Geoscience Reference
In-Depth Information
3. Simulations of Northern Indian Ocean Tropical
Cyclone Activity
We are exploring several methods to simulate TC activity in the Indian Ocean
region using dynamical models. Our starting point for the results described
here is a global atmospheric model forced by observed time-varying SSTs,
which is used to simulate and assess the frequency of tropical cyclones. Zhao
et al. (2009) showed that by using such a model with a horizontal grid scale of
50 km, the frequency and interannual variability of tropical cyclones with
intensities above a threshold of 33 m/s are simulated quite realistically, especially
in the Atlantic basin, but also in the Northwest Pacific basin. However, as seen
in Fig. 5, the simulation for the North Indian Ocean is not as close to observed
as in the Atlantic, with particular too many storms simulated with winds
exceeding 33 m/s in the Arabian Sea and too few in the Bay of Bengal.
As a second downscaling step, we take each of the individual tropical
cyclones (winds exceeding 17 m/s) from the Zhao et al. (2009) simulation and
downscale those storm cases into the GFDL hurricane model using the models
and approach described in Bender et al. (2010). (Actually we use here a slightly
modified version of the Zhao et al. global model that has quite similar simulation
characteristics to the original one in Zhao et al. 2009). This allows us to examine
the simulated vs. observed distribution of more intense tropical cyclones in the
basin using the downscaling framework. The GFDL hurricane model has a
grid-spacing over five times as small as in the original global model (as fine as
9 km) and also includes dynamical ocean coupling to simulate the potentially
significant impact of cold waves beneath intense tropical cyclones.
The global model simulates too many tropical cyclones over the Arabian
Sea as compared to the observations (Fig. 5) which is reflected in the distribution
of tropical storms (winds exceeding 17 m/s) from our downscaling runs. In
addition, the occurrence of storms exceeding 33 m/s (Fig. 6) is excessive in
both the Arabian Sea and Bay of Bengal. There are almost four times as many
simulated as observed storms of this class. These biases in tropical storm counts
in the global model carry through to the hurricane model. That is, the over-
estimate of storm frequency in the hurricane model simulations is largely pre-
determined by the over-estimation of such storms in the global model.
Finally, Fig. 7 focusses on even more intense North Indian Ocean tropical
cyclones storms, with maximum surface winds exceeding a threshold value of
50 m/s. The model simulates about 60% more of such storms than observed,
with a more pronounced bias in the Arabian Sea region.
Figure 8 shows the observed and modelled intensity distribution of the
storms that were downscaled into the GFDL hurricane model for the North
Indian Ocean basin. The simulation shows a pronounced “peaked distribution”
bias where there are too few weak and strong storms and too many storms in
the middle of the distribution (35-55 m/s). The causes of this overly peaked
Search WWH ::
Custom Search