Observing System experiments with ECWMF operational ocean analysis (ORA-S3)
Observing System Simulation Experiments for COSMIC-II
description
Transcript of Observing System Simulation Experiments for COSMIC-II
Observing System Simulation Experiments for
COSMIC-IIUCAR COSMIC Team
FORMOSAT-3/COSMIC Follow-on Mission Planning
• NSC/NSPO and NOAA are discussing a possible collaboration on the FORMOSAT-3 Follow-On Mission, which is called “COSMIC-II.”
• Preliminary design calls for 12 low-Earth orbiting satellites, each carrying an advanced receiver to track thee navigation systems, including GPS, GALILEO, and GLONASS.
• Observing system simulation experiments (OSSE) are useful to assess the potential impacts of the COSMIC-II mission, and to assist in constellation design.
Three Key Questions:
• What are the potential impacts of FORMOSAT-3 Follow-on on the prediction of typhoons in the vicinity of Taiwan?
• What is the optimal design of the Follow-on Mission? [This needs to be looked at from global, regional weather prediction, climate, and space weather perspectives.]
• What are the relative performance of FORMOSAT-3 Follow-on compared with the existing FORMOSAT-3 mission?
Two Possible Configurations
• Option A: – 8 satellites placed at 72 degree inclination angle– 4 satellites placed at 24 degree inclination angle
• Option B:– 12 satellites placed at 72 degree inclination angle
Distribution of RO soundings in a day
FORMOSAT-3/COSMIC COSMIC-IIA COSMIC-IIBG A B
Different color shows availability of RO soundings at different hours of the day.
COSMIC - 6 x 72o
COSMIC-IIA - 8 x 72o + 4 x 24o
COSMIC-IIB - 12 x 72o
Data Density for FORMOSAT-3, COSMIC-IIA, COSMIC-IIB
COSMIC-IIA Provides a much more even data density around the globe.Enhanced data density over the tropics is important for typhoon prediction.
Selected Case: Typhoon Shanshan (2006)
• Min. pressure of 920 hPa on Sep 15, 2006
Nature Run from MM5 with ECMWF Initial condition
• Typhoon track recurvature was reproduced• The track is close to the best track• Note that no official intensity observation is available
Forecast Experiment Design
• WRF-Var (3D-Var) /WRF with GFS IC and LBC– B.E.: generated from one month forecast
of September, 2006 (NMC)– 169 x 157 x 38, ptop = 10 hPa– Assimilation performed on 36-km grid,
1hr update cycle, over two-day period.
Convention Data Are Assimilated As Well
• Upper Air sounding (SOUND)• Observation (SYNOP)• Satellite Cloud track Wind (SATOB)• The horizontal and temporal distribution of these data are
consistent with actual observations (location/time, and with realistic errors )
Synop SOUND upper air Satellite wind (SATOB)
Simulated RO Refractivity Data
• For GPSRO refractivity, the observation errors vary with height and latitude (Chen and Kuo, 2005).
• RO soundings are simulated from the nature run as local value, no ray tracing simulation.
GPS OBS_ERR Based on Chen-Kuo
0
4
8
12
16
20
0 0.5 1 1.5 2 2.5 3 %N
KM obs_err_new lat 0 lat 90
Percentage of GPS Ref observations
Equator
pole
Height
Intensity Forecast
• Intensity Forecast Performance: C+A > C+B > C+G > C
C+G
C+A
C+B
C
Nature
Intensity Forecast Improvement against control (conventional data) forecast
• FORMOSAT-3 only shows modest improvement over Control.• COSMIC-IIA shows significant improvements over control.• COSMIC-IIA is superior to COSMIC-IIB
Percentage of improvement of intensity forecast
Percent Improvement relative to the control:
P.I. = (Error of Exp. C minus Error of Exp. X)/Error of Exp. C
Track Forecast Errors
• Performance: C+A > C+B > C+G > C
C+G
C+A
C+B
C
Track Forecast ErrorsImprovement against control (conventional data) forecast
Percentage of improvement of track forecast
• FORMOSAT-3 only shows visible improvement over Control.• COSMIC-IIA shows significant improvements over control.• COSMIC-IIA is superior to COSMIC-IIB
6 hour Integrated Precipitation Forecast
No precipitation system was developed in C, C+G, and C+B, only C+A
C+G C+A C+BN C
6 hour Integrated Precipitation (cont’d)
• Precipitation in C+A is much closer to that in nature run
C+G C+A C+BN C
6 hour Integrated Precipitation (cont’d)
• Late development of precipitation in C, C+G, and C+B, but in wrong locations
C+G C+A C+BN C
6 hour Integrated Precipitation (cont’d)
• All precipitation is stronger, but C+A still show better location
C+G C+A C+BN C
Summary and Conclusions
• COSMIC-IIA gives a much more uniform data distribution globally, compared with COSMIC-IIB.
• Data density is important for typhoon prediction:– FORMOSAT-3: < 1 over 500 km x 500 km– COSMIC-IIA: > 8 over 500 km x 500 km– COSMIC-IIB: < 4 over 500 km x 500 km
Typhoon Forecast Improvements• We perform two-day data assimilation, followed
with three-day forecast for FORMOSAT-3, COSMIC-IIA, and COSMIC-IIB.
• Compared with the Control (without RO data) COSMIC-II gives far superior results.
Intensity forecast Track forecast
FORMOSAT-3 8.1 25.0
COSMIC-IIA 43.3 79.1
COSMIC-IIB 26.0 39.5
Summary and Conclusions• COSMIC-IIA also gives significantly better
precipitation forecasts, in terms of rainfall intensity and distribution.
• The Option A design will greatly benefit the prediction of severe weather events over the Taiwan area, including typhoon, Mei-Yu, and mesoscale convective systems.
Data Density for COSMIC and COSMIC-II Options: A, B, C, and D
COSMIC - 6 x 72o
COSMIC-IIA - 8 x 72o + 4 x 24o
COSMIC-IIB - 12 x 72o
COSMIC-IIC - 6 x 72o + 6 x 24o
COSMIC-IIB - 4 x 72o + 8 x 24o
Intensity Forecast
Track Forecast Errors
Intensity Forecast Improvement against control (conventional data) forecast
• Intensity: C+D ~ C+C ~ C+A > C+B > C+G > (C) in evident
Percentage of improvement of intensity forecast
Track Forecast ErrorsImprovement against control (conventional data) forecast
• Track forecast: C+A ~ C+D > C+C > C+B > C+G > (C) in evident
Percentage of improvement of track forecast
6 hour Integrated Precipitation
2A: 8x72 + 4x24, 2B: 12x72 + 0x24, 2C: 6x72 + 6x24, 2D: 4x72 + 8x24
C+G C+A C+BN C+C C+D
Future Work• Additional cases that affect Taiwan area:
– Sinlaku (2008), Jangmi (2008), and other T-PARC and U.S. cases– Mei-Yu convective systems and heavy rainfall events
• More realistic simulation of observations:– Ray tracing simulation of RO observations– Take into account available ground stations, and data latency
• Use different data assimilation systems:– Global and Regional NCEP GSI system– WRF/DART ensemble data assimilation system
• Collaborate with NCEP and ECMWF on global OSSEs and observing system experiments (OSE).
The COSMIC Team
Bill Kuo: P.I. on OSSE StudyTed Iwabuchi: WRF-3D-Var and ForecastBill Schreiner: Mission simulationZaizhong Ma: Observation simulationsTae-Kwon Wee:Nature runYong-Run Guo: WRF-Var and Obs. Simulation
This work is funded by Dr. Jay Fein at NSF.