Abstract

In this study, observational and numerical modeling analyses based on the Weather Research and Forecasting Model (WRF) are used to investigate the impact of urbanization on heavy rainfall over the Milwaukee–Lake Michigan region. The authors examine urban modification of rainfall for a storm system with continental-scale moisture transport, strong large-scale forcing, and extreme rainfall over a large area of the upper Midwest of the United States. WRF simulations were carried out to examine the sensitivity of the rainfall distribution in and around the urban area to different urban land surface model representations and urban land-use scenarios. Simulation results suggest that urbanization plays an important role in precipitation distribution, even in settings characterized by strong large-scale forcing. For the Milwaukee–Lake Michigan region, the thermodynamic perturbations produced by urbanization on the temperature and surface pressure fields enhance the intrusion of the lake breeze and facilitate the formation of a convergence zone, which create favorable conditions for deep convection over the city. Analyses of model and observed vertical profiles of reflectivity using contoured frequency by altitude displays (CFADs) suggest that cloud dynamics over the city do not change significantly with urbanization. Simulation results also suggest that the large-scale rainfall pattern is not sensitive to different urban representations in the model. Both urban representations, the Noah land surface model with urban land categories and the single-layer urban canopy model, adequately capture the dominant features of this storm over the urban region.

1. Introduction

Urban modification of precipitation has been extensively investigated in the past few decades. Observational and modeling studies point to three main classes of mechanisms that result in urban modification of precipitation: urban heat island (UHI) effects (e.g., Bornstein and Lin 2000; Dixon and Mote 2003; Oke 1982), urban canopy effects (e.g., Chen et al. 2011a; Miao et al. 2009; Zhang et al. 2011), and urban aerosol effects (e.g., Jin et al. 2010; Ntelekos et al. 2009; Rosenfeld 2000). Surveys of urban precipitation modification research are presented in Lowry (1998), Shepherd (2005), Collier (2006), and Shepherd et al. (2010b).

The setting for examining urban modification of precipitation in this study is the Milwaukee, Wisconsin, metropolitan region, which is located along the western boundary of Lake Michigan (Fig. 1). Previous studies (see, e.g., Ntelekos et al. 2008; Shepherd et al. 2010a) have pointed to the importance, and complexities, of examining urban modification of precipitation along major land–water boundaries like Lake Michigan. Previous research (Yang et al. 2013) has shown striking spatial heterogeneities of rainfall over the urban core of the Milwaukee region for flood-producing storms.

Fig. 1.

(a) Domain configurations of the model and the observation networks: (left) domains 1–3 with (right) zoom on domain 3. Red box in domain 3 denotes the urban focus domain. Colors represent the elevation (m MSL) within the domains. (b) Land use for the urban focus domain. MMSD rain gauge locations and Menomonee River gauging station are shown. Two dashed black boxes denote the urban core and rural area near the city, which are used for analyses in section 5. Line A–B crosses the urban core and is located approximately 43.05°N.

Fig. 1.

(a) Domain configurations of the model and the observation networks: (left) domains 1–3 with (right) zoom on domain 3. Red box in domain 3 denotes the urban focus domain. Colors represent the elevation (m MSL) within the domains. (b) Land use for the urban focus domain. MMSD rain gauge locations and Menomonee River gauging station are shown. Two dashed black boxes denote the urban core and rural area near the city, which are used for analyses in section 5. Line A–B crosses the urban core and is located approximately 43.05°N.

In this paper, we focus on numerical modeling studies of a typical heavy rainfall event (see the descriptions in section 3a) over the Milwaukee metropolitan region on 22–23 July 2010 using the Weather Research and Forecasting (WRF) Model. For this region, land–water interactions, especially the local circulations induced by differential heating, are a special focus of numerical modeling studies.

Ntelekos et al. (2008) found a pronounced impact of Chesapeake Bay on the storm evolution and rainfall distribution for an organized thunderstorm system over the Baltimore metropolitan region. Shepherd et al. (2010a) examined the role of the sea-breeze circulation in the development of a convergence zone over the western fringe of Houston, which is responsible for elevated precipitation. For coastal/lakeside cities, the heat island circulation (HIC) can be an important factor for the sea-/lake-breeze circulation. Ohashi and Kida (2002) showed that the nonlinear effects of costal urban regions can increase the penetrating speed of the sea-breeze front. Freitas et al. (2006) found that the UHI can produce a convergence zone in the center of São Paulo, Brazil, a coastal city, which then accelerates the propagation of the sea breeze. Lo et al. (2007) and Lin et al. (2008) obtained similar results over the Pearl River delta region in China and northern Taiwan, respectively. More recently, Carter et al. (2012) simulated a sea-breeze day over Houston using WRF coupled with an urban canopy model (UCM). Their findings reinforced the aforementioned results. However, Keeler and Kristovich (2012) analyzed the lake-breeze front propagation (during April–September of 2005) through Chicago and its surrounding areas (including Milwaukee) by means of radar and surface observations. They found that the lake-breeze front could be slowed because of the UHI effect. Their conclusions are supported by Sarkar et al. (1998). The contrasting response of lake- and sea-breeze circulations in different settings to the UHI points to the need for additional study.

Instead of examining storm systems that initiate and develop in close proximity to the urban region (e.g., Miao et al. 2011; Shem and Shepherd 2009), we focus on storm systems that have long lifetimes and produce heavy rainfall over large areas. Shepherd et al. (2011) provided detailed analyses of the Atlanta floods of 2009. In that study, they pointed out the potential influence of urban land cover interactions on the heavy flood-producing rainfall event.

Yang et al. (2013) present analyses of 18 storms that resulted in the largest flash floods in the Milwaukee region. As noted above, one of the conclusions of that study is that extreme flood-producing rainfall is concentrated in the urban core of Milwaukee. The second conclusion of that study (see also Smith et al. 2012; Thorndahl et al. 2013, manuscript submitted to Atmos. Res.) is that storms that produce flash floods in Milwaukee are organized thunderstorm systems that are heavy rainfall producers over a large region of the upper Midwest. This motivates the study of urban modification of rainfall for storm systems with strong synoptic-scale forcing (see section 3 for details on how we characterize strong synoptic-scale forcing).

The objectives of this paper are to examine the mechanisms of urban modification of rainfall under strong synoptic-forcing conditions. We focus especially on the interactions of the UHI and the lake breeze and their possible roles in modifying the precipitation distribution for organized convective systems. Our analyses are motivated by the following hypotheses: 1) urbanization plays an important role in affecting the rainfall distribution by altering the storm evolution, even under strong synoptic forcing; 2) for the Milwaukee–Lake Michigan region, the urban effect enhances the intrusion of the lake breeze and facilitates the formation of convergence zone within the city; 3) local urban forcing on atmospheric processes affects local rainfall distribution, but the effects diminish rapidly with distance from the urban area; 4) coupled atmosphere–land surface modeling with either the Noah land surface model (LSM; with simple urban treatment) or Noah coupled with a single-layer urban canopy model can capture the key impacts of urban terrain on storm features quite well. The methodology used to examine these hypotheses is based mainly on numerical simulations with contrasting model configurations and physics options. We also employed a storm-tracking algorithm (Dixon and Wiener 1993) to examine Lagrangian properties of both observed and simulated storms based on 3D radar reflectivity fields. In addition, we synthesize properties of the ensemble of storms that pass over the study region using the contoured frequency by altitude displays (CFADs; see Yuter and Houze 1995).

The paper is organized as follows. In section 2, we introduce the study region, data, model configuration, and experimental setup. Analyses of both observed and simulated storm properties are presented in section 3. In section 4, the model is validated through comparisons with multiple datasets. We analyze and discuss the simulation results in section 5. A summary and conclusions are presented in section 6.

2. Data and methodology

a. Study region, observations, and storm-tracking algorithm

The complete study area covers the upper Midwest of the United States (denoted as domain 1 in Fig. 1a). Elevation ranges from less than 100 to more than 600 m MSL. The Milwaukee metropolitan area is located in the center of the area, along the western boundary of Lake Michigan. The Menomonee River basin drains a large portion of the city, with a drainage area of 319 km2 (Yang et al. 2013). An “urban focus” domain was also established (Fig. 1b), which centers over the urban area. Detailed analyses will be given for this domain.

There are well-maintained observational networks in the research area. Automated surface observing system (ASOS) stations are distributed in domain 1 (26 sites), which provide time series observations of air temperature and specific humidity (at the height of 2 m) that we use for model validation. The Milwaukee Metropolitan Sewerage District (MMSD) maintains a network of 21 rain gauges. The rainfall observations are used for bias correction of radar rainfall fields.

Three-dimensional volume scan radar reflectivity fields were obtained from the KMKX Weather Surveillance Radar-1988 Doppler (WSR-88D; see radar location in Fig. 1) for the 22–23 July 2010 storm period. We used the Hydro–Next Generation Weather Radar (NEXRAD) system to convert the reflectivity fields into rainfall fields with 15-min interval and 1-km spatial resolution (e.g., Smith et al. 2012; Wright et al. 2012). Bias correction was implemented for the estimated rainfall fields as in Yang et al. (2013).

In addition to rainfall estimation, 3D radar reflectivity fields also played an important role in analyzing storm properties. We used the Thunderstorm Identification Tracking, Analysis and Nowcasting (TITAN) storm-tracking algorithms (Dixon and Wiener 1993) to examine the evolution of storm features for the life cycle of the storm.

The velocity–azimuth display (VAD) algorithms of Matejka and Srivastava (1991) are used to develop vertical profiles of the horizontal wind from KMKX Doppler velocity observations at 5–6-min time intervals. We use these observations to examine boundary layer winds during the storm and to assess WRF simulations of boundary layer wind profiles.

b. Model configuration

The WRF is a fully compressible, nonhydrostatic, mesoscale model. The Advanced Research version WRF (ARW) version 3.3 was used in this study. Three one-way nested domains were developed (as shown in Fig. 1). The horizontal grids are 100 × 100, 181 × 172, and 250 × 250, with horizontal grid spacing of 9, 3, and 1 km, respectively (Fig. 1). The outer domain covers the upper Midwest of the United States, while the inner domain (domain 3) centers over the Milwaukee metropolitan area and includes a portion of Lake Michigan. The vertical grids contain 55 sigma levels, and the upper boundary is set at 100 hPa. The time steps for the three domains are 30, 10, and 10/3 s, respectively. Initial and boundary conditions for all the simulations are provided by the North American Regional Reanalysis dataset (NARR; see http://www.emc.ncep.noaa.gov/mmb/rreanl/ for more details) with 32-km spatial resolution and 3-h temporal resolution.

We also incorporated high-spatial-resolution (30 m) land use data [National Land Cover Data (NLCD); see http://www.mrlc.gov/nlcd2006.php for more details] into the model. This dataset has four urban categories—open space, low intensity, medium intensity, and high intensity (see Fig. 1b)—and provides better representation of urban surface properties than earlier land cover datasets. In addition, the new land use dataset provides the necessary urban parameters (e.g., albedo, building height, and roughness length) for the single-layer UCM (Chen et al. 2011a; Kusaka et al. 2001). The physics options used in the model are summarized in Table 1. Details and additional references for these options can be found in WRF’s user manual (Skamarock et al. 2008).

Table 1.

WRF physics options.

WRF physics options.
WRF physics options.

c. Experimental setup

Three different runs were implemented in this study. For the control run, we use all the physics options and domain configurations as shown in section 2b and Table 1. For the second run, the UCM was turned off, with the other physics options and domain configurations kept the same as the control run. This simulation will be referred to hereafter as the NOUCM run. In this run, a simple urban treatment (the bulk roughness approach) was used in the default Noah LSM (Chen and Dudhia 2001a,b). The difference between NOUCM and control run lies in the calculation of the surface fluxes over an urban grid. When UCM is not used, the urban grid is treated as 100% pervious (with urban properties) by the Noah LSM; on the contrary (UCM is used, as in control run), the grid is the combination of impervious and vegetated surfaces. The UCM is coupled to Noah through a parameter “urban fraction” (Chen et al. 2011a). The Noah LSM treats the nonurban part (vegetated surfaces of the grid), while the UCM calculates surface fluxes for the impervious surfaces.

A NOURBAN run was also developed for which all urban land in the city was replaced with cultivated crops, one of the dominant land use types in the vicinity of the city. All of the simulations are initiated at 0000 UTC on 21 July 2010 and run until 0000 UTC on 25 July 2010.

3. The 22 July 2010 storm over Milwaukee

a. Rainfall and flood

The 22–23 July 2010 storm produced extensive rainfall over the states of Iowa, Illinois, and Wisconsin, especially along the Wisconsin–Illinois border. This rainfall event caused serious flood damage in Wisconsin, and the total loss for southern Wisconsin was about 27.7 million U.S. dollars, with 24.1 million in Milwaukee County alone (see http://www.crh.noaa.gov/mkx/?n=072210_severe).

The storm total rainfall field developed from bias-corrected Hydro-NEXRAD rainfall fields is shown in Fig. 2a. There is a narrow band of heavy rainfall extending from west to east along the Wisconsin–Illinois border, and the maximum value was more than 270 mm. The accumulated rainfall of this event has almost the same spatial pattern as that of the 21 August 2002 event in the same region (Thorndahl et al. 2013, manuscript submitted to Atmos. Res.). In addition, Thorndahl et al. (2013, manuscript submitted to Atmos. Res.) compared other statistics of this rainfall event (including the contributing areas with different heavy rain rates, accumulated rainfall amount within different time intervals, and spatiotemporal correlation structures) with other severe storm events based on a sample of the 50 largest storm events over Wisconsin during 1996–2010 (see Table 1 of their paper). The spatiotemporal properties of this event match the characteristics of the heavy storm events over Wisconsin. Maddox et al. (1979) found that most flash floods from intense rainfall events in the Great Plains were related to convective storms that moved over the same area repeatedly. These storms were positioned near the midtroposphere, large-scale ridge position. The 22–23 July 2010 storm event in this study reflects the general features of heavy rainfall systems over this region (e.g., Thorndahl et al. 2013, manuscript submitted to Atmos. Res.) and is one of the typical flash flood–producing storms over the Milwaukee region (Yang et al. 2013).

Fig. 2.

Storm total rainfall fields (mm): (a) based on bias-corrected radar rainfall field under the umbrella of the KMKX WSR-88D radar station (see Fig. 1 for location), (b) simulation results from the control, (c) NOUCM, and (d) NOURBAN runs.

Fig. 2.

Storm total rainfall fields (mm): (a) based on bias-corrected radar rainfall field under the umbrella of the KMKX WSR-88D radar station (see Fig. 1 for location), (b) simulation results from the control, (c) NOUCM, and (d) NOURBAN runs.

If we take a close look at the total rainfall field over the urban focus domain, a narrow rain belt can be observed over the northern boundary of the city, extending from west to east (Fig. 3). The maximum storm total accumulation of approximately 90 mm was located over the northwestern boundary of Milwaukee.

Fig. 3.

Storm total rainfall (mm) over the urban focus domain for the 22–23 July 2010 storm based on bias-corrected Hydro-NEXRAD rainfall fields. The short black line near the top left and the triangles represent the tracks of the storm cells passing through the urban area.

Fig. 3.

Storm total rainfall (mm) over the urban focus domain for the 22–23 July 2010 storm based on bias-corrected Hydro-NEXRAD rainfall fields. The short black line near the top left and the triangles represent the tracks of the storm cells passing through the urban area.

The maximum rainfall rate averaged over the 319 km2 region was approximately 60 mm h−1 and occurred between 2100 and 2200 UTC on 22 July. The discharge at the Menomonee River stream gauging station reacted rapidly to the rain impulse and reached its peak within 3 h (figure not shown). The peak unit discharge at this gauging station was approximately 1 m3 s−1 km−2, which is the third largest flood peak in the past decade (2000–2010; see Yang et al. 2013 for more details).

b. Synoptic environment

The synoptic environment at 1200 UTC on 22 July 2010 preceding storm development was characterized by high pressure over the southeastern United States and strong moisture transport around the high-pressure ridge into southern Wisconsin (figure not shown). Isentropic analyses based on NARR reanalysis fields (Fig. 4) illustrate large-scale moisture transport and support for vertical motion over the region of heavy rainfall. In Fig. 4 we show specific humidity, wind, and pressure fields on the 315-K isentropic surface (defined as a surface of constant potential temperature) for the period from 0900 UTC on 22 July to 0300 UTC on 23 July.

Fig. 4.

Synoptic conditions at (a) 0900, (b) 1500, and (c) 2100 UTC 22 July and (d) 0300 UTC 23 July on the 315-K isentropic surface. Wind velocity (full barb = 10 m s−1), specific humidity (shaded, kg kg−1), and pressure (contours, hPa, interval 50 hPa).

Fig. 4.

Synoptic conditions at (a) 0900, (b) 1500, and (c) 2100 UTC 22 July and (d) 0300 UTC 23 July on the 315-K isentropic surface. Wind velocity (full barb = 10 m s−1), specific humidity (shaded, kg kg−1), and pressure (contours, hPa, interval 50 hPa).

The storm period was characterized by continental-scale transport of moisture to the heavy rainfall region of Wisconsin (Fig. 4). Preceding the initiation of heavy rainfall (see 1500 UTC analysis shown in Fig. 4b), the 315-K isentropic surface shows a humidity maximum of 13 g kg−1 in Nebraska, “upstream” of the region that subsequently experienced heavy rain. The region of peak humidity was linked to strong moisture transport into the heavy rainfall region, as shown in the wind field for the 315-K surface. Parcels moving from the region of peak humidity to the heavy rainfall area experienced substantial isentropic lift, reflected in the decrease in pressure from more than 750 hPa in the humidity maximum to less than 650 hPa over Milwaukee (Fig. 4). The gradients of pressure on the 315-K surface increased over the course of the storm period (Fig. 4; in particular, note the 2100 UTC analyses in Fig. 4c). Wind speed in the moist conveyor belt was about 20 m s−1 over Nebraska, Kansas, Iowa, and Wisconsin at 2100 UTC. This strong moist flow was the direct cause for the widespread storms and floods in the upper Midwest of the United States.

The prestorm environment was also characterized by a large region of high convective available potential energy (CAPE). In Fig. 5, we show the CAPE field at 15 UTC from the NARR reanalysis. There are large north–south gradients in the CAPE field with values of 2000 J kg−1 just south of the Wisconsin–Illinois border decreasing to values less than 500 J kg−1 in southern Wisconsin.

Fig. 5.

CAPE (J kg−1) at 1500 UTC 22 July 2010 based on NARR field. Black box denotes the location of Wisconsin–Illinois border area.

Fig. 5.

CAPE (J kg−1) at 1500 UTC 22 July 2010 based on NARR field. Black box denotes the location of Wisconsin–Illinois border area.

c. Storm characteristics

We examined the structure, motion, and evolution of the 22–23 July 2010 storm using the TITAN storm-tracking algorithms. The track of the principal storm element affecting the Milwaukee metropolitan region is shown in Fig. 3, which is based on the centroids locations of the storm volume. The storm initiated in northwestern Wisconsin and propagated to the southeast, passing through the Milwaukee metropolitan region between 2040 and 2200 UTC on 22 July. There was a slowing and an alteration of direction over the northern part of the city, and the location corresponds to the heavy rainfall region as revealed in the shaded background total rainfall field.

Figure 6 shows the temporal evolution of the storm elements based on four storm variables. Maximum reflectivity increased gradually, exceeded 60 dBZ when the storm approached the city and remained above 60 dBZ for the rest of the storm life cycle. The height of maximum reflectivity was increasing, and the average height is 6 km. It varied with the height of the echo top, which was also increasing, and the average top height was 12 km with a maximum of 18 km, which occurred around 2200 UTC on 22 July (figure not shown). The increasing echo top height is an indication of strong convection (see Smith et al. 1996 for similar storm-tracking analyses). The storm speed exhibited small variation and large values, with an average speed of approximately 70 km h−1. The dominant storm motion direction was southeasterly during this period. The storm system experienced increased frequency of splits and mergers, as reflected in the time series of the number of simple tracks (simple storms that have no splits or mergers), as it passed the urban area. Our results are consistent with the study of Niyogi et al. (2011), who examined 91 summertime thunderstorm cases over Indianapolis. They found 60% of the storms change structure as they pass the urban area and storms tend to split closer to upwind urban region and merge downwind.

Fig. 6.

Time series of (a) max reflectivity (dBZ), (b) height of max reflectivity (km), (c) number of simple tracks (dimensionless), and (d) storm speed (km h−1) for the storm. Shaded area highlights the period when the storm is passing the Milwaukee metropolitan area from 2040 to 2200 UTC 22 July. Black line represents the tracking results based on the observed radar reflectivity, while the dots are from the generated reflectivity field of the WRF control run. The interval for the time coordinate is 1 h.

Fig. 6.

Time series of (a) max reflectivity (dBZ), (b) height of max reflectivity (km), (c) number of simple tracks (dimensionless), and (d) storm speed (km h−1) for the storm. Shaded area highlights the period when the storm is passing the Milwaukee metropolitan area from 2040 to 2200 UTC 22 July. Black line represents the tracking results based on the observed radar reflectivity, while the dots are from the generated reflectivity field of the WRF control run. The interval for the time coordinate is 1 h.

4. Model validation

The control simulation (Fig. 2b) captured key elements of the extreme rainfall distribution for the 22–23 July 2010 storm. In Fig. 2b, we show the accumulated rainfall distribution over domain 3 for the control run. The narrow band of extreme rainfall along the Wisconsin–Illinois border is well captured, as compared to the bias-corrected radar rainfall field (Fig. 2a). The maximum value for the simulation, 300 mm, is slightly larger than the maximum accumulation in the radar rainfall analysis. The mean bias, root-mean-square error (RMSE), and spatial correlation between the radar and WRF simulation rainfall fields are −8.1, 36.6, and 0.46 mm, respectively. The spatial correlation is statistically significant at the 95% confidence level. In addition, the model simulation captured large-scale features of the storm environment, especially the features controlling extreme moisture transport from the Gulf of Mexico to the heavy rainfall region (figure not shown).

We also provide comparisons of storm structure between observed and simulated reflectivity fields for three time periods (Fig. 7). The model captures the timing and organization of the storm well, with only a small time offset between the model and observed fields. The magnitudes of the radar reflectivity field are underestimated by the model, which will produce less precipitation, as will be discussed below.

Fig. 7.

Storm evolution represented by the reflectivity field (dBZ, at the height of 4.6 km) from (top) KMKX radar observations and (bottom) WRF simulations. Three snapshots during the evolution: (left) 1900, (middle) 2000, and (right) 2200 UTC 22 July.

Fig. 7.

Storm evolution represented by the reflectivity field (dBZ, at the height of 4.6 km) from (top) KMKX radar observations and (bottom) WRF simulations. Three snapshots during the evolution: (left) 1900, (middle) 2000, and (right) 2200 UTC 22 July.

Three indices are chosen to validate the simulation results of surface air temperature and specific humidity, for which we have continuous observations from 26 ASOS stations covering the entire storm period (Table 2). The time series are separated into two subseries: daytime period [0800–2000 local time (LT)] and nighttime period (2000–0800 LT of the following day).

Table 2.

Verification statistics for the WRF simulations with the observations by ASOS stations. These statistics are averaged for the whole simulation period (from 0000 UTC 21 July to 0000 UTC 25 July) and averaged for all the stations within the domain.

Verification statistics for the WRF simulations with the observations by ASOS stations. These statistics are averaged for the whole simulation period (from 0000 UTC 21 July to 0000 UTC 25 July) and averaged for all the stations within the domain.
Verification statistics for the WRF simulations with the observations by ASOS stations. These statistics are averaged for the whole simulation period (from 0000 UTC 21 July to 0000 UTC 25 July) and averaged for all the stations within the domain.

In general, nighttime simulation results are somewhat better than daytime results for the two surface variables. Simulated specific humidity time series are more consistent with the observations during the daytime than nighttime, while the opposite is true for air temperature. Both air temperature and specific humidity are generally overestimated in the model, which likely plays a role in the overstimulation of rainfall at a larger scale (Fig. 2b).

The evolution of the vertical wind profile in the lower atmosphere is illustrated in Fig. 8a through hourly VAD wind profiles derived from KMKX Doppler velocity observations. The VAD wind is interpolated onto the same vertical levels as used in the model simulations (Fig. 8a). The wind is predominantly from northwest before the onset of the storm, but veers to southerly at lower levels (below 1500 m) and southwesterly at higher levels (above 1500 m) on 22 July when the storm approached the region. The wind speed at levels above 1500 m exceeds 25 m s−1 for portions of the storm and is about 15 m s−1 on average for the entire period. Model simulations exhibit similar evolution of wind profile characteristics (Fig. 8b). Although the wind speed is underestimated in the levels above 1500 m on 22 July and the direction is more southerly in the model rather than southwesterly at the end of the day, the pattern matches with the VAD wind profile quite well in lower levels (below 1500 m) during the entire period.

Fig. 8.

Time series of horizontal wind profiles at the KMKX WSR-88D station: (a) VAD and (b) WRF profiles. Vertical arrows represent southerly wind with the other directions in a clockwise order. Wind speed (m s−1) is represented by the various colors.

Fig. 8.

Time series of horizontal wind profiles at the KMKX WSR-88D station: (a) VAD and (b) WRF profiles. Vertical arrows represent southerly wind with the other directions in a clockwise order. Wind speed (m s−1) is represented by the various colors.

Time series of simulated hourly rain rate averaged over the urban area (Fig. 9) are generally in good agreement with observations. The timing of peak rain rate agrees well with the bias-corrected radar observations. The magnitude of peak rainfall rate is somewhat underestimated in the model simulation and the magnitude of the underestimation is dependent on the details of the land surface model representation (see section 5 for additional details).

Fig. 9.

Time series of hourly rain rate (mm h−1) averaged over urban region (denoted by the black box over the city in Fig. 1b). The interval for the time coordinate is 6 h.

Fig. 9.

Time series of hourly rain rate (mm h−1) averaged over urban region (denoted by the black box over the city in Fig. 1b). The interval for the time coordinate is 6 h.

In addition to direct comparisons of model and observed rainfall, we examined evolution of the simulated storm based on storm-tracking analyses using the TITAN algorithms. In Fig. 6, we compare simulated time series of maximum reflectivity, height of maximum reflectivity, storm speed, and number of storm elements (simple tracks) with observed time series based on 3D volume scan reflectivity observations. The analyses in Fig. 6 are presented for the dominant complex track (storms that include all simple tracks within the entire lifetime of the storm) passing over the Milwaukee region. The simulated storm is somewhat more intense than the observed storm, especially as the storm approaches the urban area. The simulated storm has peak reflectivity values between 65 and 70 dBZ; the observed storm has sharply increasing peak reflectivity values with peak values of approximately 65 dBZ over the urban area. Although the maximum reflectivity simulated by the model is higher than the radar-based results, the height of maximum reflectivity is well simulated (Fig. 6b). Height of 45-dBZ echo top (figure not shown) is approximately 13 km in both observations and simulation. Both the number of storm elements (i.e., simple tracks in TITAN) and the storm speed are well simulated by the model, with consistent trends and identical values (Figs. 6c,d). The simulated precipitation flux in the model simulation (figure not shown) is larger than the observations because of storm area being larger in the simulation than in observations. Instead of concentrating the rain zone over the city, the model spreads rain to a larger area in and around the city.

In general, the storm structure as well as its evolution is well represented by the model simulations. Validation analyses suggest that numerical experiments in which we perturb surface parameterizations can provide useful insights into the urban modification of rainfall.

5. Model simulation results

Analyses in this section are based on three simulations for the 22–23 July 2010 storm. In addition to the control simulation examined in the previous section, we present results from two simulations representing modification to the land surface parameterization (see discussion of experimental setup in section 2). The NOURBAN simulation represents pre-urban conditions in the Milwaukee region. The NOUCM simulation is used to illustrate the contrast between two representations of land surface processes in urban regions.

Time series of latent heat and sensible heat flux (Figs. 10a,b), spatially averaged over the urban area (denoted as the black box over the city in Fig. 1b), highlight the most striking contrasts among the three simulations. Peak diurnal values of latent heat flux for the NOURBAN simulation are more than twice as large as the corresponding values for the NOUCM simulation. The situation reverses for sensible heat flux. Latent heat and sensible heat flux time series for the control simulations, which include the UCM, are between the extremes of the NOUCM and NOURBAN simulations. Timing of the minima and maxima in the diurnal cycle of latent heat and sensible heat fluxes are the same for the three runs. Unlike the peak magnitudes during daytime, there are not substantial differences for the nighttime fluxes. The differences of the heat fluxes (sensible and latent heat flux) are similar to previous studies (e.g., Shem and Shepherd 2009; Zhang et al. 2009).

Fig. 10.

Time series of the control, NOURBAN, and NOUCM runs for (a) latent heat flux (W m−2), (b) sensible heat flux (W m−2), (c) air temperature at 2 m (K),(d) specific humidity at 2 m (kg kg−1), (e) skin temperature (K), and (f) difference of skin temperature (K) between urban and rural region. All these variables except (f) are spatially averaged over the urban region (denoted as the black box over the city in Fig. 1b). The interval for the time coordinate is 12 h. Only 0000 UTC of each day is labeled on the coordinate.

Fig. 10.

Time series of the control, NOURBAN, and NOUCM runs for (a) latent heat flux (W m−2), (b) sensible heat flux (W m−2), (c) air temperature at 2 m (K),(d) specific humidity at 2 m (kg kg−1), (e) skin temperature (K), and (f) difference of skin temperature (K) between urban and rural region. All these variables except (f) are spatially averaged over the urban region (denoted as the black box over the city in Fig. 1b). The interval for the time coordinate is 12 h. Only 0000 UTC of each day is labeled on the coordinate.

For the NOURBAN scenario, the urban land in the city was replaced with cultivated crops, which produce greater soil moisture availability and evapotranspiration (Fig. 10a), while for the two urban runs (control and NOUCM), albedo is smaller than for the NOURBAN run, which results in larger net incoming radiation. The differences in latent and sensible heat flux between the control and NOUCM run can be attributed to the different urban representations in the model. In the NOUCM run, the bulk roughness approach is used for each urban grid, which results in overestimation of the urban heat flux compared to the single-layer UCM (Fig. 10b).

For surface air temperature and specific humidity, the differences between the control run and NOUCM run are not as large as the differences between the two urban runs and NOURBAN run (Figs. 10c,d). This implies that these variables are more sensitive to surface properties than to details of the urban representations in the model.

Time series of skin temperature averaged over the urban area for three runs (Fig. 10e) illustrate the magnitude of the UHI effect. The two urban runs (control and NOUCM) produce skin temperature values that are 4–5 K higher than skin temperature values for the NOURBAN run during the middle of the day. There is not, however, a big difference between the control and NOUCM run. We examine the UHI effect as the difference of spatially averaged skin temperature over urban and rural areas (denoted as two black boxes in Fig. 1b). With this definition, the NOUCM UHI effect is a little larger than the UHI effect for control run (Fig. 10f). The difference of skin temperature for the two urban runs is about 4–6 K during the day and 0–2 K during the night. As expected, the NOURBAN run yields skin temperatures over the modified land that are, on average, equal to the rural temperatures. We used skin temperature to be the proxy of UHI rather than air temperature, simply because it is a more direct variable to depict the radiation properties of land surface (Jin 2012; Jin and Dickinson 2010). The UHI intensity in our study is slightly higher than the results of previous studies (e.g., Bornstein and Lin 2000; Niyogi et al. 2011).

In Fig. 11 we show cross sections of vertical wind profiles along line A–B (denoted in Fig. 1b) at 2100 UTC on 22 July, which is right before the onset of the storm over the city. Strong updrafts were located over the western fringe of the city for the two urban runs, with peak vertical wind velocities of more than 3.5 m s−1 in some regions (Figs. 11a,b). For the NOUCM run, this updraft was paired with a strong downdraft to the east of the cross section, while a weaker downdraft zone was located to the west of the inland urban boundary in the control run. Both urban runs included strong updrafts over the urban area, in contrast to the NOURBAN run (Fig. 11c). There are also strong contrasts between the surface pressure fields between the two urban runs and the NOURBAN simulation (Fig. 12). Compared to the two urban runs, the NOURBAN run exhibits a higher sea level pressure field over the urban area (Fig. 12c).

Fig. 11.

Cross sections of the vertical wind profile along line A–B (see Fig. 1b) at 2100 UTC 22 July for three runs: (a) control, (b) NOUCM, and (c) NOURBAN. Shading represents the magnitude of the vertical wind (m s−1) and the black horizontal bar is the location of the urban area.

Fig. 11.

Cross sections of the vertical wind profile along line A–B (see Fig. 1b) at 2100 UTC 22 July for three runs: (a) control, (b) NOUCM, and (c) NOURBAN. Shading represents the magnitude of the vertical wind (m s−1) and the black horizontal bar is the location of the urban area.

Fig. 12.

Sea level pressure on at 2100 UTC 22 July over the urban focus region for three runs: (a) control, (b) NOUCM, and (c) NOURBAN. The urban boundary is denoted by the thin irregular white line. Shading represents the pressure (hPa).

Fig. 12.

Sea level pressure on at 2100 UTC 22 July over the urban focus region for three runs: (a) control, (b) NOUCM, and (c) NOURBAN. The urban boundary is denoted by the thin irregular white line. Shading represents the pressure (hPa).

Our findings are consistent with the study of Shepherd et al. (2010a). They found similar vertical motion and pressure fields. The pressure field perturbation is linked to surface convergence and vertical motion fields over the urban area, as discussed below.

Because Milwaukee is located along the western boundary of Lake Michigan, the lake-breeze dynamics play a role in the formation of convergence zones. In Fig. 13, we show cross sections of the vertical profile of the lake breeze along line A–B (see Fig. 1b) at 2100 UTC on 22 July (prior to the period of peak rainfall, as in Figs. 11, 12). Although the wind field at a larger scale is southwesterly at this time (Fig. 8), the lake-breeze intrusion can still be seen along this cross section.

Fig. 13.

Cross sections of vertical lake-breeze profile along line A–B (see Fig. 1b) at 2100 UTC 22 July for three runs: (a) control, (b) NOUCM, and (c) NOURBAN. Arrows are the synthesis of the vertical and east–west wind components. The color of the arrows represents the wind speed (m s−1). Color bars at the bottom of the panels represent different regions: rural (yellow), urban (red), and lake (blue). Black diagonal lines highlight the transition fronts of the wind direction.

Fig. 13.

Cross sections of vertical lake-breeze profile along line A–B (see Fig. 1b) at 2100 UTC 22 July for three runs: (a) control, (b) NOUCM, and (c) NOURBAN. Arrows are the synthesis of the vertical and east–west wind components. The color of the arrows represents the wind speed (m s−1). Color bars at the bottom of the panels represent different regions: rural (yellow), urban (red), and lake (blue). Black diagonal lines highlight the transition fronts of the wind direction.

The penetration of the lake breeze reached the inner land boundary of the city for the two urban runs (Figs. 13a,b). For the control run, the penetration forms a wedge-shaped region in the cross section, while the region is much wider in the NOUCM run, attaining a height of more than 1200 m. This is in agreement with the higher sensible heat flux simulated in the NOUCM run, which should generate a stronger lake breeze. For the NOURBAN simulation, the lake breeze does not penetrate inland over the urban area of Milwaukee. Another noticeable feature of the cross-sectional wind profile is that the westerly wind in the NOURBAN run is stronger than for the two urban runs (Fig. 13). This may be caused by urban roughness effects (Cheng and Chan 2012; Chen et al. 2011b) or the counteraction of the strong lake-breeze intrusion with the background wind from inland.

Lake-breeze intrusion enhances the low-level convergence zone over the urban area and provides favorable conditions for convection. Figure 14 shows the low-level (960 hPa) water flux field 1 h before the occurrence of peak rain rate. There is a convergence zone over the northwestern boundary of the city, extending to the northeast in the control run. For the NOUCM run, the convergence zone is located within the city boundary. In contrast to the urban runs, the convergence zone is concentrated northeast of the city for the NOURBAN simulation.

Fig. 14.

Water flux field at 960-hPa geopotential height and rain rate at 2200 UTC 22 July over the urban focus region for three runs: (a) control, (b) NOUCM, and (c) NOURBAN. Shading represents rain rate (mm h−1) and the white arrows, the water flux (m s−1 kg kg−1). The urban boundary is denoted by the thin irregular black lines. The dashed boxes denote the water flux convergence zones.

Fig. 14.

Water flux field at 960-hPa geopotential height and rain rate at 2200 UTC 22 July over the urban focus region for three runs: (a) control, (b) NOUCM, and (c) NOURBAN. Shading represents rain rate (mm h−1) and the white arrows, the water flux (m s−1 kg kg−1). The urban boundary is denoted by the thin irregular black lines. The dashed boxes denote the water flux convergence zones.

Water flux convergence zone is coupled with peak rain rate zone, as revealed in Fig. 14. Maximum hourly rain rate was more than 80 mm h −1 for the NOUCM run and about 60 mm h −1 for the control run over the urban area. The storm center was shifted away from the city in NOUBAN run, although peak rain rates were comparable to the urban runs.

Thermal properties of urban surfaces and the resulting UHI effect are important elements of the Milwaukee rainfall anomaly (see Yang et al. 2013 for analyses of the Milwaukee rainfall anomaly). Changes of surface properties and urban representations in the model produce a large perturbation in skin temperature, which is directly related to the surface heat flux. Elevated surface temperatures and strong sensible heat flux promote development of surface pressure perturbations that interact with the lake-breeze circulation of Lake Michigan. The feedback of updrafts, which reduced surface pressure, enhances the lake-breeze intrusion, which is initiated by the lake–land thermal gradients. The confluence of lake breeze and inland wind facilitates the formation of the low-level convergence zone, which supports the development of deep convection. Both urban runs produced distinct convergence zones and high peak rain rate over the urban area. In contrast, the storm center was shifted away from the city in NOURBAN run.

Model simulations suggest that urbanization does not significantly alter the dynamics of precipitating cloud systems, as reflected in the ensemble of vertical profiles of reflectivity over the urban area. In Fig. 15a, we show the CFADs (see Yuter and Houze 1995) of simulated reflectivity fields from the control simulation over the Milwaukee metropolitan region (denoted by the black box over the city in Fig. 1b) during the time period from 1800 to 2300 UTC on 22 July. The CFAD for the NOURBAN run (Fig. 15b) exhibits slightly broader reflectivity distribution in the middle of the troposphere, but otherwise, there are not sharp contrasts between the control and NOURBAN simulations. We include the CFAD for the larger land domain (the control run) in Fig. 15c for reference.

Fig. 15.

CFADs (color shading represents frequency in percent). Reflectivity from (a) control run, (b) NOURBAN run over urban region (denoted as the black box over the city in Fig. 1b), and (c) control run but over a larger land domain.

Fig. 15.

CFADs (color shading represents frequency in percent). Reflectivity from (a) control run, (b) NOURBAN run over urban region (denoted as the black box over the city in Fig. 1b), and (c) control run but over a larger land domain.

The distributions of total rainfall do not exhibit noticeable differences at a larger scale. As illustrated in Fig. 2, the accumulated rainfall distributions have the same pattern for all three runs, with maximum value (about 300 mm) located along the Wisconsin–Illinois border and much wider rain belts compared to the radar rainfall field. The total rainfall distribution at a larger scale is not sensitive to the local-scale differences of surface properties and urban representations.

6. Summary and conclusions

In this paper, we examine a storm that produced heavy rainfall (storm total accumulations exceeding 80 mm) and flash flooding in Milwaukee. The area of heavy rainfall in Milwaukee was embedded in a much larger region of extreme rainfall over the upper Midwest of the United States, with peak accumulations exceeding 300 mm. Observational and numerical modeling experiments are used to examine urban impacts on rainfall for a storm system with strong large-scale forcing and extreme rainfall over a large region. Major findings are summarized below.

  1. The control simulation with WRF captures key elements of the rainfall distribution for the 22–23 July 2010 storm, including the elongated region of peak rainfall (exceeding 300 mm) along the Wisconsin–Illinois border, rainfall accumulations over Milwaukee, and timing of peak rainfall over Milwaukee. The structure and evolution of storm characteristics are also captured in the control simulation, as represented by storm-tracking analyses of observed and simulated 3D reflectivity fields.

  2. Analyses of NARR reanalysis fields and WRF simulations show that continental-scale moisture transport was a key element of the heavy rainfall distribution over the upper Midwest for the 22–23 July 2010 storm system. Isentropic uplift along the plume of moisture transport provided support for development of deep convection over the heavy rainfall region.

  3. Even under conditions of strong large-scale forcing described in item 2 above, urbanization can alter the local distribution of heavy rainfall. Two sensitivity experiments (NOUCM and NOURBAN) were carried out to examine the role of urbanization for heavy rainfall distribution under large-scale forcing. The two urban simulation cases produce a strong urban heat island (UHI) perturbation on temperature and pressure fields in the atmospheric boundary layer. These features enhance the intrusion of lake breeze, which facilitates the formation of convergence zone and provides favorable conditions for deep convection over the urban region. The surface skin temperature over the urban area is 5 K higher than the rural area to the west of Milwaukee.

  4. Analyses of vertical structure of reflectivity for the simulated storms using contoured frequency by altitude displays (CFADs) suggest that urbanization does not alter the dynamics of thunderstorm cells that pass over the urban region. These results point to the primary role of atmospheric boundary layer processes in determining urban impacts on heavy rainfall.

  5. The large-scale rainfall pattern is not sensitive to local-scale urban forcing as represented by the Noah LSM and urban canopy model in the NOUCM and control simulations. The storm total rainfall distribution for the three simulations show negligible differences in the region of heavy rainfall. However, our results also revealed the importance of proper urban representations in capturing the dynamics of sensible and latent heat flux, as supported by previous studies (Chen et al. 2011a; Carter et al. 2012). Since only one case is examined in this study, caution should be exercised when generalizing the results. Future studies should examine a broader sample of rainfall events with different synoptic conditions and should further test the impact of urban representations on rainfall distribution. In addition, aerosol–cloud interactions should be examined for the Milwaukee region to assess the effects of urban aerosol on deep convection and heavy rainfall (as in Ntelekos et al. 2009).

Acknowledgments

This study was supported by the National Science Foundation of China (NSFC 51179084, 51222901), the special foundation of Ministry of Water Resources, China (Project 201001004), the U.S. National Science Foundation (Grant CBET-1058027), the NOAA Cooperative Institute for Climate Science, and the Willis Research Network. The authors would like to acknowledge that the numerical experiments in this paper were performed at the TIGRESS high-performance computer center at Princeton University, which is jointly supported by the Princeton Institute for Computational Science and Engineering and the Princeton University Office of Information Technology.

REFERENCES

REFERENCES
Bornstein
,
R.
, and
Q.
Lin
,
2000
:
Urban heat islands and summertime convective thunderstorms in Atlanta: Three case studies
.
Atmos. Environ.
,
34
,
507
516
,
doi:10.1016/S1352-2310(99)00374-X.
Carter
,
M.
,
J. M.
Shepherd
,
S.
Burian
, and
I.
Jeyachandran
,
2012
:
Integration of lidar data into a coupled mesoscale–land surface model: A theoretical assessment of sensitivity of urban–coastal mesoscale circulations to urban canopy parameters
.
J. Atmos. Oceanic Technol.
,
29
,
328
346
,
doi:10.1175/2011JTECHA1524.1.
Chen
,
F.
, and
J.
Dudhia
,
2001a
:
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
,
doi:10.1175/1520-0493(2001)129<0569:CAALSH>2.0.CO;2.
Chen
,
F.
, and
J.
Dudhia
,
2001b
:
Coupling an advanced land surface–hydrology model with the Penn State–NCAR MM5 modeling system. Part II: Preliminary model validation
.
Mon. Wea. Rev.
,
129
,
587
604
,
doi:10.1175/1520-0493(2001)129<0587:CAALSH>2.0.CO;2.
Chen
,
F.
, and
Coauthors
,
2011a
:
The integrated WRF/urban modelling system: Development, evaluation, and applications to urban environmental problems
.
Int. J. Climatol.
,
31
,
273
288
,
doi:10.1002/joc.2158.
Chen
,
F.
,
S.
Miao
,
M.
Tewari
,
J.-W.
Bao
, and
H.
Kusaka
,
2011b
:
A numerical study of interactions between surface forcing and sea breeze circulations and their effects on stagnation in the greater Houston area
.
J. Geophys. Res.
,
116
,
D12105
,
doi:10.1029/2010JD015533.
Cheng
,
C. K. M.
, and
J. C. L.
Chan
,
2012
:
Impacts of land use changes and synoptic forcing on the seasonal climate over the Pearl River delta of China
.
Atmos. Environ.
,
60
,
25
36
,
doi:10.1016/j.atmosenv.2012.06.019.
Collier
,
C. G.
,
2006
:
The impact of urban areas on weather
.
Quart. J. Roy. Meteor. Soc.
,
132
,
1
25
,
doi:10.1256/qj.05.199.
Dixon
,
M.
, and
G.
Wiener
,
1993
:
TITAN: Thunderstorm Identification, Tracking, Analysis, and Nowcasting—A radar-based methodology
.
J. Atmos. Oceanic Technol.
,
10
,
785
797
,
doi:10.1175/1520-0426(1993)010<0785:TTITAA>2.0.CO;2.
Dixon
,
P. G.
, and
T. L.
Mote
,
2003
:
Patterns and causes of Atlanta’s urban heat island-initiated precipitation
.
J. Appl. Meteor.
,
42
,
1273
1284
,
doi:10.1175/1520-0450(2003)042<1273:PACOAU>2.0.CO;2.
Freitas
,
E. D.
,
C. M.
Rozoff
,
W. R.
Cotton
, and
P. L. S.
Dias
,
2006
:
Interactions of an urban heat island and sea-breeze circulations during winter over the metropolitan area of São Paulo, Brazil
.
Bound.-Layer Meteor.
,
122
,
43
65
,
doi:10.1007/s10546-006-9091-3.
Jin
,
M.
,
2012
: Developing an index to measure urban heat island effect using satellite land skin temperature and land cover observations. J. Climate,25, 6193–6201, doi:10.1175/JCLI-D-11-00509.1.
Jin
,
M.
, and
R. E.
Dickinson
,
2010
:
Land surface skin temperature climatology: Benefitting from the strengths of satellite observations
.
Environ. Res. Lett.
,
5
,
044004
,
doi:10.1088/1748-9326/5/4/044004.
Jin
,
M.
,
J. M.
Shepherd
, and
W.
Zheng
,
2010
:
Urban surface temperature reduction via the urban aerosol direct effect: A remote sensing and WRF model sensitivity study
.
Adv. Meteor.
,
2010
, 681587, doi:10.1155/2010/681587.
Keeler
,
J. M.
, and
D. A. R.
Kristovich
,
2012
:
Observations of urban heat island influence on lake-breeze frontal movement
.
J. Appl. Meteor. Climatol.
,
51
,
702
710
,
doi:10.1175/JAMC-D-11-0166.1.
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
,
doi:10.1023/A:1019207923078.
Lin
,
C.-Y.
,
F.
Chen
,
J. C.
Huang
,
W. C.
Chen
,
Y. A.
Liou
,
W. N.
Chen
, and
S.-C.
Liu
,
2008
:
Urban heat island effect and its impact on boundary layer development and land–sea circulation over northern Taiwan
.
Atmos. Environ.
,
42
,
5635
5649
,
doi:10.1016/j.atmosenv.2008.03.015.
Lo
,
J. C. F.
,
A. K. H.
Lau
,
F.
Chen
,
J. C. H.
Fung
, and
K. K. M.
Leung
,
2007
:
Urban modification in a mesoscale model and the effects on the local circulation in the Pearl River delta region
.
J. Appl. Meteor. Climatol.
,
46
,
457
476
,
doi:10.1175/JAM2477.1.
Lowry
,
W. P.
,
1998
:
Urban effects on precipitation amount
.
Prog. Phys. Geogr.
,
22
,
477
520
.
Maddox
,
R. A.
,
C. F.
Chappell
, and
L. R.
Hoxit
,
1979
:
Synoptic and meso-α scale aspects of flash flood events
.
Bull. Amer. Meteor. Soc.
,
60
,
115
123
,
doi:10.1175/1520-0477-60.2.115.
Matejka
,
T.
, and
R. C.
Srivastava
,
1991
:
An improved version of the extended velocity-azimuth display analysis of single-Doppler radar data
.
J. Atmos. Oceanic Technol.
,
8
,
453
466
,
doi:10.1175/1520-0426(1991)008<0453:AIVOTE>2.0.CO;2.
Miao
,
S.
,
F.
Chen
,
M. A.
LeMone
,
M.
Tewari
,
Q.
Li
, and
Y.
Wang
,
2009
:
An observational and modeling study of characteristics of urban heat island and boundary layer structures in Beijing
.
J. Appl. Meteor. Climatol.
,
48
,
484
501
,
doi:10.1175/2008JAMC1909.1.
Miao
,
S.
,
F.
Chen
,
Q.
Li
, and
S.
Fan
,
2011
:
Impacts of urban processes and urbanization on summer precipitation: A case study of heavy rainfall in Beijing on 1 August 2006
.
J. Appl. Meteor. Climatol.
,
50
,
806
825
,
doi:10.1175/2010JAMC2513.1.
Niyogi
,
D.
,
P.
Pyle
,
M.
Lei
,
S. P.
Arya
,
C. M.
Kishtawal
,
M.
Shepherd
,
F.
Chen
, and
B.
Wolfe
,
2011
:
Urban modification of thunderstorms: An observational storm climatology and model case study for the Indianapolis urban region
.
J. Appl. Meteor. Climatol.
,
50
,
1129
1144
,
doi:10.1175/2010JAMC1836.1.
Ntelekos
,
A. A.
,
J. A.
Smith
,
M. L.
Baeck
,
W. F.
Krajewski
,
A. J.
Miller
, and
R.
Goska
,
2008
:
Extreme hydrometeorological events and the urban environment: Dissecting the 7 July 2004 thunderstorm over the Baltimore MD metropolitan region
.
Water Resour. Res.
,
44
, W08446,
doi:10.1029/2007WR006346.
Ntelekos
,
A. A.
,
J. A.
Smith
,
L.
Donner
,
J. D.
Fast
,
W. I.
Gustafson
Jr.
,
E. G.
Chapman
, and
W. F.
Krajewski
,
2009
:
The effects of aerosols on intense convective precipitation in the northeastern United States
.
Quart. J. Roy. Meteor. Soc.
,
135
,
1367
1391
,
doi:10.1002/qj.476.
Ohashi
,
Y.
, and
H.
Kida
,
2002
:
Local circulations developed in the vicinity of both coastal and inland urban areas: A numerical study with a mesoscale atmospheric model
.
J. Appl. Meteor.
,
41
,
30
45
,
doi:10.1175/1520-0450(2002)041<0030:LCDITV>2.0.CO;2.
Oke
,
T. R.
,
1982
:
The energetic basis of the urban heat island
.
Quart. J. Roy. Meteor. Soc.
,
108
,
1
24
, doi:10.1002/qj.49710845502.
Rosenfeld
,
D.
,
2000
:
Suppression of rain and snow by urban and industrial air pollution
.
Science
,
287
,
1793
1796
,
doi:10.1126/science.287.5459.1793.
Sarkar
,
A.
,
R. S.
Saraswat
, and
A.
Chandrasekar
,
1998
: Numerical study of the effects of urban heat island on the characteristic features of the sea breeze circulation. Proc. Indian Acad. Sci., Earth Planet. Sci.,107, 127–137, doi:10.1007/BF02840463.
Shem
,
W.
, and
M.
Shepherd
,
2009
:
On the impact of urbanization on summertime thunderstorms in Atlanta: Two numerical model case studies
.
Atmos. Res.
,
92
,
172
189
,
doi:10.1016/j.atmosres.2008.09.013.
Shepherd
,
J. M.
,
2005
:
A review of the current investigations of urban-induced rainfall and recommendations for the future
.
Earth Interactions
,
9
,
doi:10.1175/EI156.1.
Shepherd
,
J. M.
,
M.
Carter
,
M.
Manyin
,
D.
Messen
, and
S.
Burian
,
2010a
:
The impact of urbanization on current and future coastal precipitation: A case study for Houston
.
Environ. Plann. B: Plann. Des.
,
37
,
284
304
,
doi:10.1068/b34102t.
Shepherd
,
J. M.
,
J. A.
Stallins
,
M. L.
Jin
, and
T. L.
Mote
,
2010b
: Urbanization: Impacts on clouds, precipitation, and lightning. Urban Ecosystem Ecology, Agron. Monogr., No. 55, Amer. Soc. of Agronomy, 1–27, doi:10.2134/agronmonogr55.c1.
Shepherd
,
M.
,
T.
Mote
,
J.
Dowd
,
M.
Roden
,
P.
Knox
,
S. C.
McCutcheon
, and
S. E.
Nelson
,
2011
:
An overview of synoptic and mesoscale factors contributing to the disastrous Atlanta flood of 2009
.
Bull. Amer. Meteor. Soc.
,
92
,
861
870
,
doi:10.1175/2010BAMS3003.1.
Skamarock
,
W. C.
, and
Coauthors
,
2008
: A description of the Advanced Research WRF version 3. NCAR Tech. Note NCAR/TN-475+STR, 113 pp. [Available online at http://www.mmm.ucar.edu/wrf/users/docs/arw_v3.pdf.]
Smith
,
J. A.
,
M. L.
Baeck
,
M.
Steiner
, and
A. J.
Miller
,
1996
:
Catastrophic rainfall from an upslope thunderstorm in the central Appalachians: The Rapidan storm of June 27, 1995
.
Water Resour. Res.
,
32
, 3099–3113,
doi:10.1029/96WR02107.
Smith
,
J. A.
,
M. L.
Baeck
,
G.
Villarini
,
C.
Welty
,
A. J.
Miller
, and
W. F.
Krajewski
,
2012
:
Analyses of a long-term, high-resolution radar rainfall data set for the Baltimore metropolitan region
.
Water Resour. Res.
,
48
, W04504,
doi:10.1029/2011WR010641.
Wright
,
D. B.
,
J. A.
Smith
,
G.
Villarini
, and
M. L.
Baeck
,
2012
:
Hydroclimatology of flash flooding in Atlanta
.
Water Resour. Res.
,
48
, W04524,
doi:10.1029/2011WR011371.
Yang
,
L.
,
J. A.
Smith
,
D. B.
Wright
,
M. L.
Baeck
,
G.
Villarini
,
F.
Tian
, and
H.
Hu
,
2013
:
Urbanization and climate change: An examination of nonstationarities in urban flooding
.
J. Hydrometeor.
,
14
,
1791–1809, doi:10.1175/JHM-D-12-095.1.
Yuter
,
S. E.
, and
R. A.
Houze
Jr.
,
1995
:
Three-dimensional kinematic and microphysical evolution of Florida cumulonimbus. Part II: Frequency distributions of vertical velocity, reflectivity, and differential reflectivity
.
Mon. Wea. Rev.
,
123
,
1941
1963
,
doi:10.1175/1520-0493(1995)123<1941:TDKAME>2.0.CO;2.
Zhang
,
C. L.
,
F.
Chen
,
S. G.
Miao
,
Q. C.
Li
,
X. A.
Xia
, and
C. Y.
Xuan
,
2009
:
Impacts of urban expansion and future green planting on summer precipitation in the Beijing metropolitan area
.
J. Geophys. Res.
,
114
,
D02116
,
doi:10.1029/2008JD010328.
Zhang
,
D. L.
,
Y. X.
Shou
,
R. R.
Dickerson
, and
F.
Chen
,
2011
:
Impact of upstream urbanization on the urban heat island effects along the Washington–Baltimore corridor
.
J. Appl. Meteor. Climatol.
,
50
,
2012
2029
,
doi:10.1175/JAMC-D-10-05008.1.