Numerical Experiments of Tropical Cyclone Seasonality over the Western North Pacific Dong-Kyou Lee School of Earth and Environmental Sciences Seoul National University, Korea Contributors: Suk-Jin Choi, Cheon-Sil Jin and Dong-Hyun Cha June 5-6, 2012
Contents 1. Motivation and Objective 2. Methodology (Data, Model, Experiment) 3. Results from Numerical Experiments 4. Concluding Remarks
1. Motivation and Backgrounds The prediction of tropical cyclone (TC) seasonality is one of important issues in meteorological research and operation. A regional climate model (RCM) using global climate model (GCM) prediction outputs can take advantage of detail physical processes and high-resolution grids by dynamical downscaling technique. In recent years, sub-seasonal and seasonal prediction has been discussed together with numerical weather prediction and climate prediction. The Weather Research and Forecasting (WRF) model can be employed to improve the seasonal forecast of high-impact weather features such as extreme precipitation and tropical cyclone. In this preliminary study, we attempt to simulate seasonal features of tropical cyclones over the western North Pacific Ocean in 2002.
Concept of RCM Large-scale flow Convection In higher resolution, detailed topography and sophisticated physical processes can add meso-scale features (mainly related to convection) to large-scale flows of coarse global data (e.g., GCM data and global analysis) where convective activities can not or weakly be resolved.
- A RCM is nested within a GCM in order to locally increase the model resolution GLOBAL Hemispheric scale Coupled AOCM Coarser resolution GCM provides initial and lateral boundary conditions to drive RCM MESOSCALE/REGIONAL Downscaling Regional climate model Land/Atmosphere interaction Regional hydrology Higher resolution RCM provides mesoscale and regional detail information
Objective Understanding of sub-seasonal and seasonal prediction performance of tropical cyclone in the WRF model forced by large-scale data or global model outputs In this study, we investigate - Behaviors of scales between large-scale fields and modelgenerated tropical cyclones in regional model - Role of dynamical downscale method and slab ocean model implemented in the WRF model - Effect of different initial data and dynamical downscaling from long-term global model data
2. Methodology WRF-based RCM - Domain : Model Configuration Model prototype NCAR/WRF/ARW Vertical layers (top) 36 layers (50 hpa) Horizontal grids 30 km Horizontal grids number 240 X 200 (7200 X 6000km) Integral time-step 90 sec Cumulus convection Kain-Fritch Explicit moisture WSM3 Simple Ice PBL YSU Short / Long Radiation RRTM/Dudhia Land surface NOAH LSM - Data : NCEP Final (FNL) analysis and NCEP/DOE R-2 reanalysis data for IC, BC RSMC Tokyo-Typhoon Center best track data 3-h 0.25 0.25 TRMM precipitation data (3B42 product) Daily 0.25 0.25 AMSR+AVHRR OISST data HadGEM2 global data (a historial run)1.875x1.25, 38 layers
Spectral nudging (SN) Keep model-generating small-scale details, prevent model from drift, and provide the added-value (Waldron et al. 1996, von Storch et al. 2000, Miguez-Macho et al. 2004) Spectral Nudging u t f = F ( X ) + w ( u u ) η a L f L Subscript L : Large-scale fields filtered out small-scale features Subscript f : Forecast Subscript a : Analysis (driving data) Subscript η: Vertically varied value F(X) : Model operator Grid nudging u t f = F ( X ) + w( u u ) a f
Cut-off wavelength (1000km) Uf (RCM) Ua (FNL analysis) UfL (RCM) UaL (FNL) +
Slab ocean model (SOM) Simulate the ocean temperature feedback in the mixed layer ocean model (1-d single-layer model, Pollard 1973; Emanuel et al., 2004; Davis et al., 2008) which is only concerned on the first order negative feedback of wind-driven ocean mixing Z h Z h ρ : Density of seawater h : Mixed layer depth u : Mixed layer velocity τ : Vector wind stress C : Heat capacity of seawater T : Temperature (i : initial) U T Entrainment into the mixed layer is given by bulk Richardson number(r=1)
Detection of TCs (Nguyen and Walsh, 2001, Oouchi et al., 2006) 1. Local minimum of sea level pressure (SLP) 2. Surface wind > 17.5 m/s (over Severe Tropical Strom (RSMC scale)) 3. Relative vorticity at 850 hpa > 3.5 X 10-5 /s 4. Warm-core criterion : ΔT Δ T +Δ T +ΔT K sum 300 500 700 1.5 5. Maximum wind speed at 850 hpa near the point is larger than that at 300 hpa 6. Temperature anomaly at 300 hpa > Temperature anomaly at 850 hpa 7. Lifetime > 1 day
Simulation A seasonal TC event over western North Pacific (WNP) is simulated on the WRF model during 3 month period of June 18-September 28, 2002 The 2002 TC season is chosen because the annual TC frequency is 26. This is very close to the 30-year climatic mean occurrence of 26.09
3. Results ACTS Symposium 5 June 2012
Cross correlation coefficient of 850 hpa geopotential height (GPH) Mean synoptic fields (200hPa Jet (shaded), 5880hPa GPH (contour), 850hPa Wind) OBS CTL
3. Results Effect of SOM To investigate the ocean temperature feedback in RCM, SOM is applied and modified to simulate seasonal TC activity for shortterm integration. SOM does not include the mechanism of recovering a mixed layer depth for long-term simulation. In this study, the temperature and depth of ocean mixed layer are adjusted to observation with 1-day memory, because near-inertial motions dominate the response of ocean mixed layer current to TC passage on the time scale of order 1 day (Price 1981).
Mean SST difference and TC tracks Mean difference : 0.26 FNL SOM
Frequency : 12 ACE index : 146.0 RSMC CTL Frequency : 25 ACE index : 223.0 CTL SOM Frequency : 20 ACE index : 214.0 SOM
3. Results Effect of SPN FNL analysis CTL CTL SPN SPN
Temperature profiles over the tropical ocean (5-25N, 130-160E) SPN CTL - SPN The mean atmospheric profiles over tropical ocean show conditionally unstable condition in the lower troposphere. The increased T and Q at lowest levels in CTL induce more conditionally unstable conditions which provide more favorable generation of Tropical cyclones. θ θ e
Comparison between CTL and SPN 3-month mean synoptic fields (200hPa Jet, shaded, 5880hPa GPH, contour, 850hPa Wind) OBS CTL SPN SPN recovers the western North Pacific high, low-level cyclonic circulation and upper-level westerly wind patterns
Initial position, number and ACE of TCs OBS(RSMC) CTL SPN The number of TCs [ACE index (10 4 kt 2 )] OBS 12 [146.0] CTL 25 [223.0] 200% [150% ] SPN 16 [142.0] The accumulated cyclone energy (ACE) index is defined as the sum of the squares of the 6-hourly maximum wind speed for all simulated TCs while they are at least tropical storm strength (Bell et al.,2000)
3-month mean precipitation (mm/day) OBS CTL SPN SPN reduces overestimated rainfall over the tropical ocean and simulates rainfall in Southern China.
Probability density of TC tracks OBS(RSMC) CTL SPN Note that there are two main tracks in PDF
Frequency : 12 ACE index : 146.0 Frequency : 25 ACE index : 223.0 Frequency : 16 ACE index : 142.8
EOF first mode eigenvectors of 6-hourly 925 hpa wind vector OBS(14.1%) CTL(12.6%) SPN(13.6%) PC Timeseries OBS CTL SPN
Temporal variation of 850 hpa GHT pattern correlation between observation and simulations CTL SPN red : < 960hPa Without the adjustment of large-scale field, simulated atmospheric flows are considerably deviated from the boundary forcing and relatively strong TCs of central pressure less than 960 hpa substantially affects simulated atmospheric patterns.
CTL SPN
Difference fields (CTL-SPN) Windspeed at 850hPa UST Compared to SPN, the increase of windspeed in CTL results in the increase of frictional velocity (UST), which enhances latent heat flux (LH) and moisture flux (QFX) from the sea surface through the PBL process. QFX LH So, T and Q at the lowest levels are increased in CTL.
Minimum central pressure of TC (Saffir-Simpson scale) OBS CTL SPN
Relation between central pressure and life time of TC ACTS Symposium 5 June 2012
3-month mean spectral variances of the kinetic energy at 200 hpa and the ratio of CTL to SPN (CTL/SPN) CTL SPN Cut-off Wave number = 7 Typical TC radius = WN 15~20 = WL 350km~500km Largest spectral variance difference occurs in the large-scale regime below cutoff wave. Small-scale circulations are readily affected by the large kinetic energy of spurious large-scale circulations influenced by TC without the adjustment of large-scale field.
Simulated cloud-top T SNP GMS Image
Without spectral nudging, westerly monsoon flows are strengthened, the subtropical Pacific high is weakened, and tropical cyclones tend to have large radii of track curvatures, much deepening central pressure and shifted eastward locations.
Effect of different initial times Experiments at 3 initial times (12, 15 and 18 June 2002) The number of TCs Initial data 20020612 18 Initial data 20020615 17 Initial data 20020618 16
Probability density of TC tracks 020612 020615 020618
Mean of 3 simulations ACTS Symposium 5 June 2012
From HadGEM2 data (2002) First position and number of TCs HadGEM2 WRF SLP < 1010 hpa, 12 TCs SLP < 1000 hpa, 14 TCs
Probability density of TC tracks HadGEM2 WRF
Relation between central pressure and life time of TC Hadgem2 WRF
4. Concluding Remarks ACTS Symposium 5 June 2012 Season-long simulated tropical cyclones considerably modifies large-scale environment in the model domain so that westerly monsoon flows are strengthened and the subtropical Pacific high is shifted toward eastward in the model domain. Simulated frequency, intensity and tracks of tropical cyclones are significantly improved, with large-scale flow fields improved in the model. It is important to have accurate large-scale (forecast) fields. The detection method of simulated tropical cyclones (Oouchi et al, 2006) works well to apply to model data of 30km horizontal resolution in this study. This study shows that dynamical and physical scale interactions between internal disturbances generated in model integration and large-scale flow forcing need to be properly treated for season-long simulations of tropical cyclone.
Thank you! ACTS Symposium 5 June 2012