Abstract

The 2009 Atlanta flood was a historic event that resulted in catastrophic damage throughout the metropolitan area. The flood was the product of several hydrometeorological processes, including moist antecedent conditions, ample atmospheric moisture, and mesoscale training. Additionally, previous studies hypothesized that the urban environment of Atlanta altered the location and/or overall quantities of precipitation and runoff that ultimately produced the flood. This hypothesis was quantitatively evaluated by conducting a modeling case study that utilized the Weather Research and Forecasting Model. Two model runs were performed: 1) an urban run designed to accurately depict the flood event and 2) a nonurban simulation where the urban footprint of Atlanta was replaced with natural vegetation. Comparing the output from the two simulations revealed that interactions with the urban environment enhanced the precipitation and runoff associated with the flood. Specifically, the nonurban model underestimated the cumulative precipitation by approximately 100 mm in the area downwind of Atlanta where urban rainfall enhancement was hypothesized. This notable difference was due to the increased surface convergence observed in the urban simulation, which was likely attributable to the enhanced surface roughness and thermal properties of the urban environment. The findings expand upon previous research focused on urban rainfall effects since they demonstrate that urban interactions can influence mesoscale hydrometeorological characteristics during events with prominent synoptic-scale forcing. Finally, from an urban planning perspective, the results highlight a potential two-pronged vulnerability of urban environments to extreme rainfall, as they may enhance both the initial precipitation and subsequent runoff.

1. Introduction

Because of a complex combination of both social and physical factors, urban areas are particularly vulnerable to flooding. The pressures of urbanization have increased flood zone occupancy, especially among vulnerable individuals (Houston et al. 2011), while simultaneously altering the surface properties that largely govern rainfall–runoff processes (Leopold 1968; Debbage and Shepherd 2018). A striking example of this urban vulnerability to extreme rainfall occurred in Atlanta, Georgia, during late September 2009, as the city suffered from a historic flooding event (McCallum and Gotvald 2010). The flood occurred as a result of a prolonged period of intense precipitation from 16 to 22 September. The maximum 24-h rainfall total of 534.16 mm was observed west of downtown Atlanta in Douglas County on 21 September (NWS Atlanta 2014). Throughout metropolitan Atlanta, the sustained intense rainfall overwhelmed urbanized watersheds. Streamflow at numerous gauges across the region approached or exceeded the discharge associated with a 100-yr flood (McCallum and Gotvald 2010). In terms of societal impacts, the flood was categorized as “catastrophic” according to the damage-based flash flood severity index of Schroeder et al. (2016). Unfortunately, almost a dozen fatalities were reported and flood damage claims approached $500 million (McCallum and Gotvald 2010; NWS Atlanta 2014).

The extreme flooding was the product of several different meteorological processes that were consistent with the broader flood climatology of the region (Gamble and Meentemeyer 1997; Shepherd et al. 2011). At the synoptic scale, a weak cutoff low stalled over the lower Mississippi Valley on 17 September and resulted in substantial precipitation prior to the heaviest rainfall. The high antecedent moisture conditions produced by the cutoff low likely exacerbated the flood peaks observed several days later. Throughout the flood event, there was also persistent southerly and easterly flow in the region associated with a surface low pressure system over the lower Mississippi Valley and a surface high pressure system over the Great Lakes, respectively. This circulation pattern advected atmospheric moisture from the Gulf of Mexico and Atlantic Ocean into the greater Atlanta region for several days, which provided ample water vapor for the most intense precipitation events. At the mesoscale, training occurred, possibly due to interactions with the topography, and contributed to the localized pockets of anomalously high precipitation. Finally, Shepherd et al. (2011) hypothesized that boundary layer interactions with the urban land cover of Atlanta may have enhanced the overall quantity of precipitation and/or influenced its location. This hypothesis appeared plausible because the largest precipitation totals were located downwind of Atlanta, primarily in Douglas County and Cobb County, which is where urban rainfall enhancement would likely occur based upon the framework established by Shepherd et al. (2002).

The hypothesis that the urban environment of Atlanta influenced the rainfall observed during the 2009 flood was also informed by previous research focused on urban modifications of precipitation and convective activity. These studies date back at least to Horton (1921), who provided anecdotal evidence that urban environments were preferential areas for convective activity. Subsequent observational and climatological analyses produced by the Metropolitan Meteorological Experiment (METROMEX; Huff and Changnon 1973; Changnon et al. 1977) largely supported Horton’s hypothesis. Specifically, METROMEX identified a 9%–17% increase in warm-season precipitation over background values due to urban effects within and 15–55 km downwind of the city (Huff and Changnon 1973). Over the past four decades, studies validated and extended the original findings of METROMEX by more conclusively identifying the casual mechanisms responsible for the urban rainfall effect during both the warm and cold season (Burian and Shepherd 2005; Johnson and Shepherd 2018). Urban-induced alterations of precipitation have generally been attributed to one or a combination of the following mechanisms (Shepherd 2005): 1) the increased surface roughness within urban environments enhancing surface convergence (Thielen et al. 2000); 2) the different thermal properties of the city producing unstable atmospheric conditions through the creation, enhancement, and/or displacement of mesoscale circulations (Shepherd and Burian 2003); 3) elevated aerosol concentrations altering cloud microphysical processes by providing an abundant source of cloud condensation nuclei (CCN; Molders and Olson 2004; Schmid and Niyogi 2017); 4) urban irrigation and industrial activities increasing low-level atmospheric moisture availability (Shepherd et al. 2002; Diem and Brown 2003); and 5) the built environment acting as a barrier that bifurcates existing storm systems (Bornstein and Lin 2000). Both observational and modeling studies have identified the capability of these mechanisms to alter convection, precipitation, and lightning activity within and downwind of Atlanta (Bornstein and Lin 2000; Dixon and Mote 2003; Diem and Mote 2005; Rose et al. 2008; Shem and Shepherd 2009; Ashley et al. 2012; Haberlie et al. 2015; Debbage et al. 2015; McLeod et al. 2017).

The overarching goal of this study was to quantitatively evaluate the hypothesis of Shepherd et al. (2011) that interactions with the urban environment influenced the location and/or quantity of precipitation during the 2009 Atlanta flood. A better understanding of the degree to which urbanization influenced the flood will help inform policies aimed at making the city more resilient to extreme rainfall, which has been identified as a research priority in Georgia due to the state’s rapidly urbanizing population (Rudd et al. 2018). Additionally, this paper provides an extension of past research that has primarily addressed urban precipitation effects under synoptically benign conditions (i.e., warm season convection) by analyzing possible urban-induced alterations of precipitation during an event with prominent synoptic-scale forcing. Although several modeling studies have indicated that urban environments can influence the spatial distribution and quantity of precipitation even under strong large-scale forcing (Ntelekos et al. 2008; Yang et al. 2014; Yu and Liu 2015), they have generally focused on shorter-lived extreme rainfall events during the warm season. Therefore, the complexities of urban rainfall modification during sustained synoptic-scale forcing merit further investigation. It is important to note that urbanization is certainly not the primary contributor to large-scale precipitation systems, such as those responsible for the 2009 Atlanta flood. However, the urban environment may play an important secondary role through localized enhancements of convection and precipitation within the synoptically driven event.

By conducting a numerical weather modeling case study, this paper aimed to answer the following research questions: 1) how did the urban environment of Atlanta influence the spatiotemporal characteristics of runoff and precipitation during the 2009 flood, and 2) what physical mechanisms were potentially responsible for any urban modifications? The subsequent section describes the data and methods used to assess how the hydrometeorological characteristics of the flood were potentially modified by the urban environment. Section 3 presents the results of the numerical modeling experiments. Finally, section 4 summarizes the major findings and explores the potential urban planning implications of the study.

2. Data and methodology

a. WRF Model configuration

Because of the complexity of the 2009 Atlanta flood, it was difficult to conclusively discern the second-order influences of the urban environment from observational records. Therefore, this study relied upon the Weather Research and Forecasting (WRF) Model to conduct a controlled modeling experiment. WRF is a fully compressible, nonhydrostatic, mesoscale model with advanced dynamics, physics, and numerical schemes that is supported by the National Center for Atmospheric Research (NCAR; Skamarock et al. 2008). Specifically, version 3.8.1 of the Advanced Research WRF (WRF-ARW) was used. The WRF-ARW Model was selected because it has been previously utilized to identify urban modifications of precipitation and convective activity in North American cities (e.g., Shem and Shepherd 2009; Ntelekos et al. 2008).

The modeling analysis focused on the Atlanta Metropolitan Statistical Area (MSA), which is located in the southeastern portion of the United States and consists of 28 individual counties (Figs. 1a,b). A two-way nested modeling domain was created to simulate the 2009 Atlanta flood. The outer domain covered a majority of the southeastern United States at a 10-km resolution while the inner domain was centered over Georgia and had a resolution of 2 km (Fig. 1c). The North American Regional Reanalysis (NARR) was used to initialize the model, and it provided the boundary conditions throughout the simulations. NARR was selected primarily because of its high temporal resolution of 3 h and moderate spatial resolution of 32 km (Mesinger et al. 2006). The simulations were initialized on 1200 UTC 13 September, which provided ample spinup time prior to the heaviest precipitation events, and concluded on 1200 UTC 23 September. A 30-s time step was used in both the inner and outer domains. Table 1 summarizes the WRF-ARW physics options selected in the simulations. The physics, except the cumulus parameterization, were the same for both the outer and inner domain to minimize inconsistencies at the boundary. For the cumulus parameterization, the Kain–Fritsch scheme was used in the outer domain, but parameterization was not necessary within the inner domain since the resolution was sufficient to explicitly resolve convection.

Fig. 1.

The Atlanta study region: (a) position of Atlanta within the southeastern United States, (b) detailed map of the Atlanta MSA, and (c) location of the outer and inner WRF modeling domains used to simulate the 2009 Atlanta flood.

Fig. 1.

The Atlanta study region: (a) position of Atlanta within the southeastern United States, (b) detailed map of the Atlanta MSA, and (c) location of the outer and inner WRF modeling domains used to simulate the 2009 Atlanta flood.

Table 1.

WRF-ARW physics options used in the simulations.

WRF-ARW physics options used in the simulations.
WRF-ARW physics options used in the simulations.

b. Modeling experiments: Urban and nonurban simulations

Two different model runs, an urban and nonurban simulation, were performed to evaluate the influence of the urban environment on the 2009 Atlanta flood. In both the urban and nonurban simulations, the configurations described in the previous section remained the same, but the land use scenarios varied. For the urban run, the default Moderate Resolution Imaging Spectroradiometer (MODIS) land use in WRF developed by the International Geosphere–Biosphere Programme (IGBP; Table 2) was augmented within the inner model domain by including urban land use data from the 2011 National Land Cover Database (NLCD; Homer et al. 2015; Fig. 2a). The four NLCD urban classes were reclassified into three categories (i.e., low-intensity residential, high-intensity residential, and commercial/industrial/transportation) to provide the necessary urban parameter information for the single-layer urban canopy model (SLUCM) used to depict urban-related meteorological processes (Kusaka et al. 2001; Chen et al. 2011).

Table 2.

Characteristics of the land use/land cover (LULC) categories included in the modified IGBP MODIS Noah classification scheme.

Characteristics of the land use/land cover (LULC) categories included in the modified IGBP MODIS Noah classification scheme.
Characteristics of the land use/land cover (LULC) categories included in the modified IGBP MODIS Noah classification scheme.
Fig. 2.

Land use scenarios used for the (a) urban and (b) nonurban WRF simulations. The LULC category numbers correspond to the modified IGBP MODIS Noah classification scheme outlined in Table 2. (c) Location of the downwind area of interest included in the areal averages as well as the cross sections in the northern and southern portions of Atlanta.

Fig. 2.

Land use scenarios used for the (a) urban and (b) nonurban WRF simulations. The LULC category numbers correspond to the modified IGBP MODIS Noah classification scheme outlined in Table 2. (c) Location of the downwind area of interest included in the areal averages as well as the cross sections in the northern and southern portions of Atlanta.

The SLUCM is coupled to the Noah land surface model (LSM; Chen and Dudhia 2001) and provides a more realistic representation of heat, momentum, and water vapor exchanges within urban environments by incorporating a simplified urban canyon geometry and considering the shadowing, reflection, and trapping of radiation. The depiction of urban surface roughness is also more detailed by using the SLUCM. In the Noah LSM, all urban land use is parameterized with a surface roughness length of 0.8 m, but in the SLUCM each urban category is associated with a different building height (Table 2). Finally, to provide a more spatially heterogeneous depiction of the urban land use, the urban fraction parameter (FRC_URB2D) was specified using percent developed imperviousness data from the 2011 NLCD rather than relying upon the default WRF lookup tables.

The urban footprint of Atlanta was removed in the nonurban simulation and replaced with natural vegetation, which is a standard modeling technique used to assess urban influences on meteorological phenomena (e.g., Ntelekos et al. 2008; Niyogi et al. 2011; Carter et al. 2012; Wang et al. 2015). Similar to previous WRF modeling studies of Atlanta (Shem and Shepherd 2009; Zhou and Shepherd 2010), the NLCD urban land use within the inner domain was replaced with a cropland/natural vegetation mosaic (class 14) because it is representative of the dominant land cover types surrounding the metropolitan area (Fig. 2b). The cropland/natural vegetation mosaic class was also selected, rather than a forest land cover category, because its lower surface roughness length of 0.14 m enabled an evaluation of how urban surface roughness influenced precipitation (e.g., Zhong and Yang 2015; Table 2). Finally, since the urban environment was excluded from the nonurban run, the SLUCM was not coupled to the Noah LSM and FRC_URB2D was not specified.

c. Analysis of WRF simulation output

Several techniques were used to analyze the output from the two simulations and identify any urban influences on the hydrometeorological characteristics of the flood. The analysis focused on 19–22 September since these days incorporated the most intense precipitation. First, the precipitation from the urban simulation was compared with three sets of observations to ensure that the model adequately captured the dominant features of the storm event. The spatial and temporal distribution of the modeled rainfall was evaluated using the National Centers for Environmental Prediction (NCEP) Stage IV quantitative precipitation estimates (QPEs; Nelson et al. 2016). The NCEP Stage IV QPEs provide gridded Multisensor Precipitation Estimates (MPEs), which combine Doppler radar estimates and observations from station gauges at a spatial resolution of approximately 4 km (Seo 1998; Seo et al. 2010). The overall average root-mean-square error (RMSE) of the MPE throughout the eastern United States is 8.13 mm, although this varies for individual storm events (Wootten and Boyles 2014). Both hourly and daily MPE datasets were utilized in the model validation. Comparisons were also drawn between the simulated precipitation and hourly observations obtained from the Fulton County Airport (KFTY) ASOS station. Lastly, the modeled output was compared to the average daily precipitation totals from the Community Collaborative Rain, Hail and Snow Network (CoCoRaHS; Cifelli et al. 2005) stations located in Douglas County and southern Fulton County.

A quantitative accuracy assessment was performed in addition to qualitative visual comparisons. Specifically, the RMSE was calculated for the daily precipitation totals on 20–22 September by comparing the observed MPE values with the predicted values from WRF. Due to the different projections and spatial resolutions of the MPE dataset and WRF Model output, each WRF precipitation value within the inner domain was compared to the value at the nearest MPE grid cell. Lower RMSE values suggest that the WRF Model more accurately depicted the storm event.

After evaluating the model performance, potential urban modifications were detected by comparing maps of daily precipitation between the urban and nonurban simulations. Areal averages of precipitation from the two simulations were also analyzed, which provided a quantitative assessment. The region included in the areal averages incorporated a majority of Douglas County and the southern portions of Cobb and Fulton counties (Fig. 2c). This specific location was selected because it contained the heaviest rainfall and also represented the area downwind of Atlanta where Shepherd et al. (2011) hypothesized urban rainfall enhancement occurred. Although this region west of Atlanta is not commonly downwind of the city in terms of the broader climatology, the area is consistent with the “flow regime dependent” downwind area concept introduced by McLeod et al. (2017) because the predominant surface flow during the flood event was southeasterly.

To further highlight any discrepancies between the urban and nonurban model runs, difference plots were created for the surface runoff. Maps were also produced for the 10-m wind field, surface convergence, friction velocity, sensible heat flux, and 2-m air temperature to help identify the underlying physical mechanisms responsible for any urban influences on precipitation. Finally, the evolution of the updrafts associated with the rainfall were examined by creating cross sections of vertical velocity along two transects that spanned the city of Atlanta (Fig. 2c), which is a common approach used to assess urban influences on vertical motion (e.g., Niyogi et al. 2011; Wang et al. 2015).

3. Results

a. WRF Model validation

The model performance was evaluated by comparing the daily precipitation accumulation from the urban simulation with the 24-h MPE totals for 20–22 September 2009. The model appeared to capture the major elements of the rainfall distribution each day (Fig. 3). The areas with the greatest precipitation totals on 20 September in northwest Georgia, northeast Georgia, and west of Atlanta were accurately depicted in the model. More detailed features were also resolved, including the narrow band of heavy precipitation west of Atlanta that stretched north of Interstate 20 (I-20) roughly along the border of Paulding and Cobb County. Additionally, the model captured the pocket of intense precipitation over the southwestern quadrant of Atlanta. On 21 September, the urban simulation again adequately captured the regions of intense precipitation in northwest and northeast Georgia. The model also resolved the heavy rainfall directly over the northern and western portions of Atlanta as well as the region of intense precipitation west of the city in Douglas and Cobb counties. Finally, the heavy precipitation within the vicinity of Interstate 575 (I-575) was accurately depicted by the urban simulation on 22 September.

Fig. 3.

Daily precipitation totals from the WRF urban simulation, MPE, and WRF nonurban simulation for 1200 UTC (left) 20 Sep, (center) 21 Sep, and (right) 22 Sep.

Fig. 3.

Daily precipitation totals from the WRF urban simulation, MPE, and WRF nonurban simulation for 1200 UTC (left) 20 Sep, (center) 21 Sep, and (right) 22 Sep.

Although the overall spatial distribution and magnitude of precipitation were well represented in the urban model run, several discrepancies emerged. First, the simulation underestimated precipitation around the boundary of the inner domain, which was likely due to the cumulus parameterization being utilized in the outer domain but not in the inner domain. Since the predominant surface flow was southeasterly, Atlanta was positioned further away from the southeastern domain boundary to prevent possible edge effects from influencing the area of potential urban rainfall enhancement downwind of the city. Second, the precipitation totals appeared to be slightly larger in the urban simulation on 20 September relative to the MPE but smaller on 22 September. This disparity could be due to the inherent limitations of the MPE dataset, which include the errors associated with radar beam geometry, assumptions of ZR relationships, and gauge undercatch (Smith et al. 1996; Sieck et al. 2007; Kitzmiller et al. 2013; Wootten and Boyles 2014). The differences are also possibly indicative of a minor timing discrepancy, as the WRF simulation may have initiated the heaviest rainfall earlier. Despite these inconsistencies, the general spatial distribution and quantity of precipitation predicted by the urban model run was in reasonable agreement with the MPE daily totals.

Additional model validation was performed, which focused explicitly on the region downwind of Atlanta where urban rainfall enhancement potentially occurred. The hourly simulated precipitation was averaged within the downwind area of interest (Fig. 2c) and compared to hourly observations from the KFTY ASOS station and hourly MPE precipitation values averaged over approximately the same region. The cumulative precipitation time series indicated that the urban simulation initiated intense rainfall earlier than the MPE and KFTY observations (Fig. 4a). However, the model did accurately depict two distinct periods of heavy precipitation similar to the KFTY observations (Fig. 4a). Although it may appear that the KFTY station recorded three periods of intense rainfall, this was largely an artifact since the abrupt lull in precipitation on 21 September between 0900 and 1500 UTC was due to a gauge malfunction. The values for a majority of the ASOS variables were reported as missing during these 6 h, and the METAR indicated that system maintenance was required.

Fig. 4.

Comparison of the (a) cumulative and (b) hourly precipitation from the WRF urban simulation averaged within the downwind area of interest with several proximate observations.

Fig. 4.

Comparison of the (a) cumulative and (b) hourly precipitation from the WRF urban simulation averaged within the downwind area of interest with several proximate observations.

A time series of hourly precipitation highlighted similar temporal discrepancies, as the simulated periods of intense rainfall displayed slightly different magnitudes and often preceded the peaks in the MPE record (Fig. 4b). Previous WRF modeling studies have encountered comparable issues regarding the nuanced timing and magnitude of individual rainfall peaks within broader intense precipitation events (Wang et al. 2015). Importantly, as noted by Wang et al. (2015), since both the urban and nonurban simulations displayed similar temporal characteristics, the timing discrepancies could be considered systematic model error, which would not necessarily affect the model comparisons.

The cumulative precipitation time series was also used to examine how well the model resolved the overall precipitation accumulation in the region of potential urban rainfall enhancement. The urban simulation performed satisfactorily, as the cumulative precipitation closely aligned with the areal-averaged MPE and CoCoRaHS data (Fig. 4a). Specifically, the model predicted a total accumulation of 266 mm, which was 11 and 15 mm less than the cumulative precipitation recorded by the CoCoRaHS and MPE, respectively. Because of the missing data during the second period of intense rainfall, the cumulative precipitation total of 205 mm reported at KFTY was likely a severe underestimate.

Finally, the RMSE of the daily precipitation totals predicted by the urban simulation on 20–22 September ranged from 31 to 40 mm. These values were comparable to previous WRF modeling studies of extreme precipitation events (Pennelly et al. 2014), which suggests that the model performance for the 2009 Atlanta flood was adequate. For the 3-day period of interest, the RMSE of the urban simulation was lower than that of the nonurban simulation, indicating that the inclusion of the urban land use improved the model performance. Overall, the realistic depiction of the major elements associated with the 2009 Atlanta flood by the urban simulation suggests that the modeling experiment (i.e., removing the urban footprint of Atlanta) provided a substantive assessment of how urbanization influenced the hydrometeorological characteristics.

b. Comparison of precipitation and runoff between the urban and nonurban simulations

The predicted daily precipitation accumulation from the nonurban simulation was also mapped to provide a qualitative comparison between the urban and nonurban model runs (Fig. 3). On 20 September, the nonurban simulation failed to capture the intense precipitation in Douglas County south of I-20. Instead, the model appeared to incorrectly shift this pocket of heavy rainfall northward. The nonurban simulation also erroneously predicted a broad swath of intense precipitation northwest of Atlanta parallel to Interstate 75 (I-75), which was not observed in the MPE. Most importantly, the nonurban simulation failed to resolve the narrow band of heavy precipitation that stretched north of I-20 roughly along the border of Paulding and Cobb County.

Similar discrepancies were observed on 21 September. The nonurban simulation again underestimated the quantity of precipitation in Douglas County and southern Fulton County. Additionally, the model failed to capture the intense precipitation directly over the northern and western quadrants of Atlanta. The nonurban simulation also overestimated precipitation south of Atlanta between LaGrange and Macon during this period. Although overestimates in this region were observed in the urban model, they were more pronounced in the nonurban model. Finally, the differences between the two simulations were less notable on 21 September, as both models underestimated the daily total precipitation due to the earlier initiation of the extreme rainfall in WRF. The nonurban simulation, however, still produced less precipitation in the area of potential urban enhancement, particularly within Interstate 285 (I-285) over northwest Atlanta. Overall, the visual differences between the two model outputs suggest that the urban environment played a notable role in governing both the location and the overall quantity of precipitation.

Averaging the precipitation from each simulation within the area of interest downwind of Atlanta (Fig. 2c) provided a quantitative assessment of the potential urban rainfall enhancement. Plots of cumulative rainfall revealed that the nonurban simulation underestimated the precipitation during both periods of intense rainfall (Fig. 5a). Overall, the nonurban model predicted a precipitation total of 172 mm during the 5-day period analyzed, which was approximately 100 mm less than the cumulative precipitation produced by the urban model. The underestimation by the nonurban simulation was so severe that its total cumulative precipitation was less than that reported by KFTY, which malfunctioned during the second phase of intense rainfall. The substantial differences revealed by this quantitative assessment further suggest that interactions with the urban environment of Atlanta likely amplified precipitation in Douglas County and southern Fulton and Cobb counties.

Fig. 5.

Comparison of the (a) cumulative and (b) hourly precipitation averaged within the downwind area of interest between the urban and nonurban WRF simulations.

Fig. 5.

Comparison of the (a) cumulative and (b) hourly precipitation averaged within the downwind area of interest between the urban and nonurban WRF simulations.

To further assess the differences between the urban and nonurban simulations during the two periods of intense rainfall, plots of hourly precipitation were created (Fig. 5b). The first phase of intense precipitation began at 0000 UTC 20 September. The urban simulation predicted greater quantities of precipitation throughout the majority of this phase of rainfall, with a maximum hourly difference of approximately 10 mm. During the second period of intense rainfall, the urban and nonurban simulations predicted similar precipitation values for the first several hours. The most notable dissimilarities occurred during the latter half of the storm’s evolution with the urban simulation predicting greater quantities of precipitation. However, the maximum hourly difference during the second phase of intense precipitation was only 5 mm. Overall, the varying magnitude of the differences between the urban and nonurban simulations during each phase of intense rainfall suggests that different physical mechanisms may have been responsible for the urban precipitation enhancements.

The notable disparities in precipitation had clear implications in terms of the surface runoff and subsequent flooding. To visualize these discrepancies, difference maps were created for surface runoff by subtracting the values predicted by the nonurban model from the urban model. During the first phase of intense precipitation, a well-defined positive surface runoff anomaly approaching 80 mm was observed north of I-20 roughly along the border of Paulding and Cobb County (Fig. 6a). This was likely due to the nonurban simulation failing to resolve the narrow band on intense precipitation that developed downwind of Atlanta. Anomalies of a greater magnitude occurred during the second phase of heavy rainfall, as the urban simulation produced 120 mm of additional surface runoff directly over southwest Atlanta (Fig. 6b). Since the surface runoff anomalies exceeded the differences in precipitation between the urban and nonurban simulation during each phase of intense rainfall, they cannot be attributed to urban precipitation enhancement alone. Instead, the enhanced surface runoff in the urban simulation was likely the combined effect of additional rainfall, saturated soils, and the impervious nature of the urban land use. This highlights a potential two-pronged vulnerability of cities to extreme rainfall, as interactions with the urban environment may enhance both the quantity of precipitation and the subsequent surface runoff (Yang et al. 2013; McLeod et al. 2017).

Fig. 6.

Difference in the surface runoff between the urban and nonurban simulations during the (a) first and (b) second phase of intense precipitation. Positive values represent greater surface runoff in the urban simulation.

Fig. 6.

Difference in the surface runoff between the urban and nonurban simulations during the (a) first and (b) second phase of intense precipitation. Positive values represent greater surface runoff in the urban simulation.

c. Potential mechanisms responsible for urban modifications

To identify the potential mechanisms responsible for the observed urban rainfall enhancement downwind of Atlanta, the 10-m wind field, surface convergence, friction velocity, vertical velocity, sensible heat flux, and 2-m air temperature from both simulations were examined. The 10-m wind field was superimposed on the modeled land use to highlight potential surface flow interactions with the urban environment. Prior to the first phase of intense precipitation on 20 September at 0000 UTC, a clear boundary developed downwind of Atlanta along the urban–rural interface in the urban simulation (Fig. 7a). Importantly, this boundary was not observed when the urban footprint of Atlanta was removed from the model. The enhanced surface roughness within the urban environment was likely responsible for these differences, as the city appeared to bifurcate the surface flow and subsequently enhance convergence at the downwind urban–rural interface. At 0100 UTC, this boundary was still well pronounced along the border of Paulding and Cobb County north of I-20 in the urban simulation while it was largely nonexistent in the nonurban simulation. The failure of the nonurban model to resolve the surface flow boundary at the urban–rural interface likely contributed to its underestimation of precipitation downwind of Atlanta.

Fig. 7.

Differences in the (a) 10-m wind field and (b) divergence at the beginning of the first phase of intense precipitation between the (left) urban and (right) nonurban WRF simulations.

Fig. 7.

Differences in the (a) 10-m wind field and (b) divergence at the beginning of the first phase of intense precipitation between the (left) urban and (right) nonurban WRF simulations.

Divergence/convergence fields were computed using the u and υ wind components to further visualize how the urban environment potentially enhanced near-surface convergence prior to the first period of extreme rainfall on 20 September. At 0000 UTC, several strong, linear bands of convergence (i.e., negative divergence) developed downwind of Atlanta along the western urban–rural interface in the urban simulation (Fig. 7b). The linear extent and strength of this convergence zone was not fully captured in the nonurban simulation. Additionally, along the eastern urban–rural boundary a noticeable ribbon of divergence was observed in the urban simulation, which was less pronounced in the nonurban simulation. This band of divergence along the eastern edge of Atlanta supports the hypothesis that the surface flow was bifurcated by the city, due to the enhanced surface roughness within the urban environment (e.g., Bornstein and Lin 2000; Niyogi et al. 2011), which ultimately contributed to the enhanced converged observed at the downwind urban–rural interface.

The general pattern of enhanced convergence downwind of Atlanta along the urban–rural interface was persistent for several hours in the urban simulation. This suggests that interactions with the urban environment helped establish and sustain a linear convergence zone downwind of Atlanta, which contributed to the greater rainfall totals observed in Cobb, Paulding, and Douglas counties during the first period of intense precipitation. Previous studies have similarly indicated that urban land use can enhance convergence along the urban–rural interface rather than directly over the urban core (Shem and Shepherd 2009).

Analysis of friction velocity also indicated that the surface roughness of Atlanta was a critical mechanism that contributed to the bifurcation of the surface flow and enhanced downwind convergence in the urban simulation. Greater friction velocity values were observed within the urban footprint of Atlanta in the urban simulation at 0100 UTC (Fig. 8a), which suggests that the flow was more turbulent within the city at the onset of the first phase of intense precipitation. The mechanical turbulence associated with the increased surface roughness of urban environments has been widely documented to increase surface convergence (e.g., Thielen et al. 2000). Urban influences on the thermal environment appeared to be of less importance during the first phase of intense precipitation as modest differences in the sensible heat flux and 2-m temperature were observed between the urban and nonurban models. Urban alterations of air temperature and sensible heat were likely moderated to some extent by the rainfall that occurred prior to the first phase of intense precipitation at 1800 UTC.

Fig. 8.

Differences in the (a) friction velocity and (b) vertical velocity along the northern transect at the beginning of the first phase of intense precipitation between the (left) urban and (right) nonurban WRF simulations.

Fig. 8.

Differences in the (a) friction velocity and (b) vertical velocity along the northern transect at the beginning of the first phase of intense precipitation between the (left) urban and (right) nonurban WRF simulations.

Finally, vertical velocity cross sections were created for both the urban and nonurban simulations to visualize how the urban environment of Atlanta potentially altered the circulation patterns. For the first period of intense rainfall, the vertical circulations associated with the narrow band of intense precipitation north of I-20 along the border of Cobb and Paulding County were examined using the northern cross section (Fig. 2c). At 0100 UTC 20 September, the urban simulation predicted notable vertical velocities exceeding 6 m s−1 at the downwind urban–rural interface while the maximum vertical velocity predicted by the nonurban simulation at the same location was less than 4 m s−1 (Fig. 8b). Additionally, the updraft observed in the nonurban model exhibited less vertical development. Similar differences were observed for the following hours, as the vertical velocities at the downwind urban–rural interface were greater in the urban simulation. The weaker updraft present in the nonurban simulation was also less stationary and progressed farther eastward by 0300 UTC.

Overall, the differences in the vertical velocities suggest that the enhanced downwind convergence in the urban simulation supported an earlier formation of updrafts, which were stronger and more persistent along the urban–rural interface. The capability of urban environments to enhance vertical velocities and the resulting convection has been documented by previous modeling studies that performed similar urban versus nonurban comparisons (Niyogi et al. 2011). The vertical velocity discrepancies at least partially explain the greater rainfall totals observed along the border of Cobb and Paulding County in the urban model.

A similar analysis was conducted for the second period of intense rainfall that began at 0000 UTC 21 September. Initially, the differences in the 10-m wind field were not as pronounced between the two simulations. A boundary developed in the urban model over the southwest portion of Atlanta that stretched southwestward along Interstate 85 (I-85) at 0200 UTC 21 September. In the nonurban simulation, a similar boundary was present, although it was shifted toward the northwest. Importantly, by 0300 UTC, the boundary progressed eastward in the nonurban model while in the urban simulation it was still located directly over the city, roughly parallel to I-85 (Fig. 9a). The less-persistent boundary observed over Atlanta in the nonurban model likely contributed to its underestimation of precipitation within the city during the second period of intense rainfall.

Fig. 9.

Differences in the (a) 10-m wind field and (b) divergence during the second phase of intense precipitation between the (left) urban and (right) nonurban WRF simulations.

Fig. 9.

Differences in the (a) 10-m wind field and (b) divergence during the second phase of intense precipitation between the (left) urban and (right) nonurban WRF simulations.

Differences in convergence during the second phase of extreme precipitation mirrored the discrepancies observed in the 10-m wind field. At 0200 UTC 21 September, both models initially resolved pockets of convergence over Atlanta. However, the nonurban simulation predicted the areas of intense convergence would progress eastward beyond I-285 by 0300 UTC, while in the urban model a crescent shaped band of convergence was still present directly over the city (Fig. 9b). This suggests that interactions with the urban environment may have helped sustain the convergence zone directly over Atlanta.

Analysis of the sensible heat flux and 2-m air temperature indicated that urban influences on the thermal environment potentially contributed to the stationary nature of the convergence zone during the second phase of intense precipitation. The sensible heat flux within Atlanta was ~60 W m−2 greater in the urban simulation relative to the nonurban model prior to the second phase of intense rainfall at 2300 UTC (Fig. 10a). Several hours earlier at 2100 and 2200 UTC, these differences were even more pronounced, as the sensible heat fluxes within Atlanta in the urban simulation were ~100–125 W m−2 greater. These differences in the sensible heat flux between the urban and nonurban simulations within Atlanta were comparable to similar modeling studies that have documented urban influences on precipitation (Shem and Shepherd 2009; Shepherd et al. 2010).

Fig. 10.

Differences in the (a) sensible heat flux, (b) 2-m air temperature, and (c) vertical velocity along the southern transect during the second phase of intense precipitation between the (left) urban and (right) nonurban WRF simulations.

Fig. 10.

Differences in the (a) sensible heat flux, (b) 2-m air temperature, and (c) vertical velocity along the southern transect during the second phase of intense precipitation between the (left) urban and (right) nonurban WRF simulations.

Importantly, the surface energy balance leading up to the second phase of intense precipitation was less influenced by antecedent rainfall, as the simulated reflectively and hourly precipitation plots (Fig. 4b) highlighted that little precipitation occurred over Atlanta during the preceding 6–8 h. This enabled more pronounced surface heating, especially compared to the first period of intense rainfall, which was preceded by precipitation at 1800 UTC 19 September (Fig. 4b). The lack of antecedent precipitation likely explains the more important role of the thermal environment during the second phase of intense precipitation since the diurnal heating of urban surfaces was not disrupted by afternoon precipitation. The highest sensible heat fluxes in the urban model were observed directly over downtown Atlanta and Hartsfield–Jackson Atlanta International Airport, which further demonstrated the important influence of intense urban land use on the surface energy balance prior to the second period of intense rainfall.

Plots of 2-m air temperature illustrated that an urban heat island (UHI) was also evident at 0000 UTC 21 September before the second phase of intense precipitation began (Fig. 10b). The core of the UHI was displaced toward the northwest of downtown Atlanta, which was expected due to the southeasterly flow. Collectively, the higher sensible heat fluxes and UHI depicted in the urban simulation likely contributed to the development of a sustained UHI circulation that supported the stationary convergence zone across Atlanta. Unfortunately, it was challenging to assess the role of urban surface roughness during the second phase of precipitation using friction velocity since the large values observed over Atlanta in both the urban and nonurban simulations were likely influenced by the individual storm elements.

For the second period of intense rainfall, an analysis of the vertical circulations within the convergence zone located directly over Atlanta was performed using the southern cross section (Fig. 2c). During the first 3 h of rainfall from 0000 to 0200 UTC, the vertical velocities were greater in the urban simulation relative to the nonurban simulation. The updraft was marginally stronger in the nonurban simulation by 0300 UTC, but it had shifted eastward and was no longer located over the urban core of Atlanta (Fig. 10c). Conversely, in the urban model, two updrafts were still present, with one located closer to the downwind urban–rural interface and one positioned directly over the city. The streamlines highlighted that low-level convergence potentially contributed to the stronger updraft observed over the city in the urban simulation at 0300 UTC. The cross-section results suggest that urban influences on the thermal environment during the second period of intense rainfall contributed to the quasi-stationary nature of the convection, which Shepherd et al. (2011) identified as an important contributor to the prodigious precipitation totals.

Overall, interactions with the urban environment appeared to alter the surface flow regime substantially and support the formation of persistent and well-defined boundaries, which were conducive for extreme precipitation. During the first phase of intense precipitation, the greater surface roughness of the urban environment was of primary importance while urban thermal effects were likely more influential during the second period of intense rainfall. This may explain the greater urban rainfall enhancement observed during the first period of intense precipitation, as the urban environment appeared to bifurcate the surface flow and support the development of a broad convergence zone at the downwind urban–rural interface. Conversely, the urban thermal effects observed during the second period helped sustain a convergence zone over Atlanta, which primarily enhanced rainfall during the latter stages of the storm’s evolution. These findings are consistent with Zhang et al. (2017) regarding how the location of urban precipitation enhancements relates to the underlying causal mechanisms. Specifically, Zhang et al. (2017) concluded that urban surface roughness enhances precipitation outside the urban core whereas a stronger UHI effect tends to amplify precipitation directly over the city. This important distinction mirrors the different locations and physical mechanisms responsible for the urban rainfall enhancements observed in the urban simulation during the first and second phases of intense precipitation, respectively.

4. Conclusions and urban planning implications

By performing the first WRF modeling case study of the 2009 Atlanta flood, this paper explored how the hydrometeorological characteristics of the flood were potentially influenced by the urban environment. A base run urban simulation was initially conducted, which appeared to accurately capture the general spatial patterning and overall quantity of precipitation when compared with ASOS, MPE, and CoCoRaHS measurements. A nonurban simulation was then performed by replacing the urbanized land use with natural vegetation. Finally, both qualitative and quantitative comparisons were made between the urban and nonurban models to assess the influence of urbanization on the flooding event.

The comparative analysis revealed noticeable discrepancies between the two simulations in the area downwind of Atlanta where Shepherd et al. (2011) hypothesized that urban rainfall enhancement occurred. Specifically, the nonurban model underestimated the cumulative precipitation in this region by approximately 100 mm. The substantial differences in precipitation combined with the impervious land cover of the urban environment produced notable disparities in the surface runoff between the urban and nonurban simulations as well. In some regions of Atlanta, the urban simulation predicted 120 mm of additional surface runoff. Overall, the findings indicated that the urban environment of Atlanta played a notable role in governing the location and quantity of precipitation and runoff that ultimately produced the record-breaking 2009 flood.

Several different physical mechanisms were potentially responsible for the stark contrasts between the urban and nonurban simulations. Interactions with the urban environment appeared to alter the surface flow regime, which invigorated and sustained areas of convergence. During the first period of intense precipitation, the elevated surface roughness within the urban environment appeared to bifurcate the surface flow and enhance convergence at the downwind urban–rural interface. For the second phase of intense rainfall, the greater sensible heat fluxes and formation of a UHI in the urban simulation likely helped sustain convergence directly over Atlanta. These two mechanisms acted to produce stronger and more persistent updrafts in the urban simulation during each period of intense precipitation, which amplified the rainfall totals downwind of and directly over downtown Atlanta. By quantitatively evaluating the urban rainfall effect during the 2009 Atlanta flood and identifying potential explanatory mechanisms, this study largely confirms the hypothesis of Shepherd et al. (2011) that urbanization substantially increased not only the surface runoff but also the quantity of precipitation that produced the flood. Furthermore, the findings suggest that interactions with the urban environment can alter the mesoscale hydrometeorological characteristics of a storm event even if it is synoptically driven.

From an urban planning perspective, these findings highlight a potential two-pronged vulnerability of urban environments to extreme rainfall. Although it is widely understood that impervious surfaces can increase surface runoff during extreme rainfall events (e.g., Yang et al. 2013; Debbage and Shepherd 2018), the capability of urban environments to also amplify the initial precipitation is not broadly acknowledged. Fortunately, this two-pronged vulnerability may indicate that mitigation strategies aimed at reducing the influence of urbanization on runoff have previously unrecognized synergistic benefits. For example, while incorporating green infrastructure within a city can enhance infiltration and mitigate increases in surface runoff due to urbanization (Kim and Park 2016), such strategies could also potentially reduce the surface roughness within the urban environment and mitigate the UHI effects that are often responsible for urban rainfall enhancement. This highlights the importance of green infrastructure as a possible synergistic policy solution due to its well-recognized runoff benefits and its potential to reduce urban rainfall enhancement as well. Additional research, however, is necessary to more fully understand the scales at which the potential cobenefits of urban green infrastructure are maximized. Importantly, the National Research Council has examined many facets of this topic and provided specific recommendations regarding how to operationalize such research in urban planning and flood management arenas (National Research Council 2012).

Of course, the model findings are sensitive to the WRF specifications used in this study as well as the unique characteristics of the storm itself. Although the case study approach enabled a detailed assessment of the urban influences on precipitation during the 2009 Atlanta flood, considering additional cases will be imperative to conclusively determine the commonality of urban rainfall enhancement during large-scale flooding events. For example, the 2015 Texas–Oklahoma flood would likely provide another informative case study. Future modeling efforts that utilize higher spatial resolutions, more complex representations of urban land use based upon urban climate zones, and multilayer urban canopy models may also enable further insights regarding the physical mechanisms responsible for urban rainfall enhancements during floods. Despite these avenues for additional research, the present study contributes to an emerging body of literature that suggests urban environments can influence the magnitude and/or location of precipitation during events with prominent synoptic-scale forcing (e.g., Ntelekos et al. 2008; Yang et al. 2014; Yu and Liu 2015; Reames and Stensrud 2018).

Acknowledgments

Funding for this research was provided by a grant (2017GA373B) from the Georgia Water Resources Institute. The authors are thankful for the feedback provided by three anonymous reviewers that helped improve the quality of the initial manuscript. Finally, the authors are appreciative of fruitful discussions with Marcus Williams, Bradford Johnson, Paul Miller, and Kyle Mattingly.

REFERENCES

REFERENCES
Ashley
,
W. S.
,
M. L.
Bentley
, and
J. A.
Stallins
,
2012
:
Urban-induced thunderstorm modification in the southeast United States
.
Climatic Change
,
113
,
481
498
, https://doi.org/10.1007/s10584-011-0324-1.
Bornstein
,
R.
, and
Q.
Lin
,
2000
:
Urban heat islands and summertime convective thunderstorms in Atlanta: Three case studies
.
Atmos. Environ.
,
34
,
507
516
, https://doi.org/10.1016/S1352-2310(99)00374-X.
Burian
,
S. J.
, and
J. M.
Shepherd
,
2005
:
Effect of urbanization on the diurnal rainfall pattern in Houston
.
Hydrol. Processes
,
19
,
1089
1103
, https://doi.org/10.1002/hyp.5647.
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
, https://doi.org/10.1175/2011JTECHA1524.1.
Changnon
,
S. A.
,
F. A.
Huff
,
P. T.
Schickedanz
, and
J. L.
Vogel
,
1977
: Summary of METROMEX, Volume 1: Weather anomalies and impacts. Illinois State Water Survey Bulletin 62, 260 pp., https://www.isws.illinois.edu/pubdoc/B/ISWSB-62.pdf.
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
, https://doi.org/10.1175/1520-0493(2001)129<0569:CAALSH>2.0.CO;2.
Chen
,
F.
, and Coauthors
,
2011
:
The integrated WRF/urban modelling system: development, evaluation, and applications to urban environmental problems
.
Int. J. Climatol.
,
31
,
273
288
, https://doi.org/10.1002/joc.2158.
Cifelli
,
R.
,
N.
Doesken
,
P.
Kennedy
,
L. D.
Carey
,
S. A.
Rutledge
,
C.
Gimmestad
, and
T.
Depue
,
2005
:
The Community Collaborative Rain, Hail, and Snow network: Informal education for scientists and citizens
.
Bull. Amer. Meteor. Soc.
,
86
,
1069
1077
, https://doi.org/10.1175/BAMS-86-8-1069.
Debbage
,
N.
, and
J. M.
Shepherd
,
2018
:
The influence of urban development patterns on streamflow characteristics in the Charlanta Megaregion
.
Water Resour. Res.
,
54
,
3728
3747
, https://doi.org/10.1029/2017WR021594.
Debbage
,
N.
,
J.
McLeod
,
J.
Rackley
,
L.
Zhu
,
T.
Mote
, and
A. J.
Grundstein
,
2015
:
The influence of point source aerosol emissions on atmospheric convective activity in the vicinity of power plants in Georgia, USA
.
Pap. Appl. Geogr.
,
1
,
134
142
, https://doi.org/10.1080/23754931.2015.1012430.
Diem
,
J. E.
, and
D. P.
Brown
,
2003
:
Anthropogenic impacts on summer precipitation in Central Arizona, U.S.A
.
Prof. Geogr.
,
55
,
343
355
, https:/doi.org/10.1111/0033-0124.5503011.
Diem
,
J. E.
, and
T. L.
Mote
,
2005
:
Interepochal changes in summer precipitation in the Southeastern United States: Evidence of possible urban effects near Atlanta, Georgia
.
J. Appl. Meteor.
,
44
,
717
730
, https://doi.org/10.1175/JAM2221.1.
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
, https://doi.org/10.1175/1520-0450(2003)042<1273:PACOAU>2.0.CO;2.
Gamble
,
D. W.
, and
V.
Meentemeyer
,
1997
:
A synoptic climatology of extreme unseasonable floods in the southeastern United States, 1950–1990
.
Phys. Geogr.
,
18
,
496
524
, https://doi.org/10.1080/02723646.1997.10642632.
Haberlie
,
A. M.
,
W. S.
Ashley
, and
T. J.
Pingel
,
2015
:
The effect of urbanisation on the climatology of thunderstorm initiation
.
Quart. J. Roy. Meteor. Soc.
,
141
,
663
675
, https://doi.org/10.1002/qj.2499.
Homer
,
C. G.
, and Coauthors
,
2015
:
Completion of the 2011 National Land Cover Database for the conterminous United States-Representing a decade of land cover change information
.
Photogramm. Eng. Remote Sensing
,
81
,
345
354
, https://doi.org/10.14358/PERS.81.5.345.
Horton
,
R. E.
,
1921
:
Thunderstorm-breeding spots
.
Mon. Wea. Rev.
,
49
,
193
, https://doi.org/10.1175/1520-0493(1921)49<193a:TS>2.0.CO;2.
Houston
,
D.
,
A.
Werritty
,
D.
Bassett
,
A.
Geddes
,
A.
Hoolachan
, and
M.
McMillan
,
2011
: Pluvial (rain-related) flooding in urban areas: The invisible hazard. Joseph Rowntree Foundation, 96 pp., https://www.jrf.org.uk/report/pluvial-rain-related-flooding-urban-areas-invisible-hazard.
Huff
,
F. A.
, and
S. A.
Changnon
,
1973
:
Precipitation modification by major urban areas
.
Bull. Amer. Meteor. Soc.
,
54
,
1220
1232
, https://doi.org/10.1175/1520-0477(1973)054<1220:PMBMUA>2.0.CO;2.
Johnson
,
B.
, and
J. M.
Shepherd
,
2018
:
An urban-based climatology of winter precipitation in the northeast United States
.
Urban Climate
,
24
,
205
220
, https://doi.org/10.1016/j.uclim.2018.03.003.
Kim
,
H. W.
, and
Y.
Park
,
2016
:
Urban green infrastructure and local flooding: The impact of landscape patterns on peak runoff in four Texas MSAs
.
Appl. Geogr.
,
77
,
72
81
, https://doi.org/10.1016/j.apgeog.2016.10.008.
Kitzmiller
,
D.
,
D.
Miller
,
R.
Fulton
, and
F.
Ding
,
2013
:
Radar and multisensor precipitation estimation techniques in National Weather Service hydrologic operations
.
J. Hydrol. Eng.
,
18
,
133
142
, https://doi.org/10.1061/(ASCE)HE.1943-5584.0000523.
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
, https://doi.org/10.1023/A:1019207923078.
Leopold
,
L. B.
,
1968
: Hydrology for urban planning—A guidebook on the hydrologic effects of urban land use. Geological Survey Circular 554, USGS, 26 pp., https://pubs.usgs.gov/circ/1968/0554/report.pdf.
McCallum
,
B. E.
, and
A. J.
Gotvald
,
2010
: Historic flooding in northern Georgia, September 16-22, 2009. USGS Fact Sheet 2010-3061, 3 pp., http://pubs.usgs.gov/fs/2010/3061/pdf/fs2010-3061.pdf.
McLeod
,
J.
,
J. M.
Shepherd
, and
C. E.
Konrad
,
2017
:
Spatio-temporal rainfall patterns around Atlanta, Georgia and possible relationships to urban land cover
.
Urban Climate
,
21
,
27
42
, https://doi.org/10.1016/j.uclim.2017.03.004.
Mesinger
,
F.
, and Coauthors
,
2006
:
North American Regional Reanalysis
.
Bull. Amer. Meteor. Soc.
,
87
,
343
360
, https://doi.org/10.1175/BAMS-87-3-343.
Molders
,
N.
, and
M. A.
Olson
,
2004
:
Impact of urban effects on precipitation in high latitudes
.
J. Hydrometeor.
,
5
,
409
429
, https://doi.org/10.1175/1525-7541(2004)005<0409:IOUEOP>2.0.CO;2.
National Research Council
,
2012
: Urban Meteorology: Forecasting, Monitoring, and Meeting Users’ Needs. National Academies Press, 190 pp., https://doi.org/10.17226/13328.
Nelson
,
B. R.
,
O. P.
Prat
,
D. J.
Seo
, and
E.
Habib
,
2016
:
Assessment and implications of NCEP stage IV quantitative precipitation estimates for product intercomparisons
.
Wea. Forecasting
,
31
,
371
394
, https://doi.org/10.1175/WAF-D-14-00112.1.
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
, https://doi.org/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
, https://doi.org/10.1029/2007WR006346.
NWS Atlanta
,
2014
: 5th Anniversary Atlanta Flood. NOAA, 1 pp., http://www.weather.gov/images/ffc/events/SeptFlood2009/September2009_infographic.png.
Pennelly
,
C.
,
G.
Reuter
, and
T.
Flesch
,
2014
:
Verification of the WRF model for simulating heavy precipitation in Alberta
.
Atmos. Res.
,
135–136
,
172
192
, https://doi.org/10.1016/j.atmosres.2013.09.004.
Reames
,
L. J.
, and
D. J.
Stensrud
,
2018
:
Influence of a Great Plains urban environment on a simulated supercell
.
Mon. Wea. Rev.
,
146
,
1437
1462
, https://doi.org/10.1175/MWR-D-17-0284.1.
Rose
,
L. S.
,
J. A.
Stallins
, and
M. L.
Bentley
,
2008
:
Concurrent cloud-to-ground lightning and precipitation enhancement in the Atlanta, Georgia (United States), urban region
.
Earth Interact.
,
12
,
1
30
, https://doi.org/10.1175/2008EI265.1.
Rudd
,
M. A.
, and Coauthors
,
2018
:
Climate research priorities for policy-makers, practitioners, and scientists in Georgia, USA
.
Environ. Manage.
,
62
,
190
209
, https://doi.org/10.1007/s00267-018-1051-4.
Schmid
,
P. E.
, and
D.
Niyogi
,
2017
:
Modeling urban precipitation modification by spatially heterogeneous aerosols
.
J. Appl. Meteor. Climatol.
,
56
,
2141
2153
, https://doi.org/10.1175/JAMC-D-16-0320.1.
Schroeder
,
A. J.
, and Coauthors
,
2016
:
The development of a flash flood severity index
.
J. Hydrol.
,
541
,
523
532
, https://doi.org/10.1016/j.jhydrol.2016.04.005.
Seo
,
D. J.
,
1998
:
Real-time estimation of rainfall fields using radar rainfall and rain gage data
.
J. Hydrol.
,
208
,
37
52
, https://doi.org/10.1016/S0022-1694(98)00141-3.
Seo
,
D. J.
,
A.
Seed
, and
G.
Delrieu
,
2010
: Radar and multisensor rainfall estimation for hydrologic applications. Rainfall: State of the Science, Geophys. Monogr., Vol. 191, Amer. Geophys. Union, 79–104.
Shem
,
W.
, and
J. M.
Shepherd
,
2009
:
On the impact of urbanization on summertime thunderstorms in Atlanta: Two numerical model case studies
.
Atmos. Res.
,
92
,
172
189
, https://doi.org/10.1016/j.atmosres.2008.09.013.
Shepherd
,
J. M.
,
2005
:
A review of current investigations of urban-induced rainfall and recommendations for the future
.
Earth Interact.
,
9
, https://doi.org/10.1175/EI156.1.
Shepherd
,
J. M.
, and
S. J.
Burian
,
2003
:
Detection of urban-induced rainfall anomalies in a major coastal city
.
Earth Interact.
,
7
, https://doi.org/10.1175/1087-3562(2003)007<0001:DOUIRA>2.0.CO;2.
Shepherd
,
J. M.
,
H.
Pierce
, and
A. J.
Negri
,
2002
:
Rainfall modification by major urban areas: Observations from spaceborne rain radar on the TRMM satellite
.
J. Appl. Meteor.
,
41
,
689
701
, https://doi.org/10.1175/1520-0450(2002)041<0689:RMBMUA>2.0.CO;2.
Shepherd
,
J. M.
,
M.
Carter
,
M.
Manyin
,
D.
Messen
, and
S.
Burian
,
2010
:
The impact of urbanization on current and future coastal precipitation: A case study for Houston
.
Environ. Plann.
,
37B
,
284
304
, https://doi.org/10.1068/b34102t.
Shepherd
,
J. 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
, https://doi.org/10.1175/2010BAMS3003.1.
Sieck
,
L. C.
,
S. J.
Burges
, and
M.
Steiner
,
2007
:
Challenges in obtaining reliable measurements of point rainfall
.
Water Resour. Res.
,
43
,
W01420
, https://doi.org/10.1029/2005WR004519.
Skamarock
,
W. C.
, and Coauthors
,
2008
: A description of the Advanced Research WRF version 3. NCAR Tech. Note NCAR/TN-475+STR, 113 pp., https://doi.org/10.5065/D68S4MVH.
Smith
,
J. A.
,
D. J.
Seo
,
M. L.
Baeck
, and
M. D.
Hudlow
,
1996
:
An intercomparison study of NEXRAD precipitation estimates
.
Water Resour. Res.
,
32
,
2035
2045
, https://doi.org/10.1029/96WR00270.
Thielen
,
J.
,
W.
Wobrock
,
A.
Gadian
,
P. G.
Mestayer
, and
J.-D.
Creutin
,
2000
:
The possible influence of urban surfaces on rainfall development: A sensitivity study in 2D in the meso-γ-scale
.
Atmos. Res.
,
54
,
15
39
, https://doi.org/10.1016/S0169-8095(00)00041-7.
Wang
,
J.
,
J.
Feng
, and
Z.
Yan
,
2015
:
Potential sensitivity of warm season precipitation to urbanization extents: Modeling study in Beijing-Tianjin-Hebei urban agglomeration in China
.
J. Geophys. Res. Atmos.
,
120
,
9408
9425
, https://doi.org/10.1002/2015JD023572.
Wootten
,
A.
, and
R. P.
Boyles
,
2014
:
Comparison of NCEP multisensor precipitation estimates with independent gauge data over the eastern United States
.
J. Appl. Meteor. Climatol.
,
53
,
2848
2862
, https://doi.org/10.1175/JAMC-D-14-0034.1.
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
, https://doi.org/10.1175/JHM-D-12-095.1.
Yang
,
L.
,
J. A.
Smith
,
M. L.
Baeck
,
E.
Bou-Zeid
, and
S. M.
Jessup
,
2014
:
Impact of urbanization on heavy convective precipitation under strong large-scale forcing: A case study over the Milwaukee–Lake Michigan region
.
J. Hydrometeor.
,
15
,
261
278
, https://doi.org/10.1175/JHM-D-13-020.1.
Yu
,
M.
, and
Y.
Liu
,
2015
:
The possible impact of urbanization on a heavy rainfall event in Beijing
.
J. Geophys. Res. Atmos.
,
120
,
8132
8143
, https://doi.org/10.1002/2015JD023336.
Zhang
,
Y.
,
S.
Miao
,
Y.
Dai
, and
R.
Bornstein
,
2017
:
Numerical simulation of urban land surface effects on summer convective rainfall under different UHI intensity in Beijing
.
J. Geophys. Res. Atmos.
,
122
,
7851
7868
, https://doi.org/10.1002/2017JD026614.
Zhong
,
S.
, and
X. Q.
Yang
,
2015
:
Ensemble simulations of the urban effect on a summer rainfall event in the Great Beijing Metropolitan Area
.
Atmos. Res.
,
153
,
318
334
, https://doi.org/10.1016/j.atmosres.2014.09.005.
Zhou
,
Y.
, and
J. M.
Shepherd
,
2010
:
Atlanta’s urban heat island under extreme heat conditions and potential mitigation strategies
.
Nat. Hazards
,
52
,
639
668
, https://doi.org/10.1007/s11069-009-9406-z.

Footnotes

© 2019 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).