The impact of ENSO on tropical cyclone activity over western North Pacific using very high resolution AGCM

Nowadays, the hazardous impact of tropical cyclones to the human beings increases rapidly. For the preparation to the devastating disaster, the accurate seasonal prediction of tropical cyclones requires using global model. Even though there was an exceptional study using 20km mesh global model (Oouchi 2005), until now, the simulation of tropical cyclone has been usually produced by low resolution global model (Bengtsson et al 1995, Camargo and Sobel 2004) ; more or less T42 (about 300km). Because the intense wind speed and precipitation occurs in a very small area, the low resolution is inimical to simulate realistic tropical cyclone. Therefore very fine grid is essential component of tropical cyclone simulation.

For simulating realistic tropical cyclones, we used very high resolution version of SNUAGCM (Kim et al., 1998). The horizontal resolution is 1536*768 grid points and its horizontal mean mesh size is about 22.66 km at 30° degree. SNUAGCM adopts three-dimensional hydrostatic primitive equations on sphere with normalized pressure coordinate in the finite volume dynamical frame. And we use simplified Arakawa-Schubert cumulus convection scheme, large-scale condensation scheme based on Letreut and Li(1991), Bonan's Land Surface Model, non-local PBL/vertical diffusion (Holtslag and Boville 1993), and modified CCM3 slab ocean/sea-ice model. It was mainly developed for the seasonal prediction and its standard resolution is T42L20 for spectral version and about 300km for finite volume version.

We performed the integrations forced by observational monthly mean sea surface temperatures (OISST, Reynolds et. al. 2002) for the period from 15 April to 31 October at each 1997 El Nino/1999 La Nina and each case has 6 ensembles. The reason why we chose these cases is that 1997 El Nino and 1999 La Nina is a recordable strong ENSO period so these periods are more conducive to draw the impact on tropical cyclone activity forced by SST anomaly relative to the other effects, e.g., QBO and MJO. For inspecting the time evolutionary detail features of tropical cyclones, the frequency of output is set to 6 hour and it is a snap shot at every time intervals. Targeted area is western North Pacific basin.

We used 850hPa relative vorticity, surface or 10m wind speed, the temperatures at 700, 500, 300hPa, and sea level pressure for detection of tropical cyclones. After detected by these variables satisfying a certain criteria simultaneously, the all points are aligned and we got the tracks of tropical cyclones. We analyzed the features of TC by three aspects; (1) structure, (2) genesis, and (3) tracks.

SLP reveals closed concentric rings around the center of the minimum and the gradient toward the center increases as close to the axis. The most intense TC among the all simulated storms has minimum SLP of 946.44hPa. This value is close to real TC of the category 3~4, and that gives an implication that high resolution model can reproduce the intensity close to the real TC. The precipitation is also concentrated in the axis and the azimuthal distribution which was mentioned by Frank (1977) is similarly revealed. We also examine the vertical structure of wind, temperature and moisture anomalies in order to inspect the thermodynamic characteristics of TC in mature stage. The warm core, radial and tangential wind structures in vertical are well-simulated in the view of magnitude and spatial distribution.

As a result of analyzing statistics of all simulated TCs, the mean wind speed and SLP of El Nino year, 1997, are stronger than the La Nina year, 1999, and this result retains a great similarity with the previous observational studies in western North Pacific (Wang and Chan 2002). Although the magnitudes of these variables are weaker than the observation at all ensembles, it can be said that the different characteristics between El Nino and La Nina is somewhat well-simulated. Furthermore, ensemble mean numbers of TCs at each year are very much like that of RSMC (Regional Specialized Meteorological Center, Tokyo Tropical Cyclone Center) observation.

The locations of all TCs cover each location of the observed, and the mean location changes between strong El Nino and La Nina argued by Wang and Chan (2002) are well-simulated. On the other hand, model can not reproduce well far east (central Pacific) and far west (south China sea) TCs from the mean location. And the mean locations of simulated TCs are shifted to the eastward and slightly southward relative to the observed mean. This difference from the observation is inferred to come from the model bias. The model simulates the location of large vorticity areas, in which TCs are generated by convergence of trade winds blown form both hemispheres, toward eastward than observation. This bias makes the mean location of simulated genesis to be eastward relative to the observation.

The simulated tracks are also similar to the observation best tracks. A fair number of tracks of the simulated TCs is confined to the southward of 30°N not to be extended to the mid-latitude. But the recurvate tracks during 1997 are revealed more clearly than during 1999 and the tracks of 1999 which to lean toward westward of 140°E relative to 1997 are also clearly represented. By the analysis of the tropical cyclone passage frequency (TPF), the difference between 1997 and 1999 is confirmed distinctly.

We also examined the integrated intensity using accumulated cyclone energy (ACE). ACE includes all of the lifetime, intensity (commonly defined value by maximum sustained wind speed), and frequency (total number of genesis), so it is a good measurement of TC activity as a whole in a certain period. As a matter of course, 1997 El Nino case in which the frequency is larger and the lifetime is longer than 1999 La Nina, has larger intensity.