Use of ensemble simulations to evaluate the urban effect on a localized heavy rainfall event in Tokyo, Japan

Recently, localized heavy rainfall over highly urbanized areas has caused severe damage in Japan. Although studies have indicated that the presence of urban areas can intensify rainfall, the chaotic noise caused by differences in initial conditions can change the urban effect. Therefore, the usability of ensemble simulation methods for urban effects on a single localized heavy rainfall event must be clarified to synthesize state-of-the-art observations and numerical model studies. This study examined the difference in a localized heavy rainfall event under different initial conditions. We demonstrated the advantage of ensemble experiments using a cloud-resolving model to evaluate the urban effect. In this study, we focused on a localized heavy rainfall event that occurred over Tokyo on 5 August 2008 and killed five people working in a drainpipe. Two ensemble experiments were performed according to the lagged average forecasting (LAF) method. Each experiment had six members, with initial conditions spanning 6 h from 02 to 07 Local Standard Time (LST) on 5 August 2008. The model was integrated continuously through 18 LST for each run. The model had a horizontal grid resolution of 2 km. The first experiment (CTRL) used the actual land cover, anthropogenic heat, and geometric structures. The second experiment (PDDY) used an imaginary land cover in which the urban area was changed to paddy fields. Comparison of the ensemble means showed that the peak value of the maximum hourly rainfall decreased and the peak time was delayed in PDDY compared with CTRL. The smaller peak value and delayed peak time of rainfall in PDDY seemed to be caused by the weaker updraft associated with near-surface convergence, which was caused by weaker surface heating. Regardless of the difference caused by the initial conditions, the decrease in the peak value and the delay in the peak time were significant according to the t-test. Although the area was limited compared with near-surface temperature and horizontal wind convergence, a significant reduction in rainfall was also seen over the heavily urbanized area of Tokyo in the horizontal distribution of the difference in accumulated rainfall. These results indicate that although differences in the initial condition can change the modeled urban effect on rainfall, ensemble simulation methods can be used to examine the significance of the urban effect on a single localized heavy rainfall event.