Tropical cyclone genesis frequency over the western North Pacific simulated in WCRP CMIP3 CGCMs Satoru Yokoi (CCSR, UT) Yukari N. Takayabu (CCSR, UT) Johnny C. L. Chan (City U. Hong Kong)
Motivation How may tropical cyclone (TC) frequency change in warmer climate? Many previous studies reported that global total TC number will decrease. But, quantitative, probabilistic, and regional prediction is yet to be achieved. It seems necessary to adopt multi-model and multi-ensemble approach. 20C 20C 21C 21C Difference Sugi et al. (2002, JMSJ) Bengtsson et al. (2007, Tellus)
CGCM output available at PCMDI Under CMIP3 (Phase 3 of Climate Model Intercomparison Project) Daily-mean 3-D atmospheric variables of current and global warming climate simulations are available at PCMDI (Program for Climate Model Diagnosis & Intercomparison). 12 CGCMs Data period: 40 years (current) & 200 years (GW experiment in total). Horizontal resolution of the CGCMs is T106 (1.1) at the utmost, and mainly T42 (2.8) or T63 (1.9). Studies on TC intensity is difficult, but frequency is possible. At first, we have to check CGCM performance on TCs.
Purpose Evaluate CGCMs’ performance in simulating TC genesis frequency (TCGF) over the western North Pacific, with focusing on horizontal distribution and seasonal march, …, and discuss possible reasons for revealed biases from the viewpoint of reproducibility of environmental conditions that affect TCGF.
CGCMs and data 12 CGCMs that 3-D daily-mean atmosphere data are available at PCMDI. 20th century climate simulations (40 years) are analyzed. Daily mean u, v, T, and monthly mean u, v, T, H, and SST. Data JMA (Japan Meteorological Agency) best track data (1977-2006) ECMWF 40-years reanalysis (1982-2001) NOAA OISST (1982-2001) BCCR-BCM2.0, CGCM3.1(T63), CNRM-CM3, CSIRO-Mk3.0, CSIRO-Mk3.5, ECHAM5/MPI-OM, FGOALS-g1.0, GFDL-CM2.0, GFDL-CM2.1, INGV-SXG, MIROC3.2(hires), MRI-CGCM2.3.2
Definition of TC-like disturbance (1) z850 local maximum (=TC center) zt. (cyclonic vortex) (2) (T300 at the center)-(environmental T300) Tt. (warm core) (3) Conditions 1 & 2 are satisfied at least 3 time steps. (4) Genesis point is over the ocean. (5) At genesis time, maximum wind speed is greater at the 850-hPa level than at the 300-hPa level. (exclusion of extra-tropical cyclones) Threshold values (zt, Tt) are determined independently for each model. They are determined in order that meridional distribution of annual TCGF agrees with the best track data.
Annual TCGF Models Obs. Unit: TCG number in 55 box in 10 years (1) (2) (3) (4) (5) (6) (7) (8) (9) (A) (B) (C) Models Obs. Unit: TCG number in 55 box in 10 years
RMSE of annual TCGF Low High Performance 1 2 3 4 5 6 7 8 9 A B C 3.5 3 2.5 2 1.5 1 0.5 Performance High 1 2 3 4 5 6 7 8 9 A B C High performance models: 2, 4, 5, 6, A Moderate : 1, 3, 8, 9, C Low : 7, B
High performance models (a) Model 1 (c) Model 3 (e) Model 5 (b) Model 2 (d) Model 4 Obs. 2 4 6 8 10 12 [(5x5)-1(10yrs)-1] These models reproduce maxima over the South China Sea and east of Philippines, and relatively low TCGF east of 160E.
Seasonal march in TCGF (110-150E) Model 1 Model 3 Model 5 July-September Black: Obs. Color: models FGA Model Obs. Model 2 Model 4 General winter-summer contrast is reproduced. 4 models overestimate TCGF in early summer (May-June). All models underestimate TCGF in mature summer (July-September), especially to the south of 15N. FGA
Genesis potential (GP) index Proposed by Emanuel & Nolan (2004). Calculate using monthly-mean data. Dynamic component Thermodynamic component Vorticity term Shear term PI term Humidity term h850: absolute vorticity @850hPa |u850-u200|: vertical wind shear Vpot: potential intensity (Bister & Emanuel 2002) H700: relative humidity @700hPa
Seasonal march in GP (110-150E) Model 1 Model 3 Model 5 normalized Model Obs. Model 2 Model 4 GP pattern represents TCGF quite well. Overestimation in May-June and/or underestimation in July-September.
Decomposition of GP FGA mean Model 1 Overestimate Model 3 Model 5 Underestimate Model 2 Model 4 AV: Vorticity term SH: Shear term PI : Potential intensity term RH: Humidity term Two dynamic components (AV & SH) are causes of seasonal bias for all of the 5 models.
Monsoon trough latitude May-June July-September z850>0 (trough) Obs. Pale blue tone & white contour: positive z850 Red vector: u850-u200 Black contour: |u850-u200| Mod. 3 A majority of TCs are generated over the monsoon trough. (MJ) Simulated trough is over the FGA, while the observed one is to its south. Overestimation of TCGF. (JAS) Simulated trough is located 15-25N, while the observed one is just over the FGA. Underestimation of TCGF (especially to the south of 15N).
Seasonal migration of monsoon trough z850 in 120-150E, positive only, tone: observation. Model 1 Model 3 Model 5 Model 2 Model 4 The trough migrates northward too fast and reaches too north.
Ridge of subtropical high May-June July-September Obs. Pale blue tone: positive z850 Black contour: SLP Red line: Observed ridge Black points: Simulated ridge Mod. 3 Simulated ridge of subtropical high over western Pacific is too north compared with observation.
Summary I evaluated performance of 12 CGCMs in simulating TCGF over the western North Pacific, using 20th-century climate simulations. 5 models reproduce annual TCGF distribution realistically. Although these 5 models reproduce general winter-summer contrast, they overestimate TCGF in early summer and underestimate it in mature summer. Diagnosis of environmental conditions using GP suggests that the biases are due to unrealistic seasonal migration of monsoon trough; it travels northward too fast and reaches too north.