Urban wind resource assessment in changing climate: Case study

Abstract

Urban wind resource assessment in changing climate has not been studied so far. This study presents a methodology for microscale numerical modelling of urban wind resource assessment in changing climate. The methodology is applied for a specific urban development in the city of Toronto, ON, Canada. It is shown that the speed of the southwest winds, that is, the most frequent winds increased for .8 m s⁻¹ in the period from 1948 to 2015. The generated wind energy maps are used to estimate the influences of climate change on the available wind energy. It is shown that the geometry of irregularly spaced and located obstacles in urban environments has to be taken into consideration when performing studies on urban wind resource assessment in changing climate. In the analysed urban environment, peak speeds are more affected by climate change than the mean speeds.

Keywords

Urban winds wind energy climate change Toronto sustainability computational fluid dynamics wind resource assessment

Introduction

An accurate, and rather, simple definition of climate is that it represents the average of weather over time and space. The Earth’s climate has always been changing. Although defining climate is easy, understanding the causes as well as the effects of climate changes is challenging indeed. This difficulty is caused by variety of terrestrial and extra-terrestrial climate factor that constantly interact among each other. As over 99% of the Earth’s energy comes from solar radiation (Black and Flarend, 2010), the astronomical climate factors, such as the shape of the Earth’s orbit around the Sun, tilt of Earth’s axis and precession are the main drivers of the Earth’s climate, as described in 1920 by Milanković’s theory of ice age cycles (Milanković, 1920, 1930). These astronomical factors influence the amount of the Sun’s energy reaching the Earth. However, these factors have large return periods and hence can be considered as being constant over the time periods of several hundreds of years or so (e.g. the period for precession of Earth’s orbit is around 23,000 years while the other two factors have even larger periods).

Changes in the concentration of greenhouse gases and the reflectivity of Earth’s atmosphere and surface are the most important terrestrial factors which can disrupt the Earth’s energy balance. Changes in concentration of the greenhouse gasses (water vapour (H₂O), carbon dioxide (CO₂), methane (CH₄) and chlorofluorocarbons (CFCs)) affect the amount of heat retained by Earth’s atmosphere. Numerous studies (e.g. Cook et al., 2013; Hansen et al., 2011; Trenberth, 2009) suggest that human activities have altered the concentration of CO₂ in the atmosphere and thus resulted in the on-going climate change. Reflectivity of Earth’s atmosphere and surface has also been changing due to changes in land use and land cover such as deforestation, desertification and urbanization.

Rapid urbanization is a global phenomenon. In 2014, 54% of the world’s population inhabited urban areas (United Nations, 2014). The same study reported that the percentages have been even higher in economically developed countries. Therefore, estimating the effects of climate change in urban environments is of great interest.

From mechanical and thermodynamical points of view, urban environments represent rough surfaces with variety of sources and sinks of heat. Air flows in urban environments are very complex due to a large number of irregularly located and spaced obstacles. Compared to rural areas, the winds in cities are characterized by larger values of turbulence intensity and smaller mean wind speed. Climate change analysis in urban environments is very challenging mainly due to three factors. First, complicated climate change feedback loops and mechanisms, as well as the relationship between different feedbacks, are not very well understood. For example, the role of clouds in different climate change scenarios is not fully known. However, this difficulty concerning the representation of clouds in climate models is not restricted only to urban environments, but to the climate change modelling in general. The second factor has already been mentioned and it is the complexity of wind fields in cities. Namely, it is difficult to perform any wind resource assessment study in urban environments due to large number of buildings and other objects which act as obstacles to the incoming wind. For that reason, the urban boundary layers are very complex and highly turbulent close to the surface. Lastly, an additional factor which complicates climate change analysis in cities is the change in the city landscape as well as the fact that cities are growing in size, population and built-up surface. What is a park or a field today might be a built-up area in the future. We, here, resume to only consider the climate change aspects and not the changes in the city landscape.

Wind is one of the meteorological variables mostly influenced by climate change. A number of studies have shown that the near-surface winds are slowing down across the globe (McVicar et al., 2012; Romanić et al., 2015a). Vautard et al. (2010) reported that wind speeds in the northern mid-latitudes declined between 5% and 15% in the period 1979–2008. They further concluded that 25%–60% of these negative trends are caused by the increase in surface roughness, which is mostly due to increased urbanization in the last several decades. Holt and Wang (2012), however, showed that the westerly winds at 80 m level in the Great Lake region in Canada had pronounced positive trends in the period from 1979 to 2009 (~.24 m s⁻¹ over 10 years).

This study investigates the potential influence of climate change on wind resources in an urban environment. To the authors’ knowledge, this is the first study that aims to relate the urban wind resource assessment with the long-term wind speed trends. The proposed method is based on combining microscale computational fluid dynamics (CFD) simulations with the long-term wind speed trends calculated from the National Centers for Environmental Prediction/National Center for Atmospheric Research (NCEP/NCAR) Reanalysis 1 data (Kalnay et al., 1996).

The detailed description of data, test site and methodology used to relate long-term wind speed trends to available wind resources is presented in section ‘Data, methodology and numerical setup’. The results and critical discussion are given in section ‘Results and discussion’. The concluding remarks follow in the last section.

Data, methodology and numerical setup

Site

The analysis of urban wind resource assessment in changing climate is conducted on a city block located in Toronto, ON, Canada. The site had been used as the 2015 Pan American Games Athletes’ Village (PanAm Village, Figure 1). The construction works on the site were finished in the first half of 2015. PanAm Village is located between Bayview Avenue and Cherry Street and from north of Front Street to a rail corridor north of the shores of Lake Ontario (Latitude 43°39′12″N, Longitude 79°21′16″W). The PanAm Village is bound by the urban area from the north and east sides while the southern and western sides are mostly exposed to land with low terrain roughness neighbouring the lake. The athletic village is composed of 22 building blocks positioned along the existing streets.

Figure 1.

Location of PanAm Village in Toronto. Winds coming in from southwest direction (240°) are shown. The arrow in the lower-left corner indicates the north direction.

Data

The wind data used in this study are extracted from the global NCEP/NCAR Reanalysis 1 dataset (Kalnay et al., 1996). The reanalysis data are the result of an advanced analysis/forecast system that performs data assimilation using past data from 1948 to the present. The data have 2.5° latitude by 2.5° longitude spatial resolution over the whole globe and are available at three different temporal resolutions (4 times a day, daily, and monthly). The mean daily values for the two wind components (zonal, u, and meridional, v) are extracted for the PanAm Village site from the entire global database. The extracted data cover the time period from 1 January 1948 to 1 January 2015 (67 years of data which corresponds to 24,473 data records).

The wind data are given at the sigma .995 level (also known as the near-surface level, σ_.995). Sigma, σ, is a vertical coordinate used in many meteorological weather forecast models. The coordinate is defined as a ratio of the pressure at a given point in the atmosphere to the surface pressure underneath it. Thus, σ = 1 at the surface and σ = 0 at the top of the atmosphere. A level where σ is equal to .995 and therefore represents the level in the atmosphere where the pressure is .995 times the surface pressure. Wind speed at σ_.995 level is calculated from the wind components as $V_{σ . 995} = \sqrt{u^{2} + v^{2}}$ . Note that σ_.995 actually represents σ_0.995 level, but the zero in the symbol is omitted for simplicity. Wind direction at the σ_.995 level (D_σ_.995) is obtained from the two wind components using the four-quadrant inverse tangent, namely

D_{σ . 995} (u, v) = \frac{180}{π} (2 \tan^{- 1} \frac{u^{2} + v^{2} - v}{u} + π)

(1)

where the factor 180/π is used to convert radians to degrees.

Wind rose of direction and intensity, as well as histograms of the wind probability density function (PDF) for the PanAm Village site is shown in Figure 2. It can be observed that winds coming in from the third quadrant are the most frequent. These winds were active in more than 50% of the time. The southwest winds (240° direction, $V_{σ . 995}^{240 °}$ ) were blowing in approximately 18% of the time. The orientation of this wind direction in respect to the PanAm Village is portrayed in Figure 1. It can be seen that the southwest winds are coming in to the PanAm Village from Lake Ontario. Winds blowing from the first and fourth quadrant were rare. Besides being the most frequent winds, Figure 2(a) and histograms in Figure 2(b) and (c) show that the $V_{σ . 995}^{240 °}$ winds were also the strongest winds (in 50% of the time $V_{σ . 995}^{240 °} > 6.04 m s^{- 1}$ ).

Figure 2.

(a) Wind rose with speed distribution, (b) wind histogram for all wind directions and (c) wind histogram for southwest winds ( $V_{σ . 995}^{240 °}$ ). All data are for the PanAm Village and at the σ_.995 level.

Due to the dominance of the $V_{σ . 995}^{240 °}$ winds over the PanAm Village site, we chose to perform the wind resource assessment analysis for this wind direction. The methodology described in this article, however, can be extended and applied to any number of wind directions.

Hereafter, the following notation will be used. $X_{σ . 995}^{240 °}$ will represent any variable X given at σ_.995 level when $V_{σ . 995}^{240 °}$ winds were observed. Geometric height of the σ_.995 level (z_σ_.995) is computed utilizing the barometric formula rearranged to be solved for the height, that is

z_{σ . 995} = \frac{R * \bar{T}}{M g} \ln (\frac{p_{s}}{p_{σ . 995}})

(2)

Here, $R^{*} = 8.3145 J {mol}^{- 1} K^{- 1}$ is the universal gas constant, M = .029 kg mol⁻¹ is the molar mass of air, g = 9.81 m s⁻² is the gravitational acceleration, p_σ_.995 = .995·p_s is the pressure at the σ_.995 level (in Pa), p_s is the surface pressure (in Pa) and $\bar{T} = (T_{2 m} + T_{σ . 995} / 2)$ (in K) is the average temperature in the layer between the surface and σ_.995 level. In order to calculate $z_{σ . 995}$ from equation (2), the following variables have also been obtained from the reanalysis dataset: p_s, T_2m and $T_{σ . 995}$ . All downloaded data are stored in the self-describing NetCDF files and the total size of the downloaded data is 13.5 MB.

Finally, the air density ( $ρ_{σ . 995}$ ) is calculated from the ideal gas law as follows

ρ_{σ . 995} = \frac{p_{σ . 995}}{R T_{σ . 995}}

(3)

where R = 286.9 J kg⁻¹ K⁻¹ is the gas constant for dry air.

Methodology and numerical setup

The methodology to estimate the climate change influences on available wind resources in PanAm Village is based on the magnitude and sign of the linear trend of the $V_{σ . 995}^{240 °}$ winds (Figure 3(a)). Recall, however, that wind speeds are given at the σ_.995 level, which has a variable geometric height (see section ‘Data’) that depends on the surface pressure. For that reason, the trend of $z_{σ . 995}^{240 °}$ had to be calculated as well (Figure 3(b)) in order to properly interpret the observed trend of $V_{σ . 995}^{240 °}$ . The trend in air density, $ρ_{σ . 995}^{240 °}$ , is shown in Figure 3(c).

Figure 3.

Time series and calculated trend lines of (a) $V_{σ . 995}^{240 °}$ , (b) $z_{σ . 995}^{240 °}$ , (c) $ρ_{σ . 995}^{240 °}$ and (d) $T_{σ . 995}^{240 °}$ .

The Mann–Kendall non-parametric test for trend (Kendall, 1970; Mann, 1945) and Sen’s slope estimator (Sen, 1968) are used to detect and estimate strength of trends in the time series. These two methods are widely used in many meteorological, climatological as well as wind trend analysis studies (e.g. Romanić et al., 2015a; Tyrlis and Lelieveld, 2013). The two-tailed Mann–Kendall test inspects the null hypothesis of the absence of trend in the time series at the s significance level (in this study, s = .05). Sen’s slope estimator calculates and applies the median slope among all the slopes determined by all pairs of data points. The linear equations of the trend lines are given in figures in which trends are analysed.

The trend lines are used as the inputs for the CFD analysis. Namely, one CFD analysis of wind resources in PanAm Village is performed using the values of $V_{σ . 995}^{240 °}$ , $z_{σ . 995}^{240 °}$ and $ρ_{σ . 995}^{240 °}$ at the beginning of each of these three trend lines. In other words, the offset values on the trend lines are used as inputs for the first CFD simulation. The values at the end points on the trend lines are used as the numerical inputs for the second CFD analysis. Difference in the computed wind resources between the first and the second CFD simulation is an estimate of the climate change influence on the wind resources over the PanAm Village. The employed methodology therefore links the long-term wind speed changes over the PanAm Village site with the complexity of flows in that urban environment without taking into consideration the changes in urban planning during the same period of time.

The numerical simulations of wind resources in PanAm Village are performed using a CFD software STAR-CCM+^® (Version 9.04), developed by CD-adapco. The numerical setup of the model used in this study is described in Romanić et al. (2015b). Namely, the steady-state Reynolds-averaged Navier–Stokes (RANS) equations with the k-ω shear stress transport (SST) turbulence model (Menter et al., 2003) are used to simulate turbulent flow over PanAm Village. The k-ω SST turbulence model has been used in many wind engineering studies (e.g. Jubayer and Hangan, 2014; Romanić et al., 2015b). The domain is set to be thermally homogeneous with constant air density and the Coriolis effect is neglected. The discretization of the RANS equations is achieved applying the second-order accuracy with the pressure–velocity coupling through the segregated flow model. The semi-implicit method for pressure-linked equations (SIMPLE) method (Pletcher et al., 2011) is utilized to update the solution between successive iterations.

The inflow velocity profiles are calculated applying the power law on $V_{σ . 995}^{240 °}$ values at the σ_.995 level, namely

V_{g r}^{240 °} = V_{σ . 995}^{240 °} {(\frac{z_{g}}{z_{σ . 995}^{240 °}})}^{α_{s}}

(4)

Here, $V_{g r}^{240 °}$ is the gradient wind speed for the 240° wind direction. Boundary layer height of $z_{g} = 500 m$ is being selected. The power law exponent, $α_{s}$ , for suburban environment is set to be .34 (Kaltschmitt et al., 2007). Turbulence intensity profiles are calculated following the ESDU 83045 guidelines (ESDU, 2002), as implemented in Romanić et al. (2015b). The inflow velocity and turbulence intensity profiles are presented in Figure 4.

Figure 4.

(a) Inlet velocity profiles and (b) inlet turbulence intensity profiles.

The generated polyhedral mesh contains 4.37 million cells in the computational domain sized according to the COST guideline (Franke et al., 2004) (see Figure 5). The computational domain is sized in respect to the height of the tallest building in PanAm Village (h = 100 m). Detailed description of the mesh as well as the grid independency analysis is given in Romanić et al. (2015b).

Figure 5.

Extension of the computational domain with boundary conditions (Romanić et al., 2015b).

Results and discussion

Observed trends

Figure 3 shows that $V_{σ . 995}^{240 °}$ winds at the PanAm Village site increased by .8 m s⁻¹ in the period from 1948 to 2015. This speed-up corresponds to a 14.2% increase in the offset wind speed. The observed trend is statistically significant with the p value <10⁻⁴ (see Table 1). Important to note is that Sen’s slopes of $V_{σ . 995}^{240 °}$ are statistically significant even at the 95% confidence level. $z_{σ . 995}^{240 °}$ , $ρ_{σ . 995}^{240 °}$ and $T_{σ . 995}^{240 °}$ , on the other hand, were all trendless during the same time period.

Table 1.

Evaluation statistics of the trend analysis study.

Variable	Mann–Kendall coefficient	Hypothesis test	p value	Sen’s slope (m s⁻¹ per record)	Sen’s slope with 95% confidence intervals (m s⁻¹ per record)
Variable	Mann–Kendall coefficient	Hypothesis test	p value	Sen’s slope (m s⁻¹ per record)	Lower	Upper
$V_{σ . 995}^{240 °}$	.0533	1	<10⁻⁴	1.7 × 10⁻⁴	1.1 × 10⁻⁴	2.3 × 10⁻⁴
$z_{σ . 995}^{240 °}$	−.0091	0	.3536	−1.6 × 10⁻⁵	−5.0 × 10⁻⁵	1.8 × 10⁻⁵
$ρ_{σ . 995}^{240 °}$	.0029	0	.7634	1.5 × 10⁻⁷	−8.4 × 10⁻⁷	1.1 × 10⁻⁶
$T_{σ . 995}^{240 °}$	−.0056	0	.5670	−6.8 × 10⁻⁵	−3.0 × 10⁻⁴	−1.6 × 10⁻⁴
$u_{σ . 995}^{240 °}$	.0492	1	<10⁻⁴	1.4 × 10⁻⁴	8.4 × 10⁻⁵	1.9 × 10⁻⁴
$v_{σ . 995}^{240 °}$	.0487	1	<10⁻⁴	8.6 × 10⁻⁵	5.2 × 10⁻⁵	8.6 × 10⁻⁴
Meridional $Δ T_{σ . 995}^{240 °}$	.0415	1	<10⁻⁴	1.1 × 10⁻⁴	5.7 × 10⁻⁵	1.6 × 10⁻⁴
Zonal $Δ T_{σ . 995}^{240 °}$	.0319	1	.0011	1.0 × 10⁻⁴	4.1 × 10⁻⁵	1.6 × 10⁻⁴

A trend analysis of the wind speed components (zonal, $u_{σ . 995}^{240 °}$ ), and meridional, $v_{σ . 995}^{240 °}$ ) is further performed in order to better understand the observed trend of the $V_{σ . 995}^{240 °}$ winds. Results are depicted in Figure 6 and Table 1. Both wind components have statistically significant positive trends. The positive trend in zonal wind speeds (1.4 × 10⁻⁴ m s⁻¹ per record) is considerably stronger than the trend in meridional wind speeds (8.6 × 10⁻⁵ m s⁻¹ per record). Holt and Wang (2012) concluded that the increase in zonal wind speed components in the Great Lake region is due to the positive trend of the westerly winds. They also noticed that the strength of mid-latitude jet stream increased over the last several decades. The westerly belt and the mid-latitude jet stream are the key elements of the general circulation of the atmosphere.

Figure 6.

Trend analysis of (a) zonal and (b) meridional components of the $V_{σ . 995}^{240 °}$ winds.

The long-term trends of wind speeds could also be caused or altered by the trends of the horizontal temperature gradients. The relationship between trends of wind speed and trends of temperature is relatively straightforward. Namely, winds are caused by differences in temperature which in turn result in pressure differences (i.e. pressure gradients). A long-term trend of temperature gradients would therefore be accompanied with the corresponding wind speed changes. For that reason, the meridional and zonal temperature gradients over 5° latitude and 5° longitude are calculated at the PanAm Village site (Figure 7 and Table 1). The positive trend in the period 1949–2015 is statistically significant for both meridional and zonal temperature gradients. The meridional temperature gradient, however, is more pronounced than the zonal temperature gradient. This result is in accordance with the previous finding that the zonal wind component ( $u_{σ . 995}^{240 °}$ ) has experienced stronger positive trend in the analysed 67-year long period.

Figure 7.

Trend analysis of (a) meridional and (b) zonal temperature gradients.

Wind resource assessment study

The results of the wind resource assessment analysis at the PanAm Village site are shown in Figure 8. As previously explained in section ‘Methodology and numerical setup’, flow fields in Figure 8(a) and (b) are based on the $V_{σ . 995}^{240 °}$ values from the first and the last point on the trend line in Figure 3(a), respectively. Wind speeds in Figure 8 are given at 5 m above all surfaces (roads, roofs and walls) in PanAm Village. The height of 5 m is chosen for the reason that most urban wind turbines are either vertical axes wind turbines (VAWT) or horizontal axis wind turbines (HAWT) with low hub heights which are typically below 10 m above the roof surface.

Figure 8.

Velocity field at 5 m above building and ground surfaces calculated using the trend line in Figure 3a. Wind fields based on (a) the first (i.e. the offset) value of the $V_{σ . 995}^{240 °}$ trend line and (b) the last points on the trend line. Flow is along x-axis.

The resulting wind field portrayed in Figure 8 can be compared with the long-term wind speed trend in Figure 3(a). It can be seen that the difference between the area-averaged speeds on the isosurface positioned at 5 m above all surfaces is for .25 m s⁻¹ smaller than the corresponding long-term increase in the V^240°_σ.995 $V_{σ . 995}^{240 °}$ winds (see Figure 3(a) and Table 1). The spatial average of velocities is the standard method to address the problem of horizontal inhomogeneity of flows in urban environments (Cheng and Castro, 2002; Kastner-Klein and Rotach, 2004; Romanić et al., 2015b). These results lead to the conclusion that the long-term increase in the $V_{σ . 995}^{240 °}$ winds above PanAm Village is a function of complexity of the urban environment. The influence of urban environment on the flow is incorporated twofold. First, the non-linearity of the power law (equation (4)) and the adequate values of the power law exponent for urban or sub-urban environments are necessary to vertically extrapolate velocities from one height to another. Due to the non-linearity of the power law and the requirement of no-slip condition at the surface, the calculated wind speed trends will be a function of the height above ground. Second, small scale features of urban environments (e.g. buildings, trees, cars) are not directly accounted for in the reanalysis data. The presented CFD analysis, on the other hand, is capable of capturing some of these features and therefore their non-linear effects on the flow field in the analysed urban block. The above discussion and the presented results indicate that an accurate modelling of the complex nature of urban winds is an important prerequisite for adequate urban wind resource assessment in changing climate.

The difference between surface-averaged velocities between the two model runs is .55 m s⁻¹, as indicated in Figure 8(a) and (b). The maximum (peak) velocities, on the other hand, differ for 1.2 m s⁻¹. This observation is in accordance with the literature. Namely, Cheng et al. (2012) investigated a possible impact of climate change on wind gusts in Ontario, Canada. They concluded that the hourly wind gusts above 28 m s⁻¹ in the period 2081–2100 are expected to be 10%–15% greater than the observed gusts for the period 1994–2007. It seems that the peak velocities are more influenced by the climate change than the mean velocities.

The observed long-term increase in the mean wind speed could be beneficial for the urban wind energy utilization. Wind turbines employed in urban environments are typically VAWT (Romanić et al., 2015b). For this reason, the observed increase in the peak wind speed and hence turbulence intensity should not considerably affect their performances.

Increased ventilation effect due to higher wind speeds results in better air quality in cities. Namely, higher wind speeds carry away the pollutants from their origin faster and more efficient. The increase in the turbulence intensity contributes in diluting pollutants. The observed increase in peak winds, however, could have negative effects on the wind turbine performance and building design from a structural point of view. Higher peak speeds and increased turbulence intensity lead to more pronounced wind loadings and wind-induced dynamic responses.

Finally, it should be noted here that the presented analysis takes into account only one wind direction and no changes in urban coverage. For the complete picture of the connection between climate change and urban winds, a planning growth as well as all wind directions should be considered. These factors shall be investigated in the next step of this research. Furthermore, the future work should include the thermal effects, atmospheric stability and increased vertical air flow due to local heating.

Concluding remarks

This study analysed the long-term wind speed trend above the 2015 Pan American Games Athletes’ Village (PanAm Village), located in Toronto, Ontario, Canada. Based on the NCEP/NCAR Reanalysis 1 data (Kalnay et al., 1996), it is shown that the speed of the predominant winds at PanAm Village, the southwest winds, increased for .8 m s⁻¹ in the period from 1948 to 2015.

The increase in the zonal wind component was larger than the increase in the meridional wind component. Air temperature and air density, on the other hand, were trendless. An additional analysis showed that both zonal and meridional temperature gradients above the PanAm Village had positive trends during the same period. This finding indicates that the observed positive trends of wind speeds might be caused by the positive trends of the temperature gradients.

A wind resource assessment analysis based on the velocity data obtained from the trend analysis is performed using the CFD tool. The results show that the long-term wind speed changes do not affect the mean velocity field at PanAm Village as much as they influence the peak values. It has been demonstrated that randomly spaced and located objects in urban environments are factors that have to be accounted for in urban wind resource studies in changing climate.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship and/or publication of this article.

References

Black

Flarend

(2010) Alternative Energy (1st edn). Santa Barbara, CA: Greenwood.

Cheng

. (2012) Possible impacts of climate change on wind gusts under downscaled future climate conditions over Ontario, Canada. Journal of Climate 25(9): 3390–3408.

Cheng

Castro

(2002) Near wall flow over urban-like roughness. Boundary-Layer Meteorology 104(2): 229–259.

Cook

Nuccitelli

Green

. (2013) Quantifying the consensus on anthropogenic global warming in the scientific literature. Environmental Research Letters 8(2): 24024.

ESDU (2002) Strong Winds in the Atmospheric Boundary Layer: Pt. 2: Discrete Gust Speeds (83045). London: ESDU International PLC.

Franke

Hirsch

Jensen

. (2004) Recommendations on the use of CFD in wind engineering. Cost Action C 14: C1.

Hansen

Sato

Kharecha

. (2011) Earth’s energy imbalance and implications. Atmospheric Chemistry and Physics 11(24): 13421–13449.

Holt

Wang

(2012) Trends in wind speed at wind turbine height of 80 m over the contiguous United States using the North American regional reanalysis (NARR). Journal of Applied Meteorology and Climatology 51(12): 2188–2202.

Jubayer

Hangan

(2014) Numerical simulation of wind effects on a stand-alone ground mounted photovoltaic (PV) system. Journal of Wind Engineering and Industrial Aerodynamics 134: 56–64.

10.

Kalnay

Kanamitsu

Kistler

. (1996) The NCEP/NCAR 40-year reanalysis project. Bulletin of the American Meteorological Society 77(3): 437–471.

11.

Kaltschmitt

Streicher

Wiese

(2007) Renewable Energy – Technology, Economics and Environment (XXXII). Berlin; Heidelberg; New York: Springer. Available at: http://www.springer.com/us/book/9783540709473 (accessed 15 September 2014).

12.

Kastner-Klein

Rotach

(2004) Mean flow and turbulence characteristics in an urban roughness sublayer. Boundary-Layer Meteorology 111(1): 55–84.

13.

Kendall

(1970) Rank Correlation Methods. London: Griffin.

14.

McVicar

Roderick

Donohue

. (2012) Global review and synthesis of trends in observed terrestrial near-surface wind speeds: Implications for evaporation. Journal of Hydrology 416–417: 182–205.

15.

Mann

(1945) Nonparametric tests against trend. Econometrica 13(3): 245–259.

16.

Menter

Kuntz

Langtry

(2003) Ten years of industrial experience with the SST turbulence model. In: Proceedings of the fourth international symposium on turbulence, heat and mass transfer (Turbulence heat and mass transfer series), Antalya, 12–17 October, pp. 625–632. Begell House. Available at: http://cfd.mace.manchester.ac.uk/flomania/pds_papers/file_pds-1068134610Menter-SST-paper.pdf (accessed 23 September 2014).

17.

Milanković

(1920) Théorie mathématique des phénomènes thermiques produits par la radiation solaire. Paris: Gauthier-Villars et Cie.

18.

Milanković

(1930) Mathematische Klimalehre und astronomische Theorie der Klimaschwankungen. Berlin: Gebrüder Borntraeger.

19.

Pletcher

Tannehill

Anderson

(2011) Computational Fluid Mechanics and Heat Transfer (3rd edn). Boca Raton, FL: CRC Press.

20.

Romanić

Ćurić

Jovičić

. (2015a) Long-term trends of the ‘Koshava’ wind during the period 1949–2010. International Journal of Climatology 35(2): 288–302.

21.

Romanić

Rasouli

Hangan

(2015b) Wind resource assessment in complex urban environment. Wind Engineering 39(2): 193–212.

22.

Sen

(1968) Estimates of the regression coefficient based on Kendall’s Tau. Journal of the American Statistical Association 63(324): 1379–1389.

23.

Trenberth

(2009) An imperative for climate change planning: Tracking Earth’s global energy. Current Opinion in Environmental Sustainability 1: 19–27.

24.

Tyrlis

Lelieveld

(2013) Climatology and dynamics of the summer etesian winds over the eastern Mediterranean. Journal of the Atmospheric Sciences 70(11): 3374–3396.

25.

United Nations (2014) World Urbanization Prospects: The 2014 Revision. New York: United Nations. Available at: http://esa.un.org/unpd/wup/Highlights/WUP2014-Highlights.pdf (accessed 20 January 2015).

26.

Vautard

Cattiaux

Yiou

. (2010) Northern hemisphere atmospheric stilling partly attributed to an increase in surface roughness. Nature Geoscience 3(11): 756–761.