Numerical Experiment

The empirical results presented above reveal the significant effect on the amplitude and variance of the ISO. This section presents numerical simulations performed to see whether similar effects could be realized in the numerical models. Two models were used in this study. The first was the Purdue regional model (PRM), which had been used for the study of various mesoscale phenomena (e.g. Sun et al., 1991; Hsu and Sun, 1994; Sun and Chern, 1993) and climate simulation (e.g. Bosilovich and Sun, 1998, 1999; Hsu et al., 2004; Sun et al., 2004). The second one was the National Taiwan University's general circulation model (NTUGCM; Hsu et al., 2001).

The resolution for the PRM was 60 km in the horizontal and 28 levels in the vertical. The model domain is 90°E-160°E and 0-40°N, with a 15-point buffer zone on the lateral boundary. The ECMWF advanced analysis at 1.125° resolution was used as the initial condition and the lateral boundary condition. The latter was updated every 6 hours. The resolution for the NTUGCM was T42 in the horizontal and 13 levels in the vertical. The ECMWF basic analysis at 2.5° resolution was used as the initial condition.

Both models show a certain degree of ability to simulate TCs in the first few days. One example for the NTUGCM is shown in Fig. 11 for the forecast with initial condition at 00Z

16 June. The 24-hour hindcast [Fig. 11(a)] was able to simulate the location of TCs, although with a much weaker amplitude. The TC vortex looks larger and smoother because of the low resolution. In the 48-hour and 72-hour hind-casts [Figs. 11(b) and 11(c)], the simulated TC-like vortex is still evident, but lags behind the observed TC vortex. The 48-hour and 72-hour hindcasts generally show large track bias and smaller amplitudes.

It appears that both models are able to produce satisfactory simulation in the 24-hour hindcast. To evaluate the TCs' effect on the simulated ISV, different numerical experiments were performed. A series of 24-hour hindcasts were performed daily, using both models, from 1 June to 31 October 2004. They yielded a dataset of the control experiment in JJASO, which was used for diagnostics like the real data. Three series of the NTUGCM experiment were performed. The first is a series of 24-hour hindcasts starting with the observed initial conditions. The second is the same as the first, except that TCs have been removed from the initial condition. The third is a 24-hour hindcast experiment, starting with the observed initial condition, in which the vor-ticity in the lower troposphere was artificially enhanced where the TCs are located. The vor-ticity was enhanced at the grid points, where the surface pressure was lower than 980 hPa and the 900hPa/850hPa/700hPa mean vor-ticity was larger than 0.00005 s-1. The vorticity at 900hPa/850hPa/700hPa was multiplied by

Figure 12. Variance of the 32—76-day filtered 850 hPa vorticity for (a) the observed and the (b) control, (c) TC-removed, and (d) vorticity-enhanced simulations using the NTUGCM. The contour interval is 0.5 X 10_10 s~2.

1.15/1.01875/1.00555, respectively. These three experiment series were performed daily from 1 June to 31 October, and the 32-76-day bandpass filter was applied to the three simulation datasets. Both the simulated and the observed ISV of the 850 hPa vorticity are shown in Fig. 12. The observed ISV, which is plotted at the NTUGCM spatial resolution, looks much smoother because of the low spatial resolution [2.5° by 2.5°; Fig. 12(a)]. The control experiment [Fig. 12(b)] was able to reproduce a realistic spatial distribution of the ISV with a slightly weaker magnitude. In contrast, the TC-removed experiments [Fig. 12(c)] failed completely to simulate the maximum variance along the TC tracks. The vortex-enhanced experiments [Fig. 12(d)], on the contrary, not only simulated well the ISV distribution but also enhanced the magnitude to the observed level. A comparison between the results obtained from the three experiment series indicates that the presence and enhancement of TC-like disturbance, although fluctuating at a much higher frequency, enhance the ISV.

A series of heating-enhanced hindcast experiments using the PRM were performed, by adding a prescribed heating profile to the simulated TC-like vortices. The heating profile mimics the Q1 profile (not shown), which was calculated from the reanalysis at those grid points where category-4 typhoons passed. The profile was characterized by the maximum heating at 600 hPa, and the decreasing linearly upward and downward. However, the heating below 800 hPa was kept constant, to obtain better simulation. During the simulation, the heating was multiplied by 2 and held constant at the center of TC-like vortices, and exponentially decreased outward for 10 grid points. Since this was an idealized experiment, the prescribed heating is the same for all cases at all times, disregarding the different sizes and strengths of the vortices. The purpose was to artificially and significantly enhance the strength of the TC-like vortices in the model and, through a comparison with the controlled experiments, to evaluate whether the TCs can enhance the ISV in the simulation.

The 32-76-day band-pass filter was applied to the 850 hPa vorticity of the control and heating-enhanced hindcast experiments to extract the intraseasonal fluctuation. The control experiment was able to reproduce the overall distribution of the ISV, but the values are only about 1/3 of the observed variance. This is because the simulated TC-like vortices are weaker than the observed, and tend to weaken quickly. The simulated variance in the heating-enhanced experiments is raised to the equivalent level of the observed magnitude, while maintaining a realistic spatial distribution. This contrast indicates that the enhanced TC-like disturbance in the regional model also contributed to enhancing the ISV. The results of numerical experiments confirm the hypothesis, which was proposed based on the empirical results, that the presence of TCs in clusters can enhance the ISV.

0 0

Post a comment