Data assimilation, Real-time forecasting and high-resolution simulation during the 2011 CINDY/DYNAMO field campaign Tim Li and X. Fu (UH), Kunio Yoneyama and Tomoe Nasuno (JAMSTEC) Real time observational support Real-time MJO prediction using U. Hawaii hybrid coupled GCM Hindcast experiments during 2004-2008 have been conducted (Fu et al. 2009, 2010) A new initialization strategy WRF 3DVar/4DVar data assimilation during CINDY/DYNAMO campaign period Strategy: Combine in-situ observations with remotely sensing satellite measurements to generate regular grid (~10km) data A similar attempt during TCS-08 campaign was conducted using WRF 3DVar/4DVar (Li et al. 2009) NICAM global cloud-resolving (3.5km ~ 7 km) simulations JII: Initial condition and validation from the assimilation products Process-oriented study to reveal MJO initiation mechanisms (extratropical forcing vs. encircunavigation mode vs. local processes)
ISO Intensity in 14 IPCC AR-4 Climate Models OBS ECHAM CNRM Lin et al. (2006) East-West Direction 29HURRICANES, Tucson, May 10-14, 2010
NP-ISO is well Simulated by UH_HCM HCM UH_HCM OBS ECHAM-4 AGCM + UH_IOM Fu, Wang, Li, and McCreary, 2003, J. Climate
Experimental ISO Forecasts with UH_HCM Target Period: May-October 2004-2008 Forecast Interval: Every 10 days, totally 16 forecasts 10 Ensembles: Perturbations are 10% of daily differences Integration Length: 60 days Initial Conditions: R1/R2; ERA-Interim and modified reanalyses Skill Measure: Anomaly Correlation Coefficient over Southeast Asia and global tropics. 29HURRICANES, Tucson, May 10-14, 2010
ISO Forecast Skills in 2004 Summer (1 July 1986 30 June 1993) Westward Eastward Observed OLR ISO intensity in the reanalysis is 2-3 factors smaller! Reanalysis OLR Shinoda et al. (1999)
ISO Forecast Skills in 2004 Summer (May-October) ACC of Filtered U850 ACC of Filtered Rainfall Fu et al. (2009, GRL) Challenge in real-time application: no time filtering! 29HURRICANES, Tucson, May 10-14, 2010
filtered (shad) vs. non-filtered (cont) DJF MJO_U850 1990-2009 Pattern corr. = 0.81 RMSE = 1.6 m/s
WRF 3DVar Data Assimilation during TCS-08 Field Campaign Case: Super-typhoon Sinlaku, Sept. 7-21, 2008 (minimum pressure: 929hPa) Date: In-situ data: USAF C-130 dropsonde data(including T, Td, RH, U-wind, V-wind) NRL P-3 dropsonde data (including T, Td, RH, U-wind, V-wind) DLR Falcon Doppler Wind Lidar data (including U-wind, V-wind) DLR Falcon flight level data (including T, Td, RH, U-wind, V-wind) Satellite data: AIRS temperature and humidity QSCAT surface wind derived from Seawinds instrument AMSU temperature TRMM rainfall rate Other data: Synthetic observations final analysis (1 degree by 1 degree)
1010 1005 1000 995 990 985 NOGAPS The Profile of Pmin across TC center at 0906 TCDI/3DVar 5 980-540 -486-432 -378-324 -270-216 -162-108 -54 0 54 108 162 216 270 324 378 432 486 540 40 35 30 25 20 15 10 NOGAPS TCDI/3DVar The Profile of Speed across TC center at 0906 0-540 -486-432 -378-324 -270-216 -162-108 -54 0 54 108 162 216 270 324 378 432 486 540 The Profile of Pmin across TC center at 1006 1010 1000 990 980 970 960 NOGAPS TCDI/3DVar 950-540 -486-432 -378-324 -270-216 -162-108 -54 0 54 108 162 216 270 324 378 432 486 540 The Profile of Speed across TC center at 1006 35 30 25 20 15 10 NOGAPS 5 TCDI/3DVar 0-540 -486-432 -378-324 -270-216 -162-108 -54 0 54 108 162 216 270 324 378 432 486 540 1010 1000 990 980 970 960 950 940 The Profile of Pmin across TC center at 1112 The Profile of Speed across TC center at 1112 50 45 40 NOGAPS 35 TCDI/3DVar 30 25 20 15 NOGAPS 10 TCDI/3DVar 5-540 -486-432 -378-324 -270-216 -162-108 -54 0 54 108 162 216 270 324 378 432 486 540 0-540 -486-432 -378-324 -270-216 -162-108 -54 0 54 108 162 216 270 324 378 432 486 540
750 800 (A) The Speed Profile at South-East Quadrant at 1006 850 OBS 900 NOGAPS TCDI/3DVar 950 5 10 15 20 25 30 35 40 45 750 800 (B) TheSpeedProfileatNorth-West Quadrant at 1006 80 850 OBS 900 NOGAPS TCDI/3DVar 950 5 10 15 20 25 30 35 40 45 Top: 850hPa total wind speed 750 (C) 800 The Speed Profile at South-West Quadrant at 1006 Right: comparison of wind speed profile at 1006 900 850 OBS 900 NOGAPS TCDI/3DVar 950 5 10 15 20 25 30 35 40 45
750 800 (A) The Speed Profile at South-East Quadrant at 1112 850 OBS 900 NOGAPS TCDI/3DVar 950 5 10 15 20 25 30 35 40 45 50 55 60 TheSpeedProfileatNorth-West Quadrant at 1112 750 (B) 800 850 OBS 900 NOGAPS TCDI/3DVar 950 5 10 15 20 25 30 35 40 45 50 55 60 Top: 850hPa total wind speed 800 The Speed Profile at South_West Quadrant at 1112 750 (C) Right: comparison of wind speed 900 profile at 1112 850 OBS 900 NOGAPS TCDI/3DVar 950 5 10 15 20 25 30 35 40 45 50 55 60
Data Assimilation Strategy: Combine in-situ ship/island/flight observations and satellite observations Cover the entire campaign period Provider regular-grid grid (~10 km) 3-hourly reanalysis products for analysis, model initialization, and validation
Idealized AGCM experiments: What initiates MJO in IO? sensitivity experiments: No_restore: control exp Restore_nsbound Restore_TATL Restore_TAL_nsbound Bottom: Eastwardpropagating MJO energy spectrum
Thanks Diamond Head
Reanalysis-1 underestimates ISO (2004 Summer) TRMM 3B42 GPCP Reanalysis Rainfall _R1 2-3 factors smaller _R1 Eastward Propagation Northward Propagation 29HURRICANES, Tucson, May 10-14, 2010
Forecast Sensitivity Experiments EXP. _R_f Initial Conditions AC 1 + Filtered 2 _R_unf AC + Filtered + High-frequency 2* _ R _ f AC + 2*Filtered ed 3*_R_f 3*_R_unf AC + 3*Filtered AC + 3*Filtered + High-frequency 1 AC (Annual Cycle) => mean + first three harmonics 2 Filtered => 30-90-day variability 3 High-frequency => Total - AC - Filtered 29HURRICANES, Tucson, May 10-14, 2010
Daily Rainfall in 2004 Summer along the Equator R1 (10 o S-10 o N) 29HURRICANES, Tucson, May 10-14, 2010
Initialized on May 31, 2004 with Enhanced ISO in R1 29HURRICANES, Tucson, May 10-14, 2010
750 800 (A) The Speed Profile at East Quadrant at 0906 850 OBS 900 NOGAPS TCDI/3DVar 950 0 5 10 15 20 25 30 750 800 (B) The Speed Profile at North Quadrant at 0906 850 OBS 900 NOGAPS TCDI/3DVar 950 0 5 10 15 20 25 30 Top: 850hPa total t wind speed 800 750 (C) The Speed Profile at West Quadrant at 0906 Right: comparison of wind speed profile at 0906 850 OBS 900 NOGAPS TCDI/3DVar 950 0 5 10 15 20 25 30
1990-2009 DJF MJO OLR forecast skill 1.0 0.8 0.6 MJO pattern corr (DJF_OLR) Year 90 indicates mac m30d m30a5 forecast period from Dec 1989 to Feb 1990. 0.4 0.2 0.0 40.0 30.0 20.00 10.0 0.0 90 90 91 91 92 92 93 93 94 94 95 95 96 96 97 98 99 9 00 0 01 02 03 MJO rmse (DJF_OLR) mac m30d m30a5 97 98 99 00 01 02 03 04 04 05 05 06 06 07 07 08 08 09 09 avg avg mac: remove annual cycle (minus 90d low-pass filrering) i m30d: remove interannual component ( minus previous 30d average) m30a5: remove synoptic component (previous 5d average) Process of removing interannual components shows significant impact on the forecast skills in El Nino years (1991/1992 and 1997/1998).
1990-2009 DJF MJO U850 forecast skill 1.0 0.8 MJO pattern corr (DJF_U850) mac m30d m30a5 0.6 0.4 0.2 0.0 4.0 3.0 90 91 92 93 94 95 96 97 98 99 00 01 02 03 MJO rmse (DJF_U850) 04 05 06 07 08 09 av vg mac m30d m30a5 p MJO U850 forecast shows similar il results to OLR. Removing interannual components are important for El Nino years. 20 2.0 1.0 0.0 90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 avg
1990-2009 JJA MJO OLR forecast skill 1.0 0.8 MJO pattern corr (JJA_OLR) mac m30d m30a5 0.6 0.4 0.2 0.0 40.0 89 For the MJO forecast in boreal summer, the 8effects of removing interannual components MJO rmse (JJA_OLR) is not as large as it in mac m30d m30a5 wintertime. 90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 av vg 30.0 20.00 10.0 0.0 89 90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 avg
1990-2009 JJA MJO U850 forecast skill 1.0 0.8 MJO pattern corr (JJA_U850) mac m30d m30a5 0.6 0.4 0.2 0.0 4.0 3.0 2.0 89 90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 av vg MJO rmse (JJA_U850) MJO U850 forecast shows similar il results to OLR. During summertime, removing interannual components for the MJO forecast is not such important as it in winter. p mac m30d m30a5 1.0 0.0 89 90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05 06 07 08 09 avg
rmse 19.0 17.0 15.0 13.0 11.0 9.0 Test of averaged periods for interannual and synoptic components MJO forecast skill (DJF_OLR) m30a5 m60a5 m30a10 m60a10 pattern corr. 0.4 0.5 0.6 0.7 0.8 0.9 rmse MJO forecast skill (DJF_U850) 2.4 2.2 m30a5 m60a5 m30a10 m60a10 2.0 18 1.8 1.6 1.4 1.2 pattern corr. 1.0 0.4 0.5 0.6 0.7 0.8 0.9 rmse 19.0 17.0 15.0 13.0 11.0 9.0 MJO forecast skill (JJA_OLR) m30a5 m60a5 m30a10 m60a10 pattern corr. 0.4 0.5 0.6 0.7 0.8 0.9 rmse MJO forecast skill (JJA_U850) 2.4 2.2 m30a5 m60a5 m30a10 m60a10 2.0 1.8 1.6 1.4 1.2 1.0 pattern corr. 0.4 0.5 0.6 0.7 0.8 0.9 Optimal periods for interannual and synoptic components are previous 30day and previous 5day averages, respectively.
filtered (shad) and non-filtered (cont) DJF MJO_OLR 1990-2009 Pattern corr. = 0.76
filtered (shad) and unfiltered (cont) JJA ISO_OLR Pattern corr. = 0.73 RMSE = 11.15 W/m2
filtered (shad) and unfiltered (cont) JJA ISO_U850 Pattern corr. = 0.76 RMSE = 1.39 m/s