Analysis of Urban Effects in Oklahoma City using a Dense Surface Observing Network

Xiao-Ming Hu Center for Analysis and Prediction of Storms, and School of Meteorology, University of Oklahoma, Norman, Oklahoma

Search for other papers by Xiao-Ming Hu in
Current site
Google Scholar
PubMed
Close
,
Ming Xue Center for Analysis and Prediction of Storms, and School of Meteorology, University of Oklahoma, Norman, Oklahoma

Search for other papers by Ming Xue in
Current site
Google Scholar
PubMed
Close
,
Petra M. Klein Center for Analysis and Prediction of Storms, and School of Meteorology, University of Oklahoma, Norman, Oklahoma

Search for other papers by Petra M. Klein in
Current site
Google Scholar
PubMed
Close
,
Bradley G. Illston Oklahoma Mesonet, Oklahoma Climatological Survey, University of Oklahoma, Norman, Oklahoma

Search for other papers by Bradley G. Illston in
Current site
Google Scholar
PubMed
Close
, and
Sheng Chen Key Laboratory of Beibu Gulf Environmental Evolution and Resources Utilization, Guangxi Teachers Education University, Ministry of Education, Nanning, Guangxi, China, and Hydrometeorology and Remote Sensing Laboratory, and School of Civil Engineering and Environmental Science, University of Oklahoma, Norman, Oklahoma

Search for other papers by Sheng Chen in
Current site
Google Scholar
PubMed
Close
Full access

Abstract

Many studies have investigated urban heat island (UHI) intensity for cities around the world, which is normally quantified as the temperature difference between urban location(s) and rural location(s). A few open questions still remain regarding the UHI, such as the spatial distribution of UHI intensity, temporal (including diurnal and seasonal) variation of UHI intensity, and the UHI formation mechanism. A dense network of atmospheric monitoring sites, known as the Oklahoma City (OKC) Micronet (OKCNET), was deployed in 2008 across the OKC metropolitan area. This study analyzes data from OKCNET in 2009 and 2010 to investigate OKC UHI at a subcity spatial scale for the first time. The UHI intensity exhibited large spatial variations over OKC. During both daytime and nighttime, the strongest UHI intensity is mostly confined around the central business district where land surface roughness is the highest in the OKC metropolitan area. These results do not support the roughness warming theory to explain the air temperature UHI in OKC. The UHI intensity of OKC increased prominently around the early evening transition (EET) and stayed at a fairly constant level throughout the night. The physical processes during the EET play a critical role in determining the nocturnal UHI intensity. The near-surface rural temperature inversion strength was a good indicator for nocturnal UHI intensity. As a consequence of the relatively weak near-surface rural inversion, the strongest nocturnal UHI in OKC was less likely to occur in summer. Other meteorological factors (e.g., wind speed and cloud) can affect the stability/depth of the nighttime boundary layer and can thus modulate nocturnal UHI intensity.

Supplemental information related to this paper is available at the Journals Online website: http://dx.doi.org/10.1175/JAMC-D-15-0206.s1.

Corresponding author address: Dr. Xiao-Ming Hu, Center for Analysis and Prediction of Storms, University of Oklahoma, 120 David L. Boren Blvd., Norman, OK 73072. E-mail: xhu@ou.edu

Abstract

Many studies have investigated urban heat island (UHI) intensity for cities around the world, which is normally quantified as the temperature difference between urban location(s) and rural location(s). A few open questions still remain regarding the UHI, such as the spatial distribution of UHI intensity, temporal (including diurnal and seasonal) variation of UHI intensity, and the UHI formation mechanism. A dense network of atmospheric monitoring sites, known as the Oklahoma City (OKC) Micronet (OKCNET), was deployed in 2008 across the OKC metropolitan area. This study analyzes data from OKCNET in 2009 and 2010 to investigate OKC UHI at a subcity spatial scale for the first time. The UHI intensity exhibited large spatial variations over OKC. During both daytime and nighttime, the strongest UHI intensity is mostly confined around the central business district where land surface roughness is the highest in the OKC metropolitan area. These results do not support the roughness warming theory to explain the air temperature UHI in OKC. The UHI intensity of OKC increased prominently around the early evening transition (EET) and stayed at a fairly constant level throughout the night. The physical processes during the EET play a critical role in determining the nocturnal UHI intensity. The near-surface rural temperature inversion strength was a good indicator for nocturnal UHI intensity. As a consequence of the relatively weak near-surface rural inversion, the strongest nocturnal UHI in OKC was less likely to occur in summer. Other meteorological factors (e.g., wind speed and cloud) can affect the stability/depth of the nighttime boundary layer and can thus modulate nocturnal UHI intensity.

Supplemental information related to this paper is available at the Journals Online website: http://dx.doi.org/10.1175/JAMC-D-15-0206.s1.

Corresponding author address: Dr. Xiao-Ming Hu, Center for Analysis and Prediction of Storms, University of Oklahoma, 120 David L. Boren Blvd., Norman, OK 73072. E-mail: xhu@ou.edu

1. Introduction

Because of the different properties of the urban surface, temperatures over urban areas are typically higher than over the surrounding rural areas, a phenomenon well known as the urban heat island (UHI; Oke 1976, 1981, 1982; Arnfield 2003). During the past few decades, many studies have been conducted to observe and document UHI magnitude/intensity in cities around the world. UHI intensity is often quantified as the difference in temperature of near-surface air or land surface between urban site(s) and surrounding rural site(s). Most of these previous studies were limited to discussing paired urban–rural differences, either between one urban and one rural site or between urban mean and rural mean temperature (e.g., Gedzelman et al. 2003; Wienert and Kuttler 2005; Yow and Carbone 2006; Alonso et al. 2007; Basara et al. 2008; Miao et al. 2009; Steeneveld et al. 2011; Husain et al. 2014; Hu and Xue 2016).

Relative to its temporal characteristics, the spatial characteristics of UHI intensity (UHII) are much less investigated, largely due to difficulties and costs of deploying multiple instruments with enough density across the urban area (Basara et al. 2010; Chapman et al. 2015). Because the in situ weather stations in urban areas are usually sparse (Kim and Baik 2005; Stone 2007; Tan et al. 2010; Basara et al. 2011; Muller et al. 2013; Fenner et al. 2014), UHII at the spatial scale of a city is sometimes derived from remotely sensed land surface temperatures (Voogt and Oke 2003; Fung et al. 2009; Nichol et al. 2009; Zhou and Shepherd 2010; Peng et al. 2012; Winguth and Kelp 2013). However, there are inherent issues associated with remotely sensed UHII. Satellite-derived land surface temperature only account for the radiant temperatures of surfaces seen by the radiometer, and the data correspond to averages across the area of a pixel (Roth et al. 1989). As a consequence, roofs, treetops, and open horizontal areas are oversampled and vertical surfaces and areas below tree crowns are neglected in remotely sensed UHII. The physical properties, and radiative and turbulent environments of facets that are well represented by remote sensing data, often differ from those that are undersampled (Arnfield 2003). Thus, the remote sensing poorly samples the true surface temperature within the city. As a result, remotely sensed land surface temperature UHII and in situ measured air temperature UHII can have very distinctly different behaviors, such as different diurnal variation (Cui and de Foy 2012). While air temperature UHII is reported to be stronger at night than in the day in many studies (Arnfield 2003; Souch and Grimmond 2006), remotely sensed land surface UHI shows stronger intensity during daytime than during nighttime (Roth et al. 1989; Imhoff et al. 2010; Schwarz et al. 2011; Jin 2012; Klok et al. 2012; Peng et al. 2012; Zhao et al. 2014). Thus, further investigation of spatial and temporal characteristics of UHII using consistent, quality observations of air temperature is required to improve the understanding of the impacts of urbanization (Bottyán and Unger 2003; Grimmond 2006; Huang et al. 2008; Basara et al. 2010; Grimmond et al. 2010; Chen et al. 2012; Muller et al. 2013; Schatz and Kucharik 2014; Dou et al. 2015). The information regarding spatial distribution of air temperature UHI of a city can also help provide heat information at a neighborhood scale for use in future detailed public health analyses and heat hazard mitigation strategies (Heusinkveld et al. 2014; Chapman et al. 2015; Debbage and Shepherd 2015).

Several studies have shown that UHII is dictated by the intrinsic characteristics of a city (Oke 1981, 1982; Unger 2004; Grimmond 2007; Hart and Sailor 2009; Georgakis et al. 2010; Ryu and Baik 2012; Adachi et al. 2014; Barlow 2014) and modulated by external meteorological factors (Morris and Simmonds 2000; Hu et al. 2013b). Oke et al. (1991) used a simple energy balance model to assess the relative importance of the commonly stated intrinsic causes of UHI under calm and cloudless conditions, including anthropogenic heat, thermal properties/moisture availability of the materials of the city, street canyon geometry, and urban greenhouse gases. The first three of these were identified as the main intrinsic causative factors contributing to the UHII in a modeling study conducted by Ryu and Baik (2012). A quantitative attribution of various contributions to UHII is estimated in Zhao et al. (2014) using a surface energy balance analysis. Low efficiency at dissipating heat from urban surfaces due to larger aerodynamic resistance is diagnosed to be the dominant contributor to daytime UHII in cities in the humid southeast and south-central United States (including Oklahoma), which coincides roughly with the Koppen–Geiger temperate climate zone. Improved understanding of causative factors contributing to the UHII is still needed for UHI mitigation management (Hidalgo et al. 2010; Loughner et al. 2012; Clinton and Gong 2013; Li et al. 2013; Theeuwes et al. 2013; Li et al. 2014; Zhao et al. 2014).

Conflicting results have been reported for the seasonal variation of UHII (Arnfield 2003; Yang et al. 2013). Using the air temperature–defined UHI, some previous studies (e.g., Magee et al. 1999; Steinecke 1999; Montávez et al. 2000; Jonsson 2004; Kim and Baik 2005; Hinkel and Nelson 2007; Zhou and Shepherd 2010; Memon et al. 2011; Yang et al. 2013) reported that UHIs are the weakest in summer and strongest in winter, while other studies found that UHIs are best developed in summer or warm half of the year (Karl et al. 1988; Schmidlin 1989; Klysik and Fortuniak 1999; Philandras et al. 1999; Morris et al. 2001; Fortuniak et al. 2006; Camilloni and Barrucand 2012; Fenner et al. 2014; Schatz and Kucharik 2014; van Hove et al. 2015). Most studies based on analysis of satellite-derived land surface temperature reported that UHII over many cities was significantly higher in summer than in winter, for example, Imhoff et al. (2010) and Peng et al. (2012) for cities of the United States, K. C. Wang et al. (2007) and Zhang et al. (2005) for Beijing, Li et al. (2012) for Shanghai, Meng and Liu (2013) for Jinan, and Zhang et al. (2010) for other cities. Factors governing the seasonal variation of UHII need further investigation.

Oklahoma City (OKC), Oklahoma (35.468 762°N, 97.516 304°W), spans ~1610 km2. It is one of the 10 largest cities by land area in the United States. The terrain of OKC is quite flat. The temporal variation of UHII in OKC has been investigated in observational studies (Basara et al. 2008; Basara et al. 2010; Klein 2012) for a few short-term periods (mostly for July 2003). Studies with research and weather prediction models further focused on evaluating the skill of these models in predicting UHII with urban canopy layer parameterization schemes of different levels of complexity (Liu et al. 2006; Lemonsu et al. 2009; Hu et al. 2013b; Husain et al. 2013). During 2007 and 2008, a dense network of atmospheric monitoring sites [i.e., the OKC Micronet (OKCNET)] was deployed across the OKC metropolitan area (Basara et al. 2011), which provided data for long-term monitoring of UHII in OKC at the spatial scale of the city. Thus, given the dense meteorological observations across the OKC metropolitan area, the objectives of this study are to 1) illustrate the spatial pattern of UHII in the OKC metropolitan area, 2) investigate the diurnal and seasonal variations of UHII in OKC, and 3) diagnose how UHI formation is related to land surface and boundary layer processes. Since previous studies have shown that UHII is closely related to wind speed (Fast et al. 2005; Hu et al. 2013b), some of its spatiotemporal characteristics are also discussed.

2. Data and methods

Our analysis focuses on the OKC metropolitan area (Fig. 1). Embedded within the OKC urbanized area is a well-defined central business district (CBD) that spans ~20 km2, with the average building height being around 50–70 m and the tallest building being 152 m high (during the years studied). This study utilized surface meteorological data collected at the stations around the OKC metropolitan area from two networks, that is, the OKCNET and the Oklahoma Mesonet (site locations in Fig. 1; Brock et al. 1995; McPherson et al. 2007).

Fig. 1.
Fig. 1.

(a) Location of the OKCNET stations (red dots) and seven surrounding Oklahoma Mesonet sites (i.e., ELRE, GUTH, KIN2, MINC, NRMN, WASH, and SPEN in blue dots). The CBD of OKC is also marked. The background shade corresponds to the land-use categories derived from the USGS 2006 NLCD at a spatial resolution of 30 m, in which the urban land use was divided into three categories: low-intensity residential (category 31), high-intensity residential (category 32), and commercial/industrial (category 33). In the CBD, symbols overlap due to the high spatial density of the OKCNET stations. (b) Spatial distribution of built-up area fraction also derived from the NLCD.

Citation: Journal of Applied Meteorology and Climatology 55, 3; 10.1175/JAMC-D-15-0206.1

The OKCNET is an operational network designed to improve atmospheric monitoring across the OKC metropolitan area, which was officially commissioned in November 2008 (Basara et al. 2011). OKCNET consists of a total of 39 stations with an average spacing of approximately 3 km, including 36 stations mounted on traffic signals at a height of ~9 m above ground level (AGL) and 3 stations in OKC (i.e., OKCN, OKCW, and OKCE) deployed following the protocols of the Oklahoma Mesonet. Basara et al. (2010, 2011) provided detailed information about the siting of the OKCNET sites and their classification using different criteria proposed in the literature. Basara et al. (2010) used a simplified classification for a composite analysis of UHI in OKC for a heat wave episode, in which OKCNET sites are grouped into three categories (i.e., urban, suburban, and rural) based on the surrounding land cover characteristics. A more detailed climate-based classification of urban sites, that is, seven urban climate zones (UCZs), was proposed by Oke (2004). The urban OKCNET stations fall into UCZ classifications UCZ1 and UCZ2 (intensely developed urban), the suburban OKCNET stations fall into UCZ classifications UCZ4–UCZ7 (highly developed, medium, or low-density urban, suburban to semirural), and the rural OKCNET stations fall into UCZ classification UCZ7 (semirural; Basara et al. 2010; Basara et al. 2011).

The Oklahoma Mesonet is a rural network of 120 meteorological stations with minimal influences from urban landscapes (McPherson et al. 2007; Basara et al. 2008). Each of the Oklahoma Mesonet stations is located within a fenced 100 m2 plot of land and measures more than 20 environmental variables, including wind at 2 and 10 m, air temperature at 1.5 and 9 m AGL, and shortwave radiation, that are used in the analysis of this study. The average temperature at 9 m AGL at the seven Oklahoma Mesonet sites surrounding the OKC metropolitan area [i.e., El Reno (ELRE), Guthrie (GUTH), Kingfisher (KIN2), Minco (MINC), Norman (NRMN), Washington (WASH), and Spencer (SPEN); see Fig. 1] are calculated as background rural temperature following the approach of Basara et al. (2008) and Klein (2012). The UHI intensity at each OKCNET site is defined as the difference between the temperature at the OKCNET site and the background rural temperature. Note that the average elevation difference between the OKCNET sites (≈370 m above sea level) and the seven rural Oklahoma Mesonet sites (≈363 m above sea level) is approximately 7 m. Using mean temperature at the seven Oklahoma Mesonet sites as background rural temperature provides a more robust measure of UHI intensity and minimizes the inherent variability between rural sites that can impact the magnitude of UHI values (Hawkins et al. 2004; Sakakibara and Owa 2005; Hunt et al. 2007; Lee and Baik 2010; Mohsin and Gough 2012). For example, as a result of the urbanization in recent years, the NRMN station was at the edge of an urban area according to the U.S. Geological Survey (USGS) 2006 National Land Cover Data (NLCD; Fig. 1), where the measured rural temperature may be biased high. On the other hand, the ELRE site to the west of the metropolitan area is known to experience more rapid nighttime cooling than other nearby Oklahoma Mesonet sites and often has a low bias (Hunt et al. 2007).

Richardson number Ri, an indicator of dynamic stability of an air layer, considers the effects of both wind shear and buoyancy on the atmospheric stability. To investigate the diurnal cycle of near-surface atmospheric stability and its impact on UHI development, the Ri at Oklahoma Mesonet sites around OKC was estimated using (Bodine et al. 2009)
e1
where g is the acceleration due to gravity, Γd = 0.01°C m−1 is the dry adiabatic lapse rate, T9m and T1.5m are the air temperatures measured at 9 and 1.5 m AGL, and u2m and u10m are the wind speed at 2 and 10 m AGL, respectively. The height differences between the measurement levels are ΔzT = 7.5 m for air temperature and Δzu = 8.0 m for wind speed.

One of the surrounding Oklahoma Mesonet sites, KIN2, was deployed in March 2009. A few OKCNET sites were decommissioned in November 2010, that is, reliable temperature data over the urban area were no longer available from thereon. Thus, the investigation of urban effects focuses on the period of April 2009–October 2010 in this study. Precipitation has been reported to reduce the difference in the heating/cooling rate between the urban and rural areas, thus suppressing UHI development (Chow and Roth 2006; Lee and Baik 2010). Time periods with precipitation at any OKCNET sites or the seven surrounding Oklahoma Mesonet sites were not considered in this study to avoid the impact of precipitation on UHII. Since the spatial distribution of UHII is to be examined, all the sites need to have the same measurement period; so only the time periods when data from all the sites were available between April 2009 and October 2010 (total 6937 h) were considered when investigating urban–rural differences. On the other hand, the Oklahoma Mesonet has been continuously operational. The temperature inversion between 1.5 and 9 m AGL at the seven Oklahoma Mesonet sites surrounding OKC were examined for a longer period (i.e., April 2009–December 2012) to diagnose possible seasonal variations of UHI intensity.

In addition to values at spatially distributed observation sites, spatial patterns of UHII and wind speed were analyzed using the kriging interpolation method. The kriging method is popular in mapping meteorological and chemical variables based on station observation data in diverse applications (e.g., Chen et al. 2014; Smoliak et al. 2015; Zou et al. 2015). The kriging function embedded in the Interactive Data Language (IDL, version 8.4) was used in this study as in Chen et al. (2014).

3. Results

a. Mean spatial distributions and diurnal variations of UHI and wind speed

1) UHI intensity and relationship with spatial distributions of wind speed

The spatial distributions of UHII and wind speed are overlaid on top of land-use categories in Fig. 2. The strongest UHI intensity was mostly confined around the CBD during both daytime and nighttime except for a hotspot of OKCNET site KNW103 during nighttime (Fig. 2b). The spatial maps of UHI intensity produced by the kriging method illustrate elongated heat plumes north of the CBD during both daytime and nighttime (Figs. 3a,b), which can be explained by downwind transport of heat by the predominant southerly winds.

Fig. 2.
Fig. 2.

Spatial distribution of (a),(b) mean UHII (defined as the difference between temperature at each OKCNET site and rural background temperature computed as average temperature at the seven Oklahoma Mesonet sites) and (c),(d) wind speed during (top) daytime (i.e., 0900–1700 CST) and (bottom) nighttime (i.e., 2200–0500 CST) in April 2009–October 2010. The background shade shows again the land-use categories derived from the 2006 NLCD. In the CBD, not all station names are shown.

Citation: Journal of Applied Meteorology and Climatology 55, 3; 10.1175/JAMC-D-15-0206.1

Fig. 3.
Fig. 3.

Spatial patterns of (a),(b) mean UHII and (c),(d) wind speed computed with a kriging interpolation method (using the data shown in Fig. 2) during (top) daytime and (bottom) nighttime in April 2009–October 2010. As reference, the location of some of the sites is indicated by the station name.

Citation: Journal of Applied Meteorology and Climatology 55, 3; 10.1175/JAMC-D-15-0206.1

While urban effects on temperature are quite widely discussed in the literature, less information is available about urban effects on wind patterns (Klein 2012; Klein and Galvez 2014). Spatial patterns of mean wind speed around OKC were examined using the OKCNET observations (Figs. 2c,d). The corresponding analyses based on the kriging method are also shown in Figs. 3c and 3d. Unlike the elongated feature of UHI intensity, wind speeds showed a more concentric pattern around the CBD. Wind speed was slowest during both daytime and nighttime around the CBD, where the UHI intensity was the strongest. Having the largest wind speed reduction centered around the CBD suggests a strong impact of surface roughness that is largest in this area. Roughness also affects the aerodynamic resistance of sensible heat transfer. Larger (smaller) roughness leads to stronger (weaker) vertical mixing and more (less) efficiency to transfer sensible heat from the land surface to the atmospheric boundary layer (ABL) (Lee et al. 2011).

In the humid southeast and south-central United States (including Oklahoma), the rural land is densely vegetated, owing to ample precipitation, and is aerodynamically rough. Zhao et al. (2014) argue that in this humid region, sensible heat transfer from the surface to the ABL is more efficient in rural than in urban areas, leading to relatively lower (higher) surface temperatures in the rural (urban) area. Such a process is attributed in Zhao et al. (2014) to largely explain MODIS-derived daytime land surface UHII in this region, which is termed as the roughness warming theory. The roughness warming theory was derived when comparing the aerodynamic resistance/roughness between rural land-use categories and a single urban category (i.e., the intraurban variation of roughness is ignored). Given the intraurban variation of roughness due to different building densities and heights, according to the roughness warming theory, aerodynamically smoother urban areas should experience stronger daytime UHII than aerodynamically rougher urban areas (e.g., CBD), assuming that the ratio of roughness length of momentum to heat remains roughly constant across the entire urban area (Moriwaki and Kanda 2006; Kato et al. 2008; Sugawara and Narita 2009; De Ridder et al. 2012). However, this is not the case for OKC. On the contrary, the aerodynamically roughest CBD area experienced the highest UHII (Figs. 2a,b, 3a,b). Thus, the roughness warming theory of Zhao et al. (2014) used for explaining the land surface UHI may not explain the air temperature UHI observed in this study. Other intrinsic causative factors, for example, thermal properties of urban surfaces (Grimmond and Oke 1999; Ogoli 2003; Liu et al. 2006; Zhu et al. 2009; Bohnenstengel et al. 2011; Yang et al. 2013), anthropogenic heat (Ichinose et al. 1999; Fan and Sailor 2005; Grossman-Clarke et al. 2005; Schlünzen et al. 2010), reduction in evaporative cooling due to impervious surfaces (Taha 1997), and trapping of longwave radiation by urban buildings (Arnfield 2003) must have played more important roles in contributing to the air temperature UHII in OKC, as suggested in a modeling study conducted by Ryu and Baik (2012).

Detailed diurnal variation of UHI is still subject to debate (Memon et al. 2009; Imhoff et al. 2010; Hu et al. 2013b; Zhao et al. 2014). A better understanding of UHIIs’ temporal variation could help to diagnose its causative factors (Hu et al. 2013b). Mean diurnal variation of UHII between April 2009 and October 2010 at OKCNET sites was calculated after removing the periods when precipitation occurred or data from certain sites were missing. All the sites show a prominent diurnal variation of UHII with higher values during nighttime (Figs. 4a,b). The UHII normally increased rapidly around the early evening transition (EET), that is, 1900–2000 central standard time (CST), and then stayed at a roughly constant level throughout the night until early the next morning when the convective boundary layer developed. Note that since the sunset and sunrise time varies, the transition time in the diurnal variation of UHII is different in different seasons. Such an issue will be discussed in section 3b. The characteristics of diurnal variation of UHII shown in Fig. 4a are consistent with existing studies for a few other cities, such as Bucharest (Tumanov et al. 1999), Paris (Lemonsu and Masson 2002), New York City (Gedzelman et al. 2003), Orlando (Yow and Carbone 2006), London (Bohnenstengel et al. 2011; Chemel and Sokhi 2012; Barlow et al. 2015), Thessaloniki, and Athens (Giannaros and Melas 2012; Giannaros et al. 2013). In all of these studies, UHI intensity was quantified using near-surface air temperature. Reversed diurnal variation of UHII (i.e., higher UHII during daytime than nighttime) was reported when it was quantified using remotely sensed land surface temperature (Imhoff et al. 2010; Peng et al. 2012; Zhao et al. 2014). As discussed in the introduction, remotely sensed land surface temperatures and UHIIs must be carefully interpreted and do not necessarily characterize well urban temperatures. Another advantage of quantifying UHII using near-surface air temperature is that ambient air temperature is directly related to public health (Basara et al. 2010; Tan et al. 2010; Oswald et al. 2012). Future studies to further investigate the two different quantifications of UHII (i.e., using air temperature and land surface temperature) and combine their advantages are greatly needed (Schwarz et al. 2012).

Fig. 4.
Fig. 4.

(a) Mean diurnal variation of UHII at each OKCNET site during April 2009–October 2010. The OKCNET sites are classified into three categories (i.e., urban, suburban, and rural) with different colors based on the neighborhood land cover characteristics surveyed in Basara et al. (2010). (b) As in (a), but the OKCNET sites are classified into different UCZs defined by Oke (2004).

Citation: Journal of Applied Meteorology and Climatology 55, 3; 10.1175/JAMC-D-15-0206.1

The rapid increase of UHII around the EET is likely related to rapid changes of mean flow and turbulence in the ABL. Around sunset, with the rapid decline of solar radiation and sustained radiational cooling of the surface, upward sensible heat fluxes decreased and convective eddies subsided quickly. Consequently, near-surface atmospheric stability increased quickly as indicated by the rapid increase of mean Ri at the Oklahoma Mesonet sites around OKC (Fig. 5). At 1900 CST, the mean Ri became larger than 0.2, which is considered a quite stable condition (Banta et al. 2003; Galperin et al. 2007). Turbulent kinetic energy (TKE) in the upper part of the ABL decays quickly during the EET, and this part of the ABL becomes the residual layer, which often becomes decoupled from the newly formed stable layer near the surface (Acevedo and Fitzjarrald 2001; Acevedo et al. 2012; Bonin et al. 2013; Rizza et al. 2013; Klein et al. 2014). Rapid reduction of near-surface wind speed around 1800 CST (Fig. 6) is an indication of decoupling of the near-surface stable boundary layer from the upper layers with larger horizontal momentum. Impacts of surface heterogeneities (urban vs rural) are enhanced in the shallow, near-surface, stable boundary layer (Godowitch et al. 1987; Acevedo and Fitzjarrald 2001). Rural, near-surface temperature drops quickly in the shallow, stable boundary layer due to radiational cooling of the surface. In contrast, the longwave radiative heat loss at street level in urban areas is reduced due to multiple reflections among urban buildings (Oke 1981). Meanwhile, heat stored in the urban materials during the day begins to release during the EET (Ogoli 2003; Harman and Belcher 2006; Liu et al. 2006; Zhu et al. 2009; Bohnenstengel et al. 2011; Zajic et al. 2011; Yang et al. 2013). Together with anthropogenic heat, this extra heat in the urban region is released into and confined in the urban boundary layer (Ichinose et al. 1999; Fan and Sailor 2005; Grossman-Clarke et al. 2005; Rizwan et al. 2008; Schlünzen et al. 2010; Kotthaus and Grimmond 2014a,b; Barlow et al. 2015). Consequently, urban near-surface temperatures decrease slowly in a more neutral urban boundary layer induced by the stronger turbulent vertical mixing due to higher heat emissions and rougher surface in cities (Clarke 1969; Oke 1987; Uno et al. 1988; Martilli et al. 2002; Nelson et al. 2011; Salamanca et al. 2014; Bohnenstengel et al. 2015). As a result of different cooling rates at rural and urban sites for 2–3 h, UHII increased quickly during the EET and stayed at a roughly constant level for the rest of the night (Figs. 4a,b).

Fig. 5.
Fig. 5.

Mean diurnal variation of Richardson number Ri and shortwave radiation at 10 Oklahoma Mesonet sites around OKC (i.e., OKCE, OKCN, OKCW, MINC, SPEN, ELRE, GUTH, KIN2, NRMN, and WASH) during April 2009–October 2010.

Citation: Journal of Applied Meteorology and Climatology 55, 3; 10.1175/JAMC-D-15-0206.1

Fig. 6.
Fig. 6.

Mean diurnal variation of wind speed (WSPD) at OKCNET sites during April 2009–October 2010.

Citation: Journal of Applied Meteorology and Climatology 55, 3; 10.1175/JAMC-D-15-0206.1

Previous studies suggested that the spatial variability of temperature in rural areas played a very important role in determining the UHII (Hawkins et al. 2004). A question arises as to how important the EET cooling rate in rural areas is for UHI development (Yow and Carbone 2006). The relationship between rural background temperature change rate during the EET (2 h before sunset) and that of UHI intensity immediately after the EET (2 h after sunset) are thus further examined (Fig. 7). The mean UHI intensity at the OKCNET “urban” sites (i.e., KCB101–109; Basara et al. 2010) is chosen for analysis. Different sunset times during different months are accounted for. The statistic investigation based on 2009/10 data demonstrates that the near-surface cooling rate in the rural area around OKC during the EET shows a significant correlation with the UHI intensity at early evening with a correlation coefficient of −0.63 (Fig. 7). The mean UHI intensity at early evening roughly represents the mean nighttime UHI intensity (Fig. 4a). Thus, the near-surface cooling rate in the rural area during the EET is confirmed to play a critical role in determining the nocturnal UHI intensity with strong (weak) EET cooling rate likely leading to strong (weak) nocturnal UHI.

Fig. 7.
Fig. 7.

Correlation between rural cooling rate during the EET (i.e., 2 h before sunset) and average UHII at early evening (2 h after sunset) at OKCNET urban sites during April 2009–October 2010.

Citation: Journal of Applied Meteorology and Climatology 55, 3; 10.1175/JAMC-D-15-0206.1

Similar to the cooling rate during the EET, the heating rate during the morning transition was also larger in the rural than urban area (figure not shown), which is consistent with previous studies (Oke 1987). The average absolute values of temperature change rate during the early evening and morning transition are shown in Fig. 8. The temperature change rate shows a concentric distribution with the lowest rate located over OKC’s CBD and increases from urban to rural zones (Fig. 8), indicating the moderating effects of urban to diurnal temperature change. The temperature change rate during the evening and morning transition may provide an alternate way to estimate the magnitude of urban footprints.

Fig. 8.
Fig. 8.

Spatial distribution of average temperature change rate during the early evening and morning transition based on (a) spatially distributed observations and (b) kriging analysis.

Citation: Journal of Applied Meteorology and Climatology 55, 3; 10.1175/JAMC-D-15-0206.1

As discussed above, UHI intensity exhibited a large spatial variation over OKC (Figs. 2, 3, 4). UHI intensity over OKCNET sites varied between 0.4° and 2.1°C during nighttime, while it varied between −0.2° and 1.0°C during the daytime (Fig. 4a). The large spatial variation of UHI intensity over OKC may be primarily caused by the different surface structures and cover at each site, as also reported for other cities (e.g., Shahgedanova et al. 1997; Eliasson and Svensson 2003; Alcoforado and Andrade 2006; Kolokotroni and Giridharan 2008; Bohnenstengel et al. 2011; Brandsma and Wolters 2012; Oswald et al. 2012; Suomi and Kayhko 2012; van Hove et al. 2015). Classifying all the sites in cities with different surface structure and cover into a single urban category may be inadequate to accurately quantify UHI in terms of its intensity and variation (Stewart 2011; Stewart and Oke 2012). According to the three-category (i.e., urban, suburban, and rural) classification of Basara et al. (2010), most of the sites starting with KCB (except KCB110) are located in the CBD and their surface type is classified as urban. These urban sites experienced higher UHI intensities than most of the remaining OKCNET sites during both daytime and nighttime (Figs. 2, 3, 4). Four OKCNET sites (i.e., KSW101, KSW110, KSE102, and KNE103) are classified as “rural” surface types in Basara et al. (2010), and these sites experienced lower UHI intensities than most of the other OKCNET sites (Figs. 2, 3, 4). The large spatial variation of UHI intensity across OKC suggests that determining UHI intensity using the temperature difference between a certain urban site and a certain rural site may lack objective meaning and climatological relevance. These results thus justify the proposal by a few recent studies (e.g., Basara et al. 2010) of classifying measurement sites into detailed (to certain degrees) urban categories for objective UHI quantification and description.

The Basara et al. (2010) three-category classification is further compared with the more detailed classification of seven different UCZs defined by Oke (2004). The UHII diurnal variation shows a large variability within individual UCZs (Fig. 4b), while the three-category classification better captures the differences in the UHII characteristics among the OKCNET sites. We have thus decided to use the classification into rural, suburban, and urban sites as proposed by Basara et al. (2010) in this study.

2) Wind speed and implications for treatments in numerical models

Because of the diurnal variation of vertical coupling strength (strong coupling in daytime and weaker coupling at nighttime), the diurnal cycle of surface wind was prominent, exhibiting a maximum (minimum) in the daytime (nighttime) (Fig. 6). During the daytime, stronger downward transport of boundary layer momentum led to stronger surface winds than are observed at night. Wind speeds at each individual site were largely determined by the roughness of each site (Kamal et al. 2015) and wind climatology in its surrounding environment. Two OKCNET sites on the west side of OKC, KNW104 (suburban) and KSW101 (rural) with relatively low roughness being located on dispersed settlement and grassland (Basara et al. 2011), experienced strong wind speeds (Figs. 2c,d, 9). The climatological east-to-west gradient of wind speed in the presence of dominant southerly wind in Oklahoma (Song et al. 2005) is another factor to explain the large wind speeds at these sites. Such an east-to-west gradient of southerly wind speed in Oklahoma was previously noticed in both observation and modeling results (Hu et al. 2013a; Hu et al. 2013b).

Fig. 9.
Fig. 9.

Mean diurnal variation of wind speed difference between OKCNET sites and the seven surrounding Oklahoma Mesonet sites during April 2009–October 2010.

Citation: Journal of Applied Meteorology and Climatology 55, 3; 10.1175/JAMC-D-15-0206.1

Urbanization can have two different effects on surface wind: (i) enhanced surface roughness reduces surface wind speeds (Kamal et al. 2015), while (ii) stronger downward momentum transport due to enhanced vertical mixing increases surface speeds (Y. Wang et al. 2007; Hu et al. 2013b). Wind speeds at OKCNET sites (except KNW104 and KSW101) were always smaller than at the surrounding rural Oklahoma Mesonet sites (Fig. 9), which suggests that the effects of the enhanced urban surface roughness were dominant. The urban effect on wind in terms of the magnitude of wind speed reduction was more prominent during the daytime than during nighttime. The reductions of near-surface wind speed in the CBD area are as large as 3.5 m s−1 during daytime, whereas the magnitude of wind speed reduction decreases to 2 m s−1 during nighttime (Fig. 9).

Figures 2, 3, and 9 imply that the near-surface wind speed in urban areas is typically weaker than over the surrounding rural areas due to the dominant effects of enhanced roughness in urban areas. However, in three-dimensional simulations with the Weather Research and Forecasting (WRF) Model, higher values of near-surface wind speed are at times simulated over urban than over rural areas (e.g., during the early evening transition; see Fig. A1 in the online supplemental material). These results, which are likely unrealistic according the wind speed data observed at the OKCNET sites, suggest that the vertical transport of momentum in mesoscale models may be overestimated. The advancement of model capability to handle vertical transport of momentum has been slower compared to the advancement of model capability to handle transport of scalars because most previous research efforts in terms of vertical transport had been focused on scalars rather than momentum (Frech and Mahrt 1995; Storm et al. 2009; Hu et al. 2013a; Ngan et al. 2013; Draxl et al. 2014; Gutiérrez et al. 2015).

b. Seasonal variation of UHI

Seasonal variation of UHI intensity in OKC during April 2009–October 2010 (Fig. 10a) did not show a clear warm–cold season contrast as reported in other cities (e.g., Magee et al. 1999; Steinecke 1999; Montávez et al. 2000; Jonsson 2004; Kim and Baik 2005; Hinkel and Nelson 2007; Zhou and Shepherd 2010; Memon et al. 2011; Yang et al. 2013). Hu et al. (2013b) concluded that rural temperature inversion strength can serve as an indicator of nocturnal UHI intensity based on the analysis of temperature data in July 2003 and model simulation results. Thus, temperature inversion strength (defined as the temperature difference between 1.5 and 9 m AGL) at the seven surrounding Oklahoma Mesonet sites is also examined (Fig. 10b). The temperature inversion strength had a similar variation as that of UHII and did not show a clear warm–cold season contrast. Monthly variation of percentiles (median, 25%/75%, and 5%/95%) of daily mean nocturnal UHII was also similar to those of nocturnal rural inversion strength (Fig. 11), even though the monthly percentiles may not be statistically significant in certain months due to many missing values because of precipitations (see Fig. 10). The UHII and surrounding rural inversion strength had a significant correlation during nighttime (2200–0500 CST) with a correlation coefficient of 0.79 (Fig. 12a). As discussed above, urban effects including reduced outgoing longwave radiation, extra heat, and stronger roughness lead to a more neutral and relatively thicker urban boundary layer and slower temperature decrease in contrast to rapid temperature decrease in the shallow stable rural boundary layer during the EET. Stronger rural temperature inversion is normally associated with a shallower and more stable boundary layer and allows urban effects to manifest more prominently with higher UHI intensity (Hu et al. 2013b), effectively explaining the positive correlation between rural inversion strength and UHI intensity. Two exceptional points (i.e., 29 March 2010 and 29 October 2010) stand out in Fig. 12a. Two reasons were responsible for these exceptions: first, extremely strong inversion (~8°C between 1.5 and 9 m AGL) occurred at the ELRE Oklahoma Mesonet site on these nights. Similar significant nocturnal inversion at the ELRE site in presence of clam winds, low humidity, and clear skies was previously noticed and reported by Hunt et al. (2007); second, nocturnal warming events occurred at the MINC Oklahoma Mesonet site. Various nocturnal warming events may occur at certain rural stations for a few reasons (White 2009; Nallapareddy et al. 2011; Hu et al. 2013c; Hu and Xue 2016), including cold front passages. Detailed investigation of the rural nocturnal warming events is beyond the scope of this study.

Fig. 10.
Fig. 10.

Time series of (a) UHII at OKCNET urban sites and (b) rural average temperature inversion at the seven surrounding Oklahoma Mesonet sites during April 2009–October 2010.

Citation: Journal of Applied Meteorology and Climatology 55, 3; 10.1175/JAMC-D-15-0206.1

Fig. 11.
Fig. 11.

Monthly variation of median, 25%/75%, and 5%/95% percentiles of mean nocturnal (2200–0500 CST) (a) UHII and (b) rural inversion strength around OKC during April 2009–October 2010.

Citation: Journal of Applied Meteorology and Climatology 55, 3; 10.1175/JAMC-D-15-0206.1

Fig. 12.
Fig. 12.

Correlations between daily nocturnal (2200–0500 CST) UHII at the OKCNET urban sites and (a) rural inversion strength and (b) rural wind speed during April 2009–October 2010.

Citation: Journal of Applied Meteorology and Climatology 55, 3; 10.1175/JAMC-D-15-0206.1

Given the significant correlation between rural inversion strength and nocturnal UHII, further investigation of a longer term of rural inversion strength can help diagnose the seasonal variation of nocturnal UHI intensity. Thus, longer-term Oklahoma Mesonet data (April 2009–December 2012) are examined. Monthly variation of median, 25%/75%, and 5%/95% percentiles of daily mean nocturnal rural inversion strength around OKC during this longer period is shown in Fig. 13. The years 2011 and 2012 were exceptionally dry (Ramsey et al. 2014). There were only a few precipitation events during these years; thus, more inversion data are counted and the statistics are more reliable than those in Fig. 11b. The maximum median nocturnal inversion occurred in October. The standard deviations of rural inversion were smaller and the extreme values (indicated by 95% percentile) of rural inversion were lower in June and July (Fig. 13). Given the relationship between rural inversion and UHI intensity discussed above, the strongest nocturnal UHI was less likely to occur in June and July.

Fig. 13.
Fig. 13.

Monthly variation of median, 25%/75%, and 5%/95% percentiles of mean nocturnal (2200–0500 CST) rural inversion strength around OKC during April 2009–December 2012.

Citation: Journal of Applied Meteorology and Climatology 55, 3; 10.1175/JAMC-D-15-0206.1

Wind speed has been reported to modulate UHI intensity in various cities (Morris et al. 2001; Unger et al. 2001; Fast et al. 2005; Steeneveld et al. 2011). The correlation between rural wind speed and nocturnal UHI intensity during April 2009–October 2010 was examined. Nocturnal UHI intensity normally decreased with increased wind speed with a correlation coefficient of −0.72 (Fig. 12b). Larger wind speed leads to stronger turbulence and stronger mechanical vertical mixing, which reduce or eliminate rural background nocturnal temperature inversion. Since rural temperature inversion is a good indicator of nocturnal UHI intensity as discussed above, larger wind speeds decrease UHI intensity. Other processes (e.g., clouds) also play roles in modulating UHI intensity (Morris et al. 2001; Rosenzweig et al. 2005; Yow and Carbone 2006; Hoffmann et al. 2012), which can partially explain the scattering of the data points in Fig. 12b. Unfortunately, cloud data are not available from OKCNET and Oklahoma Mesonet sites.

Even though a clear warm–cold season contrast in the magnitude of UHII in OKC was not discerned during April 2009–October 2010 (Fig. 10a), seasonal variation of the timing of UHI was prominent (Fig. 14). Figure 14 shows the mean UHII at OKCNET sites as a function of month and time of the day. Note comparison of UHI intensity between different months in Fig. 14 is not meaningful since the number of available data during each month is different (see the data availability in Fig. 10a after removing the periods with precipitation and missing data at certain sites). Figure 14 further confirms the prominent diurnal variation of UHI intensity; that is, nocturnal UHI was strong, while daytime UHI was weak. The timing of onset/subsiding of nocturnal UHI showed a clear monthly variation. The time span of nocturnal UHI was short during warm months, while it was relatively longer during cold months. The onset timing of nocturnal UHI roughly followed the sunset time. This further confirms the critical roles played by the physical processes during the EET in the development of UHI.

Fig. 14.
Fig. 14.

Mean UHII over the OKCNET sites as a function of month and time of the day.

Citation: Journal of Applied Meteorology and Climatology 55, 3; 10.1175/JAMC-D-15-0206.1

4. Conclusions and discussion

Using the data recorded from a dense surface observing network, that is, the Oklahoma City Micronet, during April 2009–October 2010, observed spatial distribution of UHII over Oklahoma City is investigated. UHII exhibited a large spatial variation over OKC. The widely varied UHII over OKC is partially explained by the different surface structure and cover at each site. The large variation of UHII across the urban area suggests that determining UHII using the temperature difference between an individual urban–rural site pair may lack objective meaning. It is better to classify measurement sites into detailed categories for objective UHI quantification and description as suggested by recent studies (e.g., Basara et al. 2010).

During both daytime and nighttime, the strongest UHII was mostly confined around the central business district where surface roughness is the highest in the OKC metropolitan area. These results do not corroborate the roughness warming theory of Zhao et al. (2014), according to which aerodynamically smoother urban areas would experience stronger daytime UHII than aerodynamically rougher urban areas (e.g., CBD).

UHII of OKC increased prominently around the early evening transition and stayed at a fairly constant level through the night. The boundary layer processes during the EET played a critical role in determining the nocturnal UHII in the absence of disturbances such as precipitation. Associated with the rapid decline of solar radiation during the EET, a stable boundary layer develops close to the ground. Rural temperatures in the shallow stable boundary layer decrease quickly due to radiative cooling. Meanwhile, heat stored in the urban building materials during the day released rapidly, together with the anthropogenic heat emissions, and heated the urban boundary layer. This extra heat, together with reduced outgoing longwave radiation (due to wall reflection, etc.) and elevated roughness, led to a more neutral urban boundary layer, in which temperature near the surface decreased slower than in the rural stable boundary layer. As a result of different cooling rates between urban and rural, UHII increased rapidly during the EET. The near-surface cooling rate in the rural area during the EET regulated the nocturnal UHII with a correlation coefficient of −0.63. Factors such as wind speed and clouds may have affected the stability/depth of the background nocturnal boundary layer, thus modulating UHII.

Nocturnal rural temperature inversion strength had a similar day-to-day variation as that of UHII. The nocturnal UHII and surrounding rural inversion strength are significantly correlated with a correlation coefficient of 0.79. A stronger rural inversion normally means a shallower surface layer and a larger EET temperature decrease in the rural area than in the urban area, which leads to a stronger UHII. The rural inversion strength did not show a clear warm season–cold season contrast during April 2009–October 2010. Thus, warm season–cold season contrast of UHII in OKC was not prominent during this period. Analysis of a longer term (April 2009–December 2012) of rural inversion strength suggested that the strongest nocturnal UHI in OKC was more likely to occur in months other than June and July. Seasonal variation of the timing of UHI was prominent, with a shorter (longer) time span of nocturnal UHI during warm (cold) months, which is directly linked to the sunset and sunrise timings.

Though not shown here, surface ozone (O3) was removed during the EET because of the deposition and chemical reactions in the stable boundary layer. A shallower stable boundary layer normally led to a quicker surface O3 reduction. The O3 removal rate during the EET showed a good correlation with the rural cooling rate. Thus, the characteristics of certain chemical species such as O3