With half of the world’s population residing in cities, urban-atmospheric phenomena is becoming more intense. In Southeast Asia countries, capital cities are characterized by complex building morphology. Building morphologies, represented by aerodynamic surface roughness and displacement height, induce significant drag on winds within and above the urban canopy. Coupling of urban canopy models such as the single-layer urban canopy model (UCM) with the Weather Research and Forecasting Model (WRF) could adequately represent urban atmospheric phenomena. However, UCM still tends to overestimate near-surface wind speed and unable to reproduce the drag effect from buildings (Nelson et al., 2015) possibly due to the lack in accuracy of urban parameter distribution.
A new urban aerodynamic roughness parameterization (Kanda et al., 2013), integrated with UCM, was found to improve WRF-simulated near-surface wind speeds in built-up areas of Tokyo (Varquez et al., 2014). The concept behind the parameterization is that aerodynamic roughness parameters, displacement height (d) and roughness length for momentum (z0), can be estimated from building morphological parameters; plane area index (λp) (ratio of total building plane area to the total floor area), frontal area index (λf) (ratio of total building frontal area to the total floor area), average building height (Have), maximum building height (Hmax), and standard deviation of building heights (σH). Building morphological parameters are currently estimated from actual 3D building data, which unfortunately is unavailable for most cities such as that of developing countries. Hence, modelling a realistic urban atmospheric environment in developing cities especially in Southeast Asia is very challenging. In this paper, we propose a globally-applicable method for estimating the distribution of urban morphological parameters from satellite images. WRF simulations including the satellite-derived urban parameter distribution were conducted for Jakarta, Indonesia.
2. Estimation of 1-km λp
Landsat8 satellite images were used. Land use classification was conducted at Jakarta, Tokyo, and Istanbul to produce land use distribution in 30-m resolution using supervised classification and later assembled in 1-km resolution grid. Using ratio of occupied area of each urban class to total area in 1-km grid, we defined 3 variables or urban ratio: rU_med, rU_hi, and rU_ind which denote medium density, high density, and commercial urban ratio. Ordinary Least Squares (OLS) method was utilized to determine fitting function between urban ratios and real λp value for 3 megacities: Tokyo, Istanbul, and Jakarta. Here, urban ratios were used as 3 independent variables and real λp value as dependent variables. Fitting function (Equation 1) created by OLS gives high linear similarity with >0.86 Pearson R correlation value for three cities (Figure 1). λf later calculated from Kanda et al. 2013 (Equation 2).
λp_predict = α1(rU_med) + α2(rU_hi) + α3(rU_ind) (Equation 1)
where: α1 = 0.01~0.03, α2 = 0.4~0.5, α3 = 0.1~0.13
λf_predict = 1.42(λp_predict)2 + 0.4(λp_predict) (Equation 2)
3. Estimation of 1-km Have, σH, and Hmax
To predict 1-km Have, 3 satellite images were used: 30-m ASTER GDEM, 7.5 arc-second GMTED2010 – and calibrated nightlight images – 1-km Nightlight Defense Meteorological Satellite Program (DMSP) satellite image. Using these images, fitting analysis using feedforward artificial neural network (ANN) were conducted for 2 different major cities in Japan, Tokyo and Nagoya. This method requires subtraction of maximum value of ASTER GDEM –considered as building-included topography- with mean value of GMTED2010 –considered as actual land topography- on the corresponding 1-km grid, these data called ASTER-GMTED. Two ANN function were created for each city: ASTER-GMTED is used as first input neuron and nightlight as second input neuron. The output neuron is each city’s real Have obtained from PASCO Mapcube data. Inter-comparison using Tokyo ANN function for Nagoya and vice versa were done to define ANN general function. It showed that Nagoya ANN function works well for both Tokyo and Nagoya (Figure 2) compared with Tokyo ANN function. Hence, Nagoya ANN function were applied to define in Jakarta. σH and Hmax were calculated using Kanda et al. 2013 equation (2) and (3).
4. Estimation of d and z0
The last two aerodynamic parameters d and z0 were estimated following Kanda et al. 2013.
5. WRF simulation using satellite-derived urban parameters in Jakarta Megacity
Two simulation cases were conducted to analyze the influence of urban parameterization on Jakarta urban climate. The first case utilized the satellite-derived distributed urban parameters (SAT), and the second utilized uniformly-assumed urban parameters from the default UCM (DEF). The completed urban morphology and aerodynamic parameters were processed as WRF geographical boundary. Two meteorological station in central Jakarta (KMY) and Jakarta bay (TPR) were used as validation.
Results show that DEF and SAT have similar performance in predicting diurnal temperature variability in Jakarta and both of them are very close to observation value. However, SAT defines better wind speed than DEF as depicted by the lower rmse and bias values for all stations in SAT. SAT has wind speed rmse value of 1.62 and 1.40 for KMY and TPR station respectively, while DEF results rmse value of 2.87 and 2.61 for KMY and TPR respectively for nighttime when the peak of UHI happened. SAT was better in resolving wind speed due to an improved representation of displacement height and roughness length especially at densely built-up areas (Figure 3).
6. Conclusion
The WRF simulation results showed improved accuracy on simulated near-surface winds when the distributed urban parameters estimated from globally-available datasets were used. Thus, satellite-derived urban parameters can be a promising substitute to estimating distributed urban parameters during the absence of real building morphological information.
References
Kanda, M., Inagaki, A., Miyamoto, T., Gryschka, M. and Raasch, S. (2013) ‘A New Aerodynamic Parametrization for Real Urban Surfaces’, Boundary-Layer Meteorology, 148(2), pp. 357–377.
Nelson, M.A., Brown, M.J., Halverson, S.A., Bieringer, P.E., Annunzio, A., Bieberbach, G. and Meech, S. (2015) ‘A case study of the weather research and forecasting model applied to the joint urban 2003 Tracer field experiment. Part 1: Wind and turbulence’, Boundary-Layer Meteorology, 158(2), pp. 285–309.
Varquez, A.C.G., Nakayoshi, M. and Kanda, M. (2014) ‘The Effects of Highly Detailed Urban Roughness Parameters on a Sea-Breeze Numerical Simulation’, Boundary-Layer Meteorology, 154(3), pp. 449–469.
Acknowledgement
This research was supported by the Environment Research and Technology Development Fund (S-14) of the Ministry of the Environment, Japan.
![](images/flag.png)
![](images/blue_ribbon.png)