In recent years, typhoon track forecast has been improved, though there has been no improvement in intensity forecast. Possible reasons of the problem include defects of forecast models and uncertainty in initial values used in forecasts. The kinetic energy source of typhoon is the latent heat of water vapor supplied from the ocean. Therefore, there is a strong relationship between sea surface temperature (SST) and maximum potential intensity of typhoon, and a small difference in SST may cause large differences in the maximum intensity. However, SSTs given as initial conditions in numerical models have some types of uncertainty.

Wind stress imposed on the upper ocean by a typhoon can limit the intensity primarily through wind-induced mixing of the upper ocean and subsequent cooling of the sea surface. Decreasing in SST due to the passage of a typhoon may be predicted by a vertical one-dimensional (1-D) process. However, there are few studies examining the ocean latent heat flux using a 1-D ocean model on real typhoons.

The purpose of this study is to clarify the sensitivity of typhoon intensity to initial SST values for Typhoon Mindulle (2021) using CReSS (Cloud Resolving Storm Simulator), a cloud-resolving model that incorporates a vertical 1-D ocean model.

One control experiment and two sets of sensitivity experiment were performed. The sensitivity experiment set I examined the sensitivity of SST bias in the initial SST values. Eight experiments were conducted with biases ranging from -2 to +2 °C. The results indicate that there is a clear negative correlation between the initial SST bias and lifetime-minimum central pressure (LMCP), and a positive correlation between the initial SST bias and the intensifying rate in the development period from the initial time to the time of the LMCP. These results confirm that the maximum intensity of typhoon strongly depends on the initial SST bias.

The sensitivity experiment set II examined the intensity sensitivity to perturbations in initial SST data. Five experiments were conducted with uniform random perturbations with amplitudes ranging from ±0.01 to ±2.0 °C. No correlation was found between the amplitude of the perturbations and the LMCP or the intensifying rate until the LMCP. This indicates that the forecast error of the typhoon intensity does not depend on the amplitude of random error of the initial SST values. In all experiments of sensitivity experiment set II, the central pressure was about 5 hPa higher than in the control experiment. Therefore, no matter how much the initial SST noise is reduced, there remains uncertainty of about 5 hPa.