Jordan G. Powers Kevin W. Manning. Mesoscale and Microscale Meteorology Laboratory National Center for Atmospheric Research Boulder, CO

Transcription

1 Jordan G. Powers Kevin W. Manning Mesoscale and Microscale Meteorology Laboratory National Center for Atmospheric Research Boulder, CO

2 Background : Model for Prediction Across Scales Global atmospheric NWP model designed to simulate down to the cloud-resolving scales Global mesh required Regional refinements supported No limited-area capability (e.g., ) yet Uniform Mesh setup in AMPS: Variable-resolution mesh w/refinement over Antarctica Variable Resolution Mesh

3 Recent Northern Hemisphere Evaluation Mesoscale Reference Physics Suite Northern Hemisphere Anomaly Correlation (20-80N) 60 km/15 km variable resolution mesh

4 AMPS and Model Configurations Model Setups domains: AMPS configuration 30/10/3.3/ 1.1-km domain: 60-km global / 15-km Antarctic Vertical levels: : 61 : 46 Model Tops: : 31 km (10 mb) : 30 km ( 12 mb) Terrain height shaded AMPS Domains

5 Methodology (cont d) Differences in and : Practical Constraints Prevent Identical Setups (i) Resolution: 10 km (& finer) v. 15 km (ii) Terrain data: RAMP2 () (200 m) v. GTOPO30 () (30 sec) (iii) Data assimilation & reanalysis: Yes No (iv) Subsurface temperature initialization: Cycled GFS (v) Physics: Not all physics packages/versions available to Different schemes PBL: MYJ YSU Microphysics: WSM-5 WSM-6 SW rad: Goddard RRTMG Common LSM: Noah Cu: Kain-Fritsch LW rad: RRTMG Surface layer: Eta

6 Methodology (cont d) Periods Examined (i) Winter (Jul. Aug. 2016): Surface (ii) Summer (Dec. Jan ): Surface (iii) Autumn (Apr. May 2017): Upper air Model Verification Data: AWS (~80 sites) for surface Radiosonde (12 sites) for upper-air Fields examined: T, Wind speed, Pressure, RH, U, V Statistics: Bias, RMSE, Corr

7 Forecast Comparison: & (Hr 96) 1200 UTC 5 June 2017 (1200 UTC 1 June init) SLP: int= 4 mb 3-hourly precip shaded Correspondence of synoptic and mesoscale systems and precip areas Analysis Evolution similar through 96 hrs (72 hrs more generally)

8 Verification Results Surface McMurdo Temperature Winter Summer Obs Stat significant difference Forecast hr & : Cold biases both seasons

9 McMurdo Temperature Stats Winter Summer RMSE better w/ significance Bias Winter Bias: = -3.6C = -4.8C RMSE: = 5.1C = 6.6C Bias : Lower bias, RMSE for both seasons & : Cold biases, both seasons Summer Bias: = -2.1C = -3.3C RMSE: = 3.2C = 4.4C Winter Increased error in both models

10 South Pole Temperature Winter Obs Summer Poor analysis? : Warm bias winter : Warm summer : Summer Analysis error issue

11 South Pole Temperature Stats Winter Summer Grid shutoff Winter Bias: = -1.3C = 3.0C RMSE: = 4.9C = 6.5C : Better winter Bias Summer Bias: = 1.6C = 0.1C RMSE: = 2.5C = 2.5C Both models: Winter Increased error : Better summer

12 Regional Surface T ( C) and Wind Speed (ms -1 ) Errors better better Summer T RMSE WS RMSE Ross Is East Antarctica Plateau/Pole Queen Maud Land West Antarctica Ant. Peninsula Both models: Winter performance dropoff Winter T RMSE WS RMSE Ross Is East Antarctica Plateau/Pole Queen Maud Land West Antarctica Ant. Peninsula

13 Sfc Wind Speed RMSE Winter Performance varies by region Summer finer grid Wind Speed RMSE better: Ross Ice Shelf, Ross Is. region, West Antarctica, Peninsula better: Plateau, East Antarctica, DML

14 McMurdo Upper Air Verification (Hr 48) Temperature Wind Speed Stat significant difference T: Low biases for both models / better lower trop Wind speed: better lower trop / better upper trop/lower strat

15 South Pole Upper Air Verification (Hr 48) Temperature Wind Speed Stat significant difference T: generally better, but difference ( tho stat significant) is minimal WS: better lower trop and strat, better middle trop Overall: T, WS biases are small, with no big differences between models

16 Upper Air Verification Hr 48 RMSE differences: Size proportional to improvement T 700 mb Better Model T 500 mb Wind Speed 700 mb Wind Speed 500 mb Overall: Balanced performance (mix of results aloft)

17 Summary Assessment Synoptic/mesoscale forecast evolution & progs in step w/producing same structures Fcsts close though 72 hr, solution divergence afterward Sfc verification overall better statistically than Both, degrade in winter forecasts shows an analysis issue 2-m T Upper-air verification Overall: Model error differences generally small: Balanced performance aloft relative performance (winds) tends to improve with height Sfc physics, terrain resolution differences of less influence

18 Moving Forward (cont d) Plans: Interoperability of and Capability to process data to initialize/drive Git repositories Git version control software for source code management for both Development of common physics repository: Availability of packages to code repository Common physics repository code repository Development of inter-model common physics driver (CPD): Supports model interoperability and use of physics from many modeling systems (e.g., GFS, FV3) Common post-processing and graphics tools Model source code checked out on file system