1. The 2nd International Workshop on GPM Ground Validation TAIPEI, Taiwan, 27-30 September 2005 GV for ECMWF's Data Assimilation Research Peter Bauerpeter.bauer@ecmwf.int ECMWF, Reading, UK NWP and data assimilation Validation options Requirements for GV

2. The 2nd International Workshop on GPM Ground Validation TAIPEI, Taiwan, 27-30 September 2005 What will NWP systems focus on in the GPM timeframe?

3. The 2nd International Workshop on GPM Ground Validation TAIPEI, Taiwan, 27-30 September 2005 Example of NWP Prediction Skill Development Improvement of model spatial/temporal resolution due to increased computer power. Improvement of physical parameterizations (diabatic, land/ocean-atmosphere etc.). Increased satellite data usage! 2-day skill improvement Elimination of NH-SH discrepancy

4. The 2nd International Workshop on GPM Ground Validation TAIPEI, Taiwan, 27-30 September 2005 Expected Future Developments in NWP ECMWF now:- 40 km (T511/T159), 60 model levels; - Two analysis suites (6-hour, 12-hour window); - Two 10-day forecasts initialized at 00 and 12 UTC; - 50-member EPS (50 km, T255); - Radiances/products from ~20 different satellite sensors assimilated; - Assimilation of rain-affected radiances operational since 28/06/2005. ECMWF late 2005:- 25 km (T799/T255), 91 model levels; - Two 14-day forecasts initialized at 00 and 12 UTC. ECMWF ~2010:- 15 km (EPS 30 km); - towards longer assimilation window analyses; - towards unified ensemble prediction system (medium-range, monthly, seasonal); - towards coupled data assimilation (land-ocean-atmosphere); - towards environmental monitoring; - towards focus on severe weather forecasting. General: - NWP systems will become much better in (physically) resolving even meso- scale synoptic systems. - NWP systems will become much better in assimilating cloud and rain affected observations (see recent JCSDA workshop on Cloud and Precipitation, http://www.jcsda.noaa.gov/CloudPrecipWkShop/program.html ).

5. The 2nd International Workshop on GPM Ground Validation TAIPEI, Taiwan, 27-30 September 2005 Why do observations in cloud and precipitation have potential?

6. The 2nd International Workshop on GPM Ground Validation TAIPEI, Taiwan, 27-30 September 2005 Forecast sensitivity to Cloud and Rain-affected Observations Current systems produce rather good precipitation forecasts without assimilating any (!) direct precipitation or cloud observation. However: There are indications that key analysis errors occur in areas that are influenced by clouds and precipitation. ANAL @ t 0 + 2 days Model H ANAL @ t 0 FCST @ t 0 + 2 days Sensitivity of FCST error Adjoint Model H* to perturbations FCST error @ t 0 + 2 days in ANAL @ t 0 Optimum perturbations to minimize FCST error KEY ANALYSIS ERROR TRACKING (Rabier et al. 1996, Klinker et al. 1998)

7. The 2nd International Workshop on GPM Ground Validation TAIPEI, Taiwan, 27-30 September 2005 Mean Dec 1999 600 hPaT-perturbations modifiedT-perturbations modified T-perturbationsby high cloud coverby low cloud cover Mean profile of Dec 1999 Mean Dec 1999 high cloud cover Mean Dec 1999 low cloud cover T-perturbations Forecast sensitivity to Cloud and Rain-affected Observations (McNally 2002)

8. The 2nd International Workshop on GPM Ground Validation TAIPEI, Taiwan, 27-30 September 2005 What are the validation options for clouds and precipitation in NWP systems?

9. The 2nd International Workshop on GPM Ground Validation TAIPEI, Taiwan, 27-30 September 2005 Direct vs. Indirect Validation of NWP-System Performance: Small-scale Direct validation of ECMWF cloud-cover analyses with LITE observations Indirect validation of ECMWF cloud/rain profiles with ground-based (ARM) 35-GHz radar observations ECMWF LITE ECMWF ARM Latitude Time [h] Z e [dBZ] Cloud water (solid) and ice (dashed) mixing ratio [g/kg] (Lopez et al. 2005)

10. The 2nd International Workshop on GPM Ground Validation TAIPEI, Taiwan, 27-30 September 2005 Direct validation of ECMWF Indirect validation of ECMWF precipitation forecasts with cloud/rain fields with SSM/I BMRC rain gauge analyses 19.35 GHz (h) observations ECMWF SSM/I ECMWF CTRL ECMWF EXP BMRC mm Direct vs. Indirect Validation of NWP-System Performance: Large-scale

11. The 2nd International Workshop on GPM Ground Validation TAIPEI, Taiwan, 27-30 September 2005 What could be the specific GV requirements?

12. The 2nd International Workshop on GPM Ground Validation TAIPEI, Taiwan, 27-30 September 2005 Conceptual Model of GV for Data Assimilation Purposes Rain assimilation at ECMWF became operational in June 2005: 1) 1D-Variational retrieval of integrated moisture using SSM/I radiances over ocean. 2) 4D-Variational assimilation of integrated moisture in analysis system. (Bauer et al. 2005a, b) Main validation requirements: Cost function minimization in variational assimilation calculates:  J (x) = B -1 (x-x b ) + H T R -1 [H(x) - y o ] where R includes modelling (of H) (also representativeness) and observation errors. H comprises moist physical parameterizations and radiative transfer models but R is required in units K because y o consists of radiances! Initialize single column Validate model (H) forecastPerform validation over long time with 3-d observations of with radiometric and other series and different forecast lengths: T, q, u, v, …. observations: model error small direct estimation of R in units K representativeness error estimate areas of improvement for H model error growth estimate Single column model experiment at GV site:

13. The 2nd International Workshop on GPM Ground Validation TAIPEI, Taiwan, 27-30 September 2005 Potential Ground Validation Requirements NWP systems perform analyses to optimally initialize model forecasts of the entire atmospheric state: Example 1: bad moisture analyses will produce bad cloud and precipitation forecasts. Example 2: good moisture analyses with bad diabatic models will produce bad cloud and precipitation forecasts. Ideally, the entire 3-D meteorological environment should be observed to validate the initial conditions, the model parameterizations, and the observation operators that are employed in data assimilation. Large-scale (satellites, networks): Direct validation with derived products from independent observations (example: PR, GPCP, Cloudsat products, …). Indirect validation with radiances (operational, simple, requires interpretation but prepares for radiance assimilation). Small-scale (GV): Accurate and continuous (to address representativeness issue) observations of derived key meteorological parameters and radar/radiometry (example: conceptual model).