Numerical investigation of summer sea breeze using updated urban aerodynamic parameters in Kanto Region

I. Introduction

Figure 1 Map of Kanto Region (Final Domain)

Various researches in the past for Kanto region, Japan, (refer to Fig. 1) understand the importance of wind circulation to urban heat island effect. During summer, heat stagnates in Saitama region due to the heat advected by sea breeze coming from Sagami Bay, Tokyo Bay, and the eastern coastline of Chiba and Ibaraki. This effect is intensified by the existence of sea breeze delaying cities like Kanagawa, Tokyo, and Chiba city. Aside from the direct effect to heating in urban areas, the urban/sea interaction occasionally generates summer thunderstorms. Thus, urban parameterization in numerical modelling is an on-going challenging for urban meteorologists.

In the past decades, the concept of displacement height, d, and roughness length for momentum, z0m, to represent urban cities in modelling have grown in significance and sophistication. Aside from the equations summarized by Grimmond and Oke (1999), a new LES-derived empirical formulation was developed by Kanda et al. (2013), which consider the effect of a calculation grid's maximum building height, and building height standard deviation.

This paper tackles on the incorporation of a more realistic d and z0m into the Single-layer Urban Canopy Model of the Weather Research and Forecasting (WRF) model, the model validation, and advances in the understanding of sea/urban aerodynamic interaction.

II. WRF Model Incorporation

           1-m. resolution mapcube datasets by CAD CENTER (www.cadcenter.co.jp) were acquired for the 23 wards of Tokyo. These were used to estimate the 1-km resolution d and z0m. For the whole of Kanto region, NASA ASTER Global Digital Elevation Map (DEM) at 30-m, which captures building heights at low-resolution when subtracted with Japan Geospatial Information Authority of Japan (GSI) DEM, were used. After corrections and estimates from previously acquired statistics between plane and frontal area indices, a distribution of d and z0m were acquired.

            d, which affects the starting point from the ground of the logarithmic wind profile, was added to the natural topography used in WRF's surface boundary. On the other hand, z0m was directly used by the UCM, replacing the default values called from the WRF urban parameters table. Large d and z0m values are noticeable along the coasts of Tokyo, Kanagawa, and Chiba with decreasing values towards in-land. In highly urbanized areas like Tokyo, values could exceed to up to ten times the default values for the urban canopy model.

III. Numerical Setting

            Three (3) simulation cases were conducted on September 14, 2011 (one day spin-up period) when a sea breeze flow was apparent (even from cloud images captured by a high resolution stationary satellite). The first case (CNTL) uses the default package of WRF-ARW version 3.3.1 where urban dominant grids use fix d (5.71m) and z0m (0.33m). The second case (SDLC) uses a modified WRF-ARW version 3.3.1 where urban fraction in the grid is considered in the UCM along with distributed d, z0m¸anthropogenic heat/vapor emission, and sky-view factor. The final case (VEGE) uses the default package of WRF-ARW version 3.3.1 with all urban areas replaced with grasslands (no UCM called). The output horizontal resolution is 4.8 km for the first domain and 1.2 km for the target domain. 28 vertical levels are outputted with the first atmospheric level set to an average of 30-m throughout.

            The effect of anthropogenic emission and sky-view factor indirectly affects the sea breeze flow field but significantly affects the energy/heat balance in the model. With this in mind, d and z0m remains to dominate the aerodynamics of sea breeze so the paper limits on the direct impacts of d and z0m.

            For the common boundary conditions and physics, high-resolution vegetation fraction and sea surface temperature were acquired from MODIS data. 3-hr Japan Meteorological Agency (JMA) Reanalysis data were used for meteorological boundary and 3-hr (interpolated from 6 hr) NCEP FNL data were used for additional surface boundaries.

I.V. Results

Figure 2 Selected Observation Points for Validation

Figure 3 Left: Calculated RMSE for wind speed. Right: Trend of simulated wind speed at To1

            A trial simulation was conducted for two months from August 1 to September 30, 2011 with last two days of July as spin-up period. Dry days were analysed comprising 681 hours (28 days). From Fig. 3, SDLC has least RMSE for all observation points even at points farther from the coastline with lesser roughness parameters. From the trend of wind speed at Tokyo (To1), it can clearly be observed that the CNTL and VEGE generally overestimates near surface wind speed (30-m. above ground). However, the simulation of wind speed at night is still underestimated by all cases.

V. Summary of Sea Breeze Simulation Findings

            On September 14, 2011, sea breeze starts penetrating from Sagami Bay and Tokyo Bay at around 1100 JST. For sea breeze penetration coming from Tokyo bay, VEGE simulated, as expected, the fastest sea breeze flow. CNTL then simulates a slightly delayed front with almost the same trailing wind speed. Until late in the afternoon, SDLC remains to have the least wind speed distribution near the surface with the sea breeze speeding up leeward from rough areas. This effect could be simulated up to an approximate height of 1000-m. at 1500 JST in Tokyo.

            The opposite could be observed for the sea breeze coming from Sagami bay. Due to the large roughness parameters and treatment of urban fraction, SDLC has larger temperature in-land leading to slightly earlier sea breeze advancement at areas with lesser z0m and d.

            SDLC simulated clouds represent well rapidscan stationary satellite images compared with VEGE (worst cloud simulated cloudline) and CNTL (slightly shifted north). This simulates a better stagnation region (Saitama) in the Kanto region.