Table 1 lists the characteristics of the brick kilns sampled and Fig. 1 shows the kilns. A description of their operation processes as well as the sampling location for each kiln is presented in the Supplement. The MK2 kiln and the traditional-campaign kiln were measured in El Refugio, a community of brick producers located in the periphery of Leon, Guanajuato. The traditional-fixed kiln was measured in a separate community of brick producers in Abasolo, Guanajuato. Measurements took place during the dry season on 12–16 March 2013. Close collaboration with the local authorities and the brick producers' associations allowed us to establish an agreement that other kilns would not be fired during the measurement period to minimize the influence from nearby sources. The selected kilns were operated by experienced brick producers under real-world operating conditions, with fuels types and practices they commonly use.

A random sample of 60 bricks were identified, measured, and weighed before the firing took place for each kiln. At the end of the firing, these same bricks were again measured, weighed, and sent to a laboratory to test their mechanical resistance and water absorption content following the corresponding NMX-C-404-ONNCCE-2012 Mexican standard (ONNCCE, 2012). Samples of fuels and raw materials were collected before the firing to determine carbon content and heating value of combustion for the fuels. The determination was carried out with an elemental analyzer PE-2400 Series 1 and a microbalance. An acetanilide standard was used to calibrate the equipment and obtain the sample's carbon content. The heating value of combustion was determined using ASTM standards (ASTM 1995) with a Parr 1108 calorimetric pump operating with an excess of oxygen to assure complete combustion of the sample. The results of these analyses are presented in Tables S1–S3 in the Supplement. Four thermocouples were installed at each of the lower, middle, and upper levels to obtain three cross sections and to determine the temperature profile inside the kilns during their operation. These three levels were defined as follows: 0.4 m above the combustion chamber for the lower level, 0.3 m below the last layer of bricks for the upper level, and half the distance between the lower and upper levels for the middle level.

Two sampling techniques were used to obtain the emission factors of pollutants generated from the brick kilns. In the sampling probe technique, a temporary scaffolding was built on the side of the kiln for equipment and technicians, and a probe was installed on top of the kiln and connected to a sensor sampling train containing real-time sensors and filters for PM collection. This sampling technique is possible due to the relatively low velocities of the exhaust so that an isokinetic flow train is not required (Weyant et al., 2016). During the few seconds right after exiting the kiln and before they are well-mixed downwind, the emission plumes on top of the kiln can vary substantially in intensity and composition. This implies that the location of the sampling probe on top of the kiln is of key importance to the representativeness of the filter measurement. To account for this effect, the sampling probe was mounted on a rotating crane that was continuously spinning slowly on top of the kiln (see Fig. 1).

An inertial mass separator with a cut-point of 2.5 µm was used to obtain the PM2.5 fraction of PM collected on 47 mm diameter quartz filters. The PM2.5 filters were replaced approximately once an hour depending on the pressure drop on the sampler. After the samplings, the filters were thermally stabilized and sent to the laboratory for gravimetric and EC and OC composition analysis using thermal-optical transmittance (TOT) and reflectance (TOR) analysis (Chow et al., 2004) using the IMPROVE_A protocol (Chow et al., 2007). Although the EC measured by the thermal-optical methods is not technically considered as BC (Petzold et al., 2013), in this paper we refer to EC by TOR as a surrogate of BC as the light-absorbing carbon in the measured PM. Since the collection filters were heavily loaded and had homogenous deposits, analyses of anions (chlorides, nitrates, sulfates) and cations (ammonium and water-soluble sodium and potassium) by ion chromatography were performed. Laboratory analyses showed that field blank concentrations were low in relation to those in source samples, averaging < 5 % for OC and < 0.1 % for BC.

Exhaust flow in the sampling train was measured using a piston flowmeter and directed to a continuous emissions monitoring system (CEMS) with a Fourier-transform infrared spectrometer (FTIR) to measure CO2 and CO and to a flame ionization detector (FID) analyzer to measure total gaseous organic compounds (TOGs). Instrument specifications and sampling calibration protocols are described in Tables S4–S5 in the Supplement. The gaseous carbon concentration in standard conditions is used in the carbon mass balance method together with the measured carbon content of the fuels (see Tables S2–S3) to obtain fuel-based emission factors (EFfuel, g(kgfuel)-1) of a pollutant (p) emitted (Thomson et al., 2016) as shown in Eq. (). EFfuel,p=pCO2MCMCO2+COMCMCO+OC+[BC]wc In Eq. (), wc (g(kgfuel)-1) represents the measured effective fuel carbon fraction in dry basis; MC, MCO2, and MCO represent the molecular weight of carbon, CO2 and CO, respectively. Energy-based emission factors (EFenergy, gMJ-1) and brick-based emission factors (EFbrick, g(kgbrick)-1) are calculated using EFfuel, the measured effective fuel heating value in dry basis (wf, MJ (kg fuel)-1), and the specific energy consumption (SEC, MJ(kgbrick)-1), respectively, as shown in Eqs. () and (). EFenergy,p=EFfuel,pwf-1EFbrick,p=EFenergy,pSEC In Eq. (), SEC is calculated by multiplying the fuel mass consumption rate (kg (fuel day)-1) by wf and dividing by the brick production rate (kg (bricks day)-1).

The second technique used to sample the kilns was based on the tracer ratio method in which the emission rate of the targeted source is obtained by simultaneously measuring in real-time the above-background concentrations of the species of interest and of a selected gas tracer with a known release rate that is co-located at the emission source (Lamb et al., 1995). This method is based on the fundamental assumption that a relatively unreactive mixture of gases emitted from a common location experiences a quasi-perfect co-dispersion and equivalent dilution through the atmosphere. The tracer ratio method does not quantify the dispersion of air pollutants or the spatial representation of the brick kilns' emission plumes, but is used to quantify the emission rates of co-emitted pollutants from a single source. The source's emission rate (ls-1, standard conditions) can be estimated using the relationship between above-background concentrations of the species p emitted and the tracer Ct multiplied by the known tracer's release flow rate Rt (ls-1, standard conditions) as shown in Eq. (): ERp=[p][Ct]Rt. Using the scaffolding built for the measurements, the tracer was released at a constant rate close to the top of the kiln so that the kiln's emissions and the released tracer were simultaneously transported downwind and measured by the instruments onboard the Aerodyne Mobile Laboratory (AML). For this study nitrous oxide (N2O) and ethyl acetate (C4H8O2) were used as tracers for the measurement of the ER due to their low atmospheric reactivity and the ability of the AML to measure their concentrations very accurately and with high sensitivity. The tracer gas was released from a compressed gas cylinder of pure N2O located in a separate vehicle. The flow rate was controlled with an MKS Instruments mass flow controller (MFC), which was calibrated against a traceable Drycal mass flowmeter several times over the course of the measurement campaign. A 3/8′′ polyethylene tube extended from the mass flow controller to the desired location, allowing the cylinder and MFC to be located in a close but safe distance from the kiln. Mass flow rates were digitally recorded and manually logged. The C4H8O2 emission tracer was generated by bubbling air through a bottle containing the compound. While it was co-located with the known N2O emission, its direct release rate was uncertain. Thus, only the N2O tracer was used to quantify brick kiln emission rates. The C4H8O2 served as an auxiliary tracer identified when the AML was downwind of plumes from the kiln of interest, rather than from other sources in the area. Furthermore, it independently diagnosed the tracer plume characteristics directly with the instrumentation used to measure VOCs of interests, as described below.

The AML incorporates real-time data acquisition and data display capabilities so that in situ decisions by the investigators can be made to move the laboratory in and out of the emission plumes that are identified by tracer detection. This is a key element for the successful application of this technique since the dilution and advection of the kiln emissions are dictated by local meteorological conditions that can vary in short timescales. The mobile laboratory was typically positioned between 20 and 100 m from the kiln during the tracer ratio measurements. The tracer ratio method allows the unequivocal identification of emission plumes from the targeted kiln at various time periods of its operation process; this in turn allows the further application of the mass carbon method to the identified plumes following Eqs. (1–3) to obtain fuel, energy, and brick-based emission factors that can be compared to those obtained with the filter-based technique.

The instrumentation onboard the AML included a soot particle aerosol mass spectrometer (SP-AMS) developed by Aerodyne Research Inc. (Onasch et al., 2012), which measured BC and OC using laser-induced incandescence of absorbing soot particles to vaporize both the coatings and BC cores of exhaust soot particles within the ionization region of the AMS (Dallman et al., 2014). The SP-AMS also measured other inorganic PM components including nitrates, sulfates, ammonium, and chlorides corresponding to a particle size range of 50–600 nm. The SP-AMS is able to handle high concentrations of particulate matter routinely. It is frequently used in source studies where organics and black carbon may be higher than 100 µgm-3 (Massoli et al., 2012; Zavala et al., 2017). The SP-AMS acquires data in 1 s mode during which it obtains an average mass spectrum sampling from 12 to 1000 m/z. The mass spectrum is then processed and high-resolution fits are applied to peaks (e.g., C3 at m/z 36 or C3H7 at m/z 43) to distinguish between BC and organics. All fit peaks are summed, counted as a particular species, and then that species is quantified for each second. There are two vaporizers simultaneously heating particles so that gas-phase molecules are then available to ionize by reaction with electrons. The laser vaporizer heated particles with a 1064 nm laser while the conventional AMS vaporizer was also present, and after passing through the laser vaporizer the particle beam impacted the conventional vaporizer which was heated to 600 ∘C. Inorganic species such as sulfates and nitrates are not vaporized by the 1064 nm laser but were vaporized by a conventional heater. In this study, we refer to PM emission factors obtained with the AML as the sum of BC, OC and inorganic components simultaneously measured with the SP-AMS.

The AML measured N2O, CH4, C2H6, SO2, CO, and acetylene (C2H2) using tunable infrared laser differential absorption spectrometers (TILDAS); NO/NOy were measured using a Thermo Electron 42i chemiluminescent detector modified for fast-response; a LiCor 6262 nondispersive infrared (NDIR) instrument measured CO2; and a proton transfer reaction mass spectrometry (PTR-MS) using H3O+ as the ionization reagent was operated in multiple ion detection mode to measure selected VOCs (Rogers et al., 2006). Species measured with the PTR-MS included methanol, acetonitrile, acetaldehyde, acetone, benzene, toluene, acetic acid, ethyl acetate, C2-benzenes (sum of C8H10 isomers: xylenes, ethylbenzene, and benzaldehyde), and C3-benzenes (sum of C9H12 isomers and C8H8O isomers). Calibrations of these instruments were checked using certified gas standards. Other instruments onboard the mobile laboratory included a global positioning system (GPS), a sonic anemometer, and a video camera. Further details on the AML instruments' detection limits and sensitivities are presented in Table S5 of the Supplement.

The physical and chemical changes occurring in the bricks during the cooking process are associated with the burning, smoldering, and cooling stages, which are in turn determined by changes in thermal energy transfer rates within the kiln, and are closely related to the final quality of the cooked bricks. In describing the brick-cooking process, we define the burning stage as the time passed since the firing starts until the feeding of fuel is stopped, the smoldering stage as the time when the maximum temperature at the top of the kiln is reached minus the burning time, and the cooling stage as the time when the temperature at the bottom of the kiln reaches a stable minimum minus smoldering time. The temporal profiles of temperature at the lower, middle, and upper levels of the kilns and the brick-cooking stages are shown in Fig. 3. The corresponding rates of heating and cooling are obtained as the time derivatives of the temperature profiles.

The data show that the cooking of bricks results from vertical transfer of thermal energy inside the kiln starting from the beginning of the burning stage when temperatures at the bottom layers rise quickly with very high heating rates. In general, higher temperatures are reached inside the traditional-fixed kiln, followed by the traditional-campaign and the MK2 kilns. The bricks located in the middle and upper layers of the kiln start their cooking process only after sufficient thermal energy is transferred from the bottom layer. Interestingly, in the case of the MK2 and the traditional-campaign kilns this can occur during the burning stage, but for the traditional-fixed kiln the cooking of bricks at the middle and upper layers occur only during the smoldering and cooling stages.

During the burning stage at the bottom of the kiln, the heating rate is much higher and smoother in the case of the traditional-fixed kiln compared to the traditional-campaign kiln, whereas the MK2 kiln shows highly variable but overall decreasing heating rates. This critical difference in the heating process at the burning stage is likely due to the physical arrangement of bricks and the design of the kiln. The traditional-fixed kiln seems to be particularly efficient in its vertical thermal energy transfer inside the kiln as temperatures in the middle and upper levels reach similarly high values (and at comparable heating rates) as those at the bottom even after the burning stage has finished. The primary effect of the initial period of the burning stage is to remove all the remaining moisture from the bricks. At the beginning of the process this is done only at the bottom layers as temperatures do not reach high values in the middle and upper layers until much later. Once the moisture is removed and the temperatures continue rising, the carbonaceous organic material contained in the clay is removed by combustion. The raw materials for the three kilns are comparable in mass and type of clay used, but the traditional-campaign and the MK2 kilns use about 3.5 wt % of manure whereas the traditional-fixed kiln use 8 wt % of sawdust (see Table 1). These materials are additives that the brick producers use during the clay preparation process, mixing them with water and crushing them until the mixture is ready for molding. These organic additives effectively act as internal fuel during the brick-cooking process and affect the quality and mechanical condition of the bricks (Martínez and Jiménez, 2014).

As temperatures continue to rise, the hydroxyl groups that are combined with the chemical compounds forming the clay begin the process of dehydroxylation, which effectively releases water and other volatile compounds at about 450 ∘C (Osornio-Rubio et al., 2016). Figure 3 shows that the traditional-fixed kiln reaches dehydroxylation much faster than the MK2 and traditional-campaign kilns. At about 573 ∘C (Tv in Fig. 3) the silica contained in the clay changes its α-quartz structure to a β-quartz structure, effectively expanding the volume of the clay (Heaney and Veblen, 1991). If the temperatures throughout the brick are not homogenous around Tv, cracks in the brick can form due to the mechanical stress of different volume expansion (Weyant et al., 2016).

Above Tv the clay begins the actual vitrification process in which clay particles melt to form a glassy bond, ultimately giving strength to the brick during the cooling stage. Brick producers have learned by experience the importance of not extending the vitrification process more than what is needed as overheating may distort the shapes of the bricks. Similarly, if the vitrification is not achieved homogenously within the brick, the mechanical resistance and thus the quality of the final product will be smaller. The time that the bricks are exposed to temperatures above Tv is 1.9 and 1.2 times larger for the traditional-campaign kiln compared to the traditional-fixed and MK2 kilns, respectively. Therefore, of the three kilns the traditional-fixed kiln exposes the bricks to temperatures above Tv for much shorter periods of time. In addition, the time integrals of the temperature profiles above Tv of the traditional-fixed kiln are at least half the magnitude of the corresponding time integrals for the traditional-campaign and MK2 kilns, indicating that much less thermal energy is transferred inside the traditional-fixed kiln for vitrification.

The environmental performance of the brick kilns can be assessed in terms of the relative magnitude of the emission factors during the brick production process. The use of fuel-based emission factors to compare brick kilns' performance is adequate when similar fuels are used among different kilns and when bricks have similar physical characteristics. In contrast, energy-based emission factors are adequate comparison indicators when fuels types are substantially different because they take directly into account the effective heating value of the fuels employed. Brick-based emission factors are adequate comparison indicators between kilns when the mass and size of the bricks produced are substantially different. Nevertheless, regardless of the type of emission factor used, an integrated assessment of the brick kilns' performance should also incorporate other parameters, such as energy efficiency, fuel consumption, combustion efficiency, production time, and the quality of bricks produced, among others.

Figure 4 shows an inter-comparison of the relative performance of the three sampled kilns along with the specific energy consumption, fuel consumption, modified combustion efficiency, and measured brick's mechanical resistance as a surrogate for the bricks' quality. In order to compare the relative environmental performance of the three kilns, we have normalized the fuel-based emission factors with the corresponding average of the three kilns for each pollutant in Fig. 4. Regardless of the base (fuel mass, brick mass, or energy) employed, the normalization effectively allows the simultaneous comparison of emissions factors for multiple pollutants that differ by orders of magnitude while providing information on their relative magnitudes.

The results show that the traditional-fixed kiln had lower modified combustion efficiency, lower fuel (wood) consumption, and slightly higher specific energy consumption compared to the other two kilns. The results of the measured mechanical resistance of the bricks produced are shown in Table S8 of the Supplement. The traditional-fixed kiln also produced bricks with an average mechanical resistance almost half of that compared to the traditional-campaign kiln, in agreement with the much higher time integral of the temperature profile above Tv for the traditional-campaign kiln and suggesting a more efficient vitrification process. Thus, although bricks from the traditional-fixed and MK2 kilns complied with the Mexican standard, the much higher mechanical resistance in the traditional-campaign kiln indicates that its bricks were produced with higher quality.

Low combustion efficiency is related to higher pollutant emissions produced during incomplete combustion. The traditional-fixed kiln had the highest emissions factors for CO, BC, as well as CH4, C2H6, CH3OH, C2H3N, and C2H4O but emitted substantially smaller inorganic PM components. Conversely, CO, BC, and OC emission factors were much smaller for the MK2 kiln compared to the traditional-campaign and traditional-fixed kilns, but had the highest inorganic PM components. Although the latter minimally contribute in mass to the overall emissions, ionic species may be important contributors to chemical processes in the atmosphere involving wet deposition. The measured SEC values were similar for the three kilns, with 10 % variation among them, because the fuels used had similar heating values. In addition, since the combustion efficiency for the MK2 and the traditional-campaign kilns are somewhat similar in magnitude, the results indicate that the traditional-campaign kiln produced bricks of much higher quality while performing more efficiently in energy consumption and combustion efficiency than the other kilns.

Very few studies are available on the chemical characteristics of emission factors for brick kilns. Previous work by Christian et al. (2010) includes measurements of multiple gases and PM composition for three traditional-fixed kilns in Mexico that used wood waste products as fuel. Of the five types of kiln designs measured by Rajarathnam et al. (2014) and Weyant et al. (2014) in India and Vietnam, only the downdraft kiln type used wood as fuel while the rest used mostly coal. Both the zigzag and clamp kilns measured by Stockwell et al. (2016) in Nepal also used coal as fuel. Jayarathne et al. (2017) recently reported the particle-phase results of the same kilns measured by Stockwell et al. (2016). Of these studies, Stockwell et al. (2016) and Christian et al. (2010) report fuel-based energy factors whereas Rajarathnam et al. (2014) and Weyant et al. (2014) report energy-based emission factors, allowing a proper inter-comparison with our results. Tables 3 and 4 show a comparison of the energy-based and fuel-based emission factors, respectively, with those obtained in other studies.

Table 3 shows that the specific energy consumption for brick kilns using coal as fuel in the studies of Rajarathnam et al. (2014) and Weyant et al. (2014) is much smaller than for those measured in this study, due to the much higher energy density content of coal vs. wood. SO2 emission factors for coal-firing kilns are higher than those of wood-firing kilns, with coal having a larger sulfur content than wood. Nevertheless, the major difference between emissions factors among the kilns seems to be caused by the kiln design. The improved designs for the zigzag and vertical shaft kilns are related to substantially smaller emission factors than the other kilns, indicating large environmental benefits by the use of more efficient brick kiln technologies. Thus, addressing the complex economic, social, and technical barriers surrounding the adoption of more efficient technologies can produce substantial environmental and health benefits.

The emission factors in this study are closer to the values reported for the downdraft kiln by Rajarathnam et al. (2014) and Weyant et al. (2014) and to the results by Christian et al. (2010) due to similarities in kiln designs and fuels (wood) employed. However, there are differences in the emission factors that suggest substantial inter-variability of emissions even when fuels and kilns designs are similar. The average BC and OC emission factors obtained in this study for the traditional-fixed kiln of 0.54 and 0.14 g(kgfuel)-1, respectively, are within the lower range of the values reported by Christian et al. (2010), whereas the corresponding BC and OC energy-based emission factors are 2–12 times lower than those reported by Weyant et al. (2014) for the downdraft wood-fueled kiln. In the study of Weyant et al. (2014), the sample streams were diluted and cooled before measuring whereas our filter-based measurements were not diluted. Similarly, in the study of Stockwell et al. (2016) the emissions were sampled downwind of the stack after natural dilution and cooling. As gas-to-particle mass transfer processes are likely to occur under strong temperature gradients, different sampling techniques can further contribute to observed differences. In this study, condensation of emitted semi-volatile VOCs between the top of the kiln and the sampling location of the mobile laboratory downwind of the plume is possible due to the strong temperature gradient, potentially adding organic content to the measured OC. However, quantification of this effect is beyond the scope of this study.

The MK2 and the traditional-campaign kilns presented similar average MCE values (0.94–0.96) that were higher than for the traditional-fixed kiln (0.91–0.92). This is reflected in the much higher CO emission factors for the traditional-fixed kiln in comparison with the other two kilns, indicating overall smaller combustion efficiency. In our study the BC/OC ratios were 5.2, 0.9, and 3.8 for the MK2, traditional-campaign, and traditional-fixed kilns, respectively, whereas the corresponding BC/OC ratios in Christian et al. (2010) ranged from 5.29 to 8.15. Methane, methanol, and acetic acid fuel-based emission factors for the traditional-fixed kiln are 3–5, 2–8, and 5 times higher, respectively, than those reported by Christian et al. (2010). These higher emission factors are consistent with the lower average MCE of 0.910 obtained in this study compared to the average MCE of 0.968 for the traditional-fixed kiln reported by Christian et al. (2010). As a comparison, Stockwell et al. (2016) reported a much higher average MCE value of 0.994 for the zigzag coal-fueled brick kiln sampled.

Overall, the comparison of the results in this study with the available literature reports indicate that there is substantial variability among brick kiln designs and fuel types. The observed variability is also the result of the combination of materials, fuels, kiln types, and operational practices that brick producers use. However, due to the small sampling size, it is not possible to infer from the data the contribution of fuel types and kiln design to the overall variability of emissions during brick production. Therefore, although both the traditional-campaign and traditional-fixed kilns are widely used in Mexico, caution should be taken in generalizing the results to other brick production regions with different fuels and operation practices. The results from this study are not intended to provide definitive generalizations of the brick making process, but to help in understanding the effects of different kiln designs and fuels on gaseous- and particulate-phase emissions from brick kilns. Nevertheless, since the number of studies with chemical composition of brick kiln emissions is so small, the results of this study represent valuable additions to the current literature.