1. Introduction
The global population has been systematically moving from rural to urban areas, currently accounting for 54% of the total population, and it is expected to rise to 70% by the year 2050 (Seto and Shepherd 2009). Urban environments are often very polluted, and the latest global report of the World Health Organization (Lim et al. 2012) highlights the number in premature deaths due to air pollution. Past and more recent studies have indicated and confirmed that urbanization on its own is capable of modifying the energy balance at the surface by suppressing evaporation and increasing turbulent heat fluxes, surface heating, and 2-m air temperatures with respect to surrounding rural areas (Oke 1982; Shepherd 2005; Seto and Shepherd 2009). This condition, particular to urban areas, is known as an urban heat island (UHI), a phenomenon that has been studied over the years for many cities in the world and that has been linked to changes in meteorological conditions over and around cities.
Observational studies of temperate urban locations found that UHI intensity decreased with surface wind speed and cloud cover (e.g., Kidder and Essenwanger 1995). In fact, under clear skies and low wind speeds, the UHI intensity reached more than 6°C (Unwin 1980). Synoptic conditions also affect the intensity of the UHI under anticyclonic conditions, which can exacerbate an already large UHI effect (Unger et al. 2001). The UHI intensity also depends on the size of the population and the type and extent of the land use (Hogan and Ferrick 1998), and it is enhanced during the warm half of the year (Schmidlin 1989). The UHI effect tends to be more intense at night, but on occasions it can become negative under moist conditions and larger thermal admittances by the rural areas (Tereshchenko and Filonov 2001). Many idealized and realistic numerical studies have focused on the UHI and the interaction with the overlying atmosphere and its downwind effects (e.g., Kimura and Takahashi 1991; Baik 1992; Baik et al. 2001; Thielen et al. 2000). Perhaps the most interesting aspect of a UHI over a heavily polluted area is its ability to change the thermodynamic (surface heat fluxes) and microphysical (aerosols) state of the air mass that comes in contact with it.
Precipitation anomalies have been reported downwind of industrial sites [e.g., Changnon (1968) for Chicago], an indication that urban surface characteristics modify the stability of the overlying atmosphere. To confirm these early findings, an extensive 5-yr field campaign was performed to collect information from a dense rain gauge network in St. Louis, Missouri, a city located over a relatively flat terrain and substantial industrial activity (Changnon et al. 1991). Low-level wind measurements ahead of a precipitation event (rather than storm motion) were used to separate the downwind area into four quadrants. Changnon et al. (1991) found that during fall and spring seasons, the most affected quadrants downwind were consistent with the expected maximum in precipitation and frequency of rainfall events caused by deep and organized convection. Rainfall estimates from satellite-based measurements were used to confirm globally that precipitation perturbations occurred over and downwind of major cities (Shepherd and Burian 2003). Recent observational studies have accumulated evidence that unequivocally pointed to the ability of urban barriers to change the divergence of near-surface winds over and downwind of city centers. Bornstein and LeRoy (1990) were first to note storm bifurcation in New York City using radar measurements. A maximum of radar reflectivity over the city was associated with convection initiation, in near-calm wind conditions during summer. Later in the day, traversing thunderstorms organized into two maxima around the city edges and produced precipitation maxima downwind. The built-up urban barrier is thought to generate this divergence flow effect. More recent studies in other cities such as Phoenix (N. Selover 1997, meeting presentation), Atlanta (Bornstein and Lin 2000), Indianapolis (Niyogi et al. 2011), and Beijing (Dou et al. 2015) confirmed this phenomenon. A good review of urban induced or modified precipitation can be found in Shepherd (2005). Han et al. (2014) recently reviewed the vast literature on this subject that has accumulated over the years. They summarized the causes of precipitation perturbations downwind of urban sites as 1) existence of a UHI, 2) large roughness differences between urban and rural areas, and 3) changes in aerosol concentrations between urban and rural areas. Several modeling studies were devoted to attack these three aspects, individually or in combination. One such example was the paper by Rozoff et al. (2003), who compared results of experiments made with the Regional Atmospheric Modeling System (RAMS) to understand how the urban surface modified the initiation of convective storms downwind of the city edges. They found that the inclusion of an urban canopy model was fundamental to explain the initiation of deep convection and precipitation on the leeward side of the city and downwind of the city, respectively. Pathirana et al. (2014) studied the sensitivity to the size of the urban area and its effect on extreme precipitation and found that three out of four tropical cities responded with an increase in extreme precipitation, although the authors did not specify if intense events occurred over or downwind of the city centers. In a more realistic setting, Li et al. (2013) studied the sensitivity of urban surface representation by including three different types of urban density and the effects of aerosols on precipitation. The authors found similar sensitivities to urban representation compared to different microphysical schemes. However, they found that low-level jets were not properly represented and had a deleterious effect on the timing of precipitation initiation. The precipitation peak occurred about 4–5 h before the observed peak, and the different numerical experiments showed little variation between microphysical schemes. Nevertheless, other studies indicated that changes in the microphysical structure of convective clouds can lead to changes in precipitation timing and intensity. Rosenfeld et al. (2008) show evidence suggesting that polluted clouds can have a higher vertical development, intensifying and possibly delaying precipitation. However, these results are somewhat simplistic, since they are based on adiabatic parcels and may not represent the full complexities of convective motions. Numerical studies show that an increase in aerosol concentrations can lead to either enhancement or suppression of precipitation, seemingly dependent on aerosol loading and aerosol indirect effect (Storer et al. 2010; Ntelekos et al. 2009).
Mexico City and its surrounding suburbs constitute a megacity of about 20 million inhabitants who are regularly affected by polluting gases and aerosol particles. While air quality has improved significantly in recent years, the health standards for ozone and particulates are still often exceeded (Molina et al. 2010). Mexico City is located in a high-altitude basin (2260 m MSL) surrounded by large mountains on three of four sides. Before the arrival of Europeans to America, most of the basin was covered by lakes, which were drained during colonial times. Only a single lake currently exists and covers a very small area while the urban area has spread to almost the complete basin. Because of its tropical location (19°N), Mexico City is characterized by a wet season from June to October and a dry season from November to May. During the wet season the rain is predominantly produced by large convective clouds, with a large component of orographic forcing and modulation by tropical meso- and synoptic-scale phenomena. The sparse precipitation that falls during the dry season is frequently associated with cold fronts (Klaus et al. 1999).
The large urban growth of Mexico City has led to a well-documented UHI effect (Jauregui 1993, 1997) and precipitation in the city has also been recorded for decades (Jauregui and Romales 1996). Different trends in the monthly mean precipitation were reported at different locations within the basin between 1941 and 1985: a steady increase in the western urban station (5 mm decade−1) and no increase in the eastern station (suburban, less than 20 km east-northeast of the western station). Moreover, Jauregui and Romales (1996) presented evidence of changes in the timing of intense events (defined as >20 mm h−1) during summer precipitation at the urban site. Intense events were more frequent between 1900 local time (LT; where LT = UTC − 6 h) and midnight in the 1940s and shifted to being more frequent between noon and 1800 LT in the 1960s. Magaña et al. (2003) analyzed extreme events (as defined by their probability density functions) from 1993 to 2000 in five regions of the basin. Furthermore, Magaña et al. (2003) identified the thresholds in each region and showed that these events occurred more frequently to the west because of orographic forcing and did not relate this condition to urbanization.
In the present study, the observational analysis of Jauregui and Romales (1996) and Magaña et al. (2003) was extended to the year 2008 and evaluates the variability in the timing of the most intense precipitation [>20 mm h−1, as defined by Jauregui and Romales (1996)]. The observed changes in the frequency and timing of the intense precipitation events are evaluated by postulating two alternative hypotheses, both related to the increased urbanization and anthropogenic activities in the basin:
the changes in frequency and timing of intense precipitation events are due to changes in the microphysical properties of clouds because of enhanced aerosol concentrations and thus, enhanced cloud condensation nuclei (CCN), and
the changes in frequency and timing of intense precipitation events are due to changes in land use in the Mexico City region.
The objective of this study is to present more current observations of intense precipitation events in the basin and to test the proposed hypotheses using a mesoscale numerical model. This study aimed to provide proof of concept for the two hypotheses. The first hypothesis is addressed through a methodology that involves changing the droplet number concentration within the mesoscale model and thus modifying the development of precipitation and the timing of intense events. The second hypothesis is tested by synthetically reducing the urban area in the mesoscale model and evaluating potential changes in the timing of intense events related to changes in surface friction and momentum fluxes.
2. Methodology
a. Observational data
The dataset used in this study was obtained from the Water Authority in Mexico City (Sistema de Aguas de la Ciudad de México) for the period 1993–2008 and consisted of about 60 rain gauge stations (shown in Fig. 1) with tipping-bucket pluviometers that recorded continuously. This was the same dataset analyzed by Jauregui and Romales (1996) and Magaña et al. (2003) for earlier periods. The data were averaged to derive 1-h accumulated rainfall values per station. Figure 1 shows all the rain gauge stations analyzed and highlights three different groups of four stations considered representative of different regions in the city. The blue dots (northeast) indicate stations that are located close to the airport and at the mean basin level (2260 m MSL) and include the original suburban station used by Jauregui and Romales (1996). The red dots (west) indicate stations that are located in the western region of the city and at a few hundred meters above the basin level and also include the urban station used by Jauregui and Romales (1996). The large black dots indicate stations located where major changes in land use have occurred in the last few decades. It is important to note that none of these stations can be considered rural within the period of this study.
Geographical location of the basin where Mexico City is located (political boundaries of the capital and surrounding states are shown by black lines). The complex topography that surrounds the basin is indicated by the gray lines (contour interval = 300 m). The gray (green) area corresponds to urbanized (vegetated) area that remains constant in CTRL and LULC experiments. The yellow area is the urban class changed to vegetation for LULC experiments. Smaller regions (dashed lines) labeled as z1 and z2 indicate the regions over which calculations of the intense precipitation events were carried out. Circles represent pluviometric stations; bigger circles and colors highlight the stations used for three different sections: northeast (blue), east (black), and west (red).
Citation: Journal of Hydrometeorology 16, 4; 10.1175/JHM-D-14-0081.1
The data for the whole period (1993–2008) were divided into 5-yr subperiods to evaluate the evolution of precipitation: 1993–97, 1998–2002, and 2003–07. For further stratification of the data, only intense precipitation events of more than 20 mm h−1 were considered, since these are the events studied by Jauregui and Romales (1996) and correspond to the basinwide average intensity of extreme events determined by Magaña et al. (2003). Those are the events responsible for serious flooding of city streets and the collapse of the drainage systems.
b. Numerical model experiments
Simulations were performed with the Weather Research and Forecasting (WRF) Model, version 3.4 (Skamarock et al. 2008), for 30 consecutive days during the rainy season, corresponding to the month of September. The wet season in south-central Mexico is dominated by a bimodal precipitation distribution, with a first maximum in June, followed by a relative minimum in July and August (called the midsummer drought; Karnauskas et al. 2013, and references therein). The second peak in precipitation occurs in September. Simulations were performed for 10 different Septembers (from 2002 to 2011) to capture the interannual variability of precipitation and intense events.
WRF is a nonhydrostatic fully compressible model that uses a terrain-following vertical coordinate system. Shortwave and longwave radiation computations are performed with the Rapid Radiative Transfer Model (RRTM). A nonlocal turbulence first-order scheme developed at Yonsei University (Hong et al. 2006) is used to represent processes within the planetary boundary layer (PBL). The surface layer scheme used in conjunction with the above PBL scheme is based on the Monin–Obukhov similarity theory (Skamarock et al. 2008). The Noah land surface model (LSM; Chen and Dudhia 2001) uses four soil layers (10-, 30-, 60-, and 100-cm thickness) to predict soil moisture, soil temperature, soil water/ice, and snow cover. Land use (vegetation type), land cover (vegetation fraction), soil texture, albedo, surface roughness, and soil hydraulic properties are obtained from a lookup table and 2D fields (Chen and Dudhia 2001). The root zone of the Noah LSM is located in the first 100 cm below ground. Microphysical processes are represented by the scheme proposed by Thompson et al. (2008), which allows for the interactions of five different hydrometeor species: cloud water, rainwater, cloud ice, snow, and graupel. This microphysical scheme does not explicitly include the activation of CCN. All five species of hydrometeors in this scheme are assumed to have gamma distributions, except snow, which depends on ice water content and temperature. Each distribution has specific parameters that have been determined by empirical studies predominantly from observations in midlatitude precipitation systems (Thompson et al. 2008). In particular, the total cloud droplet number concentration is a parameter that can be changed in this scheme. To evaluate the effect of increasing CCN concentrations on precipitation development, the total cloud droplet number concentration is increased as a proxy of larger amounts of CCN present in a polluted atmosphere. The default value of the droplet number concentration was increased by a factor of 2 to synthetically represent increased activation from enhanced aerosol pollution. No other categories of hydrometeors were modified explicitly, but since precipitation development involves interactions between hydrometeors, the change in the cloud water category induces changes in all other categories.
Few studies have determined CCN concentrations in Mexico City (Montañez and García-García 1993; Baumgardner et al. 2004; Wang et al. 2010) and none of them have reported measurements during the wet season. During the polluted dry months, CCN concentrations can reach up to 4000 cm−3; such polluted events are characterized by the presence of anticyclones, subsidence, and very reduced ventilation within the basin, so that cloud formation is very much suppressed. Deep clouds develop during the wet season since radar observations indicate that precipitation echo tops can reach 15 km MSL (Novo and Raga 2013). Under such atmospheric conditions that favor deep convection, the PBL height is ill-defined and vertical transport will dilute CCN concentrations in a much larger volume than during the polluted dry season. Moreover, during the wet season, in-cloud scavenging processes play a major role in reducing CCN concentrations through cycles of activation, droplet growth, collisions, and then evaporation due to entrainment, until precipitation development. Therefore, a factor of 2 increase in the CCN considered in the polluted simulations is reasonable until measurements are available during the wet season.
Specifically, a concentration of 600 cm−3 cloud droplets [typical of clean continental cases, as considered by Pruppacher and Klett (1997)] was assumed to represent background aerosol atmospheric conditions (CTRL simulations), while a concentration of 1200 cm−3 cloud droplets represented enhanced activation due to a polluted atmosphere (EXP simulations). The value of this parameter is defined at the beginning of the simulation and remains constant throughout.
Figure 2 shows the three domains used for all experiments. The grid spacing for the larger domain (d01) is 9 km, while 3 and 1 km are used for the nested domains (d02 and d03, respectively). The model was set to include 25 vertical levels, with 10 levels within the first 1000 m above the ground. Two-way interaction among nested domains was adopted and cumulus parameterization was used only in the 9-km domain (labeled d01, see Fig. 2). The North American Regional Reanalysis (NARR; Mesinger et al. 2006) fields were used to obtain the boundary and initial conditions for the model. In this experimental design, sea surface temperature was not explicitly included and only skin temperature from the NARR dataset was used to initialize the simulations. “Nudging” of the large scales was not considered in the simulations, since the main objective is to study the sensitivity of extreme precipitation to changes in cloud droplet concentrations (as a proxy of CCN concentrations) and the change in urban area.
Computational domains for the simulations with WRF, with 9-, 3-, and 1-km (d01, d02, and d03, respectively) horizontal resolution. The dashed rectangle inside d03 represents the area shown in Fig. 1.
Citation: Journal of Hydrometeorology 16, 4; 10.1175/JHM-D-14-0081.1
To address the proof of concept for the proposed hypotheses, two subdomains within the innermost domain (d03) were chosen for the analysis of the results, as indicated in Fig. 1 by the box in dashed lines. Refer to Fig. 1 for a close-up of the spatial distribution of the topography (contours shown as thin black lines) within these subregions. These subregions are selected for an in-depth analysis, as well as to relate to the location of the rain gauge stations within Mexico City. Note that the subdomain z1 (Fig. 1) covers a large flat area of the basin, which includes most of the urban area; subdomain z2 is smaller than z1 and is located at the foothills of the western mountain range, but accounts for only a small variation in elevation. This choice of subregions is important since, as Magaña et al. (2003) pointed out, the topographic forcing of precipitation in the basin is significant and needs to be minimized in the analysis.
The U.S. Geological Survey (USGS) surface boundary conditions are the default dataset in WRF for land use and are used in the CTRL simulation. To assess the role of changes to the land use and land cover due to urbanization, the urban area in USGS was reduced synthetically by about 50% and replaced by evergreen broadleaf forest [land use–land cover (LULC) simulations] over the southern region of the basin where Mexico City is located (Fig. 1). Table 1 shows the properties of the two categories modified in this study. These artificial changes in LULC represent a smaller urban distribution that does not exactly correspond to earlier historical times and is consistent with the proof of concept design used in this study. According to the USGS, in WRF the class evergreen broadleaf forest is part of the vegetation type that surrounds Mexico City, and the area left after removing the urban area is filled with this class in the LULC simulations. Note that in these simulations the microphysical scheme were not modified, keeping it the same as in the CTRL simulation. The proposed land use can potentially alter surface energy fluxes, surface albedo, roughness length, and soil moisture, simultaneously. The LULC experiment is designed to test the model with a reduced urban area and a cleaner atmosphere, represented by a reduced droplet number concentration (as in the CTRL experiment).
Surface properties for urban and evergreen broadleaf categories used by the Noah LSM. The variable Z0 is surface roughness and the green vegetation factor is noted by GVF.
As shown in Table 1, there is a factor of 2 increase in roughness length from the urban canopy category to the evergreen broadleaf forest category, from 0.5 to 1.0 m, respectively. The changes are performed in the USGS 24 land-use lookup table provided with WRF (Chen and Dudhia 2001). Current values of surface roughness length for the urban canopy (0.5 m) are consistent with the observation that Mexico City conforms to a rather smooth surface (e.g., values less than 0.5 m), such as those used in new and established urban areas [see Gero et al. (2006), their Table 2 and references therein]. Mexico City is considered even smoother by some authors, such as Cui and de Foy (2012), who used 0.25 m.
3. Results
a. Observations
The average precipitation over the wet season was estimated using the four stations considered representative of the three different regions in the city, mentioned in section 2 and shown in Fig. 1: northeast (NE; blue dots), west (W; red dots), and east (E; large black dots). Figure 3 presents the time series from 1993 to 2008 in these three regions and clearly shows the much larger precipitation amounts in the west, related to orographic forcing. These 15 yr of data show large interannual variability but no significant trends, as opposed to the trends shown between 1945 and 1981 by Jauregui and Romales (1996) in the west station (urban) in their study.
Average accumulated rain per year for selected stations in the east (black), northeast (blue), and west (red).
Citation: Journal of Hydrometeorology 16, 4; 10.1175/JHM-D-14-0081.1
The majority of intense precipitation events occur primarily during the wet season from June to October, which is the period analyzed in this study. Figure 4 shows the changes that have occurred in the intensity of precipitation within the basin by presenting the spatial distribution of the number of events in each of the measuring stations. The size of the circles indicates the number of events with precipitation intensity larger than 20 mm h−1 (larger circles correspond to larger values), and the gray shading of the circles represents the amount of precipitation associated with those intense events (darker gray shades represent more accumulated precipitation). Each row in the figure corresponds to one of the 5-yr periods in which the observational dataset was divided, and each of the columns indicates the 6-h period in which those events occurred (the few events that occurred in the morning were not included in the analysis). The first notable result is that the period between 1200 and 1800 LT is characterized by fewer intense events than the period between 1800 and 2400 LT, indicated in Fig. 4 by the size of the circles. Moreover, the events between 1200 and 1800 LT have less accumulated precipitation, indicated by the lighter shades of gray, than in the later period. A more in-depth look at the changes in the 5-yr periods (the different rows in Fig. 4) reveals that there have been subtle but systematic changes both in the number of intense events as well as in the amount of precipitation associated with them. From 1200 to 1800 LT it is clear that the precipitation events have decreased in frequency as well as in the accumulated amount in the stations located in the western sector of the city. A somewhat different picture emerges from the changes in the spatial distribution of intense events later in the day (1800–2400 LT, right column in Fig. 4). Whereas larger amounts again are associated with events in the west, related to orographic forcing, changes in the frequency and accumulation have occurred in the center and eastern sectors of the city. These changes are opposite to those observed earlier in the day. The signal of an increase in both variables is particularly clear in the center, eastern, and southeastern sectors of the city. Table 2 shows the number of intense events detected on each group of rain gauge stations (west, east, and northeast). There, a steady decrease in the number of intense events occurred between 12 and 18 h (36% on average) with a corresponding increase between 19 and 24 h (27% on average) for the three 5-yr periods.
Spatial distribution of the number of intense precipitation events (>20 mm h−1). The size of the circles indicates the number of intense events and the gray shading corresponds to the amount of accumulated rain in those intense events. Each row corresponds to the 5-yr period in the label. Each column corresponds to a 6-h period during the day in local time (left column from 1200 to 1800 LT; right column from 1800 to 2400 LT).
Citation: Journal of Hydrometeorology 16, 4; 10.1175/JHM-D-14-0081.1
Total intense events >20 mm h−1 detected on each group of rain gauge stations: west, east, and northeast.
To further highlight differences in the observations, the number of intense events in the four stations within each of the three sectors described above (E, W, and NE) was averaged to produce hourly histograms. Figure 5 shows the histograms corresponding to the hourly average number of intense events in each sector (E, W, and NE) and for the three periods into which the whole dataset has been subdivided. The average number of intense events in the west (red curve) shows a maximum between 1900 and 2000 LT in the period 1993–97. This single maximum becomes broader and there is doubling in the number of events observed at 2100 LT from 1993–97 to 2003–07. The signal is not as clear for the two other sectors in the city, but it is still discernible. For example, the eastern stations (black curve) show a decrease in early afternoon events at 1500 and 1600 LT in the most recent period (2003–07) while showing a monotonic increase in average number of events up to 2100 LT. The NE stations also indicate mode events between 2000 and 2100 LT in the recent period 2003–07 compared to 1993–97. Note that these changes in the timing of the average number of intense events occur in the context of unmodified mean total precipitation (from June to October), as was shown in Fig. 3.
Time evolution of the average number of intense precipitation events (>20 mm h−1) for the two regions in Mexico City indicated by the colored circles in Fig. 1. The different lines correspond to different sectors within the city: black line to the east, red line to the west, and blue line to the northeast. The three 5-yr periods into which the dataset has been subdivided for analysis are shown.
Citation: Journal of Hydrometeorology 16, 4; 10.1175/JHM-D-14-0081.1
Data from the Global Precipitation Climatology Centre (GPCC) obtained for six decades from 1950 to 2009 (not shown) indicate a clear interdecadal variability in annual accumulated precipitation for the entire country. However, certain features are maintained throughout the whole 60 yr, particularly for the central plateau where the region of study is located. The north of Mexico is characterized by much less precipitation and only in the 1990–99 decade shows slightly more precipitation during the wet season compared to other decades. Unfortunately, it is not possible to evaluate the large-scale variability in the intense precipitation events (>20 mm h−1) since the global datasets do not provide this type of information. Mexico City is located at approximately 19°N and 99°W, in the region of a large gradient of precipitation within the country. Maps of precipitation indicate that not much appreciable change is observed in the study region for several decades (not shown). In the more recent past, this GPCC dataset appears to indicate a slight increase in precipitation, which is not observed in the stations within the city (see Fig. 3).
b. Simulations with WRF
1) Response to changes in droplet number concentration
Figure 6 (top) shows the spatial distribution (in d03) of the ensemble average of surface precipitation rate (mm h−1) for the 10 Septembers simulated in the CTRL experiments. As was shown in the observations, precipitation in the basin where Mexico City is located shows a large spatial gradient related to the complex topography (shown as thin black contours in Fig. 1) that surrounds the city. More intense precipitation is observed over mountain peaks and less over the central to eastern part of the basin, and the model appears to capture the general spatial pattern of the observed gradient (see Fig. 4). Figure 6 (middle) shows the relative percentage difference of monthly mean precipitation rate calculated as relative percentage difference = [(EXP − CTRL)/CTRL] × 100. The first thing to note from Fig. 6 (middle) is that most of the computational domain has negative values (blue shades) of about 20%–40%, suggesting a reduction in precipitation consistent with the proposed hypothesis related to microphysical effects, particularly in the mountainous regions where most of the precipitation falls (see Fig. 6, top). Nevertheless, a region of positive values (around 10%–20%) is clearly seen southeast of the basin. While certainly some regions within d03 exist with more intense precipitation in the EXP simulation (with enhanced droplet concentration), the domain average of the frequency distributions of precipitation intensity (not shown) shows a modest decrease in the frequency of intense events (less than 10%); this is consistent with the hypothesis of reduced precipitation under conditions of enhanced droplet concentrations. Note that over and around the main mountain ranges (particularly on the western side of the Mexico City basin) negative anomalies in average precipitation intensity (Fig. 6, middle) are produced in the context of very similar orographic lifting, as indicated by nearly equal values of vertical velocities and surface winds in both EXP and CTRL simulations (not shown). The main mechanism for decreased precipitation is the competition for water vapor between the larger number of droplets in order to reach sizes where drop–drop collisions are more efficient in the lower regions of the clouds where temperatures are above freezing. A larger fraction of supercooled droplets are lofted higher in the clouds where they can then participate in collisions with ice crystals and growing snowflakes (from ice–ice collisions) and also give rise to graupel. A decrease in the precipitation over the higher elevations in the EXP simulations leads to changes in the near-surface winds, since fewer density currents (or cold pools) would be flowing downslope under more polluted conditions. This would result in a decrease in the convergence regions where the downslope flow meets the mostly northerly flow into the basin. The reduction in the downslope flow would lead to a decrease in the deceleration of the flow into the basin, resulting in an increase contribution to the cyclonic vortex present in the southeast region of the basin (see Fig. 7, top). This mechanism will be further explained in the next subsection.
Spatial distribution in d03 of the ensemble average of precipitation. (top) Precipitation for the CTRL simulation, (middle) precipitation percentage anomaly between EXP and CTRL calculated as {[(EXP − CTRL)/CTRL] × 100}, and (bottom) precipitation percentage anomaly between LULC and CTRL calculated as {[(LULC − CTRL)/CTRL] × 100}.
Citation: Journal of Hydrometeorology 16, 4; 10.1175/JHM-D-14-0081.1
Spatial distribution in d03 of the ensemble average of wind fields. (top) Surface wind fields (vectors and colored shading; m s−1) at 10 m for CTRL experiment and terrain height (white contours; m). (bottom) Wind speed percentage differences (colored shading) between LULC and CTRL calculated as {[(LULC − CTRL)/CTRL] × 100} and terrain height (black contours; m).
Citation: Journal of Hydrometeorology 16, 4; 10.1175/JHM-D-14-0081.1
2) Response to the change in land use
Figure 6 (bottom) presents results as the relative percentage difference between CTRL and LULC simulations (calculated as in the previous section, but for the LULC) for the ensemble average of surface precipitation over 10 yr of simulations for the month of September. Note that the spatial pattern of the difference in precipitation rate shows a decrease between 10% and 20% over the high topography to the west (where most of the precipitation falls; see Fig. 6, top) and a more generalized increase in the basin, particularly in the southeast region, of between 20% and 30%.
Early in the morning, surface wind patterns (not shown) reveal a northeasterly component that impinges on the windward side of the western barriers. Northerly winds to the south and southeast of the basin are also part of the simulated circulation (not shown). Wind patterns change notably during the afternoon, particularly in the southeast region of the basin. Figure 7 (top) shows the surface wind field average between 1500 and 2200 LT for the CTRL experiment. A northeasterly wind is the dominant average feature observed in the northern region of the basin in the CTRL experiments corresponding to the month of September. At 10 m above the surface, most of the northeasterly flow that enters the basin between 1500 and 2200 LT is deflected toward the southeast by the orographic barriers that surround the basin. A cyclonic vortex is clearly seen within the basin, with its center located at about 19°16′N, 99°5′W in the ensemble average simulations.
This cyclonic circulation was first hinted at from observations by Raga and Lemoyne (1996) and shown by Doran et al. (1998) and was also evidenced in results of mesoscale simulations by Bossert (1997) and Fast and Zhong (1998) and more recently by De Foy et al. (2005, 2006) and Fast et al. (2007). The high mountains that surround the basin on three of its sides allow a predominant northerly flow through the open side into the basin and force the air into a circulation pattern that is not typically seen in other cities. Note that in the case of Mexico City, there is no clear evidence of a region that can be considered downwind of the urban area.
Changes in land use also cause distinct surface wind speed differences shown here as a relative percentage difference between the LULC and CTRL simulations in Fig. 7 (bottom). The most prominent feature is precisely seen in the region where the forest has replaced the urban area in the LULC simulation. A 25% reduction in wind speed is seen in the LULC simulation over this area close to the foothills, where an increase in latent heat flux and a decrease in sensible heat flux are also found (not shown).
The reduction in surface winds, over the region of land-use change, elicits a rather complex response in atmospheric circulation not only locally, but one that propagates over much of the entire innermost domain. Increased winds are seen feeding into the cyclonic vortex shown in the ensemble average (Fig. 7, top), located in the southeastern section of the basin. This enhanced horizontal vortex in the LULC simulation leads to changes in the spatial distribution of the vertical velocity. Updrafts at about 650 hPa (not shown) over the area with modified surface conditions in the LULC simulation are larger to the west and smaller to the east compared to the CTRL simulation. Positive differences in vertical velocities become predominant at upper levels at the basinwide scale; this is consistent with larger precipitation values in the LULC simulation (see Fig. 6, bottom) over the southeastern region, where the cyclonic vortex is found.
In the east–west direction, closer examination of the southern part of the basin reveals a spatial pattern of enhancement in surface wind speed, surrounded by decreases in wind speed to the east and to the west (Fig. 7, bottom). Thus, over the windward side of the western mountains, surface winds decelerate and reduce the vertical velocity and precipitation in the LULC simulation (Fig. 6, bottom). On the windward side of the southern mountains there is an enhancement of vertical velocity and precipitation, and over the eastern side (the lee side viewed from the adjacent valley to the east) there is a region where both deceleration and acceleration help enhance horizontal surface wind field convergence, creating larger vertical velocities and precipitation over the center of the cyclonic vortex (Fig. 6, bottom).
These results suggest that there are at least two main aspects to consider when discussing the effects of land-use changes in this very complex orographic setting. First, there is an obvious influence from the orography combined with atmospheric stability that forces part of the northerly airflow over the western mountains and deflects the majority of the flow toward the southeast, strengthening the cyclonic vortex. Second, the localized increase in surface roughness (which decreases surface wind speeds locally) helps explain a reduction in precipitation over the mountains to the west where formation of clouds occurs on the upper windward slopes. Over the southern mountains, on the other hand, the wind accelerates both to the east and to the south in connection with the afternoon cyclonic vortex. This produces the largest increase in precipitation over the southeastern region of the basin and explains the increase in vertical velocities and is consistent with the increased intense events of the precipitation observed in the eastern sector of the basin.
A conceptual view of the dynamic controls of precipitation development within the basin emerges where decelerated wind fields over the land-use change produce less vigorous ascent over the windward side of the western mountain range. This in turn produces weaker subsidence to the southeast and thus a larger probability of occurrence of ascent and cloud formation under these conditions.
3) Temporal analysis of intense events
Figure 8 (top) presents the diurnal evolution of the fraction of intense precipitation events determined within 1 h for the innermost domain (d03) for CTRL (solid line), EXP (dashed line), and LULC (dotted line) simulations. All lines correspond to the 10 Septembers ensemble average for each experiment and the gray shading depicts the intermember variability of the CTRL simulation only. The majority of simulated intense events occur after local noon, consistent with the observations presented earlier. In all cases, there is a steady increase in intense events for a couple of hours after local noon, then certain stabilization (a shoulder) for a few hours, with all lines showing a maximum later in the evening. In Fig. 8 (top), the curves for the CTRL and the EXP simulations show the same general behavior, with differences among experiments that can be of the same order of magnitude as one standard deviation in the CTRL simulations. For instance, at about 1900 LT, the CTRL curve is just 20% larger than that of the EXP simulation, while the maximum frequency in CTRL is almost the same as that of the LULC experiment at this time.
Diurnal evolution of the ensemble average fraction of intense precipitation events. (top) Intense precipitation events for d03 and three sets of experiments CTRL (solid), EXP (dashed), and LULC (dotted). (middle) As in (top), but for subdomain z1. (bottom) As in (top), but for subdomain z2. The shading indicates the variability between individual members of CTRL simulations in all plots.
Citation: Journal of Hydrometeorology 16, 4; 10.1175/JHM-D-14-0081.1
Figure 8 (middle) shows the fraction of intense precipitation events in the subdomain z1 (see Fig. 1), which encompasses the whole urban area with not much change in elevation. Figure 8 (middle) shows more intermember variability in the frequency of intense events observed in this smaller area than in d03 (Fig. 8, top). The peak observed at 1900 LT in the CTRL simulation has shifted to 1600 LT in the EXP and LULC simulations, and the “shoulder” seen in Fig. 8 (top) has now disappeared from the diurnal evolution. A hint of a secondary maximum is observed at 1900 LT for the LULC simulation but is clearly not present in the EXP simulation.
Changes similar to those seen in Fig. 8 (middle) are observed in Fig. 8 (bottom) for subdomain z2, which is much smaller than subdomain z1 (see Fig. 1), albeit with much larger interannual variability (due to the smaller area considered in z2). Both the EXP and LULC show earlier peaks in frequency than in the CTRL simulation, which is not as clearly seen in d03. It is interesting to note that in subdomain z2, the EXP simulation shows the development of a secondary maximum between 2000 and 2100 LT.
4. Discussion
The analysis of intense precipitation events observed from 1993 to 2007, in 5-yr intervals, indicates that there are marked differences in the trends depending on the area considered within the basin. The results of Jauregui and Romales (1996) for the 1940s at the western station (their Fig. 4) show a clear majority (70%) of intense events between 1900 LT and local midnight. But that pattern was reversed in the decade from 1960 to 1969, with only 10% of intense events at that period of the day. This remarkable difference in diurnal behavior was attributed by the authors to the development of a UHI. However, their results for the decade 1980–89 suggested that the intense events were increasing in frequency again in the later period of the day, accounting for 20%. To compare directly with the results presented by Jauregui and Romales (1996) in their Fig. 4, we have degraded the time resolution of observed results (in 5-yr intervals from 1993 to 2007) and combined the intense events into two 6-h segments: from local noon to 1800 LT and from 1900 LT to local midnight. Our analysis of the 15 yr after 1993 clearly indicates (not shown) that the frequency of intense events between 1900 LT and local midnight has continued to increase. The period 2003–07 shows that the majority (between 60% and 70%) of intense events is seen between 1900 LT and local midnight, constituting a remarkable change in diurnal pattern in the context of continued increase of the urban area. This diurnal pattern of intense precipitation events, regardless of the sector (W, E, and NE) considered within the urban area is comparable to the observations that Jauregui and Romales (1996) showed for the 1940s.
The hypotheses that we put forward to identify the processes potentially responsible for these changes in the diurnal trend of intense precipitation events (>20 mm h−1) were evaluated using a 10-member ensemble simulation with a mesoscale model at high resolution. The model results indicate that observations can be represented by the CTRL simulation in a semi-idealized way consistent with our proof of concept. This CTRL simulation then would correspond to the case of increased urbanization but lower droplet concentrations (as proxy for cleaner air). This is an interesting result because the urban area of Mexico City has continued to increase throughout the decades, but the pollution was at its worst during the 1980s, when changes in the formulation of gasoline were introduced (Bravo and Torres 2000; Raga et al. 2001). The government introduced some measures to curb air pollution (SEDEMA 2012), which have slowly reduced the ambient concentrations of several primary and secondary gases as well as of aerosol particles formed by gas-to-particle conversion. While this may be circumstantial, it is nevertheless worth noting.
There are two contrasting features in the spatial distribution of intense precipitation when comparing the observed periods 1993–97 and 2003–07 (see Fig. 4). The first noteworthy result is that more intense events (and more accumulated precipitation) are observed between 1200 and 1800 LT in the western sector of the basin in the earlier period 1993–97 than in 2003–07 (cf. Fig. 4, top-left and bottom-left). Second, the opposite trend is observed for intense events between 1900 and 2400 LT; for example, an increase is observed in the west and, most notably, in the southeast of the basin during the period 2003–07 compared to the earlier period 1993–97 (cf. Fig. 4, top-right and bottom-right).
Some of the modeling results presented here are consistent with the shifts in the spatial and diurnal pattern of intense events described from the observations. In particular, the ensemble average of intense precipitation events (>20 mm h−1) simulated peaks earlier in the LULC simulation (with a smaller urban area) than the CTRL simulation (with a larger urban area) by about 3 h (see Fig. 8). The changes consistent with a larger urban area result in a delay in the timing of the intense events, shifting them toward the evening.
The aerosol particles affect the production of hydrometeors within clouds in the basin and the net precipitation patterns. An experiment simulating a polluted scenario was performed by doubling the total cloud droplet concentrations (EXP), albeit keeping the same urbanized area as the CTRL simulation (large urban area). The domainwide average precipitation rate is decreased, as expected, and is consistent with the hypothesis that an enhanced droplet number concentration would lead to less precipitation, since most of the precipitation falls over the mountains and that is where the largest reductions are seen (Fig. 6, middle). However, results from this experiment indicate that intense events peak earlier in a polluted atmosphere than those in the cleaner CTRL simulation, and this result appears to be inconsistent with the hypothesis that larger droplet concentrations would result in a delay in the onset of intense precipitation. However, one possible explanation for this result may lay in the actual microphysical processes and hydrometeor interactions that give rise to precipitation in the simulations compared to the actual reality. While we do not have in situ observations (e.g., with aircraft) inside the clouds that develop in the Mexico City basin, it is very likely that most of the precipitation that reaches the surface as liquid drops originates from “cold rain” processes, which involves the solid phase. Precipitating clouds from the first exploratory study using radar observations (Novo and Raga 2013) indicate cloud tops in excess of 10–12 km, well above the freezing level (on average located at 4.9 km MSL, corresponding to 2–2.5 km above cloud base), so that the largest volume of the cloud is at temperatures below freezing and the cold rain processes dominate the production of precipitation. The radar cannot detect polarization, so hydrometeor phase cannot be determined. However, the echoes derived from radar are consistent with the presence of graupel/hail within the convective cores and not with the presence of snow within the clouds. One possible explanation for the shift toward earlier precipitation in the EXP simulation is that the increase in supercooled cloud droplets is actually helping to accelerate the production of small hail and, particularly, of snow. A larger number of cloud droplets will enhance the probability of collisions with ice crystals and freeze on contact and thus lead to a faster production of graupel and large amounts of snow. An examination of the vertical distribution of the snow mixing ratio fields in the different simulations show generalized snow in the simulated clouds, not consistent with the radar observations (Novo and Raga 2013). The parameterization of Thompson et al. (2008) used for these simulations has been “tuned” to produce the observed snow in midlatitude frontal systems and mesoscale squall lines that develop over the United States, and it probably overestimates the production of snow in tropical convection (R. Rassmussen 2012, personal communication). In this context, the simulated precipitation would be the result of melting vast quantities of snow efficiently produced from supercooled water, and a shift forward in time might be expected. This pathway would not likely be the one responsible for precipitation in the clouds that develop in the Mexico City basin, but lack of in situ and polarized radar data prevents us from confirming this statement.
Under polluted conditions, EXP simulations show a decrease in precipitation over high elevations, possibly leading to fewer density currents (or cold pools) that flow downslope into the basin. Therefore, it is expected that convergence regions would decrease where cold air meets mostly northerly flow into the basin. These conditions would favor the intensity of the cyclonic vortex in the southeast region of the basin (see Fig. 7, top). So the polluted experiment also gives rise to changes in circulation that affect the timing of precipitation, reinforcing the cyclonic vortex located in the southeast sector of the basin.
These results from the EXP simulations are consistent with the changes that Jauregui and Romales (1996) showed between the 1940s and 1960s, but not with the observational results between 1993 and 2007. However, note that the amount and type of aerosol emissions have changed dramatically from the 1980s onward in Mexico City, leading to much cleaner conditions currently than in the 1980s.
Reductions in total precipitation were observed in the EXP simulations, related to the increase in concentration of smaller droplets within clouds increasing their lifetime and delaying the production of orographically induced rainfall. In turn, this reduction in orographic precipitation results in decreased density currents and modifies the convergence regions within the basin. In the land-use simulations, the reduction of wind speed by an increase in surface roughness length of afforested area reduces vertical velocities over the western mountains facing the basin. As a result, lesser subsidence develops over the east and south regions of the basin, which allows more ascent in this region and larger production of rainfall. While the experiments were proposed to address different mechanisms, in the complex topography where Mexico City is located, it seems that the response of circulations is the one that dominates in both experiments.
5. Conclusions
In this study, we present the evolution of intense precipitation events in the large area that encompasses Mexico City and its surrounding suburbs, based on observations and numerical simulations. From observations we conclude that there has been a shift in the time of the maximum frequency of intense events in the 15 yr analyzed, and that in the more recent 5-yr period (2003–07) the timing has reverted to the situation reported by Jauregui and Romales (1996) in the 1940s when 70% of intense events occurred between 1900 LT and local midnight. Note that in the 1960s the fraction of intense events between 1900 LT and local midnight had decreased to only 10% (Jauregui and Romales 1996). That study tried to link this remarkable behavior of the timing of the intense events to the UHI effect in Mexico City, where larger convergence areas promote higher precipitation downwind of the city in the afternoon.
From simulations we conclude that increasing the droplet number concentration in the microphysical scheme (as proxy for more polluted atmosphere, EXP simulation) results in suppression of monthly mean precipitation rates over most of the computational domain, but particularly over the mountainous regions where most of the precipitation falls. Moreover, this experiment suggests that microphysical processes have the potential to shift the timing of precipitation to earlier times when a more polluted atmosphere is considered. Reducing the urban area, on the other hand, also results in a reduction of monthly mean precipitation rates, although different in magnitude and spatial pattern from those found in the EXP simulation. The change in land use produced a very complex response in atmospheric circulation within and outside of the affected region, which was clearly modified by orography. Noteworthy is the enhancement of the cyclonic vortex in the southeastern sector of the basin, which plays an important role in local ascent and in modifying precipitation. In general, the patterns of precipitation anomaly LULC minus CTRL are the result of a single perturbation performed over a small region that used to be part of the urban area.
In summary, the ensemble average of the simulations performed indicates that changes in the microphysics and land use both affect the timing of intense precipitation events: the reduction of the urban area and the increase in the number concentration of aerosols both shift the timing of intense precipitation events earlier in the afternoon with respect to the CTRL simulation. This is a remarkable result, given the context of large interannual variability of summer precipitation where the complex orography plays an important forcing role. In reality, Mexico City has experienced a large urban growth with increased pollution (although the last decades have seen a limit in ambient aerosol concentrations because of emission controls). The urban growth leads to a delay in intense precipitation events and the increased pollution leads to earlier events. While the ensemble average results of the sensitivity experiments (EXP and LULC) indicate a shift in the timing of intense precipitation events, note that analysis of results from individual years (not presented in this study) show partial or complete cancellation of these two effects. Further investigation is needed to isolate what other physical factors may also affect the development of precipitation in large urban regions. The observations seen in the context of the idealized simulations performed in this study seem to indicate that the changes in land use and land cover that delay intense events dominate over the microphysical aspects in this complex basin.
Over the years, urban areas have been growing at an explosive rate, creating larger areas of impervious materials that result in larger runoff and increased vulnerability to flash flooding. Extreme precipitation during the wet season drives the city drainage networks to their limits, decreasing response times of urban watersheds. In developing countries, research in this area is becoming a real necessity given the growth of urbanization, which will continue for any foreseeable future.
Acknowledgments
This study was partially funded by Consejo Nacional de Ciencia y Tecnología de México (CONACyT) under Grant SEP-CONACyT 154729. Carlos A. Ochoa is grateful to the Secretaría de Ciencia Tecnología e Innovación del Distrito Federal, México for the financial support. NCAR Command Language (NCL) (UCAR/NCAR/CISL/VETS 2013) was used for WRF-ARW data processing and visualization.
REFERENCES
Baik, J.-J., 1992: Response of a stably stratified atmosphere to low-level heating—An application to the heat island problem. J. Appl. Meteor., 31, 291–303, doi:10.1175/1520-0450(1992)031<0291:ROASSA>2.0.CO;2.
Baik, J.-J., Kim Y.-H. , and Chun H.-Y. , 2001: Dry and moist convection forced by an urban heat island. J. Appl. Meteor., 40, 1462–1475, doi:10.1175/1520-0450(2001)040<1462:DAMCFB>2.0.CO;2.
Baumgardner, D., Raga G. B. , and Muhlia A. , 2004: Evidence for the formation of CCN by photochemical processes in Mexico City. Atmos. Environ., 38, 357–367, doi:10.1016/j.atmosenv.2003.10.008.
Bornstein, R. D., and LeRoy G. M. , 1990: Urban barrier effects on convective and frontal thunderstorms. Extended Abstracts, Fourth Conf. on Mesoscale Processes, Boulder, CO, Amer. Meteor. Soc., 120–121.
Bornstein, R. D., and Lin Q. , 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.
Bossert, J. E., 1997: An investigation of flow regimes affecting the Mexico City region. J. Appl. Meteor., 36, 119–140, doi:10.1175/1520-0450(1997)036<0119:AIOFRA>2.0.CO;2.
Bravo, H. A., and Torres R. J. , 2000: The usefulness of air quality monitoring and air quality impact studies before the introduction of reformulated gasolines in developing countries. Mexico City, a real case study. Atmos. Environ., 34, 499–506, doi:10.1016/S1352-2310(99)00382-9.
Changnon, S. A., 1968: The La Porte anomaly: Fact or fiction? Bull. Amer. Meteor. Soc., 49, 4–11.
Changnon, S. A., Shealy R. T. , and Scott R. W. , 1991: Precipitation changes in fall, winter, and spring caused by St. Louis. J. Appl. Meteor., 30, 126–134, doi:10.1175/1520-0450(1991)030<0126:PCIFWA>2.0.CO;2.
Chen, F., and Dudhia J. , 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, doi:10.1175/1520-0493(2001)129<0569:CAALSH>2.0.CO;2.
Cui, Y. Y., and de Foy B. , 2012: Seasonal variations of the urban heat island at the surface and the near-surface and reductions due to urban vegetation in Mexico City. J. Appl. Meteor. Climatol., 51, 855–868, doi:10.1175/JAMC-D-11-0104.1.
De Foy, B., and Coauthors, 2005: Mexico City basin wind circulation during the MCMA-2003 field campaign. Atmos. Chem. Phys., 5, 2267–2288, doi:10.5194/acp-5-2267-2005.
De Foy, B., Clappier A. , Molina L. T. , and Molina M. J. , 2006: Distinct wind convergence patterns in the Mexico City basin due to the interaction of the gap winds with the synoptic flow. Atmos. Chem. Phys., 6, 1249–1265, doi:10.5194/acp-6-1249-2006.
Doran, J. C., and Coauthors, 1998: The IMADA-AVER boundary layer experiment in the Mexico City area. Bull. Amer. Meteor. Soc., 79, 2497–2508, doi:10.1175/1520-0477(1998)079<2497:TIABLE>2.0.CO;2.
Dou, J., Wang Y. , Bornstein R. , and Miao S. , 2015: Observed spatial characteristics of Beijing urban climate impacts on summer thunderstorms. J. Appl. Meteor. Climatol., 54, 94–105, doi:10.1175/JAMC-D-13-0355.1.
Fast, J. D., and Zhong S. Y. , 1998: Meteorological factors associated with inhomogeneous ozone concentrations within the Mexico City basin. J. Geophys. Res., 103, 18 927–18 946, doi:10.1029/98JD01725.
Fast, J. D., and Coauthors, 2007: A meteorological overview of the MILAGRO field campaigns. Atmos. Chem. Phys., 7, 2233–2257, doi:10.5194/acp-7-2233-2007.
Gero, A. F., Pitman A. J. , Narisma J. T. , Jacobson C. , and Pielke R. A. , 2006: The impact of land cover change on storms in the Sydney basin, Australia. Global Planet. Change, 54, 57–78, doi:10.1016/j.gloplacha.2006.05.003.
Han, J.-Y., Baik J.-J. , and Lee H. , 2014: Urban impacts on precipitation. Asia-Pac. J. Atmos. Sci., 50, 17–30, doi:10.1007/s13143-014-0016-7.
Hogan, A. W., and Ferrick M. G. , 1998: Observations in nonurban heat islands. J. Appl. Meteor., 37, 232–236, doi:10.1175/1520-0450(1998)037<0232:OINHI>2.0.CO;2.
Hong, S.-Y., Noh Y. , and Dudhia J. , 2006: A new vertical diffusion package with an explicit treatment of entrainment processes. Mon. Wea. Rev., 134, 2318–2341, doi:10.1175/MWR3199.1.
Jauregui, E., 1993: Mexico City’s urban heat island revisited. Erdkunde, 47, 185–195, doi:10.3112/erdkunde.1993.03.03.
Jauregui, E., 1997: Heat island development in Mexico City. Atmos. Environ., 31, 3821–3831, doi:10.1016/S1352-2310(97)00136-2.
Jauregui, E., and Romales E. , 1996: Urban effects on convective precipitation in Mexico City. Atmos. Environ., 30, 3383–3389, doi:10.1016/1352-2310(96)00041-6.
Karnauskas, K. B., Seager R. , Giannini A. , and Busalacchi A. J. , 2013: A simple mechanism for the climatological midsummer drought along the Pacific coast of Central America. Atmósfera, 26, 261–281, doi:10.1016/S0187-6236(13)71075-0.
Kidder, S. Q., and Essenwanger O. M. , 1995: The effect of clouds and wind on the difference in nocturnal cooling rates between urban and rural areas. J. Appl. Meteor., 34, 2440–2448, doi:10.1175/1520-0450(1995)034<2440:TEOCAW>2.0.CO;2.
Kimura, F., and Takahashi S. , 1991: The effects of land-use and anthropogenic heating on the surface temperature in the Tokyo metropolitan area: A numerical experiment. Atmos. Env., 25B, 155–164, doi:10.1016/0957-1272(91)90050-O.
Klaus, D., Jauregui E. , Poth A. , Stein G. , and Voss M. , 1999: Regular circulation structures in the tropical basin of Mexico City as a consequence of the urban heat island effect. Erdkunde, 53, 231–243, doi:10.3112/erdkunde.1999.03.04.
Li, D., Bou-Zeid E. , Baeck M. L. , Jessup S. , and Smith J. A. , 2013: Modeling land surface processes and heavy rainfall in urban environments: Sensitivity to urban surface representations. J. Hydrometeor., 14, 1098–1118, doi:10.1175/JHM-D-12-0154.1.
Lim, S. S., and Coauthors, 2012: A comparative risk assessment of burden of disease and injury attributable to 67 risk factors and risk factor clusters in 21 regions, 1990–2010: A systematic analysis for the Global Burden of Disease Study 2010. Lancet, 380, 2224–2260, doi:10.1016/S0140-6736(12)61766-8.
Magaña, V., Pérez J. , and Méndez M. , 2003: Diagnosis and prognosis of extreme precipitation events in the Mexico City basin. Geofis. Int., 42, 247–260.
Mesinger, F., and Coauthors, 2006: North American Regional Reanalysis. Bull. Amer. Meteor. Soc., 87, 343–360, doi:10.1175/BAMS-87-3-343.
Molina, L. T., and Coauthors, 2010: An overview of the MILAGRO 2006 campaign: Mexico City emissions and their transport and transformation. Atmos. Chem. Phys., 10, 8697–8760, doi:10.5194/acp-10-8697-2010.
Montañez, R. A., and García-García F. , 1993: Some urban and meteorological effects on the production of cloud condensation nuclei in Mexico City. Atmósfera, 6, 39–49.
Niyogi, D., Pyle P. , Lei M. , Arya S. P. , Kishtawal C. M. , Shepherd M. , Chen F. , and Wolfe B. , 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.
Novo, S., and Raga G. B. , 2013: The properties of convective storms in central Mexico: A radar and lightning approach. Atmósfera, 26, 461–472, doi:10.1016/S0187-6236(13)71088-9.
Ntelekos, A. A., Smith J. A. , Donner L. , Fast J. D. , Gustafson W. I. , Chapman E. G. , and Krajewski W. F. , 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.
Oke, T. R., 1982: The energetics basis of the urban heat island. Quart. J. Roy. Meteor. Soc., 108, 1–24, doi:10.1002/qj.49710845502.
Pathirana, A., Denekew H. B. , Veerbeek W. , Zevenbergen C. , and Banda T. A. , 2014: Impact of urban growth-driven land use change microclimate and extreme precipitation—A sensitivity study. Atmos. Res., 138, 59–72, doi:10.1016/j.atmosres.2013.10.005.
Pruppacher, H. R., and Klett J. D. , 1997: Microphysics of Clouds and Precipitation. Kluwer Academic, 954 pp.
Raga, G. B., and LeMoyne L. , 1996: On the nature of air pollution dynamics in Mexico City—I. Nonlinear analysis. Atmos. Environ., 30, 3987–3993, doi:10.1016/1352-2310(96)00122-7.
Raga, G. B., Baumgardner D. , Castro T. , Martínez-Arroyo A. , and Navarro-Gonzalez R. , 2001: Mexico City air quality: A qualitative review of gas and aerosol measurements (1960–2000). Atmos. Environ., 35, 4041–4058, doi:10.1016/S1352-2310(01)00157-1.
Rosenfeld, D., Lohmann U. , Raga G. B. , O’Dowd C. D. , Kulmala M. , Fuzzi S. , Reissel A. , and Andreae M. , 2008: Flood or drought: How do aerosols affect precipitation? Science, 321, 1309–1313, doi:10.1126/science.1160606.
Rozoff, M., Cotton W. , and Adegoke J. O. , 2003: Simulation of St. Louis, Missouri, land use impacts on thunderstorms. J. Appl. Meteor., 42, 716–738, doi:10.1175/1520-0450(2003)042<0716:SOSLML>2.0.CO;2.
Schmidlin, T. W., 1989: The urban heat island at Toledo, Ohio. Ohio J. Sci., 89, 38–41.
SEDEMA, 2012: Cálidad del Aire en la Ciudad de México, Informe 2011. Secretaria del Medio Ambiente del Distrito Federal, 164 pp. [Available from Dirección de Monitoreo Atmosférico, SEDEMA, Plaza de la Constitución 1, 3rd Floor, Centro, Mexico City, Mexico.]
Seto, K., and Shepherd J. M. , 2009: Global urban land-use trends and climate impacts. Curr. Opin. Environ. Sustainability, 1, 89–95, doi:10.1016/j.cosust.2009.07.012.
Shepherd, J. M., 2005: A review of current investigations of urban-induced rainfall and recommendations for the future. Earth Interact., 9, doi:10.1175/EI156.1.
Shepherd, J. M., and Burian S. J. , 2003: Detection of urban-induced rainfall anomalies in a major coastal city. Earth Interact., 7, doi:10.1175/1087-3562(2003)007<0001:DOUIRA>2.0.CO;2.
Skamarock, W., and Coauthors, 2008: A description of the Advanced Research WRF version 3. NCAR Tech. Note NCAR/TN-475+STR, 113 pp., doi:10.5065/D68S4MVH.
Storer, R. L., van den Heever S. C. , and Stephens G. L. , 2010: Modeling aerosol impacts on convection under differing storm environments. J. Atmos. Sci., 67, 3904–3915, doi:10.1175/2010JAS3363.1.
Tereshchenko, I. E., and Filonov A. E. , 2001: Air temperature fluctuations in Guadalajara, Mexico, from 1926 to 1994 in relation to urban growth. Int. J. Climatol., 21, 483–494, doi:10.1002/joc.602.
Thielen, J., Wobrock W. , Gadian A. , Mestayer P. G. , and Creutin J.-D. , 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, doi:10.1016/S0169-8095(00)00041-7.
Thompson, G., Field P. R. , Rasmussen R. M. , and Hall W. D. , 2008: Explicit forecasts of winter precipitation using an improved bulk microphysics scheme. Part II: Implementation of a new snow parameterization. Mon. Wea. Rev., 136, 5095–5115, doi:10.1175/2008MWR2387.1.
UCAR/NCAR/CISL/VETS, 2013: The NCAR Command Language (NCL). Version 6.1.2, UCAR, doi:10.5065/D6WD3XH5.
Unger, J., Sumeghy Z. , and Zoboki J. , 2001: Temperature cross-section features in an urban area. Atmos. Res., 58, 117–127, doi:10.1016/S0169-8095(01)00087-4.
Unwin, D. J., 1980: The synoptic climatology of Birmingham's heat island. Weather, 35, 43–50, doi:10.1002/j.1477-8696.1980.tb03484.x.
Wang, J., Cubison M. J. , Aiken A. C. , Jimenez J. L. , and Collins D. R. , 2010: The importance of aerosol mixing state and size-resolved composition on CCN concentration and the variation of the importance with atmospheric aging of aerosols. Atmos. Chem. Phys., 10, 7267–7283, doi:10.5194/acp-10-7267-2010.
