April 8, 2010 Meeting Summary
In-Hyuk Kwon presented his work titled "Tropical Cyclone Initialization with Spherical High-Order Filter and Idealized Three-Dimensional Bogus Vortex." In his introduction, In-Hyuk mentioned that the main purpose of this presentation was to introduce a new tropical cyclone (TC) initialization method and evaluate this method using the WRF model. The new TC initialization method used a similar strategy to that in the GFDL model's hurricane initialization with some differences. While GFDL replaces a poorly resolved TC in the analysis with a more realistic bogus vortex, the method used here generates a 3-dimensional (3-D) bogus vortex using empirical functions instead of a spin-up vortex. In-Hyuk's method also separates the TC component from the analysis field using a spherical high-order filter with a cut-off scale that can be varied based on the TC scale. In-Hyuk also used the final analysis from GDAS (FNL) for the global analysis and interpolated it to high-resolution (18km grid resolution). The TC-initialized global data was generated from the interpolated global analysis and the 3-D bogus vortex.
Next, In-Hyuk explained the TC initialization method by first presenting some information on the structure of an ideal 3-D bogus vortex. Based on work from Holland (1984) and Fujita (1952), it was determined that surface wind and pressure associated with the TC structure are well-represented with empirical formulas. In-Hyuk's work wanted to use information from the TC to define empirical formulas for a 3-D bogus vortex. In-Hyuk noted that, based on work by Frank (1977), in a hurricane, the geopotential height at lower altitudes has a bigger negative geopotential deviation. Also, the maximum positive deviation occurs around 150 hPa, and the geopotential height deviation gradually decreases above 150 hPa. In-Hyuk then explained that an empirical function can be used to calculate the geopotential deviation between the surface pressure and the pressure at 150 hPa. The empirical function used has two forms: one for when the pressure is less than or equal to pressure at the level of maximum positive geopotential difference (about 150 hPa) and another for when the pressure is greater than that at the level of maximum positive geopotential difference. In the function, S is the variable used to determine the temperature deviation at the surface pressure while tau is the parameter used to determine the shape of the vertical profile. In-Hyuk then detailed the procedure for generating a bogus vortex. The bogus vortex consists of four variables: geopotential deviation, temperature deviation, tangential wind, and radial wind. Beginning with the geopotential deviation, the temperature deviation and gradient wind are found and from the gradient wind, the tangential wind and radial wind are found. In-Hyuk then showed the geopotential deviation (upper left - a), temperature deviation (upper right - b), tangential wind (lower left - c), and radial wind (lower right - d) for Typhoon Nari at 00UTC on September 14, 2007. The geopotential deviation (a) was negative in the center of the storm, the temperature deviation (b) indicated a warm core near 300 hPa, the tangential wind (c) showed a reduction near the PBL with cyclonic to anticyclonic circulation in the upper troposphere, and the radial wind (d) showed lower level convergence and upper level divergence. In-Hyuk then illustrated the structure-adjustable vortex by presenting plots of sigma for small and strong Typhoon Nari, larger Typhoon Wipha, and weaker storm Kajiki. In-Hyuk also noted that the vertical velocity was dependent on the storm structure.
Then, In-Hyuk discussed the TC initialization with a 3-D bogus vortex. Using a horizontal filter, the analysis is separated into the large scale (basic) and small scale (disturbance) with the tropical cyclone being considered the vortex on the environmental flow. The horizontal filter response can be defined as how much amplitude is applied to a specific scale. Looking at Kurihara's filter response values, the larger cut-off scale (Mc = 18) is greatly diminished and near zero. For his work, In-Hyuk used a spherical high-order filter from Cheong et al. (2002, 2004) and the same cut-off scale was used for each level of a certain storm.
In-Hyuk next explained the procedure for TC initialization with an ideal 3-D bogus vortex. Beginning with the high resolution global analyses and the bogus vortex, the analyses is split into basic and disturbance scales using a horizontal filter with a TC scale. Through horizontal filtering, the bogus vortex is separated into the vortex disturbance. Next, a streamfunction is used to determine the TC domain from the disturbance part of the analysis. The vortex disturbance is merged with the separate parts of the analysis through a matching function used to provide a smooth transition from the vortex disturbance to the analysis disturbance with an increasing radius. This provides the TC-initialized data and the relocation of the TC center. In-Hyuk noted that for his work, only small-scale features were modified. Next, an example of TC-initialized global data was presented. Before TC initialization, the FNL data cannot resolve Typhoon Nari (right image), however, after TC initialization, Nari is intensified with no discontinuity. An example showing an animation of the cloud and rain water and trajectory of Typhoon Mindulle was also presented.
In-Hyuk next presented results on the effect of TC initialization on typhoon prediction, first for Typhoon Nari. The WRF (ARW v18.104.22.168) model was used for evaluation of the TC-initialization effect with a 60-hr simulation. Nari was a small typhoon with strong intensity that rapidly developed and then decayed and had large track and intensity errors from the KMA (Korean Meteorological Administration) and RSMC (Regional Specialised Meteorological Center). In the results plots shown, CTRL indicates the model test without TC initialization (plots on left) and BG indicates the model test with initialization (plots on right). The FNL analysis was used as a boundary condition for this work. In-Hyuk pointed out that at the initial time, the BG geopotential height deviation (a) has a larger amplitude than that for CTRL. The temperature deviation for BG is also more concentrated in the central storm region for BG than for CTRL. The BG meridional wind (c) is intensified while that for the CTRL is weak while the zonal wind (d) for BG is more realistic and that for the CTRL is abnormally large. For the 12-hr simulation, the structure for geopotential height and temperature deviation is similar to the initial state with an amplitude increase, and the zonal and meridional winds for BG have also increased. Looking at the track and intensity prediction, where BG values are in red and CTRL values are in green, the BG track and intensity are much closer to best track values (in black). Next, In-Hyuk presented the track errors for the years 2005, 2006, and 2007 comparing RSMC (in blue) to BG. Clearly, BG track errors were lower than those for RSMC, however, In-Hyuk noted that this was not a fair comparison because BG used the analysis as boundary conditions. The intensity error plots showed similar results to those for track errors comparing RSMC to BG.
In-Hyuk concluded by presenting his summary. He mentioned that this new TC initialization method is developed and intended to improve TC track and intensity prediction with TC-initialized high-resolution global data generated and evaluated using the WRF model. The 3-D bogus vortex is carefully designed and simple to generate with input from tcvital data, ambient pressure, and mean surface temperature. In-Hyuk mentioned that the weak point in this study was that vertical tilt for storms was not taken into account, and this is more of a problem for weak TC cases. Applying this work to hurricane relocation involves using a first guess vortex instead of a bogus vortex and the TC component would be extracted from the background and moved to the correct position.