The Impact of Hobby Rocket Motors on Atmospheric Fine Particle Zinc Concentrations

Zinc concentrations in hobby rocket exhaust aerosols were measured by ICP-MS to be approximately 300 mg/g, revealing that hobby rocket motors emit particulate matter highly enriched in zinc. Zinc is a commonly used indicator species in receptor models for atmospheric particulate matter and is assumed to be emitted only by well-established point sources and mobile sources. The potential impact of the ephemeral particle-bound zinc from hobby rockets on atmospheric particulate zinc concentrations was estimated using a Gaussian puff model. The results from the model were compared to 24-hour averaged particulate zinc concentration data collected in three US urban centers. Potential impacts several kilometers downwind of the rocket launch site were found to be significant. Clearly the impact of ephemeral and unregistered emissions such as those from hobby rockets needs to be considered when using zinc in source apportionment models.


INTRODUCTION
Trace metals are widely used as source indicators for atmospheric particulate matter in receptor modeling studies (Chueinta et al., 2000, Ramadan et al., 2000, Song et al., 2001, Shah et al., 2004, Zhou et al., 2004, Bennet et al., 2005, Lough et al., 2005, Schauer et al., 2006).In most cases, the sources which are considered in a model are limited to well-established stationary or mobile sources that are known to emit large volumes of particles of a previously determined trace metal composition.It is often assumed that sources that emit smaller masses of particles, and which may also be ephemeral in nature, are not sufficiently enriched in the indicator species of interest to significantly contribute to the chemistry of particles collected at a receptor.
Zinc is widely used as an indicator species for aerosol emissions from smelters, municipal waste incinerators, automobiles and motorcycles (Chueinta et al., 2000, Ramadan et al., 2000, Song et al., 2001, Shah et al., 2004, Zhou et al., 2004, Bennet et al., 2005, Lough et al., 2005, Schauer et al., 2006).Hobby rocket fuel is predominantly made of zinc and ammonium perchlorate (NH 4 ClO 4 ).Therefore, it is reasonable to assume that the particulate matter emitted by hobby rockets is likely to be greatly enriched in zinc and chloride.Emissions of particles that are greatly enriched in zinc can significantly impact 24-hour averaged PM 2.5 (particles less than 2.5 μm in diameter) and PM 10 (particles less than 10 μm in diameter) zinc concentrations without greatly impacting total particle mass.As is the case with many small point sources, hobby rockets are not expected to contribute significantly to the zinc content of atmospheric particles on a regional scale, but may significantly impact a sampler that is downwind of a location where rockets are being launched.Size distributions of hobby rocket particle emissions have not previously been characterized.This information is necessary for determining the exact influence of the emitted particles on PM 2.5 and PM 10 concentrations.
The aim of this study was to reinforce the idea that sources emitting relatively small masses of particles highly enriched in an indicator species may bias receptor models.The size distribution and chemical fingerprint data of the particle emissions from hobby rockets were measured and these results were used to estimate the potential impact of hobby rockets on 24-hour averaged zinc concentrations in atmospheric particulate matter at three US urban centers.

METHODOLOGY
Four types of hobby rocket motors were discharged into a residence chamber and samples of the diluted exhaust particulate matter were collected on filters.The specific models of rocket used for the experiments are shown in Table 1 along with manufacturer and number of motors in each composite.All are small enough not to be subject to FAA or other governmental restrictions, so are freely available to the general public on the open market.
The residence chamber used for sampling and collection has previously been described in detail (Hildemann et al., 1989, Hildemann et al., 1991a, b) and will only be briefly discussed here.A schematic of the apparatus is presented in Fig. 1.The residence chamber was an upright cylinder made of stainless steel, 1.5 m in height and 0.25 m in internal diameter.The rocket motors were discharged into the top of the residence chamber through an access port in the side wall.In some experiments, several motors were discharged into the residence chamber at once to provide a composite sample.The particulate matter in the residence chamber was then diluted with ambient air in order to permit filter collection of the particulate matter.Composite samples were made for several tests to increase the mass of diluted particulate matter collected by the sampling trains.The diluted exhaust aerosols were sampled using two PM 2.5 cyclones (Thermo-Andersen Instruments) and two Micro-Orifice Uniform Deposit Impactors (MOUDI; MSP Corporation, Minneapolis, MN).The PM 2.5 cyclones and MOUDIs were operated at 8.0 liters per minute (L/min) and 30 L/min, respectively, using downstream vacuum pumps controlled by critical orifices.The chamber was backfilled with room air.Hills, NY) were used in the other.Each MOUDI contains 12 stages that provide cutpoints ranging from 56 nm to 18 µm, along with an after filter to collect particles with aerodynamic diameters < 56 nm.In our analyses, we considered only the substrates that collected particles between < 56 nm and 5 μm in order to make a direct mass and chemical comparison with the samples collected by the PM 2.5 cyclones.
The filters were analyzed for: i) total mass; ii) ions such as chloride, sulfate, nitrate and ammonium; iii) trace metals, and; iv) elemental and organic carbon.Fig. 1 shows where in the sampling train each type of filter was used, and what analyses were performed on them.Teflon substrates were used in both PM 2.5 cyclone # 1 and MOUDI # 1 to measure the total and sizedependent mass and trace metal characteristics of the particles.Quartz-fiber filters were used in PM 2.5 Cyclone # 2 to determine the mass, ion (Cl -, SO 4 2-, NO 3 -and NH 4 + ), elemental carbon and organic carbon content of the particles (Schauer et al., 2003, Lough et al., 2005).Aluminum foil substrates were used in MOUDI # 2 to determine size-dependent ion contents of the particles.
Loading blanks were performed for all of the PM 2.5 cyclone filter holders and all of the stages of both MOUDIs, which were subsequently used in blank correcting the data.
All substrates were cleaned before sampling.Teflon substrates were leached with trace metal grade 2N HCl and 2N HNO 3 by sequentially drawing the acids through a filter holder.The substrates were then washed with 18MΩ (Milli-Q) deionized water, then finally air-dried in a HEPA-filtered laminar flow hood.The cleaned Teflon substrates were transported and stored in acid leached polystyrene Petri dishes.Once loaded with the clean substrates the Petri dishes were tripled bagged and handled using trace metal clean handling techniques.All stages of the cleaning processes for the trace metal substrates and Petri dishes were conducted in a trace metals clean room.Quartz substrates were used for total mass, ion and EC/OC analyses.To prepare them for use, the substrates were baked for 12 hours at 550°C and stored in polystyrene Petri dishes lined with foil also baked to 550°C.Aluminum foils used for mass and ion analyses were baked for 12 hours at 550°C before use.All Teflon, quartz and aluminum substrates designated for mass analyses were preweighed after being cleaned.
The metal compositions of the particles were determined by Inductively Coupled Plasma Mass Spectrometry (ICP-MS, PQ Excell, ThermoElemental).Particulate matter deposited on the Teflon substrates from Cyclone # 1 and MOUD1 # 1 was dissolved by microwave-assisted acid digestion (HCl, HNO 3 , HF) using techniques developed for the analysis of trace metals in ambient aerosol samples.These techniques permit sample solubilization with minimal contamination (Schauer et al., 2006, Lough et al., 2005).The digestates were analyzed for a large suite of metals (refer to Supporting Information), although only Zn, Na, Ca, Mg, Al, K, Fe, Cr, Ni, Cu, Ti, Pb, Mn, Se, Pd, Cd, Zr, Ag and V were detected.Most elements were measured using standard plasma conditions (1350 watts forward power) and a desolvating microconcentric nebulizer (MCN6000, CETAC Technologies Inc.).Lighter elements (Na, K, Ca, Al, Mg and Fe) are impacted by polyatomic interferences in standard plasma mode and were therefore measured under cool plasma/shielded torch conditions.Propagated uncertainties from the ICP-MS analysis were calculated from the standard deviation of field blank measurements and standard deviation of replicate analyses of Standard Reference Materials (SRMs, NIST).To verify metal recoveries during analyses, 3 different SRMs were digested and analyzed in duplicate to give 6 recovery standards with every batch of 25 samples.
Mass analyses of samples on preweighed filters were performed using a microbalance (Metler MX-5) in a temperature and humidity controlled room following a 24-hour equilibration period.
The average measurement precision was approximately 9%.
Sulfate, nitrate, chloride and ammonium were analyzed by ion chromatography (IC) after a water extraction of the quartz and foil substrates.
Elemental and organic carbon (EC/OC) were measured with a thermo-optical analyzer (Sunset Laboratory, Inc., Forest Grove, OR) using the NIOSH 5040 method (Schauer et al., 2003).
Mass and chemical composition data for particles collected on the PM 2.5 cyclone filters are available for all of the experiments performed, while mass and chemical data from particles collected using the MOUDIs are restricted to experiments 6-11 and 8, respectively.All of the analyses were performed at the Wisconsin State Laboratory of Hygiene, Madison, WI.
To determine whether or not hobby rockets could cause a measurable interference in an atmospheric particle sampler we used a Gaussian puff model to simulate the emissions from a hobby rocket launch and calculated the mass of particulate zinc that would be deposited on an air sampling filter during a 24-hour collection cycle.The result from this calculation was compared with zinc contents of urban PM 10 collected in Milwaukee, WI, Waukesha, WI, and Denver, CO (Lough, 2004).The Milwaukee sampling site (MKE) was located on top of the 16 th Street Community Center, a three-story building close to downtown on the corner of 16 th and Greenfield streets.The area is mixed residential and business district with constant road traffic.
The Waukesha sampling site (WKA) was located 20 km west of the Milwaukee site on Cleveland Street close to the four-lane suburban highway 18/JJ.The sampler was located on a platform elevated 1 m above the ground.The area is mixed residential and industry with the sampling site located right next to a vehicle body repair workshop.The PM 10 emissions of zinc listed for the Milwaukee, WI-Chicago, IL corridor in the EPA Toxics Release Inventory (TRI) were dominated by emissions from non-ferrous smelting and steel mills, but also included smaller zinc emissions from electric power stations, metal plating and coating businesses, and lubricant manufacturers (USEPA, 2001).
The Denver sampling site (DEN) was located on a two-story building in a mixed residential and industrial area of central Adams County.The PM 10 zinc emissions listed in the TRI for the Denver area were from electric power stations and metal plating facilities (USEPA, 2001).In addition, the Denver site was located near road traffic.Both cities are considered representative of urban centers with industrial areas and have high traffic densities Samples were collected using 24-hour sampling cycles combined with the EPA 6 th day sampling schedule for one year, providing approximately 50 samples from each site.The average zinc concentrations in PM 10 at the MKE, WKA and DEN sites were 0.018 ± 0.015 μg/m 3 , 0.022 ± 0.022 μg/m 3 and 0.033 ± 0.03 μg/m 3 , respectively (Lough, 2004).
A reflective Gaussian puff model was used to assess the zinc interference from rocket launches.A reflective, rather than an absorbing, model was chosen, as it was believed that the dry deposition of the majority of the particles detected in the exhaust aerosol would be minimal over the advection distances of interest.To test this hypothesis, the ratio of the mean particle diameter deposition velocity to the eddy correlation coefficient, described by Eqs. ( 1) and ( 2), was calculated (Seinfeld and Pandis, 1998).Therefore, the transport of all of the particles in the puff could be approximated well by a reflecting puff model.The reflecting Gaussian puff model is presented in Eq. ( 3) (Seinfeld and Pandis, 1998): where C is the concentration of zinc in kg/m 3 , q is the emission rate of zinc from the rocket in kg/s, u is the wind speed in m/s, and σ x , σ y and σ z are the Pasquill-Gifford dispersion parameters in the x, y and z axes.The coordinates x, y and z describe the center of the puff at emission when t = 0, whereas x', y' and z' are the puff coordinates of the puff center sometime after emission (i.e. when t = t').
The modeling was performed using the Aerotech F20-7W White Lightning motor (Econojet, 2004) for which an emission rate had to be determined.This was achieved by combining information about the propellant formulation, assumptions regarding the conversion of zinc to particulate matter and assuming that the entire exhaust plume left by the rocket could be reasonably modeled as a point source emitted at the midpoint of the exhaust trail.The height of the exhaust plume was calculated to be 100 m using the thrust curve and rocket mass provided by Aerotech.The Aerotech F20-7W White Lightning motor had a propellant load of 0.03 kg, 7% (2.1 g) of which was believed to be zinc (Stuart Davies, Michigan Tripoli Rocketry Association, pers. comm., 2005).Fuel formulations are proprietary so it was not possible to obtain the exact zinc content in the motors used.Our data indicate that the vast majority of the emitted particle mass is composed of the zinc and chloride from the ammonium perchlorate.Assuming that all of the zinc is converted to particles, only 26% of the entire fuel load forms particles in the rocket exhaust.
A neutral atmospheric lapse (stability class D) was used for the base case.The sampler was assigned an inlet altitude of 5 m to simulate a 2-m tall sampler on a raised platform.The exhaust plume is modeled as a puff assuming that negligible amounts of the PM will deposit on the ground and obstacles between the launch site and sampler.Annual plots of daily average wind speeds for Milwaukee (thus Waukesha also) over the past 3 years were used to determine an annual average wind speed of approximately 5 m/s (GLERL, 2005).The annual average wind speeds in Denver, CO were determined to be approximately 4 m/s (NCDC, 2001).Therefore, a wind speed of 5 m/s was used to represent all three urban environments.
The particulate zinc concentrations at the puff center were determined and converted into masses of zinc that would be collected on the filter in the sampler at the different distances from the launch site.To do this, the time of exposure of the sampler to the puff was calculated from the wind speed and puff dispersion along the direction of the puff travel, σ x (units of m).

RESULTS AND DISCUSSION
A comparison of the zinc, chloride, ammonium and total PM 2.5 masses determined from the MOUDI1 and PM 2.5 cyclones revealed no statistical differences, indicating good agreement between the two techniques.
The bulk chemistries of the exhaust particles from the White Lightening, Blue Thunder, Black Jack and Estes rocket motors are shown in Fig. 2a.The compositions of the White Lightning, Blue Thunder and Black Jack rockets were dominated by zinc and chloride with mass concentrations of 250-300 mg of Zn per gram of particle mass and approximately 200-300 mg of chloride per gram of particle mass.The propellant compositions for these rockets were assumed to be a mixtures of zinc, ammonium perchlorate and a polymeric binder, analogous to the aluminum, ammonium perchlorate and polymeric binder fuel used in the commercial solid rocket motors, SRMs (Rutter, 2002).The chemistry of the exhaust particles generated by the Estes motor was dominated by potassium and sulfate, with mass concentrations of approximately 300 mg/g and 500 mg/g respectively.These combustion products imply that fuel for the Estes rocket was black powder (carbon, potassium nitrate, and sulfur).Fireworks typically use black powder fuel rather than a fuel based on zinc.Zinc emissions from black powder fuels are demonstrated by the Estes rocket (see Fig. 2a) to be negligible.Previous publications have measured particles emitted from fireworks and did not report Zn as a major component (Liu et al., 1997, Kulshrestha et al., 2004, Drewnick et al., 2006).
The elements and ions shown in Fig. 2a are presented in probable parent compounds in Fig. 2b.
Since aluminum oxide (Al 2 O 3 ) and hydrochloric acid (HCl) are the exhaust products of the aluminum-based SRMs (Rutter, 2002), zinc and chloride are assumed to be present as predominantly as zinc oxide (ZnO) and HCl, although the ammonium present in Fig. 2a suggests that some unburnt ammonium perchlorate might have been present in the particles.The reconstructed mass in the White Lightning, Blue Thunder and Black Jack rockets accounted for about 60-85% of the total mass in the zinc-based rockets, leaving 15-40% of the mass as chemically unresolved.The unresolved mass could have been due to water generated from the ammonium perchlorate and the polymeric binder combustion which together probably constitute 80% of the fuel mass (Rutter, 2002;Stuart Davies, Michigan Tripoli Rocketry Association, pers. comm., 2005).Over 80% of the mass is accounted for in the Estes rocket as potassium sulfate (K 2 SO 4 ), leaving less than 20% unresolved.Negligible concentrations of elemental and organic carbon (EC and OC) were detected in particles collected from all of the rockets.Emission profiles showing the chemical fingerprints of each rocket type are provided in Table 2.The total mass size distributions obtained from MOUDIs 1 and 2 for experiments 6-11 are presented in Figs. 3 (a-f).The size distributions are normalized to the total mass for ease of comparison with the chemically speciated size distributions.The six figures present data from two experiments on each model of the zinc based rockets.The size distributions from MOUDIs 1 and 2 agree well with themselves and between each other over the course of the White Lightning and Black Jack experiments, showing a high level of precision.The size distributions collected in both of the Blue Thunder experiments are more tightly distributed than the other rocket measurements; around 1 μm.A disagreement between MOUDIs 1 and 2 is apparent in Fig. 3a at approximately 1 μm, which is not present in the duplicate experiment in Fig. 3b.It is not known why this disagreement occurred.Overall, the size distribution data for the Blue Thunder size distributions compare well with those presented for the other rocket measurements.In each figure, the ultrafine particles in the nucleation mode (mode 1) generated by gas-to-particle conversion were not clearly visible.Most of the particle mass was detected in the accumulation mode (mode 2) which was comprised of agglomerates of smaller mode 1 particles.The mean diameters and variances determined from each of the accumulation modes in the MOUDI samples are shown in Table 3. Good agreement is seen between the mean diameter and variances of the total masses and the zinc and chloride data.All of the size distributions in Fig. 3 show that the particles in hobby rocket exhaust are less than 2.5 μm in diameter, and therefore, are capable of impacting PM 2.5 and PM 10 samplers.
Fig. 3 (a-f).The comparison of the total mass size distributions collected by MOUDIs 1 and 2 for the Black Jack, White Lightning and Blue Thunder motors.The size distributions of zinc and chloride ions from the MOUDI filters used to sample Black Jack exhaust aerosol are presented in Fig. 4 and are described with mean diameters, p D and variances, σ 2 in Table 3. Small amounts of ammonium and trace species were detected on the MOUDI substrates, but were omitted from this graph for the purposes of clarity.Fig. 4 shows a single mode at approximately 1μm and agrees well with the total mass size distributions in Fig. 3 (a-f).Authors who have performed in-situ measurements of full-sized solid rocket motor (SRM) exhaust plumes in the stratosphere have typically seen bimodal and possibly trimodal size distributions (Ross et al., 1999, Rutter, 2002, Whitefield, 1997)  To determine whether or not hobby rockets could cause a bias in an atmospheric particle sampler we used a Gaussian puff model to simulate the emissions from a hobby rocket launch and calculated the mass of particulate zinc that would be deposited on a filter during a 24-hour collection cycle.The masses of particulate zinc from one rocket exhaust puff that would be collected by a low volume PM 2.5 or PM 10 sampler are presented in Fig. 5 as atmospheric concentrations, and are compared with the 24-hour averaged urban background concentrations of particulate zinc.A significant impact of particulate zinc from rocket exhaust was defined as when the rocket particulate zinc was 10% greater than the mass collected from the background air.This criterion was chosen under the assumption that an impact in the filter loadings of 10% or greater would represent a strong source contribution to a receptor; and therefore, produce an unacceptable error in the receptor model if the source were not accounted for.Sensitivity analyses were performed on the Gaussian puff model revealing that the stability class introduced the greatest uncertainty into the data.The largest daytime deviations from the base case results were for the atmospheric stability class C (unstable lapse rate) which returned particulate zinc concentrations that were between 80% and 15% those of the base case, depending on the distance of the sampler from the launch site.Nighttime deviations from the base case were explored with stability class E (stable lapse rate) which returned particulate zinc concentrations that were a factor of 1.5 to 2 higher than the base case.The data presented in Fig. 5 show the impact that one F20-7W motor, one of the smallest available, would cause on a 24-hour averaged filter sample collected in a low volume PM 10 sampler located at distances less than 20 km from the launch site.The rocket exhaust puff contains enough zinc particulate to significantly impact the atmospheric particulate zinc concentrations up to 20 km distance from the launch site under all of the stability classes studied.
Under stability classes D and E the impact is significant at distances greater than 20 km from the launch site.
Motors with propellant loadings larger than 30 g are available, but are subject to FAA regulations (MN TRA, pers comm., 2005).Therefore these motors are only used at rural launches organized by the Tripoli Rocket Associations (TRA) in each state.Such high-powered rocketry meetings occur as regularly as once a month and normally launch at least 20 to 30 rockets in a day.Some clubs have one or two meetings a year which are three or four days long during which time over 1000 launches may occur.The high-powered rockets typically use motors with propellant loads of 200-500 g, but on occasion may contain 1-5 kg of propellant (WI and MI TRA, pers. comm. 2005).Fuel formulations are proprietary (Aerotech, pers. comm., 2004) so without scientific tests it is unknown whether the fuels of these larger rockets are based upon zinc, aluminum (the typical fuel used in the full scale SRMs), or some other chemical.To assess the impact of aluminum-based motors on a sampler, aluminum contents of atmospheric particles measured in Milwaukee and Waukesha by Lough (2004) were used as a reference.
Average background particulate aluminum concentrations were approximately two and a half times those of particulate zinc in both locations, and were probably due to soil material (Ramadan et al., 2000, Song et al., 2001, Zhou et al., 2004).Therefore particles from a small rocket with an aluminum-based fuel might also cause sampling interference which could produce errors in receptor model results.Furthermore, both small monthly meetings and larger weekend events would cause large interferences with background PM 10 zinc and PM 10 aluminum data at distances far greater than 20 km downwind of the event site, meaning that nearby rural or urban locations could be significantly impacted by zinc and aluminum from hobby rockets.Such interferences with zinc and aluminum signals from background particulate matter could potentially cause large impacts on receptor model calculations.

CONCLUSIONS
This study illustrates how a small, unregistered and ephemeral source of particulate matter that is enriched in an atmospheric particulate indicator species, might significantly impact receptor model calculations performed using data from 24-hour averaged filter samples.The illustration presented in this study uses the zinc content of particulate matter collected from the exhaust plumes of hobby rocket motors to determine the downwind impact of the particulate zinc on a sampler.A Gaussian puff model was used to estimate the dispersion of the exhaust plume as it is transported from the launch site to the sampler.The results presented imply that the use of zinc as an emission indicator species might be unreliable if hobby rockets are being launched upwind of the sampler.This raises the more general question of whether there are other unregistered sources that emit particulate matter which is highly enriched in other commonly used indicator species.Should such sources be identified they would need to be characterized to determine whether their emissions could interfere with receptor modeling performed using impacted 24 hour averaged PM 2.5 and PM 10 filter samples.The minor metals size distributions had a strong mode 1 at 0.1 μm in which all of the metals were present.Only potassium predominated in mode 2. The trace metals size distribution was more bimodal in character than any of the size distributions previously discussed.In contrast to the size distribution of the minor metals, the strongest of the trace metal modes was mode 2 at 1 μm, which also corresponded to the strongest mode in the zinc, chloride and total mass size distributions.Fig. S2 reflected size distributions of particles emitted by the full size SRMs with modes at 0.1 μm, and 0.9 μm (Whitefield, 1997, Ross et al., 1999, Rutter, 2002).The detection of a nucleation mode and an accumulation mode, suggested that the particle formation occurred by nucleation from gaseous combustion products followed by agglomeration.

Fig. 1 .
Fig. 1.Schematic of apparatus used to collect particles emitted by rockets.
(URG, Corporation, Chapel Hill, NC)  were used on each cyclone to obtain mass and chemical speciation data.Teflon substrates (Teflo ® , 2 μm, 47 mm Pall Life Sciences, East Hills, NY) were used in one PM 2.5 cyclone, whilst pre-baked quartz substrates (47mm Pall Life Sciences, East Hills, NY) were used in the other.Likewise, pre-baked aluminum-foil substrates were used in one MOUDI whilst Teflon substrates (47 mm Pall Life Sciences, East where ν d is the particle settling velocity, K zz is the eddy correlation coefficient of the atmosphere in the vertical direction and σ z is the Pasquill-Gifford parameter in the vertical direction.For a perfectly reflecting puff, β = 0, meaning that turbulent eddy diffusion dominates over gravitational deposition in the motion of the particle in the vertical direction.Conversely, for a perfectly absorbing puff, β = ∞.Values of β were calculated for particles at the minimum and maximum size limits of the size distributions presented in Figs.3 and 4to assess whether or not a reflecting or an absorbing puff should be used to model the rocket emissions.Particle deposition velocities (ν d ) of 3 × 10 -8 m/s and 8 × 10 -2 m/s were calculated for particles 50 nm and 5 µm in diameter.A K zz value of 500 m at 1km was calculated for a stability class D, where σ z = 31.5 m.The β values for 50 nm and 5 µm particles were calculated to be 6 × 10 -11 1/ms and 2 × 10 -6 1/ms, respectively.The low β values of the 50 nm and 5 µm particles showed that the vertical movement of all of the particles in the puff was dominated by turbulent eddy diffusion.

Fig. 2a .
Fig. 2a.The bulk chemical compositions of the exhaust particles from the four different models of rocket motors used in this study.

Fig. 2b .
Fig. 2b.Reconstructed bulk chemistry for each of the rocket motor types studied.The chlorine species were assumed to be comprised of hydrochloric acid and unburnt ammonium perchlorate.
Figs. 3 or 4. The minor and trace metal size distributions in Appendix 1 show the bimodal nature of the size distribution more clearly.

Fig. 4 .
Fig. 4. The size-resolved bulk chemical composition of exhaust particles from the Black Jack motors.

Fig. 5 .
Fig. 5.The impact of zinc particulate matter from a hobby rocket launch on 24 hour averages of PM 10 ambient zinc measurements as the exhaust puff is transported away from the launch site at a wind speed of 5 m/s under a range of atmospheric stabilities.

Fig. S2 .
Fig. S2.The trace metal composition of the exhaust particles from the Black Jack motors.

Table 1 .
Rocket information, particle collection technique and composite information for each experiment.

Table 2 .
Emission Profiles for Blue Thunder, White Lightening, Black Jack and Estes rockets.Data not in bold are not statistically different than zero.

Table 3 .
Mean diameters, p D and variances σ 2 for modes 1 and 2 in each of the size distributions presented in Figs.3 and 4.