Classifying and zoning wind environments in metro station areas at the city scale: A case study of Nanjing

Abstract

While the “3D” principles―density, diversity, and design―remain central to Transit-Oriented Development (TOD), their effectiveness in metro station areas is increasingly constrained by the complex interplay of natural and built environmental factors. Among these, wind conditions are critical yet often overlooked. This study proposes a zoning optimization framework based on a dual-dimensional model combining wind speed (aerodynamic force) and the Frontal Area Index (FAI, morphological resistance). Using WRF/CALMET simulations and GIS-based modeling, localized wind data were obtained for 146 metro stations in Nanjing. K-means clustering and spatial autocorrelation analyses identified four representative wind environment types. Key findings: the mean station wind speed is 3.08 m/s―pedestrian-comfortable yet slightly below the city mean―and the mean FAI is 1.01, indicating generally high aerodynamic resistance. About 50% of stations fall within FAI 0.79—1.21; higher FAIs cluster in historic cores, elevated values occur in both flat and hilly terrains via different mechanisms, and lower values appear near large open spaces. Clustering yields four wind-environment types from “low-speed—high-FAI” to “high-speed—low-FAI.” The resulting wind-zoning map provides a robust basis for differentiated zoning and climate-responsive TOD planning.

Graphical abstract

Keywords

Metro station areas / Wind speed / Frontal Area Index (FAI) / Wind environment zoning

1 Introduction

1.1 Rethinking TOD guidelines in the context of urban wind environments

Since Calthorpe introduced the concept of Transit-Oriented Development (TOD) in 1993, developing mixed-use metro station areas―combining residential, commercial, and office functions―has become a common global approach (Calthorpe, 1993). In 1997, Cervero and colleagues proposed the classic “3D” principles of TOD: Density, Diversity, and Design (Cervero and Kockelman, 1997). Over time, these principles have become core guidelines for TOD planning, aiming to increase development intensity around stations, promote functional integration, and improve walkability to enhance transit efficiency.

Standard TOD principles are not always easily applicable across diverse natural and built environments. The wind environment is a notable example of such contextual complexity. In low-wind, enclosed areas, high building density can block natural ventilation corridors, causing heat buildup and poor air quality (Yang et al., 2020a). By contrast, in areas near rivers or lakes with strong natural ventilation, open space and pedestrian designs should make better use of terrain and water bodies to guide and enhance wind flow, ensuring continuous ventilation path (Yang et al., 2020b).

Therefore, TOD planning must be adjusted to local wind conditions to improve the built environment of metro station areas. This leads to a key research question addressed in this paper: In what ways can wind environment characteristics be systematically evaluated and zoned in metro station areas at the city scale to support the development of adaptive and context-sensitive TOD strategies?

1.2 Key metrics and evaluation approaches for analyzing urban wind environments

To address the aforementioned challenges, it is essential to first review the commonly employed indicators and evaluation methods used in planning practice for assessing urban wind environments.

(1)　Key Indicators and Assessment Framework for Urban Wind Environment in Planning

Urban wind environment indicators can generally be categorized into three main types. The first category comprises meteorological indicators, including wind speed, wind direction, wind frequency, and turbulence intensity. These fundamental parameters are typically obtained through meteorological observations or numerical simulations and serve as essential inputs for wind environment modeling and analysis. For instance, average wind speed indicates the ventilation capacity of a given area (Oke, 1988); prevailing wind direction (van Doorn et al., 2000) is crucial for identifying potential urban ventilation corridors; and wind frequency and turbulence intensity provide insight into the variability and stability of local airflow patterns (Zou et al., 2021). Moreover, these indicators are instrumental in defining boundary conditions and validating the accuracy of simulation outcomes.

The second category comprises morphology-based indicators, which assess the influence of three-dimensional urban form on airflow dynamics. These indicators typically encompass parameters such as FAI, building height variation, sky view factor (SVF) (Yuan and Chen, 2011), and porosity ratio (Bitog et al., 2011). They are used to quantify the degree to which the built form either obstructs or channels airflow, thereby contributing to the understanding of localized wind environment formation. Among these, FAI is one of the most widely adopted metrics for evaluating the effects of building orientation and spatial layout on wind flow, particularly in relation to pressure distribution and aerodynamic resistance on windward surfaces (Xu and Gao, 2022a). Originally introduced by Lettau (1969), FAI is defined as the ratio of the projected frontal area of buildings facing the wind direction to the corresponding ground surface area over which it is measured.

The third category includes comfort- and safety-oriented indicators, which focus on evaluating the impact of wind conditions on human experience and pedestrian-level safety. These typically involve parameters such as wind comfort thresholds (Janssen et al., 2013), wind hazard levels (Mudd et al., 2014), and pedestrian wind speed exceedance rates (Bu et al., 2009). Such indicators are commonly derived from empirical standards (e.g., Lawson criteria) (Lawson and Ad, 1977) or comfort models that define acceptable wind speed ranges for various pedestrian activities―such as sitting, standing, or walking (Du et al., 2022). These metrics are particularly important in urban design and public space planning, where ensuring wind comfort and minimizing wind-induced hazards are key to creating livable and safe environments.

Comprehensive urban wind environment evaluation frameworks enable multi-scale assessment of wind quality for diverse planning scenarios. Pedestrian-level comfort and safety are typically evaluated by combining field measurements and numerical simulations at 1.5—2 m height, with classification based on established comfort—safety criteria (Moonen et al., 2012; Du et al., 2017). Ventilation corridor identification integrates meteorological data and morphological parameters in GIS, coupling wind accessibility with “wind resistance surfaces” from building density, SVF, FAI, porosity, and height metrics, and extracting corridors via minimum-cost path models (Liu et al., 2022). In high-density areas, Hong Kong’s framework centers on the wind velocity ratio (VRw = Vp/V∞) to gauge building-induced wind attenuation (Ng et al. 2011), while Zhao et al.’s (2019) green building system, grounded in China’s Green Building Evaluation Standard (GB/T 50378-2014), employs key indicators such as SARVL and WVALR, incorporating background wind data and multi-direction simulations for systematic performance evaluation.

The essence of comprehensive wind environment evaluation frameworks lies in the judicious selection of indicators aligned with research objectives, the wind field characteristics of the study area (including spatial scale and prevailing wind patterns), and the specific application context in urban planning. For city-scale classification and zoning of wind environments in metro station areas, the primary challenge is identifying an evaluation framework that accommodates a dual-scale requirement. At the macro scale, this entails examining regional wind regimes and elucidating how topography and the broader urban spatial structure influence airflow distribution. At the micro scale, the emphasis shifts to assessing how morphological features―such as building configurations, street canyons, and open spaces―modulate local airflow dynamics and pedestrian-level wind conditions.

(2)　Evaluation Approaches for Assessing Urban Wind Conditions

The commonly used methods for evaluating urban wind environments can be broadly classified into three main categories. The first category is on-site measurement, which involves the direct collection of wind data through field instruments such as anemometers, ultrasonic wind sensors, and meteorological stations. Although this method offers high accuracy and reliability, its application is often constrained by significant costs and logistical complexities, making it unsuitable for large-scale evaluations across multiple metro station areas.

The second category is numerical simulation, which encompasses both mesoscale meteorological models and micro-scale Computational Fluid Dynamics (CFD) models. At the regional or city scale, mesoscale meteorological models―such as the Weather Research and Forecasting (WRF) model and the Regional Atmospheric Modeling System (RAMS)―are widely employed (Potty et al., 2012). These models typically operate at a spatial resolution of around 1 km × 1 km, enabling the simulation of broad-scale wind patterns influenced by topography, land use, and atmospheric dynamics (Steppeler et al., 2003; Wang and Chen, 2010; Jiménez and Dudhia, 2012). At a finer spatial scale, particularly in areas smaller than 1 km², CFD models are more commonly applied. These models provide high-resolution representations of micro-scale airflow patterns, capturing detailed wind behavior around individual buildings, street canyons, and other complex urban geometries. CFD simulations are particularly valuable in urban ventilation assessment, wind comfort analysis, and design optimization at the block level.

To improve the resolution and accuracy of urban wind environment simulations, researchers have increasingly adopted multi-scale modeling approaches that integrate regional-scale meteorological simulations with block-scale microscale models. This strategy enables the coupling of large-scale atmospheric dynamics with localized airflow patterns, thereby enhancing both spatial precision and contextual relevance. Representative examples of such composite modeling frameworks include KAMM/WAsP, WRF/WAsP, AROME/WAsP, WRF/CALMET, WRF/CFD, and WRF/LES (Deng et al., 2023; Díaz-Chávez et al., 2024).

(1) WRF/UCM and WRF/SLUCM methods: These approaches enhance the accuracy of mesoscale simulations by integrating the Urban Canopy Model (UCM) or the Single-Layer Urban Canopy Model (SLUCM) into the WRF framework. They rely heavily on high-precision urban surface and building morphological data. However, their spatial resolution typically remains around 1 km × 1 km, which limits their ability to capture fine-grained urban form and street-level wind variations (Blocken, 2015).

(2) WRF/CALMET method: This composite approach couples large-scale meteorological data generated by WRF with localized terrain and surface inputs within CALMET, enabling more efficient wind field downscaling. It can achieve spatial resolutions of up to 100 m × 100 m, allowing for reasonably detailed representation of local wind environments. However, CALMET’s simplified surface parameterization lacks explicit building morphology, which may reduce its reliability in dense urban settings (Deng et al., 2023).

(3) WRF/CFD methods: These methods involve performing separate Computational Fluid Dynamics (CFD) simulations for individual metro station areas, using WRF outputs as boundary conditions. They offer high spatial accuracy and strong adaptability for complex urban geometries. Nevertheless, the application of WRF/CFD across hundreds of station areas can be computationally intensive and data-demanding, posing significant challenges for large-scale implementation (Tang et al., 2021).

A comparative summary of these commonly used multi-scale modeling approaches―including their spatial resolution, data input requirements, and typical application contexts―is provided in Table 1.

The third category of method is parameterized evaluation based on GIS platforms. This approach relies on aerodynamic principles and parameterized analysis techniques to assess the urban wind environment using existing building footprint and terrain data rapidly and systematically (He et al., 2020). It employs indicators such as the FAI and SVF to directly quantify the influence of urban morphology on airflow obstruction and channeling. Unlike on-site measurements or large-scale numerical simulations, this method does not require costly instrumentation or intensive computational resources, offering advantages in efficiency, scalability, and cost-effectiveness. These characteristics make it particularly well-suited for evaluating the wind environments of large numbers of metro station areas simultaneously. However, GIS-based parameterization typically assumes a uniform background wind field, which limits its ability to account for local meteorological variability. This simplification implies that the method’s accuracy may be constrained when applied to complex or large-scale urban environments involving multiple metro station area areas.

1.3 Research gap and study contributions

Despite advances in urban wind environment research, few studies have developed scalable yet precise evaluation frameworks capable of accommodating the diverse spatial contexts of metro station areas. A key methodological challenge, therefore, is to establish a comprehensive wind environment evaluation framework and select an approach that balances scalability with analytical precision, enabling the classification of wind environment types and the formulation of context-specific planning strategies for multiple station areas across the city.

The main contributions of this study are as follows:

First, this study adopts wind speed and FAI as core indicators for evaluating the wind environment of metro station areas. Wind speed reflects the driving force of the wind field and directly determines ventilation performance, while FAI integrates factors such as micro-topography, building height, orientation, and spacing, providing an objective measure of how urban morphology interferes with airflow. Compared to indicators such as wind comfort coverage, wind speed and FAI are based on clear physical mechanisms, do not rely on subjective threshold settings, and offer greater consistency and operability in large-scale, multi-site studies.

Second, we develop an evaluation method that integrates WRF/CALMET mesoscale meteorological simulations with GIS-based parametric modeling. In this framework, WRF/CALMET generates localized wind field data for each metro station area, which are then applied as boundary conditions for calculating the FAI. This approach balances computational efficiency with spatial resolution, thereby enhancing the reliability of station-scale wind environment assessments. Moreover, it introduces an innovative weighted FAI that incorporates the distribution of inflow wind directions―fundamentally differing from the conventional FAI definition.

Third, using Nanjing as a case study, we identified several distinct wind environment types and examined their spatial distribution across the metro network. Based on these findings, we developed a wind environment zoning scheme for metro station areas at the city scale, aimed at supporting the adaptation of standard TOD planning principles to local microclimatic conditions.

2 Material and method

2.1 Case overview

Nanjing is located in the transitional zone between the Chang Jiang (Yangtze R.) floodplain and the Ning-Zhen-Yang hilly region, encompassing a variety of landforms including flat alluvial plains, low hills, and river valleys (Fig. 1). The city contains several major water bodies, such as the main channel of the Chang Jiang (Yangtze R.), Xuanwu Lake, and Qinhuai River, which together form typical waterfront ventilation corridors. At the junctions between hills and plains, extreme wind conditions often occur, including accelerated wind speeds and vortex accumulation. At the same time, with the rapid progress of urbanization, old and new urban areas in Nanjing have become increasingly intertwined. From historic districts in the old city to newly developed districts and high-density modern central business areas, the city exhibits a highly diverse urban form and a complex mix of land uses.

Nanjing’s first metro line began operations on May 15, 2005. As of October 2022, the Nanjing Metro system comprises 11 lines: Lines 1, 2, 3, 4, 10, S1, S3, S6, S7, S8, and S9. These lines collectively include 146 stations and extend over a total length of 429.1 km, servicing 11 districts in Nanjing (Fig. 2).

According to TOD theory, the metro station area is typically delineated with a radius of 500 m, extending to 800 m for interchange stations. This delineation considers the surrounding road network and current land use. An example of this process is illustrated in Fig. 3.

This study focuses on Nanjing because of its complex and locally variable wind environment, which results from a combination of distinctive climatic conditions, diverse topographic features, and a highly urbanized setting. These characteristics provide a representative context for exploring the ventilation and thermal comfort challenges faced by TOD planning under varying microclimatic conditions.

In terms of the temporal dimension, Nanjing is located in a subtropical monsoon climate zone and is predominantly influenced by southwesterly winds during the summer. In August 2022, the city recorded an average temperature of 31.1 °C, along with reduced precipitation and increased solar radiation. These factors intensified the urban heat island effect and placed additional stress on thermal comfort. Selecting August as the study window allows the proposed planning strategies, which are developed under extreme heat and high ventilation demand, to offer broader applicability in similar urban contexts.

2.2 Methodological framework

This study employs an integrated framework that combines meteorological numerical simulations with parametric analysis to evaluate the urban wind environment comprehensively. Initially, 30-m resolution Digital Elevation Model (DEM) and land use data are input into the WRF/CALMET modeling platform to simulate and diagnose the wind field at the municipal scale. Following this, high-resolution wind field outputs are processed using ArcGIS to calculate the FAI across the study area. By coupling wind speed data with the FAI, a composite wind environment evaluation model is developed, supporting the classification of metro station types according to their wind environment characteristics. Subsequently, spatial autocorrelation methods―specifically Join Count statistics and Local Moran’s I―are utilized to assess the clustering patterns, distribution trends, and hotspot areas of these station types. The results inform the delineation of wind environment zoning maps for metro station areas. The primary technical roadmap is illustrated in Fig. 4.

2.3 Data processing and simulation

2.3.1 WRF model

The data processing employs the WRF model, version 4.2. The initial fields and lateral boundary conditions are sourced from ERA5 reanalysis data. The analysis period was August 2022, with a 24-h spin-up preceding the analysis period. The model domain was centered at 31.95°N, 118.77°E, using a three-level one-way nested grid at 9 km, 3 km, and 1 km resolutions, with dimensions of 120 × 120, 166 × 166, and 154 × 323 grid points, respectively. The physics suite included Thompson microphysics (mixed-phase, single-/double-moment) (Thompson et al., 2004), RRTMG longwave/shortwave radiation (Iacono et al., 2008), the Noah-MP land surface model (Niu et al., 2011), the YSU nonlocal K-profile planetary boundary layer scheme (Hong and Lim, 2006), and the Kain—Fritsch deep-convection parameterization (Kain, 2004).

2.3.2 CALMET model

CALMET comprises a diagnostic wind-field module and a micrometeorological module (Scire et al., 2000). The initial-guess wind field was derived from the innermost (1 km) WRF domain and combined with routine surface and upper-air observations. CALMET ingested 30 m Digital Elevation Model (DEM) and GlobeLand30 land-use data, resampled to a 100 m computational grid (Table 2). Kinematic terrain-flow and slope-flow adjustments, together with terrain-blocking corrections, were applied to refine the preliminary wind field. Iterative procedures included: (1) spatial interpolation of mesoscale model outputs to the target grid; (2) smoothing with terrain adjustment to suppress spurious small-scale noise while preserving synoptic-scale structures; (3) vertical-velocity computation from the mass-continuity equation with terrain forcing; and (4) mass-consistent divergence minimization to ensure three-dimensional physical balance. These steps were repeated until convergence criteria were satisfied, yielding the final diagnostic wind field. Using surface observations and mesoscale model outputs, the micrometeorological module also calculates key boundary-layer parameters―surface heat flux, planetary boundary-layer height, friction velocity (u*), convective velocity scale (w*), and Monin—Obukhov length (L)―to characterize turbulence and atmospheric stability.

2.3.3 Extraction and processing of wind environment simulation results

The CALMET simulations were driven by 1 km-resolution WRF outputs, ensuring full retention of mesoscale meteorological features. Although topography and land cover data were originally at 30 m resolution, the wind field output was set to 100 m resolution to balance computational efficiency with spatial accuracy, and to avoid noise arising from terrain/land-surface resolution exceeding that of the driving meteorological fields.

Data from the WRF and CALMET models were processed using Python tools for reading, interpolation, and post-processing. Mean wind speeds for August were calculated, and hourly wind direction data for August were extracted. This wind field data was then processed using ArcGIS 10.4, resulting in 612,400 data points representing average wind speed across the Nanjing City area. These data points, containing average wind speed information, served as the raw data for spatial interpolation. The Kriging interpolation tool from the Geostatistical Analyst toolbox in ArcGIS 10.4 was employed to perform spatial interpolation of wind speed data, thus obtaining detailed wind speed information for both the Nanjing City area and the metro station areas.

2.4 FAI calculation based on WRF/CALMET simulations

Utilizing the hourly wind direction data for August simulated by the WRF-CALMET model, this study extracted wind direction data for station areas and calculated wind frequency data for 16 directions at each station.

Wind frequency refers to the percentage of the total number of observations of a particular wind direction in a given period of time. It is expressed as a percentage of the cumulative total number of occurrences of each of the different wind directions (including calm wind). When counting the frequency of various wind directions over the period of time. The calculation formula is as follows:

\begin{array}{}\text{(1)}& A_m=\frac{K_m}{{\displaystyle \sum _{m=1}^{16}{K}_m+C}}\mathrm{,}\end{array}

where A_m is the frequency of wind occurrence in the m-th direction;K_m is the number of times the wind has been recorded in the m-th direction;C is the number of occurrences of calm wind.

The FAI is defined as the total area of buildings projected into the plane normal to the approaching wind direction A_F divided by the plan area of the study site A_T (Wong et al., 2010). In this study, the FAI was calculated using the digital elevation model (DEM), building data, and meteorological data. And the 16 compass orientation method was used, where P_θ represents the wind frequency in the θ direction (Ng et al., 2011).

The formula for calculating the FAI is as follows (Grimmond and Oke, 1999; Burian et al., 2002; Yang et al., 2019):

\begin{array}{}\text{(2)}& FAI={\displaystyle \sum _{\theta =1}^N{\lambda }_{f(\theta )}\times P_{\theta }}\mathrm{,}\end{array}

where λ_f(θ) represents the ratio of the building’s windward area perpendicular to a particular wind direction P _θ to the area of the building plot. P_θ represents the frequency winds from direction θ. A larger λ_f(θ) indicates greater morphological (aerodynamic) resistance; a smaller value indicates weaker resistance. N represents 16 wind direction.

The directional FAI λ_f(θ) is defined as follows and illustrated in Fig. 5.

\begin{array}{}\text{(3)}& {\lambda }_{f(\theta )}=\frac{A_F}{A_T}=\frac{n\cdot b_{f(\theta )}{h}_{f(\theta )}}{A_T}\mathrm{.}\end{array}

A_T represents total plan area of the study site; A_F refers to the total area of each building or slope projection on the plane perpendicular to the incoming wind direction; b_f(θ) denotes the width of a façade unit’s projection on the plane perpendicular to wind direction θ. h_f(θ) denotes the height of a façade unit’s projection on the plane perpendicular to wind direction θ. n denotes the number of occupied façade units on the projection plane under wind direction θ. In the grid-based implementation, the study area is subdivided into fine units; each unit intersected by a building façade projection is counted as one effective unit, and the total number of such units yields n.

2.5 Cluster analysis and spatial correlation analysis

Based on the obtained average wind speed and FAI for each station area, this study will analyze their respective numerical characteristics and spatial distribution features. Kmeans clustering analysis will then be applied to comprehensively evaluate the wind environment of the station areas. The optimal number of clusters will be determined through preliminary calculations using the Elbow Method (SSE) and the Silhouette Coefficient method. In addition, Join Count statistics and Local Moran’s I will be calculated to analyze the spatial clustering patterns of different station types.

3 Results

3.1 Validation of simulation results

According to the WRF-CALMET simulation results, from August 1 to 31, 2022, the average wind speed within the Nanjing area was 3.34 m/s, with a maximum wind speed of 4.61 m/s and a minimum wind speed of 2.57 m/s. To quantitatively compare the simulation results with the observed data, the daily average wind speed data from the National Meteorological Data Center are used. The error statistics between the simulated daily average wind speeds (using the WRF-CALMET model) and the observed values are calculated and presented in a Taylor diagram. Additionally, due to the high randomness and fluctuation of wind speed on short time scales, wind speeds may exhibit significant variations even under the same meteorological conditions. This could lead to considerable fluctuations between the simulated and observed data on smaller time scales. Therefore, prior to accuracy analysis, outliers in the hourly wind speed data are identified and processed.

The Taylor diagram for the period from August 1 to 31, 2022, is based on observational data from five meteorological stations in Jiangsu Province, Nanjing: Lishui, Liuhe, Pukou, Nanjing, and Gaochun (Fig. 6). In the process of constructing the Taylor diagram, the standard deviation of the Nanjing station’s observed values is chosen as the reference standard deviation for normalizing the simulated values of the other stations, in order to align the comparison across multiple sites. In the Taylor diagram (Fig. 7), the radial distance from the origin represents the standard deviation (SD) of the simulated values, and the cosine of the angle between the line connecting the origin and the data point and the horizontal axis represents the correlation coefficient (R) between the simulated and observed data. The green arc indicates the root mean square error (RMSE) between the simulated and observed values, and the arc length from the point to the green origin (reference value) represents the RMSD (Root Mean Square Deviation) value for that particular point.

By comparing with the observed data, the Taylor diagram and comparison table (Table 3) for the 10-m wind speed (V₁₀) of the observed values and CALMET simulated values are presented. The correlation coefficients between the observed values and simulated values at the five meteorological stations are relatively high (all >0.4), indicating that the coupling of WRF/CALMET for wind field diagnosis at the city scale is reasonably effective. Among the five stations, the Nanjing meteorological station shows the best simulation performance, with a small root mean square error (RMSE) and a correlation coefficient (R) approaching 0.7, indicating a strong correlation. In contrast, the simulations for the Lishui and Pukou meteorological stations are relatively poor, which can be attributed to the complexities of the surface topography and local microclimate effects.

3.2 Overall wind environment characteristics of Nanjing

Influenced by the summer monsoon, the prevailing wind direction in August was primarily from the south and southeast (Fig. 7(a)). The citywide average wind speed was 3.35 m/s, with an average of 3.12 m/s in newly developed areas and 2.91 m/s in the old urban core. Overall, wind speeds in Nanjing show a typical spatial pattern, with lower values in the central urban areas and higher values in the suburban regions. Within the old city, wind speeds are generally lower, while the northern and southern parts of the city experience relatively higher wind speeds, revealing a gradual decrease from the outer areas toward the center (Fig. 7(b)).

The spatial distribution of wind speeds is closely related to both topography and the built environment. On one hand, areas such as the Chang Jiang (Yangtze R.) and Shijiu Lake, which lack urban structures and terrain obstructions, exhibit higher wind speeds. Locations near large water bodies, such as Chang Jiang (Yangtze R.) and Xuanwu Lake, also experience enhanced wind conditions due to the influence of river and lake breezes. These natural elements contribute to better wind environments compared to the more enclosed urban core. On the other hand, terrain elevation also plays a key role in shaping wind patterns. Higher-elevation areas tend to have slightly stronger wind speeds, as elevated terrain experiences less surface friction. In the central plains of Nanjing, where the area is surrounded by low mountains, wind speeds are significantly reduced. In built-up districts such as the central city, Lishui, and Gaochun, the combination of undulating terrain and dense construction further restricts airflow, leading to lower wind speeds in these urbanized zones.

3.3 Wind environment analysis of metro station areas

3.3.1 Wind speed features

An analysis of wind speed data from 146 metro stations across Nanjing reveals an average wind speed of 3.08 m/s (Fig. 8), which is slightly lower than the citywide mean yet exceeds the average wind speeds observed in newly developed districts. According to Yuan and Ng (2012), this wind speed falls within the range considered comfortable for pedestrians. Among all surveyed stations, Jiangxinzhou Station on Metro Line 10―situated on an island in the Chang Jiang (Yangtze R.)―records the highest wind speed at 4.09 m/s. Conversely, Minggugong Station on Metro Line 2, located in the eastern sector of the historic urban core near the southwestern slopes of Purple Mountain, exhibits the lowest wind speed at 2.67 m/s. The majority of stations fall within the 2.8—3.4 m/s range, with the distribution approximating a normal curve and displaying minimal variance. The median wind speed is approximately 3.0 m/s.

Spatially, the wind speed distribution in metro station areas exhibits the following characteristics (Fig. 9):

(1) Influence of the Built Environment: Metro station areas within the historic urban core generally exhibit lower average wind speeds compared to those located in outer suburban regions. This disparity is primarily attributed to the dense built environment and narrow street canyons typical of the old city, which significantly obstruct and diminish airflow. Representative stations within the historic core, uch as The Confucius Temple (V_mean = 2.78 m/s) and Daxinggong (V_mean = 2.81 m/s), demonstrate this trend …

(2) Proximity to Water Bodies: Metro stations located near large bodies of water tend to register higher average wind speeds. For instance, stations adjacent to the Chang Jiang (Yangtze R.), such as Jiangxinzhou (V_mean = 4.09 m/s), Lvboyuan (V_mean = 3.60 m/s), and Linjiang (V_mean = 3.46 m/s), and those near Xuanwu Lake, including Xuanwumen (V_mean = 3.35 m/s), consistently display elevated wind conditions. This can be attributed to the reduced presence of both natural and built obstructions in these areas, resulting in minimal wind resistance. Additionally, the smooth surface of large water bodies enables uninterrupted airflow and supports the formation of river breezes, which are driven by thermal differentials between land and water surfaces. However, this wind-enhancing effect is not as evident near smaller water bodies such as Mochou Lake and Jiulong Lake, where spatial and thermal scales are insufficient to generate significant wind amplification.

(3) Topographical Effects: Topography plays a crucial role in shaping local wind conditions. Stations situated on flat terrain―such as Nanjing Lukou International Airport Station (V_mean = 3.47 m/s) and Mozhoudonglu Station (V_mean = 3.43 m/s)―typically record higher average wind speeds compared to those in hilly or mountainous areas like Jiangwangmiao (V_mean = 3.12 m/s) and Xiaolingwei (V_mean = 2.70 m/s). In flat landscapes, the absence of substantial elevation changes allows for sustained, unobstructed wind flow. In contrast, the complex topography of hilly regions induces airflow disruptions―including updrafts, turbulence, and flow deflection―thereby reducing overall wind velocities.

3.3.2 FAI of metro station areas

(1)　WRF—CALMET-Derived Wind Frequency as Input for FAI Evaluation in Metro Station Zones

Based on hourly wind direction data simulated by the WRF—CALMET model for the month of August, this study calculates the prevailing wind frequency for each metro station area. Using Metro Line 10 as a case study, the wind frequency rose diagram demonstrates that the overall trend in wind direction across the stations closely aligns with the citywide baseline pattern (Fig. 10). Nonetheless, notable discrepancies emerge among stations oriented in different directions, highlighting localized variations. These directional differences offer critical boundary conditions that enhance the precision of FAI calculations.

(2)　FAI Characteristics of Nanjing Metro Station Areas

The overall wind environment across Nanjing’s metro station areas is marked by considerable aerodynamic resistance, as indicated by consistently high FAI values. According to Xu and Gao (2022b), FAI values less than or equal to 0.5 suggest minimal aerodynamic interference from urban morphology, whereas values exceeding 0.5 reflect increasing wind resistance due to build form characteristics. In this study, the average FAI across 146 metro station areas is 1.01. The highest FAI is recorded at Dachang Station (FAI = 1.98), while the lowest occurs at Wangjiawan Station (FAI = 0.31), reflecting substantial spatial variability. The overall distribution of FAI approximates a normal curve with moderate central tendency (Fig. 11): approximately 50% of the stations fall within the range of 0.79—1.21, with a median of 0.99.

From a spatial perspective, FAI values in Nanjing’s metro station areas exhibit the following characteristics (Fig. 12):

(1) Higher FAI Values in Historic Urban Cores: Older districts tend to exhibit elevated FAI values due to denser building configurations and more enclosed urban forms, which intensify airflow disruption. For example, stations such as Andemen (FAI = 1.57) and Shanghai Road (FAI = 1.56) are located in areas where surrounding buildings generate significant turbulence and aerodynamic interference. In contrast, stations in newly developed zones, such as Nanjing Lukou International Airport Station (FAI = 0.56), are situated in more open, low-density environments, resulting in reduced morphological resistance to wind.

(2) Elevated FAI in Both Flat and Hilly Terrains, Driven by Distinct Mechanisms: Stations in flat urban areas, such as Xinjiekou (FAI = 1.59), show high FAI values due to the concentration of high-rise, high-density development that significantly impedes wind flow. Conversely, in hilly regions such as Jubaoshan (FAI = 1.82) and Jiangwangmiao (FAI = 1.84), complex topographical conditions introduce additional aerodynamic disturbances, leading to increased wind resistance even in the absence of extreme building density.

(3) Lower FAI in Areas with Large Open Spaces: Stations surrounded by expansive plazas or green spaces generally report lower FAI values. For instance, Nanjing South Railway Station (FAI = 0.86) and Baijiahu Station (FAI = 0.70) benefit from nearby open areas that facilitate wind dispersion and reduce physical obstructions. These spatial configurations enhance local ventilation and contribute to a more favorable wind environment.

3.4 Classification of metro station types based on wind environment characteristics

To categorize the 146 metro stations in Nanjing according to their wind environment characteristics, this study applied the K-means clustering algorithm, focusing on metrics such as average wind speed and FAI (Fig. 13).

Cluster 1 is characterized by relatively “Low Wind Speed and Medium FAI” values. The average wind speed is 2.91 m/s, and the average FAI is 0.98, covering a total of 67 stations. These stations are mainly located at the centers of newly developed urban areas, often on the periphery of the historical core. Typical examples include Olympic Stadium East, Xianlin Center, and Xinglong Street. While these areas experience slightly higher wind speeds due to their open location outside the old city, the development intensity at the center of new towns tends to be high. As a result, their FAI remains moderate, indicating that building morphology imposes a relatively strong obstruction to airflow.

Cluster 2 features “the Highest average Wind Speed and the Lowest FAI” values among all clusters, with an average wind speed of 3.35 m/s and an average FAI of 0.64. This type includes 35 stations, typically located in newly developed areas or along the Chang Jiang (Yangtze R.). Unlike Type 1, the stations in Type 2 are mostly situated at the outer edges of new districts. These peripheral locations have even higher wind speeds and lower building density, resulting in less morphological obstruction and thus lower FAI values. Representative stations include Yuzui, Xiaolongwan, and Konggangxincheng.

Cluster 3 is composed of stations with “the Highest average Wind Speed and the Medium FAI” values. The average wind speed is 3.36 m/s, and the FAI is 1.09. This group includes 15 stations, typically located near rivers, lakes, hilly terrains, or in distant suburban areas. Representative examples include Xuanwumen, Zhongshan Lake, and Cuipingshan. The favorable wind conditions in these areas are primarily attributed to the presence of natural features such as water bodies and hills. However, the development of these naturally attractive areas also leads to increased building intensity, resulting in moderate obstruction to airflow.

Cluster 4 is defined by “Low Wind Speed and High FAI” values, with an average wind speed of 2.98 m/s and an average FAI of 1.51. This cluster consists of 29 stations, including typical examples such as Xinjiekou and Zhangfuyuan. These stations are mostly located in areas with dense urban development or complex terrain, both of which contribute to reduced wind speed and elevated wind resistance.

3.5 Spatial distribution of different clusters

Spatial analysis confirms that the distribution of all four metro station types in Nanjing exhibits statistically significant clustering, reflecting distinct non-random spatial patterns. To assess these patterns, a Univariate Local Moran’s I analysis was conducted using the GeoDa spatial analysis platform, with Local Indicators of Spatial Association (LISA) maps generated accordingly. As the original classification variable comprises four categorical station types, direct application of spatial autocorrelation methods could introduce bias or misinterpretation. To ensure methodological compatibility, the data were preprocessed by transforming the multi-class variable into a series of dummy variables, each assigning a value of 1 to stations of a specific type and 0 to all others. This approach converted the data into a Boolean format suitable for spatial econometric analysis, enabling the accurate identification of spatial clusters and local outliers.

To further interpret these patterns and synthesize the spatial clustering and local heterogeneity of station types, the LISA results were refined and systematically categorized to construct a targeted classification map (Fig. 14). This map highlights key spatial associations along metro lines. Stations identified as “High-High” were retained to denote areas of pronounced spatial concentration, serving as core clusters for each station type. “High-Low” stations were also included, as they represent isolated occurrences within generally unclustered zones, thus revealing local heterogeneity. In contrast, “Low-High” and “Low-Low” categories were excluded from the final visualization due to their limited interpretive value in understanding meaningful spatial aggregation or differentiation.

The resulting classification map reveals distinct spatial distributions and clustering intensities among the four station clusters across the Nanjing metropolitan area. Figure 14 illustrates the spatial hotspots, regional delineations, and overall distribution of each station cluster. Notably, Clusters 1 and 4 exhibit clear spatial concentrations, while Clusters 2 and 3 show relatively weaker clustering patterns. Cluster 1 stations are primarily concentrated in the southern part of the historic city and the northern bank of the Chang Jiang (Yangtze R.) (Pukou area), whereas Cluster 4 stations are mostly located within the dense built environment of the historic urban core.

4 Discussion and conclusion

This study investigated the wind environment characteristics of metro station areas in Nanjing and proposed a two-dimensional evaluation framework integrating wind speed (as an indicator of aerodynamic force) and FAI (as a measure of morphological resistance). By coupling mesoscale meteorological simulation (WRF/CALMET) with GIS-based spatial modeling, wind field data were derived for 146 metro station areas. K-means clustering and spatial autocorrelation analyses were employed to classify wind environment patterns, resulting in the identification of four representative station clusters. A spatially explicit zoning map was developed to support context-sensitive planning and design strategies.

4.1 Key findings and planning implications

The results reveal that the wind environment of metro station areas is jointly shaped by meteorological conditions, topography, and built form. Influenced by the summer monsoon, the prevailing wind directions in Nanjing during August are primarily from the south and southeast. The citywide average wind speed is 3.35 m/s, with 3.12 m/s in newly developed areas and 2.91 m/s in the old urban core, showing a typical spatial pattern of decreasing wind speed from the suburbs toward the city center. Higher wind speeds are mainly observed along the Chang Jiang (Yangtze R.), Shijiu Lake, Xuanwu Lake, and other waterfront areas as well as elevated terrains, while lower wind speeds are concentrated in the old city, Lishui, and Gaochun, where terrain undulation and dense built-up areas prevail.

Across the 146 metro stations, the average wind speed is 3.08 m/s with a median of approximately 3.00 m/s, and variance is small. Overall, metro station areas exhibit lower average wind speeds than the citywide level, lower than those of newly developed areas but higher than those of the old urban core. Stations near large water bodies generally experience higher wind speeds, whereas those at the foot of Zijin Mountain or within enclosed street canyons in the old city tend to have lower wind speeds. Stations located on flat terrain show higher wind speeds compared to those in hilly areas.

The FAI is generally high, with an average of 1.01 and a median of 0.99 across the 146 stations, and about 50% of the values fall within the range of 0.79—1.21. These values are generally higher than the mean of 0.5. Spatially, high-density old city areas and complex terrain zones present higher FAI values, while stations near large squares or green spaces exhibit lower FAI values. Overall, the wind environment of Nanjing metro station areas shows significant spatial differentiation: suburban, waterfront, and elevated stations are characterized by higher wind speeds and lower resistance, whereas old city and topographically complex areas suffer from lower wind speeds and higher resistance.

Cluster analysis delineates four wind environment types, spanning from “low speed—high FAI” to “high speed—low FAI,” each reflecting distinct spatial and morphological characteristics. This study also develops a wind environment zoning map for metro station areas in Nanjing, offering a technical foundation for implementing differentiated management and climate-responsive TOD planning. While traditional TOD principles emphasize high density, functional diversity, and pedestrian accessibility within a 500-m radius, this research suggests the need to adapt such strategies to local wind conditions.

In zones classified as “Low Wind Speed + High FAI” (Cluster 4―characterized by dense development and complex terrain―further intensification would likely worsen airflow obstruction. In these areas, TOD strategies should incorporate greater building setbacks, reduce building mass along prevailing wind directions, and optimize street canyon geometry to enhance ventilation. Conversely, “High Wind Speed + Low FAI” zones (Cluster 2, typically located at the urban periphery, possess abundant wind resources but lack spatial organization. These areas risk becoming underutilized “wind resource wastelands” without deliberate planning. Interventions should focus on establishing continuous ventilation corridors and adopting directional layouts that guide wind flow through urban fabric. Preserving key ventilation pathways around metro stations is also essential to maintain uninterrupted airflow transmission.

4.2 Contributions and innovation

This study substantiates three key contributions to wind environment research and metro station planning. First, it demonstrates the utility of wind speed and FAI as integrated indicators capturing both aerodynamic and morphological dimensions of urban ventilation performance. Their foundation in physical principles, combined with operational simplicity, makes them suitable for consistent application in large-scale spatial analyses.

Second, the integration of WRF/CALMET simulation with GIS-based parametric modeling enables accurate, localized wind assessments while preserving computational efficiency. This methodological synthesis strengthens the spatial applicability and technical reliability of wind environment evaluation frameworks.

Third, the classification of wind environment clusters across the Nanjing metro system, followed by the creation of a zoning map, provides a practical planning tool. It enables TOD strategies to be calibrated based on environmental variability, offering actionable guidance for optimizing urban form in relation to local wind conditions.

4.3 Limitations and future work

Despite its contributions, the study has several limitations. Wind field validation was conducted using data from five meteorological stations, which may be insufficient for capturing microclimatic variations, particularly within the dense central urban area where only one station is available. This constraint may limit the representativeness and robustness of model validation. Future research should incorporate more extensive on-site measurements and expand the spatial coverage of observational data to strengthen model calibration and validation. Additionally, exploring seasonal and diurnal wind variations could further enhance the applicability of the framework for year-round planning and operational management.

References

[1] Bitog, J., Lee, I.-B., Hwang, H.-S., Shin, M.-H., Hong, S.-W., Seo, I.-H., Mostafa, E., Pang, Z., 2011. A wind tunnel study on aerodynamic porosity and windbreak drag. For. Sci. Technol. 7 (1), 8—16.

[2] Blocken, B., 2015. Computational fluid dynamics for urban physics: importance, scales, possibilities, limitations and ten tips and tricks towards accurate and reliable simulations. Build. Environ. 91, 219—245.

[3] Bu, Z., Kato, S., Ishida, Y., Huang, H., 2009. New criteria for assessing local wind environment at pedestrian level based on exceedance probability analysis. Build. Environ. 44 (7), 1501—1508.

[4] Burian, S., Brown, M., Linger, S., 2002. Morphological Analyses Using 3D Building Databases. Los Angeles, California. LANL Report LA-UR-02-781.

[5] Calthorpe, P., 1993. The next American Metropolis: Ecology, Community & the American Dream. Princeton Architectural Press.

[6] Cervero, R., Kockelman, K., 1997. Travel demand and the 3Ds: density, diversity, and design. Transport. Res. Transport Environ. 2 (3), 199—219.

[7] Deng, X.W., Cao, Q., Wang, L.C., Wang, W., Wang, S., Wang, S.Q., Wang, L.Z., 2023. Characterizing urban densification and quantifying its effects on urban thermal environments and human thermal comfort. Landsc. Urban Plann. 237.

[8] Díaz-Chávez, L., Melendez-Surmay, R., Arregocés, H.A., 2024. Urban heat island intensity in coastal cities of northern Colombia using landsat data and WRF/UCM model. Case Stud. Chem. Environ. Eng. 9, 100617.

[9] Du, S., Zhang, X., Jin, X., Zhou, X., Shi, X., 2022. A review of multi-scale modelling, assessment, and improvement methods of the urban thermal and wind environment. Build. Environ. 213, 108860.

[10] Du, Y., Mak, C.M., Kwok, K., Tse, K.-T., Lee, T.-C., Ai, Z., Liu, J., Niu, J., 2017. New criteria for assessing low wind environment at pedestrian level in Hong Kong. Build. Environ. 123, 23—36.

[11] Grimmond, C.S.B., Oke, T.R., 1999. Aerodynamic properties of urban areas derived from analysis of surface form. J. Appl. Meteorol. 38 (9), 1262—1292.

[12] He, B.J., Ding, L., Prasad, D., 2020. Urban ventilation and its potential for local warming mitigation: a field experiment in an open low-rise gridiron precinct. Sustain. Cities Soc. 55.

[13] Hong, S.-Y., Lim, J.-O.J., 2006. The WRF single-moment 6-class microphysics scheme (WSM6). Asia-Pacific J. Atmos. Sci. 42 (2), 129—151.

[14] Iacono, M.J., Delamere, J.S., Mlawer, E.J., Shephard, M.W., Clough, S.A., Collins, W.D., 2008. Radiative forcing by long-lived greenhouse gases: calculations with the AER radiative transfer models. J. Geophys. Res. Atmos. 113 (D13).

[15] Janssen, W.D., Blocken, B., Van Hooff, T., 2013. Pedestrian wind comfort around buildings: comparison of wind comfort criteria based on whole-flow field data for a complex case study. Build. Environ. 59, 547—562.

[16] Jiménez, P.A., Dudhia, J., 2012. Improving the representation of resolved and unresolved topographic effects on surface wind in the WRF model. J. Appl. Meteorol. Climatol. 51 (2), 300—316.

[17] Kain, J.S., 2004. The Kain—Fritsch convective parameterization: an update. J. Appl. Meteorol. 43 (1), 170—181.

[18] Lawson, T., Ad, P., 1977. The Effects of Wind on People in the Vicinity of Buildings. Cambridge University Press, pp. 605—622.

[19] Lettau, H., 1969. Note on aerodynamic roughness-parameter estimation on the basis of roughness-element description. J. Appl. Meteorol. Climatol. 8 (5), 828—832.

[20] Liu, X., Huang, B., Li, R., Zhang, J., Gou, Q., Zhou, T., Huang, Z., 2022. Wind environment assessment and planning of urban natural ventilation corridors using GIS: shenzhen as a case study. Urban Clim. 42, 101091.

[21] Moonen, P., Defraeye, T., Dorer, V., Blocken, B., Carmeliet, J., 2012. Urban physics: effect of the micro-climate on comfort, health and energy demand. Front. Architech. Res. 1 (3), 197—228.

[22] Mudd, L., Wang, Y., Letchford, C., Rosowsky, D., 2014. Hurricane wind hazard assessment for a rapidly warming climate scenario. J. Wind Eng. Ind. Aerod. 133, 242—249.

[23] Ng, E., Yuan, C., Chen, L., Ren, C., Fung, J.C.H., 2011. Improving the wind environment in high-density cities by understanding urban morphology and surface roughness: a study in Hong Kong. Landsc. Urban Plann. 101 (1), 59—74.

[24] Niu, G.Y., Yang, Z.L., Mitchell, K.E., Chen, F., Ek, M.B., Barlage, M., Kumar, A., Manning, K., Niyogi, D., Rosero, E., 2011. The community Noah land surface model with multiparameterization options (Noah-MP): 1. Model description and evaluation with local-scale measurements. J. Geophys. Res. Atmos. 116 (D12).

[25] Oke, T.R., 1988. Street design and urban canopy layer climate. Energy Build. 11 (1), 103—113.

[26] Potty, J., Oo, S.M., Raju, P.V.S., Mohanty, U.C., 2012. Performance of nested WRF model in typhoon simulations over West Pacific and South China Sea. Nat. Hazards 63 (3), 1451—1470.

[27] Scire, J., Robe, F., Fernau, M., Yamartino, R., 2000. A User’s Guide for the CALMET Meteorological Model (Version 5). Earth Tech. Inc., 01741. Concord, MA, USA.

[28] Steppeler, J., Doms, G., Schättler, U., Bitzer, H.W., Gassmann, A., Damrath, U., Gregoric, G., 2003. Meso-gamma scale forecasts using the nonhydrostatic model LM. Meteorol. Atmos. Phys. 82 (1—4), 75—96.

[29] Tang, S., Huang, S., Yu, H., Gu, M., Tang, J., 2021. Impact of horizontal resolution in CALMET on simulated near-surface wind fields over complex terrain during Super Typhoon Meranti (2016). Atmos. Res. 247.

[30] Thompson, G., Rasmussen, R.M., Manning, K., 2004. Explicit forecasts of winter precipitation using an improved bulk microphysics scheme. Part I: description and sensitivity analysis. Mon. Weather Rev. 132 (2), 519—542.

[31] Van Doorn, E., Dhruva, B., Sreenivasan, K.R., Cassella, V., 2000. Statistics of wind direction and its increments. Phys. Fluids 12 (6), 1529—1534.

[32] Wang, W.Y., Chen, N., 2010. WRF/CALMET-Based wind field simulation of mountain regions. J. Southwest Jiaot. Univ. 45 (6), 990—996.

[33] Wong, M.S., Nichol, J.E., To, P.H., Wang, J., 2010. A simple method for designation of urban ventilation corridors and its application to urban heat island analysis. Build. Environ. 45 (8), 1880—1889.

[34] Xu, F., Gao, Z., 2022a. Frontal area index: a review of calculation methods and application in the urban environment. Build. Environ. 224, 109588.

[35] Xu, F.S., Gao, Z., 2022b. Frontal area index: a review of calculation methods and application in the urban environment. Build. Environ. 224.

[36] Yang, J., Shi, B., Shi, Y., Marvin, S., Zheng, Y., Xia, G., 2020a. Air pollution dispersal in high density urban areas: research on the triadic relation of wind, air pollution, and urban form. Sustain. Cities Soc. 54, 101941.

[37] Yang, J., Wang, Y., Xiao, X., Jin, C., Xia, J., Li, X., 2019. Spatial differentiation of urban wind and thermal environment in different grid sizes. Urban Clim. 28, 100458.

[38] Yang, L., Liu, X., Qian, F., 2020b. Research on water thermal effect on surrounding environment in summer. Energy Build. 207, 109613.

[39] Yuan, C., Chen, L., 2011. Mitigating urban heat island effects in high-density cities based on sky view factor and urban morphological understanding: a study of Hong Kong. Archit. Sci. Rev. 54 (4), 305—315.

[40] Yuan, C., Ng, E., 2012. Building porosity for better urban ventilation in high-density cities - a computational parametric study. Build. Environ. 50, 176—189.

[41] Zhao, S., Li, Q., You, X.-Y., 2019. A Comprehensive Index for Wind Environment Evaluation of Green Building Arrangement. IOP Publishing, 022071.

[42] Zou, J., Yu, Y., Liu, J., Niu, J., Chauhan, K., Lei, C., 2021. Field measurement of the urban pedestrian level wind turbulence. Build. Environ. 194, 107713.

Rights & permissions

2095-2635/2025 The Authors. Publishing services by Elsevier B.V. on behalf of KeAi Communications Co. Ltd.