1. Introduction
The National Center for Atmospheric Research (NCAR) has developed a real-time, multigrid (grid increments of 0.5–45 km), four-dimensional data-assimilation (RTFDDA) and forecasting system (Cram et al. 2001) for meteorological support of testing at U.S. Army Test and Evaluation Command (ATEC) test ranges. The RTFDDA system, based on the fifth-generation Pennsylvania State University–NCAR Mesoscale Model (MM5), has been running operationally at five Army test ranges for over 4 yr. It is designed to cycle at a frequency of 1–12 h, with each cycle producing a model-based analysis and a forecast of up to 48-h duration. The data-assimilation system is based on Newtonian relaxation (Stauffer and Seaman 1994), and produces dynamically consistent model-assimilated datasets that are used as estimates of current conditions as well as for model initial conditions. In this paper, in addition to forecasts, analyses are verified against independent observations because they are important for providing input to short-range transport and diffusion (T&D) calculations. In addition to its use by the Army test ranges, the RTFDDA system has supported the missions of other U.S. Department of Defense and government agencies.
In this paper, we discuss the ability of the RTFDDA system to produce model-assimilated datasets and forecasts in support of the Joint Urban 2003 (JU2003) Atmospheric Dispersion Study, held in Oklahoma City (OKC), Oklahoma, in July 2003. Even though the JU2003 study focused on building-scale effects on T&D in the central region of OKC (Allwine et al. 2004), accurate operational city-scale predictions of boundary layer (BL) structures were needed for planning intensive observation periods. Therefore, during the pre-JU2003 period, considerable effort was devoted to evaluating and improving MM5’s data-assimilation methods and model physics schemes in order to better represent the urban atmospheric environment. This paper describes two model improvements relevant to the prediction of BL structures and urban effects. The first is related to MM5’s commonly used Medium Range Forecast (MRF) model BL parameterization (Hong and Pan 1996), and the second involves the representation of urban land use/substrates in the NCEP–Oregon State University–Air Force–Hydrologic Research Laboratory (Noah) land surface model (LSM; Ek et al. 2003). As discussed in more detail in section 2, most of the BL parameterization schemes available in the MM5, including the MRF, tend to underestimate the daytime near-surface wind speeds. Also, the MRF scheme tends to develop the convective BL too early in the day, and it becomes excessively deep. Furthermore, the standard Noah LSM in the MM5 (Chen and Dudhia 2001; Ek et al. 2003) has a simplified representation of cities, such that model tests showed only a very marginal urban heat island (UHI) effect for OKC.
As with any moderate- to large-sized city, OKC has sufficient area that land–atmosphere interaction can affect the bulk, city-scale, dynamic and thermal evolution of the BL, and hence the T&D of airborne material over the city. Thus, correctly modeling the bulk urban effects on the atmosphere were crucial for the success of the RTFDDA system’s support for this field experiment. There is much research aimed at understanding and modeling the complex land–atmosphere interaction that affects atmospheric dynamics and thermodynamics in urban areas, where methods involve field observations (Rotach 1995; Doran et al. 2002; Allwine et al. 2002), as well as parameterizing the processes in mesoscale models (Brown 2000; Grimmond and Oke 2002; Masson 2000). For example, using idealized environments, the improved knowledge and modeling technology gained from the above-mentioned and other studies have been used to investigate various aspects of the UHI and urban-induced momentum mixing (Uno et al. 1989; Masson 2000; Grimmond and Oke 2002; Martilli et al. 2002; Dupont et al. 2003; Otte et al. 2004).
One approach for modeling the bulk effects of urban areas on the atmosphere is to use an urban canopy model (UCM; also known as urban canopy parameterization) to represent the model grid cell–averaged effect of the building structures on the dynamics and thermodynamics. Such UCMs parameterize the ensemble characteristics of the urban morphology, but individual buildings and street canyons are not explicitly represented (Brown 2000; Masson 2000; Martilli et al. 2002; Otte et al. 2004). At NCAR, an effort is under way to couple a UCM that is based on the work of Kusaka et al. (2001), with the community mesoscale Weather Research and Forecasting (WRF) model (Chen et al. 2004). This type of UCM considers the geometry of buildings and roads to represent the radiation trapping and wind shear in the urban canopy. Such an approach requires detailed urban land use datasets and the input of a number of parameters that define the urban geometry, where these parameters need to be calibrated for each individual city. In this operational support mission for the JU2003 field program, it was necessary to balance urban model complexity and data input requirements with the need to set the model up quickly. Therefore, we adopted an approach of modest complexity in which the Noah LSM in the public release version of MM5, version 3.4, was enhanced to capture the primary influences of the urban area (see section 3 for details).
Section 2 describes the modifications made to the MRF BL parameterization, and in section 3 the refinements to the Noah LSM are summarized. Section 4 contains an objective verification of the improvements to the MRF BL parameterization. In section 5 a comparison is made between model products and field program data, Oklahoma Mesonet data from locations surrounding the city, and conceptual models of the urban BL. The last section, section 6, provides a summary of this study.
2. Modification of the MM5 MRF BL parameterization
a. Known deficiencies with the current MRF BL parameterization
The MRF BL scheme, commonly used by the MM5 community for research and operational mesoscale weather forecasting, was selected for use in the RTFDDA system for its efficiency and its comparable overall performance relative to other BL schemes available in MM5. It is worth noting that the MRF parameters were defined by Hong and Pan (1996) using a single case in which the main objective was to improve coarse-resolution precipitation forecasts. A known problem with this parameterization has been that it tends to overestimate the daytime BL depth, and seriously underestimate the near-surface winds (Zhang and Zheng 2004). These problems can seriously impact the quality of T&D calculations that are based on MM5 output. For example, Cheng et al. (2003) reported that MM5 simulations with the MRF BL scheme persistently overpredicted the BL depth over the Houston–Galveston, Texas area. Also, in sensitivity experiments with five BL parameterizations available in MM5, for a period with three consecutive days of clear skies over the central plains, Zhang and Zheng (2004) found that all of the BL schemes underestimated the daytime near-surface winds, and that the MRF scheme performed the most poorly. They point out that there is a general lack of understanding of BL momentum transport processes because previous research on BL and land surface modeling has primarily focused on thermal and moisture fluxes.
Statistical verification of the BL analyses and forecasts of the ATEC operational RTFDDA systems has revealed the same problem for various geographic regions and for all seasons. Figure 1 shows the bias errors of the wind speed at 10 m above ground level (AGL) for 10–12-h forecasts (i.e., forecasts with a length of 10–12 h) on the fine mesh of a nested grid (1.1- or 3.3-km grid increment, depending on the ATEC range). These are averaged statistics for five test ranges, as verified against the range observations (total of 71 mesonet stations) during August 2002. Despite large geographic and climatic differences, the RTFDDA forecasts underpredicted the daytime wind speeds at all ranges.
In summer, the daytime BL is characterized by free-convective mixing of sensible and latent heat, as well as momentum. Winds near the surface are affected by the downward mixing of the higher momentum from above, and the upward mixing of lower momentum from below. Thus, an underprediction of daytime near-surface winds can result from either an underestimate of the downward momentum flux, as speculated by Zhang and Zheng (2004), or an overestimate of surface stress.
An analysis of RTFDDA results confirms the problem with the W* formulation in Eq. (5). On a typical summer day with clear skies, MM5 simulations, with the Noah LSM scheme and the lowest model computation level at ∼15 m AGL, can produce differences in θυg and θυa of 10–25°C during the daytime. For a difference of 16°C, Eq. (5) will yield a W* of about 8 m s−1, which is larger than U in most weather situations. This difference will lead to an overestimate of the friction velocity u*, resulting in excessively weak near-surface winds, and, unfortunately, with the excessively weak near-surface winds, the free-convective mixing W* becomes, relatively speaking, even more dominant in the surface-layer flux computation.
b. Incorporating Beljaars’s formulation into the MRF BL parameterization
c. A new method for calculating the BL height
From Eq. (6), an accurate computation of W* will still rely on estimates of the BL depth, as well as the surface heat and moisture fluxes. Recall that one of the known problems with the original MRF BL scheme is its overestimate of the BL height. Analysis of the RTFDDA output indicates that the MRF BL scheme frequently produces premature BL development (1–3 h), which results in excessively deep BLs (by a few hundred meters to more than 1 km) in the morning. These BL depth errors not only affect the BL thermodynamic structure, but also directly result in an overestimate of W*. To improve the surface flux computation in the free-convective BL, the MRF BL depth needed to be improved.
It is of interest to point out that using the Ribl to estimate BL heights is conceptually consistent with the approach in more complicated, higher-order BL turbulence closure schemes, where BL heights are diagnosed based on turbulence kinetic energy (TKE), which is predicted at the model grid points. These local TKE values represent the intensity of the local turbulence, which is similar to using the Ribl. Numerical experiments run on the same cases with different BL parameterizations showed that the BL heights calculated with the revised MRF BL scheme are very close to those diagnosed by the more sophisticated and computationally demanding Meller–Yamada–Janjic TKE scheme (not shown).
Note that the Noah LSM uses a skin temperature for heat transfer calculations that is defined at the ground surface, and not at the height of the roughness length. Hence, the Monin–Obukhov similarity theory cannot be directly applied. The effect of the molecular sublayer is to act as a resistance to fluxes, and it is equivalent in this sense to a thermal roughness length that is smaller than the roughness length for momentum. Zilitinkevich (1995) proposed relating this ratio to the roughness Reynolds number, and Chen et al. (1997) found that this approach can improve the surface heat flux and skin temperature simulations. This scheme was introduced into the MM5 by Chen and Dudhia (2001). It is restated here because it is of great importance to incorporate this refinement along with the MRF BL modifications described previously.
3. Modification to the representation of urban areas in the Noah LSM
Properly modeling urban effects on the atmosphere was crucial for the successful use of RTFDDA to support the JU2003 experiment. In the model configuration used, the urban area occupied roughly 20% of the 1.5-km grid and 70% of the 0.5-km grid (see Fig. 2). At this horizontal resolution, the bulk influence of the urban land surface forcing on the BL should be representable, as should effects related to the irregular shape of the urban area. Niyogi et al. (1999) and Holt et al. (2006) indicate the importance of small-scale land surface heterogeneity to BL development. Although urban parameterization schemes with varying degrees of complexity have been coupled to many versions of MM5 (Otte et al. 2004; Dupont et al. 2003; Brown 2000), they are not yet available in the public release of MM5. In addition, it is not clear what level of complexity in the urban land use treatment should be incorporated. For instance, Taha (1999) prefers a simple approach, and points out that a large amount of detail in complex urban models may be lost when averaging back to a coarse model grid. Eventually, the WRF Noah LSM will be coupled with an advanced, one-layer urban canopy model (UCM) that is based on Kusaka et al. (2001). For the work described here, a simple bulk representation of urban areas was incorporated in the standard MM5 and WRF Noah LSM (Chen and Dudhia 2001). Given the simplicity, efficient computations, and encouraging results (shown below) of these improvements to the Noah LSM, we believe this relatively simple approach could be a good alternative to the more complex UCM for operational forecasting for urban areas on these scales. In fact, the modifications introduced here are already being used by others (Grossman-Clarke et al. 2005; D. Byun 2004, personal communication; D. Stauffer 2004, personal communication).
The Noah LSM urban enhancements include increasing the roughness length from 0.5 m to 0.8 m to better account for the drag resulting from buildings. The roughness length of 0.8 m used here is slightly lower than the 1.0 m used by Atkinson (2003), in consideration of the generally low buildings in OKC. The surface albedo was reduced from 0.18 to 0.15, where this reduction accounted for the shortwave radiation trapping in the urban canyons [The new value is consistent with the average for urban areas suggested by Pielke (1984), and the same value was suggested by Atkinson (2003).] The volumetric heat capacity was increased to 3.0 × 106 J m−3 K−1 and the soil thermal conductivity to 3.24 W m−1 K−1. These two values are larger than those for the prevailing concrete/asphalt materials (about 2.0 × 106 J m−3 K−1 and 2.0 W m−1 K−1, respectively), to roughly reflect the heat storage resulting from building walls. Last, the green vegetation fraction was reduced to 0.05, and the available urban soil water capacity was decreased to reduce evaporation. At present, there is no proper estimate of the bulk water capacity in urban regions. We adjusted these parameters experimentally, prior to the RTFDDA support for the JU2003, to keep the surface latent heat flux small when there is no rain. In a few UHI models (e.g., Martilli et al. 2002; Otte et al. 2004), the surface evaporation is ignored. Note that these physical properties were not chosen to “tune” the model solutions to the JU2003 field program data and other data used in this study; they were based on estimates made prior to the verification process.
4. Pre-field-program verification of the performance of the improved MRF BL parameterization
Testing of the modified MRF BL scheme and Noah LSM was conducted during the preparation stage for the JU2003 field experiment. Thus, datasets that were available prior to the experiment are used here for verification.
a. Verification during a 3-day period in the Central Plains: 26–28 May 2003
A period with clear skies over the central plains, from 26 to 28 May 2003, was chosen because of the strong growth of the BL. Five nested grids were used, with grid increments of 40.5, 13.5, 4.5, 1.5, and 0.5 km (Fig. 2). The RTFDDA system was cycled every 3 h, and for each cycle a final analysis of present conditions and a 15-h forecast were generated. The model physics, aside from the modified MRF BL, were the same as those described in Liu et al. (2002). The RTFDDA analyses and forecasts were verified against hourly near-surface observations and 12-hourly radiosondes. Here, only the surface verification of the analyses and 10–12-h forecasts are presented.
First, simulations were conducted to verify the new BL height calculation scheme. The RTFDDA-forecast BL heights for 0500–1800 local time (LT) 28 May 2003, calculated using the new and old (original) methods, were interpolated from model domain 1 to the locations of three profilers located in Kansas and operated by the Argonne National Laboratory (ANL). The low-power signal-to-noise ratio (SNR) of the wind profilers may be used to infer BL turbulence properties. Coulter and Holdridge (1998) and LeMone et al. (2000) describe how to define BL heights according to the values and vertical gradient of the SNR. Essentially, the BL heights are located immediately above the maximum in the vertical gradient of the SNR. It should be noted that, because the SNR is dependent on local moisture gradients, the estimates might differ from the real BL height in some instances, for example, during the late afternoon when the BL is collapsing. Thus, one needs to be cautious when estimating the BL height with the wind profiler SNR returns. Figure 3 shows the comparison between the BL height evolution of the RTFDDA model runs with the old and new BL schemes, and SNR fields from ANL profilers located at Beaumont, Whitewater, and Oxford, Kansas. The BL calculated with the original approach develops too early and becomes too deep during the daytime. In contrast, the new approach performs better, even though, arguably, the BL heights are underpredicted in the early afternoon.
Next, a comparison was conducted of the performance of the old and the new MRF BL schemes for the period from 0000 UTC 27 May to 0000 UTC 29 May 2003. Figure 4 shows the verification of 10–12-h wind speed forecasts at 10 m AGL from the two model versions, based on over 1400 Meteorological Terminal Air Reports (METAR) and other near-surface observations in model domain 1 for the series of forecasts initiated every 3 h over the 2-day period. Again, values were interpolated from the model grid to the locations of the observations. The original scheme considerably underestimated daytime winds at 10 m AGL, with a maximum bias of −2.5 m s−1 occurring at about 1500 LT for the 10–12-h forecasts and a maximum bias of −1.6 m s−1 during the day for the analyses (Fig. 4a). The wind speed RMS errors (Fig. 4b) have maxima during the day. The modified MRF BL scheme reduces these daytime weak-wind forecast biases by more than 70% at some times. In addition, the new scheme also improved, albeit to a lesser degree, the nocturnal wind-speed forecast. With the revised MRF BL scheme, the RMS errors during the night and day are of a similar magnitude.
Although the major goal of the MRF BL modification was to correct the daytime weak-wind bias near the surface, the other near-surface variables also benefit to a lesser degree. Figure 5 shows that the new scheme improves the daytime and early evening 10-m-AGL wind direction and 2-m-AGL temperature predictions. At some times, the wind direction is improved by 10°. Humidity forecasts have a similar improvement (not shown). Verification of forecasts from the finer grid meshes (Fig. 2) shows improvements very similar to those just described for domain 1.
b. Verification for Aberdeen, Maryland, for a 1-month period
In contrast to the above evaluation of the modified MRF BL scheme for a few case study days, in this section we examine its performance in the context of an operational mesoscale forecasting system at the ATEC Aberdeen Test Center in Maryland. Statistics for the month of August 2002, before the changes were made, are compared with those for August 2003, after they were made. Figure 6 compares the monthly mean 10-m- AGL wind speed bias and RMS errors on the middle grid of a three-grid nest, where the grid increment is 10 km. Again, the verification is done by interpolating the model output to all surface stations. As with the previous case study results, Fig. 6a shows a significant improvement in the 10-m-AGL daytime wind speed bias with the new MRF BL, with the monthly mean 4–6-h forecast biases for 1200 LT decreasing from 2.2 to 0.5 m s−1. Similar improvements can be seen in the daytime RMS wind speed errors in Fig. 6b, with nighttime benefits also being evident. Other variables show improvements similar to those computed for the case studies above.
5. Application of the enhanced model to the JU2003 field program period, and assessment of performance
In this section, analyses from the enhanced model are compared with data for the JU2003 field program period, and the ability of the model to replicate observed BL properties is assessed. Note that it is reasonable to employ model-generated analyses because the field program data with which they are compared were not assimilated in the operational system. Because urban effects are more identifiable during clear days with weak synoptic forcing, model verification in this section focuses on eight clear-sky days (4, 5, 15, 16, 17, 18, 24, and 25 July) during that period.
a. Verification of near-surface variables
Verification data used here are from the JU2003 field program and the Oklahoma Mesonet (Brock et al. 1995). Three observation stations were selected for use (Fig. 2): Portable Weather Information Display System (PWIDS) 14 (PW14) represents an urban setting, and the sensor is mounted on a light pole in the central OKC business district. PWIDS is a lightweight mobile surface atmospheric measurement station. It measures wind speed, wind direction, temperature, and relative humidity. The SPEN station is in the suburbs to the northeast of the city, and MINC represents rural conditions and is to the southwest of the city. The PW14 station was operated by the U.S. Army Dugway Proving Ground for the field program, and SPEN and MINC are permanent Oklahoma Mesonetwork stations (Brock et al. 1995). The model near-surface variables on domain 4 were interpolated to the station locations. The prevailing surface flow over the OKC region is typically from the south in July, and this was the case for all the clear-sky days used here, except for 24 July, on which a northeasterly-to-southeasterly flow was observed. The following subsections describe comparisons of simulated and observed temperature and specific humidity. Model winds are not verified here because the measurement location in the urban area is within a street canyon, and thus the data are not representative of the scales that are simulated by the model
1) Near-surface temperature
The diurnal evolution of the observed and model-analyzed temperatures at 10 m AGL at the three stations is presented in Fig. 7 for 16 (southerly flow) and 24 (easterly flow) July 2003. The verification results for the other six clear-sky days are generally similar to those for the 16 July case. Also, Table 1 presents, for each of the eight dates, the observed and predicted 1) amplitude of the diurnal temperature variation and 2) urban–rural and urban–suburban temperature differences for 0600 and 1400 LT. The 1400 LT time was chosen to reflect conditions in the warm BL because precipitation was observed late in the afternoon for a few of the cases, which locally influenced temperatures. The following general conclusions about model performance can be obtained from Fig. 7 and Table 1:
The timing of the simulated and observed temperature minima and maxima coincide well (Fig. 7). The observed small midmorning difference in the temperatures among the three locations is generally represented by the model (Fig. 7). The small midmorning urban–rural temperature difference is a well-document phenomenon (Oke 1987). After sunrise, urban regions heat up more slowly than rural areas because of the larger heat capacity and conductivity of the urban surface.
The diurnal temperature ranges were simulated well (Table 1 and Fig. 7). For each of the three locations, 24 July had the largest observed and forecast diurnal range of the 8 days. For 5 July, the model correctly forecast the smallest diurnal amplitude of the 8 days at all three of the locations. Also, the ranking of the three locations in terms of the 8-day-averaged diurnal range is the same based on the model and the observations: The rural location had the largest, the suburban location had the smallest, and the urban site was intermediate. Individual errors were frequently in the range of a few tenths of a degree, with the average error being 0.5°–0.6°C, or about 5% of the observed range.
The simulated temperature excess of the urban site relative to the other two nonurban sites (the UHI) is reasonable for most days at the indicated times, but the error can occasionally be significant (Table 1). For example, on the 3 days out of the 8 with the largest observed urban–rural nocturnal UHI temperature excess (16, 17, and 18 July), the model also predicted the largest UHI effect. The UHI effect is correctly simulated to be larger at night (2.0°–4.0°C urban–rural difference), relative to other times of the day (1.0°–2.5°C during the day). The simulated urban–rural temperature difference, averaged over the 7 days with southerly flow, is 2.54°C, which is only 0.11°C less than observed. Forecasts with misplaced isolated precipitation were occasionally responsible for temperature errors.
The observations show that the nocturnal temperature at the rural site is warmer than that at the suburban site on 24 July, whereas the observed relationship has the opposite sign on 16 July (Fig. 7) and the other 6 days. The model captured the reversal on this day. On the 24th, the observed and predicted large-scale winds veered during the night from northeasterly to southeasterly, which meant that the rural site to the southwest of the city could have been influenced by the advection of the UHI warm anomaly, the “thermal shadow,” and therefore remained warmer than the upwind suburban location.
The forecast temperatures have a 1°–3°C positive bias (Fig. 7).
It is encouraging that the model simulations verified reasonably well, even though the conventional Eta Model soil moisture and temperature fields were used to initialize RTFDDA, and the complexities of the urban radiation budget were not seriously treated.
2) Near-surface humidity
Relative humidity (RH) differences between urban and rural areas result from both the UHI effect and specific humidity differences. Because of strong city-to-city variability in the amplitude of the UHI effect, it is difficult to state a typical urban–rural RH contrast. However, observations for individual cities have shown contrasts that vary from 5% to 30%, depending on season and time of day. For the dates studied here, 8-day-averaged observed and simulated 10-m urban RH deficits (relative to the rural location) are 16% and 14%, respectively, for 0600 LT. Specific humidity derived from JU2003 observations shows the urban location to be 0–2 g kg−1 drier than the rural one, depending on the time of day and date. This is not inconsistent with documented urban–rural contrasts of 0.4 g kg−1 for European cities (Yoshino 1975) and 0.1 g kg−1 for New York City, New York, in March (Clark et al. 1985). The model’s specific humidity shows urban–rural differences of about that sign and magnitude for most dates and times. The model also does well at representing the observed diurnal cycle of specific and relative humidity in terms of phase and amplitude. The JU2003 data for the three sites typically show the double maximum that is documented in the literature for specific humidity, with one in the midmorning and one in the early evening (Geiger 1966). The model solution picks up this observed double maximum, as well as the observed daily range of 3–5 g kg−1. Figure 8 shows the observed and analyzed specific humidity for 16 July, where the model greatly overestimated the amplitude of the evening maximum at the rural site, but otherwise did reasonably well at capturing the main features of the diurnal cycle.
b. Verification of the simulated urban BL development
During JU2003, three BL wind profilers (915 MHz) from ANL, the Pacific Northwest National Laboratory (PNNL), and the University of Oklahoma (OU) operated in the central urban area of OKC (see Fig. 2 for locations of the ANL and PNNL systems). The vertical profiles of the SNR from the ANL profiler are shown in Fig. 9 for each of the eight study days, along with the model BL height forecasts at the locations of the ANL and PNNL profilers. The SNR from the PNNL profiler is very similar to that shown because of their proximity (6 km) and the fact that the sampled volumes of air overlap. The model reasonably simulated the urban BL in terms of the diurnal evolution of the BL height. In particular, the model nicely captured the growth rates of the convective BL from sunrise to early afternoon, as can be seen by comparing the slope of the model solutions and that of the SNR gradient. With some exceptions, the model also simulated the nighttime shallow BLs well, which are affected by the prevailing low-level jets (LLJs), the UHI, and the surface radiational cooling. Furthermore, the model seems to represent some of the day-to-day variation in the BL growth. For example, the difference in the growth rates on 4 and 15 July are captured. It is worth repeating that in the late afternoon and evening when the surface has cooled, the BL collapses but residual turbulence still exists. Thus, the maximum gradient in the SNR at this time may not indicate the BL height. Last, it is of interest to understand why the forecast BL heights at the two nearby locations differ by as much as they do on some days, with the BL height at the PNNL location in the southern downtown area generally being less than at the ANL location in the northern part of the city. We speculate that the upwind PNNL location is influenced more by the shallower BL that advects into the urban area from the south. In contrast, the warm urban surface causes the simulated BL to deepen in the southerly flow, and therefore the BL is deeper in the northern part of the city.
c. Simulated UHI dynamics and thermodynamics
In this section, the three-dimensional (3D) structure of the UHI and its effects on the BL are analyzed using model output. Because observations are insufficient to define the 3D structure on the metropolitan area scale, comparisons will be made with existing conceptual models of the UHI. Because the point comparisons previously discussed show that the model reasonably represents selected, observed BL features, the model-produced analyses where there are no data can be used here with some confidence.
The RTFDDA analyses on model domain 4 for the eight clear-sky days were composited, and the surface fluxes and lower-tropospheric variables were analyzed to document the features of the UHI and the associated BL structure. Figure 10 shows that the simulated nocturnal UHI effects are prominent in the average. The temperature at 2 m AGL for local midnight is about 2°–3°C warmer in the urban area than in surrounding rural regions (Fig. 10a), where the shape of the thermal excess corresponds to the shape of the urbanized area. There is some downwind displacement of the warm anomaly, however. Figure 10b indicates that the simulated urban surface sensible heat fluxes are less negative than in the surrounding rural area, and that they approach zero in some places. For individual dates, the nocturnal urban sensible heat flux varies between zero and −40 W m−2, which agrees well with observations taken by Grimmond et al. (2004) at a few urban locations during JU2003. Also, on a few dates there were urban regions where small upward heat fluxes prevailed for most of the night. This is consistent with the general discussion of urban surface fluxes in Grimmond and Oke (2002), and with the observation of positive nocturnal surface sensible heat fluxes of 40–50 W m−2 at an urban site during JU2003 (Gouveia et al. 2004).
One major difference between urban and rural landscapes is that the urban surface is much drier, and there is less transpiration. The distributions of the 8-day-averaged simulated specific humidity and RH at 15 m AGL are given in Fig. 11 for 0000 and 1200 LT. It can be seen that the OKC urban area is characterized by a dry core, during both day and night, with a RH anomaly of about 10% and a specific humidity anomaly of about 0.5–1.5 g kg−1. Obviously, the RH deficit reflects the effects of both the positive temperature anomaly and the negative specific humidity anomaly. The dry anomaly is displaced tens of kilometers downwind.
The model simulation for 24 July illustrates an interesting possible impact of the urban area on a LLJ. Figure 12a shows the model-simulated BL height for 0000 LT, with rural values of about 60 m and urban center values in excess of 250 m AGL. On this night, there was an LLJ, which is typical in the summer over the Great Plains (Wu and Raman 1998; Banta et al. 2002), that was revealed by lidar observations during JU2003 (Wang et al. 2004; Zajic et al. 2004). The 1200 UTC sounding from OU in Norman, Oklahoma, observed a 10+ m s−1 speed maximum at 600 m AGL. The model analysis (Fig. 12b) shows a maximum LLJ wind speed of about 10 m s−1 in the rural area and <8 m s−1 over the urban area. This lower speed over the city is presumably because of both the larger roughness elements and the slightly more unstable surface layer there. The lower speeds over the city and the deeper BL show characteristics similar to those of the supercritical flow and the associated compression fan observed and modeled by many (e.g., see Haack et al. 2001). The simulated wind speed maximum is much closer to the surface than was observed, even though the low-level wind enhancement extended through a depth of over 500 m above the surface.
Large daytime differences in rural–urban BL development can also be seen. Figure 13 shows the 1200 LT BL height on the 1.5-km grid (domain 4) and cross sections of vertical wind speed, horizontal divergence, and equivalent potential temperature θe. The daytime BL depth over the OKC area was 300–500 m greater than that over rural areas (Fig. 13a), where the urban anomaly was significantly displaced downwind (to the north) of the surface forcing. The deeper BL in the eastern part of the domain is apparently unrelated to the urban area. Also, organized mesoscale circulations develop as a result of the differential heating between the urban and rural areas (Figs. 13b,c), forming a convergence–divergence couplet in the lowest 2 km and upward motion exceeding 0.5 m s−1 from 1.0 to 1.5 km over the city. The urban anomaly in equivalent potential temperature is also apparent (Fig. 13d).
Finally, even with the simple representation of the urban area, the mesoscale model was able to simulate qualitatively reasonable features of the urban BL. Although the general properties of the simulated three-dimensional structures are conceptually consistent with results from other field experiments and previous model results (Oke 1987; Comrie 2000; Gallo and Owen 1999; Hafner and Kidder 1999), better three-dimensional observational datasets are needed for metropolitan and surrounding areas to verify these structures properly.
6. Summary and conclusions
The MM5-based mesoscale data-assimilation and forecasting system developed by NCAR for operational use at Army test ranges provided operational modeling support for the OKC JU2003 field program. In particular, the model-predicted variables were used by forecasters to help to plan for intensive observation periods. Because this was an urban T&D study, of special importance was the accurate prediction of the prevailing wind field in the BL. Thus, in preparation for this model application, some known MM5 deficiencies related to the prediction of BL height and winds were corrected. This paper describes these improvements; shows the benefit to forecast skill that results, based on pre-field-program verification studies using BL data; and then evaluates the skill of the operational system during the field program.
The model improvements involved implementation of a better representation of surface-layer momentum fluxes, a more accurate diagnosis of the BL depth, and a bulk parameterization of urban substrate properties in the LSM. Evaluation of the benefits of these changes prior to the field program showed that the well-known problem of a large negative bias in the near-surface wind speed during the day was eliminated, with daytime errors now comparable to those at night. Small improvements also occurred in the simulated near-surface temperature and wind direction during the day. In addition, profiler SNR data from Kansas indicated that the BL height simulations were improved as well.
Selected operational model products available during the field program were then compared with observations. Model-based analyses and forecasts of near-surface temperature and specific humidity showed reasonably accurate contrasts between urban, suburban, and rural locations. The predicted phase and range of the diurnal temperature wave was handled well, as was the more complex diurnal pattern of specific humidity. On one of the days studied, the model also reproduced what was apparently an observed nocturnal thermal shadow downwind of the urban area. Wind forecasts were not compared with observations from the urban area because street canyon winds are not comparable with those from the model, which represent grid-box averages. The model BL growth rate over the city corresponded well, on most days, with that inferred from profiler SNR data.
Last, the model-analyzed 3D thermal, humidity, and wind anomalies associated with the urban area were examined. Because the model forecasts verified well where data were available, it is reasonable to have some confidence in the model realizations of the 3D structures that cannot be well observed. Based on a multiday composite, the nocturnal urban temperature was about 2.5°C warmer than that of the surrounding rural area, and the nocturnal surface heat fluxes were 20–30 W m−2 larger in the city. The enhanced nocturnal mixing and BL height over the city reduced the speed of an LLJ on one night.
Perhaps one of the most interesting, and possibly surprising, implications of these results is that a mesoscale model with a simple treatment of the land surface physics and properties can reasonably replicate the bulk effect on the BL of an urban area. This is not to say that the specific effects of buildings and street canyons, such as radiation trapping, are not important to the bulk effect of the city on the energy and moisture budgets, but that first-order effects can be represented reasonably well with simple treatments of the land surface that do not include urban canopy models.
Acknowledgments
The development of the RTFDDA system was funded by the U.S. Army Test and Evaluation Command through an Interagency Agreement with the National Science Foundation. Support for the JU2003 operational modeling described here was provided by the Army Test and Evaluation Command and the Defense Threat Reduction Agency (DTRA). Jeff Basara’s analysis of the field program data was funded by DTRA and by National Aeronautics and Space Administration New Investigator Award NA17RJ1277.
REFERENCES
Allwine, K. J., J. H. Shinn, G. E. Streit, K. L. Clawson, and M. Brown, 2002: Overview of Urban 2000—A multiscale field study of dispersion through an urban environment. Bull. Amer. Meteor. Soc., 83 , 521–536.
Allwine, K. J., M. J. Leach, L. W. Stockham, J. S. Shinn, R. P. Hosker, J. F. Bowers, and J. C. Pace, 2004: Overview of Joint Urban 2003—An atmospheric dispersion study in Oklahoma City. Preprints, Symp. on Planning, Nowcasting, and Forecasting in the Urban Zone, Seattle, WA, Amer. Meteor. Soc., CD-ROM, J7.1.
Atkinson, B. W., 2003: Numerical modeling of urban heat-island intensity. Bound.-Layer Meteor., 113 , 285–310.
Banta, R. M., R. K. Newsom, J. K. Lundquist, Y. L. Pichugina, R. L. Coulter, and L. Mahrt, 2002: Nocturnal low-level jet characteristics over Kansas during CASE-99. Bound.-Layer Meteor., 105 , 221–252.
Beljaars, A. C. M., 1995: The parameterization of surface fluxes in large-scale models under free-convection. Quart. J. Roy. Meteor. Soc., 121 , 255–270.
Brock, F. V., K. C. Crawford, R. L. Elliott, G. W. Cuperus, S. J. Stadler, H. L. Johnson, and M. D. Eilts, 1995: The Oklahoma Mesonet: A technical overview. J. Atmos. Oceanic Technol., 12 , 5–19.
Brown, M., 2000: Urban parameterization for mesoscale models. Mesoscale Atmospheric Dispersion, Z. Boybey, Ed., Wessex Press, 193–225.
Chen, F., and J. Dudhia, 2001: Coupling an advanced land surface–hydrology model with the Penn State/NCAR MM5 modeling system. Part I: Model implementation and sensitivity. Mon. Wea. Rev., 129 , 569–585.
Chen, F., Z. I. Janjic, and K. Mitchell, 1997: Impact of atmospheric surface-layer parameterizations in the new land-surface scheme of the NCEP Mesoscale Eta model. Bound.-Layer Meteor., 85 , 391–421.
Chen, F., Y. Liu, H. Kusaka, M. Tewari, and J-W. Bao, 2004: Utilizing the coupled WRF/LSM/urban modeling system with detailed urban classification to simulate the urban heat island phenomena over the greater Houston area. Preprints, Fifth Urban Environment Conf., Vancouver, BC, Canada, Amer. Meteor. Soc., CD-ROM, 9.11.
Cheng, F., D. Byun, and S. Kim, 2003: Sensitivity study of the effects of land surface characteristics on meteorological simulations during the TexAQS2000 period in the Houston-Galveston area. Preprints, 13th PSU/NCAR Mesoscale Model Users’ Workshop, Boulder, CO, NCAR, 2.7.
Clark, E. C., R. D. Bornstein, and Y. T. Tam, 1985: Current and potential anthropogenic moisture effects on the New York City planetary boundary layer. J. Air Pollut. Control Assoc., 35 , 831–833.
Comrie, A. C., 2000: Mapping a wind-modified urban heat island in Tucson (with comments on integrating research and undergraduate learning). Bull. Amer. Meteor. Soc., 81 , 2417–2431.
Coulter, R. L., and D. H. Holdridge, 1998: A procedure for the automatic estimation of mixed layer height. Proc. Eighth Atmospheric Radiation Measurement (ARM) Program Science Team Meeting, Tucson, AZ, Department of Energy Office of Energy Research, 177–180.
Cram, J. M., Y. Liu, S. Low-Nam, R-S. Sheu, L. Carson, C. A. Davis, T. T. Warner, and J. F. Bowers, 2001: An operational mesoscale RT-FDDA analysis and forecasting system. Preprints, 18th Conf. on Weather Analysis and Forecasting and 14th Conf. on Numerical Weather Prediction, Ft. Lauderdale, FL, Amer. Meteor. Soc., CD-ROM, J2.11.
Deardorff, J. W., 1972: Numerical investigation of neutral and unstable planetary boundary layer. J. Atmos. Sci., 29 , 91–115.
Doran, J. C., J. D. Fast, and J. Horel, 2002: The VTMX 2000 Campaign. Bull. Amer. Meteor. Soc., 83 , 537–551.
Dupont, S., T. L. Otte, and J. K. S. Ching, 2003: Simulation of meteorological fields within and above urban and rural canopies with a mesoscale model (MM5). Bound.-Layer Meteor., 113 , 111–158.
Ek, M. B., K. E. Mitchell, Y. Lin, E. Rogers, P. Grunmann, V. Koren, G. Gayno, and J. D. Tarpley, 2003: Implementation of Noah land surface model advances in the National Centers for Environmental Prediction operational Mesoscale Eta model. J. Geophys. Res., 108 .8851, doi:10.1029/2002JD003296.
Gallo, K. P., and T. W. Owen, 1999: Satellite-based adjustments for urban heat island temperature bias. J. Appl. Meteor., 38 , 806–813.
Geiger, R., 1966: The Climate near the Ground. Harvard University Press, 611 pp.
Gouveia, F. J., M. J. Leach, and J. H. Smith, 2004: Measurements of net radiation, ground heat flux and surface temperature in an urban canyon. Preprints, Symp. on Planning, Nowcasting, and Forecasting in the Urban Zone, Seattle, WA, Amer. Meteor. Soc., CD-ROM, J7.4.
Grell, G., J. Dudhia, and D. Stauffer, 1993: A description of the Fifth-Generation Penn State/NCAR Mesoscale Model (MM5). NCAR Tech. Note NCAR/TN-398+STR, 122 pp.
Grimmond, C. S. B., and T. R. Oke, 2002: Turbulent heat fluxes in urban areas: Observations and a local-scale urban meteorological parameterization scheme (LUMPS). J. Appl. Meteor., 41 , 792–810.
Grimmond, C. S. B., H-B. Su, B. Offerle, B. Crawford, S. Scott, S. Zhong, and C. Clements, 2004: Variability of sensible heat fluxes in a suburban area of Oklahoma City. Preprints, Symp. on Planning, Nowcasting, and Forecasting in the Urban Zone, Seattle, WA, Amer. Meteor. Soc., CD-ROM, J7.2.
Grossman-Clarke, S., J. A. Zehnder, W. L. Stefanov, Y. Liu, and M. A. Zoldak, 2005: Urban modifications in a mesoscale meteorological model and the effects on near-surface variables in an arid metropolitan region. J. Appl. Meteor., 44 , 1281–1297.
Haack, T., S. D. Burk, C. Dorman, and D. Rogers, 2001: Supercritical flow interaction within the Cape Blanco–Cape Mendocino orographic complex. Mon. Wea. Rev., 129 , 688–708.
Hafner, J., and S. Q. Kidder, 1999: Urban heat island modeling in conjunction with satellite-derived surface/soil parameters. J. Appl. Meteor., 38 , 448–465.
Holt, T., D. Niyogi, F. Chen, K. Manning, M. A. LeMone, and A. L. Qureshi, 2006: Effect of land–atmosphere interactions on the IHOP 24–25 May 2002 convection case. Mon. Wea. Rev., 134 , 113–133.
Hong, S-Y., and H-L. Pan, 1996: Nonlocal boundary layer vertical diffusion in a Medium-Range Forecast model. Mon. Wea. Rev., 124 , 2322–2339.
Kusaka, H., H. Kondo, Y. Kikegawa, and F. Kimura, 2001: A simple single-layer urban canopy model for atmospheric models: Comparison with multi-layer and slab models. Bound.-Layer Meteor., 101 , 329–358.
LeMone, M. A., and Coauthors, 2000: Land–atmosphere interaction research, early results, and opportunities in the Walnut River Watershed in southeast Kansas: CASE and ABLE. Bull. Amer. Meteor. Soc., 81 , 757–779.
Liu, Y., and Coauthors, 2002: Performance and enhancements of the NCAR/ATEC mesoscale FDDA and forecasting system. Preprints, 15th Conf. on Numerical Weather Prediction, San Antonio, TX, Amer. Meteor. Soc., 399–402.
Martilli, A., A. Clappier, and M. W. Rotach, 2002: An urban surface exchange parameterization for mesoscale models. Bound.-Layer Meteor., 104 , 261–304.
Masson, V., 2000: A physically-based scheme for the urban energy budget in atmospheric models. Bound.-Layer Meteor., 94 , 357–397.
Niyogi, D., S. Raman, and K. Alapaty, 1999: Uncertainty in specification of surface characteristics, Part ii: Hierarchy of interaction—Explicit statistical analysis. Bound.-Layer Meteor., 91 , 341–366.
Oke, T. R., 1987: Boundary Layer Climates. 2d ed. Routledge and John Wiley and Sons, 435 pp.
Otte, T. L., A. Lacser, S. Dupont, and J. K. S. Ching, 2004: Implementation of an urban canopy parameterization in a mesoscale meteorological model. J. Appl. Meteor., 43 , 1648–1665.
Pielke, R. A., 1984: Mesoscale Meteorological Modeling. Academic Press, 612 pp.
Rotach, M. W., 1995: Profiles of turbulence statistics in and above an urban street canyon. Atmos. Environ., 29 , 1473–1486.
Stauffer, D. R., and N. L. Seaman, 1994: Multiscale four-dimensional data assimilation. J. Appl. Meteor., 33 , 416–434.
Sykes, R. I., D. S. Henn, and W. S. Lewellen, 1993: Surface-layer description under free-convection conditions. Quart. J. Roy. Meteor. Soc., 119 , 409–421.
Taha, H., 1999: Modifying a mesoscale meteorological model to better incorporate urban heat storage: A bulk-parameterization approach. J. Appl. Meteor., 38 , 466–473.
Uno, I., H. Ueda, and S. Wakamatsu, 1989: Numerical modeling of nocturnal urban boundary layer. Bound.-Layer Meteor., 49 , 77–98.
Wang, Y., D. Ligon, E. Creegan, C. Williamson, C. Klipp, M. Felton, and R. Calhoun, 2004: Turbulence characteristics over an urban domain observed by Doppler lidars. Preprints, 16th Symp. on Boundary Layers and Turbulence, Portland, ME, Amer. Meteor. Soc., CD-ROM, 6.3.
Wu, Y., and S. Raman, 1998: The summertime Great Plains lower level jet and the effect of its origin on moisture transport. Bound.-Layer Meteor., 88 , 445–466.
Yoshino, M. M., 1975: Climate in a Small Area. University of Tokyo Press, 549 pp.
Zajic, D., H. J. S. Fernado, M. Princevac, and R. Calhoun, 2004: Flow and turbulence in urban canopies. Preprints, Fifth Conf. on Urban Environment, Vancouver, BC, Canada, Amer. Meteor. Soc., CD-ROM, 6.2.
Zhang, D-L., and R. A. Anthes, 1982: A high-resolution model of the planetary boundary layer—Sensitivity tests and comparisons with SESAME-79 data. J. Appl. Meteor., 21 , 1594–1609.
Zhang, D-L., and W. Zheng, 2004: Diurnal cycles of surface winds and temperatures as simulated by five boundary-layer parameterizations. J. Appl. Meteor., 43 , 157–169.
Zilitinkevich, S., 1995: Non-local turbulent transport: Pollution dispersion aspects of coherent structure of convective flows. Air Pollution Theory and Simulation, H. Power, N. Moussiopoulos, and C. A. Brebbia, Eds., Vol. I, Air Pollution III, Computational Mechanics Publications, 53–60.
Bias errors of wind speed at 10 m AGL for 10–12-h forecasts on the fine mesh of a nested grid (either 1.1- or 3.3-km grid increment). These are averaged statistics for five Army test ranges, as verified against the range observations (71 mesonet stations) during August 2002. Plots are shown for the Cold Regions Test Center (CRTC) in Alaska, Dugway Proving Ground (DPG) in Utah, White Sands Missile Range (WSMR) in New Mexico, Yuma Proving Ground (YPG) in Arizona, and Aberdeen Test Center (ATC) in Maryland.
Citation: Journal of Applied Meteorology and Climatology 45, 7; 10.1175/JAM2383.1
(left) The RTFDDA grid configuration, terrain height in domain 1, and (right) land use in domain 4. The urban areas (gray shades in domains 4 and 5) of OKC and Norman are shown. The A and P symbols indicate the locations of the ANL and PNNL profilers, respectively. The M, S, and W symbols show the locations of the following three surface stations used in the model verification: Oklahoma Mesonet stations MINC and SPEN, and the Dugway Proving Ground PWIDS 14, respectively.
Citation: Journal of Applied Meteorology and Climatology 45, 7; 10.1175/JAM2383.1
The signal-to-noise ratio (color) for the ANL 915-MHz wind profilers in (top) Beaumont, (middle) Whitewater, and (bottom) Oxford, KS, observed from 1800 LT 27 May to 1800 LT 28 May 2003. Superimposed are the BL heights of the RTFDDA 0–13-h forecasts beginning at 0500 LT, interpolated to the profiler locations, with the old (black) and new (red) surface-layer physics and BL height diagnosis schemes.
Citation: Journal of Applied Meteorology and Climatology 45, 7; 10.1175/JAM2383.1
Comparison of the old and new MRF BL schemes for the period from 0000 UTC 27 May to 0000 UTC 29 May 2003 in terms of (a) bias and (b) RMS errors for analyses (Anal) and 10–12-h forecasts (f10–12) of 10-m-AGL wind speed, based on over 1400 METAR and other observations in model domain 1. Forecasts were initiated every 3 h over the period. Values were interpolated from the model grid to the locations of the observations. Because a 48-h period is used for verification, each data point on the abscissa represents an average of two analyses or forecasts that are valid at that time of day. The verification is performed by horizontally interpolating the model output to the observation locations, and using similarity theory to extrapolate downward from the lowest model computation level at ∼15 m AGL to the level of the observation.
Citation: Journal of Applied Meteorology and Climatology 45, 7; 10.1175/JAM2383.1
Same as Fig. 4, but for the 10-m-AGL wind direction and 2-m-AGL temperature RMS errors.
Citation: Journal of Applied Meteorology and Climatology 45, 7; 10.1175/JAM2383.1
Comparison of the monthly mean 10-m-AGL wind speed bias and RMS errors of analyses (Anal) and 4–6-h forecasts (f4–6) from a 10-km grid increment grid of the operational mesoscale forecasting system at the ATEC Aberdeen Test Center, MD, for August 2002, before the MRF physics changes were made, and for August 2003, after they were made. The verification is performed by horizontally interpolating the model output to the observation locations, and using similarity theory to extrapolate downward from the lowest model computation level at ∼15 m AGL to the level of the observation.
Citation: Journal of Applied Meteorology and Climatology 45, 7; 10.1175/JAM2383.1
The diurnal evolution of the (right) observed and (left) model-analyzed ∼10-m-AGL temperatures at the urban, rural, and suburban stations for 16 (southerly flow) and 24 (easterly flow) Jul 2003. The verification results for the other six clear-sky days are generally similar to those for the 16 Jul case. Note that PW14 observations are at 10 m AGL while MINC and SPEN observations are at 9 m AGL. Model values are obtained at the observation locations by horizontally interpolating the model output to the observation locations and using similarity theory to extrapolate downward from the lowest model computation level at ∼15 m AGL to the level of the observation.
Citation: Journal of Applied Meteorology and Climatology 45, 7; 10.1175/JAM2383.1
Same as Fig. 7, but for specific humidity and the 16 July case only.
Citation: Journal of Applied Meteorology and Climatology 45, 7; 10.1175/JAM2383.1
Vertical profiles of the SNR from the ANL profiler for each of the eight study days, with the model-analyzed BL height at the locations of the ANL (red) and PNNL (black) profilers.
Citation: Journal of Applied Meteorology and Climatology 45, 7; 10.1175/JAM2383.1
Composite 2-m-AGL (a) temperature (°C) and (b) surface sensible heat flux (W m−2), based on RTFDDA analyses, for the nine clear-sky days in July 2003, valid at 0000 LT. The isopleth interval for temperature in (a) is 0.5°C and for sensible heat flux in (b) is 10 W m−2. Model temperatures are obtained at 2 m AGL by using similarity theory to extrapolate downward from the lowest model computation level at ∼15 m AGL. The OKC urban area is marked with thick solid lines.
Citation: Journal of Applied Meteorology and Climatology 45, 7; 10.1175/JAM2383.1
Mean 2-m-AGL (a), (c) RH (isopleth interval is 5%) and (b), (d) specific humidity (isopleth interval is 0.5 g kg−1) at (a), (b) 0000 and (c), (d) 1200 LT, based on RTFDDA analyses. Model humidities are obtained at 2 m AGL by using similarity theory to extrapolate downward from the lowest model computation level at ∼15 m AGL. The OKC urban area is marked with thick solid lines.
Citation: Journal of Applied Meteorology and Climatology 45, 7; 10.1175/JAM2383.1
The 0000 LT 24 Jul 2003 RTFDDA-analyzed (a) BL height on domain 4 with a 1.5-km grid increment, and (b) wind speed perpendicular to the cross section AB. The isopleth interval for the BL height in (a) is 20 m and the interval of wind speed in (b) is 1 m s−1. The OKC urban area is marked with thick solid lines in (a).
Citation: Journal of Applied Meteorology and Climatology 45, 7; 10.1175/JAM2383.1
The 1200 LT 5 Jul 2003 RTFDDA-analyzed (a) BL height on domain 4 with a 1.5-km grid increment (isopleth interval of 150 m), and vertical cross sections of (b) vertical motion (isopleth interval of 0.1 m s−1), (c) horizontal divergence (isopleth interval of 10 × 10−5 s−1), and (d) θe (isopleth interval of 2 K) along the line AB in (a). The OKC urban area is marked with thick solid lines in (a). The approximate extent of the urban area is indicated in the cross sections.
Citation: Journal of Applied Meteorology and Climatology 45, 7; 10.1175/JAM2383.1
Comparison of the simulated and observed 10-m temperature differences between urban (PW14) and rural (MINC) locations and between urban and suburban (SPEN) locations at the designated times, and the range in the diurnal temperature variation at the three stations. The observations are to the right of the slashes, and the simulated values are to the left (24 Jul, the date with easterly winds, is not included in the mean in the bottom row).
