Mexico City and the biogeochemistry of global urbanization

Mexico City is far advanced in its urban evolution, and cities in currently developing nations may soon follow a similar course. This paper investigates the strengths and weaknesses of infrastructures for the emerging megacities. The major driving force for infrastructure change in Mexico City is concern over air quality. Air chemistry data from recent field campaigns have been used to calculate fluxes in the atmosphere of the Valley of Mexico, for compounds that are important to biogeochemistry including methane (CH4), carbon monoxide (CO), nonmethane hydrocarbons (NMHCs), ammonia (NH3), sulfur dioxide (SO2), nitrogen oxides (NOx and NOy), soot, and dust. Leakage of liquified petroleum gas approached 10% during sampling periods, and automotive pollutant sources in Mexico City were found to match those in developed cities, despite a lower vehicle-to-person ratio of 0.1. Ammonia is released primarily from residential areas, at levels sufficient to titrate pollutant acids into particles across the entire basin. Enhancements of reduced nitrogen and hydrocarbons in the vapor phase skew the distribution of NOy species towards lower average deposition velocities. Partly as a result, downwind nutrient deposition occurs on a similar scale as nitrogen fixation across Central America, and augments marine nitrate upwelling. Dust suspension from unpaved roads and from the bed of Lake Texcoco was found to be comparable to that occurring on the periphery of the Sahara, Arabian, and Gobi deserts. In addition, sodium chloride (NaCl) in the dust may support heterogeneous chlorine oxide (ClOx) chemistry. The insights from our Mexico City analysis have been tentatively applied to the upcoming urbanization of Asia.


Introduction
The Valley of Mexico contains what is arguably the world's largest metropolis with 20 million people coexisting in a space of 1000 km 2 (UN, 1992;Villareal et al., 1996). The con®ning topography of the basin contributes to pollutant loadings that are among the most severe known (Jauregui, 1971;MARI, 1994).
Recent ®eld campaigns are providing an unprece-dented characterization of the composition in the basin atmosphere (Doran et al., 1998;Edgerton et al., 1999). While Mexico City's air environment has been found to resemble peak smog years in cities of the Western United States (Seinfeld, 1989;Finlayson Pitts and Pitts, 1997), pollution extremes have been reached in the Valley of Mexico which may be more relevant to studies of`megaurban' zones of the developing world. Many new, large cities are expected to emerge across the globe in the near future. For example, whereas just 33% of Asia's population of 3 billion people currently live in urban areas, Organization for Economic Cooperation and Development (OECD) urbanization levels of 75±80% could be attained within two generations (UN, 1992(UN, , 1994. In this work, we investigate Mexico City from a biogeochemical perspective, at a time when cities are becoming increasingly important within the earth system. Emission and ventilation rates are calculated for selected biogeochemically important species (hydrocarbons, carbon monoxide, sulfur dioxide, nitrogen compounds, and particulates) and the results are interpreted in terms of regional to global biogeochemical cycling. Strengths and weaknesses of the megainfrastructure are assessed, and the results are tentatively applied to the industrialization of Asia.

Data base and manipulations
Mexico City air quality has been studied in depth by individual groups at both US and Mexican universities and atmospheric science institutions (e.g. Nickerson et al., 1992;Miranda et al., 1994;Rowland, 1995, Riveros et al., 1995;Vega et al., 1995). Several recent research campaigns in Mexico City have emphasized the coupling of complex meteorological and chemical phenomena (MARI, 1994;Doran et al., 1998;Edgerton et al., 1999). Here, we summarize chemical and meteorological results from the recent campaigns together with our estimates of the burdens, emission rates, and ventilation rates of the key chemical species in the Valley of Mexico. Measurements from the latest international campaign in March, 1997, have been supplemented by results from (1) a standard air quality monitoring system (UN, 1992;Goddard, 1996), (2) earlier measurement intensives (MARI, 1994), and (3) the work of individual Mexican and international atmospheric science groups. The key chemical and meteorological parameters were then manipulated to yield mass¯uxes through the Valley system. The analysis reported here has been restricted to the win-ter±spring season when the data are most abundant.
The data used to generate surface emission and ventilation rates in the Valley of Mexico are given in Table 1. The`remote concentrations' were drawn from the literature (e.g. Penner et al., 1991Penner et al., , 1994Jae, 1992;Schlesinger and Hartley, 1992;Benkovitz et al., 1994;Prather et al., 1995;Tegen and Fung, 1995) and are estimates of background concentrations in the local region. The`urban concentrations' are averages for the mid-afternoon, from samples that were taken over the built-up area of the city. The chemical removal of the gas phase species (`chemistry time constant') was estimated using rate constants from standard tabulations (Atkinson and Lloyd, 1984;Demore et al., 1990). For the reaction rate calculations, concentrations of the chief oxidants were ®xed at rough values that have been observed and modeled for daytime polluted atmospheres (O 3 =100 ppbv; OH=10 7 radicals cm À3 ). The`scale heights' are those for eddy removal within an analytical one-dimensional framework (Elliott et al., 1997a). Concentrations at the 2 km height (`2 km conc') were constructed from the scale heights, from full photochemical modeling (MARI, 1994;Elliott et al., 1998) and from aircraft measurements (Nickerson et al., 1992;MARI, 1994;Vidal and Raga, 1998). For the short-lived species (NO x ), horizontal distributions across the Valley (`% coverage') were estimated from surface wind speeds and the chemical loss rate. The percent areal coverage for all the longer-lived species is 100% (Table 1). The`transport time constant' is the residence time for air in the Valley; its value of 0.75 d was derived by applying ®rst order decay kinetics to the Lagrangian parcel export experiments of Fast and Zhong (1998). The overall rate of species removal (tons d À1 ) is considered to be a diurnal average of (1) the sum of chemical and meteorological processing during the daytime photoperiod, and (2) transport loss alone at night. The removal time constant (`average time constant') is simply the reciprocal of the average diurnal rate constant. Fractional ventilation from the basin (`% vented' 6 100) was characterized as the meteorological removal rate divided by the total loss for each species.
Burdens of the individual substances were computed by treating the Valley of Mexico as a 50 Â 50 Â 2 km vessel containing 750 mb pressure. The concentration of each substance was assumed to decrease linearly with height in the enclosure, unless the scale height for the species was less than 2 km (see Table 1). The 2 km basin depth was determined by the height of a chimney eect blocking upper level¯ow over the surrounding ridgetops during the daytime (Bossert, 1997). At night, purging can occur down to the average ridge altitude of about 1 km AGL. Valley-wide ventilation and chemical loss rates were calculated from the burden and generalized removal constants for each species. Top down emissions are those required to balance the Valley-wide ventilation and chemical loss. Bottom up emissions are estimated from the accumulation of each species under the low nocturnal to morning inversion, but only when the time resolution of the data is sucient. The inversion altitude was set at 200 m AGL (MARI, 1994;Elliott et al., 1997a).
The calculated emissions compared favorably with Mexican and international tabulations (PICCA, 1990;TUV, 1992;MARI, 1994), although the discrepancy between the bottom up and top down¯uxes was a factor of two in either direction. The error in the downwind deposition calculations includes error propagation and is therefore larger. We note that the averaged concentrations, burdens, and ventilation rates presented here have been constructed very indirectly, and that the chemical transformations and¯ow within the Valley have been handled using only ®rst order kinetics and simple transport models. Because the analysis was based on data primarily from surface measurements, our interpretations of Mexico City air chemistry could be signi®cantly improved in future studies by placing more emphasis on aircraft platforms and on regional chemistry/aerosol transport modeling. In addition, geostationary satellites will soon provide 10 km resolution for several components of urban euent (Fishman, 1991).

Local air chemistry
The ozone and particulate levels that were measured in Mexico City are comparable to values from the inland Los Angeles area during the peak years of the 1970s. For example, Stage 1 ozone episodes (O 3 exceeding 200 ppbv over a 1-h averaging period) were customary for many dozens of days per year in the Los Angeles of the 1970s, over much of its basin area (NRC, 1991). On heavy smog days in Los Angeles, total suspended particulate (TSP) aerosol masses greater than 100 mg m À3 were often measured. In March 1997, O 3 exceeded 150 ppb over most of the urban area in the Valley of Mexico on half of the sampling days, and particulate matter of less than 2.5 mm (PM-2.5) averaged 40 mg m À3 overall.
By analogy with megaurban zones of the developed world (see Sillman et al., 1990a,b;Russell and Odman, 1993), we expect that the Mexico City euent will be augmented by neighboring plumes. In particular, Puebla, Toluca, and Cuernavaca are nearby cities of population greater than 1 million. In air exchanging with upper levels at the 2 km chimney height, the concentrations of reactive escaping photochemicals such as NMHCs, NO y , O 3 and sulfur oxides (SO x ) were found to be orders of magnitude greater than trosphospheric background levels (Table 1). For example, at the transition to geostrophic winds, O 3 approached 300 ppb on winter afternoons (MARI, 1994;Vidal and Raga, 1998). Within the basin, NMHC losses balance O 3 production at the usual yields with a NO x eciency of 5±10 (Liu et al., 1987;NRC, 1991;Jacob et al., 1993a,b). Volatile organics remaining in the air carry ppm-level potential ozone production capacity. Nitrogen oxides and SO 2 also have well-established eects on oxidant and cloud condensation nucleus (CCN) ®elds as they dilute into remote air masses (Liu et al., 1987;Crutzen, 1988;Benkovitz et al., 1994;Schwartz and Andreae, 1996). An upper limit O 3¯u x of 5000 tons d À1 was established over the photoperiod, for the eddy ventilation piston velocity of 10 cm s À1 and over the 50 Â 50 km area. Indeed, the O 3 emanating directly from the Valley of Mexico adds 30±50 Dobsons to the tropospheric column density, and the high levels of O 3 Table 1 For selected substances within the Valley of Mexico atmosphere: concentrations (`conc.', ppb for individual gases, ppbC for NMHC, mg m À3 for ®ne particles), removal time constants (`t', days), scale distances (km), areal coverages (%), ventilation fractions (%), burdens (tons for individual gases and for particles in the ®ne mode, tons C for NMHC, tons NO for NO x , tons N for NO y ) and¯uxes (tons day À1 ). The methane burden is eective over background may be detectable in the Total Ozone Mapping Spectrometer (TOMS) data stream. Extrema in the Mexico City measurements suggest that nonstandard oxidation and removal mechanisms exist for species including NMHCs, peroxyacetyl nitrate (PAN), NH 3 , nitrates, particulates, and ClO x . Half of the NMHCs emitted within the basin are also degraded there, because of the combination of high oxidant concentrations with a relatively long threequarter day residence time for air in the Valley of Mexico. In this special, relative sense Mexico City is cleaner than counterparts such as Los Angeles, which ventilates in half a day. The local air in Mexico City is nonetheless rich in hydrocarbons. Total organic matter in the gas phase was found to average nearly 1 ppmC through the urban zone, with a morning maximum of 4 ppmC at many sites (Fig. 1). In addition, the concentrations of organic nitrates (which are hydrocarbon oxidation products) are enhanced by kinetic and mass action eects.
Total ammonia emissions within the Valley of Mexico (Table 1) appear to be comparable to those of greater Los Angeles of the 1970s, despite a dierent source distribution for the two cities. In Southern California, half of the airborne reduced nitrogen is volatilized from feedlots, with a few percent from domestic animals (Cass et al., 1982;Russell et al., 1983). By contrast, the two fractions are reversed in the Valley of Mexico, and domestic animals rather than feedlots make a larger contribution to NH 3 volatilization.
The latest NH 3 inventories in the Valley attribute 75% of the emissions to dogs, cats, and rats in residential areas, together with partially treated human waste (Osnaya . Because human and domestic animal inputs dominate the NH 3 emissions, the source of ammonia is necessarily connected with areas of intense human activity. The baseline of NH 3 lies at 20 ppb across the urban zone (Table 1), and falls o only slowly towards the foothills. The concentration of NH 3 is sucient to titrate photochemically generated acids ubiquitously into ®ne mode (PM-2.5) aerosols (Russell et al., 1983(Russell et al., , 1985. Taken together, the NMHCs and NH 3 shift the odd nitrogen (NO y ) from a traditional nitric acid (HNO 3 ) reservoir towards species that deposit more slowly, including ®ne particles and PAN (McRae and Russell, 1984;Russell et al., 1985. Peroxyacetyl nitrate concentrations regularly reached 30 ppb on the grounds of the Mexican Petroleum Institute, located in the north-center of the city (Fig. 2). PAN has exhibited rapid afternoon declines in Mexico City that may be heterogeneous in nature and may rely on surfactant coatings (Ravishankara, 1997). Such surface chemistry would return carbon and nitrogen to aerosol accessibility. Nitrates also withhold carbon from the secondary organic aerosol, which accounts for one third of the ®ne mass aerosol.
In Mexico City, PM-10 contains a high concentration of chloride, ranging from 1 to 10 mg m À3 in the Xalostoc and Nezahualcoyotl neighborhoods (Fig. 3). The bed of Lake Texcoco has tentatively been ident-i®ed as the chloride source. Dry sediment covers much of the eastern half of the Valley of Mexico, as a result of hydrological reengineering during centuries of European rule (Ezcurra, 1991). At night, NO x reacts with NaCI to give photosensitive ClNO 2 , and ultimately ClO x (Finlayson Pitts, 1993;Abbatt and Waschewsky, 1998). By contrast with marine environments, however, the Mexican chlorine chemistry will not be nitrogen limited.

Biogeochemical considerations
Several of the substances we have investigated couple major biogeochemical cycles to the Mexico City infrastructure. The cycling of substances within Mexico City has been studied often, but has not been interpreted from a biogeochemical perspective. Here, we use urban air measurements to quantify the release of pollutants from the Valley of Mexico and to investigate their in¯uence on regional and global biogeochemical cycling.
Methane (CH 4 ) is a prime example of a species whose release from Mexico City can in¯uence larger scale biogeochemical cycles. Methane is a strong greenhouse gas and it regulates the tropospheric oxidizing capacity (Cicerone and Oremland, 1988;Crutzen, 1988). The CH 4 that is emitted into the atmosphere of Mexico City is exported to the free troposphere. The metropolis of Mexico City consumes approximately 500 PJ of energy per year, of which 20% is derived from natural gas (UN, 1992;Villareal et al., 1996). Usage is 5000 metric tons of CH 4 per day. If the average of the top down and bottom up emission estimates in Table 1 were entirely attributable to losses from the natural gas infrastructure, the leak rate would be 10%. The ®gure constitutes an upper limit because the local source distribution of CH 4 (including contributions from land®lls, sewers and automobiles) has yet to be determined. It is nevertheless of interest to contrast the leak rate with loss estimates of CH 4 along its global industrial life cycle. Leakage for overall gas production is only 1% from venting and¯aring at extraction sites, and 0.1±0.3% along most pipelines (Muller, 1992). Only in former Eastern Block countries are the percentage loss ®gures believed to approach double digits. Direct human inputs of CH 4 associated with the natural gas industry (50 Mton yr À1 ) represent about a tenth of the world sources of 500 Mton yr À1 (Khalil and Rasmussen, 1983;Cicerone and Oremland, 1988). Losses from megacities that rely heavily upon natural gas thus have the potential to add signi®cantly to the planet's CH 4 inventory.
The various governments administering Mexico City have emphasized the implementation of lique®ed petroleum gas (LPG) systems in residential areas (MARI, 1994;Blake and Rowland, 1995;Villareal et al., 1996). LPG was deemed a safe domestic fuel because it can be transported by truck through the earthquake prone basin. The Mexican LPG is half butane (iso and normal; C 4 H 10 ) and half propane (C 3 H 8 ) by moles (Blake and Rowland, 1995). LPG comprises another 20% of energy use in Mexico City, with 5400 metric tons consumed daily. Source pro®ling demonstrates that LPG is the dominant propane source. By contrast, automobiles make only small contributions to the concentrations of C 3 ±C 4 alkanes. The propane emissions (Table 1) indicate that the lique®ed gases show a substantial leakage which is consistent with handling losses of almost 10%. However, because photochemistry in the Valley of Mexico is NO x -limited rather than hydrocarbon limited, minimizing the leakage of hydrocarbons from LPG will have only a minimal eect on local O 3 levels. Carbon monoxide, NO x , and a subset of the NMHCs are among the gas phase vehicular pollutants with biogeochemical signi®cance. As with O 3 and aerosols, the concentrations of CO, NO x and NMHCs in Mexico City resemble those from peak pollution years in Los Angeles. The similarity may be crudely explained by two factors: the population density of Mexico City is several times that of Los Angeles, while the auto-to-person ratio is correspondingly lower (UN, 1992;MARI, 1994). Low incomes and ecient public transportation are continuing to keep the auto-to-person ratio low.
The global nitrogen cycle is second only to the carbon cycle in terms of in¯uence on the climate (Jae, 1992). Here, we consider Mexico City in terms of its contribution toward nutrient transport to ecosystems throughout the region. In Los Angeles, nitrogen deposition is attributable mainly to HNO 3 (McRae and Russell, 1984;Russell et al., 1985). By contrast, the depositional depletion of airborne nitrogen is expected to be less in the Valley of Mexico than in Los Angeles. This is partially because of reaction of nitric acid with an ample supply of NH 3 to create ammonium nitrate aerosol. Unlike HNO 3 , ammonium nitrate does not readily deposit and thereby enables nitrogen to be transported out of the Valley of Mexico. The gas phase organic nitrates also sequester nitrogen atoms.
Ammonia injections into the Mexico basin are traceable to the local food supply. Emissions and venting (50 metric tons d À1 ; Table 1) are roughly comparable to NH 3 losses from fertilizer application in the agricultural region associated with Mexico City (Schlesinger and Hartley, 1992;Matson et al., 1997Matson et al., , 1998. Because local and international agronomists are working to cut production costs by maximizing fertilizer use eciency during intensive wheat farming in the central and northern portions of the country (Matson et al., 1998), the urban volatilization pathway will increase in relative magnitude. The urban waste treatment infrastructure in Mexico City appears to lie in between developed and rural analogs in its eectiveness. For example, waste treatment in Mexico is less eective than in US cities, but NH 3¯u xes are nevertheless low compared to the emissions from rural areas (Dentener and Crutzen, 1994).
Oxidized nitrogen emissions from Mexico City ®gure quite dierently from ammonia/reduced nitrogen in regional to global budgets. Even at the low urban vehicle-to-person ratio in Mexico City, the per capita release of NO y surpasses the global average (Jae, 1992;Prather et al., 1995). The NO y concentration in the basin ventilate (2 km height) is ®ve times higher than that of NH 3 (Table 1). During the daytime, the NO y concentration is about 35 ppbv at the transition into the upper level¯ow. The mobilization of nitrogen in Mexico City is thus leading to signi®cant export from the Valley. Nearby cities also contribute to a plume of urban euent which is perhaps 100 km in breadth and dilutes slowly in the horizontal (Giord, 1982). Given standard tropospheric eddy diusivities (Thompson and Cicerone, 1982;Liu et al., 1984), 20 ppb of NO y may contact the surface downwind. At a generalized deposition velocity of 1 cm s À1 (McRae and Russell, 1984), 5 Â 10 11 N atoms cm À2 s À1 could enter regional soils. By contrast, nitrogen ®xation in central Mexico is on the order of 1 Â 10 11 N atoms cm À2 s À1 (Schlesinger, 1997), and the sum of background NH 3 and NO y deposition is similar (Penner et al., 1991;Schlesinger and Hartley, 1992;Dentener and Crutzen, 1994). Our simple analysis of the Mexico City plume chemistry suggests that it is likely to alter regional soil nitrogen cycling. The impact of NO y on regional soil chemistry will become increasingly important in the future if the vehicle-to-person ratio in Mexico City increases.
The marine eects of Valley ventilation are dicult to gauge because of the intervening overland distance between Mexico City and the ocean. The Caribbean Sea behaves oligotrophically (Berger and Wefer, 1991), and at oceanic Red®eld ratios (Red®eld, 1963;Takahashi et al., 1985) 1 Â 10 11 new N cm À2 s À1 enter the euphotic zone from deeper waters below. The plume emanating from the urban highlands of Mexico could well augment the input of nitrogen into the Caribbean. The Gulf of Mexico is ®ve times more productive than the Caribbean (Berger and Wefer, 1991), and it often lies along the dispersion trajectory from Mexico City (Bossert, 1997). As a result, nitrogen emissions from the Valley of Mexico may also impact the Gulf of Mexico.
Primary aerosol has been inventoried only for the ®ne mode (PM-2.5) in Table 1, in order to facilitate comparisons with the ammonium salts and organics. For dust, the PM-10 mass concentration is 5 times higher than the ®ne mode dust value that is shown in Table 1. A mode radius of 4 mm is applicable for PM-10 dust (Tegen and Fung, 1995), and sedimentation is minimal. The PM-10 dust may be considered inert within the basin, and an emission of 0.015 kg m À2 yr À1 is calculated. This emission rate is comparable to (1) natural suspension from the Mexican plateau, (2) contributions from local land use change, and (3) peripheral suspension from the belt of dry soils running from the Sahara through the Middle East to Central Asia (Penner et al., 1994;Tegen and Fung, 1995). In Mexico City, the source of the PM-10 dust may be purely mechanical as a result of urbanization. New neighborhoods regularly arise in the form of squatter communities lacking in government-sponsored roads or sanitation (Frieden, 1965;Varley, 1987). Throughways in Xalostoc and Nezahualcoyotl have remained unpaved as township populations move into the millions. These suburbs record TSP levels in the hundreds of mg m À3 (Jauregui, 1989;UN, 1992;Villareal et al., 1996). Eects of the dry lake bed which supplies ClO x have yet to be quanti®ed. The Mexico City atmosphere is quite basic so that an assessment of metal solubilization may be needed. For example, much of the literature on photodissolution of iron from dust assumes an acid coating (Zhuang et al., 1992a,b). Soot particles are con®ned to the ®ne mode (Penner et al., 1994;Cooke and Wilson, 1996), and concentrations in the Valley and its euent are sucient to lower extinction lengths in the visible range. Indeed, black carbon and the urban haze in general already compromise the strength of the Mexican tourist industry (Bravo et al., 1988;Vasilyev et al., 1995).
There are several areas in which progress has been made in controlling pollution in the Valley of Mexico. Sources of sulfur dioxide and of lique®ed petroleum gas leaks have been identi®ed, and removal/rehabilitation eorts have been made (MARI, 1994;Blake and Rowland, 1995). A strong motivator for controlling pollutant levels lies outside any earth system issue, and may be as basic as visibility: the perception of haze may be the most likely motivating factor for the tightening of urban infrastructures. Indeed, climatologists have pointed out that the global aerosol may be controlled early simply because it is tangible and¯eeting (Rind, 1998).

The urbanization of Asia
We now discuss the implications of these ®ndings in a global context. The next burst of urbanization is predicted in Eastern Asia and is expected to involve almost half of the earth's population (UN, 1992(UN, , 1994OSTP, 1997). In the People's Republic of China (PRC), a strong intranational economic gradient is driving a migration towards the coast. As urban incomes rise along the seaboard, the demand for readily available coal is declining while imports of cleaner petroleum gases rise. Increased global anthropogenic inputs of CH 4 and certain NMHCs could occur if the 10% leakages of LPG that were estimated in Mexico City applied to commercial gas infrastructures across coastal Asian cities.
The Chinese vehicle¯eet is expanding exponentially, particularly in the cities (WB and CMC, 1994;Sathaye et al., 1994;Elliott et al., 1997b). Vehicle-to-person ratios are currently negligible but are projected to reach the Mexican value of 0.1 in 2020, based on estimates of gross domestic product (GDP) evolution within China. The release of combustion products will rise accordingly, and dust suspension may occur in dry areas of the coastal plains, as has been found in Western China . Not only in Latin America but globally as well, automobiles have typically entered new and developing markets before highways can be built to accommodate them (Sathaye et al., 1994;Smith et al., 1994), thereby leading to increased levels of dust suspension. The Chinese atmosphere of the next century is likely to be rich in both ammonia and acids, so that the pH of its aerosol cannot easily be predicted.
In the Valley of Mexico, the interplay of gases in a pollutant-rich atmosphere stabilizes against nitrogen deposition, thereby facilitating long-range transport. As megacities in China begin to mobilize nitrogen on an unprecedented scale, Asian ecosystems are also expected to be placed at risk (Galloway, 1996). Managed ecosystems of Asia must feed the coming megapopulation, and it is clear that they will be severely taxed. While harvests must increment at 1% per year to keep pace with a growing population, leaf damage by O 3 and SO 2 is likely to lower crop yields on the order of 10% (Chameides et al., 1994). Further, aerosols may be suciently re¯ective and adsorptive to limit crop growth by aecting photosynthetically available radiation Zhou et al., 1994). Clearly, the magnitude of these problems will be largest in the vicinity of the Asian population corridor. Because the economic gradient leads eastward in the PRC, regions such as Japan, the North Paci®c, and even Western North America lie downwind and also may be aected by pollution originating from Eastern China (Elliott et al., 1997b,c).
As the urbanization of the PRC progresses, China may leapfrog certain aspects of Western development. For example, Chinese society has traditionally relied on the bicycle-rail combination for mass transit, and still has the option to retain the strategy (Zhihao, 1990). Similarly, the ecient metro of Mexico City has been credited in part for stabilizing its auto-to-person quotient. However, pressure for a rapid update of the transportation system in China is tremendous, and Japanese/Western auto manufacturers are aggressively seeking to ®ll the vacuum (WB and CMC, 1994;Elliott et al., 1997b,c). By contrast, because returns in oil exploration are diminishing, the total recoverable reserves may be insucient to supply a large Asian eet of vehicles (Kerr, 1998).

Summary
The Valley of Mexico is arguably the world's most populated and heavily polluted urban zone, and it may serve as a partial model for the growth of megacities in other areas of the developing world. The air chemistry of Mexico City is relatively well studied, and we have used local air quality surveys to estimate the emissions, processing, and ventilation of major carrier species.
The air in the Valley of Mexico resembles peak smog episodes in cities of the Western United States during the 1970s. However, the Mexico City pollution is considered to be extreme in several senses. Leakage rates of 10% were calculated for the NMHC components of LPG. Losses may also be high from the natural gas system, but individual CH 4 sources need to be quanti®ed for the situation to be properly assessed. Automotive emissions of CO, NO x , and some NMHCs are already comparable to the most polluted parts of the US and Europe, despite a lower vehicleto-person ratio in Mexico City. Further, the emissions will rise if the vehicle¯eet expands. The release of ammonia in the Valley of Mexico occurs from disperse sources because domestic animals make a much larger contribution to NH 3 volatilization than emissions from feedlots. An excess of NH 3 forces photochemically generated acids into ®ne mode particles. Together, the NMHCs and NH 3 shift the nitrogen reservoir (NO y ) towards species which deposit more slowly, such as ®ne particles and PAN. Partly as a result, there is a mobilization of the urban nitrogen whereby the concentration of NO y at 2 km is about 35 ppb (vs <1 ppb in remote areas), and signi®cant export from the Valley occurs. Finally, chloride originating from dry sediments of the former Lake Texcoco coexists with concentrations of 100 ppb of NO y in Mexico City, raising the possibility of inland ClO x photochemistry.
The export of pollutants from the Valley of Mexico is expected to impact local, regional and global biogeochemical cycles. If the LPG leakage rates in Mexico City apply to commercial gas infrastructures in newer megacities of the developing world, the next generation megacities taken as a group may add signi®cantly to the planet's CH 4 inventory. In addition, O 3 emanating directly from the Valley of Mexico adds 30±50 Dobsons to the tropospheric column density, and the enhancement may be detectable in the TOMS data stream. The mobilization of oxidized nitrogen may lead to a downwind deposition that is several times larger than nitrogen ®xation or the background atmospheric source in Central America. Emissions from Mexico City are therefore likely to aect regional soil nitrogen cycling. Similarly, sizeable nitrogen deposition into the Caribbean Sea and Gulf of Mexico appears likely. Massive mineral dust suspension from the Val-ley¯oor is a result of reliance on unpaved roads in suburbs such as Xalostoc and Nezahualcoyotyl. The dust generation on an areal basis is similar to rates on the edges of the Sahara, Arabian, and Gobi deserts.
Over the next few generations, Asia is expected to dominate the human dimension of geochemical change. Rising incomes in the coastal zone of China will lead to the demand for fossil fuels that are cleaner than coal. The transport and use of natural and lique-®ed petroleum gases is therefore likely to occur, though the leak rates in the Valley of Mexico indicate that caution should be applied. The Asian vehicle-toperson quotient is rising rapidly and should reach 0.1 within a few decades. As a result, per capita NMHC and NO x emissions may surpass Latin American levels if vehicle maintenance is not established as a priority. Nitrogen mobilization from Eastern China is expected to impact both continental and oceanic ecosystems that lie downwind in the North Paci®c (Japan, the central gyre, and North America). The recent economic diculties in Asia may lead its next generation of megacities along a high-impact biogeochemical pathway, similar to that of Mexico City. However, it is also possible that Asia will improve on certain aspects of Western city design by centralizing energy consumption and minimizing reliance on the automobile. As in Mexico City, infrastructure improvements in Asia will likely be driven by popular concerns over health and visibility.
Any success in our approach as it has been applied to Mexico City is due in part to the very basin which creates the local pollution problem. Analyses such as the one oered here for Mexico City may be successfully conducted in other such metropolitan areas, and the approach could be extended to create a statistically useful earth system data base. For example, populous, con®ned cities are found throughout Latin America and Asia, and include Sao Paolo (Brazil) and Chongqinq (PRC). Further, the infrastructure deductions we compile may apply across the developing world, regardless of topography. Coordination with local air quality studies is recommended as a way to bootstrap biogeochemical research to other developing megacities.
collaborators researching aspects of global biogeochemistry. Related projects include simulation of marine ecodynamics in eddy resolving ocean circulation models, and analysis of new strategies for the capture and sequestration of greenhouse gases. Elliott, Dubey and McNair have backgrounds in air chemistry and simulations thereof. Bossert is a meteorologist specializing in microscale to mesoscale atmospheric modeling and simulation. F. A. Smith is an ecologist in the University of New Mexico biology department with an interest in studying the dynamics of the global city. Smith is originally a mammalogist and has published high pro-®le research on the Pleistocene paleoecology of the American Southwest. She is presently an ocer of the American Society of Mammalogists and a grantee of the National Center for Ecological Analysis and Synthesis.