Recent performance statistics for AMPS real-time forecasts

32
Recent performance statistics for AMPS real-time forecasts Kevin W. Manning National Center for Atmospheric Research NCAR Earth System Laboratory Mesoscale and Microscale Meteorology Division Boulder, CO NCAR is sponsored by the National Science Foundation AMPS is sponsored by the NSF Office of Polar Programs and the NSF UCAR and Lower Atmosphere Facilities Oversight Section Workshop on Polar Simulations with the Weather Research and Forecasting (WRF) Model 02-03 Nov 2011, Columbus, OH 1

description

Recent performance statistics for AMPS real-time forecasts. Kevin W. Manning – National Center for Atmospheric Research NCAR Earth System Laboratory Mesoscale and Microscale Meteorology Division Boulder, CO – NCAR is sponsored by the National Science Foundation – - PowerPoint PPT Presentation

Transcript of Recent performance statistics for AMPS real-time forecasts

Page 1: Recent performance statistics for AMPS real-time forecasts

1

Recent performance statistics for AMPS real-time forecasts

Kevin W. Manning–

National Center for Atmospheric ResearchNCAR Earth System Laboratory

Mesoscale and Microscale Meteorology DivisionBoulder, CO

NCAR is sponsored by the National Science Foundation

–AMPS is sponsored by the NSF Office of Polar Programs

and theNSF UCAR and Lower Atmosphere Facilities Oversight Section

–Workshop on Polar Simulations with the Weather Research and Forecasting (WRF) Model

02-03 Nov 2011, Columbus, OH

Page 2: Recent performance statistics for AMPS real-time forecasts

2

AMPS real-time forecasts• 6 two-way interactive grids:

– 45-km / 15-km grids to 120 forecast hours– 5-km / 1.67-km grids to 36 forecast hours

• Initial conditions– GFS 0.5-degree analyes used as first guess for WRFDA 3D variational data assimilation step

(domains 1 and 2)– Sea-ice conditions from near real-time SSM/I daily global ice concentration (NSIDC) (25-km grid)

• Domain 1 boundary conditions– GFS 0.5-degree forecast updated at 6-hour intervals

• WRF options– 44 vertical levels; lowest half level ~ 12 m above surface; 12 layers below ~1 km above surface– Microphysics: WSM 5-class scheme– LW Radiation: RRTMG– SW Radiation: Goddard SW scheme– Surface Layer Physics: Monin-Obukhov (Janjic Eta) scheme– Land Surface: Noah Land-surface model; 4 subsurface layers– PBL Physics: MYJ (Eta) TKE scheme

Workshop on Polar Simulations with the Weather Research and Forecasting (WRF) Model

Page 3: Recent performance statistics for AMPS real-time forecasts

3Workshop on Polar Simulations with the Weather Research and Forecasting (WRF) Model

Page 4: Recent performance statistics for AMPS real-time forecasts

4

AMPS statistics – Summertime Surface

• Nov-Dec-Jan 2010/2011 season– Temperature– Pressure– Wind

• Surface station reports– From GTS– From University of Wisconsin – Antarctic Meteorological Research Center

(AMRC)• Three regions

– Ross Ice Shelf– East Antarctic plateau– East Antarctic coastal

• Older WRF version 3.0.1.1 with Polar modifications

Workshop on Polar Simulations with the Weather Research and Forecasting (WRF) Model

Page 5: Recent performance statistics for AMPS real-time forecasts

5

Summer T – Ross Ice Shelf

Mean Statistics -- ~ 15 stations

BiasRMSECorrelation (-1.0 to 1.0)

Workshop on Polar Simulations with the Weather Research and Forecasting (WRF) Model

Forecast Hour (0 – 120)

Forecast Hour (0 – 120)

Page 6: Recent performance statistics for AMPS real-time forecasts

6

Summer T – East Antarctic Plateau

Mean Statistics ~ 6 stations

BiasRMSECorrelation (-1.0 to 1.0)

Workshop on Polar Simulations with the Weather Research and Forecasting (WRF) Model

Forecast Hour (0 – 120)

Forecast Hour (0 – 120)

Page 7: Recent performance statistics for AMPS real-time forecasts

7

Summer T – East Antarctic Coastal

Mean Statistics -- ~10 stations

BiasRMSECorrelation (-1.0 to 1.0)

Workshop on Polar Simulations with the Weather Research and Forecasting (WRF) Model

Forecast Hour (0 – 120)

Forecast Hour (0 – 120)

Page 8: Recent performance statistics for AMPS real-time forecasts

8

Summer P – Ross Ice Shelf

Mean Statistics -- ~ 15 Stations

BiasRMSECorrelation (-1.0 to 1.0)

Workshop on Polar Simulations with the Weather Research and Forecasting (WRF) Model

Forecast Hour (0 – 120)

Forecast Hour (0 – 120)

Page 9: Recent performance statistics for AMPS real-time forecasts

9

Summer P – East Antarctic Plateau

Mean Statistics -- ~ 6 stations

BiasRMSECorrelation (-1.0 to 1.0)

Workshop on Polar Simulations with the Weather Research and Forecasting (WRF) Model

Forecast Hour (0 – 120)

Forecast Hour (0 – 120)

Page 10: Recent performance statistics for AMPS real-time forecasts

10

Summer P – East Antarctic Coastal

Mean Statistics -- ~10 stations

BiasRMSECorrelation (-1.0 to 1.0)

Workshop on Polar Simulations with the Weather Research and Forecasting (WRF) Model

Forecast Hour (0 – 120)

Forecast Hour (0 – 120)

Page 11: Recent performance statistics for AMPS real-time forecasts

11

Summer Wind Speed – Ross Ice Shelf

Mean Statistics -- ~ 15 stations

BiasRMSECorrelation (-1.0 to 1.0)

Workshop on Polar Simulations with the Weather Research and Forecasting (WRF) Model

Forecast Hour (0 – 120)

Forecast Hour (0 – 120)

Page 12: Recent performance statistics for AMPS real-time forecasts

12

Summer Wind Speed – East Antarctic Plateau

Mean Statistics -- ~ 6 stations

BiasRMSECorrelation (-1.0 to 1.0)

Workshop on Polar Simulations with the Weather Research and Forecasting (WRF) Model

Forecast Hour (0 – 120)

Forecast Hour (0 – 120)

Page 13: Recent performance statistics for AMPS real-time forecasts

13

Summer Wind Speed – East Antarctic Coastal

Mean Statistics -- ~10 stations

BiasRMSECorrelation (-1.0 to 1.0)

Workshop on Polar Simulations with the Weather Research and Forecasting (WRF) Model

Forecast Hour (0 – 120)

Forecast Hour (0 – 120)

Page 14: Recent performance statistics for AMPS real-time forecasts

14

• Summertime Summary– Warm bias on plateau. Mixed temperature bias in other

regions.– Warming trend in East Antarctic plateau and Ross Ice Shelf

regions. Very little temperature trend for coastal stations.– Plateau stations show greatest temperature error growth

(RMSE) over 120 hours. Little temperature error growth (RMSE) for Ross Ice Shelf and coastal regions.

– Pressure statistics show high correlation in all three regions, but steady error (RMSE) growth.

– Low pressure bias increasing in time over Ross Ice Shelf.– Slight high wind-speed bias

Workshop on Polar Simulations with the Weather Research and Forecasting (WRF) Model

Page 15: Recent performance statistics for AMPS real-time forecasts

15

AMPS Behavior – Wintertime Surface

• May-Jun-Jul 2011• Newer WRF version 3.2.1 with Polar

Modifications

Workshop on Polar Simulations with the Weather Research and Forecasting (WRF) Model

Page 16: Recent performance statistics for AMPS real-time forecasts

16

Winter T -- Ross Ice Shelf region

Mean Statistics -- ~21 stations

BiasRMSECorrelation (-1.0 to 1.0)

Workshop on Polar Simulations with the Weather Research and Forecasting (WRF) Model

Forecast Hour (0 – 120)

Forecast Hour (0 – 120)

Page 17: Recent performance statistics for AMPS real-time forecasts

17

Winter T -- East Antarctic Plateau

Mean Statistics – ~13 stations

BiasRMSECorrelation (-1.0 to 1.0)

Workshop on Polar Simulations with the Weather Research and Forecasting (WRF) Model

Forecast Hour (0 – 120)

Forecast Hour (0 – 120)

Page 18: Recent performance statistics for AMPS real-time forecasts

18

Winter T -- East Antarctic Coastal

Mean Statistics -- ~15 stations

BiasRMSECorrelation (-1.0 to 1.0)

Workshop on Polar Simulations with the Weather Research and Forecasting (WRF) Model

Forecast Hour (0 – 120)

Forecast Hour (0 – 120)

Page 19: Recent performance statistics for AMPS real-time forecasts

19

Winter P -- Ross Ice Shelf region

Mean Statistics -- ~21 stations

BiasRMSECorrelation (-1.0 to 1.0)

Workshop on Polar Simulations with the Weather Research and Forecasting (WRF) Model

Forecast Hour (0 – 120)

Forecast Hour (0 – 120)

Page 20: Recent performance statistics for AMPS real-time forecasts

20

Winter P -- East Antarctic Plateau

Mean Statistics -- ~13 stations

BiasRMSECorrelation (-1.0 to 1.0)

Workshop on Polar Simulations with the Weather Research and Forecasting (WRF) Model

Forecast Hour (0 – 120)

Forecast Hour (0 – 120)

Page 21: Recent performance statistics for AMPS real-time forecasts

21

Winter P -- East Antarctic Coastal

Mean Statistics -- ~15 stations

BiasRMSECorrelation (-1.0 to 1.0)

Workshop on Polar Simulations with the Weather Research and Forecasting (WRF) Model

Forecast Hour (0 – 120)

Forecast Hour (0 – 120)

Page 22: Recent performance statistics for AMPS real-time forecasts

22

Winter Wind Speed -- Ross Ice Shelf region

Mean Statistics -- ~21 stations

BiasRMSECorrelation (-1.0 to 1.0)

Workshop on Polar Simulations with the Weather Research and Forecasting (WRF) Model

Forecast Hour (0 – 120)

Page 23: Recent performance statistics for AMPS real-time forecasts

23

Winter Wind Speed – East Antarctic Plateau

Mean Statistics -- ~13 stations

BiasRMSECorrelation (-1.0 to 1.0)

Workshop on Polar Simulations with the Weather Research and Forecasting (WRF) Model

Forecast Hour (0 – 120)

Forecast Hour (0 – 120)

Page 24: Recent performance statistics for AMPS real-time forecasts

24

Winter Wind Speed -- East Antarctic Coastal

Mean Statistics -- ~15 stations

BiasRMSECorrelation (-1.0 to 1.0)

Workshop on Polar Simulations with the Weather Research and Forecasting (WRF) Model

Forecast Hour (0 – 120)

Forecast Hour (0 – 120)

Page 25: Recent performance statistics for AMPS real-time forecasts

25

• Wintertime summary– In each region, a mix of high and low temperature biases

– average bias near zero.– Over plateau, strong signal of initial condition warm bias

that the model quickly corrects.– Larger temperature error (RMSE) growth, larger reduction

of temperature correlation, than we saw in summer.– As in summer, pressure statistics show significant error

growth (RMSE)– High wind speed bias, more notable than in summer.

Workshop on Polar Simulations with the Weather Research and Forecasting (WRF)

Model

Page 26: Recent performance statistics for AMPS real-time forecasts

26

Subsurface Temperature Initialization

• BPRC Antarctic results differ from AMPS results– BPRC results show cold bias in summer over East Antarctic Plateau– AMPS results show warm bias in summer over East Antarctica Plateau

• One possibly significant difference in the AMPS and BPRC is the initialization of subsurface temperature fields– AMPS cycles the subsurface temperature from one forecast to the next

• High resolution details• In balance with WRF physics• Subject to model drift

– BPRC initializes subsurface temperature fields using a 40-year annual mean air temperature analysis at deep ice layers, and a 40-year monthly mean air temperature analysis at the shallowest subsurface layer• No spin-up required• Low resolution

• Could this account for the different results for forecasts of air temperature?

Workshop on Polar Simulations with the Weather Research and Forecasting (WRF) Model

Page 27: Recent performance statistics for AMPS real-time forecasts

27

Two experiments

• Cycled subsurface temperature fields (CYCLE)• Subsurface temperature fields initialized from

monthly mean and annual mean temperatures (MEANT)

• Two 72-hour forecasts per day, in the AMPS 45km/15km configuration, from 10 Jan through 06 Feb 2011 (about 4 weeks).– The CYCLE conditions have been spun up for about 6

weeks, starting from AMPS real-time fields (i.e., already using the real-time cycled conditions) from 01 Dec 2010.

Workshop on Polar Simulations with the Weather Research and Forecasting (WRF) Model

Page 28: Recent performance statistics for AMPS real-time forecasts

28Workshop on Polar Simulations with the Weather Research and Forecasting (WRF) Model

MEANT Level 2 (-0.25 m) T

CYCLE Level 2 (-0.25 m) Ice T

DifferenceLevel 2 T (-0.25 m)CYCLE – MEANT

Averages at Forecast hour 00

Page 29: Recent performance statistics for AMPS real-time forecasts

29Workshop on Polar Simulations with the Weather Research and Forecasting (WRF) Model

MEANT Level 4 (-1.5 m) Ice T

CYCLE Level 4 (-1.5 m) Ice T

DifferenceLevel 4 T (-1.5 m)CYCLE – MEANT

Averages at Forecast hour 00

Page 30: Recent performance statistics for AMPS real-time forecasts

30Workshop on Polar Simulations with the Weather Research and Forecasting (WRF) Model

MEANT 2-m air Temperature

CYCLE 2-m air Temperature

Difference2-m air TCYCLE – MEANT

Averages at Forecast hour 72

Page 31: Recent performance statistics for AMPS real-time forecasts

31Workshop on Polar Simulations with the Weather Research and Forecasting (WRF) Model

Page 32: Recent performance statistics for AMPS real-time forecasts

32

Short-term plans for sea-ice code in WRF, particularly Noah LSM

• Currently, sea-ice code is scattered throughout the Noah LSM code– Difficult to trace the sea-ice processes through the code– Difficult to develop or replace

• Plan: Pull sea-ice code out of Noah-LSM, and make the Noah sea-ice treatment its own separate WRF module– Easy to trace the sea-ice processes through the code– Easy to develop or replace

• A place to link up with more sophisticated sea-ice schemes or models– Easy to use with other LSM options (e.g., the new Noah-MP

LSM)

Workshop on Polar Simulations with the Weather Research and Forecasting (WRF) Model