Investigation of the uncertainty of initial and boundary conditions in the hindcasts of flow fields over urban areas using a mesoscale numerical weather prediction model

Numerous studies have been conducted using mesoscale numerical weather prediction models to analyze the thermal and wind environments of urban areas around the world. However, weather predictions are highly sensitive to initial conditions. This is the so-called "butterfly effect." In this study, the effects of initial and boundary conditions upon the reproduction accuracy for wind environments (particularly for the sea breeze) over the Fukuoka metropolitan area were investigated by varying (i) the objective analysis data used for the initial and boundary conditions and (ii) the length of the spin-up calculation period. The results of the simulation were compared with surface observations and Doppler LIDAR observations of the airflow field. The accuracy is thought to decrease under an increase in the length of the spin-up calculation period; however, this was not the case in this study. In terms of the differences in the objective analysis data, no clear trend was observed from the comparisons with surface observations; however, slight differences were observed when comparing with the Doppler LIDAR. The lagged average forecasting method (which reduces the uncertainty of the initial conditions by shifting the calculation start time) was also adopted; the ensemble-averaged results showed a good correspondence with the observations.