4>EPA
EPA/600/R-12/734 December 2012 www.epa.gov/research
United States
Environmental Protection
Agency
Atmospheric Light Detection
and Ranging (LiDAR) Coupled
With Point Measurement Air
Quality Samplers to Measure
Fine Participate Matter (PM)
Emissions From Agricultural
Operations:
The Los Banos CA Fall 2007
Tillage Campaign
RESEARCH AND DEVELOPMENT
-------
-------
Atmospheric Light Detection and
Ranging (LiDAR) Coupled With Point
Measurement Air Quality Samplers to
Measure Fine Particulate Matter (PM)
Emissions From Agricultural
Operations:
The Los Banos CA Fall 2007 Tillage Campaign
David J. Williams1, Sona Chilingaryan2, and Dr. Jerry Hatfield3
1 U.S. Environmental Protection Agency, Environmental Sciences Division,
National Research Laboratory, RTP, NC
2U.S. Environmental Protection Agency, Air Division, Region 9, San Francisco, CA
3U.S. Department of Agriculture, Agricultural Research Service, National Soil Tilth Laboratory, Ames, IA
Although this work was reviewed by EPA and approved for publication, it may not necessarily reflect
official Agency policy. Mention of trade names and commercial products does not constitute
endorsement or recommendation for use.
U.S. Environmental Protection Agency
Office of Research and Development
Washington, DC 20460
-------
Report Prepared By:
Dr. Gail Bingham4, Jennifer Bowman5, Dr. Christian Marchant6,
Dr. Randal Martin7, Kori Moore8, Dr. Philip Silva9, and Dr. Michael Wojcik
4SDL Chief Scientist & Civil Space and Environment Division Leader,
Space Dynamics Laboratory, North Logan, UT
5Senior Manager, Internal Communications ATK Aerospace Systems, Magna, UT
Imagery Scientist, National Geospatial-lntelligence Agency, Springfield, VA
7Research Associate Professor, Department of Civil and Environmental Engineering,
Utah State University, Logan, UT
Environmental Engineer, Space Dynamics Laboratory, North Logan, UT
Environmental Chemist, Animal Waste Management Research Unit,
USDA Agricultural Research Service, Bowling Green, KY
10Branch Chief for Environmental Measurement, Space Dynamics Laboratory, North Logan, UT
-------
Atmospheric Light Detection and Ranging (LiDAR) coupled with point
measurement air quality samplers to measure fine particulate matter (PM)
emissions from agricultural operations: the Los Banos CA Fall 2007 tillage
campaign.
Authors:
David Williams
Environmental Sciences Division
National Exposure Research Laboratory
U.S. Environmental Protection Agency
RTF, NC
Sona Chilingaryan
Air Division
Region 9
U.S. Environmental Protection Agency
San Francisco, CA
Dr. Jerry Hatfield
National Soil Tilth Laboratory
Agricultural Research Service
United States Department of Agriculture
Ames, Iowa
Disclaimer Notice: Although the information in this document has been funded in part by the
United States Environmental Protection Agency under Interagency Agreement: DW 12922568 to
Space Dynamics Laboratory, it does not necessarily reflect the views of the Agency and no
official endorsement should be inferred.
-------
TABLE OF CONTENTS
Table of Contents ii
List of Figures iv
List of Tables vii
Executive Summary 5
1. Background 9
1.1 Literature Review 10
2. Experiment Design 15
2.1 Site Description 15
2.2 Operation Description 18
2.3 Tillage Operation Data 22
3. Measurements and Methods 24
3.1 Overview 24
3.1.1 Meteorological Measurements 29
3.1.1.1 Eddy Covariance Measurements 30
3.1.2 Wind Profile Calculations 32
3.1.3 Soil Sampling 33
3.1.4 Air Quality Point Samplers 33
3.1.4.1 MiniVol Portable Air Sampler 34
3.1.4.2 Optical Particle Counter 35
3.1.4.3 Organic Carbon/Elemental Carbon Analyzer 37
3.1.4.4 Ion Chromatographic (1C) Analysis 38
3.1.4.5 Aerosol Mass Spectrometer 38
3.1.5 Lidar Aerosol Measurement and Tracking System 39
3.2 Modeling Software 43
3.2.1 Dispersion model software 43
-------
3.2.2 Plume Movement Prediction 46
3.3 Statistical Analysis of Data 47
4. Results and discussion 48
4.1 General Observations 48
4.1.1 Soil Characteristics 48
4.1.2 Precipitation Data 51
4.1.3 Eddy Covariance Calculations 52
4.1.4 Plume Movement Prediction 55
4.2 Aerosol Characterization Data 56
4.2.1 MiniVol Filter Sampler Data 56
4.2.2 ISC/AERMOD dispersion models 61
4.2.2.1 ISCST3 62
4.2.2.2 AERMOD 62
4.2.3 PM Chemical Analysis 63
4.2.3.1 Organic Carbon/Elemental Carbon Analyzer 63
4.2.3.2 Ion Chromatographic (1C) Analysis 65
4.2.4 Aerosol Mass Spectrometer 67
4.3 Optical Characterization Data 68
4.3.1 Optical Particle Counter 68
4.3.2 Optical to PM Mass Concentration Conversion 69
4.3.3 Lidar Aerosol Concentration Measurements 74
4.4 Fluxes and Emission Rates 76
4.4.1 Lidar based Fluxes and Emission Rates 76
4.4.2 ISCST3 Model Emission Rates 80
4.4.3 AERMOD Model Emission Rates 81
4.5 Derived Emission Rate Comparison 82
in
-------
5. Summary and conclusions 88
6. Acknowledgments 92
7. Publications 92
8. References 93
9. Appendices 98
9.1 Appendix A 98
9.2 Appendix B 100
LIST OF FIGURES
Figure 1. Shaded relief map of the State of California, USA, with the location of the selected
sample site shown by the white star. Image from geology.com [29] 16
Figure 2. Soil type overlay on an aerial photo of the fields of interest (Field A and Field B) with
field boundaries in red. Soil classifications include: 103 - Alros clay loam, partially drained; 139
- Bolfar clay loam, partially drained; 170 - Dos Palos clay loam, partially drained; and 283 -
Xerofluvents, channeled [28] 17
Figure 3. Photo taken on October 19, 2007 standing near the southern edge of Field B looking
north across fields B and A during the chisel pass of the combined operations tillage method... 18
Figure 4. Photograph of a stationary Optimizer Model 5000 with attached roller assemblies 21
Figure 5. An Optimizer Model 5000 in operation during this field experiment, with an
instrumented tower in the background 21
Figure 6. Wind rose for September and October of 2004 - 2006 as recorded by the CIMIS Station
# 56 (Los Banos) 24
Figure 7. Sample layout used for Field B, with the area tilled shown by the shaded polygon.
Created using Google Earth software 25
Figure 8. Layout used during tillage of Field A, with the area tilled shown by the shaded
polygon. Created using the Google Earth software 28
Figure 9. A wind profile calculated for the Disc 2B pass on 10/27/2007 32
Figure 10. Airmetrics MiniVol Portable Air Sampler, a closeup view and an example of field
deployment 34
IV
-------
Figure 11. Met One Instruments Optical Particle Counter (OPC) Model 9722, indicated by
arrow, setup for field deployment on a tower base with an accompanying rechargeable battery
pack and solar panel 35
Figure 12. The three wavelength Aglite lidar at dusk, scanning a harvested wheat field 39
Figure 13. The Aglite lidar retrieval algorithm flow chart, showing the input locations for the in
situ data 40
Figure 14. (A) Conceptual illustration of the method for using lidar to generate time resolved
local area particulate fluxes. (B) An example of a "staple" lidar scan over the facility showing
aerosol concentration on the three sides of the box 41
Figure 15. Soil sample collection locations in fields under study 50
Figure 16. Timeline of soil moisture levels, tillage activities, and precipitation events 50
Figure 17. Expected soil moisture levels over time due to addition by precipitation and losses
through evapotranspiration, with potentially affected tillage operations shown 51
Figure 18. Friction velocity for 10/24/2007, 13:00-15:00 hours at N Met computed as (a) 1, (b)
15 and (c) 30-minute averages 53
Figure 19. One minute average wind speed (a), aw (b), and w*(c) for Site 3 on 10/23/2007 54
Figure 20. Average measured PM concentrations, upwind and downwind, with the particle size
contributions to the total PM 59
Figure 21. Modeled period average ISCST3 results (modeled hours = 8, sample time = 7.27 hrs)
at 2m above ground level for the Disc 1 pass of the conventional tillage operations on October
23, 2007 with light north winds. The field area is outlined by the thick dashed line and sampler
locations are shown in green; contour line numerical values are in |ig/m3 62
Figure 22. Modeled period average AERMOD results (modeled hours = 8, sample time = 7.27
hrs) at 2m above ground level for Disc 1 pass of the conventional tillage operations on October
23, 2007 with light north winds. The area of operations is outlined by the thick dashed line and
sampler locations are shown in green; contour line numerical values are in |ig/m3 63
Figure 23. PM2.5 OC/EC time series concentrations as collected at the downwind AQ trailer
location. It should be noted that the raw instrument OC concentrations have been multiplied by
1.7 to account for potential non-carbon functional groups 64
Figure 24. PM2.5 organic matter and elemental carbon concentrations during specific sampling
periods (parallel to filter-based sampling) 64
Figure 25. Chemical composition of downwind PM2.5, PMio and TSP filters from the chisel pass
in the conventional tillage method, 10/25 66
-------
Figure 26. Average chemical composition of particles measured by AMS (~PMi) in Los Banos,
10/15/2007-10/17/2007 67
Figure 27. Typical mass-to-charge (m/z) data collected at Los Banos, with significantly
contributing ions and their source particles identified 68
Figure 28. Sample period average particle volume size distributions (|am3/cm3-|am) measured
from upwind (background) and downwind (background plus emissions) locations, with the
difference being the aerosol emitted by the tillage activity, (a) is the chisel operation of the
combined operations tillage method, (b) is the optimizer operation of the combined operations
tillage method, (c) is the disc 1 operation of the conventional tillage method, and (d) is the chisel
operation of the conventional tillage method 69
Figure 29. Average daily MCF with error bars representing the 95% confidence interval 71
Figure 30. PM2.5, PMio, and TSP mass concentrations retrieved from collocated lidar and OPC
during the 'stare' time series for 10/25. Data acquisition time of the lidar data is 0.5 sec while
OPCs were set to 20 sec accumulation time. Measurements were done on the upwind side of
facility (location Nl - "Pig" OPC) 72
Figure 31. PM2.5, PMio, and TSP mass concentrations retrieved from collocated lidar and OPC
during 'staple' scanning (bottom point for the range bin of the OPC, collected from each staple
shown). Data acquisition time of the lidar data point is 0.5 sec while OPCs were set to 20 sec
accumulation time. Measurements were done on the downwind side of facility (location S5 -
'Horse' OPC) on 10/23/2007 74
Figure 32. Wind speed, wind direction, upwind and downwind plume area average particulate
volume concentrations, for the October 20, 2007 Optimizer pass of the combined operation
tillage 75
Figure 33. Wind speed, wind direction, upwind and downwind plume area averaged particulate
volume concentrations for the October 23, 2007 first disc pass of the conventional tillage
operation 76
Figure 34. Lidar derived fluxes (g/s) of PM2.5, PMio, and TSP for the October 20, 2007
Optimizer pass of the combined operation tillage over the operation sample time of 2.85 hrs.... 77
Figure 35. Lidar derived fluxes (g/s) of PM2.5, PMio, and TSP for the October 23, 2007 first disc
pass of the conventional tillage operation over the operation sample time oil 21 hrs 77
Figure 36. Summed PM2.5, PMio, and TSP emission rates ± 95% confidence intervals for both
tillage methods derived from lidar flux measurements and inverse modeling using ISCST3 and
AERMOD. (* One PM2.5 emission rate missing per tillage method, therefore no total emissions
were calculated.) 83
Figure 37. Weighted average PMio concentrations (ng/m3) modeled by (a) ISCST3 and (b)
AERMOD using derived emission rates for 10/23 (modeled hours = 5, sample time = 4.24 hrs)
along the downwind vertical plane that corresponds to the lidar scanning plane. The green
VI
-------
markers show point sampler locations on the downwind side of the field. Maximum predicted
concentrations for ISCST3 and AERMOD were 49.9 and 37.7 ng/m3, respectively, at 2 m above
ground level 86
Figure 38. Lidar-measured downwind PMio concentrations (a) averaged over all valid scans over
the 4.24 hr long sample period for 10/23 (w=122) and (b) for a single vertical scan, which
demonstrates observed plume lofting 87
LIST OF TABLES
Table 1. Emission factors and uncertainties for land preparation as reported by Flocchini et al.
(2001) [3] 11
Table 2. Emission factors used by the California Air Resources Board in estimating agricultural
tilling PMio emissions [13] 11
Table 3. Conventional and conservation tillage emission rates reported by Madden et al. (2008)
for tillage in a dairy forage crop rotation [4]. ST= standard tillage method, CT = conservation
tillage method 12
Table 4. Tillage operations and dates performed for the comparison study 19
Table 5. Equipment used on the fields to perform the tillage operations 20
Table 6. Tractor run time and fuel usage as recorded by the farming company 22
Table 7. Operation data for both the conventional and combined operation tillage studies as
recorded by field personnel 23
Table 8. Sample period, total tractor operation time, and the sample period-to-tractor operation
ratio for all sample periods 23
Table 9. Summary of instruments located at each site for the combined operations tillage study of
Field B. All height given as above ground level (agl) 26
Table 10. Summary of instruments located at each site for tillage study of Field A. All heights
given as agl 28
Table 11. Number of upwind and downwind lidar scans determined to be valid for emission rate
calculations 43
Table 12. Statistics of soil characteristics measured for both fields 49
Table 13. Percent of puffs predicted to cross the lidar beam plane 56
Vll
-------
Table 14. Mean measured PM concentrations for each operation upwind and downwind of the
tillage site. Error is the 67% CI about the mean for n > 3 57
Table 15. Average (± la) fraction of TSP that was PM2.5 and PMio for each operation upwind
and downwind of tillage site, and campaign averages for upwind and downwind 58
Table 16. Upwind and downwind average concentrations ± 67% CI (for w>3) used in emission
rate calculations 61
Table 17. Averaged filter ionic analysis for upwind and downwind samples for Oct. 23rd, 25th,
and 26th, 2007 65
Table 18. Mass conversion factors estimated for each day of the tillage operations and averaged
for the whole campaign. Error values represent the 95% confidence interval 70
Table 19. Comparison of PM mass concentrations (|ig/m3) as reported by MiniVol samplers and
mean values measured by collocated OPCs and lidar at Nl (upwind) and S5 (downwind) for
10/23/2007 73
Table 20. Mean fluxes (g/s) ± 95% confidence interval from quality controlled samples for each
tillage operation 78
Table 21. Aerosol mass transfer (± 95% confidence interval) from each field (flux normalized by
operation duration and area tilled) as calculated from lidar data for all tillage operations 79
Table 22. Mean emission rates (± 95% CI for w>3) for each operation as determined by inverse
modeling using ISCST3 81
Table 23. Mean emission rates (± 95% CI for «>3) for each operation as calculated by inverse
modeling using AERMOD 82
Table 24. Calculated PMio emission rates (± 95% confidence interval) from the lidar and inverse
modeling using two dispersion models 84
Table 25. A comparison of lidar-based total calculated PMio emissions from the tillage activities
and estimated tractor exhaust, calculated based on fuel usage and an emission factor of 3.23 g
PMio/L fuel (Kean et al. [26]) 88
Table 26. A comparison of PMio emission rates herein derived and found in literature 90
Table 27. Lidar-derived particulate emissions, tillage rate, and fuel usage comparison between
conventional and combined operations tillage 91
Vlll
-------
EXECUTIVE SUMMARY
Airborne particles, especially fine particulate matter 2.5 micrometers (|im) or less in
aerodynamic diameter (PM^.s), are microscopic solids or liquid droplets that can cause serious
health problems, including increased respiratory symptoms such as coughing or difficulty
breathing, decreased lung function, aggravated asthma, development of chronic bronchitis,
irregular heartbeat, heart attacks, and premature death in people with heart or lung disease.
Concern with these effects resulting from local operations in agricultural areas is drawing
increased regulatory scrutiny and research. To investigate the control effectiveness of one of the
current San Joaquin Valley Air Pollution Control District Conservation Management Practices
(CMPs) listed for agricultural land preparation on the generation of particulate matter levels, the
U.S. EPA Environmental Sciences Division, National Exposure Research Laboratory was
awarded a Regional Applied Research Effort (RARE) project. The objectives of this study were
to discover:
1) What are the magnitude, flux, and transport of PM emissions produced by agricultural
practices for row crops where tillage CMPs are being implemented vs. the magnitude, flux, and
transport of PM emissions produced by agricultural practices where CMPs are not being
implemented?
2) What are the control efficiencies of equipment being used to implement the "combined
operations" CMP? If resources allow assessing additional CMPs, what are the control
efficiencies of the "equipment change/technological improvements" and "conservation tillage"
CMPs?
3) Can these CMPs for a specific crop be quantitatively compared, controlling for soil
type, soil moisture, and meteorological conditions?
This study used advanced measurement technologies such as atmospheric light detection and
ranging (lidar) systems coupled with conventional point measurement air quality samplers to
map PM emissions at high spatial and temporal resolutions, allowing for accurate comparisons of
the CMP under test. The purpose of this field study was to determine whether and how much
particulate emissions differ from the conventional method of agricultural fall tillage and a
"Combined Operations" Conservation Management Practice.
The test location and CMP to be evaluated were chosen in discussion with stakeholders,
regulatory agencies, and researchers. The RARE tillage experiment includes a fall tillage
sequence following the harvest of a row crop (corn, cotton, tomatoes, etc.), followed by a repeat
set of measurements at the spring harvest. The fall tillage site was near Los Banos, California
and consisted of two adjacent fields that were cultivated in cotton for the 2007 growing season
and were planned to grow similar crops in the 2008 growing season. The test fields were
adjacent on a north/south orientation. Conventional tillage operations were applied to the north
field, which contained 25.5 hectares. The combined operations tillage was applied to the south
field, which contained 51.8 hectares. Soil type distribution for both fields are dominated by soil
type 170 (Dos Palos clay loam, partially drained) and both contain small areas of soil type 103
-------
(Alros clay loam, partially drained). Both fields contain another soil type, 139 (Bolfar clay loam,
partially drained), isolated in the very corner of the north and on the eastern end the south field.
The conventional tillage method included two disc passes, separated by a chisel operation,
followed by a land plane pass. A precipitation event occurred before the land plane event, which
left the surface layer slightly moist reducing emissions for the land plane operation. The
Combined Operation CMP chosen for examination was the Optimizer1, which is designed to
perform the work of several pieces of equipment in one pass, thereby reducing the number of
passes and time spent on the tillage operation. The Optimizer was preceded by a chisel pass.
An extensive measurement system was applied during the study, which allowed two independent
methods of emission analysis, one using conventional methods, and the other using the direct -
near real-time measurements made with a lidar. Meteorological measurements included two 15.2
m towers instrumented at 5 heights with cup anemometers (2.5, 3.9, 6.2, 9.7 and 15.3 m) and
relative humidity/temperature sensors (0.9, 2.4, 3.7, 6.1, and 8.2 m) to provide profiles of wind
speed, temperature, and relative humidity. A wind vane was stationed at a height of 15.3 m on
each tower. Four sonic anemometers surrounding the site provided atmospheric turbulence data,
with two located at 11.3 m and two at 2.97 m elevations. Aerosol point samplers included both
filter and optical measuring techniques, with the filter samplers utilizing impactor heads for
aerodynamic separation. The filter-based samplers were arrayed in clusters to provide PMi,
PM2.5, PMio and TSP measurements at their various locations. The filter samplers ran for the
entire time of each operation and provided one sample per operation for both mass and chemical
analysis. The optical samplers were collocated with the filter-based samplers to provided aerosol
size distribution integrated over the 20-second sample periods. Two additional aerosol chemical
analysis systems were employed in a sampling trailer located on the downwind side of the field
under test. Samples were collected using sample ports on the upwind side of the trailer just above
and below the roof level. These samplers included an Organic Carbon/Elemental Carbon
Analyzer (OC/EC) sampling the PM2.s size fraction, and an Aerosol Mass Spectrometer (AMS).
The AMS provided chemical composition and particle size information for volatile and semi-
volatile particle components in the 0.1 - 1.0 jim size ranges in vacuum aerodynamic diameter.
Two EPA-approved models were used to assess emission rates based on the particulate sampler
data. They were the Industrial Source Complex Short-Term Model, version 3 (ISCST3) and the
American Meteorological Society/Environmental Protection Agency Regulatory Model
(AERMOD), which as of November 2005 is recommended for all regulatory applications [1][2].
Both models assume steady-state conditions, continuous emissions, and conservation of mass.
ISCST3 assumes a Gaussian distribution of vertical and crosswind pollutant concentrations
based on time averaged meteorological data. AERMOD uses continuous functions for
atmospheric stability determinations, and based on stability determines the appropriate
distribution: a Gaussian distribution for stable atmospheric conditions, and a non-Gaussian
distribution for unstable, or turbulent conditions. Final emission rates were determined using
inverse modeling coupled with observed facility-derived concentrations.
The Aglite lidar aerosol measurement and tracking system used for this experiment was a
scanning monostatic unit that uses a three-wavelength, Nd:YAG laser, emitting at 1.064 (3W),
0.532 (2W) and 0.355 (1W) um with a 10 kHz repetition rate. The lidar utilizes a turning mirror
turret to direct the beam to collect vertical aerosol slices upwind and downwind of the field to
-------
provide plume concentration images that are used to develop the 3D map of the source. This
data, when combined with wind information, enables direct assessment of the transport of
process generated aerosols off the field during the 1-minute scan time. A series of scans collected
during the operation are used to provide the time integrated emission information for each
operation. The lidar was calibrated against the point measurement systems (optical particle
counter for volume concentration and filter data for mass conversion information) for each
operation. The number of scans utilized to assess each operation varied from 47 to 122, with the
sample number reduced during some operations by light and variable wind conditions. Emission
values were determined by subtracting the upwind aerosol concentration from the downwind
aerosol concentration measured by each scan and multiplying by the wind speed.
Chemical analyses of filter catch for PM2.s, PMio, and TSP showed that measured upwind
(background) and downwind aerosols were composed of roughly the same concentrations of
ionic species and organic and elemental carbon. For both upwind and downwind sample
locations, all size fractions were dominated by an unknown mass fraction, which we attribute to
crustal sources. The nature of the data collection and analysis does not allow for separation of
local versus regional sources of collected crustal mass.
Emission data calculated for each measurement method for the conventional and conservation
tillage operations are presented. The study showed that the conservation practice under study
reduced the number of tillage passes by 50%, with similar reductions in fuel use and estimated
tractor associated PMio emissions.
The filter sampling/inverse model methods were severely challenged by the small fugitive dust
emissions with spatial and temporal variations encountered in this study and differences (based
on 67% confidence intervals not statistically significant) between average upwind and downwind
concentrations for several of the operations. However, the scanning lidar provided highly
significant emission measurements for all measurement periods.
Lidar-derived emission rates for PM2.5, PMio, and TSP by operation along with the average
tillage rate in hours per hectare were summarized from the experiment. The Combined
Operations tillage method reduced PM2.5 emission by 29%, PMio by 60%, and TSP by 25%.
Differences in total emissions per tillage treatment were significant at the 95% confidence
interval for PM25, PMio, and TSP. The time and fuel per hectare required to perform the
conservation tillage was about 40% and 50% of the conventional method. The control efficiency
of the Conservation Management Practice for particulate emissions was 0.289 ± 0.016, 0.604 ±
0.007, and 0.246 ± 0.013 for PM2.5, PMio, and TSP, respectively. The calculated tractor exhaust
emission amount was smaller than the uncertainty in the measurements indicating that the tractor
emissions were only a small part of the total aerosol emission from either operation. Tractor
emissions from the combined operations method were reduced by 50%.
The report compares the results from this study with results from previous studies found in
literature. The values from this experiment are in occasional agreement with those reported by
Flocchini et al. (2001) and Madden et al. (2008), as well as the emission factors used by CARB
to calculate area source PMio contributions from agricultural tilling [3][4][13]. While the values
from all three published studies are generally not in close agreement, they are well within the
range of the variability expected from measurements made under different meteorological and
-------
soil conditions, as demonstrated by the wide range of values in the literature. The largest
difference between the emission rates herein derived and those found in literature is that the PM
emission rates in literature for a land plane operation are much higher, up to 10 times greater,
than most other tillage operations, whereas the opposite was found in this study - the lowest
emission rates were calculated for the land plane operation. This is likely due to the presence of
residual moisture in the soil surface from a precipitation event that occurred 2 days earlier.
Therefore, the control efficiency of the Combined Operations tillage method is likely greater
under normal conditions.
-------
1. BACKGROUND
Airborne particles, especially fine particulate matter 2.5 micrometers (|im) or less in
aerodynamic diameter (PM^.s), are microscopic solids or liquid droplets that can cause serious
health problems, including increased respiratory symptoms such as coughing or difficulty
breathing, decreased lung function, aggravated asthma, development of chronic bronchitis,
irregular heartbeat, heart attacks, and premature death in people with heart or lung disease [5].
Larger particles tend to be removed from the air stream by the nose and throat before entering
the lungs [6]. The U.S. Environmental Protection Agency (U.S. EPA) has established limits for
PM2.5 and PMio (particles less than or equal to an aerodynamic diameter of 10 jim) levels in
order to protect public health as part of the National Ambient Air Quality Standards (NAAQS)
[7][8]. The U.S. EPA requires state air quality management agencies to monitor ambient PM2.5
and PMio concentrations for conditions hazardous to the population, report areas that exceed the
NAAQS beyond the allowed number of times, and establish procedures to reduce particulate
concentrations to meet the standards.
To address the problems associated with exposure to high particulate matter levels, the U.S. EPA
Environmental Sciences Division, National Exposure Research Laboratory was awarded a
Regional Applied Research Effort (RARE) project to determine the control effectiveness of
Conservation Management Practices (CMPs) for agricultural tillage using advanced
measurement technologies such as atmospheric light detection and ranging (lidar) systems. These
systems, when coupled with point measurement air quality samplers, can map PM emissions at
high spatial and temporal resolutions, allowing for accurate comparisons of various CMPs for a
variety of agricultural practices [9]. The purpose of this RARE project was to deploy an elastic
lidar system, together with a network of air samplers, to measure PM emissions from agricultural
operations in order to answer the following research questions:
1. What are the magnitude, flux, and transport of PM emissions produced by agricultural
practices for row crops where tillage CMPs are being implemented vs. the magnitude, flux, and
transport of PM emissions produced by agricultural practices where CMPs are not being
implemented?
2. What are the control efficiencies of equipment being used to implement the "combined
operations" CMP? If resources allow assessing additional CMPs, what are the control
efficiencies of the "equipment change/technological improvements" and "conservation tillage"
CMPs?
3. Can these CMPs for a specific crop be quantitatively compared, controlling for soil type,
soil moisture, and meteorological conditions?
The San Joaquin Valley was selected as the research site because it is one of five high priority
geographic areas for air quality identified in Region IX's Strategic Plan because of its size,
population, and extensive air pollution problems. In November 2008, EPA redesignated the San
Joaquin Valley to attainment for the PMio NAAQS, but sources in the Valley must continue to
implement measures that helped the District attain the PMio standard, including CMPs. The
Valley continues to violate the PM2.5 NAAQS. The CMP chosen for comparison against the
-------
conventional tillage method was the Combined Operations CMP, which is defined as a process
that combines equipment to perform several tillage operations in one pass [10]. Data collected
October 12-29, 2007 at a site near Los Banos, California are included in this report.
1.1 LITERATURE REVIEW
There are a handful of published articles pertaining to particulate matter emissions from
agricultural tilling, with the majority of the studies being performed in the state of California.
There are also several papers that collectively examine impacts of a variety of conservation
tillage practices with respect to soil characteristics, fuel consumption, cost of production, and air
emissions.
The use of an elastic lidar system by the University of California at Davis (UC Davis) to
examine dust plumes resulting from tillage activities was presented by Holmen el al.[9].
Qualitatively, the constructed system was able to track the plume emitted from the moving
source and provide a 2-D vertical, downwind map of the plume. It was observed that the plume
heights were often above the point samplers located at 10 m along the downwind plane. The
authors suggested that the best fugitive dust sampling procedures would include a combination
of elastic lidar and strategically placed point samplers.
Two papers by Holmen et al., presented in series in 2001, further discuss tillage PMio emission
rate investigations by UC Davis using filter-based mass concentration samplers and qualitative
measurements from the previously mentioned elastic lidar system [11][12]. The 24 samples
listed within the articles as being valid were collected from Fall 1996 to Winter 1998 in the San
Joaquin Valley during a wide range of environmental (temperature = 7-35 °C, relative humidity
= 20-90%, and from prior to the season's first precipitation to periods between winter storms)
and soil moisture conditions (1.5-20%). Tillage operations examined were discing, listing, root
cutting, and ripping. Calculated PMio emission rates ranged from 0 to 800 mg/m2 (0 to 6.9
Ib/acre), the mean ± one standard deviation was 152 ± 240 mg/m2 (1.4 ±2.1 Ib/acre), and the
median was 43 mg/m2 (0.4 Ib/acre). One point made by Holmen et al. [12] is that several
environmental conditions (temperature profile, relative humidity, soil moisture, etc.) can have
very significant effects on PM emissions and should be monitored and accounted for in emission
rate measurement and reporting. As a result, the reliability of direct comparisons of emission
rates measured under different environmental conditions must be carefully examined.
The studies published by researchers at UC Davis and herein previously discussed were part of a
much larger investigation of agricultural PMio emission rates in the San Joaquin Valley as
funded by the U.S. Department of Agriculture Special Research Grant Program. Findings of this
broad study are published in Flocchini et al. (2001) [3]. Table 1 presents emission factors
published in this study for different types of agricultural tillage along with the crop and time of
year. As seen in results measured by Holmen et al. [12], the emission factors reported by
Flocchini et al. for agricultural tillage were significantly influenced by environmental conditions,
such as the near-ground temperature profile, relative humidity, and soil moisture. The potential
variability with the same implement under opposing extreme environmental conditions may be
larger than the variation from the type of crop or equipment used for tilling.
10
-------
Table 1. Emission factors and uncertainties for land preparation as reported by Flocchini et al. (2001) [3].
Date
Emission Factor
(mg/m2)
Uncertainty
Date
Emission Factor
(mg/m2)
Uncertainty
Stubble Disc
10/27/1995
11/3/1995
11/3/1995
11/3/1995
11/3/1995
11/3/1995
11/3/1995
11/15/1995
11/15/1995
6/24/1997
257.7
49.3
27.4
231.0
136.7
140.8
286.1
537.9
542.2
430.0
NC
9%
470%
4%
7%
6%
5%
9%
125%
17%
11/6/1998
11/6/1998
11/6/1998
11/6/1998
11/6/1998
11/6/1998
11/6/1998
11/6/1998
11/6/1998
50.0
28.4
35.0
28.0
117.0
32.4
58.9
93.5
74.2
146%
145%
NC
10%
18%
9%
8%
9%
8%
Finish disc
11/26/1996
11/26/1996
11/26/1996
12/2/1996
124.3
142.4
97.5
91.0
3%
4%
5%
9%
12/4/1996
12/4/1996
12/4/1996
12/5/1996
9.2
0.6
3.5
-0.5
NC
NC
NC
NC
Ripping/chisel
6/24/1997
6/26/1997
6/26/1997
765.0
112.0
776.0
5%
5%
3%
6/25/1997
6/25/1997
331.0
577.0
5%
6%
Root cutting
11/16/1996
30.0
12%
11/16/1996
36.0
8%
The California Air Resources Board (CARB) developed their area source emission inventory
calculation methodologies for agricultural tillage and harvesting operations based on the report
by Flocchini et al. [3][13][14]. A summary of the resulting emission factors appears in Table 2.
The given unit for the emission factor can be explained as the mass of PMio particles released
per acre for each operation, or pass. Specific tillage operations were assigned to one of these five
categories and the displayed emission factor was used for all operations in each category.
Table 2. Emission factors used by the California Air Resources Board in estimating agricultural tilling PM10
emissions [13].
Agricultural Tilling Operation
Root cutting
Discing, Tilling, Chiseling
Ripping, Subsoiling
Land Planing & Floating
Weeding
Emission Factor
(Ib PM10/acre-pass)
0.3
1.2
4.6
12.5
0.8
(mg PM10/m2-pass)
33.6
134.5
515.6
1401.0
89.7
The U.S. EPA (2001) uses the empirically derived equation shown below to estimate the quantity
of particulate matter emitted from all agricultural tilling processes [15].
E=cxkxs' xpxa
(1)
where E = PM emission in Ibs, c = constant 4.8 Ib/acre-pass, k = dimensionless particle size
multiplier (TSP = 1.0, PMio = 0.21, PM2.5 = 0.042), s = silt content of surface soil (%), p =
11
-------
number of passes or tillings in a year, and a = acres of land tilled. The above equation was
developed to estimate TSP emissions (k = 1.0) and has since been scaled to estimate PMio and
PM2.5 emissions by using the respective k value. Average values of s are tabulated in Table 4.8-6,
U.S. EPA (2001) as a function of soil type on the soil texture classification triangle [15].
A comparison between standard tilling practices and conservation tilling (strip-till) in dairy
forage production on two farms in the San Joaquin Valley is given by Madden et al. (2008) [4].
Both strip-till and standard till operations were monitored for PMio emissions over two tillage
cycles at both farms. Results show that conservation tillage practices reduce PMio emissions
from one farm by 86% and 52% for 2004 and 2005, respectively. At the second farm,
conservation tillage emissions were reduced by 85% and 93% for 2004 and 2005, respectively.
Derived emission rates are presented in Table 3. Madden et al. attribute these reductions, in part,
to a reduced total number of passes (from 3-6 passes in standard tillage to 1 pass in conservation
tillage) and the ability for conservation tillage operations to be done under a higher soil moisture
content than standard operations.
Table 3. Conventional and conservation tillage emission rates reported by Madden et al. (2008) for tillage in a
dairy forage crop rotation [4]. ST= standard tillage method, CT = conservation tillage method.
Season/Year
Spring 2004
Spring 2005
Sweet Haven Dairy
Operation
ST: 1st discing
ST: 2nd discing (w/ roller)
ST: 3rd discing (w/roller)
ST: Planting
CT: Strip-tilling
CT: Planting
ST: 1st discing
ST: 2nd discing (w/ roller)
ST: 3rd discing (w/roller)
ST: Planting
CT: Strip-tilling
CT: Planting
Avg Emission
Factor (mg/m2)
198
1035
114
103
181
26
139
375
404
263
180
385
Barcellos Farms
Operation
ST: 1st discing
ST: 2nd discing
ST: Listing
ST: Bed discing
ST: Bed mulching
ST: Ring roller
ST: Planting
ST: Ring roller
CT: Planting
ST: 1st discing
ST: 2nd discing
ST: Circle harrow
ST: Listing
ST: Bed discing
ST: Bed mulching
ST: Planting
CT: Planting
Avg Emission
Factor (mg/m2)
259
917
615
25
89
566
96
104
394
51
123
337
466
109
384
481
130
Dust concentrations produced by agricultural implements used at a University of California
Davis research farm west of Davis, California were reported by Clausnitzer and Singer (1996)
[16]. Personal exposure samplers measuring respirable dust (RD) concentrations, particles that
may reach the alveolar region of the lungs when breathed in (with a 50% cut-point diameter of 4
|im), were mounted on implements in 22 different operations over 7 month period in 1994; only
the 18 operations with replicate samples were reported. Average RD concentrations measured on
the implement ranged from 0.33 mg/m3 for discing corn stubble to 10.3 mg/m3 for both land
planing and ripping operations. While RD concentration was heavily influenced by operation,
12
-------
other factors determined to be significant in dust production were relative humidity, air
temperature, soil moisture, wind speed, and tractor speed.
Further investigation of the data set presented by Clausnitzer and Singer (1996) [16] and of
another data set collected on a different University of California Davis research farm was
reported by Clausnitzer and Singer (2000) [17]. Both sets of data focus on RD concentrations as
measured on the agricultural implement and the analysis examined environmental influences on
the measured concentration. Again, soil moisture and air temperature were found to be
significant factors in RD production. The RD production with respect to soil moisture was well
fit by a power function, with the curve predicting RD concentrations becoming significantly
steeper below 5%. Air temperature was hypothesized to be significant in that it was a surrogate
measurement of atmospheric instability - as temperature increases near the surface, the
atmosphere becomes less stable and may carry greater quantities of dust upwards.
Baker et al. (2005) examined differences between dust concentrations resulting from standard
and conservation tillage practices in the San Joaquin Valley over a two year cotton-tomato crop
rotation, each under two different cover crop scenarios: 1) no cover crop and 2) a cover crop
forage mixture [18]. Total dust (TD), particles < 100 jim in aerodynamic diameter, and RD
samplers were stationed on the implements to collect samples in the plume. For both TD and RD,
the presence or lack of a cover crop in the standard till treatment did not seem to affect
concentrations. Summed concentrations for conservation tillage without a cover crop were about
one third of standard tillage and for conservation tillage with a cover crop they were about two
thirds for both dust fractions measured. Reductions in summed concentrations with conservation
tillage were attributed to fewer operations, including the elimination of the dustiest (discing and
power incorporation). When comparing operations common to all four treatments, tomato
planting and harvesting in conservation till produced higher concentrations than standard till
(thought to be due to increased organic matter on the surface) and concentrations during cotton
harvesting, which does not disturb the soil, were equivalent for all treatments. This study was
part of a larger effort to quantify the effects of conservation tillage in California on crop
production, soil quality, and time and resources dedicated to production as outlined by Mitchell
et al. (2008) and Veenstra et al. (2006) [19][20].
Upadhyaya et al. (2001) compared the Incorpramaster, a one-pass tillage instrument against a
conventional combination of discing twice and land planing twice based on fuel consumption,
timeliness, and effect on soil [21]. Studies on four experimental fields at UC Davis showed no
statistical difference between resulting soil conditions (bulk density changes, soil moisture
changes, and aggregate size), but the Incorpramaster used between 19 and 81% less fuel with a
mean of 50%. The time savings ranged from 67 to 83% with a mean of 72%. In most cases, two
passes with the Incorpramaster were required to achieve the same soil conditions as the four
passes in the conventional till.
Three conservation tillage methods were compared against the standard tillage method in cotton
production and reported in Mitchell et al. (2006) in terms of yield, yield quality, tractor passes,
fuel, and production costs [22]. A single field near Fresno, CA was divided in area among seven
tillage treatments: 1) standard, 2) no till/chop, 3) no till, 4) ridge till/chop, 5) ridge till, 6) strip
till/chop, and 7) strip till. Prior to both cotton growing seasons examined, a small grain wheat
was planted in the field to enhance soil properties; this crop was sprayed with herbicide, and in
13
-------
treatments 2, 4, and 6 it was chopped with a mower prior to tillage activities. In the other
treatments (1, 3, 5, and 7), the dead wheat was either incorporated by the tilling or left standing.
Yield and yield quality were statistically the same for both years for all treatments, though the
standard treatment was numerically higher the 2n year. Conservation tillage treatments reduced
tractor passes by 41% to 53% over the standard method and estimated fuel reduction was 48% to
62% for the conservation practices. The estimated overall production costs of the conservation
tillage systems were 14% to 18% lower than the conventional system. Mitchell et al. estimated,
by extrapolation from other work, that whole-tillage process particulate matter emissions would
also be decreased.
Particulate matter is released during agricultural tillage activities from both the operational
activity of the tillage implement, as well as the tractor in use. Emissions from the tractor mainly
originate from the tires in contact with the soil and the combustion engine. Attempts to quantify
the PM emitted in agricultural tractor exhaust have been made by the U.S. EPA and the
California Air Resources Board in software designed to estimate off-road engine emissions on a
county or regional scale. The U.S. EPA developed the NONROAD software program, with the
latest version distributed in 2005 [23]. Emission factors from compression ignition (diesel)
engines used in the NONROAD model are calculated by adjusting a zero-hour, steady state
measured emission factor (EFSS) for engine deterioration with operation time (DF) and a transient
adjustment factor (TAP) that accounts for variations from steady state engine loading and speed,
as shown in the following equation [24].
EFadj(PM) = EFSS x TAF x DF - SPMadj (2)
where EFadj(pM) is the adjusted PM emission factor and SpMadj is the emission factor adjustment
accounting for the use of a diesel fuel with a sulfur content different than the default
concentrations, as fuel sulfur level is known to affect PM emissions. The units for EFadj(pM), EFSS,
and SpMadj are g/hp-hr, where hp stands for horsepower, and TAF and DF are both unitless. All
four variables on the right side of Eq. 2 vary with model year and engine size, expressed in
horsepower (hp), according to measured values and/or the emission standards each model year
and engine size was designed to meet. The selection of values for steady state emission factors
and all the adjustment variables given in U.S. EPA (2004) [24] was performed using a variety of
tests and resources, including the Nonroad Engine and Vehicle Emission Study (NEVES) Report
[25], or by setting the values such that the adjusted PM emissions were equal to model year-
specific emission standards.
The California Air Resources Board (CARB) has also developed a model to forecast and
backcast daily exhaust emissions from off-road engines, including agricultural tractors, called
OFFROAD. Similarly to the U.S. EPA's NONROAD model, emission factors (EF) for each
engine size and model year are calculated based on a zero-hour emission rate (ZH) with a
deterioration factor (DF) applied to account for engine wear with use, as in Eq. 3. The derived
emission factor is then multiplied by the load factor (LF), the maximum rated average
horsepower (HP), and the amount of time the engine is active through the year (Activity) in
hr/yr.
EF = ZH + DR*CHrs (3)
14
-------
P = EF * HP *LF* Activity *CF (4)
where CHrs is the cumulative engine operation hours, P is the amount of pollutant released in
tons/day, and CF is the conversion factor from units of grams per year to tons per day. The
values for the EF and DR are derived from measured values or they are calculated based on
requirements to meet the proposed emissions limits for future years [26].
Kean et al. (2000) estimated off-road diesel engine, locomotive, and marine vehicle emissions of
NOX and PMio for 1996 based on fuel sales [27]. Diesel engine exhaust emission factors were
developed based on information provided in the development of the U.S. EPA NONROAD off-
road vehicle emissions model with supplemental information in order to calculate emissions
based on fuel consumption. A fleet-wide average PMio emission factor was determined for farm
diesel equipment to be 3.8 g/kg of fuel used, at an average mass per volume of 0.85 kg/L of
diesel fuel. Fuel sales surveys from 1996 were used to calculate regional and national emissions.
In the off-road category, which includes farm equipment, the U.S. EPA NONROAD model
calculated on average 2.3 times higher emissions, which was attributed to higher engine activity
assumed in the EPA model than represented in the reported fuel sales data.
2. EXPERIMENT DESIGN
2.1 SITE DESCRIPTION
After discussion among stakeholders, regulatory agencies, and researchers, it was collectively
determined that the appropriate type of tillage for this experiment would be the fall tillage
sequence following the harvest of a row crop (corn, cotton, tomatoes, etc.). Fall tillage activities
prepare the ground for planting the next season's crop. An appropriate site was identified near
Los Banos, California (see Figure 1) and consisted of two adjacent fields. Both fields were
cultivated in cotton for the 2007 growing season, and plans indicated that both would receive
similar crops in the 2008 growing season. Upon arrival at the site, immediately prior to the study,
both fields had been harvested and the cotton stalks had been shredded and left on the ground.
An aerial photo and soil classification map of the experimental site is shown in Figure 2. This
photo was extracted from the Merced County Soil Survey that was completed by the U.S.
Natural Resources Conservation Service (NRCS). The two fields are delineated in the photo by
red outlines. The field to the north, labeled Field A, is 254,600 m2 (62.9 acres) in area. The field
to the south, labeled Field B, is 518,400 m2 (128.1 acres). Soil type distribution is also indicated
in Figure 2 by orange lines separating different soil classifications. Both fields are dominated by
soil type 170 (Dos Palos clay loam, partially drained) and both contain small areas of soil type
103 (Alros clay loam, partially drained). Both fields contain another soil, soil type 139 (Bolfar
clay loam, partially drained) but it is isolated in the very corner of Field A and on the eastern end
of FieldB [28].
The experiment fields were surrounded on all sides by other cultivated fields, dirt access roads,
irrigation ditches, and drainage ditches. The majority of the land in the surrounding area is used
for agricultural purposes. The crops grown in the surrounding fields were cotton, tomatoes, and
alfalfa hay. Tomatoes had already been harvested and the ground had been tilled. The cotton
15
-------
100 KM 100 Miles
Figure 1. Shaded relief map of the State of California, USA, with the location of the selected sample site
shown by the white star. Image from geology.com [29].
fields were in various stages of harvest with some fields already harvested and tilled in
preparation for the next crop. Fields already harvested were bare of most standing vegetation.
Alfalfa hay was still being grown in some fields to the southeast of the site. The dirt roads were
mainly used for access to the fields and were only occasionally traveled by farm vehicles. The
ditches did not appear to be at operational flows and some appeared to contain little or no water.
The terrain surrounding the fields was relatively flat for many miles in all directions. The main
form of topographical relief was provided by the drainage and irrigation ditches, with banks of
varying elevations higher than the fields and channel bottoms which were all lower than the
agricultural fields. Several hundred meters to the north of field A was a long row of trees that ran
16
-------
Ao M: ioa eoc 5.200
rat
-------
Figure 3. Photo taken on October 19, 2007 standing near the southern edge of Field B looking north across
fields B and A during the chisel pass of the combined operations tillage method.
2.2 OPERATION DESCRIPTION
The purpose of this field study was to determine if and how much particulate emissions differ
between the conventional method of agricultural fall tillage and a "Combined Operations"
Conservation Management Practice. The Combined Operation CMP chosen for examination was
the Optimizer1, which is designed to perform the work of several pieces of equipment in one
pass, thereby reducing the number of passes and time spent on the tillage operation. The farm on
which the study was performed has been using the Optimizer for all of its fall tillage for several
years, including on both fields herein studied. Prior to that, all fields were tilled using the
conventional method [30].
The conventional tillage method was employed in field Field A and the combined tillage
operations were used in Field B. The operations performed in each method are shown in order in
Mention of a specific tradename or manufacturer does not imply endorsement or preferential treatment by the
USDA-ARS or Space Dynamics Laboratory or Utah State University.
18
-------
Table 4, with their corresponding dates. It should be noted that in the conventional tillage
method, the second pass with the disc (Disc 2) was monitored twice due to tillage equipment
malfunctions on the first day (Disc 2A), causing long periods of inactivity during the sampling
period. On the second day, Disc 2B, the rest of the field not tilled during Disc 2A was finished;
there was minimal overlap of the two areas tilled between Disc 2A and 2B. However, it should
be noted that a third disc pass between the second disc pass and the land plane pass may be
performed, depending on the resulting soil conditions from the previous passes.
Table 4. Tillage operations and dates performed for the comparison study.
Sequence
Operation
Date
Combined Operations Tillage (Field B)
1
2
Chisel
Optimizer
10/19/2007
10/20/2007
Conventional Tillage (Field A)
1
2
3
4
5
Disc 1
Chisel
Disc 2A
Disc 2B
Land Plane
10/23/2007
10/25/2007
10/26/2007
10/27/2007
10/29/2007
In this study, the use of the Optimizer reduced the number of passes by two, not counting the
extra day in the conventional tillage due to equipment malfunctions. However, it is possible in
some circumstances, based on field conditions, to eliminate the chisel pass before the Optimizer
and further reduce the number of tillage passes. Therefore, the Optimizer may potentially reduce
tillage passes by 2-4 in comparison with conventional practices.
The tractors and implements used during all the tillage operations are listed in Table 5 by date
and operation. During all but the land plane operation, two tractors were pulling identical
implements, with one usually providing a straight line based on an on-board global positioning
system (GPS) unit for the second operator. Both tractors pulling the Optimizer implement,
studied on 10/20, were equipped with on-board GPS units. A separate handheld GPS unit was
placed on the tractor designated as Tractor 1 for each day and actively logged the tractor's
position over time. It should be noted that on 10/23, 10/25, and 10/26 there were mechanical
problems with one or both of the tractors that resulted in periods during each of those sample
runs where there was only one or no tractors operating. On 10/23, another tractor, Tractor 3, was
used to replace Tractor 2 after mechanical problems required it to stop operating for the duration
of the experiment.
19
-------
Table 5. Equipment used on the fields to perform the tillage operations.
Operation
Chisel
Optimizer
Disc 1 *
Chisel
Disc 2A
Disc 2B
Land
Plane
Date
10/19/2007
10/20/2007
10/23/2007
10/25/2007
10/26/2007
10/27/2007
10/29/2007
Tractor
l)John Deere 9520 w/ rubber track
2)John Deere 9400 w/ rubber track
l)John Deere 9620 w/ rubber track
2)John Deere 9620 w/ rubber track
1 John Deere 9520 w/ rubber track
2John Deere 9400 w/ rubber track
3 John Deere 9320 I/ 8 tires
1 John Deere 9520 w/ rubber track
2John Deere 9400 w/ rubber track
1 John Deere 9520 w/ rubber track
2John Deere 9400 w/ rubber track
1 John Deere 9520 w/ rubber track
2John Deere 9400 w/ rubber track
1 John Deere 7920
Implement (1 per tractor)
Schmeiser Chisel, 5 shank, 26" depth, 16' wide
pulling a 16' wide spiked ring roller
Optimizer 5000
Rome Full Stubble Disc, 32" blade, 16' wide
Schmeiser Chisel, 5 shank, 26" depth, 16' wide
pulling a 16' wide spiked ring roller
Rome Full Stubble Disc, 32" blade, 16' wide
Rome Full Stubble Disc, 32" blade, 16' wide
Schmeiser Tri-Plane, 24' long, 16' wide
* Note: the 9400 ran for ~ 15 minutes, then the 9320 came to replace it using the same disc set
The Optimizer, manufactured by Tillage International (Turlock, CA), is an agricultural tillage
implement that incorporates all forms of conventional tillage in a single pass. It combines
multiple soil preparation processes into one. It uses the following tools in order from front to
back: 1) two rows of inline disc units, 2) two rows of flow control reels or coulters, one after
each row of disc units, 3) optional third row of reels, rollers, or coulters 4) axle assembly with
off-set walking beam assembly, 5) three rows of off-set chisel shank and/or spring tooth
assemblies, 6) two rows of flow control baskets, 7) optional roller assemblies for light or heavy
soil, 8) optional planting unit for forage crops, and 9) optional unit for injection or application of
fertilizer or herbicide. The flow control reels perform chopping, shattering clods, and
incorporating activities. Also, attached to the axle assembly is a leveling system for consistent
function of components, which helps the Optimizer to also perform the function of a land plane
in leveling the field. Two models are available, depending on crop and horsepower requirements.
Figure 4 shows an Optimizer Model 5000 unit, the larger of the two models, with the optional
roller assemblies attached as utilized in this experiment, and Figure 5 shows one of the two
identical units in operation [31].
20
-------
Figure 4. Photograph of a stationary Optimizer Model 5000 with attached roller assemblies.
Figure 5. An Optimizer Model 5000 in operation during this field experiment, with an instrumented tower in
the background.
21
-------
Field personnel observed operations continually and recorded notes on tractor operation times,
potential contamination issues due to traffic on surrounding dirt roads, general meteorological
observations, etc.
The cooperating farming company recorded fuel usage and tractor run time, as shown in Table 4,
throughout the field study time period in order to examine tractor time and fuel consumption
differences between the conventional and combined operations methods [31]. Tractor run times
were recorded by SDL employees and included periods of time during which the tractors were
not moving for servicing, break downs, or operator break times. Fuel usage was measured by a
volume-calibrated meter on the service truck and recorded in the service log by the attending
employee. The service log and fuel meter were verified monthly for accuracy.
Table 6. Tractor run time and fuel usage as recorded by the farming company.
Operation
Date
Total tractor time (hr)
Total fuel used
(gallons)
(liters)
Combined Operation Tillage
Chisel
Optimizer
10/19
10/20
Sum
8.5
4.36
12.76
90.0
46.2
136.2
340.7
174.9
438.2
Conventional Tillage
Disc 1
Chisel
Disc 2
Land Plane
10/23
10/25
10/26-27
10/29
Sum
11.0
6.5
9.16
3.33
29.99
104.6
58.5
82.5
26.6
272.2
396.0
188.2
265.5
85.6
875.8
2.3 TILLAGE OPERATION DATA
Based on field notes and GPS data points, the total tractor run time in tractor hours (hrtractor), or
the sum of individual tractor operation times, and the area tilled per day were calculated. This
tractor run time is smaller than that reported by the cooperating farming company because it only
includes times when the tractor was moving and performing the specified operation in the field.
The tractor hours reported therefore do not include the time the tractor spent motionless with an
idling engine. From these two numbers, the tillage rate (hectares/hrtractor) and the operation rate of
the tractors (hrtractor/hectares) were calculated. Tractor operation rates were summed to provide
the total amount of time per hectare spent preparing the ground for the next season's crops
(Table 7). In this study, the tillage rate of the combined operations was 0.59 hrtractor/hectare and
the conventional tillage rate was about 2.5 times that amount at 1.56 hrtractor/hectare. It should be
noted that these numbers only include times when the tractors and implements were tilling, and
that the downtimes mentioned earlier due to mechanical failures are not included.
22
-------
Table 7. Operation data for both the conventional and combined operation tillage studies as recorded by field
personnel.
Operation
Date
Total Tractor
Time
(hrtractor)
Area
tilled
(hectares)
Tillage Rate
(hectare/hr^eto,.)
Tractor Rate
(hrtractor/hectare)
Combined Operation Tillage
Chisel
Optimizer
10/19
10/20
8.5
4.25
22.0
20.0
Sum
Average
2.6
4.7
3.7
0.38
0.21
0.59
Conventional Tillage
Disc 1
Chisel
Disc 2A
Disc 2B
Disc 2
Land
Plane
10/23
10/25
10/26
10/27
10/26-
27
10/29
11.0
6.5
3.4
5.75
9.16
3.33
24.8
19.5
10.5
14.2
24.7
8.0
Sum
Average
2.3
3.0
2.7
2.4
2.6
0.44
0.33
0.37
0.42
1.56
Fuel-use Rate
(liter/hectare)
15.5
8.7
24.2
16.0
9.7
10.7
10.7
47.1
Due to the breaks in tractor run time and the presence of multiple tractors in most cases and
single tractors in others, the ratio of the sample period length and total tractor operation time was
slightly different for each day, as shown in Table 8 below. The tractor problems on 10/26 caused
an unusually high sample time to tractor operation time ratio. The difference between total
tractor operation time and sample period time is important because the source strength will also
vary based on how many tractors, if any, are operating at a given time. All calculations of
emission rates herein undertaken have accounted for this difference in source strength with time,
with final emission rates based on time being reported as the emission rate of a single tractor.
Table 8. Sample period, total tractor operation time, and the sample period-to-tractor operation ratio for all
sample periods.
Date
10/19
10/20
10/23
10/25
10/26
10/27
10/29
Sample Time (hr)
5.33
2.85
7.27
4.24
5.52
4.09
3.49
Total Tractor Time (hr)
8.50
4.25
11.00
6.50
3.41
5.75
3.33
Sample time/Tractor time
0.63
0.67
0.66
0.65
1.62
0.71
1.05
23
-------
3. MEASUREMENTS AND METHODS
3.1 OVERVIEW
A wide variety of air quality and meteorological sampling equipment was employed during this
field study to meet the experiment objectives. These instruments are described in the following
sections along with their functions, data analysis procedures, and calibration verification
procedures which were performed before, during, and after the campaign to ensure the accuracy
of data collected.
As wind direction and wind speed are important factors in obtaining accurate and representative
data from the sampling systems to meet the project objectives, it was necessary to determine the
dominant wind direction for this period of year in the Los Banos area. Meteorological data from
the California Irrigation Management Information System (CEVIIS) database were downloaded
for October of 2004 through 2006 for the Los Banos station (#56) [33]. Based on these data the
dominant wind direction was determined to be from the west to the northwest, as shown in
Figure 6.
'NE
Wind Speed, mte
• *
5-6
I«
"
2-3
™
\
E
SW
"-,
<'10.0%
SE
Figure 6. Wind rose for September and October of 2004 - 2006 as recorded by the CIMIS Station # 56 (Los
Banos).
24
-------
Due to the orientation of the fields, the preferred dominant wind direction for sampling purposes
was from the northwest. Based on a northwest wind, instrument deployment was such that
samplers meant to measure background aerosol parameters were to the north and northwest of
the fields of interest and samplers meant to measure background plus plume parameters were to
the south and east. Due to the large size of each field, one sample layout per field was utilized
during the study. The first layout, shown in Figure 7, accommodated the sampling that took place
during the combined operations tillage in Field B on October 19th and 20th. Tillage in Field B
was completed prior to tillage being performed in Field A, the northern field. The flags indicate
instrumentation locations with location names as labeled. Table 9 summarizes the type of
instruments that were located at each site.
Figure 7. Sample layout used for Field B, with the area tilled shown by the shaded polygon. Created using
Google Earth software.
25
-------
Table 9. Summary of instruments located at each site for the combined operations tillage study of Field B. All
height given as above ground level (agl).
Instrument
Location
SI
SMetl
S2
S3
El
Nl
N2
NMet
Ul
Wl
Tethersonde
Lidar 1
AQ1
Description
1
2
4
1
5
1
5
2
1
1
1
2
6
1
1
4
1
1
2
1
1
1
2
6
1
2
1
5
1
5
2
1
1
1
2
1
2
1
1
1
1
1
1
2
1
1
1
1
-10 m tower
- OPCs @ 2 and 9 m
- MiniVols: PM10 and PM25 @ 2 and 9 m
-15 m tower
- cup anemometers (2.5, 3.9, 6.2, 9.7 and 15.3 m)
- wind vane @ 15.3 m
- temp/RH sensors (0.9, 2.4, 3.7, 6.1, and 8.2 m)
- Campbell Scientific dataloggers
- sonic anemometer @ 1 1.3 m
- energy balance system @ 2 m
-10 m tower
- OPCs @ 2 and 9 m
- MiniVols: TSP, PM10, PM25, and PMj @ 9 m; PM10 and PM25 @ 2 m
-10 m tower
- OPC @ 9 m
- MiniVols: PM10 and PM25 @ 9 m; PM10 and PM25 @ 2 m
-10 m tower
- OPC @ 9 m
- MiniVols: PM10 and PM25 @ 9 m
- sonic anemometer @ 2.7 m
- Campbell Scientific datalogger
-10 m tower
- OPCs @ 2 and 9 m
- MiniVols: TSP, PM10, PM25, and PMj @ 9 m; PM10 and PM25 @ 2 m
-2m tripod
- MiniVols: PM10 and PM25 @ 2 m
-15 m tower
- cup anemometers (2.5, 3.9, 6.2, 9.7 and 15.3 m)
- wind vane @ 15.3 m
- temp/RH sensors (0.9, 2.4, 3.7, 6.1, and 8.2 m)
- Campbell Scientific dataloggers
- sonic anemometer @ 1 1.3 m
-10 m tower
- OPC @ 9 m
- MiniVols: PM10 and PM25 @ 9 m
-10 m tower
- MiniVols: PM10 and PM25 @ 9 m (PM25 stopped working on 10/20/2007)
- sonic anemometer @ 2.97 m
- Campbell Scientific datalogger
- tethersonde data collection package with a MadgeTech PRHT sensor
- Lidar data collection system
- Davis met station for lidar operator's reference
- OPC @ 5 m
- MiniVols: PM10 and PM25 @ 5 m
- Davis met station @ 5 m
- OC/EC Analyzer (inlet @ 4.5 m)
- AMS (inlet @ 4 m)
- radio and laptop for OPC Data collection
26
-------
For the first portion of the study, while monitoring the combined operations method of tillage,
the AgLite lidar was located at position lidar 1 and used the tower Ul as the upwind reference
point. Downwind scans were made along the line of sampling towers SI, S2, and S3 and the
trailer housing other air quality sampling instruments, AQ1.
To capture the particulate emitted by various tillage operations, the scanning lidar sampled
vertical profiles from the upwind (entering aerosol mass) and downwind (exiting
particulate/aerosol mass) sides of the tillage field in combination with horizontal scans from up-
to down-wind position. These fast horizontal scans were used to monitor the height of the flux
box to make sure that no source-emitted particulate transport passes through the top of the
'staple'. However, the scanning ability of the lidar system during this field study was limited
according to U.S. Federal Aviation Administration (FAA) requirements communicated to SDL
through a letter in response to our request for permission to operate a laser at this location. This
letter permitted the use of a laser beam near ground level only at 0° N and 90° E and that all
horizontal scans had to be made at > 13° from horizontal. In accordance with these requirements,
the scanning 'staple' used throughout all the tillage operation sample periods consisted of an
upwind vertical scan at 0° N lasting 73 seconds, an elevated horizontal scan to move to the
downwind side (19 sec), two downwind vertical scans (2x73 sec = 146 sec) at about 90° E, and
another elevated horizontal scan (19 sec) to return to the upwind location to return to the start of
the sequence. The total time per 'staple' scan was 257 seconds.
To assure the quality of PM mass concentration retrievals from lidar observations, routine 'stare'
modes were programmed into the lidar measurement profile. This quality assurance step was
repeated every three 'staple' scans and held stationary for 60 sec each time.
After completing the tillage in Field B, the sampling equipment along the south, east, and west
sides of Field B were moved to equivalent locations surrounding Field A to the north in order to
monitor air quality during the conventional tillage method (samples taken from October 23-29,
2007). This second sample array is shown in Figure 8, with the flags again indicating sample
locations and the location names as labels. Table 10 summarizes the type of instruments that
were located at each site.
27
-------
Figure 8. Layout used during tillage of Field A, with the area tilled shown by the shaded polygon. Created
using the Google Earth software.
Table 10. Summary of instruments located at each site for tillage study of Field A. All heights given as agl.
Instrument
Location
S4
SMet2
S5
S6
Description
1-10 m tower
2 - OPCs @ 2 and 9 m
4 - MiniVols: PM10 and PM25 @ 2 and 9 m
1-15 m tower
5 - cup anemometers (2.5, 3.9, 6.2, 9.7 and 15.3 m)
1 - wind vane @ 15 m
5 - temp/RH sensors (0.9, 2.4, 3.7, 6.1, and 8.2 m)
2 - Campbell Scientific dataloggers
1 - sonic anemometer @ 11.3 m
1 - energy balance system @ 2 m
1-10 m tower
2 - OPCs @ 2 and 9 m
6 - MiniVols: TSP, PM10, PM25, and PMj @ 9 m; PM10 and PM25 @ 2 m
1-10 m tower
1 - OPC @ 9 m
4 - MiniVols: PM10 and PM25 @ 9 m; PM10 and PM25 @ 2 m
28
-------
Instrument
Location
E2
Nl
N2
NMet
U2
W2
Tethersonde
Lidar 2
AQ2
Trl*
Description
1-
1-
2-
1-
1-
1-
2-
6-
1-
2-
1-
5-
1-
5 -
2-
1-
1-
1-
2-
1-
2-
1-
1-
1-
1-
1-
1-
2-
1-
1-
1-
1-
2-
the
10m tower
OPC @ 9 m
MiniVols: PM10 and PM25 @ 9 m
sonic anemometer @ 2.7 m
Campbell Scientific datalogger
10 m tower
OPCs @ 2 and 9 m
MiniVols: TSP, PM10, PM25, and PMl @ 9 m; PM10 and PM25 @ 2 m
2 m tripod
MiniVols: PM10 and PM25 @ 2 m
15 m tower
cup anemometers (2.5, 3.9, 6.2, 9.7 and 15.3 m)
wind vane @ 15.3 m
temp/RH sensors (0.9, 2.4, 3.7, 6.1, and 8.2 m)
Campbell Scientific dataloggers
sonic anemometer @ 11.3 m
10m tower
OPC @ 9 m
MiniVols: PM10 and PM25 @ 9 m
10 m tower
MiniVols: PM10 and PM25 @ 9 m (PM25 stopped working on 10/20/2007)
sonic anemometer @ 2.97 m
Campbell Scientific datalogger
tethersonde data collection package with a MadgeTech PRHT sensor
Lidar data collection system
Davis met station for lidar operator's reference
OPC @ 5 m
MiniVols: PM10 and PM25 @ 5 m
Davis met station @ 5 m
OC/EC Analyzer (inlet @ 4.5 m)
AMS (inlet @ 4 m)
radio and laptop for OPC Data collection
MiniVols: PM10 and PM25 @ 2 m (* temporary location for sampling on 10/27 due to
area of the field being worked being largely to the west of most downwind towers)
The lidar was located at position lidar 2 for this second portion while the conventional tillage
method was being monitored. From this location, the upwind reference tower was U2 and the
lidar was again able to scan along the downwind side in line with the towers S4, S5, and S6 and
the air quality sampling trailer at AQ2. The same lidar 'staple' and 'stare' sequences were used
at the first location, lidar 1.
3.1.1 Meteorological Measurements
A tethersonde system from Atmospheric Instrumentation Research, Inc. (Boulder, CO) was
employed to provide vertical wind speed, wind direction, temperature, humidity, and pressure
profiles. The tethersonde meteorological package was a Model TS-3A-SP, which transmits 10
second averaged data to a receiver package (Atmospheric Data Acquisition System Model AIR-
3A) at ground level. Data were collected and stored on a laptop computer connected to the
receiver. Due to malfunctioning temperature and pressure sensors, the tethersonde package was
retrofitted with a sensor/datalogger from MadgeTech (model PRHTemp 101, Contoocook, NH).
The PRHTemp 101 averages pressure, temperature, and relative humidity at a user specified
29
-------
interval and logs the data. For this experiment, the data averaging interval was 10 seconds to
match the tethersonde averaging time. The PRHTemp 101 was launched and data were
downloaded at the end of each run on the same computer that stored the tethersonde data to
ensure that the time stamps would be synchronized. A helium filled balloon tethered to a
manually operated winch lifted the tethersonde package to heights of about 460 m above ground
level.
A Vantage Pro2 Plus weather station from Davis Instruments, Inc. (Hayward, CA) was used to
monitor wind speed, wind direction, temperature, humidity, precipitation, barometric pressure,
and solar radiation. It was located on top of the Air Quality sampling trailer, at an approximate
height of 5 m above ground level. It was wired to a datalogger and display panel inside the
trailer, which was connected to a computer for data storage.
Two 15.2 m towers were instrumented with 3-cup anemometers (model 12102) at 5 heights (2.5,
3.9, 6.2, 9.7 and 15.3 m) to provide the vertical wind speed profiles. Relative
humidity/temperature sensors (Vaisala HMP45C) from Campbell Scientific, Inc. (Logan, UT)
were also stationed at 5 heights (0.9, 2.4, 3.7, 6.1, and 8.2 m) to provide profiles of temperature
and relative humidity. A Met One Instruments, Inc. (model 024A, Grants Pass, OR) Wind Vane
was stationed at a height of 15.3 m on each tower. Data from each tower were stored as one
minute averages on Campbell Scientific, Inc. (Logan, UT) CR23X dataloggers and were
downloaded daily.
3.1.1.1 Eddy Covariance Measurements
The transport of mass, energy and scalars between a surface and the overlying layer of the lowest
part of the atmosphere (boundary layer) overwhelmingly are dominated by turbulence as
opposed to diffusion. Appropriate characterization of the turbulence transport of any mass/scalar
requires high frequency response sensors that measure the 3-dimensional components of the
wind field and measurements of mass/scalar of interest. In this study eddy covariance (EC)
systems were mounted on 4 towers surrounding the field of interest to measure components of
the turbulent flow field in conjunction with the particulate concentration measurements. EC
systems were mounted on the following towers and at the specified height above ground level, as
presented in Tables 6 and 7: Wl and W2 = 2.97 m agl, El and E2 = 2.7 m agl, and NMet,
SMetl, and SMet2 = 11.3 m agl. As can be seen two heights were near the surface (E and W
towers) and two were substantially higher (N and S meteorology towers). This was done to
capture a quasi-vertical profile of the turbulence characteristics for this particular flow field, one
near the surface and the other higher above.
The EC instrumentation was comprised of four Campbell Scientific Inc. (Logan, UT) 3-
dimensional sonic anemometers (CSAT) and four LiCOR 7500 infrared gas analyzers (IRGA).
The sensor separation for all four EC systems was 10 cm. Sampling rate was 20 Hz, all data were
stored on to a Campbell Scientific data logger (CR5000). Together the CSAT and LiCOR
measure water vapor (q\ carbon dioxide (c) concentrations and velocity components of the wind
flow in three spatial dimensions x, y, and z. In meteorology wind components in the x, y and z
directions are defined as streamwise direction u, lateral direction v and in the vertical w
respectively. These measurements were made at a scan rate of 20 Hz, 20 measurements per
second for it, v, w, q and c. All of the high frequency data were preserved for subsequent post
30
-------
processing of fluxes of latent heat (LE, evaporation), sensible heat (H), carbon dioxide (c) and
momentum (uw and vV ). Each tower was visited daily between 06:00-07:00 hours for
maintenance. The maintenance visit included exchanging compact flash cards for data storage on
the CR5000, sensor interrogation at the data logger screen to evaluate measurement status,
cleaning dust from the surface of the IRGA lens with de-ionized water and removing cob weds
from the transducer arms.
The term "flux" used herein follows the definition in physics as the number of changes in
mass/energy flow across a given surface per unit area per unit time. As represented here uw and
vw are covariances of the streamwise (u) and lateral (v) velocities with the vertical (w) velocities
which is a momentum flux. The term covariance is a statistical measure of the variance of any
two random variables observed or measured in the same mean time period. This measure is equal
to the product of the deviations of corresponding values of the two variables from their
respective means. The superscript (') represents the instantaneous velocity departures (which are
the turbulence) from the mean velocity of each of the components. The over bar ( )is a time
averaged operator and can in theory represent any time averaged period. The more common
averaging period is typically about 30 minutes. In this study we evaluate multiple time averaging
periods ranging from 1-30 minutes. In addition to the fluxes there are a large number of
statistical parameters that provide important insights to the governing processes of particulate
transport to the boundary layer which is the key issue in this study.
Specifically we focus on friction velocity (w*), the standard deviation of the vertical wind
velocity (crw), mean wind speed and direction. Collectively these parameters are used to develop
an understanding of how spatial and temporal changes of the turbulent flow field can affect the
particulate emissions from agricultural production activities. Through this understanding
improved model algorithm development can proceed by incorporating parameters that are
directly relevant to the turbulent transport of particulates during tillage operations. The critical
component is to develop a defensible approach to compute emission loading into the boundary
layer and to this end must include an operational parameterization for turbulence.
Additionally, the flux of water vapor (E) from the fields of interest was estimated using Eq. 5 in
order to determine if and how much water contributed by precipitation events was present in the
top layer of soil at the time of the tillage activity following the precipitation.
E = Wpv (5)
9 1
where the units for the flux of water vapor are in kg m" s" , w is the vertical wind, and pv is
water vapor density. Again, the superscripts (') represent the instantaneous deviation from the
average and the overbar ( ) denotes a time average. This expression is identical to the covariance
of these two properties. The raw fluxes were determined for 30 minute averages. Various
corrections were made to the initial values. A traditional coordinate rotation (Tanner and
Thurtell, 1969; and Lee et al., 2004) was performed to account for tilt errors of the anemometer,
and align the x-axis with the mean wind direction [34][35]. Then the correction reported by
Webb et al. (1980) was made for effects of density fluctuations caused by water vapor and heat
transport [36].
31
-------
3.1.2
Wind Profile Calculations
Wind profiles near the surface were calculated based on one minute averaged wind speed data
from the logarithmically spaced cup anemometers on the 15m meteorological towers located at
N Met, the meteorological tower located upwind of both tillage sites. Due to malfunctions with
the meteorological package, the wind speed data collected by the tethersonde system was not
used in the wind profile estimation above the highest cup anemometer. Instead, the logarithmic
wind speed profile fit to the tower data was extrapolated up to 250 m to estimate the wind speed
at higher elevations. An error in the code used to calculate the one minute average wind
directions from the wind vane at the top of the meteorological towers (15.3 m) was discovered in
post processing. Instead, one minute averaged horizontal wind directions calculated from the
sonic anemometer data were used in subsequent analyses. Therefore, the wind direction
measured by the sonic anemometer was applied to the entire wind speed profile. For most sample
periods, wind profiles above 150 m were not required for the lidar emission rate calculations. An
example of a calculated wind speed profile based on tower-mounted cup anemometer data is
presented in Figure 9. Averaged horizontal wind speed data from the sonic anemometer (11.3 m
agl) on N Met were compared with the nearest cup anemometer (9.7 m agl) wind speed
measurements as a quality check. Both data sets showed the same patterns and recorded wind
speeds were close (< 0.25 m/s difference), with the observed difference likely due to a
combination of vertically separated sample heights and instrument error.
150
100
D)
•S
X
50
--- Generated Wind Profile
0 Cup Anemometer Wind Speed
0.5
1.5 2 2.5
Wind Speed (m/s)
3.5
Figure 9. A wind profile calculated for the Disc 2B pass on 10/27/2007.
32
-------
3.1.3 Soil Sampling
Soil characterization involved collecting soil samples for analysis of bulk density, soil moisture,
and sand/silt/clay content. Bulk Density samples were collected prior to tillage operations along
two transects, one across each field, in both the furrow and on the ridge at each sample location.
A manual device consisting of a 7.6 cm in diameter and 7.6 cm deep cylinder was hand driven
into the soil until the top of the cylinder was level with the soil surface. Samples were removed
and placed into pre-weighed cans. Post-weights were performed in the field for determination of
wet weight. All weights were determined using a Mettler balance (Columbus, OH), Model
PM2000.
Samples for soil moisture were taken for each day of operation at random locations in the field
and collected in pre-weighed cans 7 cm in diameter and 4 cm deep. Samples were collected
immediately prior to the tillage period or shortly after commencement in areas that had not been
tilled. The can was pushed into the soil approximately 3cm and removed. The can was closed
and weighed in the field for determination of wet weight.
All soil samples were dried at the Agricultural Research Service (ARS) National Soil Tilth
Laboratory (NSTL) in Ames, IA at a temperature of 105 °C until a constant weight was achieved
(-60 hours). Samples were then weighed to determine dry weight. Calculations for soil moisture
and bulk density were performed according to the following equations as found in Doran and
Jones 1996 [34].
field water content (%) = weight of moist soil - weight of oven dried soil x 100 (6)
weight of oven dried soil
% moisture = weight of moist soil - weight of oven dried soil x 100 (7)
weight of moist soil
bulk density = weight of moist soil x (1 - field water content) (8)
volume of soil collected
9 9
where "volume of soil collected" = TI x radius x length of cylinder = 71 x 3.81 x 7.62 = 347.3
cm3 and "field water content" is the value given by Eq. 6 expressed as a fraction.
A composite of all the samples collected was made. It was analyzed for the percent of sand, silt,
and clay according to the Hygrometer Procedure, as given in Soil Sampling and Methods of
Analysis (1993) [38]. The percent of stable aggregates was also determined from the composite
sample according to the Dry-Sieve Method, as given in the Soil Sampling and Methods of
Analysis (1993) [39].
3.1.4 Air Quality Point Samplers
The suite of point air quality samplers deployed around the tillage plots to quantify both the
ambient and ambient plus operations emissions values were summarized in Table 9 and Table
10. Details of each of these sensors and their data processing methods are presented below.
33
-------
3.1.4.1 MiniVol Portable Air Sampler
Thirty MiniVol Portable Air Samplers (Airmetrics, Eugene, OR) were distributed to
gravimetrically measure the time-averaged mass concentrations of aerosols at multiple locations
surrounding the fields of interest. The MiniVol is a battery operated, ambient air sampler that
gives results that closely approximate air quality data collected by a Federal Reference Method
(FRM) sampler [40] [41]. The sampler draws air through a particle size separator, or impactor
head, and then through a filter medium [42]. The photos in Figure 10 show a closeup of a
MiniVol mounted on a rechargeable battery pack with attached impactor filter heads, and an
example of how these PM samplers were deployed for the fall tillage field experiment.
Figure 10. Airmetrics MiniVol Portable Air Sampler, a closeup view and an example of field deployment.
Particulate concentrations in the PMio, PM2.5, and PMi size fractions were measured using
impactor heads for size separation based on aerodynamic diameter, while Total Suspended
Particulate (TSP) was measured by not using an impactor head. Each PM sampler was assigned
to sample a specific size fraction at a specific location throughout the study, with location
changes made as deemed necessary. Clusters of four PM samplers were assigned to two
locations, one upwind and one downwind, in order to provide size fractionated, mass-based
particle loading distributions. One instrument malfunctioned on October 20, 2007 and was
unable to be used for further sample collection.
Filters used in the PM samplers were pre-conditioned according to the protocols outlined in 40
CFR 50 Appendix J before obtaining pre- and post-sample filter weights. Final average weights
were found using a Type MT5 Microbalance (Mettler-Toledo, Inc., Columbus, OH) located at
the Utah Water Research Laboratory (UWRL) in Logan, UT to determine three stable weights
within ± 5 jig, measured on different days. Filters were transported to and from the site and
stored on-site in dessicators to maintain filter conditioning. Flow calibrations on each MiniVol
were performed using a slant manometer prior to the study and the actual sample flow was
34
-------
adjusted daily on each instrument in order to maintain the required 5.0 L/min for accurate
particle size separation.
3.1.4.2 Optical Particle Counter
Nine Optical Particle Counters (OPCs), Model 9722 from Met One Instruments, Inc. (Grants
Pass, OR), were deployed around the study area collocated with MiniVols in order to describe
the particle count and size distribution at locations measuring background and those measuring
background plus plume aerosols. Figure 11 shows an OPC deployed for the tillage campaign,
collocated with MiniVols, with an accompanying rechargeable battery pack and solar panel.
Figure 11. Met One Instruments Optical Particle Counter (OPC) Model 9722, indicated by arrow, setup for
field deployment on a tower base with an accompanying rechargeable battery pack and solar panel.
The 9722 particle counter uses scattered light to size and count airborne particles. Particle counts
are reported in eight, user-defined channels over the user-defined sample time. For this study, the
OPCs collected samples continuously at a sample time of 20 seconds with the following channel
sizes, in units of |im: (1) 0.3-0.5, (2) 0.51-0.6, (3) 0.61-1.0, (4) 1.01-2.0, (5) 2.01-2.5, (6) 2.51-
5.0, (7) 5.01-10.0, and (8) >10.0. The data from each OPC were relayed to a single computer
over a custom radio network for storage. Inlet flows for individual OPCs were measured on-site
before and after the experiment using a Gilian Gilibrator2 Calibration System, a volumetric flow
meter. The average of the averages from each flow measurement period was used as the sample
flow throughout the field study.
Calibration of OPC particle counts was performed for each sample day in post-campaign
analysis. For this calibration, careful examination of the number concentration (number of
particles/m3) time series yielded a time period prior to or after the tillage operation during which
no apparent plumes were detected. Number concentration was chosen as the calibration point, as
35
-------
opposed to the raw particle count data, because it normalizes the raw particle counts by sample
flow (see Eq. 10); sample flow varies up to 20% between OPCs. The average number
concentration (TVy) per bin (/') for the designated calibration time for each OPC (/) was calculated,
and the mean of the averages (TV,) was used as the calibration concentration. A Counting
Correction Factor (CCFy) for each bin of each OPC was calculated for each collocated run based
on the calibration concentration for that bin according to the following equation:
Due to nearby agricultural activity contaminating the sample, the average number concentration
for the OPC located at site E2 on sample dates 10/25/2007 and 10/26/2007 was not included in
the mean TV,, and counting correction factors were not calculated for this OPC on these days. In
addition, contamination of multiple OPCs during available non-operation times on 10/20/2007
prevented an adequate calibration based on the data collected that day. In place of a daily CCFy
specific for 10/20, the average for each bin of each OPC from the other six sample days was
used.
Number concentration (TV) is a function of raw particle counts (/?), the measured average flow
rate (q), the sample time (f), and the CCFy, as shown in Eq. 10.
(10)
q xt
where the units for each variable are N = number per liter (#/cm3), p = number (#), q = cubic
centimeters per minute (cm3/min), t = minutes, and CCF is unitless. As in Eq. 9, the subscript /
represents a specified bin.
The volume concentration of aerosols based on a number concentration TV is calculated based on
the following assumptions: 1) the particles are spheres, 2) the maximum particle diameter
measured is 20 jim, and 3) the geometric mean particle diameter per bin (GMD,) is
representative of the particles in a given bin /' with the assumption of a log-normal distribution of
particle numbers. The GMD, is calculated by Eq. 11.
w (11)
where dirUpper and dijower are the diameters of the upper and lower ranges for bin /'. The assumption
of a maximum measured particle diameter must be made in order to calculate the GMD for
channel 8, which counts particles > 10 jim.
The cumulative volume concentration of aerosols (Vk) up to a particle diameter k may be
calculated using the following equation:
/d (12)
36
-------
where «(d) is the number concentration at diameter d. For application to the OPC data, Eq. 12 is
expressed in the following terms that have been previously defined:
GMDj
-------
quantification of the elemental carbon (EC) fraction. To account for non-carbon components of
the organic compounds' mass, the OC concentrations reported by the 5400 were increased by the
recommended multiplier of 1.7 [44]. A sharp-cut PM2.5 cyclonic separator was placed at the
inlet, which was located directly above the instrument on top of the trailer. A nominal flow rate
of 16.7 Lpm was maintained through the system by on-board mass flow controllers and integral
temperature and pressure sensors.
Flow checks, leak checks, and CO2 audits, as per the instrument manual, were performed and
passed at the UWRL during the week prior to departure for the field campaign and upon setup of
the instrument on October 13, 2007 [43]. Additionally, CC>2 audits in the field were administered
according to the instrument manual instructions and passed on October 20 and 29 of 2007.
3.1.4.4 Ion Chromatographic (1C) Analysis
In an attempt to more fully chemically characterize the nature of the upwind and downwind
particulate matter, ion chromatographic analysis was performed on selected filters collected via
the MiniVol systems at the air quality trailer (downwind), a second downwind location (S5 at 9
m) and a presumed upwind location (Nl at 9 m). The selected filters were from the October 23rd,
25*, and 26 observations of the conventional tillage practices (refer back to Table 8). After final
post-test weights were determined from the filters (PM2.5, PMio, and TSP) at the UWRL, the
chosen filters were sonicated with triplicate rinses in 10 mL double-distilled, de-ionized water
(DDI) and split into two aliquots of approximately 15 mL each for separate anion and cation
analysis. The anion analysis occurred within 48 hours of sonication and the cation aliquots were
stabilized with 10 uL of 0.5 M HC1 acid and analyzed within 28 days of sonication. The base 1C
system (Dionex Corporation) utilized an AS 40 Automated Sampler, CD 20 Conductivity
Detector, GP 40 Gradient Pump, and membrane suppressor. Cation quantification was
accomplished using an lonPac® CS12A cation column, a CG12A cation guard column, and a 500
|iL sample loop. The system eluent was 0.15 M H2SO4 with a 1.0 mL/min flow. Anion
concentrations were determined using an lonPac® AS11HC anion column, a AG11HC anion
guard column, and a 188 jiL sample loop. For anions, the system eluent was 30 mM NaOH with
a 1.0 mL/min flow. ACS regent grade materials were used to prepare a stock standard solutions
for each of the target ions, from which concentrations of 0.5, 1.0, and 5.0 mg/L (ppm) were
mixed to make 1C calibration curves. The ions quantified were fluoride (F~), chloride (Cl~),
nitrite (NO2~), sulfate (SO/f2), nitrate (MV), sodium (Na+), ammonium (NFL;+), potassium (K+),
magnesium (Mg+ ), and calcium (Ca+ ). Verifications of the system calibrations were performed
prior to each analysis run and roughly every 10 samples blank (DDI water) and continuing
calibration verification standards (CCV) were tested. Peak identification and data processing
were executed using Dionex PeakNet Data Chromatography software (Version 2.0).
3.1.4.5 Aerosol Mass Spectrometer
An Aerosol Mass Spectrometer (AMS) from Aerodyne Research, Inc. (Billerica, MA) was
located in the sampling trailer, with a sample port on the upwind side of the trailer just below the
roof level. The AMS provides chemical composition and particle size information for volatile
and semi-volatile particle components in the 0.1 - 1.0 jim size ranges in vacuum aerodynamic
diameter. AMS size and mass calibrations were conducted the first day of arrival at the site
(10/13/2007) using polystyrene latex spheres (PSLs) and ammonium nitrate respectively. The
AMS vaporizer was operated at higher than normal temperature (-800 °C vs. -600 °C) to attempt
38
-------
to detect some of the inorganic components in dust particles. During sampling, the AMS
integrated and saved particle composition and size data every 10 minutes.
3.1.5 Lidar Aerosol Measurement and Tracking System
The Aglite lidar system is a monostatic laser transmitter and 28-cm receiver telescope (Figure
12). The laser is a three-wavelength, 6W, Nd:YAG laser, emitting at 1.064 (3W), 0.532 (2W)
and 0.355 (1W) um with a 10 kHz repetition rate. The lidar utilizes a turning mirror turret
mounted on the top of a small trailer to direct the beam -10 to + 45° vertically and + 140°
horizontally. Lidar scan rates from 0.5 - l°/s are used to develop the 3D map of the source(s),
dependent on range and concentration of the aerosol.
Figure 12. The three wavelength Aglite lidar at dusk, scanning a harvested wheat field.
The process used to retrieve aerosol mass concentration from lidar data is illustrated in Figure
13. The details of Aglite lidar calibration and aerosol retrieval process are discussed by Marchant
[45] and Zavyalov [46]. The retrieval is as follows. First, preprocessing on the lidar data is
performed. The relationships between backscatter, extinction, volume concentration, and mass
concentration of the aerosol components are established using in-situ data measured by OPCs
and clusters of PM samplers with different separation heads (PM,t). Then, the inversion of the
lidar data is performed using a form of Klett's solution [47] for two scatterers where extinction is
proportional to backscatter. Finally, a least-squares method is used to convert backscatter values
to aerosol mass concentration using the previously established relationships.
39
-------
f Lidar Signal 1 fc-PC Datal fPM Samplers!
Preprocessing
[Particle Size
I Distribution
Klett Inversion
Aerosol
Concentration
Retrieval
Boundary \
Condition +
j-idar Ratios/
( Volume
^Concentration
Mass
\ Conversion
Factor ,
( Mass ^|
^Concentration^
Figure 13. The Aglite lidar retrieval algorithm flow chart, showing the input locations for the in situ data.
From the OPC channel counts, the particle size distribution at a single point as a function of time
may be calculated according to Eq. 15. The backscatter and extinction coefficients necessary for
solving the lidar equation are then calculated at the OPC reference point as a function of time.
Klett's inversion is used to convert the lidar signal to the optical parameters (backscatter values
in particular) of particulate emitted by the operation [47]. Having recovered backscatter values as
a function of range and wavelength using the Klett inversion, these must be converted to the
aerosol cumulative volume concentration. Expressing the particle normalized backscatter values
from the OPC (pv) and the lidar measured backscatter values (PE) in a vector form, and applying
the Moore-Penrose weighted minimum least-squares solution results in the value for particle
concentration n(z) at range z
(16)
which can be multiplied by the particle normalized volume concentration vector, resulting in the
The term W is a diagonal weighting matrix, whose diagonal elements are the expected variance
of the emission backscatter for the corresponding channel.
The retrieved aerosol volume concentration from the lidar return signal is multiplied by the
MCF,t, which was previously calculated using in-situ data (Eq. 14). At this point, the Mi fraction
of the aerosol mass concentration of the emission component is known as a function of distance.
40
-------
= MCFK-VK(z)
(18)
The concept behind our flux measurement approach is shown in Figure 14A, where the facility is
treated as one would calculate the source strength in a bioreactor. In this simplified approach, the
source strength is determined using the mean flow rate through the reactor and the difference in
reactive species concentration entering and leaving the vessel. The scanning lidar samples the
mass concentration fields entering and leaving the facility, while standard cup anemometers
provide the mean wind speed profile. If we define our box large enough so that none of the
emitted material escapes through the top or sides of the box, and the downwind side is far
enough from the facility to minimize high frequency fluctuations, the same simple relationship
found in a bioreactor applies. An example of our lidar derived concentration data is shown in
Figure 14B. The concentration plot pattern from scanning up one side, across the top, and down
the other looks like the common office staple (the paper fastener), and will be referred to as a
"staple" scan. The data from the top of the box is regularly examined to be sure that no
significant paniculate transport is passing through the top. The data for the left side panel of the
staple provides the background concentration entering the box, while that on the right provides
the background plus facility concentrations leaving the box. The integrated mass concentration
difference multiplied by the wind speed during the scan completes the flux emission calculation
by yielding a mass per unit time emission from the facility.
B
Wind Direction
Flux In
Flux Out
900
800
range 700
(meters) 600
Figure 14. (A) Conceptual illustration of the method for using lidar to generate time resolved local area
particulate fluxes. (B) An example of a "staple" lidar scan over the facility showing aerosol concentration on
the three sides of the box.
The flux calculation in the integral form can be expressed as following:
F =
(19)
where Vy is the average wind speed component, defined as parallel to the long axis of the staple
box, cos9/)Cz)(r,/z) and cosQuCu(r,h) are downwind and upwind, respectively, particle
concentration corrections for wind direction for each data point at range r and height //, and CD -
41
-------
Cu form the mass concentration difference upwind and downwind, integrated over the range
(width) and height of the sides of the staple. In our routine, Eq. 19 is discretized as:
)-Ar-A/7 (20)
where RO and R are the near and far along beam edges of the box and HO and H form the top and
bottom of the box. (In many cases, HO is set above eye level and concentration is extrapolated to
the ground to avoid illuminating personnel and animals.) The Ar-A/z term is the individual area
element for which each flux component as calculated by each step in the double summation.
Figure 14B shows an example of single scan staple faces for the Cu and CD mass concentration
values for PMio between the distances of 600 and 900 m from the lidar. Single scan differences,
of course, do not account for accumulation or depletion in the box due to wind speed variation
during a scan, or from high or low concentration pulses that may still exist in the downwind
sample. For a meaningful estimate of the facility emission, many scans are combined to achieve
a time averaged emission rate.
To perform the lidar calibration using the in-situ instruments, collocated samples are needed for
the PM sampler, OPC, and lidar. In other words, the lidar beam must be directed past an
OPC/PM sampler location and held in place for the duration of at least one OPC sample
collection time. Holding the beam stationary, usually directed next to a sample tower, is referred
to as the 'stare' mode. The 'stare' sample mode provides not only calibration, but quality
assurance of the data set as well. After data processing and PM concentrations along the beam
have been calculated, successive 'stare' data are compared against data from the reference OPC
and PM samplers to verify that the concentrations calculated from both instruments are similar
within measurement errors.
All lidar scans were quality checked to prevent the use of data that was contaminated or that was
collected during times when the tractors were not actively tilling. All scans collected during
times when no tillage activity was occurring were removed from further calculations. Upwind
scans were checked for contamination from traffic on dirt access roads or agricultural activity to
the north of the study site; upwind scans containing such contamination, as well as the
corresponding downwind scans, were removed from further emission rate calculations.
Downwind scans were not considered valid, and therefore not used in further calculations, if:
1) The corresponding wind direction was outside of ± 80° from North (0 or 360°)
2) The lidar scan contained apparent plumes from an outside source (such as
from dirt road traffic which stretches across the length of the scan at a time
when the wind was blowing perpendicular to the scan)
3) No plumes were detected
4) The tillage plume had a potentially significant portion crossing the lidar beam
closer than 500 meters to the lidar trailer.
42
-------
As the upwind scans are used as background references for specific downwind scans, only the
upwind scans that corresponded with valid downwind scans were used. Table 11 presents the
number of both total and valid upwind and downwind scans collected per day. In many cases two
or more downwind scans use the same upwind scan as a reference because multiple downwind
scans were made for each upwind scan, resulting in a little over half as many upwind scans as
there are downwind scans.
Table 11. Number of upwind and downwind lidar scans determined to be valid for emission rate calculations.
Date
10/19/2007
10/20/2007
10/23/2007
10/25/2007
10/26/2007
10/27/2007
10/29/2007
Upwind scans
Total
68
46
123
70
90
76
65
Valid
23
42
69
38
37
42
28
Downwind scans
Total
140
95
246
141
180
155
131
Valid
47
78
122
70
70
77
50
% Valid
33.6
82.1
49.6
49.6
38.9
49.7
38.2
3.2 MODELING SOFTWARE
Air dispersion modeling and the prediction of plume centerline position along the southern
boundary of the field were performed during the data analysis. The techniques for these steps are
explained in detail in the following sections.
3.2.1 Dispersion model software
The U. S. EPA has approved a number of air dispersion models for use in regulatory
applications. These are listed in Appendix W of 40 CFR Part 51 [48]; included are the Industrial
Source Complex Short-Term Model, version 3 (ISCST3) and the American Meteorological
Society/Environmental Protection Agency Regulatory Model (AERMOD), which as of
November 2005 is recommended for all regulatory applications [1][2]. Both models assume
steady-state conditions, continuous emissions, and conservation of mass. The default, and
minimum, time step available for both models is one hour. Therefore, all meteorological input
data represent one hour averages.
ISCST3 assumes a Gaussian distribution of vertical and crosswind pollutant concentrations [49].
The Gaussian plume equation uses the Pasquill-Gifford horizontal and vertical plume spread
parameters, oy and crz, respectively, shown in Eq. 21.
(21)
Cio is the 10 minute average concentration (ug/m ), Q is the emission rate (ug/s), u is the average
wind speed at release height (m/s), y is the horizontal distance of the chosen receptor from the
centerline of the plume (m), z is the height of the receptor above ground level (m) and H is the
effective stack height (m), which includes estimates of plume rise due to buoyancy and/or
momentum [49].
43
-------
ISCST3 assumes a Gaussian distribution of pollutants based on time-averaged meteorological
data. It also uses stability classes to address pollution dispersion due to atmospheric mixing.
Stability classes are typically determined by a combination of vertical temperature lapse rates
and incoming solar radiation or methods using vertical or horizontal wind variance [50].
AERMOD requires more detailed meteorological and surface characteristic information. Because
of the additional input requirements for AERMOD and the lack of an established database for
these inputs, many regulatory agencies continue to use ISCST3. The suite of meteorological
instruments employed during this field study allowed us to use both models in this study.
AERMOD uses continuous functions for atmospheric stability determinations, and based on
stability determines the appropriate distribution, a Gaussian distribution for stable atmospheric
conditions, and a non-Gaussian distribution for unstable, or turbulent conditions. AERMOD is
also better at accounting for terrain features and building downwash phenomena than ISCST3
[51]. The interface used to run the models was the commercially available ISC-AERMOD View
packaged by Lakes Environmental, Inc (Waterloo, Ontario, Canada).
Final emission rates were determined using inverse modeling coupled with observed facility-
derived pollutant concentrations. In inverse modeling, the downwind impact on pollutant
concentrations by a source is known while the emission rate is unknown. To solve for the source
emission rate, a model such as ISCST3 or AERMOD is run with the following inputs: on-site
collected meteorological data, the facility layout, the locations of pollutant sources and receptors
(samplers), and an estimated or "seed" emission rate, which can be obtained from literature.
Observed facility-derived concentrations are calculated by subtracting measured background
levels from concentrations measured at locations impacted by the source plume. Modeled
concentrations are then compared to the facility-derived concentrations at each sampler location.
The location specific ratio of the measured concentration (Cmeamred) to the modeled concentration
(Cmodeied) is multiplied by the seed emission rate (Eseed) and an average across all valid locations
is calculated to yield the source emission rate corresponding to the measured facility-derived
concentrations (EesHmated) as shown in Eq. 22.
(C "|
F — F \ measured 07}
^estimated ^seed\ ^, \^^J
\ modeled j
A seed emission rate of 50 ug/s-m2 was calculated using the AP-42 4* Edition emission factor
estimate for agricultural tillage operations, Chapter 9.1 [52], assuming soil with 50% silt content.
The equation is given as follows:
E = k(5.38)s06 (23)
where E is the emissions in units of kg/ha, k is a cumulative particle size multiplier (TSP = 1.0,
PMio = 0.21, PM2.5 = 0.042), and s is the silt content of the surface soil. A run time of 2 hours
was assumed to provide a seed emission rate based on time and the same emission rate was used
for all PM size fractions. The current edition of AP-42 (i.e. 5* Edition) does not include a
method for estimating PM emissions from agricultural tillage [53].
It should be noted that evaluations of air dispersion model accuracy have shown that models are
better at predicting concentrations over longer averaging times than shorter time periods at a
44
-------
specific location; such models were developed and optimized to predict longer term averages and
do not incorporate all of the many temporally and spatially variable factors that may affect
dispersion in the atmosphere. Models can predict the magnitude of the highest concentration
reasonably well, with a typical range of errors of ±10-40%, but do not predict the exact location
and time of the highest value well. Paired measured and modeled concentrations at a specific
location throughout the modeling domain are usually poorly correlated, which is likely due to a
combination of uncertainties in the input data that potentially may be reduced and unquantifiable
uncertainties within the model itself.
An example of errors due to uncertainty in the input parameters is that concentration errors
between 20 and 70% can result from an uncertainty of five to 10 degrees in the wind direction
that directly affects plume location, depending on atmospheric stability and the sampler/receptor
location. Uncertainty within air dispersion models, called "inherent uncertainty", is mostly due to
the simplification of complex and highly variable processes affecting dispersion in the
atmosphere. If atmospheric conditions that are used as inputs into the model (wind speed, wind
direction, mixing height, etc.) are consistent across multiple sample periods, the model would
predict the same concentration while measured concentrations could vary significantly due to
variability in the complex processes that are not accounted for in the model. These inherent
uncertainties can produce predicted ground level concentration errors of up to 50%. A more
detailed discussion of air dispersion model uncertainty and accuracy is presented in Appendix W
to Part 51 of Section 40 of the Code of Federal Regulations [48]. The report "Air Emissions from
Animal Feeding Operations: Current Knowledge, Future Needs" by the National Research
Council of the National Academies states that due to the assumptions required with Gaussian
dispersion models, the uncertainty associated with predicted concentrations are not smaller than
±50%. Additional uncertainty is introduced by stability classification and sample instruments
[54]. However, the placement of near ground-level receptors along the dominantly downwind
side of large ground level area sources, such as those used in this study, potentially may reduce
the inherent uncertainty of the predicted concentrations due to both the vertical and horizontal
proximity of the receptor to the source. Unfortunately, no discussion of inherent uncertainty in
dispersion models under such conditions is available.
Based on the issue of potentially reducible uncertainty, all of the input data for both models were
very carefully screened to reduce uncertainty in the output to the maximum extent possible, as is
subsequently described. Even with such efforts, the error in the model predicted concentrations
for this study is expected to be ±50% according to the above sources, which, when combined
with a ±20% sampling error of the difference between upwind and downwind MiniVol PM
samplers (±10% each) in Eq. 22, yields a range of error about the calculated emission rate at a
single location between -46% and ±140%. Averaging all the valid emission rates per sample
period may reduce the potential error range to -33% to ±100% by removing the sampling error of
the MiniVols.
The meteorological data were carefully screened and corrected for problems prior to
preprocessing for the dispersion models. It is this screening process that uncovered the incorrect
wind direction averaging code for the meteorological towers discussed in section 3.1.2. As
previously described, we have instead employed the horizontal, one minute averaged wind
directions from sonic anemometers. The meteorological inputs for the models generated by the
preprocessing programs were also screened for consistency with input measured values;
45
-------
adjustments were made as necessary. The effects of the uncertainty associated with the
meteorological input, however small, combined with the inherent uncertainty within the model
will be most greatly impact by those receptors that are predicted to be on the edge of the plume,
which in turn can greatly affect the calculated emission rates. Arya (1998) suggests that the
plume edge be defined as 10% of the maximum modeled concentration to minimize these effects
[55]. Therefore, emission rates calculated at locations with predicted concentrations less than
10% of the maximum predicted concentration will not be used in calculating the average
emission rate.
Meteorological data were compiled as needed from the different deployed meteorological
instruments. AERMET, the preprocessor of meteorological data for AERMOD, requires that the
surrounding land use and land cover be categorized to quantify the Bowen ratio, surface
roughness, and midday albedo. For this study, the land use on all sides of the site was classified
as all cultivated land and the default monthly October values of midday albedo (0.18), Bowen
ratio (0.7), and surface roughness (0.05 m) were used. During each of the sample runs the sky
was clear of clouds, so the amount of cloud cover was set to 0.0 for all hours of interest. The
mixing height for input into RAMMET and AERMET was set at 1000 m because all samples
were started at least 2 hours after sunrise and ended at least 2 hours before sunset. In addition, all
the receptor locations of interest were on the southern edge of the field so the exact depth of the
mixing layer during daylight hours over such a short distance at ground level was not considered
to be a significant factor. Based on the measured incoming solar radiation, vertical temperature
lapse rates, and surface wind speed, stability classes for ISC during all sample periods except one
were determined to be slightly unstable to very unstable. The exception to this was the sample
run on 10/20/2007 that had average surface wind speeds > 6.0 m/s for two of the three sample
hours, which are classified as neutral stability conditions under low to moderate incoming solar
radiation [49]. The Upper Air Estimator in the AERMET View software was used to calculate
required upper air parameters based on observed surface conditions.
Digital elevation map files with a 7.5 minute resolution were used to calculate receptor and
source elevations [56]. The terrain was not considered to be a significant factor in the modeled
concentrations as the change in elevation over the entire modeled domain (~2km x 2km) was
gradual and no greater than 2 meters. The areas tilled during the sample periods were represented
in the dispersion models by ground level area sources that varied in size and shape. The readings
of the handheld GPS tracking device located on Tractor 1 for each tillage operation, mentioned
in section 2.2, were used to develop the source areas and shapes. The GPS readings of sample
locations were used to specify discrete receptors for comparison between modeled and measured
concentrations at specific locations. In addition, a receptor grid with lOmxlOm spacing and a
flagpole height of 2.0 m above ground level was set up slightly upwind of, over, and downwind
of the area source to provide a visualization of the modeled particulate matter concentrations
resulting from the area source emissions. Elevations for receptors and source areas were
calculated and assigned by AERMAP. Table A 1 in Appendix A provides more details about the
settings used in each model. All concentrations used for emission rate calculations and presented
in this report are averages over the modeled periods.
3.2.2 Plume Movement Prediction
The crossing point along the lidar beam path of the plume from the tractor with the hand-held
GPS onboard was estimated using the tractor position with time information given by the GPS
46
-------
unit and the minute averaged wind speed and direction. Tractor position was interpolated to the
nearest minute for each GPS data point, in Universal Transverse Mercator (UTM) coordinates, to
match the time of the wind data set. In actuality the tractor(s) emitted particulate plumes
constantly as it (they) moved across the field, but for movement modeling purposes only, it was
assumed that the tractor emitted a discrete puff of particles at this location and time. The
calculated wind speed and direction at 5 meters above ground level were used to calculate where
the puff would be at the end of that minute, and the wind data from the following minute were
then used to calculate the plume position at the end of that minute based on the ending
coordinates of the previous minute. Using a series of such calculations, the coordinates and time
at which the plume crossed the lidar beam path was calculated for each minute of operation time
for each day. If the calculations showed it took longer than 20 minutes to cross the lidar beam
path the calculations were truncated. For those puffs whose calculations showed them crossing
the beam, the crossing time and location, in both UTM coordinates and distance downbeam from
the lidar, were calculated.
Predicted downwind beam-plane distances and crossing times were used as one of the quality
checks for the downwind lidar scans, especially when part or all of the plume was predicted to
cross the beam path closer than 500 meters to the lidar trailer.
It should be noted that GPS position with time was only available for one tractor. Therefore, the
plume movement predictions for the plumes from the second tractor were not made. It should
also be mentioned that no attempt to predict plume concentration or plume dispersion was made
in this model.
3.3 STATISTICAL ANALYSIS OF DATA
The overall goal of the statistical analyses is to provide confidence about the observed
differences between the conventional and conservation management practices and whether the
differences are detectable. For highly variable discrete events, such as agricultural tillage
operations, each tractor/implement pass can (and does) vary considerably from the previous
tractor pass. In this way, it is the variability of the tillage process which dominates the
calculation of the error bars for any given tillage operation for a particular field. Said another
way, the reproducibility of the measurement devices (e.g. OPC, MiniVol, lidar) is typically much
higher - often by a more than a factor of 10 - than the reproducibility of any two tractor passes.
When it makes mathematical sense to use conventional statistical definitions, we make use of
them. However the highly non-statistical (either Gaussian or Poisson) nature of the agricultural
tillage process challenges the applicability of these methods.
Therefore, statistical analysis will depend upon the parameter being evaluated. For all
parameters, a simple mean and variance will be computed over selected time intervals. For
example, data from the OPC will be compiled over time and a mean and standard deviation
calculated for each unit for each tillage operation. These means can be compared among
locations using a simple T-test for those intervals and between tillage implements to determine if
there are differences. The null hypothesis is "there are no differences in the particulate emissions
between conventional and conservation management practices."
47
-------
The aggregation of the particulate data with the wind profile and sonic anemometer data will
provide an estimate of the particulate flux within a field. This data will be expressed as
cumulative particulate emission for a specific portion of the tillage operation and then
differences between systems compared using mean comparison methods. Regression analysis
using multivariate methods will allow for the incorporation of different soil moisture conditions
or meteorological conditions.
4. RESULTS AND DISCUSSION
4.1 GENERAL OBSERVATIONS
4.1.1 Soil Characteristics
The average bulk density and soil moisture values, with la and the number of samples collected
(«), are presented in Table 12. Figure 15 provides a map of soil sample locations. The average
bulk densities (± la) for Field B were 1.47 ± 0.02 g/cm3 for the furrow and 1.37 ± 0.03 g/cm3 for
the ridge. The average bulk densities for Field A were 1.52 ± 0.06 g/cm3 and 1.34 ± 0.05 g/cm3.
It is expected that the furrow would have a higher bulk density because that is the path of the
tires of the tractors, implements, and other equipment, which are known to cause soil
compaction. That is also the point that has the highest pressure from water during flood
irrigation. Drivers of the machinery avoid impacting the ridges because that may cause a loss of
profitable growing area.
Soil moisture levels were also different between the furrow and the ridge. The furrows had the
higher moisture level in both fields at 10.3 ± 0.49 % for Field B and 11.34 ± 0.61% for Field A,
while the ridges were dryer at 9.45 ± 0.06% and 8.08 ± 0.08%. One would expect the furrows to
have more water in the soil due to shading from the sun by the ridges, exposure to slower wind
speeds due to the surface friction at the level of the ridges, and the greater tendency for fallen
biomass accumulation, as was observed in the fields under study (see Table 12). Figure 16
presents a timeline of measured soil moisture levels, along with tillage operations and
precipitation events. Three precipitation events were observed and are discussed in Section 4.1.2.
Soil moisture tended to decrease with successive disturbances caused by the tillage operations.
Analysis performed on a composite of all the samples collected yielded an average of 46% stable
aggregates and soil composition of 47% sand, 36% silt, and 17% clay. Since the soil bulk density
values from fields A and B are identical within the error of the measurement we expect these
fields to have similar characteristics that contribute to aerosol/dust generation. Similarly, as has
been shown in the literature [11][12][3][4], we expect the soil moisture content to strongly
influence the amount of aerosol/dust production from any given tillage operation.
48
-------
Table 12. Statistics of soil characteristics measured for both fields.
Bulk Density Data Summary
Bulk Density for Field B Furrow
Bulk Density for Field B Ridge
Bulk Density for Field A Furrow
Bulk Density for Field A Ridge
Soil Moisture (%M) Data Summary
%M for Field B Furrow
%M for Field B Ridge
%M for Field A Furrow
%M for Field A Ridge
%M for October 13
%M for October 19 (Chisel, Combined)
%M for October 20 (Optimizer, Combined)
%M for October 22
%M for October 23 (Disc 1, Conventional)
%M for October 25 (Chisel, Conventional)
%M for October 27 (Disc 2B, Conventional)
%M for October 29 (Land Plane, Conventional)
Mean (g/cm3)
1.47
1.37
1.52
1.34
Mean (%)
10.3
9.45
11.34
8.08
10.49
9.14
6.66
11.44
8.47
6.04
7.26
6.74
Std Dev (g/cm3)
0.02
0.03
0.06
0.05
Std Dev (%)
0.49
0.06
0.61
0.08
0.34
0.58
0.43
0.70
0.75
0.29
0.99
0.36
n
15
15
5
5
n
24
24
12
12
30
24
10
10
14
17
10
10
95% CI
0.01
0.02
0.05
0.04
95% CI
0.20
0.02
0.35
0.05
0.12
0.23
0.27
0.43
0.39
0.14
0.61
0.22
49
-------
Sample Locations
anos. CA
Mapping Grade GPS positions
Plaited by Kevin Cole
USDA AR5NSTL
November, 7, 2007
Figure 15. Soil sample collection locations in fields under study.
12 -
10 -
0
5?
£ 8
Si
3
•H 6
s
5?
4 -
2 -
n
Combined Conventional
Operations Method Method
„
CD
L1J
Q.
'u
CD
Q_
CD
UJ
Q.
'u
Q-
1 I
1 _ \
CD *~
!r, CD
u E
Q.
0
'rt "s S S 2
U LO ^ Pg ^
Q u •- 5 I Q_
;- -D
CD C
> ro
LU — 1
Q_
'u
CD
Q_
Figure 16. Timeline of soil moisture levels, tillage activities, and precipitation events.
50
-------
4.1.2 Precipitation Data
Three precipitation events occurred at the study location immediately prior to and during the
study. The first precipitation event occurred on October 11th, the first day of equipment
deployment prior to deployment of the Davis weather station with a rain collector. This weather
station later recorded two precipitation events. The second event occurred on October 16* in the
evening, totaling 1.6 mm of precipitation, and the third occurred on October 27th in the late
evening hours, totaling 1.2 mm. In summation, two precipitation events occurred prior to any
sampling and a third event on October 27* occurred between the disc 2 pass and the land plane
pass in the conventional tillage. Using the precipitation rates and totals collected and the
calculated flux of water vapor representing losses due to evaporation, the depth of precipitation
water in the soil overtime was estimated as explained in Section 3.1.1.1 [57]. The results suggest
that the accumulated water depth from the event on 10/16 had already been evaporated at the
commencement of tillage activities on 10/19, as shown in Figure 17, while there was still 0.2-0.4
mm of water left in the soil from the 10/27 precipitation event during the land plane pass of the
conventional tillage method on 10/29. This remaining moisture in the soil likely affected PM
emissions from the land plane operation. Holmen et al. [12] and Flocchini et al. [3] found that
soil moisture is an environmental variable that can have very significant effects on PM
emissions.
o
£
Q.
V)
O
Q.
re
in
3
c
o
*-
1.6 -
1.2 -
0.8 -
0.4 -
o.o
10/16 10/17 10/18 10/19 10/20 10/21 10/22 10/23 10/24 10/25 10/26 10/27 10/28 10/29 10/30
-Precip. minus ET
-Combined Operations: Chisel start
-Conventional: Land plane start
Figure 17. Expected soil moisture levels over time due to addition by precipitation and losses through
evapotranspiration, with potentially affected tillage operations shown.
51
-------
4.1.3 Eddy Covariance Calculations
Eddies containing heat, water vapor, and carbon dioxide occur in a spatially and temporally large
range of scales. These scales can range from mm to km spatially and temporally from seconds to
hours. Typically EC results are presented as 30-minute averages. This is done so as to ensure that
an appropriate number of eddies have been measured and recorded that represent appropriate
fluxes for heat, water, carbon dioxide exchange for a surface in question. In the case of
particulates, this is a very different scale both spatially and temporally. First the paniculate
emissions are in effect artificially induced by agricultural production activities, in this case fall
tillage practices. The variety of scale operations can vary widely in terms of the size of fields, the
type and number of tillage operations, soil type, soil water content, wind speed, wind direction,
surface stability conditions (meteorological condition, hot surface versus cool surface). These
types of conditions render 30-minute averages as inappropriate for estimating particulate
emission fluxes. Rather shorter averaging periods need to be incorporated as these periods will
contain more representative information relating the particulate emissions to location, boundary
layer conditions and turbulence characteristics which is a function of the surface and boundary
layer conditions. Two examples are presented in Figure 18 and Figure 19. In Figure 18 three
different averaging periods of 1, 15 and 30 minute averages are shown for the fiction velocity
(u*) which is good surrogate representation of the turbulence intensity at the surface. There are
several critical points to note that are both obvious and subtle. First shorter averaging periods
reveal increased detail in the changes of turbulence intensity. The variability clearly decreases as
the averaging periods increase thus in effect smoothing much of the effect of the turbulence. For
fluxes of heat, water vapor and carbon dioxide this is not such a critical issue if the appropriate
conditions for eddy covariance measurements are in effect, i.e. relatively extensive homogeneous
upwind surface of scalars (q, c, and 7) and surface conditions (soil or vegetated surface).
However particulate emission is an altogether different problem because these emissions are
more or less from a relatively slow moving point source (i.e. the velocity of a tractor) and thus
represents a spatially confined footprint influenced by very local conditions of turbulence and
surface stability. The same issues exist for the temporal component. Thus averaging periods
typically used with EC for water and carbon dioxide exchanges may very well be inappropriate
for particulate emissions.
In Figure 19, one minute averages of mean wind speed, standard deviation of the vertical wind
(
-------
Site 3 Oct 24,2007 1300-1500 hrs
0.2
0.18
0.16
0.14
_ 0.12
w
1 0.1
= 0.08
0.06
0.04
0.02
0
0.12
0.1
§. 0.08
0.06
0.04
0.02
* *
hnmin
(b)
13:15
13:30
13:45
14:00 14:15
hnmin
14:30
14:45
15:00
13:30
14:00 14:30
hnmin
15:00
Figure IS. Friction velocity for 10/24/2007,13:00-15:00 hours at N Met computed as (a) 1, (b) 15 and (c) 30-
minute averages.
53
-------
Site 3 Oct 23, 2007 0800-1900 hrs
hr:min
hr:min
^^^
?••?• »>X>>X>>"^^" »?••?• •
hnmin
Figure 19. One minute average wind speed (a), crw (b), and u- (c) for Site 3 on 10/23/2007.
-------
4.1.4 Plume Movement Prediction
The predicted movement of puff plumes based on GPS coordinates for one tractor and wind data
demonstrate well the variability of the winds. On some days, such as 10/20 the winds were very
consistent and the predicted plume paths all follow the same general direction (Figure 19). When
light and variable wind conditions existed for a significant portion of the sample period, as on
10/25 shown in Figure 20, plotting the predicted movements of plumes results in an unorganized
and seemingly random connection of points. This was also the case for the chisel pass of the
Combined Operations Method (10/19), which had the lowest predicted lidar beam-plane plume
crossings at just 18.4%. This is also reflected in the relatively small percentage of valid lidar
samples for the 10/19 sample period (33.6%) when compared to other sample periods (Table 11).
Table 13 below gives the percent of puffs that were calculated to cross the lidar beam-plane on
the downwind side of the fields under study.
500
? o
a:
-500
tractor path
lidar beam path
^ lidar location
x plume intersection point
modeled plume paths
500 1000
Relative East (m)
1500
2000
Figure 19. Predicted puff movements during the Optimizer pass in the Combined Operations Method (10/20).
55
-------
1000
•f 500
a:
tractor path
lidar beam path
lidar location
plume intersection point
modeled plume paths
500 1000
Relative East (m)
1500
Figure 20. Predicted puff movements during the chisel pass of the Conventional Method (10/25).
Table 13. Percent of puffs predicted to cross the lidar beam plane.
Date
10/19/2007
10/20/2007
10/23/2007
10/25/2007
10/26/2007
10/27/2007
10/29/2007
Operation
Combined: Chisel
Combined: Optimizer
Conventional: Disc 1
Conventional: Chisel
Conventional: Disc 2A
Conventional: Disc 2B
Conventional: Land Plane
Beam-plane crossings (%)
18.4
98.0
72.2
54.3
90.5
58.5
74.3
4.2 AEROSOL CHARACTERIZATION DATA
The following section contains the aerosol data collected using both the filter and optical sampler
methods.
4.2.1 MiniVol Filter Sampler Data
Observed PM2.5 concentrations from the Airmetrics MiniVol samplers ranged from 5.8 to 52.9
u,g/m3; PMio concentrations ranged from 16.3 to 165.3 u,g/m3; TSP concentrations ranged from
60.5 to 203.3 (J,g/m3. All recorded PM concentrations are presented in Table B 1 in Appendix B.
56
-------
The MDL for each sample period was calculated based on the average run time of the MiniVols,
the targeted flow of 5.0 L/min, and the minimum filter catch that could be measured in the
difference between the pre- and post-test filter weights of 5 jig. The MDL for each run differed
based on different sample durations; the average MDL value ± 1 standard deviation (n = 7) was
3.7 ± 0.9 |ig/m3, with a range of 2.3 to 4.8 |ig/m3.
All collected MiniVol samples and corresponding documents were examined for potential
sampling errors (incomplete sample time, sampler malfunction, human error, etc.) and identified
problem samples were removed from further calculations. Average MiniVol measured
upwind/background and downwind/operation-impacted concentrations by operation are shown in
Table 14. Upwind and downwind locations were separated based on wind direction and source
area. In order to determine if the differences between mean upwind and downwind PM2.5 and
PMio concentrations were significant, the 67% confidence intervals (CI) were calculated and are
shown with the averages. Confidence intervals for TSP were not calculated because only one
upwind and one downwind measurement were made each day. Mean downwind concentrations
of PM2.s, PMio and TSP averaged 96%, 134% and 160%, respectively, of those upwind.
Table 14. Mean measured PM concentrations for each operation upwind and downwind of the tillage site.
Error is the 67% CI about the mean for n > 3.
Date
10/19/2007
10/20/2007
10/23/2007
10/25/2007
10/26/2007
10/27/2007
10/29/2007
Upwind
Downwind
Upwind
Downwind
Upwind
Downwind
Upwind
Downwind
Upwind
Downwind
Upwind
Downwind
Upwind
Downwind
PM25
Hg/m3
34.4 ±3.4
24.3 ±2.1
17.9 ±1.8
27.8 ±5.3
16.1 ±0.6
11.8±1.2
38.6 ±3.5
41.4±2.1
26.7 ±2.6
24.9 ±1.8
22.1 ±1.2
16.5 ±2.0
34.8 ±2.1
32.3 ±2.4
PM10
jig/m3
41.6 ±4.4
60.6 ±10.3
28.2 ±2.0
42.1 ±6.4
39.6
59.7 ±4.2
70.5 ±10.0
78.4 ±7.8
37.2 ±2.5
52.2 ±6.1
37.4 ±7.1
52.5 ±8.5
50.1 ±2.5
51.0 ±4.3
TSP
Hg/m3
157.2
122.5
87.0
174.1
60.5
203.3
123.6
196.0
84.0
70.2
84.3
89.1
62.4
Wind
Speed
m/s
1.1
6.7
1.6
1.5
2.9
3.1
1.7
Wind
Direction
o
76
305
316
2
302
30
49
Tillage
Operation
Chisel
Optimizer
Disc 1
Chisel
Disc 2A
Disc 2B
Land Plane
In theory downwind samplers would always measure higher concentrations than upwind, with
the largest differences correlating with operations producing the most PM. During this field
campaign, however, the average upwind PM2.5 concentrations for some operations were higher
57
-------
than the average measured downwind concentrations. This could be explained by the background
locations being impacted by nearby sources, such as traffic on dirt roads or nearby tillage
operations, or upwind sampler locations not having sufficient standoff distance from the
operations. The movement of air around a tractor or other vehicle passing a sampler location will
cause turbulence; at relatively low wind speeds, this may cause plumes of PM entrained in the
turbulence structure to be sampled if the upwind instruments are not located at a sufficient
distance upwind from the vehicle. Due to the cumulative nature of the MiniVols' PM collection,
even a single exposure at these potentially high concentrations can significantly bias the final
measured concentrations. Due to spatial constraints at this site, all the background sample
locations were located along dirt access roads. Traffic on the access roads was observed during
several sample runs, likely contributing to measured concentrations above the true background
level. This phenomenon can be verified by inspecting the OPC time series data, which, due to the
short sample time of 20 seconds, can track changes in background levels as well as identify the
duration and quantity of contamination plumes. A discussion of the method used for determining
background PM concentrations for upwind contamination scenarios is presented in section 4.2.2.
The mean mass fraction of PMio and PM2.5 with respect to the measured TSP values for both
upwind and downwind samplers for each operation are presented in Table 15 and shown in
Figure 20. Upwind TSP was comprised of 28.8 ± 7.2% PM2.5 and 49.6 ± 16.4% PMi0.
Downwind TSP was comprised of 22.8 ± 15.7% PM2.5 and 50.0 ± 27.4% PMio. Overall, PM2.5
comprised 24.9 ± 13.5% of TSP and PMio comprised 49.8 ± 23.9% of TSP for the tillage
experiment. The TSP compositions along with mass concentrations are shown in Figure 20.
Table 15. Average (± lo) fraction of TSP that was PM2.5 and PM10 for each operation upwind and downwind
of tillage site, and campaign averages for upwind and downwind.
Date
10/19/2007
10/20/2007
10/23/2007
10/25/2007
10/26/2007
10/27/2007
10/29/2007
Upwind
PM2S/TSP PM10/TSP
0.22 ±0.05 0.28 ±0.04
0.21 ±0.04 0.34 ±0.03
0.28 ±0.01 0.68 ±0.08
0.31 ±0.06 0.57 ±0.14
0.32 ±0.08 0.47 ±0.05
0.31 ±0.04 0.53 ±0.21
0.39 ±0.05 0.60 ±0.08
Average Upwind
0.29 ±0.07 0.50 ±0.16
Downwind
PM25/TSP PM10/TSP
0.24 ±0.06 0.55 ±0.20
0.17 ±0.06 0.33 ±0.24
0.06 ±0.02 0.29 ±.0.5
0.21 ±0.03 0.38 ±0.13
0.20 ±0.07 0.60 ±0.31
0.52 ±0.11 0.84 ±0.21
Average Downwind
0.23 ±0.16 0.50 ±0.27
Operation
Chisel
Optimizer
Disc 1
Chisel
Disc 2A
Disc 2B
Land Plane
58
-------
250
Upwind Downwind Upwind Downwind Upwind Downwind Upwind Downwind Upwind Downwind Upwind Downwind Upwind Downwind
10/19/2007 10/20/2007 10/23/2007 10/25/2007 10/26/2007 10/27/2007 10/29/2007
Chisel Optimizer Disci Chisel Disc 2A Disc2B Land Plane
Figure 20. Average measured PM concentrations, upwind and downwind, with the particle size contributions
to the total PM.
PM produced by agricultural tillage operations tends toward larger diameter particles. According
to the U. S. EPA (1985), TSP emissions from agricultural tillage should be typically 21% PMio
and 4.2% PM2.5 [52]. This being the case, concentrations of PM2.5 should not vary greatly
between the upwind and downwind sampling locations, whereas concentrations of PMio and TSP
should be more variable, as seen in this study. As previously stated, the campaign averaged
PM2.5 downwind concentrations were 93% of those measured upwind. The average downwind
concentrations of PMio and TSP were generally significantly larger than upwind, at 131% and
137%, respectively, of averaged upwind levels. Figure 20 illustrates both the lack of significant
difference in upwind and downwind PM2.5 concentrations and the generally significant
differences between upwind and downwind PMio and TSP levels.
In order to calculate an emission rate from these concentration measurements, the mass
concentration observed at each sample location in the three size fractions that resulted from the
source activity must be known. The difference between the upwind (background) and downwind
concentrations is the result of the tillage emissions. Therefore, the downwind concentrations
were compared with the sample average upwind concentration on a location-by-location basis; a
location was included in the emission rate calculations only if the downwind concentration was
greater than the average upwind concentration plus the 67% confidence interval. However,
measured upwind concentrations were sometimes higher than downwind concentrations. This is
potentially a result of contaminated samples as discussed earlier. Ideally, the background PM
concentration for each operation was measured by an upwind tower distanced from the
operations so as not to be affected by varying wind direction, turbulent eddies created by the
operations during light and variable wind conditions, or traffic on the surrounding dirt roads.
While the impacts of such contamination events on filter-based, sample period-average MiniVol
data cannot directly be calculated, examination of time series data from a collocated OPC (based
59
-------
on 20 second sample times) provides a method to detect such contamination and estimate the
amount of contributed mass on the filter. We use a simple proportional correction to estimate the
background PM levels using the OPC time series data. These proportionally scaled PM values
can then be used for emission rate calculations. An OPC background concentration (OPCback)
may be found by taking an average of the OPC measured concentrations with any plume events
omitted. Then, the ratio of OPCback and the average OPC measured concentration for the sample
period (OPCave) can be used to scale the MiniVol concentration (Cave), which is the period
average PM concentration, to provide a background mass concentration (Chock), as in Eq. 24.
Due to the expected error of ±10% on the reported MiniVol concentrations, the calculated Chock
values were used in place of Cave for average upwind calculations only if the difference between
the PMio and TSP OPCback and OPCave concentrations were greater than 10%. In such cases, the
calculated PM2.5 Cback values were also used in place of Cave. It should be noted that data which
use the proportional scaling to estimate the upwind MiniVol concentrations are based entirely
upon OPC data.
The OPC-based proportional scaling method was applied to upwind locations for the chisel pass
of the combined operations tillage method and disc pass 1 and disc pass 2B of the conventional
tillage method. The proportionally calculated Chock values (feasibility was based on the ability to
identify and separate plume events) were then used as background PM concentrations in
determining facility produced concentrations. For these three sample periods, the statistical
significance of the differences between average upwind and downwind concentrations was
determined using the Chock values. In instances where the upwind TSP concentration was greater
than the downwind concentration (10/19 and 10/29), the method for determining PM background
levels using a collocated OPC was applied to the downwind measurement location. The derived
TSP Chock concentration was then used as the background value for TSP emission rate
calculations.
Table 16 presents the average upwind and downwind measured PM concentrations used for
comparison to determine which downwind locations could be used to calculate emission rates via
inverse modeling. Upwind averages that include background PM concentrations calculated using
OPC data are designated using an asterisk (*) and the background TSP concentrations calculated
using a collocated downwind OPC and MiniVol monitoring TSP are marked with a double
asterisk (**). Some of the downwind averages in Table 16 are slightly different than those
presented in Table 15 as those locations that were only slightly impacted by the operation under
study were removed from emission rate calculations. Such locations for removal were identified
through examination of reported MiniVol concentrations, OPC time series data, and model-
predicted impacts (i.e. the dispersion models predicted a concentration less than 10% of the
maximum weighted average for the sample period). Mean downwind concentrations of PM2.5,
PMio and TSP used for emission rate calculations were 101%, 152% and 183%, respectively, of
upwind levels. Comparisons of upwind average concentrations plus the calculated 67%
confidence interval versus location-by-location downwind locations determined that emission
60
-------
rates could be calculated for PMio during each sample period, TSP for all but 10/26 which is a
result of a sampling problem, and four of the seven sample periods for PM2.s.
Table 16. Upwind and downwind average concentrations ± 67% CI (for n>3) used in emission rate
calculations.
Date
10/19/2007
10/20/2007
10/23/2007
10/25/2007
10/26/2007
10/27/2007
10/29/2007
Upwind
Downwind
Upwind
Downwind
Upwind
Downwind
Upwind
Downwind
Upwind
Downwind
Upwind
Downwind
Upwind
Downwind
PM25
Hg/m3
32.9* ±3.0
25.7 ±1.8
17.9 ±1.8
27.8 ±5.3
15.4* ±0.8
11.8± 1.2
38.6 ±3.5
41.4±2.1
24.6 ±2.1
26.7 ±2.0
21.7* ±1.0
18.3 ±2.0
35.2 ±1.5
34.0 ± 1.6
PM10
Hg/m3
39.8* ±3.6
65.1 ±11.3
28.2 ±2.0
42.1 ±6.4
41. 8* ±3.0
63.1 ±3.1
70.5 ±10.0
78.4 ±7.8
41.3 ±2.2
54.7 ±7.4
26.1* ±4.5
62.5 ±9.6
49.2 ± 1.6
57.3 ±3.3
TSP
Hg/m3
81.0**
122.5
87.0
174.1
60.5
203.3
123.6
196.0
84.0
63.5*
84.3
53.5**
62.4
Tillage
Operation
Chisel
Optimizer
Disc 1
Chisel
Disc 2A
Disc 2B
Land Plane
* Adjusted PM concentration using OPC data
** Background level calculated using OPC and MiniVol data from the downwind location
4.2.2 ISC/AERMOD dispersion models
To better understand the emission locations and local rates, it is useful to apply dispersion
models to the filter data. The model settings and seed emission rate of 50 jig/s-m2 as explained in
section 3.2.1 were used in both air dispersion models and often produced estimated
concentrations higher than those measured, but generally within an order of magnitude.
However, the modeled concentrations represent only facility produced pollutant and do not
include background aerosol levels. Thus, to compare the modeled concentrations to field
measurements, the measured background PM concentrations must be subtracted from the
measured downwind concentrations.
61
-------
4.2.2.1 ISCST3
Modeled average concentrations by ISCST3 ranged from 0.0 to 683.0 ug/m3 with the highest
concentrations typically modeled at a height of 2 m on the southern edge of the tillage sites,
although this varied slightly with shifting wind directions. Figure 21 shows an example of
ISCST3 modeled concentrations at 2m above ground level for the Disc 1 pass as part of the
conventional tillage operations with 1.6 m/s average north winds.
120O
40
20
(M9/m3)
1600
550
500
1450
1400
350
300
1250
1200
150
100
50
0
400 600
800 1000 1200 1400 1600
E(m)
Figure 21. Modeled period average ISCST3 results (modeled hours = 8, sample time = 7.27 hrs) at 2m above
ground level for the Disc 1 pass of the conventional tillage operations on October 23, 2007 with light north
winds. The field area is outlined by the thick dashed line and sampler locations are shown in green; contour
line numerical values are in jig/m3.
4.2.2.2 AERMOD
Modeled average concentrations by AERMOD ranged from 0.0 to 438.3 ug/m3 with the highest
concentrations typically modeled at a height of 2 m on the southern edge of the tillage sites,
although this varied slightly with shifting wind directions. Concentrations predicted by
AERMOD were generally lower than those predicted by ISCST3 at the same emission rate,
meaning that AERMOD predicts greater dispersion of the particles under most conditions found
in this study. Figure 22 shows an example of AERMOD modeled concentrations at 2m above
ground level for the same Disc 1 pass shown in Figure 21.
62
-------
120O
40
20O
(M9/m3)
600
550
500
1450
1400
350
300
250
200
150
100
50
0
400 600
800 1000 1200 1400 1600
E(m)
Figure 22. Modeled period average AERMOD results (modeled hours = 8, sample time = 7.27 hrs) at 2m
above ground level for Disc 1 pass of the conventional tillage operations on October 23, 2007 with light north
winds. The area of operations is outlined by the thick dashed line and sampler locations are shown in green;
contour line numerical values are in jig/m3.
4.2.3 PM Chemical Analysis
4.2.3.1 Organic Carbon/Elemental Carbon Analyzer
The PM2.s OC/EC time series data collected at the downwind Air Quality trailer (AQT) is shown
in Figure 23. As can be seen the PlV^.s-associated elemental carbon was typically quite low,
averaging 0.5 ug/m3 ± 0.04 ug/m3 (at the 95% confidence interval) and showed relatively little
variability. This would suggest that the site was not significantly impacted by typical EC sources
such as biomass or diesel combustion and is more of a regional phenomenon. The observed
organic matter concentrations, derived by multiplication of the raw OC concentrations by 1.7
(refer back to section 3.1.4.3), were seven-to-eight times the EC concentrations averaging 3.8
|-ig/m3 ± 0.23 ng/rn3. Although the PM2.5 organic component seemed to vary more and
occasionally show greater concentration spikes, these episodes generally occurred during non-
test events. During observational periods of the agricultural testing, those time periods when co-
located filter-based PM samples were also collected, the elemental carbon PM2.5 concentrations
varied from 0.1 to 0.9 ng/m3, while the organic matter PM2.5 concentrations varied from 1.4 to
6.1 |-ig/m3. On average, the carbon-related material accounted for 28.4 percent of the observed
(downwind) PM2.5 mass (see Figure 24).
63
-------
-•-Organic Matter
-•- Elemental Carbon
10/15 10/16 10/17 10/18 10/19 10/20 10/21 10/22 10/23 10/24 10/25 10/26 10/27 10/28 10/29 10/30 10/31
Figure 23. PM2.5 OC/EC time series concentrations as collected at the downwind AQ trailer location. It
should be noted that the raw instrument OC concentrations have been multiplied by 1.7 to account for
potential non-carbon functional groups.
6 -
"S3 5
i; 4 -
3 -
2 -
Organic Matter
I Elemental Carbon
^ SF
-------
4.2.3.2 Ion Chromatographic (1C) Analysis
In addition to the near-real time carbon compositional analysis, water soluble ionic analysis was
performed on selected PM2.5, PMio, and TSP filters collected at the AQ trailer, as well as one
upwind (Nl) and downwind location (S2), to more completely account for the near source
particle chemical composition. It should be kept in mind that due to the availability of only one
real-time PM carbon system, ambient PM carbon content was only determined for one site
(AQT) and one size fractionation (PIVb.s). However, since the total (EC/OC) carbon fraction was
relatively consistent across the field tests, the assumption can be made the EC and OC
concentrations are conserved across the sampling fields, or in other words, the observed PM
carbon component may be more related to regional sources as opposed to the local tillage
operations.
All of the predicted ions where observed except nitrite which was not detected (n.d.) in any of
the samples. As might be expected if the presumed sources are more regional in nature, the ionic
species showed very little mass concentration differences between the upwind (Nl) and
downwind (AQT and S2) sample locations (see Table 17). On average, chloride, sulfate, and
nitrate were the most dominant anions, while sodium, ammonium, and potassium were the
dominant cations.
Table 17. Averaged filter ionic analysis for upwind and downwind samples for Oct. 23rd, 25th, and 26th, 2007.
Upwind
(Nl) PM25
Upwind
(Nl) PM10
Upwind
(Nl) TSP
Downwind
(S2, AQT)
PM25
Downwind
(S2, AQT)
PM10
Downwind
(S2, AQT)
TSP
F
(Hg/m3)
0.3
0.4
0.3
0.3
0.2
0.6
cr
(Hg/m3)
1.6
1.6
1.2
0.9
1.0
1.2
NO2
(Hg/m3)
n.d.
n.d.
n.d.
n.d.
n.d.
n.d.
scv2
(Hg/m3)
0.3
1.6
0.3
0.2
0.2
0.7
NO3"
(Hg/m3)
0.4
0.7
0.8
0.4
0.6
1.2
Na+
(Hg/m3)
1.4
1.6
1.6
1.1
1.0
2.0
NH4+
(Hg/m3)
0.4
0.8
1.0
0.5
1.0
1.4
K+
(Hg/m3)
0.6
0.7
0.6
0.4
0.5
0.9
Mg+2
(Hg/m3)
0.0
0.9
0.3
0.1
0.1
0.2
Ca+2
(Hg/m3)
0.3
1.0
1.0
0.4
0.6
1.4
65
-------
Nationwide fine PM composition has been shown by Malm (2000) to be dominated by five
aerosol classes: sulfates, nitrates, organic carbon, light-absorbing (elemental) carbon, and crustal
elements [58]. Through the 1C analysis of filter samples and the direct EC/OC measurements,
four of the five dominant PM types have been quantified; these four known PM types combine to
form only a few ng/m3. By mass balance, the remainder of the PM mass (on the order of tens of
Hg/m3) is composed of unanalyzed, and therefore unknown, constituents and is most likely
composed of insoluble crustal elements. Therefore, it can be assumed that the remaining fine
particle collected mass was likely dominated by soil-based particles. The bar chart in Figure 25
graphically demonstrates this crustal element dominance, shown as the "unknown" component in
the summation bars, at both the upwind and downwind sampling locations and across all size
ranges, with the relative contribution also increasing with particle size. The unknown, or crustal
portions, may have a combination of local and regional sources, which cannot be separated based
on the nature of the data collection and analysis. Magliano et al. (1999) found crustal/geological
elements comprised 49% to 66% of fall PMio measurements during the 1995 Integrated
Monitoring Study (EVIS95) in the San Joaquin Valley around Corcoran, CA, a small town also in
a highly agricultural area about 100 miles southeast of Los Banos, CA [59]. Results from the
IMS95 showed a factor of 2 variability in the mass contributed by geological sources, which
were assumed to be dominated by dirt road traffic and agricultural operations, between
monitoring sites that suggests strong site-specific, local geological influences. Over the three
specified days analyzed, on average, the upwind PM2.5, PMio, and TSP filter samples contained
59.4%, 69.2%, and 86.1% "unknown" species, respectively. Similarly, the downwind
"unknown" fractions were 69.4%, 77.9%, and 86.6% for PM2.5, PMio, and TSP, respectively.
250
Unknown
Organic C
Elemental C
Calcium
Magnesium
Potassium
Ammonium
Sodium
Nitrate
Sulfate
Nitrite
Chloride
Fluoride
Nl(9)2.5 Nl(9) 10 N1(9)TSP 52(9)2.5 52(9)10 52(9) TSP AQT2.5 AQT10
Figure 25. Chemical composition of downwind PM2.5, PM10 and TSP filters from the chisel pass in the
conventional tillage method, 10/25.
66
-------
4.2.4 Aerosol Mass Spectrometer
The aerosol mass spectrometer sampled at the Los Banos site from October 15, 2007 through
October 26, 2007. However, due to instrument malfunctions, data was only successfully
collected from 10/15-10/18/2007, prior to all the tillage operations and measurements. The mass
calibration of the instrument was drifting with time even within a saving period. This makes the
spectra difficult to interpret and very problematic to quantify, as it was necessary to go through
each spectrum by hand to verify the resolution of the ion peaks. The data may be able to yield
information from the period of 10/18-10/21/2007, but the data acquired after that appear
impossible to interpret. During the three days for which data meeting quality assurance tests was
obtained, overall mass concentrations ranged from -3-15 jig m"3. The chemical composition of
PMi particles as measured by the AMS during that time was dominated by ammonium nitrate
and organic matter. Ammonium nitrate made up -49.5% of the particulate mass, organic matter
made up 39%, and ammonium sulfate -9% (see Figure 26 and Figure 27). A calculation of mole
ratio shows complete neutralization between the basic species (ammonium ion) when compared
to acidic ones (nitrate, sulfate, and chloride ions). The chemical composition of the organic
material as measured in the mass spectra appears to be a mix of oxygenated species (likely
regional secondary oxidation products) and hydrocarbon-like compounds (probably a primary
aerosol source such as combustion emissions.) The oxygenated species are dominant as would be
expected in an environment dominated by secondary reaction chemistry.
NH4
• NO3
• SO4
• Org
• Cl
Figure 26. Average chemical composition of particles measured by AMS (~PMi) in Los Banos, 10/15/2007
10/17/2007.
67
-------
ioj-
10*-
10 n
10"-
|Air (m/z 28. 32)]
Nitrate (m/z 30. 46)|
Oxygenated organic matter (e.g. m/z 44, 55
[Hydrocarbon organic matter (e.g. m/z91, 115)
Polycyclic Aromatic Hydrocarbons (e.g. m/z 152. 166. 202) |
BO 100 120 14D 16D 1BO 2DO 220 240 260 2BD 30D 320
2D 40
Figure 27. Typical mass-to-charge (m/z) data collected at Los Banos, with significantly contributing ions and
their source particles identified.
4.3 OPTICAL CHARACTERIZATION DATA
4.3.1 Optical Particle Counter
The distribution of optical particle counters surrounding the fields of interest at two heights
provided the ability to examine particle emissions by number and size, as well as to see the time
series of PM concentrations through application of the MCF.
Examples of particle volume size distributions measured during tillage operations (dV/dd is the
change in aerosol volume concentration normalized by the change in particle diameter), in units
of |im3/cm3/|im, are presented in bar graph form in Figure 28. Each graph shows the background
particle volume distribution, a volume distribution of aerosols downwind of the source, and the
difference between the two that is the volume distribution of the emitted particles. It should be
mentioned that by examining these data based on volume concentration, the large particles have
a more visible effect than if number concentrations were examined. However, viewing the data
in this way is advantageous because it is analogous to mass - the shape of and difference
between the curves remains the same and only the scale changes as the transition to mass
concentration is made. Therefore, the given bar graphs show that the greatest volume (and mass)
of emitted aerosols is in the large particle range above a diameter greater than about 2.5 jim. As
seen in the reported PM sampler levels, the greatest contribution by tillage activities was in the
and TSP measurements.
68
-------
(a)
Combined Operations: Chisel
(b)
Combined Operations: Optimizer
Diameter dim)
Diameter (|im)
(c)
Conventional: Disci
(d)
Conventional: Chisel
Diameter (|im)
Diameter (|im)
Figure 28. Sample period average particle volume size distributions (|im3/cm3-|im) measured from upwind
(background) and downwind (background plus emissions) locations, with the difference being the aerosol
emitted by the tillage activity, (a) is the chisel operation of the combined operations tillage method, (b) is the
optimizer operation of the combined operations tillage method, (c) is the disc 1 operation of the conventional
tillage method, and (d) is the chisel operation of the conventional tillage method.
4.3.2 Optical to PM Mass Concentration Conversion
A critical factor in converting the lidar data from number density (volume) data to mass
concentration fields is the derivation of the Mass Conversion Factor (MCF). This step was
described in Eq. 14. The daily average MCF calculated from the distribution parameters
measured by OPCs, from both the upwind and downwind sides combined, are shown in Figure
28. Sample period average particle volume size distributions (|im3/cm3-|im) measured from
upwind (background) and downwind (background plus emissions) locations, with the difference
being the aerosol emitted by the tillage activity, (a) is the chisel operation of the combined
operations tillage method, (b) is the optimizer operation of the combined operations tillage
method, (c) is the disc 1 operation of the conventional tillage method, and (d) is the chisel
operation of the conventional tillage method.
69
-------
The particle size distributions measured by OPCs upwind and downwind of the source were used
in lidar retrievals to estimate the range dependent Vk (see Eq. 17). Collocated OPC and PM
sampler data were used to estimate the MCF as described in section 3.1.4.2. This in turn was
used to convert lidar-measured particle concentration to mass concentration units, which can be
compared with the PM sampler measurements.
MCF values estimated as a mean value for each day and for the whole campaign are presented in
Table 18, with the daily means ± the corresponding 95% confidence interval (CI) as in Figure 29.
Day to day variation in the MCF is not fully understood, but is likely due to changes in
background aerosol sources and composition, as the point samplers collected ambient aerosol for
a much larger fraction of the time than the tillage plume. The relatively high PM2.s MCF values
calculated for 10/20 is due to a drop in small particle (< 1.0 jim) counts by all the OPCs that day,
with respect to other days, while PM2.5 concentrations measured by PM samplers did not show a
similar drop. The reason for this phenomenon is unknown at this time.
Table 18. Mass conversion factors estimated for each day of the tillage operations and averaged for the whole
campaign. Error values represent the 95% confidence interval.
Date
PM25
PM10
TSP
Average
± 95% CI
n
Average
± 95% CI
n
Average
± 95% CI
n
Daily MCF (g/cm3)
19-Oct
3.40 ±
0.63
7
1.42±
0.14
7
2.51
2
20-Oct
4.90 ±
1.89
6
1.71 ±
0.50
7
2.77
2
23-Oct
2.38 ±
0.39
7
1.34 ±
0.16
6
0.63
1
25-Oct
2.16 ±
0.24
8
1.29 ±
0.20
7
1.09
2
26-Oct
2.92 ±
0.42
6
1.40±
0.18
7
1.03
1
27-Oct
2.57 ±
0.57
8
1.37 ±
0.59
8
0.84
2
29-Oct
2.73 ±
0.24
7
1.61 ±
0.10
6
1.16
2
Overall
MCF
(g/cm3)
2.95 ±
0.35
49
1.44±
0.13
48
1.53 ±
0.51
12
70
-------
7 -
6 -
I 4
LL.
L)
§ 3
2 -
1 -
0
-PM2.5
PM10
-TSP
19-Oct 20-Oct 23-Oct 25-Oct 26-Oct 27-Oct 29-Oct
Sample Date
Figure 29. Average daily MCF with error bars representing the 95% confidence interval.
After lidar measurements were converted to PM concentrations, a quality assurance step of
comparing collocated PM sampler, OPC, and lidar concentrations during 'stare' modes was
performed. The time series of routine lidar stares were plotted with the time series data from the
calibration OPC to ensure that trends and concentrations are the same in both data sets (Figure
30). It should be noted that lidar measurements were taken every 0.5 seconds while the OPCs
recorded 20 second samples. Therefore, the 40 lidar measurements are presented for every OPC
measurement.
71
-------
Upwind OPC vs. Lidar: PM,,,.
Upwind OPC vs. Lidar: PM
80
70
60
50
"l40
^.
30
20
10
0
13:41 14:10 14:38
Time
Upwind OPC vs. Lidar: TSP
Figure 30. PM2.S, PM10, and TSP mass concentrations retrieved from collocated lidar and OPC during the
'stare' time series for 10/25. Data acquisition time of the lidar data is 0.5 sec while OPCs were set to 20 sec
accumulation time. Measurements were done on the upwind side of facility (location Nl - "Pig" OPC).
To compare lidar and OPC retrievals with collocated PM sampler measurements, the lidar and
OPC time series were averaged and a 95% confidence interval was calculated over the
corresponding MiniVol sampling time. Results from these calculations for 10/23 are presented in
Table 19. The OPC (20 sec) and lidar (0.5 sec) collect data at a much higher rate than the
MiniVols (7.25 hrs for this sample run) and are able to capture the temporal variability of the
background aerosol concentration. On many days during this field trial the measured background
variability exceeded the uncertainty of lidar retrievals [46].
72
-------
Table 19. Comparison of PM mass concentrations (jig/m3) as reported by MiniVol samplers and mean values
measured by collocated OPCs and lidar at Nl (upwind) and S5 (downwind) for 10/23/2007.
Upwind
PM sampler (Nl)
Upwind PM sampler
average ± 67% CI
OPC (Nl) ± 95% CI
Lidar ± 95% CI
Downwind
PM sampler (S5)
Downwind PM sampler
average ± 67% CI
OPC (S5) ± 95% CI
Lidar ± 95% CI
PM2.5 (jig/m3)
17.0
15.5 ±0.8
13. 9 ±0.2
13. 8 ±0.2
9.9
11.7 ±1.2
12.8 ±0.2
41.7 ±9.0
PM10 Oig/m3)
35.9
40.0 ± 3.0
54.5 ±3.9
45. 9 ±0.9
74.5
62.9 ±3.0
63. 5 ±3.1
193.7 ±47.7
TSP (jig/m3)
60.5
60.5
65.6 ±6.3
60.1 ± 1.4
203.3
203.3
97.0 ±13.0
297.7 ± 76.6
A similar comparison of OPC and lidar time series data measured from downwind of the tillage
field is shown in Figure 30. The spikes of the concentrations (especially of PMio and TSP) are
due to the dust plume generated by tillage operations crossing this OPC. Spikes are rarely
observed on the upwind side of the field (upwind spikes are associated with road traffic).
As a quality check, comparison analysis of collocated lidar, OPC, and MiniVol data was done for
all days of campaign. An example of the lidar retrieved PM concentration vs. one of the OPCs
(Horse) is shown in Figure 31. In general, the OPC and lidar data averaged for the PM sampler
acquisition time are in good agreement with mass concentrations measured by the collocated PM
samplers. Discrepancies between point source instruments (PM samplers and OPC) and lidar are
due to inherently different measurement techniques. Point instruments capture particle
concentration at a single point (a few cm3) with a small volumetric flow (1.6xlO~5 to 8.3xlO~5
m3/s). The lidar acquires information in a volume of ~3 m3 for each bin along the laser beam for
each sample (2 samples per second in this experiment), and thus can capture a spatially variable
plume while the plume may miss the point sensors. The best agreement is observed when the
lidar is compared with PM sampler data averaged over several locations along the up- or down-
wind side of the tillage field. For similar reasons, using MCF values averaged over the whole
campaign yields larger discrepancies between collocated lidar and OPC/PM sampler PM
concentration data than daily averaged MCF values. Based on these observations, a daily
averaged MCF is used for conversion of lidar 'staple' data used for flux and emission rate
calculations.
73
-------
OPC vs. Lidar:
25
PM2.5
OPC vs. Lidar: PM_,
120
20
10:34
Time
10:34
Time
OPC vs. Lidar: TSP
Figure 31. PM2.5, PM10, and TSP mass concentrations retrieved from collocated lidar and OPC during
'staple' scanning (bottom point for the range bin of the OPC, collected from each staple shown). Data
acquisition time of the lidar data point is 0.5 sec while OPCs were set to 20 sec accumulation time.
Measurements were done on the downwind side of facility (location S5 - 'Horse' OPC) on 10/23/2007.
4.3.3 Lidar Aerosol Concentration Measurements
Examples of the lidar-derived upwind and downwind plume area average volume concentrations
used in the flux calculations shown below, are shown in Figure 32 and Figure 33. The two top
panels show the profile averaged wind speed and direction values used in the flux calculation,
with the third and fourth panels showing the area averaged volume concentrations measured
upwind (Cu) and downwind (Co) in |j,g3/cm3 (see Eq. 17). Wind speeds and directions for some
of the days that measurements were made were light and variable, a wind condition that is
known to challenge our flux measurement method. The concentrations derived from the lidar
scans to be used in flux calculations were carefully quality controlled to assure that the upwind
and downwind measurements were not contaminated by road traffic or mixed air flows that did
not represent the operation under test. Quality controlled data are presented in these plots and in
the table summaries. The quality control rejects experimental data on the basis of three
conditions that violate the process flux measurement assumptions:
74
-------
a. Wind direction > ± 80° at the time of the scan from the optimal wind direction of
360° based on the lidar location
b. External fugitive dust entering from the upwind side of the field due to traffic or other
non-stationary anomaly (e.g. dust devil).
c. Contamination of the reference OPC from the upwind side of the field.
Wind directions used to screen the lidar derived aerosol concentrations to be included in the flux
and emission measurements were limited to +/-800 of magnetic North. It was assumed that the
upwind concentration measurements would be more uniform than the downwind measurements
and more downwind scans were made than those of the upwind conditions. In Figure 32 and
Figure 33, the gaps in the upwind concentration are due to this sampling plan.
Windspeed
(m/s)
Wind Direction
20
10
0
50
0
-50
100
Upwind Cone.
(iun3/cm3) 1 Q
100
Downwind Cone.
(niti3/cm3) ln
x TSP
O PMt
+ PM
2.5
05
® ©
•-H--H-
to
+ -H-
XXX... X
•H. ++
1-
++
+"1"
t+**..»—
++ +H+
10
20 30
40 50
60
70
Figure 32. Wind speed, wind direction, upwind and downwind plume area average particulate volume
concentrations, for the October 20,2007 Optimizer pass of the combined operation tillage.
75
-------
15
10
Windspeed
(m/s) 5
0
100
Wind Direction 0
-100
Upwind Cone.
(mn3/cm3)
Downwind Cone.
(iiiti3/cm3)
100
100
* TSP
PM
2.5
HHH£ * '4+*+ W+H 4H+ H+-H-' +H+.
t
0
20
40
60
80
100
120
Figure 33. Wind speed, wind direction, upwind and downwind plume area averaged particulate volume
concentrations for the October 23,2007 first disc pass of the conventional tillage operation.
4.4 FLUXES AND EMISSION RATES
Emission rates were calculated using the lidar measurements, and compared with the model
results for comparison to previous studies.
4.4.1
Lidar based Fluxes and Emission Rates
The combination of'staple' and 'stare' measurements, as described in Section 3.1.5, of the mass
concentration distribution measurements from the upwind and downwind sides of the field were
performed continuously during the each tillage operation of the field campaign. Figure 34 and
Figure 35 show calculated net flux measurements for sequential lidar measurements taken during
the Optimizer pass in the combined operation tillage method on 10/20/2007 and the initial
discing pass of the conventional tillage operation on 10/23/2007, respectively. The net flux is the
product of the plume area averaged volume concentration (Co-Cu) difference multiplied by the
daily average MCF (see Table 18) and the component of the wind velocity that is perpendicular
to the lidar beam. This quantifies the mass passing through the lidar's vertical scan per unit time.
76
-------
30 40 50
Sample Number
Figure 34. Lidar derived fluxes (g/s) of PM2 s, PM10, and TSP for the October 20, 2007 Optimizer pass of the
combined operation tillage over the operation sample time of 2.85 hrs.
4
3
& 2
X
i 1
0
Q.B
0.6
0.4
0.2
D
-0.2
60
Sample Number
Figure 35. Lidar derived fluxes (g/s) of PM2.S, PM10, and TSP for the October 23, 2007 first disc pass of the
conventional tillage operation over the operation sample time of 7.27 hrs.
77
-------
Net fluxes were calculated using up- and downwind concentration measurements averaged over
each vertical scan and using average wind information for the time of the individual scan. Single
scan differences, of course, do not account for accumulation or depletion in the measurement box
due to wind speed variation during a scan, for input background variation, or for storage in or
flushing of the flux box due to the existing large scale wind eddy structure (i.e. we do not attempt
to measure the same air mass at the upwind and downwind scans). The process is assumed to be
a continuous emission source and several scans are required to achieve a meaningful mean
estimate of the facility emission. For calculation efficiency, the flux was calculated through the
downwind surface first and then the upwind flux, differencing the flux rather than concentration.
The plume area was chosen manually by observing each downwind scan and highlighting the
area to be included in the calculation. Choosing an area that fully includes the source plume but
not a lot of extra area eliminates the need to spend resources calculating for pixels which do not
contribute significant flux.
Using the series of flux measurements collected during each tillage operation, similar to those in
Figure 34 and Figure 35, the mean fluxes were calculated for all days and are shown in Table 20
with respective 95% confidence intervals. The error bars used here for the lidar data denote our
confidence in the mean due to the scan to scan variability due to the wind transport process
occurring on that day, and not to the precision of the accuracy of the individual measurements.
Measurement precision, as reported by Bingham et al. [60] show measurement accuracy on the
order of 0. Ig/s. This is evident in the initial scans shown in Figure 35, where transport across the
downwind plane was at or below the lidar system detection limit. Standard deviations of the
measurement sequences can be of the same magnitude as the fluxes under light and variable
wind conditions, with some scans showing very light, diffuse plumes crossing the lidar plane
followed by very dense ones.
Table 20. Mean fluxes (g/s) ± 95% confidence interval from quality controlled samples for each tillage
operation.
Operation
Combined : Chisel (10/19/07)
Combined: Optimizer (10/20/07)
Conventional: Disci (10/23/07)
Conventional: Chisel (10/25/07)
Conventional: Disc 2A (10/26/07)
Conventional: Disc 2B (10/27/07)
Conventional: Land Plane (10/29/07)
PM2.5 (g/s)
0.33 ±0.15
0.43 ±0.10
0.13 ±0.02
0.30 ±0.08
0.41 ±0.12
0.22 ±0.05
0.09 ±0.02
PM10 (g/s)
0.49 ±0.23
0.56 ±0.13
0.62 ±0.12
0.66 ±0.17
0.72 ±0.21
0.53 ±0.13
0.15 ±0.04
TSP (g/s)
1.91 ±0.88
2.23 ±0.51
1.00 ±0.19
1.96 ±0.50
1.45 ±0.43
0.90 ±0.22
0.22 ± 0.06
The relatively high uncertainty in the flux data of 10/19 is due to the relatively small number of
valid samples and the high variability of those samples due to the light and variable wind
conditions that were present on that day. This higher uncertainty is evident in the summary
calculations of the chisel operation for that day. The flux data presented in Table 20 were
multiplied by the total tractor operation time to yield a total mass emitted, and then normalized
by area tilled to calculate emission rates presented in Table 21. While the lidar-measured fluxes
for the Optimizer pass in Table 20 are higher than the chisel pass of the same treatment, the
emission rates reported in Table 21 for the Optimizer are lower due to the fact that the Optimizer
pass treats about two times as much area in the same amount of time as the chisel pass.
78
-------
The 95% confidence interval of the total mass emitted per day was calculated by using the
sample statistics (standard deviation and number count) for each day. The 95% confidence
interval for the total emission, on the other hand, was derived using the assumption that the
average emission rate of each day can be treated as a random variable. The variance of the
average emission rate of each day was assumed to equal the variance of the sample set divided
by the number of samples for that day. The 95% confidence interval for the daily mean was
calculated by assuming a Gaussian distribution and calculating the interval in which the daily
mean falls with a 95% probability. The variance of the total emission for each technique was
calculated as the sum of the variances of each operation. The 95% confidence interval was then
calculated again assuming a Gaussian distribution.
The lowest emission rate for each PM size fraction among the investigated operations was
derived for the land plane operation in the conventional tillage method. It should be noted that
emission rates and factors available in literature for land planning are much higher than all other
activities; in the CARB document listing emission factors for agricultural tillage land planning
has an emission factor ten times that of discing, tilling, and chiseling. This relationship between
the emission rates for land planing and discing/tilling/chiseling was not seen in this study, which
is likely due to the remaining water in the surface soil from the precipitation event that occurred
the evening of October 27th. The summed PM2.s, PMio, and TSP emission rates for the Combined
Operation and Conventional methods all have a statistically significant difference at the 95%
confidence interval calculated as described in the previous paragraph.
In summary, Aglite - in combination with OPC and PM samplers - indicates that the Optimizer
makes more concentrated plumes than does a conventional tillage rig, this is largely because the
Optimizer moves twice as much soil per tractor pass than a conventional implement. However,
since the Optimizer requires fewer tractor passes to till a field, the total amount of dust generated
is smaller. Therefore the emission rate for a field tilled by the Optimizer is smaller than for
conventional tillage.
Table 21. Aerosol mass transfer (± 95% confidence interval) from each field (flux normalized by operation
duration and area tilled) as calculated from lidar data for all tillage operations.
Operation
Combined: Chisel
Combined: Optimizer
Sum for Combined
Operation Method
Conventional: Disc 1
Conventional: Chisel
Conventional: Disc 2
Conventional: Land plane
Sum for Conventional
Method
PM25
(mg/m2)
45.3 ± 13.1
32.5 ±5.0
77.8 ± 14.0
20.4 ±2.6
35.8 ±5.9
39.5 ± 11.2
13.8 ±3.9
109.5 ± 13.5
PM10
(mg/m2)
69.0 ±19.9
42.7 ±6.6
111.6 ±20.9
99.7 ±12.5
79.5 ±13.1
80.7 ±20.5
21. 9 ±6.2
281.9 ± 28.0
TSP
(mg/m2)
265. 9 ±76.6
169.9 ±26.2
435.8 ± 80.9
159.8 ±20.0
235.1 ±38.8
149.3 ±40.3
33.4 ±9.4
577.6 ± 60.1
79
-------
4.4.2 ISCST3 Model Emission Rates
The individual downwind measured concentrations were evaluated to determine if they were
greater than the average upwind concentration plus the 67% confidence interval for each
operation. In locations where the downwind concentration was higher and the model predicted
concentration was greater than 10% of the maximum modeled concentration, emission rates were
calculated using the source-produced concentrations (downwind minus upwind), modeled
concentrations and the seed emission rate as shown in Eq. 22. The resulting emission rates are
shown in Table 22. To compare combined and conventional operations, the emission rates of
individual operations were normalized by the sample time. The mass per unit area of PM emitted
by the individual operations could then be summed to provide total mass emitted from the
combined and conventional tillage operations. The 95 percent confidence intervals reported in
Table 22 were calculated the same way as those for the lidar emission rates (see Section 4.4.1).
On two of seven days there were not any measured PM2.5 concentrations greater than the average
upwind concentration plus the 67% confidence interval. Therefore, no PM2.5 emission rates were
calculated for these sample periods. The second disc pass was carried out over two sample
periods, October 26 and 27* , due to farm equipment malfunctions on the 26*, as stated
previously. Single emission rate values for PM2.5 and PMio for the disc 2 pass were calculated by
averaging the emission rates on a mass per unit area basis across these two sample periods. Due
to the absence of a valid downwind TSP sample for October 26th and the model predicted
concentration at the downwind TSP sample location being about 5% of the maximum predicted
concentration on October 27*, the TSP emission rate for the disc 2 pass was calculated by
assuming that the PMi0/TSP emission rate ratio observed during the disc 1 pass of 0.18 was
representative of disc passes under similar conditions and then dividing the disc 2 PMio emission
rate of 204.2 mg/m2 by 0.18 to yield a TSP emission rate of 1130.5 mg/m2 for the operation. For
the land plane operation on 10/29, attempts to calculate the source impacts on the downwind TSP
sampler using a collocated OPC as described in Section 4.2.1 yielded a TSP emission rate
slightly less than the PMio emission rate for the same operation. Since PMio is a subset of TSP
and the PMio emission rate cannot physically be greater than the TSP emission rate, the TSP
emission rate was assigned the same value as the PMio emission rate as a conservative estimate
of actual TSP emissions. As seen in the lidar-based emission rates, the land plane operation
emission rates were lower than those for all the other operations, which is likely due to the
residual soil moisture from the precipitation event that occurred on October 27 .
The PMio emission rate for the chisel pass of the conventional tillage method is based on
calculations at two points, and therefore does not have a reported 95% confidence interval. The
difference between the summed PMio emission rates for the Combined Operations Method and
the Conventional Method are not significant at the 95% confidence interval. The Combined
Operations summed TSP emission rate was less than 25% of the Conventional Method, but the
statistical significance could not be determined. For PM2.s emission rates were not summed
because one data point for each method was not calculated.
80
-------
Table 22. Mean emission rates (± 95% CI for w>3) for each operation as determined by inverse modeling
using ISCST3.
Operation
Combined: Chisel
Combined: Optimizer
Sum for Combined Operation Method
Conventional: Disc 1
Conventional: Chisel
Conventional: Disc 2
Conventional: Land plane
Sum for Conventional Method
PM25
(mg/m2)
-
54.6 ± 86.0
-
-
63.7 ±81.0
18.2 ±5.7
36.3 ±32.4
-
PM10
(mg/m2)
103.6 ±83.4
79.0 ±94.9
182.5 ± 126.4
99.5 ± 64.0
122.5
204.2 ± 240.8
45.4 ±32.3
472.0 ± 289.5
TSP
(mg/m2)
180.1
321.3
501.3
550.8
296.7
1130.5
45.4*
2023.3
4.4.3
* Set equal to PM10 emission rate as a conservative estimate of actual TSP emissions
AERMOD Model Emission Rates
The inverse modeling technique was also applied to predicted concentrations from the
AERMOD air dispersion model; the calculated emission rates are shown in Table 23, with the
reported 95% confidence intervals calculated in the same manner as those for the lidar emission
rates (see Section 4.4.1). AERMOD emission rates were determined using the same techniques
described for the calculations of emission rates from ISCST3, including the calculation of the
TSP emission rate for the disc 2 pass based on the PMi0/TSP emission rate ratio for the disc 1
pass. However, with respect to the TSP emission rate for the land plane operation, a calculated
TSP emission rate greater than the PMio emission rate was achieved using the OPC based
background calculations and was used as the TSP emission rate instead of setting the TSP value
equal to the PMio value as was done in the ISCST3 emission rate calculations.
As in the ISCST3 data set, the PMio emission rate for the chisel pass of the conventional tillage
method is based on calculations at two points, and therefore does not have a reported 95%
confidence interval. As seen in the lidar- and ISCST3-based emission rates, the land plane
operation emission rates were lower than those of all the other operations, which is likely due to
the residual soil moisture from the precipitation event that occurred on October 27th. The
difference between the summed PMio emission rates for the Combined Operations Method and
the Conventional Method are not significant at the 95% confidence interval. The Combined
Operations summed TSP emission rate was about 25% of the Conventional Method, but the
statistical significance could not be determined. PM2.5 emission rates were not summed because
one of the data points for each method was not calculated.
81
-------
Table 23. Mean emission rates (± 95% CI for w>3) for each operation as calculated by inverse modeling using
AERMOD.
Operation
Combined: Chisel
Combined: Optimizer
Sum for Combined Operations Method
PM25
(mg/m2)
-
62.3 ± 96.4
-
PM10
(mg/m2)
185.3 ± 159.2
90.9 ±106.3
276.2 ± 191.4
TSP
(mg/m2)
322.6
367.9
690.5
Conventional: Disc 1
Conventional: Chisel
Conventional: Disc 2
Conventional: Land plane
Sum for Conventional Method
-
33.3 ±113.9
23.4 ±6.6
38.6 ±45.5
-
119.2 ±57.2
160.4
147.8 ±90.9
44.8 ± 10.8
472.2 ± 109.7
981.0
411.2
1216.6
58.4
2666.9
4.5 DERIVED EMISSION RATE COMPARISON
Two emission rate determination approaches were employed in this study to calculate three
different sets of emission rates in order to quantify differences between conventional tillage
methods and a combined operations tillage method using the Optimizer. Emission rates
calculated by the two models and the lidar, with their respective error estimates, are shown in
Figure 36 for PM2.5, PMio and TSP. Due to the measured downwind PM2.5 concentrations not
being statistically different from the upwind average for two of the seven sample periods, total
emissions were not calculated using ISCST3 and AERMOD models. Error estimates were unable
to be calculated for the model-derived TSP emission rates due to the number of upwind and
downwind TSP concentrations measurements limited to 1 each per day. The limited number of
points available to calculate model-derived emission rates, while carefully collected and
analyzed, show conclusively the limitations of trying to use even large arrays of point samples to
measure the emissions associated with weak fugitive dust sources having spatial and temporal
variations. The upwind and downwind concentration differences due to plume impact on the
scattered samplers on some days were simply below the detection limit of the sampling system,
especially for the PIVb.s size fraction. The lidar system, however, effectively sampled the vertical
downwind plane and measured time-resolved plume characteristics for each operation at each
particulate size fraction.
The advantage that lidar brings to these measurements is that lidar samples far more of the air
volume over the field than does an array of point samplers. Point samplers are stationary and
depend on the plume to pass over their specific location, and even then they only sample the few
cm3 near their inlet. Lidar, on the other hand, interrogates an entire "curtain" of air from ground
level up to 1000 m in height. In this way, there is no part of the plume that can exit the field
without the lidar seeing it. Lidar also addresses artifacts due to plume inhomogeneity. For
example, if a particularly concentrated portion of an otherwise very diffuse plume passes over a
OPC/MiniVol station, the ISCST3/AERMOD models will set the concentration of the entire
plume according to a smooth Gaussian function. On the other hand, what lidar will see the small
"hot spot" as a discontinuity in an otherwise very diffuse cloud. What lidar does that models
cannot is perform a physical integration of the actual plume.
82
-------
140
£ 120 -
"SB
Q.
100
80 -
60 -
40 -
20 -
NSC*
AERMOD * Lidar
Combined
Operations
Conventional
800
ISC AERMOD "Lidar
Combined Conventional
Operations
3000
Combined
Operations
Conventional
Figure 36. Summed PM2.5, PM10, and TSP emission rates ± 95% confidence intervals for both tillage methods
derived from lidar flux measurements and inverse modeling using ISCST3 and AERMOD. (* One PM2.5
emission rate missing per tillage method, therefore no total emissions were calculated.)
83
-------
While plume modeling is an invaluable tool for investigating different dispersion and transport
phenomena, it is no substitute for experimental data - especially data that provides full field
coverage and is highly spatially resolved. Models take as input a few sparse point
measurements, assign a plume shape, propagate the plume and then integrate the result. ISCST3
and AERMOD do not yet account for microturbulence or plume discontinuities, features that
lidar excels at identifying. In the tillage operations discussed in this report, the emission source
is (1) moving and (2) is small in comparison to the size of the field. It is not surprising then, that
emission modeling based on an array of few point samplers will yield different, and predictably
larger, emission rates than lidar-based emission measurements. By the same reasoning the error
associated with emission estimates should be smaller for lidar than for modeled emissions.
A summary table of the PMio emission rates calculated from each approach is given in Table 24
for comparison. The calculated PMio emission rates demonstrate that for total mass of PMio per
unit area the Combined Operations method produced between 40% and 60% as much as the
conventional method based on all employed measurement systems. The emission rate in mg/m2
of the Optimizer pass is less than the emission rate of the chisel pass in the same treatment for all
three emission rate calculations, despite the fact it resulted in higher downwind concentrations,
as measured by the lidar and point samplers. One reason for this is that the tillage rate
(hectares/hr) of the chisel pass was about half that of the Optimizer pass. As seen in Figure 36
and Table 21, the lidar calculated PM2.5, PMio, and TSP emissions were statistically different
between tillage treatments at the 95% confidence interval.
Table 24. Calculated PM10 emission rates (± 95% confidence interval) from the lidar and inverse modeling
using two dispersion models.
Date
10/19/2007
10/20/2007
Sum
Operation
Combined: Chisel
Combined: Optimizer
Combined Operations
Emission Rates (mg/m2)
Lidar
69.0 ±19.9
42.7 ±6.6
111.6 ±20.9
ISCST3
103.6 ±83.4
79.0 ±94.9
182.5 ± 126.4
AERMOD
185.3 ±159.2
90.9 ±106.3
276.2 ± 191.4
10/23/2007
10/25/2007
10/26-27/2007
10/29/2007
Sum
Disc 1
Chisel
Disc 2
Land Plane
Conventional
99.7 ±12.5
79.5 ±13.1
80.7 ±20.5
21.9 ±6.2
281.9 ± 28.0
99.5 ± 64.0
122.5
204.2 ± 240.8
45.4 ±32.3
472.0 ± 289.5
119.2 ±57.2
160.4
147.8 ±90.9
44.8 ±10.8
472.2 ± 109.7
In general, the largest difference between the three sets of calculated emission rates is between
the two models and the lidar. This is likely due to the difference in techniques - inverse
modeling using air dispersion models versus a mass balance approach to lidar flux
measurements. The models are limited to time steps and meteorological averages of 1 hour and
the types of sources that can be used (point source, volume source, area source, or line source),
which means that a compromise must be made when dealing with a small, moving area source
such as is the case in agricultural tillage. Also, the temporal and spatial resolution of the lidar
allows it to see micro-scale variations in plume characteristics and movement such as plume
strength, frequency, lofting and detachment from the surface, wind direction effects and wind
speed effects. The models must assume a constant emission from a ground level area source,
while the lidar sees the plume movement as the tractor moves across the field. This kind of
micro structure cannot be captured by the long term sampling required for implementation of
84
-------
ISCST3 and AERMOD, consequently ISCST3 and AERMOD are incapable of generating fine
levels of spatial and temporal detail. Examples of such differences in technique are evident when
comparing single scan and sample period average PM concentrations measured by the lidar with
average modeled concentrations at a vertical plane corresponding to the lidar scan area. Figure
37 presents the average modeled concentrations in ng/m3 for the chisel pass of the conventional
tillage method as determined by (a) ISCST3 and (b) AERMOD. It should be noted that
AERMOD better accounts for actual turbulence characteristics, which is the cause of the slight
differences between the two graphs. In both cases, the highest concentrations are modeled at just
above ground level with an exponential drop-off in concentration with increasing height above
ground level. Figure 38(a) presents the average lidar measured concentrations contributed by the
source (downwind concentrations minus the average upwind concentrations) for the same
operating period, while Figure 38(b) shows a single scan in which a significantly dense plume
has lofted above and is nearly detached from the surface. The plumes measured by the lidar were
detected much higher than the model predicts, as evident by the higher than background
concentrations measured at heights up to and exceeding 100 m in the average lidar measured
concentrations. As the point sensors were deployed near the surface (at 2 and 9 meters)
downwind of the source, these higher plumes were not sampled.
Another potentially important factor to consider is the exhaust emissions from the tractor
engines. Exhaust PMio emissions from the tractors were estimated in order to determine if their
contribution to total emissions could be significant. Using the fuel consumption information
provided by the cooperating farming company and the PMio emission rate as a function of fuel
used given by Kean et al. [27], the tractor exhaust contribution was calculated to be between
4.6% and 18.5% as shown in Table 25 when compared to the lidar-derived total PMio emissions.
We have chosen to use the lidar-derived emission values in Table 25 because they are the
smallest of the three emission values and will therefore be most sensitive to the effect of tractor
exhaust; tractor exhaust would be an even smaller fraction of the emission rates derived from
modeling. The calculated tractor exhaust contribution was greater than 10 percent of the overall
emission rates on only the land plane pass of the conventional tillage method. (Note: this day had
unusually low tillage associated emissions due to residual surface moisture from the precipitation
event on 10/27, as shown in Section 4.1.2.) Referring to the list of tractors used by operation
found in Table 3 and field notes recorded by personnel, the tractor used for the land plane pass
was the smallest and oldest of those used, while the other tractors were manufactured within the
last several years. The emission rate given by Kean et al. was calculated as an average for the
entire agricultural off-road fleet in 1996. Since that time, reductions in exhaust emissions have
been implemented, which is evident in the assigned emission rates that decrease with increasing
model year in the U.S. EPA's NONROAD emissions model [25], that would result in actual PM
emissions being overestimated. The estimated PMio exhaust emission data indicate that tractor
exhaust emissions should be a measurable contributor to total process particulate emission, but
the actual contribution may be too small to measure. Since the estimated tractor exhaust
emission amounts are smaller than the corresponding lidar measurement uncertainties for each
operation, the contribution of engine exhaust emissions to the total measured emissions is
swamped by the high intrinsic variability of tillage operation emissions. These data show that
the tractor emissions were only a small part of the total aerosol emissions from all operations,
and that the combined operations tillage produced 55% less tractor emissions.
85
-------
(a) ISCST3
700 900 1100
Distance from Lidar(n)
(b) AERMOD
700 900 1100
Distance from Lidar(m)
1300
(|jg/m3)
50
40
30
20
10
H
(|jg/m3)
50
40
30
20
10
0
Figure 37. Weighted average PM10 concentrations (|ig/m3) modeled by (a) ISCST3 and (b) AERMOD using
derived emission rates for 10/23 (modeled hours = 5, sample time = 4.24 hrs) along the downwind vertical
plane that corresponds to the lidar scanning plane. The green markers show point sampler locations on the
downwind side of the field. Maximum predicted concentrations for ISCST3 and AERMOD were 49.9 and
37.7 |ig/m3, respectively, at 2 m above ground level.
86
-------
(a)
10/23/2007 PM Downwind Averages minus Average Background (30.4 jig/m"'
(b)
350
300
250
200
150
100
I
500 600 700 800 900 1000
Range (m)
10/23/2007 Single lidar downwind scan
1100 1200
•80
70
60
50
40
20
10
0
-10
-20
]250
200
150
100
-50
500 600
700
800 900
Range (m)
1000 1100 1200
Figure 38. Lidar-measured downwind PM10 concentrations (a) averaged over all valid scans over the 4.24 hr
long sample period for 10/23 (w=122) and (b) for a single vertical scan, which demonstrates observed plume
lofting.
87
-------
Table 25. A comparison of lidar-based total calculated PM10 emissions from the tillage activities and
estimated tractor exhaust, calculated based on fuel usage and an emission factor of 3.23 g PM10/L fuel (Kean
et al. [26]).
Operation
Combined: Chisel
Combined: Optimizer
Sum for Combined
Operation Method
Conventional: Disc 1
Conventional: Chisel
Conventional: Disc 2
Conventional: Land
plane
Sum for Conventional
Method
Fuel Usage
(liters)
(Bowles
Farming Co.,
2008 [32])
340.7
174.9
515.6
396.0
221.4
312.3
100.7
1030.4
Estimated PM10 Exhaust
Emissions (kg)
(g)
1100.4
564.9
1665.3
1278.9
715.3
1008.7
325.2
3328.1
(mg/m2)
5.0
2.9
7.8
5.0
3.7
4.0
4.1
16.8
Lidar-derived
PM10
Emissions
(mg/m2)
68.4 ± 19.7
43.2 ±6.6
111.5 ± 20.8
95.9 ± 12.0
80.7 ± 13.3
79.6 ±20.8
22.2 ±6.3
278.4 ± 28.1
Exhaust % of
Lidar-derived
PM10
Emissions
(%)
7.3
6.7
7.0
5.2
4.6
5.0
18.5
6.0
5. SUMMARY AND CONCLUSIONS
A Regional Applied Research Effort (RARE) project to determine the control effectiveness of
Conservation Management Practices (CMPs) for agricultural tillage was awarded to the U.S.
EPA Environmental Sciences Division, National Exposure Research Laboratory. A study was
conducted to quantify particulate emissions (PM2.5, PMio, and TSP) from conventional
agricultural tillage methods and a CMP tillage method utilizing the Optimizer during after-
harvest land preparation. The Optimizer is a tillage implement that incorporates functions from
multiple conventional tillage implements into one piece of equipment. The objective of this study
was to address three fundamental research questions:
1) What are the magnitude, flux, and transport of PM emissions produced by agricultural
practices for row crops where tillage CMPs are being implemented vs. the magnitude, flux, and
transport of PM emissions produced by agricultural practices where CMPs are not being
implemented?
2) What are the control efficiencies of equipment being used to implement the "combined
operations" CMP? If resources allow assessing additional CMPs, what are the control
efficiencies of the "equipment change/technological improvements" and "conservation tillage"
CMPs?
3) Can these CMPs for a specific crop be quantitatively compared, controlling for soil
type, soil moisture, and meteorological conditions?
The study was carried out October 19-29, 2007 in the San Joaquin Valley of California on a
commercial production farm. Two adjacent fields with the same crop and similar soil properties
-------
were selected for observation. The CMP tillage treatment involved two passes over the field: 1) a
chisel pass, and 2) an Optimizer pass. The conventional tillage treatment comprised four passes:
1) a disc 1 pass, 2) a chisel pass, 3) a disc 2 pass, and 4) a land plane pass. Tractor malfunctions
required the disc 2 pass to be measured over 2 consecutive days, while all other measurements
were made on a single day for each pass. Particulate emissions were determined using arrayed,
filter-based sampling coupled with inverse modeling, using both ISCST3 and AERMOD, as well
as mass balance using advanced scanning lidar techniques. Supporting operation characteristics
(operation time, number of tractors in operation, potential contamination issues, etc.) were
recorded and meteorological, soil characteristic, and particle count and chemical composition
measurements were made and are reported in this document. The cooperating farming company
recorded tractor operation time and fuel usage for comparison.
Differences not statistically significant (based on the upwind 67% confidence interval) between
average upwind concentrations and individual downwind values prevented PM2.5 emission rate
determination using the models for three of the seven measurement periods. However, the
scanning lidar technology employed was able to calculate emissions for all measurement periods.
Table 26 summarizes important PMio aerosol emission values found during this study along with
results from previous studies found in literature. Emission factors in units of mg/m -pass were
converted to emission rates with units of mg/m2 by assuming a single pass for comparison. The
values herein reported are in occasional agreement with those reported by Flocchini et al. (2001)
and Madden et al. (2008), as well as the emission factors used by CARB to calculate area source
PMio contributions from agricultural tilling [3][4][13]. While the values from all three published
studies are generally not in close agreement, they are well within the range of the variability
expected from measurements made under different meteorological and soil conditions, as
demonstrated by the wide range of values from Flocchini et al. (2001) [3] summarized in Table
1. Both models and the lidar used to derive emission rates in this study report low emission rates
for the land plane pass of the conventional tillage method, which is likely due to water in the soil
surface layer from the precipitation event on 10/27 that had not yet evaporated. This finding is
consistent with the influence of soil moisture and other meteorological variables found by
Flocchini et al (2001) [3].
The many differences between how to extract lidar-derived and point sampler derived emission
rates have been detailed in previous section of this report. The critical distinction between lidar
and point sampler emission rates is primarily that of a paradigm shift. Lidar functions as a broad
array of volume point samplers, essentially covering an entire field with thousands of OPCs.
The analysis of the emission rate between the two methods differs in that a point-sampler based
model uses a mathematical function to draw a picture of a plume based on a handful of data
points whereas the lidar directly sums the results from all of its virtual OPCs to determine the
extent and concentration of the plume. Derivation of emission rates using modeling techniques
is an invaluable tool for investigating different dispersion and transport phenomena, this report
shows that it is no substitute for experimental data - especially data that provides full field
coverage and is highly spatially resolved. By their nature, models smooth point sampler data
and routinely overestimate the extent of the plume, therefore yielding different, and predictably
larger, emission rates when compared to lidar-based measurements.
89
-------
Table 26. A comparison of PM10 emission rates herein derived and found in literature.
Operation
Disc 1
Chisel
Disc 2
Land Plane
Conventional Method Sum
Flocchini
et al.
(2001) [3]
*
(mg/m2)
166.1
512.2
58.5
Madden
et al.
(2008) [4]
*
(mg/m2)
161.8
612.5
CARB
(2003a)
[13]
(mg/m2)
134.5
134.5
134.5
1401.0
1804.5
This study
Lidar
(mg/m2)
99.7 ±
12.5
79.5 ±
13.1
80.7 ±
20.5
21.9 ±
6.2
281.9 ±
28.0
ISCST3
(mg/m2)
99.5 ±
64.0
122.5
204.2 ±
240.8
45.4 ±
32.3
472.0 ±
289.5
AERMOD
(mg/m2)
119.2 ±
57.2
160.4
147.8 ±
90.9
44.8 ±10.8
472.2 ±
109.7
Chisel
Optimizer
Combined Operations Method Sum
512.2
134.5
69.0 ±
19.9
42.7 ±
6.6
111.6 ±
20.9
103.6 ±
83.4
79.0 ±
94.9
182.5 ±
126.4
185.3 ±
159.2
90.9 ±
106.3
276.2 ±
191.4
* Average of available data per operation
Lidar-derived emission rates for PM2.5, PMio, and TSP by operation are presented in Table 27,
along with the average tillage rate in hours per hectare, where the hour represents total tractor
operation time, and fuel use rate. In the case where two tractors were operating at the same time,
the total tractor operation time would be twice that of the elapsed time of the operation.
Comparisons were made between the conventional and CMP tillage practices based on these
same variables and are shown in the same table. The unitless control efficiency (r|) was
calculated according to Eq. 25, which was based on a collection efficiency equation found in
Cooper and Alley (2002).
I—js-TT< J—^n
E,
(25)
CT
where ECT is the calculated emission rate for the conventional tillage method and ECOT is the
calculated emission rate for the combined operations tillage method [61]. Therefore, the control
efficiency of the CMP for paniculate emissions was 0.29 ± 0.02, 0.60 ± 0.01, and 0.25 ± 0.01 for
PM2.5, PMio, and TSP, respectively.
The tillage rate was a significant factor resulting in the Optimizer pass emission rates being
lower than the chisel pass for the Combined Operation tillage method despite higher
concentrations measured during the Optimizer pass sample. The tractor time and fuel per hectare
required to perform the CMP work were 37.8% and 48.6% of the conventional method,
respectively. A similar reduction in tractor exhaust emissions is expected to have taken place, as
well as the supporting vehicle exhaust and PM emissions from driving on dirt access roads.
90
-------
Table 27. Lidar-derived participate emissions, tillage rate, and fuel usage comparison between conventional
and combined operations tillage.
Average Emission Rates ± 95% CI (mg/m2)
PM25 PM10 TSP
Average Tillage
Rate
(hrtractor/hectare)
Average Fuel
Usage
(L/hectare)
Combined Operations Method
Chisel
Optimizer
Sum
45. 3 ±13.1 69.0 ±19.9 265.9 ±76.6
32.5 ±5.1 42.7 ±6.6 169.9 ±26.2
77.8 ±14.0 111.6 ±20.9 435.8 ±80.9
0.38
0.21
0.59
15.5
8.7
24.2
Conventional Method
Disc 1
Chisel
Disc 2
Land plane
Sum
20.4 ±2.6 99.7 ±12.5 159.8 ±20.0
35. 8 ±5.9 79.5 ±13.1 235.1 ±38.8
39.5 ±11.2 80.7 ±20.5 149.3 ± 40.3
13. 8 ±3.9 21.9 ±6.2 33.4 ±9.4
1^'t± 281.9 ±28.0 577.6 ±60.1
1*5*^
0.44
0.33
0.37
0.42
1.56
16.0
9.7
10.7
10.7
47.1
Comparison of Tillage Methods
Combined Operations /
Conventional (%)
Statistically
Significant Difference
at 95% CI
Control Efficiency, n,
± Std Dev
71.1 39.6 75.5
Yes Yes Yes
0.289 ± 0.604 ± niAx + nM*
0.016 0.007 0.246 ±0.013
37.8
Unknown
—
48.6
Unknown
—
While not directly addressing the efficacy of CMP measures, an important component of this
investigation is to assess the utility of lidar for measuring particulate emissions in an agricultural
setting. Our lidar measurements clearly indicate that lidar is an effective tool for visualizing
plumes from tillage operations. Specifically, the lidar captures far more particulate matter
suspended at heights above 20 meters than either of the two models predict. This is a critical
observation for two reasons, (1) the emission characteristics from these tillage studies cannot be
accurately represented with either ISCST3 or AERMOD analyses because these models are
being used at the limit of their designed performance and do tend to overestimate process
emissions, and (2) there is substantially more vertical transport than previously thought, which
poses larger questions about the role of PM entrainment and transport away from the tillage site
Lidar, on the other hand, is capable of sensitively interrogating aerosol concentrations at
elevations up to several thousand meters. It is clear that the incorporation of lidar measurements
is an important complement to ground based sensors because ground based sensors cannot
measure elevated plumes. ISCST3 and AERMOD would realize significant benefits if lidar-
derived information could be incorporated into their calculations.
91
-------
6. ACKNOWLEDGMENTS
The Space Dynamics Laboratory would like to thank the individuals and groups whose efforts
made this study and subsequent analysis possible. Cooperators include the National Soil Tilth
Laboratory (Dr. Jerry Hatfield, Dr. John Prueger, and Dr. Richard Pfeiffer), Utah State
University (Mark Erupe, Dr. Randy Martin, Derek Price, Dr. Phil Silva), U.S. EPA Region IX
(Sona Chilingaryan, Kerry Drake, Andrew Steckel), Dr. David J. Williams of the U.S. EPA
Office of Research and Development, the San Joaquin Valleywide Air Pollution Study Agency,
the San Joaquin Valley Ag Technical Group, the San Joaquin Valley Air Pollution Control
District (Sheraz Gill, Samir Sheikh, Patia Siong, James Sweet), California Air Resources Board
(Kevin Eslinger, Shelby Livingston, Karen Magliano) and the cooperative agricultural producers
and industry representatives. Individuals from SDL who participated in the field measurement
campaign or contributed to the report include Doug Ahlstrom, Dr. Gail Bingham, Bill Bradford,
Jennifer Bowman, Carey Hendrix, Dr. Allen Howard, Derek Jones, Richard Larsen, Christian
Marchant, Kori Moore, Andrew Pound, Shane Topham, John Weaver, Dr. Tom Wilkerson, Dr.
Michael Wojcik and Dr. Vladimir Zavyalov.
7. PUBLICATIONS
G. E. Bingham, C. C. Marchant, V. V. Zavyalov, D. J. Ahlstrom, K. D. Moore, D. S. Jones, T. D.
Wilkerson, L. E. Hipps, R. S. Martin, J. L. Hatfield, J. H. Prueger, R. L. Pfeiffer. 2009.
Lidar based emissions measurement at the whole facility scale: Method and error
analysis. Journal of Applied Remote Sensing, 3(1), 033510 [doi: 10.1117/12.829411].
C. C. Marchant, T. D. Wilkerson, G. E. Bingham, V. V. Zavyalov, J. M. Andersen, C. B. Wright,
S. S. Cornelsen, R. S. Martin, P. J. Silva, J. L. Hatfield. 2009. Aglite lidar: A portable
elastic lidar system for investigating aerosol and wind motions at or around agricultural
production facilities. Journal of Applied Remote Sensing, 3(1), 033511 [doi:
10.1117/12.829412].
M. D. Wojcik, G. E. Bingham, C. C. Marchant, V. V. Zavyalov, D. J. Ahlstrom, K. D. Moore, T.
D. Wilkerson, L. E. Hipps, R. S. Martin, J. L. Hatfield, J. H. Prueger. "Lidar Based
Particulate Flux Measurements of Agricultural Field Operations" IGARSS 2008, Boston,
Massachusetts, July 2008.
M. D. Wojcik, K. D. Moore, "Laboratory for Atmospheric and Remote Sensing (LARS)", Wind
Erosion Workshop organized by the Agriculture Research Service - Wind Erosion
Research Unit, Manhattan, Kansas, April 2008.
V. V. Zavyalov, C. C. Marchant, G. E. Bingham, T. D. Wilkerson, J. L. Hatfield, R. S. Martin, P.
J. Silva, K. D. Moore, J. Swasey, D. J. Ahlstrom, T. L. Jones. 2009. Aglite lidar:
Calibration and retrievals of well characterized aerosols from agricultural operations
using a three-wavelength elastic lidar. Journal of Applied Remote Sensing, 3(1), 033522
[doi: 10.1117/12.833365].
92
-------
8. REFERENCES
[1] U.S. EPA. 1995. User's guide for the Industrial Source Complex (ISC3) Dispersion
Models. Research Triangle Park, NC: U.S. Environmental Protection Agency, Office of
Air Quality Planning and Standards Emissions. Monitoring, Analysis Division. January,
2008. http://www.epa.gov/scram001/userg/regmod/isc3v2.pdf.
[2] U.S. EPA. 2005. Revision to the Guideline on Air Quality Models: Adoption of a
Preferred General Purpose (Flat and Complex Terrain) Dispersion Model and Other
Revisions; Final Rule. 40 CFRPart 51. Washington, D.C., U.S. Environmental Protection
Agency. January 2008. http://www.epa.gov/ttn/scram/guidance/guide/appw_05.pdf
[3] Flocchini, R.G., James, T.A., Ashbaugh, L.L., Brown, M.S., Carvacho, O.F., Holmen,
B.A., Matsumura, R.T., Trezpla-Nabalgo, K., Tsubamoto, C. 2001. Interim Report:
Sources and sinks of PMio in the San Joaquin Valley. Crocker Nuclear Laboratory, UC-
Davis, CA.
[4] Madden, N.M., Southard, R.J., Mitchell, J.P. 2008. Conservation tillage reduces PMIO
emissions in dairy forage rotations. Atmospheric Environment 42:3795-3808.
[5] Pope, C.A. 1991. Respiratory hospital admissions associated with PMIO pollution in
Utah, Salt Lake, and Cache Valleys. Archives of Environmental Health 46(2):90-97.
[6] Hinds, W. C. 1999. Aerosol Technology: Properties, Behavior, and Measurement of
Airborne Particles, 2nd Edition. John Wiley & Sons, New York.
[7] 40 Code of Federal Regulations (CFR) 50.6. National primary and secondary ambient air
quality standards for PMi0.
[8] 40 CFR 50.7. National primary and secondary ambient air quality standards for PM2.5.
[9] Holmen, B.A., Eichinger, W.E., Flocchini, R.G. 1998. Application of elastic LIDAR to
PMio emissions from agricultural nonpoint sources. Environmental Science and
Technology 32:3068-3076.
[10] Conservation Management Practices Program Report. January 2006. San Joaquin Valley
Air Pollution Control District for 2005.
[11] Holmen, B.A., James, T.A., Ashbaugh, J.L., Flocchini, R.G. 2001a. LIDAR-assisted
measurement of PMio emissions from agricultural tilling in California's San Joaquin
Valley—Part I. LIDAR. Atmospheric Environment 35:3251-2364.
[12] Holmen, B.A., James, T.A., Ashbaugh, J.L., Flocchini, R.G. 2001b. LIDAR-assisted
measurement of PMIO emissions from agricultural tilling in California's San Joaquin
Valley—Part II: Emission factors. Atmospheric Environment 35:3265-3277.
93
-------
[13] California Air Resources Board (CARB). 2003a. Area Source Methods Manual, Section
7.4: Agricultural Land Preparation.
[14] CARB. 2003b. Area Source Methods Manual, Section 7.5: Agricultural Harvest
Operations.
[15] U.S. Environmental Protection Agency (EPA). 2001. Procedures document for national
emission inventory. Criteria Air Pollutants 1985-1999. EPA-454/R-01-006.
[16] Clausnitzer, H., Singer, MJ. 1996. Respirable-dust production from agricultural
operations in the Sacramento Valley, California. Journal of Environmental Quality
25:877-884.
[17] Clausnitzer, H., Singer, MJ. 2000. Environmental influences on respirable dust
production from agricultural operations in California. Atmospheric Environment
34:1739-1745.
[18] Baker, J.B., Southard, R.J., Mitchell, J.P. 2005. Agricultural dust production in standard
and conservation tillage systems in the San Joaquin Valley. Journal of Environmental
Quality 34:1260-1269.
[19] Mitchell, J.P., Southard, R.J., Madden, N.M., Klonsky, K.M., Baker, J.B., De Moura,
R.L., Horwath, W.R., Munk, D.S., Wroble, J.F., Hembree, K.J., Wallender, W.W. 2008.
Transition to conservation tillage evaluated in San Joaquin Valley cotton and tomato
rotations. Journal of California Agriculture 62(2): 74-79.
[20] Veenstra, J.J., Horwath, W.R., Mitchell, J.P., Munk, D.S. Conservation tillage and cover
cropping influence soil properties in San Joaquin Valley cotton-tomato crop. Journal of
California Agriculture 60(3): 146-153.
[21] Upadhyaya, S.K., Lancas, K.P., Santos-Filno, A.G., Raghuwanshi, N.S. 2001. One-pass
tillage equipment outstrips conventional tillage method. Journal of California Agriculture
55(5):44-47.
[22] Mitchell, J.P., Munk, D.S., Prys, B., Klonsky, K.K., Wroble, J.F., De Moura, R.L. 2006.
Conservation tillage production systems compared in San Joaquin Valley cotton. Journal
of California Agriculture 60(3): 140-145.
[23] U.S. EPA. March 17, 2008. NONROAD Model (nonroad engines, equipment, and
vehicles). Accessed: August 21, 2008. Available:
http ://www. epa.gov/otaq/nonrdmdl .htm#techrept.
[24] U.S. EPA. 2004. Exhaust and Crankcase Emission Factors for Nonroad Engine Modeling
- Compression-Ignition. EPA420-P-04-009. April 2004.
[25] U.S. EPA. 1991. Nonroad Engine and Vehicle Emission Study - Report. EPA 460/3-91-
02. November 1991.
94
-------
[26] CARB. 1999. Emissions inventory of off-road large compression-ignited engines (>25hp)
using the new OFFROAD Emissions Model. Mail-Out #MSC99-32. December 1999.
[27] Kean, A.J., Sawyer, R.F., Harley, R.A. 2000. A fuel-based assessment of off-road diesel
engine emissions. Journal of the Air and Waste Management Association 50:1929-1939.
[28] National Resource Conservation Service (NRCS). 2009. Web Soil Survey 2.1. 13 May
2009. http://websoilsurvey.nrcs.usda.gov/app/.
[29] Geology.com. May 2009. http://geology.com/state-map/california.shtml.
[30] Cooperating producer. September 2007. Personal communication.
[31] Tillage International, Inc. August 2008. Personal communication.
[32] Cooperating producer. October 2008. Personal communication.
[33] California Irrigation Management Information System (CEVIIS). 2007. Data for Station
#56 (Los Banos) for September and October of 2004 through 2006. September 2007.
http://wwwcimis.water.ca.gov/cimis/frontStationDetailInfo.do? station!d=56&urlPicDirec
tion=W.
[34] Tanner, C. B., and G.W. Thurtell. 1969. Anemoclinometer measurements of Reynolds
stress and heat transport in the atmospheric surface layer, Final Report, United States
Army Electronics Command, Atmospheric Science Laboratory, Fort Huachuca, Arizona.
[35] Lee, X., Finnigan, J. and K.T. Paw U. 2004. Coordinate Systems and Flux Bias Error. In:
Handbook of Micrometeorology: A Guide for Flux Measurement and Analysis. Lee, X.
Massman, W. and B. Law Eds. Kluwer Academic Publishers, Boston/Dordrecht/London
[36] Webb, E. K., Pearman, G. I, Leuning, R., 1980. Correction of flux measurement for
density effects due to heat and water vapor transfer. Quarterly Journal of the Royal
Meteorological Society 106:85-100.
[37] Doran, J.W., Jones, A. 1996. Methods for assessing soil quality. SSSA Special
Publication Number 49. Soil Science Society of America. Madison, Wisconsin.
[38] Soil Sampling and Methods of Analysis, ed by M.R. Carter, Canadian Society of Soil
Science. Lewis Publishers, 1993:508-509.
[39] Ibid. 659-662.
[40] Chen, F.-L., Williams, R., Svendsen, E., Yeatts, K., Creason, J., Scott, J., Terrell, D.,
Case, M. 2007. Coarse particulate matter concentrations from residential outdoor sites
associated with the North Carolina Asthma and Children's Environment Studies (NC-
ACES). Atmospheric Environment 41:1200-1208.
95
-------
[41] Chow, J. C., Watson, J. G., Lowenthal, D. H., Chen, L.-W A., Tropp, R. I, Park, K.,
Magliano, K. A. 2006. PM2.5 and PM10 Mass Measurements in California's San Joaquin
Valley. Aerosol Science and Technology 40(10):796-810.
[42] Airmetrics MiniVol Portable Air Sampler Operation Manual v. 5.
[43] Rupprecht & Patashnick, n.d. Series 5400 Elemental Carbon/Organic Carbon Analyzer
Instrument Manual.
[44] Malm, W.C. and J.L. Hand. 2007. An examination of the physical and optical properties
of aerosols collected in the IMPROVE program. Atmospheric Environment 41:3407-
3427.
[45] Marchant, C. 2008. Algorithm Development of the AGLITE-LIDAR Instrument, MS
Thesis, Utah State University.
[46] Zavyalov, V. V., Marchant, C. C., Bingham, G. E., Wilkerson, T. D., Hatfield, J. L.,
Martin, R. S., Silva, P. J., Moore, K. D., Swasey, J., Ahlstrom, D. J., Jones, T. L. 2009.
Aglite lidar: Calibration and retrievals of well characterized aerosols from agricultural
operations using a three-wavelength elastic lidar. Journal of Applied Remote Sensing,
3(1), 033522 [doi: 10.1117/12.833365].
[47] Klett, J.D. 1985. LIDAR inversion with variable backscatter/extinction ratio. Appl. Opt.
24: 1638-83.
[48] U.S. EPA. 2009. Guideline on Air Quality Models. Appendix W. U.S. Code of Fed.
Regulations, 40 CFR Part 51. May 2009.
http://www.epa.gov/scramOO l/guidance/guide/appw_05.pdf
[49] Cooper, D.C., Alley, F.C. 2002. Air pollution control: A design approach. Waveland
Press Inc. Prospect Heights, Illinois. 611-626.
[50] Turner, D.B. 1970. Workbook of Atmospheric Dispersion Estimates. Washington, D.C.,
U.S. Environmental Protection Agency.
[51] Paine, R.J., Lee, R.F., Erode, R., Wilson, R.B., Cimorelli, A.J., Perry, S.G, Weil, J.C.,
Venkatram, A., Peters, W.D. 1998. Model evaluation results for AERMOD. Research
Triangle Park, NC:U.S. Environmental Protection Agency, Office of Air Quality
Planning and Standards Emissions, Monitoring, Analysis Division. March 2008.
http://www.epa.gov/scram001/7thconf/aermod/evalrep.pdf.
[52] U.S. EPA. 1985. Compilation of air pollutant emission factors (AP-42), Fourth Edition:
Chapter 9. Research Triangle Park, N.C. U.S Environmental Protection Agency.
[53] U.S. EPA. 1995. Compilation of air pollutant emission factors (AP-42), Fifth Edition:
Chapter 9. Research Triangle Park, N.C. U.S Environmental Protection Agency.
96
-------
[54] National Research Council, National Academies of Science. 2003. Air emissions from
animal feeding operations: Current knowledge, future needs. The National Academies
Press. Washington, D.C. 95.
[55] Arya, S.P. 1998. Air Pollution Meteorology and Dispersion. Oxford University Press.
[56] Lakes Environmental. 2009. Terrain Data: 7.5-Min DEM Native Format - United States.
June 2009. http://www.webgis.com/terr_us75m.html.
[57] Massman, W.J., Lee, X. 2002. Eddy covariance flux corrections and uncertainties in
long-term studies of carbon and energy exchanges. Agricultural and Forest Meteorology
113:121-144.
[58] Malm, W. 2000. Spatial and seasonal patterns and temporal variability of haze and its
constituents in the United States: Report III, Interagency Monitoring of Protected Visual
Environments (IMPROVE) Report, May 2000.
[59] Magliano, K.L., Hughes, V.M., Chinkin, L.R., Coe, D.L., Haste, T.L., Kumar, N.,
Lurmann, F.W. 1999. Spatial and temporal variations in PMio and PM2.s source
contributions and comparison to emissions during the 1995 integrated monitoring study.
Atmospheric Environment 33:4757-4773.
[60] Bingham, G.E., Marchant, C. C., Zavyalov, V. V., Ahlstrom, D. I, Moore, K. D., Jones,
D. S., Wilkerson, T. D., Hipps, L. E., Martin, R. S., Hatfield, J. L., Prueger, J. H.,
Pfeiffer, R. L. 2009. Lidar based emissions measurement at the whole facility scale:
Method and error analysis. Journal of Applied Remote Sensing, 3(1), 033510 [doi:
10.1117/12.829411].
[61] Cooper, D.C., Alley, F.C. 2002. Air pollution control: A design approach. Waveland
Press Inc. Prospect Heights, Illinois. 100.
97
-------
9. APPENDICES
9.1 APPENDIX A
Table A 1. Settings for the ISCST3 and AERMOD dispersion models for the tillage study in the ISC-
AERMOD View software by Lakes Environmental, Inc. All settings were held constant across the sample
periods except the source area size and shape, which changed each day, and the downwind receptor locations,
which were specific to each field studied. (— = not applicable)
Setting
ISCST3
AERMOD
Control Pathway
Dispersion Options
Output type
Plume Deposition
Pollutant
Averaging Time
Dispersion Coefficient
Terrain Height Options
Terrain Calculation
Algorithms (ISC), Receptor
Elev./Hill Hghts (AERMOD)
Regulatory Default
Concentration
None
Other - PM
Period, 1-Hr
Rural
Elevated
Simple terrain only
Regulatory Default
Concentration
—
Other - PM
Period, 1-Hr
Rural
Elevated
Run using the AERMAP
Receptor Output file
Source Pathway
Source type
Base Elevation
Release height
Emission rate
Initial Vertical Dim.
of the plume
Building downwash
Area Poly
Imported from AERMAP
0.0 m AGL
5.0E-5g/(s-m2)
Blank
None
Area Poly
Imported from AERMAP
0.0 m AGL
5.0E-5g/(s-m2)
Blank
None
Receptor Pathway
Uniform Cartesian Grid
(# Receptors: 17940)
Discrete Cartesian Receptors
(# Receptors: 17)
138x130, 10x1 Om spacing,
flagpole height z = 2.0 m AGL
Placed at sample locations, z =
2, 5, and 9 m AGL
138x130, 1 Ox 10m spacing,
flagpole height z = 2.0 m AGL
Placed at sample locations, z =
2, 5, and 9 m AGL
AERMET View Settings
Hourly Surface Data
Adjustment to Local Time
Application Station
Elevation MSL
—
—
—
Source: on-site data, mixing
height = 1000m AGL
8 hr (Pacific)
29.2m
98
-------
Setting
Upper Air Data
Mode
Sectors Parameters
Time Zone
Randomize NWS Wind
Directions
Anemometer Height
Wind direction sectors
Land Use Type
Surface parameters per
sector
ISCST3
___
___
—
—
___
___
—
AERMOD
Source: calculated based on on-
site data, mixing height = 1000
m AGL
Upper Air Estimator
8 (Pacific)
Yes
6.2 m AGL
1: Start = 0°, End = 360°
Cultivated Land
Monthly, October default
values for Midday Albedo =
0.18, Bowen Ratio = 0.7, and
Surface Roughness = 0.05 m
Meteorology Pathway
Surface Met Data
Profile Met Data
Anemometer Height
Primary Met Tower Base
Elevation above MSL
Read Entire Met Data File
Specify Data Periods to
Process
Wind Speed Categories
Source: on-site data, mixing
height = 1000m AGL
—
6.2 m AGL
—
Yes (only provided data for
sample period times)
—
Default
Source: calculated from on-site
data and default values
Source: estimated from on-site
data
___
29.2m
No
Set to sample period times,
varied by sample
Default
Output Pathway
Tabular Outputs
All
Highest values table: 1st
Highest values table: 1st
Buildings
None
None
Terrain
Calculated values from
AERMAP using 7.5 Min DEM
Calculated values from
AERMAP using 7.5 Min DEM
99
-------
9.2 APPENDIX B
Table B 1. Participate matter concentrations measured by the Airmetrics MiniVol samplers shown by date,
location, and PM size fractionation.
Location
(height in m) Size
19- 20-
Oct Oct
Upwind - Combined Operations CMP
U1 W PM10
PM25
N1 (2) PM10
PM25
Nl (9) TSP
PM10
PM25
N2 (2) PM10
PM25
36.8 27.3
28.4 19.8
37.3 31.4
38.9 21.8
157.2 87.0
50.6 27.5
42.0 13.4
* 33.1
28.3 16.4
Downwind- Combined Operations CMP
W1 W PM10
PM25
E1 (9) PMlo
PM25
81 (2) PM10
PM25
81 (9) PMlo
PM25
82 (2) PMlo
PM25
S2 (9) TSP
PM10
PM25
83 (2) PM10
PM25
83 (9) PMlo
PM25
AQ1 (5) pMiQ
PM25
37.7 21.5
15.7
43.9 65.0
27.5 43.3
* *
32.5 *
48.2 23.0
26.6 *
53.6 *
29.2 19.7
122.5 174.1
** 40.9
40.0 *
57.3 37.7
24.5 43.8
55.4 29.5
20.6 20.8
111.2 56.7
21.1 27.5
Location
(height in m) Size
23- 25- 26- 27- 29-
Oct Oct Oct Oct Oct
Upwind - Conventional tillage practices
U2 (9) PMlo
PM25
N1 (2) PMlo
PM25
Nl (9) TSP
PM10
PM25
N2 (2) PMlo
PM25
* 69.2 38.1 19.3 *
16.4 43.1 22.6 20.9 32.0
43.3 53.4 43.7 40.3 55.2
16.9 40.5 34.2 25.1 *
60.5 123.6 84.0 70.2 90.6
35.9 88.9 35.8 35.5 47.1
17.0 42.6 23.4 19.4 33.3
* * * 54.5 48.1
14.3 28.1 26.4 23.0 39.1
Downwind - Conventional tillage practices
W2 (9) PMlo
PM25
E2 (9) PMlo
PM25
84 (2) PMlo
PM25
84 (9) PMlo
PM25
85 (2) PM10
PM25
S5 (9) TSP
PM10
PM25
86 (2) PM10
PM25
86 (9) PMlo
PM25
AQ2 (5) pMiQ
PM25
Tripod 1 pMiQ
PM25
39.7 63.2 31.2 31.5 45.8
58.1 75.2 49.7 — 49.9
12.9 51.4 23.9 — 36.3
69.3 * 48.8 57.4 55.0
10.2 47.1 19.6 28.2 32.8
58.2 111.0 40.3 92.4 *
* 43.7 21.3 16.3 33.2
* 70.6 71.6 67.0 68.3
14.0 35.0 * 18.9 35.4
203.3 196.0 * 84.3 63.1
75.5 59.0 48.6 30.7 58.6
9.9 40.4 25.4 14.4 27.1
62.3 91.3 73.9 59.5 47.1
15.8 38.8 33.9 14.3 36.0
54.9 28.5 64.5 16.3 57.4
13.8 33.6 28.0 7.7 39.4
* * 20.0 * 25.5
5.8 41.5 22.0 13.9 17.8
fiS 0
— — \J^) .\J —
18.1
— Data point not collected
* Compromised by sampler or human error
** Removed due to reported PM2 5 value with ±10% error bars being greater than
collocated PM10 value ±10% (i.e., the ±10% error bars did not overlap)
100
-------
-------
&EPA
United States
Environmental Protection
Agency
Office of Research
and Development (8101R)
Washington, DC 20460
Official Business
Penalty for Private Use
$300
EPA/600/R-12/673
December 2012
www.epa.gov
Please make all necessary changes on the below label, detach
or copy and return to the address in the upper left hand corner.
If you do not wish to receive these reports CHECK HERE n!
detach, or copy this cover, and return to the address in the
upper left hand corner.
PRESORTED STANDARD
POSTAGE & FEES PAID
EPA PER MIT No, G-35
Recy cl ed/Recy cl able
Printed with vegetable-based ink on
paper that contains a minimum of
50% post-consumer fiber content
processed chlorine free
------- |