You are on page 1of 14

This article was downloaded by: [Marcel Ignatius] On: 25 September 2012, At: 23:25 Publisher: Taylor &

Francis Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

International Journal of Sustainable Building Technology and Urban Development


Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/tsub20

Comparison of STEVE and ENVI-met as temperature prediction models for Singapore context
Wong Nyuk Hien , Marcel Ignatius , Anseina Eliza , Steve Kardinal Jusuf & Rosita Samsudin
a b a a a b

Department of Building, National University of Singapore, 4 Architecture Drive, Singapore, 117566


b

Centre for Sustainable Asian Cities, National University of Singapore, 4 Architecture Drive, Singapore, 117566

To cite this article: Wong Nyuk Hien, Marcel Ignatius, Anseina Eliza, Steve Kardinal Jusuf & Rosita Samsudin (2012): Comparison of STEVE and ENVI-met as temperature prediction models for Singapore context, International Journal of Sustainable Building Technology and Urban Development, 3:3, 197-209 To link to this article: http://dx.doi.org/10.1080/2093761X.2012.720224

PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.

International Journal of Sustainable Building Technology and Urban Development Vol. 3, No. 3, September 2012, 197209

Technical Paper Comparison of STEVE and ENVI-met as temperature prediction models for Singapore context
Wong Nyuk Hiena*, Marcel Ignatiusa, Anseina Elizaa, Steve Kardinal Jusufb and Rosita Samsudinb
a

Department of Building, National University of Singapore, 4 Architecture Drive, Singapore 117566; bCentre for Sustainable Asian Cities, National University of Singapore, 4 Architecture Drive, Singapore 117566 (Received 21 August 2012) In urban areas, natural land soil has been replaced by asphalt roads and concrete buildings, which absorb and retain more heat during the day, creating the Urban Heat Island (UHI) phenomenon. Current studies show that UHI impact mitigation strategies are to increase the open spaces to allow urban ventilation and plant green cover. To complement this, a temperature prediction model could be effective for simulating and quantifying the temperature reduction for every developed strategy. This paper will look into two prediction methods: STEVE and ENVI-met. Screening Tool for Estate Environment Evaluation (STEVE) is a prediction tool which is able to calculate the Tmin, Tavg and Tmax of the point of interest for certain urban settings. The temperature at that particular point is the result of its surrounding environment within the buffer zone. Output data from STEVE will be used as a database for a Geographic Information System (GIS) to produce temperature maps. ENVI-met is a Computational Fluid Dynamics (CFD) based micro-climate and local air quality model. It calculates temperature within the interval times for 24 to 48 hours. The calculation is based on the grid (x,y) with a specied grid distance. This resolution allows analysis of small-scale interactions between individual buildings, surfaces and plants. The major differences between the models are the wind-speed variable, raster map, surface temperature and the local climate context. STEVE calculation focuses on typical calm day conditions which excludes the wind speed variable, while ENVI-met consider it as one of the parameters. The GIS raster map generated from STEVE predicted temperature is based on a buffer zone with specied diameter, while ENVI-met is based on grid pixels or cells which produces temperature maps in more detail resolution. The objective of this study is to compare both prediction models so as to understand their benets and limitations, in order to justify which model is more appropriate for a tropical urban context, and in this case Singapore. Keywords: prediction tool; ENVI-met; STEVE; urban heat island; tropical climate

Downloaded by [Marcel Ignatius] at 23:25 25 September 2012

1. Introduction The Urban Heat Island (UHI) phenomenon is a condition in which temperatures in urban areas are higher than their surroundings. This is as a result of urbanisation which increases anthropogenic heat emissions in urban centres, while water, wind and greenery that can help to cool down an urban area have concurrently decreased. Also, due to a greater area of asphalt roads and concrete buildings, more heat is absorbed and retained, while the heat reection ratio decreases. A temperature prediction model could be effective for simulating and quantifying the temperature reduction for any strategies or mitigation methods proposed. The amount of time consumed in order to measure local urban temperature by mean of simulation has been one of the main obstacles in performing urban micro-climate analysis, while the expertise of the simulation software used can also be considered as another factor. On the other hand, Screening Tool for Estate Environment Evaluation (STEVE) is offering an alternative method to predict the local urban temperature of a certain tropical urban area (STEVE was developed and built based on eld measurement data in the

Singapore context). This paper focuses on comparing two different applications for assessing temperature prediction on estate level: STEVE and ENVI-met. STEVE, developed by Dr. Steve Kardinal Jusuf from the National University of Singapore, is an application that calculates the minimum temperature (Tmin), average temperature (Tavg), and maximum temperature (Tmax) of a certain point of interest for either the existing or future condition (proposed master plan) of the estate or some specic urban area [1]. The temperature predicted at that particular point is as a result of the surrounding environment within the buffer zone. Currently, the STEVE application serves as a plug-in or tool which is used under the Geographic Information System (GIS) platform named ArcGIS. In the end, for the purpose of a more graphical urban analysis, data calculated from STEVE is used within ArcGIS to produce urban temperature maps. ENVI-met is a Computational Fluid Dynamics (CFD)based micro-climate and local air quality model [2]. The software, developed by Prof. M. Bruse from the University of Mainz and his team, is a three-dimensional

*Corresponding author. Email: marcel.ignatius@nus.edu.sg


ISSN 2093-761X print/ISSN 2093-7628 online q 2012 Taylor & Francis http://dx.doi.org/10.1080/2093761X.2012.720224 http://www.tandfonline.com

198

W.N. Hien et al. UHI can also be observed in every town and city as the most obvious climatic manifestation of urbanisation [7]. In Japan, Tokyos average temperature has increased by about 38C, and that of Osaka has increased by 28C over the past 100 years. Cumulatively, since it has been reported that global warming has raised Japans average temperature by about 18C, the temperature increase due to the UHI effect is probably about 28C degrees in Tokyo and about 18C in Osaka [8]. In Washington and Tokyo, a rate of onehalf to one degree Fahrenheit increase every 10 years was observed in the context of the summers maximum temperatures during the last 30 80 years [4]. Contextually, there were some studies conducted to investigate the Singapore UHI, as early as 1964, including one conducted by Nieuwolt [9] where it was found that there was 3.58C temperature difference between the city area and airport area. The city rural temperature differences phenomena was believed to be due to the greater absorption of solar radiation and to reduced evapotranspiration in the city. In a following study by Chia [10], the author considered the variations in cloud pattern effects on the micro-climate, where she found out that a combination of low solar radiation receipts and low wind speed together with a low cloud ceiling reduced the city rural temperature and relative humidity differences. On a more advance approach, Nichol [11] presented UHI in Singapore through the remote sensing technology, where it recorded roughly over 48C difference was observed from the satellite image of Singapore. The observation of UHI in Singapore provides evidence that the local buildings do have a great impact on the local climate. Diurnal UHI dynamics exploration was conducted by Roth [12], in the spring and summer of year 2001, and showed night time heat island magnitudes of up to 48C with the lowest temperatures observed in densely vegetated areas. A more recent study by Wong and Chen [13] investigated the severity and impact of UHI on environmental conditions and identied the possible causes by using thermal satellite imaging and mobile survey. In Singapore, the satellite image shows the UHI effect during the daytime. The hot spots are normally observed on exposed hard surfaces in the urban context, such as industrial areas, the airport and Central Business District (CBD) areas. The satellite image also shows some cool spots, which are mostly observed on the large parks, the landscape in-between the housing estates and the catchment area. 2.2 Mitigating UHI Urban morphology is a major factor that inuences the thermal behaviour in the city and creates UHI impact. Oke stated that designs for street canyons in high- and mid-latitude cities should (1) maximise shelter from wind for pedestrian comfort, (2) maximise dispersion of air pollutants, (3) minimise urban warming to reduce the need for space cooling, and (4) maximise solar access [14]. Okes analysis of

Downloaded by [Marcel Ignatius] at 23:25 25 September 2012

non-hydrostatic micro-climate model including a simple one dimensional soil model, a radiative transfer model and a vegetation model [3]. Technically, ENVI-met was designed to simulate the surface-plant-air interactions within the limited grid cells size of a certain urban environment, with a typical resolution of 0.5 to 10 m in space and 10 seconds in time. For this study, ENVI-met was deployed to calculate temperature within the interval times for 24 to 48 hours. The calculation is based on the grid (x,y) with specied grid distance. This resolution allows analysis of small-scale interactions between individual buildings, surfaces and plants. The major differences between STEVE and ENVI-met are wind-speed variable consideration, raster map output, surface temperature and the local climate context. The STEVE calculation focuses on calm day conditions which excludes the wind speed variable, while ENVI-met includes it as one of the calculation variables. The GIS raster map that is produced from the STEVE model is based on buffer zone with specied diameter (in this case 50 metre radius), while ENVI-met is based on pixels formed by the grid which will give more detailed resolution. ENVImet is able to provide surface temperature prediction at different heights while the STEVE calculation is based on a pedestrian height of 2 metres. Hypothetically, STEVE, which was developed based on 3 years local data measurement, is deemed to be more appropriate and accurate for tropical urban areas, compared to ENVI-met. The objective of this study is to give a comparison of both prediction models to understand their benets and limitations. By all means, the authors do not intend to judge if one approach is better than the other. The study is more of a justication for selecting a proper and efcient method for analysing tropical urban micro-climate (in this case, from the Singapore context). 2. Literature review

2.1 Urban heat island UHI phenomenon is a condition in which temperatures in urban areas are higher than their surroundings. Isotherms on a map show an urban area with higher temperatures emerging like a warm island oating on a cooler sea. This is the origin of the name of Urban Heat Island. Alterations of the urban surface by people results in diverse micro-climates whose aggregate effect is reected by the heat island. Building, paving, vegetation, and other physical elements of the urban fabric are the active thermal interfaces between the atmosphere and land surface. High-building density and change of street surface may lead to overheating by human energy release and absorption of solar radiation on dark surfaces and buildings [4]. The composition and structure within the urban canopy layer, which extends from the ground to about roof level, largely determines the thermal behaviour of different sites within a city [5,6].

International Journal of Sustainable Building Technology and Urban Development urban geometry in relation to air-ow patterns, radiation exchange, and thermal behaviour resulted in a range of canyon geometries and building densities that achieve a zone of compatibility. Another study done by Arneld revealed that calculation of irradiances on urban canyon walls and oors provide quantitative guidelines on street geometry [15]. He found that irradiance values generally were smaller for canyon walls than for oors, and that controlling canyon-oor irradiance was more critical at lower latitudes because of higher solar angles. North-South street orientation provided less summer and more winter canyonoor irradiance than east-west street orientation. Using the Cluster Thermal Time Constant (CTTC) model, Swain and Hoffman carried out a study where in hot regions, cities with North-South streets have weaker urban heat-island intensities (0.88C) than cities with East-West streets [16]. Their sensitivity analysis illustrated how increasing aspect ratios (canyon height to width) results in a stronger daytime cool island due to increased shade, and a weaker nocturnal heat island due to appreciably reduced maximum temperature. 2.3 Screening Tool for Estate Environment Evaluation (STEVE) STEVE has been developed based on the air temperature prediction models [1]. These prediction models were based on the empirical data collected over a period of close to 3 years as part of the development of an assessment method to evaluate the impact of estate development, which includes the assessment method of the existing greenery condition [17] and the greenery condition for a proposed master plan in an estate development [18]. In the development of the empirical model, air temperature data that had been gathered in the previous studies were combined with the most recent data, which includes estate-wide and canyon types of measurements. The measurement points cover various types of land uses, including residential, commercial, business park, education, industrial, park, open space and sport facility. The daily minimum (Tmin), average (Tavg) and maximum (Tmax) temperature of each point of measurement were calculated as the dependent variables of the air temperature prediction model. The independent variables for the models can be categorised into: Climate predictors: daily minimum (Ref Tmin), average (Ref Tavg) and maximum (Ref Tmax) temperature at reference point; average of daily solar radiation (SOLAR). For the SOLAR predictor, average of daily solar radiation total (SOLARtotal) was used in Tavg models, while average of solar radiation maximum of the day (SOLARmax) was used in the Tmax model. SOLAR predictor is not applicable for the Tmin model. These data are obtained from the weather station. Urban morphology predictors: percentage of pavement area over R 50 m surface area (PAVE), average height to

199

Downloaded by [Marcel Ignatius] at 23:25 25 September 2012

building area ratio (HBDG), total wall surface area (WALL), Green Plot Ratio (GnPR), sky view factor (SVF) and average surface albedo (ALB). These data are provided by the government agency and cross-checked by eld survey. Before the model was developed, the radius of the inuence area was determined. A radius of 50 metres was deemed as a suitable one after a series of inuence area studies comparing radius values from 25100 m (Figure 1). The temperature models were then developed by examining the variables regression coefcient values and their correlations with the dependent variables. Wind speed, one of the most common variables, was excluded in the model development, since the models focus on calm day conditions (wind speed , 3 m/s). Meanwhile another common variable, altitude was excluded from the model development since the data collected showed altitude has very little inuence on air temperature condition. In the rst stage of model development, trend analysis was carried out to identify and discuss the behaviour of the models variables (based on the data collected in eld measurement), by examining the variables regression coefcient values and their correlations with the dependent variable. Not all of the independent variables are signicant. However, it is important to analyse how these variables behave in determining the air temperature. The next stage is to develop the air temperature prediction models that use only the signicant variables. The air temperature regression models were developed based on the data collected over a period of close to 3 years. It is necessary to validate the models with another period of measurement data, in this case, with fairly clear and calm day conditions (wind speed , 3 m/s). The air temperature prediction models can be written as follows: T min 8c 4:061 0:839 Ref T min 8C 0:004 PAVE % 2 0:193 GnPR 2 0:029 HBDG 1:339E 2 06 WALL m2 T avg 8C 2:347 0:904 Ref T avg 8C 5:786E 2 05 SOLARtotal W=m2 0:007 PAVE % 2 0:06 GnPR 2 0:015 HBDG 1:311E 2 05 WALL m2 0:633 SVF T max 8C 7:542 0:684 Ref T max 8C 0:003 SOLARmax W=m2 0:005 PAVE % 2 0:016 HBDG 6:777E 2 06 WALL m2 1:467 SVF 1:466 ALB

200

W.N. Hien et al.

Downloaded by [Marcel Ignatius] at 23:25 25 September 2012

Figure 1.

Sample of urban area measurement point in inuence area radius of 25 m, 50 m, 75 m and 100 m.

Since it is impossible to put the all the theoretical background and prediction model development into this paper, the author will only underline the essential elements of STEVE, while a more detailed explanation, data validation, and its application for other studies can be read from the related papers, which can be found in [17 21]. 3. Hypotheses Both ENVI-met and STEVE have the same objective: to create a temperature map in the shape of raster images. In this case, since STEVE was developed and built based on local climate data (Singapore), in terms of predicting the temperature output, it will be more representative for reecting the real condition, compared to ENVI-met. However, since STEVE has a limitation on creating the raster image based on interpolating the result from a 100 100 m grid; ENVI-met denitely produces a better

and more accurate resolution since it simulates the result pixel by pixel and grid dependent. 4. Methodology

Comparisons between STEVE and ENVI-met were drawn on a scenario comprising ve building massings, each 48 m in height. In STEVE, the ground surface was assumed to be fully paved (Green Plot Ratio or GnPR value is zero). The same setting was also applied in ENVI-met, where the soil is set to be fully pavement (concrete). The needed boundary condition for STEVE has been described in the previous chapter, whereas several parameters from both climate and urban morphology predictors (in this case, from the 5 buildings scenario) were calculated to produce the predicted temperature. While for ENVI-met, the buildings are modelled accordingly with the

International Journal of Sustainable Building Technology and Urban Development

201

Figure 2.

Model settings on GIS using STEVE (left) and ENVI-met (right).

Downloaded by [Marcel Ignatius] at 23:25 25 September 2012

same dimensions as in the STEVE calculation (Figure 2), and for the climatic condition, the values were set within the conguration setting (CF) le, based on the reference weather data used by STEVE as well. The CF inputs needed for simulation are initial temperature (averaged from the reference data), wind speed, and relative humidity (RH). For ENVI-met, the only difference is that it does not use dynamic temperature prole, compared to STEVE which uses a 24-hourly temperature prole in order to produce Tmin, Tavg and Tmax. Because STEVE focuses on low wind speed conditions, before comparing the result from both sides the authors embarked on a mini parametric study for different wind speeds, in order to explore whether ENVI-met could simulate properly under the low wind speed condition. In the end, these boundary conditions for both methods were made as similar as possible in order to embark on an appropriate comparison study. For this study purpose, 6 March 2010 has been selected as the simulation day, since it is not just a typical calm day (suitable for STEVE calculation), but also the hottest of year 2010. Hence, the boundary condition congurations are as follow: (1) For STEVE . Based on hourly data generated by the weather station, the maximum, minimum and average temperature on 6 March 2010 are: 31.28C, 25.498C, and 27.988C respectively. . Daily solar radiation total (SOLARtotal): 5062.945 W/m2. . Solar radiation maximum of the day (SOLARmax): 683.5 W/m2. (2) For ENVI-met . Simulation date: 6 March 2010. . Initial temperature atmosphere: 301 K (or 27.988C, average temperature). . Average Relative Humidity: 70%. . Roughness length: 0.1 (for urban area). 5. 5.1 Data and analysis ENVI-met simulation for different wind speed

appears that ENVI-met simulation results are not quite accurate for the low wind-speed condition (below 1 m/s), as they show inconsistencies of wind directions. A more consistent wind direction can only be seen when the speed is set starting from 1 m/s. Therefore, in order to compare it with the STEVE result, the proposed scenario was simulated by ENVI-met with minimum wind speed of 1 m/s. 5.2 Comparison of STEVE and ENVI-met To create a raster map within ArcGIS, STEVE deployed a circular buffer zone with a 50 m radius. Then, STEVE calculated and predicted the temperature in that particular buffer centre point which was inuenced by the environment condition within. On the other hand, the ENVI-met calculation is technically based on pixels with the particular size of the grid as dened within the create model domain settings, and it uses a uniform mesh with a maximum grid size of about 300 300 35 cells with the horizontal extension ranging between 0.5 10 m and a typical vertical height of 1 to 5 metres (this study use 5 metres for each x, y, z grid cell) These two different methods appear to vary the raster map output as well (Figure 4). For instance, in the red highlighted location, STEVE displayed an equal or uniform temperature within the gaps between buildings, while ENVI-met showed temperature variation on same location. On a bigger picture, ENVI-met showed the same condition between the building gaps, while STEVE has a different temperature prediction. The explainable reason for this result is the different grid sizing for both applications; ENVI-met provides better resolution as the temperature is calculated for every grid cell, while the STEVE raster image is based on a 100 100 m grid data interpolation (based on a 50 m radius buffer zone). Thus, in order to have better resolution of the temperature map, the authors were experimenting by adding more buffer zones within ArcGIS for the STEVE calculation. To achieve a better raster map resolution, this study tried another scenario where 27 additional measurement points were deployed within the same building arrangement. While normal STEVE methodology uses an attached buffer zone, this study overlapped all those additional buffer

By observing the temperature maps from ve different scenarios regarding the wind speed condition (Figure 3), it

202

W.N. Hien et al.

Downloaded by [Marcel Ignatius] at 23:25 25 September 2012

Figure 3.

ENVI-met simulation result with different wind speed.

boundaries as a result of additional measurement points (Figure 5). The diagrams above (Figures 5 and 6) display the difference of utilising more measurement points compared to the previous attempt in order to achieve a more accurate temperature map. The highlighted area shows the difference between the two methods. In the previous

attempt, only the centre portion of the building gaps showed a low temperature while with the overlapping buffer zones, every gap in-between buildings indicated a low temperature result. Within ArcGIS, the colour coding of a certain buffer zone depends on its surrounding measurement point location (interpolation method). Therefore, to generate a detailed temperature map

International Journal of Sustainable Building Technology and Urban Development

203

Downloaded by [Marcel Ignatius] at 23:25 25 September 2012

Figure 4.

Comparison of temperature map result from both STEVE and ENVI-met.

Figure 5.

Additional buffer zone on STEVE calculation to improve the accuracy of temperature map.

prediction within a small-scale city parcel using STEVE might be misleading compared to using ENVI-met. When they are compared next to each other, the raster map generated from STEVE with the overlapping buffer zone method showed a similar pattern to the one generated from ENVI-met (Figure 7). The noticeable difference between these two lies at the maximum-minimum temperature prediction. The temperature prediction from ENVI-met displays a higher temperature (values ranging from 302.33 K to 303.05 K (29.338C to 30.358C)) compared to STEVE, where the temperature values range from 32.18C to 32.688C.

ENVI-met has been tested for different wind speed values of 1, 3, and 5 m/s. As expected, the result showed that when the wind speed increased, the temperature was reduced. The simulation displayed that for 1 m/s wind speed, the temperature ranged from 29.348C to more than 29.818C, for 3 m/s the temperature range was from 28.408C to 29.118C, and for 5 m/s the range was 27.938C to 28.638C. On the other hand, the STEVE temperature prediction for a calm day ranges from 32.18C to 32.688C.

5.3

Wind speed

As previously mentioned, the STEVE calculation focuses on calm day conditions, which excludes the wind speed variable, while the ENVI-met calculation considers this. Figure 8 shows the comparison of STEVE and ENVI-met if the initial wind speed values for the latter were varied.

5.4 Temperature prole difference It is important to note that results from both simulations show the predicted temperature from ENVI-met was constantly lower than the STEVE result. In order to explore this, another study was done to test whether ENVImet under predicts the initial temperature (since this tool is mostly used in temperate country case studies). The added scenario was to set both STEVE and ENVI-met simulating

204

W.N. Hien et al.

Downloaded by [Marcel Ignatius] at 23:25 25 September 2012

Figure 6.

Temperature map comparison as a result of overlapping buffer areas on STEVE calculation.

Figure 7.

STEVE and ENVI-met temperature map (afternoon) result comparison.

an empty land without buildings, and then the result was compared to the previous one. Figure 9 below indicates that the prediction temperature from ENVI-met for empty land (blue line) is lower than STEVE (red line). Predicted temperatures with buildings added into the parcel also indicate lower values for ENVI-met (purple line) compared to STEVE (green line). Therefore, the diagram provides a proper explanation where initial temperature for ENVI-met happens to be lower than STEVE, which means it under predicts the simulated temperature. Another observation revealed that both STEVE and ENVI-met showed similar temperature prole trends.

ENVI-met displays a smoother curve line result, mainly because it was using only an initial boundary condition to run the simulation, while STEVE implemented a full one day weather data prole (reference maximum, average, and minimum temperature) as the input data for generating the predicted temperature. For this study purpose, 6 March 2010 was the selected day, since it was not just a typical calm day, but also the hottest of 2010. Since STEVE was using the actual data, the temperature prole might be as smooth as the one which is generated by ENVI-met, since the eld measurement data tends to vary depending on the weather, while the ENVI-met simulation was running based on ideal conditions.

International Journal of Sustainable Building Technology and Urban Development

205

Downloaded by [Marcel Ignatius] at 23:25 25 September 2012

Figure 8.

STEVE and ENVI-met wind prole comparison.

Figure 9.

Comparison of STEVE tool and ENVI-met chart.

Table 1 reveals that the predicted temperature difference for the whole day for empty land between STEVE and ENVI-met could reach 3.778C maximum and 2.388C on average, while the temperature difference when buildings were added afterwards could reach 3.508C maximum and 2.268C on average. The temperature reduction from an empty land scenario to the occupied one was higher in STEVE (0.998C maximum reduction, 0.438C on average) compared to ENVI-met (0.668C maximum reduction, 0.118C on average). 5.4 Adjustment of ENVI-MET temperature prole

Comparison between STEVE and ENVI-met above shows that there is a 28C to 38C difference between the two prediction models. To achieve the most similar tempera-

ture prole between these two applications, initial temperature in ENVI-met should be raised into 28C and 38C become 302 K (298C) and 303 K (308C). The temperature prole generated by this revised ENVI-met simulation displayed a smoother curved line compared with the STEVE prole (6 March 2010 data). Therefore, to have the proper assessment on the temperature prole curved line, this study attempted to use different dates (calm and hot days), which were 28 February 2010 and 26 November 2009. Figure 10 and Table 2 below show the comparison of ENVI-met simulation with three different initial temperatures (300 K, 302 K, and 303 K) and STEVE simulation with three different dates (6 March 2010, 28 February 2010, and 26 November 2009). The results indicated that

Downloaded by [Marcel Ignatius] at 23:25 25 September 2012

206

Table 1. Empty (STEVE) Temp. diff. (STEVE) Point 4 & 5 (ENVI-met) Empty (ENVI-met) Temp. diff. (ENVI-met) STEVE-ENVI-met (empty)

Prediction temperature difference between STEVE and ENVI-met. STEVE-ENVI-met (massing)

Time (24 hrs)

Massing (STEVE)

W.N. Hien et al.

1:00 2:00 3:00 4:00 5:00 6:00 7:00 8:00 9:00 10:00 1:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00 21:00 22:00 23:00 24:00 33.09 25.85 29.21 2 0.99 2 0.13 2 0.43 29.50 23.84 26.48 30.16 23.97 26.83 2 0.66 2 0.11 2 0.35 3.77 1.16 2.38

27.26 26.92 26.55 26.19 25.85 25.71 25.88 25.76 26.41 28.30 29.82 30.75 31.51 32.10 30.62 30.65 30.85 31.15 30.57 30.48 30.11 29.48 28.99 28.63

27.47 27.11 26.72 26.35 25.99 25.85 26.02 25.89 26.58 28.56 30.30 31.45 32.37 33.09 31.29 31.32 31.57 31.93 31.23 31.12 30.67 29.90 29.28 28.90

20.21 20.19 20.17 20.16 20.14 20.13 20.14 20.14 20.17 20.26 20.49 20.70 20.86 20.99 20.67 20.67 20.72 20.78 20.66 20.64 20.55 20.41 20.29 20.27 25.03 24.85 24.64 24.48 24.30 24.13 23.96 23.84 24.89 26.04 27.18 27.87 28.66 29.31 29.44 29.5 29.35 28.97 28.17 27.24 26.61 26.04 25.67 25.34 25.34 25.06 24.82 24.61 24.42 24.24 24.09 23.97 25.22 26.43 27.69 28.46 29.16 29.71 30.06 30.16 29.99 29.50 28.43 27.51 26.90 26.41 26.00 25.65 2.13 2.05 1.90 1.74 1.57 1.61 1.93 1.92 1.36 2.13 2.61 2.99 3.21 3.38 1.23 1.16 1.58 2.43 2.80 3.61 3.77 3.49 3.28 3.25

2 0.31 2 0.21 2 0.18 2 0.13 2 0.12 2 0.11 2 0.13 2 0.13 2 0.33 2 0.39 2 0.51 2 0.59 2 0.50 2 0.4 2 0.62 2 0.66 2 0.64 2 0.53 2 0.26 2 0.27 2 0.29 2 0.37 2 0.33 2 0.31

2.23 2.07 1.91 1.71 1.55 1.58 1.92 1.92 1.52 2.26 2.64 2.88 2.85 2.79 1.18 1.15 1.50 2.18 2.40 3.24 3.50 3.44 3.32 3.29 3.50 1.15 2.29

Max Min Avg

32.10 25.71 28.77

International Journal of Sustainable Building Technology and Urban Development

207

Downloaded by [Marcel Ignatius] at 23:25 25 September 2012

Figure 10.

Comparison of STEVE tool and ENVI-met temperature chart.

ENVI-met with initial temperature set at 302 K and 303 K shows a similar prole to STEVE especially at 8.00am to 11.00am time, but the gap became bigger afterwards when the prole went to the maximum temperature. Another observation was that the STEVE predicted maximum temperature was shifting for each date. On 28 February 2010 prole, the maximum temperature occurred at 4.00pm, which was similar for the three ENVI-met proles, but the value difference between both prediction models was quite signicant, ranging from 1.888C to 4.108C.

6.

Discussions and conclusions

The rst comparison study of the temperature prediction model between STEVE and ENVI-met revealed that the ENVI-met raster map output resolution was more detailed compared to STEVE, as the ENVI-met calculation is based on each grid cell calculation while STEVE is based on the attached 100 100 m buffer zones. Thus, to achieve better raster map resolution, additional measurement points were required in STEVE. This experiment demonstrated that the overlapping buffer zone method gave a different raster map output compared to the initial attached buffer zone. Because the ArcGIS data interpolation method was used to generate the raster image, the temperature colour chart produced is measurement point location dependent. Consecutively, a raster map with more points with the overlapping buffer zone scenario appears to have a similar temperature map distribution compared to the ENVI-met result. From the urban analysis capability point of view, STEVE might be more appropriate for handling a city-

scale study area, since it does not have a limitation for the grid sizing, while ENVI-met does. In the current version, the latter has a restraint where the grid sizing has a maximum number of grid cells, which makes it impossible to simulate the micro-climate condition of a bigger scale city area with appropriate detail included. Currently, the software developer explained this is mainly because ENVI-met runs on a standard x86 personal computer running Windows XP or Vista and does at the moment not take advantage of more than one processor or distributed computing [22]. Another notable nding from the study is that the predicted temperature from ENVI-met was lower compared to STEVE. An additional scenario was completed by simulating an empty land using both STEVE and ENVImet, where lower temperature values were shown for the latter method. The output from these two scenarios indicated that initial temperatures for ENVI-met are lower than for STEVE, which means it under-predicts the temperature parameter. The 2 38C difference most probably comes from other background or other factors from the data collected from the eld for developing STEVE, whereas in ENVImet, the simulation was running in ideal conditions. In the current version, ENVI-met only considers interactions between atmosphere, buildings, soils, vegetation, and water bodies, whereas the other factors such as anthropogenic heat (which can be one of the background factors) are most probably not captured within the ENVI-met simulation. STEVE, with its capability for analysing a city-scale area for generating temperature maps, with much less time needed to conduct the simulation, and having been

208
Table 2. Time (24 hrs) 1:00 2:00 3:00 4:00 5:00 6:00 7:00 8:00 9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00 20:00 21:00 22:00 23:00 24:00 Max Min Avg

W.N. Hien et al.


Comparison of STEVE and ENVI-met temperature. STEVE (06-Mar-10) 27.47 27.11 26.72 26.35 25.99 25.85 26.02 25.89 26.58 28.56 30.30 31.45 32.37 33.09 31.29 31.32 31.57 31.93 31.23 31.12 30.67 29.90 29.28 28.90 33.09 25.85 29.21 STEVE (28-Feb-10) 27.96 27.66 27.62 27.33 26.84 26.54 26.27 26.45 26.96 28.40 30.02 31.38 32.79 33.17 33.74 34.26 34.10 33.66 32.40 31.57 30.49 29.35 28.63 28.31 34.26 26.27 29.83 STEVE (26-Nov-09) 27.02 26.77 26.50 26.13 26.00 25.81 25.63 25.68 26.76 28.42 29.92 31.28 32.36 33.36 34.43 33.86 33.94 33.56 32.29 30.80 29.55 28.77 28.14 27.87 34.43 25.63 29.37 ENVI-met on 300 K 25.34 25.06 24.82 24.61 24.42 24.24 24.09 23.97 25.22 26.43 27.69 28.46 29.16 29.71 30.06 30.16 29.99 29.50 28.43 27.51 26.90 26.41 26.00 25.65 30.16 23.97 26.83 ENVI-met on 302 K 26.8 26.54 26.3 26.1 25.92 25.76 25.61 25.5 26.71 27.59 29.11 29.92 30.63 31.19 31.54 31.65 31.49 31.01 29.81 28.92 28.31 27.83 27.43 27.09 31.65 25.50 28.28 ENVI-met on 303 K 27.5 27.24 27.01 26.80 26.62 26.47 26.33 26.21 27.42 28.12 29.79 30.64 31.35 31.91 32.27 32.38 32.22 31.74 30.48 29.60 29.00 28.52 28.12 27.79 32.38 26.21 28.98

Downloaded by [Marcel Ignatius] at 23:25 25 September 2012

developed using local temperature conditions (tropical), has several advantages compared to ENVI-met. Nevertheless, it does not mean ENVI-met is an inferior option, since it also has capabilities for analysing other things such as ow around and between buildings, turbulence, heat and vapour exchange processes, etc., whereas the current version of STEVE only focuses on local temperature prediction. In the end, in terms of utilising a better temperature prediction model, STEVE could be the better and suitable choice for studying the tropical urban area. References
[1] S.K. Jusuf and N.H. Wong, Development of empirical models for an estate level air temperature prediction in Singapore, Paper presented at Second International Conference On Countermeasures To Urban Heat Islands, Berkeley, United States (2009). [2] M. Bruse and H. Fleer, Simulating surface-plant-air interactions inside urban environments with a three dimensional numerical model, Environ. Modell. Softw. 13 (1998), pp. 373 384. [3] M. Bruse, ENVI-met, 3, available at http://www.envi-met. com. [4] H.E. Landsberg, The Urban Climate, Academy Press, New York, NY, 1981. [5] K. Takahashi, H. Yoshida, Y. Tanaka, N. Aotake, and F. Wang, Measurement of thermal environment in Kyoto City and its prediction by CFD simulation, Energy Build. 16 (2004), pp. 771 779.

[6] E.G. Mcpherson (ed.), Cooling Urban Heat Islands With Sustainable Landscapes. The Ecological City: Preserving and Restoring Urban Biodiversity, University Of Massachusetts Press, Amherst, 1994. [7] T.R. Oke, Boundary Layer Climates, Routledge, London, 1987. [8] J. Edahiro, Efforts in Japan to Mitigate the Urban Heat Island Effect, available at http://www.agrometeorology.org/ topics/new-information-for-agrometeorologists/efforts-injapan-tomitigate-the-urban-heat-island-effect. [9] S. Nieuwolt, The urban microclimate of Singapore, Journal of Tropical Geography 22 (1966), pp. 30 31. [10] L.S. Chia, Temperature and humidity observations on two overcast days in Singapore, Journal of the Singapore National Academy of Science 1 (1970), pp. 85 90. [11] J.E. Nichol, Modelling the relationship between LANDSAT TM thermal data and urban morphology, Paper presented at Proceedings of ACMS/ASPRS Annual Convention and Exposition, Baltimore, United States (1994). [12] M. Roth, Urban heat island dynamics in Singapore, (2002), available at http://courses.nus.edu.sg/course/geomr/front/ fresearch/fres.htm. [13] N.H. Wong and Y. Chen, Study of green areas and urban heat island in a tropical city, Habitat Int. 29 (2003), pp. 547558. [14] T.R. Oke, Street design and urban canopy layer climate, Energy Build. 11 (2002), pp. 103 113. [15] A.J. Arneld, Street design and urban canyon solar access, Energy Build. 14 (1990), pp. 117 131.

International Journal of Sustainable Building Technology and Urban Development


[16] H. Swaid and M.E. Hoffman, Climatic impacts of urban design features for high- and mid- latitude cities, Energy Build. 15 (1990), pp. 325 326. [17] S.K. Jusuf and N.H. Wong, An assessment method for existing greenery conditions in a university campus, Archit. Sci. Rev. 51 (2008), pp. 116 126. [18] S.K. Jusuf and N.H. Wong, GIS-based greenery evaluation on campus master plan, Landscape Urban Plan. 84 (2008), pp. 166 182. [19] N.H. Wong, S.K. Jusuf, N.I. Syai, Y. Chen, N. Hajadi, H. Sathyanarayanan, and Y.V. Manickavasagam, Evaluation of the impact of the surrounding urban morphology on building energy consumption, Sol. Energy 85 (2011), pp. 57 71. [20] N.H. Wong, S.K. Jusuf, R. Samsudin, A. Eliza, and M. Ignatius, A climatic responsive urban planning model for

209

high density city: Singapores commercial district, Int. J. Sustainable Build. Technol. Urban Dev. 2 (2011), pp. 323330. [21] N.H. Wong, N. Islam, M. Ignatius, and S.K. Jusuf, Urban greenery as mitigation strategy for urban heat island in high destiny commercial district of Dhaka, Paper presented at The Eighteenth International Seminar on Urban Form, Montreal, Canada (2011). [22] S. Huttner, M. Bruse, P. Dostal Using ENVI-met to simulate the impact of global warming on the microclimate in central European cities, Paper presented at 5th Japanese-German Meeting on Urban Climatology, Berichte des Meteorolo gischen Instituts der Albert-Ludwigs-Universita t of Freiburg, Germany (2008).

Downloaded by [Marcel Ignatius] at 23:25 25 September 2012

You might also like