Introduction

Airborne particulates play a significant role in our everyday lives. This includes the role of the particulates in the radiative balance of the atmosphere¹²³⁴⁵⁶ and their effect on human health.^{78910111213141516} These issues can be due to long-term exposure, such as lung cancer and cardiopulmonary disease,¹¹¹²¹³ or short-term exposure, such as respiratory problems and bacterial infections.^7,1415–16

To fully characterize all the roles airborne particulates play in our lives on a global scale, direct observations are needed. Satellite-based data products allow for full global coverage on a daily basis, but the data need in situ validation. Conventional ground-based air quality monitoring stations do not provide data on an appropriate spatial scale for this validation.¹⁷ Validation with traditional aerial vehicles can be challenging in the lowermost atmospheric boundary layer due to flight safety considerations. Previous studies have looked at air quality variability using conventional aircrafts.^{18192021222324} For safety reasons, the measurements are restricted to the upper levels of the boundary layer and above. Data gathered in the lower layers of the boundary layer are critical for validating satellite-based observations and data products due to the high abundance of aerosols in the boundary layer and the presence of sharp vertical gradients. Comprehensive sampling of the full aerosol size distribution and profiling of the lowermost boundary layer are rare.

Several studies have used tethered balloons to study airborne particulates.^25262728 These studies have either taken measurements at a set location across changing altitudes,^25,26 providing information on temporal and vertical variation but not horizontal, or at set locations and set altitudes,^27,28 only providing information on temporal variation. This study provides data on all three variations: vertical, horizontal, and temporal.

The motivation for this study was to prepare a validation platform for the upcoming National Aeronautics and Space Administration (NASA) missions, including the NASA Aerosol-Cloud-Ecosystems and Pre-Aerosol, Clouds, and ocean Ecosystem missions.²⁹ This study uses a remote-controlled aerial vehicle to look at the horizontal, vertical, and temporal variability of airborne particulates within the first 150 m of the atmosphere. This platform, due to its size and maneuverability, is uniquely qualified to gather the much needed data in this lower level of the boundary layer. To the best of our knowledge, this is the first time a zero-emission aerial vehicle has been used to quantify the aerosol size distribution and its vertical, horizontal, and temporal variation. Thus, this study sets the stage for the next generation of satellite validation methodologies.

Instrumentation

A large remote-controlled model airplane was built from an AMR Payload Master 100 kit. This kit was modified to use twin-electric motors mounted on the wings allowing for sample collection in undisturbed air from the nose of the aircraft. It was also modified to use a tricycle-style landing gear, instead of a tail wheel–style gear, to support the increased nose weight of the sensor package. This vehicle was originally built and flown during a methane gas measurement campaign conducted in October 2013, where it successfully completed 34 flights.

For this aerial measurement campaign, the plane needed to be able to fly a precise path at specific altitudes, and then repeat the exact flight plan multiple times over multiple days. To accomplish this, a MicroPilot MP2128LRC autopilot system was employed. The autopilot is capable of flying a complete mission from takeoff to landing with way-points set in GPS coordinates, including altitude and airspeed. This system incorporates three 3-axis gyroscopes/accelerometers, GPS, pressure altimeter, and pressure airspeed sensor onto a single circuit board.³⁰ For communication with the ground station laptop, a Microhard 900-mHz radio modem was installed on the vehicle.

To monitor particulates, a sensor package was built and installed in the twin-engine remote-controlled airplane. The package utilized a Grimm 1109 aerosol spectrometer mounted in the nose of the plane. An intake probe was mounted in the cowl to allow samples to be collected in clean, undisturbed air.

A Raspberry Pi model B+ microcomputer was used to record data from temperature, pressure, humidity, and airspeed sensors. A digital compass and GPS modules were used to calculate wind speed and direction. The GPS antennas for both the sensor package and the autopilot system were installed just behind the main wing.

A touch-screen LCD panel was installed in the front window of the airplane to display data from the Raspberry Pi sensor package. This allowed for verification of data collection from the sensor package prior to takeoff and to stop data collection after landing.

Data Collection

Four flights were conducted on December 4, 2014, starting at 12:00 pm and ending at 5:00 pm local time. Each flight lasted from 6 to 8 minutes and consisted of an oval pattern with vertices of approximately 650 m by 250 m flying at an average airspeed of 60 knots. During the first three flights, the altitude was increased after each lap, starting at approximately 30 m, in 15-m intervals up to 140 m. The fourth flight was conducted at a near-constant altitude of 60 m.

Wind speed and direction were calculated after all flights were completed by using the heading and airspeed vector and the ground speed and ground track vector. Due to differences in data collection rates between the GPS sensor and the airspeed/compass combination, the wind speed and direction were only calculated for straight path portions of the flight. Straight path was defined as portions of the flight where heading and ground track differed by less than 10°. Portions of the flight not within this limit were assumed to have a wind speed and direction equal to the mean from the previous straight path segment.

Discussion

Let us start by considering the vertical structure of the mean PM concentration in the lowest part of the boundary layer. Data from each flight were averaged into 10-m increments between the ground and 140 m. The mean PM_2.5 concentration for the first three flights was 36.3 ± 3.6 µg/m³. The small deviations suggest a very well mixed boundary layer. Figure 1 shows the summary vertical profiles for the first three flights. All three size fractions showed the highest PM concentration below 10 m. For PM_2.5, the highest concentration occurred at 9.1 m. Above 10 m, the mean PM_2.5 concentration was only 34.8 ± 2.2 µg/m³, while below 10 m the mean concentration was 40.3 ± 3.5 µg/m³. The PM₁ concentration was the most consistent, with a standard deviation of 2.0 µg/m³, while the PM_2.5 concentration had a standard deviation of 3.6 µg/m³ and PM₁₀ concentration had a standard deviation of 21.3 µg/m³.

Figure 1.

Mean concentrations of PM₁ (panel A), PM_2.5 (panel B), and PM₁₀ (panel C) as a function of altitude for the first three flights. Panels D–F are a comparison of PM₁ and PM₁₀ concentrations to PM_2.5 concentrations for the first three flight.

A key issue addressed in this paper is the characterization of aerosol variability: horizontal, vertical, and temporal. The plots in Figure 1 are the summary of many measurements, so the full PDF of aerosol distributions at each altitude is depicted by using shading to showcase the different percentiles (0%, 10%, 20%,… 100%). The overlaid line is the mean value.

The profiles show the summary value for different size fractions PM₁ (panel A), PM_2.5 (panel B), and PM₁₀ (panel C). To characterize the relative abundance of these size fractions, the ratios of PM₁/PM_2.5 (panel D) and PM₁₀/PM_2.5 (panel F) are shown. Note that the PM₁/PM_2.5 ratio is less than 1 in the sampling of the boundary layer, typically in the range of 0.7-0.95. The ratio PM₁₀/PM_2.5 (panel F) is typically larger than 1 in the range 1-6.

Temporal Variation

The first three flights had the highest observed PM_2.5 concentrations at 35.1 ± 3.5 µg/m³, 37.2 ± 3.3 µg/m³, and 36.8 ± 3.9 µg/m³, respectively, while the fourth flight saw a reduction in concentration to 23.5 ± 2.9 µg/m³, shown in Figure 2. The histogram color scale indicates the air quality according to guidelines set by the Environmental Protection Agency (EPA). Each of the four flights was separated by approximately 1.5 hours. Thus, the four distributions show how the PM_2.5 concentration changed over time. The maximum variation in mean concentration between the first three flights is only 2.2 µg/m³. This is less than the standard deviation of any of the three flights. The fourth and final flight saw a significant reduction in concentration of 13.3 µg/m³ from the previous flight. This reduction in concentration is much higher than the variation of any of the first three flights, which indicates that the reduction is not a result of the flight being conducted at a constant altitude. This flight concluded at sunset; therefore, the reduction is most likely due to the changing boundary layer height.

Figure 2.

PM_2.5 distributions for each of the four flights. Color scale indicates the air quality using EPA standards.

Air Parcel Origin

The temporal variation is in part due to where the air parcels originated. To examine the likely air parcel origin, the Hybrid Single Particle Lagrangian Integrated Trajectory Model (HYSPLIT) model³¹ was used to calculate 24-hour back-trajectories for each flight. The trajectories were calculated at 5-m increments up to 140 m and checked for any altitudes that originated on the ground within the 24 hours prior to measurement. Only the first two flights (A and B) had air parcels observed at higher altitudes that originated on the ground. Figure 3 shows the trajectories and the PM_2.5 distribution for each of the four flights. The colored traces represent the path of an air parcel at a single altitude, while the corresponding colored area outlines the region on the ground that air parcels from any altitude originated from. The distance from the region of measurement is also indicated on the map. Particulates from these areas are carried along by the wind and affect the overall measured PM concentration.

Figure 3.

Map depicting air parcel trajectories for 24 hours prior to measurement for each of the four flights.

Let us now compare Flight A, with multiple air parcel altitudes originating on the ground, and Flight C, with no air parcel altitudes originating on the ground. Flight D will be described separately due to its much lower PM concentration and because the flight plan was altered to fly at a near-constant altitude of 60 m.

The map for Flight A, Figure 4 (left), shows the air parcel trajectories over the 24-hour period prior to measurement. The blue line, as an example, is the path of the air parcels measured at 0-m altitude. The blue line is colored red to indicate when this air parcel was close to the ground. Additional red marks on the map indicate where other air parcels, at varying altitudes, are closest to the ground. The plot in the upper right shows the altitude of air parcels over the prior 24 hours; only the altitudes for air parcels that originated on the ground are included in the plot. The 0 point on the x-axis is the measurement time. This shows how long before measurement the parcel was in contact with the ground, thus how long any particulates from that area had to settle out of the atmosphere. Note the specific points notated on the trajectory and corresponding notations on the map. Point A indicates the time and location of the highest altitude reached by each of the air parcels. Points B and C indicate the time and location of air parcels origin on the ground. The plot in the lower right shows how the mean PM_2.5 concentration changes with altitude. For this plot, measurements were averaged into 5-m-altitude bins. The highest measurements are observed in the lowest two bins, point 1, equating to 5 m and 10 m. There is a second peak occurring from 35 m to 60 m, point 2, which corresponds to a different section of air parcels that originated on the ground.

Figure 4.

Map depicting air parcel trajectory over a 24-hour period prior to measurement of Flight A (left). Top right plot shows air parcel altitude as a function of time prior to measurement while the bottom right plot shows mean PM_2.5 concentration as a function of altitude.

Figure 5 shows similar information for Flight C. In the map on the left, the only time where the air is close to the ground in the 24 hours prior to measurement (red area) is at the point of measurement in the northernmost tip of the trajectory, point A. This is reflected in the back-trajectory plot in the upper right were only one altitude is indicated. Points B and C on the trajectory, and reflected on the map, indicate the time and location of air parcel peak altitudes. The mean PM_2.5 concentration is still at its highest in the 5-m and 10-m bins, point 1, but there is not another peak until the 80-m mark, point 2. This sharp peak observed at a single altitude (80 m) is in contrast to the Flight A data, which shows an increased concentration along a range of altitudes (35-60 m).

Figure 5.

Map depicting air parcel trajectory over a 24-hour period prior to measurement of Flight C (left). Top right plot shows air parcel altitude as a function of time prior to measurement while the bottom right plot shows mean PM_2.5 concentration as a function of altitude.

Vertical Variation

The largest PM concentrations are typically observed close to the ground. This study looks at how the concentration and particle size bin counts change with altitude. Figure 6 shows the altitude profile of PM₁, PM_2.5, and PM₁₀ concentrations and particle size counts. In panels A and B, PM₁ and PM_2.5 concentrations both decrease above 40 m. Conversely, PM₁₀ (panel C) has a strong concentration peak around 55 m, then returns to a fairly constant concentration above that. This peak may be a result of the low particle count. This is indicated by panel e, which shows particle count versus size for all altitude bins. For particle sizes above 3 microns, at most altitudes, the number of particles is less than 10 (the horizontal line). To see this more clearly, panels D, F, and G show particle count as a function of altitude for 0.25, 1, and 3 microns, respectively. The smaller particles have a much higher count, and larger variation, than the larger particles. However, with the much greater mass of the larger particles, a change in count of a few particles has a much greater effect on the overall concentration.

Figure 6.

Flight A particle count profiles. Panels A, B, and C show mean concentration as a function of altitude for PM₁, PM_2.5, and PM₁₀, respectively. The shading indicates the measurement percentiles while the overlaid line is the mean value. Panel E shows particle count as a function of particle size for all altitude bins. Panels D, F, and G show particle count as a function of altitude for 0.25, 1.3, and 3.5 microns, respectively.

For Flight C, Figure 7 shows how particle count for all size bins, and specific sizes, changes with altitude. In this flight, panels A, B, and C show a decrease in concentration above 20 m, with a small peak at 80 m. A small peak is also seen in the particle count for the three size bins, around 80 m, in panels d, f, and g. For this flight, particle counts for the larger particles drop below 10 particles for all altitudes (panel E), whereas Figure 6 (panel E) showed a few altitudes where all size bins maintained more than 10 particles. This is an indication of the different sources of particulates. Recall the air parcel trajectories of Figures 4 and 5. For Flight A, air parcel at multiple higher altitudes originated on the ground away from the measurement site, whereas for Flight C, no higher altitude air parcel originated on the ground within 24 hours prior to measurement. Thus, larger particles were neither lifted to the higher altitudes nor had time to settle out of the atmosphere prior to measurement.

Figure 7.

Flight C particle count profiles. Panels A, B, and C show mean concentration as a function of altitude for PM₁, PM_2.5, and PM₁₀, respectively. The shading indicates the measurement percentiles while the overlaid line is the mean value. Panel E shows particle count as a function of particle size for all altitude bins. Panels D, F, and G show particle count as a function of altitude for 0.25, 1.3, and 3.5 microns, respectively.

Horizontal Variation

Satellite-based data products have finite resolution with a typical pixel size range of 250 m-10 km. Flight D was used to examine the horizontal variation at much finer scales across a pixel of satellite imagery. To minimize any variation caused by a change in altitude, this flight was conducted at a near-constant altitude of 60 m.

Figure 8 shows mean PM_2.5 concentration, standard deviation, and number of measurement points for the gridded measurement area. The top row is gridded into a 30 m × 15 m grid to show more detail in the horizontal variance. Note that most of the grid points only have one or two measurements per cell, thus the standard deviation for the majority of the flight is 0. This is not statistically significant as each cell only has a single data point. The bottom row is gridded into a 120 m × 65 m grid to increase the number of data points in each grid cell, allowing for a more robust statistical analysis. For this flight, mean concentration ranged from 20.5 to 24.9 µg/m³ with a standard deviation range of 0.8-3.3 µg/m³. This indicates that the horizontal variance for an area of this size is of the same order as the vertical variance (3.6 µg/m³). The range of PM concentration or this area is approximately 4.4 µg/m³ which is less than twice the highest standard deviation of any grid cell. This indicates that the variation across the entire area is similar to the variation of any given point of measurement. Thus, a data pixel of this size would still accurately characterize the concentration of the entire area. The variability could be larger for a larger or more diverse area as may be covered by satellite-based observations.

Figure 8.

Mean PM_2.5 concentration, standard deviation, and number of measurements for Flight D divided into 30 m × 15 m (top) and 120 m × 65 m (bottom) grids.

Conclusion

To the best of our knowledge, this is the first time a zero-emission aerial vehicle has been used to quantify the vertical, horizontal, and temporal variations of airborne particulates in the lower region of the boundary layer. Satellite-based data products may allow for full global coverage on a daily basis, but they need in situ validation. The platform used in this study is uniquely qualified for measurements in the all-important lower boundary layer, which are necessary for this validation. This study shows the value of this approach for the next generation of satellite validation methodology using aerial vehicles.

The mean PM_2.5 concentration for three flights with varying altitude was 36.3 µg/m³, with the highest concentration occurring below 10 m altitude. The overall vertical variation was very small with a standard deviation of only 3.6 µg/m³.

PM_2.5 concentration also did not change much throughout the day with mean concentrations for the altitude-varying flights of 35.1, 37.2, and 36.8 µg/m³. The flight flown at a constant altitude had a lower concentration of 23.5 µg/m³. This reduction in concentration is much larger than the concentration variation in any of the first three flights, and therefore, the constant altitude most likely does not account for the change; however, more data are needed to draw any conclusions on the cause of the change.

To characterize variation on the satellite data subpixel scale, the fourth flight was conducted at a near-constant altitude of 60 m. Data were then sectioned into a 120 m × 65 m grid. The mean PM_2.5 concentration across the grid cells ranged from 20.5 to 24.9 µg/m³. This range is within the standard deviation of any grid cell. Thus, a data pixel with an area equivalent to the flight path area would accurately characterize the PM concentration of this region.

This system has proven to be qualified to gather the measurements needed for proper validation of satellite observations and data products. With the continued drop in cost and sensor size, this proof of concept could easily transition into multiple vehicles used on a routine basis to help validate all future missions.

Author Contributions

Conceived and designed the experiments: WAH. Analyzed the data: WAH, AGM. Wrote the first draft of the manuscript: WAH. Contributed to the writing of the manuscript: WAH, BJN. Agree with manuscript results and conclusions: WAH, DJL, BJN. Jointly developed the structure and arguments for the paper: WAH, DJL. Made critical revisions and approved final version: WAH, DJL, BJN. All authors reviewed and approved of the final manuscript.