Mo Mode del Parame meter Estima mation n with h Da Data Assi Assimilation on usi sing g NICAM AM-LE LETKF Sh Shunji Ko Kotsuki 1 , Y. Sato 2 , K. Terasaki 1 , H. Yashiro 1 H. Tomita 1 , M. Satoh 3 , and T. Miyoshi 1 1. RIKEN Center for Computational Science, Japan 2. Graduate School of Engineering, Nagoya University, Japan 3. Atmosphere and Ocean Research Institute, U. Tokyo, Japan 7 th ISDA (ISDA2019), Jan. 22, 2019 @RIKEN-CCS, Kobe
NICAM-LETKF ( Terasaki et al. , 2015; SOLA) Big Data NICAM LETKF (JAXA) Local Ensemble Transform Kalman Filter (Hunt et al. 2007) Goal: Look for effective use of precipitation measurements.
Near-real-time NICAM-LETKF :: NEXRA Running o on J JAXA’s Supercomputer ( (JSS2) Kotsuki et al. (2019, , SOLA)
Model Parameter r Estimation wi with th Data Assimilati tion
Experimental setting • Num umerical Mode del – NICAM (Satoh et al. 2008, 2014) • Horizontal : GL6 (approx. 110 km resolution) • Vertical : 38 layers up to approx. 40 km • Cumulus Parameterization : Arakawa and Shubert (1974) • La Large Scale Condensati tion : Berry (1967) • Observations ns – State estimation: PREPBUFR, AMSU-A, GSMaP estimation: ??? ??? – Parameter e • Da Data Assimilation – LETKF (Hunt et al. 2007) with 40 members • Localization: 400 km (horizontal) & 0.4 ln p (vertical) • Inflation by RTPS (α = 0.90)
Parameter Estimation in NICAM-LETKF On Online!
Estimated Parameter (large scale condensation) Berry (1967)’s LSC sche heme De Defaul ult P Parameter ρ : air density r 2 Esti Es timated Parameter B 1 l ｓ = P : precipitation rate P N l : cloud water mixing ratio + B 2 B 3 c r Nc : total # of cloud droplet l : spread : mean w/ w/o Par aram ameter DA w/ w/ Par aram ameter DA OBS OBS (GS GSMaP_Ga Gauge) 2014/06/16/00UTC mm/6h ( Kotsuki et al. , 2018; JGR)
6-h fcst OLR BIAS (vs. ERA-Interim) CT CTRL (B1=0 =0.10) w/ w/ Parameter DA ç to too cloudy OLR bias increased
Ho How can we e improve e radiation bias s with parameter r DA? 1. 1. Es Estimating ng B1 pa parameter with h LWP (f (from G om GCOMW/AM AMSR2) 2. Estimating B1 parameter spatially
Parameter Estimation impact on LWP bias CTRL (B CT (B1=0.10) Gl Glob obal Parameter Estimation on LWP improved Pa Parameter DA (Local) OBS (GCOM OBS OMW/AM AMSR-2) 2) Tropi pics: good Ex Extra-tropi pics: unde underestimated
6-h fcst OSR BIAS (NICAM − CERES) CT CTRL (B1=0 =0.10) Globa bal Parameter DA [W/m 2 ] ç le less clo loudy cl clou oudier è OSR R improved significantly OSR: Outgoing Short Wave Radiation 201501-201512
Qu Question: Is spatially-va varying parameter r beneficial?
Global and Local Parameter Estimations Global Parameter r Estimation Local Parameter r Estimation By By ETKF By By LETKF lo localiz alizatio ion LETKF LET LET LETKF lo localiz alizatio ion ・ Estimate a global constant parameter ・ Estimate spatially-varying parameter ・ no localization ・ w/ localization observation
Experiments Parameter Estimation Pa Results Re Name Name of Exp. DA DA me method Ob Obs. s. LW LWP OS OSR CT CTRL Largely La / / Ov Overp rproduced (B1=0.1) (B ov overestimated Ko Kotsuki et al. ET ETKF GSMaP GS Ov Overp rproduced / (2018, , JGR) Glob Gl obal Con onstant ET ETKF LW LWP Good Good Improved Im Parameter DA Pa ? ? Lo Local (S (Spa patial) ) LETKF LE LW LWP Pa Parameter DA (σ=2 =200km)
Estimated Parameter Field (Berry’s B1) ç slower r conversion faster r conversion è de defaul ult value ue = = 0.1 .1
Estimated Parameter Field (Berry’s B1) Pr Promising seasonality off the coast of California (s (shallow convection in summer) global estimates local estimates global estimates local estimates de defaul ult time
Parameter Estimation impact on LWP bias CT CTRL Gl Glob obal Parameter Estimation on Pa Parameter DA (Local) OBS (GCOM OBS OMW/AM AMSR-2) 2) Tr Tropics: good Ex Extra-tr tropics: u underesti timated
LOCAL Parameter Estimation impact on LWP bias CT CTRL Gl Glob obal Parameter Estimation on Lo Local P Parameter E Estimation OBS OBS (GCOM OMW/AM AMSR-2) 2)
6-h fcst OSR BIAS (NICAM − CERES) Globa bal Parameter DA Lo Local P Parameter D DA [W/m 2 ] ç le less clo loudy cl clou oudier è local lo al par aram ameter DA seems benefic ficial ial in in shallo allow-co convection regions OSR: Outgoing Short Wave Radiation 201501-201512
Summary Pa Parameter Estimation Re Results Name Name of Exp. DA DA me method Ob Obs. s. LW LWP OSR OS CTRL / / Overproduced Overestimated (B1=0.1) Kotsuki et al. ETKF GSMaP Overproduced / (2018, JGR) Gl Glob obal Con onstant ET ETKF LW LWP Good Good im improved Pa Parameter DA Local (S Lo (Spa patial) ) LE LETKF LW LWP Be Better improved im Pa Parameter DA (σ=2 =200km) Kotsuki S., Terasaki K., Yashiro H., Tomita H., Satoh M. and Miyoshi T . (2018): Online Model Parameter Estimation with Ensemble Data Assimilation in the Real Global Atmosphere: J. Geophys. Res. Atmos. , 123, 7375-7392.
Recommend
More recommend
