Added Value Generated by Regional Climate Models H. von Storch, F. Feser Institute of Coastal Research, Helmholtz Zentrum Geesthacht, Germany 29 May 1 - June th Congress of the Canadian Meteorological and Oceanographic Society (CMOS), Montreal Feser, F., B. Rockel, H. von Storch, J. Winterfeldt, and M. Zahn, 2011: Regional climate models add value. Bull. Amer. Meteor. Soc. 92: 1181–1192

Climate = statistics of weather The genesis of climate C s = f(C l, Φ s ) with C l = larger scale climate C s = smaller scale climate Φ s = physiographic detail at smaller scale von Storch, H., 1999  “downscaling”

Model variance as a function of spatial scales. The rectangles show well and insufficiently resolved spatial scales of the global and regional model.

4 Pattern correlation coefficients for V at 500 hPa between the global reanalyses and the RCM with standard forcing via the lateral boundaries and the RCM with spectral nudging global regional Nudging of the large scales

5 NCEP - RCM standard NCEP - RCM nudge March 12, 1993, 0:00 Berliner Wetterkarte RCM standardNCEPRCM nudge

Beate Müller, pers. comm. DWD analysis ) )

Added value in reconstructions using spectral nudging Improved representation of  variability at medium scales.  effect of physiographic detail (coasts)  of sub-synoptic phenomena (polar lows, medicanes, typhoons)  forcing fields for impact models (ocean waves, storm surges)

Improved presentation of variability at medium scales NCEP-driven multidecadal simulation with RCM REMO Employing spectral nudging (wind above 850 hPa, for scales > 800 km)  Usage of German Weather Service (DWD) regional analysis for a few years as reference to determine skill  Considering ratios  2 DWD /  2 NCEP and  2 DWD /  2 RCM  Determining mean spatial correlation patterns between DWD, NCEP and RCM representations, for different spatial scales.

9 Ratio of standard deviations of 2m temperature DWD-analysis / NCEP reanalysis [%], JJA 1992 – 1999, at the regional scale. Added Value Ratio of standard deviations of 2m temperature DWD-analysis / regional simulation [%], JJA 1992 – 1999, at the regional scale.

10 Positive values show added value of the regional model. 95% significant deviations are marked by a *. PCC DWD and NCEP PCC improvement/ deterioration REMO Nudge PCC improvement/ deterioration REMO Standard Pattern correlation coefficients [PCC, %] Feser, MWR 2006

Improved presentation of effect of physiographic detail NCEP-driven multidecadal simulation with RCM REMO Employing spectral nudging (wind above 850 hPa, for scales > 800 km)  Usage of Quikscat-windfields (Q) over sea.  Determining Brier Skill score B = 1 – (RCM-Q) 2 / (NCEP-Q) 2 for all marine grid boxes

12 Brier skill score using QuikSCAT level 2B12 as the “truth”, global reanalysis (NCEP reanalysis) as the reference forecast, and a regional model (SN-REMO) as forecast, Winterfeldt et al. (2010). Added Value Open Ocean: No value added by dynamical downscaling Coastal region: Added Value in complex coastal areas

Wind Speed 1998: Distribution a) Open Ocean: buoy RARH b) Coast: Light Ship Channel Percentile-percentile plots (qq-plots) of wind speed: The 99 dots represent the wind speed percentiles in steps of 1 percent. Winterfeldt and Weisse, MWR 2009

Improved representation of sub-synoptic phenomena NCEP-driven multidecadal simulation with RCM CLM Employing spectral nudging (wind above 850 hPa, for scales > 800 km) Simulation of sub-synoptic phenomena  Polar lows in the Northern North Atlantic  Polar lows in the North Pacific  Medicanes in the Mediterranean Sea  Typhoons in the West Pacific

Density of North Atlantic polar low genesis Genesis in RCM simulation constrained by NCEP reanalsysis Bracegirdle, T. J. and S. L. Gray, 2008 Zahn, M., and H. von Storch, 2008

Annual frequency of past North Atlantic polar lows Zahn and von Storch, 2008 Set-up of multi-decadal simulation NCEP/NCAR reanalysis 1/ CLM Initialised: finishing: spectral nudging of scales > 700 km σ=13

Projected changes in polar low frequency and vertical atmospheric stability A2 C20 A1B B1 Zahn and von Storch, 2010 Differences of the area and time-averaged ice- free SST and T 500-hPa over the maritime northern North Atlantic as proxy for frequency of favorable polar low conditions (CMIP3/IPCC AR4)

North Pacific Polar Low on 7 March 1977 (Chen et al., 2011) NOAA-5 infrared satellite image at 09:58UTC 7th March 1977 Cavicchia and von Storch, 2011 Medicane of January 16, 1995 ECMWF analysis CLM-COSMO, 10 km grid

19 Brier Skill Score between JMA best track data and NCEP or CCLM Sea Level Pressure 10m Wind Speed > 0 : CLM closer to best track 0 : CLM and NCEP equally close to best track < 0 NCEP closer to best track Feser and von Storch, MWR 2008 b

Improved representation of forcing fields for impact models NCEP-driven multi-decadal simulation with RCM REMO Employing spectral nudging (wind above 850 hPa, for scales > 800 km) simulation  Wind and air pressure used to drive models of sea level and circulation of marginal seas (not shown) for describing currents and sea level (North Sea “CoastDat”; not shown)  Wind used to drive models of the statistics of surface waves (ocean waves) in coastal seas (North Sea).

Red: buoy, yellow: radar, blue: wave model run with REMO winds wave direction significant wave height [days] Gerd Gayer, pers. comm., 2001

Annual mean winter high waters Cuxhaven red – reconstruction, black – observations Interannual variability of mean water levels (Weisse and Plüß 2006)

Summary: Added value …  … in medium scales. Medium scales are determined by both the large scale dynamics and the regional physiographic details (C s = f(C l,Φ s ))  More added value with large-scale constraint (spectral nudging)  Little improvement over driving large-scale fields for SLP, which is a large-scale variable, but significant improvement for structured fields like 2 m temp or coastal wind.  Dynamical downscaling works … - Large scales are hardly affected but smaller scales respond to regional physiographic detail.  Present analysis refers to reconstructions, where we can compare the results with a “truth”. For scenario simulations other approaches are needed (e.g., big brother-type)