Ensemble-based Data Assimilation of TRMM/GPM Precipitation Measurements

Transcription

1 January 16, 2014, JAXA Joint PI Workshop, Tokyo Ensemble-based Data Assimilation of TRMM/GPM Precipitation Measurements PI: Takemasa Miyoshi RIKEN Advanced Institute for Computational Science Co-Is: H. Tomita (RIKEN), M. Satoh, M. Sawada (U. Tokyo), and E. Kalnay (U. Maryland) Researcher: K. Terasaki (RIKEN) Student: G.-Y. Lien (U. Maryland)

2 Data Assimilation (DA) Observations Numerical models Data Assimilation Vaisala Data assimilation best combines observations and a model, and brings synergy.

3 Project Overview TRMM/GPM NICAM LETKF Local Ensemble Transform Kalman Filter (Hunt et al. 2007) Goal: Look for most effective use of TRMM/GPM precipitation measurements.

4 Research plans FY2013 FY2014 FY2015 Develop NICAM-LETKF Assimilate conventional observations Assimilate TRMM-based precipitation data Applying the Lien et al. (2013) approach to NICAM-LETKF with real data. Optimization of NICAM parameters Assimilate TRMM/GPM and other satellite data Applying the EFSO approach to estimate observation impacts. High-resolution analysis

5 FY2013 List of Achievements NICAM-LETKF prototype system developed Successfully tested with perfect-model OSSEs Real observations successfully assimilated NCEP PREPBUFR Reasonable performance with Glevel-6 (112 km) and Glevel-7 (56 km) resolutions Precipitation assimilation methodology explored (U. Maryland) A conceptual paper published (Lien et al. 2013) TRMM-based precipitation product (TMPA) successfully assimilated with GFS-LETKF

6 NICAM-LETKF

7 Numerical Weather Prediction (NWP) Forecast Forecast Model Simulation Analysis Observation Analysis Analysis Observation True atmosphere (Unknown) time

8 We consider the evolution of PDF Analysis ensemble mean R Obs. Analysis w/ errors An approximation to KF with ensemble representations f f f Xt1( Xt P 1) t1 m 1 FCST ensemble mean T=t0 T=t1 T=t2 T

9 Flow chart of DA (Best estimate) Initial State Simulation PDF represented by an ensemble Simulated State DA Sim-to-Obs DA conversion Sim-minus-Obs Observations

10 Flow chart of DA (Best estimate) Initial State Simulation PDF represented by an ensemble Simulated State DA Sim-to-Obs conversion Observations Sim-minus-Obs Broad-sense DA

11 NICAM-LETKF Prototype LL ICO Initial State Grid conversions NICAM ICO LL Simulated State LETKF Sim-to-Obs conversion Observations Sim-minus-Obs Broad-sense DA

12 Results from perfect-model OSSE Observing network 500-hPa Height RMSE (m) Global 20-90S 20-90N Good performance

13 With real observations (Initial: Day 0) Lat: 20 90N (m) 2011/11/01 00UTC Initial time (Day 0)

14 With real observations (Day 1) Lat: 20 90N (m) 2011/11/02 00UTC Day 1

15 With real observations (Day 5) Lat: 20 90N (m) 2011/11/06 00UTC Day 5

16 With real observations (Day 20) Lat: 20 90N (m) 2011/11/21 00UTC Day 20

17 LETKF analysis is more suitable to NICAM. NOAA Analysis NICAM initialized by ERA-I LETKF -gl6 (112km) LETKF -gl7 (56km) gl9 (14km) 30-h forecast OLR showing NICAM s spin-up

18 PRECIP ASSIMILATION

19 Challenges of precip. assimilation Previous studies (e.g., Tsuyuki 1996; Mesinger et al. 2006) succeeded in forcing the model precipitation to be close to the observed values. However, the model forecasts tend to lose their additional skill after a few forecast hours. Major difficulties (Bauer et al. 2011): 1. Linear representation of moist physical processes for variational data assimilation 2. Non-Gaussianity of precipitation variable

20 An approach to precip. assimilation Major difficulties (Bauer et al. 2011): 1. Linear representation of moist physical processes for variational data assimilation 2. Non-Gaussianity of precipitation variable Approach 1. Using LETKF (or other EnKF methods) avoids linear representation of moist physics. 2. Applying empirical Gaussian transformation 3. Considering background PDF (requiring precipitating members) Lien et al. (2013) tested the approach and obtained promising results. Lien, G.-Y., E. Kalnay, and T. Miyoshi, 2013: Effective Assimilation of Global Precipitation: Simulation Experiments. Tellus, 65A,

21 Example of precipitation distribution in DJF near Maryland Lien et al. (2013)

22 Most recent results from Guo-Yuan Lien and Eugenia Kalnay REAL TMPA AND GFS-LETKF

23 Lien and Kalnay succeeded in assimilating TMPA. 3 month time series: Analysis U (m/s) at 500 hpa, error relative to ERA int. Global Tropics Verification period Verification period NH SH Verification period Raobs Raobs_TMPA Raobs_TMPA_Log Raobs_TMPA_GT Conv Verification period Solid lines: RMS error Dashed lines: Bias

24 5-day forecasts were improved. Global Tropics 2 month average U at 500 hpa Solid lines: RMS error Dashed lines: Bias NH SH Raobs Raobs_TMPA Raobs_TMPA_Log Raobs_TMPA_GT Conv

25 What did we do? Improved QC Corr[TMPA, GFS] > 0.45 Requiring background 24 members are raining (out of 32)

26 ENSEMBLE-BASED OBS IMPACT Kunii, Miyoshi and Kalnay (2012, Mon. Wea. Rev.) Kalnay, Ota, Miyoshi and Liu (2012, Tellus) Ota, Kalnay, Miyoshi and Derber (2013, Tellus)

27 Forecast Sensitivity to Observations (FSO) Estimated observation impact Degrading TY Sinlaku With FSO approaches, observation impacts can be estimated without performing expensive data denial experiments (or OSEs). Kunii, Miyoshi, Kalnay (2012) Improving

28 Impact of TMPA on GFS forecasts Improving Degrading Lien, Kalnay (2014)

29 Research plans FY2013 FY2014 FY2015 Develop NICAM-LETKF Assimilate conventional observations Assimilate TRMM-based precipitation data Applying the Lien et al. (2013) approach to NICAM-LETKF with real data. Optimization of NICAM parameters Assimilate TRMM/GPM and other satellite data Applying the EFSO approach to estimate observation impacts. High-resolution analysis