Seminar 27
52 46 4 6 8 1 12 Forschungszentrum Karlsruhe Variation of driving data Difference of the annual mean/sum between simulation driven by ERA4 and NCEP averaged over the period 199-1999 total precipitation 1 4 3 2 1 1 - -1-1 -2-3 -4 - -1 52 46 2m - temperature 4 6 8 1 12 Differences in the small model domain up to.4 K and 1 Simulation driven by ERA4 is colder and produces less rain K 2.5 1.5 1.75.5.375.25.125 -.125 -.25 -.375 -.5 -.75-1 -1.5-2.5 7.3.27 LM-User-Seminar 27
Variation of driving data Assessment of the influence of driving data for the small model domain by calculating the root mean square error for the simulation with ERA4 and NCEP as driving data RMSE ERA4/NCEP Version 3.21 16 14 12 1 8 6 1.8.6.4 K 4 2 precipitation temperature 199 1991 1992 1993 1994 1995 1996 1997 1998 1999 year.2 Differences between ERA4 and NCEP simulation persist over the whole decade => nesting the small model domain in the different driving data will give different simulation results => simulations soon will be done with model version 3.22 to see if there is an amplification of differences due to downscaling 7.3.27 LM-User-Seminar 27
Variation of soil initialisation Known that ERA4 data does not provide best soil initialisation Possible to initialize with climatological mean soil data obtained by running the CLM for 4 years? Climatological mean was provided by Daniel Lüthi, ETH Zürich 2 Simulations: Continuous simulation from suer 1988 to 21 with ERA4 as driving data and initialization data for the soil (ERA4) Continuous simulation from suer 1988 to 21 with ERA4 as driving data and climatological mean from 4 year run as initialization data for the soil (ERA4KLIM) 7.3.27 LM-User-Seminar 27
Differences at the initial day (ERA4 ERA4KLIM).5 m.16 m 2.86m Soil temperature layer 1 Soil temperature layer 4 Soil temperature layer 8 K 1 8 Soil temperature 4-1 1 2 Soil water content layer 1 4-1 1 2 Soil water content layer 4 4-1 1 2 Soil water content layer 8 6 4 2-2 -4-6 -8-1 vol.-% 1 8 Soil moisture 4 4 4 6 4 2-2 -4-6 -8-1 -1 1 2-1 1 2-1 1 2 Strong differences between soil temperatures (up to 1K) and soil water content (up to 1 vol.-%) 7.3.27 LM-User-Seminar 27
RMSE (ERA - ERAKLIM) from 1988-21 Soil water content RMSE ERA4 / ERA4KLIM Soil water content Soil temperature RMSE ERA4 / ERA4KLIM Soil temperature RMSE (vol.-%) 6 5 4 3 2 1 W_.5 W_.16 W_2.86 RMSE (K) 2 1.5 1.5 2.5 6 T_.5 T_.16 T_2.86 87 89 91 93 95 97 99 1 3 year 87 89 91 93 95 97 99 1 3 year Strong differences in soil temperature and water content decrease during the first three years Annual cycle in difference visible over the whole decade RMSE does not vanish => simulations do not fit together 7.3.27 LM-User-Seminar 27
Difference of annual precipitation (ERA4 ERA4KLIM) 1989 21 4-1 1 2 4 Difference of annual mean 2m-temperature (ERA4 ERA4KLIM) 1989-1 1 2 4 1 Differences in annual precipitation and mean 2m-temperature are higher during the first years and do not vanish during the decade => Temperature differences between both simulations to small to decide if one initialisation gives better results than the other 7.3.27 LM-User-Seminar 27 4 2 1-1 -2-4 -1-1 1 2 4.3.2.1.1 -.1 -.1 -.2 -.3 21-1 1 2 K 1 4 2 1-1 -2-4 -1.3.2.1.1 -.1 -.1 -.2 -.3 1 8 6 4 2 RMSE ERA4/ERA4KLIM precipitation evapotranspiration 2m-temperature 1988 199 1992 1994 1996 1998 2 22 year.5.4.3.2.1 K
Conclusions 7km grid size gives better precipitation results than 14 km grid size in orographically structured region Precipitation amount is more realistic with new cloud microphysical parameterisation Strong influence of driving data on simulation results Soil initialization has a strong effect on precipitation simulation which persists over the decade Ensemble with different physical parameterisations and initial states should be generated for climate decades to quantify the uncertainty in model results One part of my PhD thesis in near future Ideas for meaningful simulations are welcome 7.3.27 LM-User-Seminar 27
7.3.27 LM-User-Seminar 27
7.3.27 LM-User-Seminar 27
Influence of resolution and model version on the simulations 7.3.27 LM-User-Seminar 27
Comparison of annual area mean of precipitation Model version V3.21_7km V3.22_7km V3.21_14km V3.22_14km Total precipitation Convective precipitation Grid scale precipitation 69.8 72.1 43.2 52.9 62.4 61.1 38.8 44.4 7.3 11. 4.4 8.5 Enhancement of area mean in convective and grid scale precipitation from 14 to 7 km Enhancement of area mean from version 3.21 to version 3.22 7.3.27 LM-User-Seminar 27
Total precipitation 2m Temperature Forschungszentrum Karlsruhe Difference model - measurements 4-1 1 2 4 ERA -1 1 2 K 1 4 3 2 1 1 - -1-1 -2-3 -4 - -1 1 4 2 1-1 -2-4 -1 4 4-1 1 2 NCEP -1 1 2 K 1 4 3 2 1 1 - -1-1 -2-3 -4 - -1 1 4 2 1-1 -2-4 -1 Precipitation strongly overestimated by the simulations in Central Europe Good agreement of temperature with measurements in both simulations in Central Europe 7.3.27 LM-User-Seminar 27
.5.5 3.21 - measurements 3.22 - measurements 47.5 6.5 7 7.5 8 8.5 9 9.5 1 1.5 1 4 3 2 1 1 - -1-1 -2-3 -4 - -1.5.5 47.5 6.5 7 7.5 8 8.5 9 9.5 1 1.5 1 4 3 2 1 1 - -1-1 -2-3 -4 - -13 7km V3.22 - V3.21.5.5 47.5 6.5 7 7.5 8 8.5 9 9.5 1 1.5 1 4 3 2 1 1 - -1-1 -2-3 -4 - -1 7.3.27 LM-User-Seminar 27