Harvesting equipment to reduce particulate matter emissions from almond harvest

Abstract

Almond harvest accounts for an estimated 12 Gg of PM₁₀ emissions in California each harvest season. Emissions from three new, “low-dust” almond harvesters (Exact Harvest Systems E4000; Flory Industries 8550; Weiss-McNair 9800 California Special) and one exhaust abatement device (Joe DiAnna, Clean Air Concept) were compared to those from a conventional harvester operating in the same orchard. Emissions of TSP and PM₁₀ trended lower for all new harvesters and were significantly lower for most harvesters (α < 0.10). Significant reductions in PM_2.5 emissions were observed from two harvesters as well. Fractionation analysis was not conducted on nut samples collected in the second year of the project, but differences observed in the composition of material that would be delivered to the huller between the Exact E4000 and conventional harvesters were functionally insignificant. The results of these tests imply that new harvest technologies are able to reduce PM₁₀ emissions from one of the largest sources in the San Joaquin Valley (SJV) of California without affecting product quality. As such, use of these new harvesters should be considered a conservation measure that would help the SJV Air Pollution Control District (SJVAPCD) meet the requirements of their PM₁₀ maintenance plan.

Implications:

The results of this research indicate that new harvesting technologies have the potential to substantially reduce PM emissions from almond harvest operations over traditional harvester designs without negatively affecting product quality. As such, use of these new harvesters could aid the SJVAPCD in maintaining its attainment status for PM₁₀ and should be considered as candidate conservation management practices for producers.

Introduction

Almond harvest accounts for an estimated 12 Gg of PM₁₀ emissions in California each harvest season and is one of the largest sources of PM₁₀ in the San Joaquin Valley (CARB, 2003). In 2010, approximately 748 Gg of almonds was harvested in California on approximately 299,000 bearing hectares with a total value of almost $2.7 billion (USDA, 2011). Over 70% (210,000 ha) of the bearing crop is located within the San Joaquin Valley (SJCV) Air Pollution Control District (SJVAPCD), which has required agricultural operations to implement Conservation Management Practices (CMPs) to reduce PM₁₀ emissions as part of the district's U.S. Environmental Protection Agency (EPA)-approved plan to bring the San Joaquin Valley (SJV) airshed into compliance with National Ambient Air Quality Standards (NAAQS) for PM₁₀. Although the SJV airshed was redesignated as being in compliance with the PM₁₀ NAAQS in November 2008, sources must continue to implement conservation measures, including CMPs, as part of the SJVAPCD's federally approved PM₁₀ maintenance plan (SJVAPCD, 2007).

During almond harvest, trees are shaken to remove almonds from the tree and allow them to dry on the ground. Several days later, sweepers move the almonds into windrows before nut pickup machines remove the product from the orchard. The current emission factor applied to all almond harvesting operations is 4,570 kg PM₁₀/km², of which 90% (4,120 kg/km²) are attributed to nut pickup, with the remaining 10% being attributed to shaking (41.5 kg/km²) and sweeping (415 kg/km²) operations.

For several years, the Almond Board of California has been investigating abatement strategies to reduce PM emissions from almond harvest operations, including changes in sweeping practices (Downey et al., 2008; Goodrich et al., 2009; Faulkner et al., 2011; Faulkner and Capareda, 2012) and harvester operation (Downey et al., 2008; Faulkner et al., 2009; Ponpesh et al., 2010; Faulkner et al., 2011). These studies have shown promise for reducing PM₁₀ emissions from sweeping and total suspended particulate (TSP) emissions from nut pickup (which affects visibility), but they have shown little promise for reducing PM₁₀ emissions from nut pickup without adversely affecting product quality.

Little work has been published considering modifications to machine design for the purpose of reducing PM emissions. Southard et al. (1997) analyzed the effect of cleaning chain length on PM generation, and Whitelock et al. (2007) developed a prototype cyclone system for removing PM from the exhaust stream of a pecan harvester. The device developed by Whitelock et al. (2007) has shown promise in pecan orchards, but the large size of the system makes application in almond orchards problematic.

Equipment manufacturers have responded to increased attention to PM emissions from almond harvesting equipment by developing “low-dust” harvesters. Flory Industries (Salida, CA) markets its 850 and 8550 models as lower emission alternatives to the traditional Flory 480 harvester based on changes in cleaning chain length, fan speed, and the location from which dust is pulled from the harvested product (Flory Industries, 2006). Similarly, Weiss-McNair (Chico, CA) markets their 9800 California Special as being “designed to help California meet all government clean air and water regulations” largely based on changes to fan speed, fan location, and cleaning chain design (Weiss-McNair, 2010). Exact Harvesting Systems (Modesto, CA) produces the E4000 harvester, which uses a water misting and brush system on the discharge of the separation fan to reduce PM emissions from the nut pickup operation and slow the speed of the discharge air (Exact Corporation, 2012). Joe DiAnna (Modesto, CA) produces a cyclonic separator (Clean Air Concept) that can be retrofit onto the discharge side of the fan of several makes and models of nut harvesters.

Low emissions harvesters must demonstrate measurable reductions in PM₁₀ emissions relative to conventional harvesters if producers are to receive CMP credits for their use. The objectives of this study were to:

1.	Compare emissions from new almond harvesting systems and retrofit abatement devices to emissions from a conventional harvester.
2.	Quantify differences in the foreign matter content of almonds harvested with the new harvesters relative to that from a conventional harvester.

Methods

Sampling location

Emissions from new harvesting systems and those from a conventional harvester (Flory 480; Flory Industries; Salida, CA) were measured in an orchard in Kern County in September 2010 and 2011. The orchard was planted in a Wasco sandy loam soil and irrigated using subsurface drip irrigation. Trees were 15–16 years old and were planted in approximately 200-m (0.125-mile) rows oriented in a north–south direction with 7.3 m (24 ft) between rows and 6.4 m (21 ft) between trees in the same row.

Almond growers commonly plant a combination of cultivars in a given orchard to achieve cross pollination. The usual combination is a Nonpareil cultivar with a “pollinator” cultivar or a Nonpareil with two “pollinator” cultivars in each orchard. In many orchards, including that in which sampling was conducted, the Nonpareil cultivars are planted every other row with the other cultivars planted on an alternating basis. The orchard in which sampling was conducted was planted with Nonpareil, Price, and Monterrey cultivars (Figure 1).

Figure 1. Layout of sampled orchard (not to scale; adapted from Faulkner et al., 2011).

Experimental design

2010

The Exact Harvesting Systems E4000 harvester (Modesto, CA) was tested against a Flory 480 harvester in September 2010. Both harvesters were pulled by John Deere 6430 tractors operating at ground speeds of 4.8 km/h (3 mph) and power takeoff (PTO) speeds of 540 rpm.

Each plot consisted of 12 tree rows (with the exception of the first two plots, which were composed of 10 tree rows each). Emissions were evaluated during harvest of the Price cultivar. During the harvest of a given tree row, windrows on both sides of the tree are used for pickup. Therefore, while each plot consisted of 12 tree rows, only six windrows were created, virtually using half of the total orchard area for harvest.

A randomized complete block experimental design was used with replication as the blocking factor. Two treatments (Exact E4000 harvester and conventional harvester) were investigated using nine replications for a total of 18 plots.

2011

The Flory Industries 8550 harvester (Salida, CA), Weiss-McNair 9800 California Special (Chico, CA), and Joe DiAnna's Clean Air Concept (Modesto, CA) retrofitted onto a Weiss-McNair 9800 California Special were each tested against a Flory 480 harvester in September 2011 during Nonpareil harvest. Each plot consisted of 10 tree rows, and all harvesters operated at a ground speed of 4.8 km/h (3 mph) and a power takeoff (PTO) speed of 540 rpm (or equivalent fan speed for the self-propelled Flory 8550). Each new harvester was tested on a separate day, so comparisons cannot be made between emissions from each of the new harvesters due to potential biases between days that result from both identified (e.g., differences in meteorological conditions or soil moisture) and unidentified sources of variance (Box et al., 2005). Within each day, a randomized complete block experimental design was used with replication as the blocking factor. Two treatments (“low-dust” harvester or retrofit and conventional harvester) were investigated using five replications for a total of 10 plots per harvester comparison.

Emissions calculations

Data collection and analysis were conducted using similar methods as in previous almond harvest emissions studies that are described elsewhere (Faulkner et al., 2009; Faulkner et al., 2011; Faulkner and Capareda, 2012). In summary, during nut pickup operations, collocated, low-volume TSP and federal reference method (FRM) PM₁₀ and PM_2.5 samplers were placed nominally upwind and downwind of each plot to measure the change in ambient PM concentrations due to harvest operations. One collocated set of samplers was located upwind of each plot while four collocated sets of samplers were spaced evenly along the width of each plot approximately 15 m (50 ft) downwind from the northern or southern edge of the plot (Figure 2). (Note: Due to equipment limitations, FRM PM_2.5 samplers were located only at sampler sets S2, S3, and S4 in Figure 2. Upwind [S5] PM_2.5 concentrations were calculated as described later.)

Figure 2. Sampler configuration (not to scale). Tree rows ran perpendicular to the “S1–S4” sampler array.

TSP concentrations were measured alongside FRM PM₁₀ and PM_2.5 samplers using samplers designed by Wanjura et al. (2005) to reduce variations in sampler flow rate that lead to high uncertainty in FRM TSP concentration measurements. PM₁₀ and PM_2.5 measurements were conducted using the same airflow control units as the TSP samplers and an FRM PM₁₀ sampling inlet (model PQ100 inlet; BGI, Inc.; Waltham, MA). PM_2.5 samplers were also equipped with a Very Sharp Cut Cyclone (VSCCA, BGI, Inc.; Waltham, MA) upstream of the filter.

The particle size distribution (PSD) of PM collected on TSP filters having more than 200 μg of PM was analyzed using a Beckman Multisizer 3 Coulter Counter (Beckman Coulter; Miami, FL). The PSD (described by a log-normal mass distribution) of each sample was measured and characterized by a mass median diameter (MMD) and geometric standard deviation (GSD) (Hinds, 1999). The MMDs were converted from equivalent spherical diameter (ESD) to aerodynamic equivalent diameter (AED) using a particle density (ρ_p) of 2.6 g/cm³ (based on soil particle samples taken from the orchard and analyzed using a pycnometer; AccuPyc 1330, Micrometrics, Norcross, GA) and a shape factor of 1.05 ( eq 1). A particle shape factor of 1.05 was used, given the slightly aspherical shape of soil particles collected from the orchard during sampling as seen using a scanning electron microscope (Figure 3):

Figure 3. Scanning electron microscope image of collected PM particles.

(1)

where AED is the aerodynamic equivalent diameter, ESD the equivalent spherical diameter, ρ_p the particle density (g/cm³), and χ the shape factor.

Upwind PM_2.5 concentrations were estimated using measured TSP concentrations and PSDs. FRM PM_2.5 samplers are designated to have a fractional efficiency curve (FEC) with a cut point of 2.5 ± 0.2 µm and a “sharp” slope that has been estimated to be equal to 1.186 (U.S. EPA, 1996). Therefore, applying such an FEC to the upwind TSP sample should allow one to estimate the concentration of particles that would be collected by an FRM PM_2.5 sampler ( eq 2):

(2)

where: C_theor is the theoretical concentration of PM collected by a size-selective PM sampler, C_TSP the concentration of total suspended particulate (TSP), f(x) the probability density function of particle size distribution of the PM, and ϵ(x) the fractional efficiency of the sampler preseparator at capturing particles of size x.

Buser et al. (2007) reported that FRM PM₁₀ and PM_2.5 samplers measure concentrations of PM₁₀ or PM_2.5 higher than those expected based on the theoretical calculations in environments dominated by large particles such as agricultural areas. Therefore, applying such an FEC likely understates the concentration of particles that would be measured by an FRM PM_2.5 sampler (Buser et al., 2008). By underestimating upwind PM_2.5 concentrations, a conservative (i.e., higher than actual) estimate of the contribution of harvest operations to downwind PM_2.5 concentrations was produced. (The maximum calculated upwind PM_2.5 concentration using this technique was 5 µg/m³.)

Because of the biases reported by Buser et al. (2007), “true” concentrations of PM₁₀ and PM_2.5 at downwind sampling locations were estimated using measured TSP concentrations and expected sampler performance characteristics (cut point of 2.5 µm and slope of 1.186 for PM_2.5 samplers; cut point of 10 µm and slope of 1.5 for PM₁₀ samplers; Hinds, 1999).

During concentration measurements, the following instruments were used to collect onsite meteorological data in an open area north of the sampled orchard:

•	A two-dimensional (2D) sonic anemometer (WindSonic1, Gill Instruments Ltd., Lymington Hampshire) was used to measure the wind speed (accuracy ± 2%) and direction (accuracy ± 3 deg.) 3m above the ground surface at a frequency of 4 Hz.
•	A three-dimensional (3D) sonic anemometer (model 81000, R. M. Young Co., Traverse City, MI) was used to collect data for use in defining the stability of the surface layer (accuracy: wind speed ± 1% rms; wind direction ± 2 deg.) at 2 m above the ground at a sampling frequency of 4 Hz.
•	A barometric pressure sensor (model 278, Setra Systems, Inc., Boxborough, MA; accuracy ± 0.25%) recorded every 5 min.
•	A temperature (accuracy ± 0.5°C) and relative humidity (accuracy ± 1.5%) probe mounted in a solar radiation shield at 2 m (HMP50, Campbell Scientific Inc., Logan, UT) recorded every 5 min.
•	Two pyranometers, one mounted face up (CMP 22, Kipp and Zonen, Delft, The Netherlands; accuracy ± 3 W/m2) and one mounted face down (CMP 6, Kipp and Zonen, Delft, The Netherlands; accuracy ± 4 W/m2), were used to measure net solar radiation at a sampling frequency of 5 min.

Additional meteorological parameters were calculated according to U.S. EPA guidance (U.S. EPA, 2004). The dimensions of each test plot and corresponding meteorological data were then used with the American Meteorological Society/Environmental Protection Agency Regulatory Model (AERMOD) to determine fluxes (µg/m²-sec) from each of the downwind samplers for each sampling period according to the protocol described by Faulkner et al. (2009). Each of the sampler sets used at each plot provided an independent measurement of concentration leading to four independent estimates of the TSP and PM₁₀ emissions flux for each plot and three independent estimates of the PM_2.5 emissions flux. These fluxes were considered repeated measurements of emissions for a given plot, such that 11 average fluxes (four TSP, four FRM PM₁₀, and three FRM PM_2.5) were used to determine the emissions from each plot. (In 2010, these emissions were then multiplied by 2 to determine emission rates, as the second half of each plot would be used for harvest of the Monterrey cultivar.)

After omitting incomplete data points due to missing meteorological parameters, insufficient mass increase of filter, and so on, analysis of variance (ANOVA) tests were conducted using the General Linear Model Repeated Measures function in SPSS (SPSS v. 14.0; SPSS, Inc., Chicago, IL). Each of the measurements from a given plot was analyzed for within-subject variation, while the two harvest treatments were analyzed for between-subject variation. The null hypothesis tested was that the means from each treatment were equal.

Size fractionation

After nuts were picked up, they were transferred to a nut cart that was pulled by the harvester. From the nut cart, almonds were transferred to a bank-out machine that carried them to a load out where some cleaning occurred as nuts were elevated into the hopper bottom trailers that carried them to the huller. A 2-gallon bucket was used to collect a sample of almonds and foreign material from each plot from the load-out stream as it entered the hopper bottom trailer. By collecting samples at this point, the sample was representative of the raw product that would be transferred to the huller. Each sample was placed in a resealable freezer bag and was transferred to a laboratory for mechanical fractionation.

Each sample from 2010 was placed in a sieve series and mechanically separated as described by Faulkner et al. (2009). With the exception of the original sample size, sieving was conducted according to ASTM Standard C 136-06 (ASTM, 2006). (The sample size required by ASTM Standard C 136-06 was too large for practical application in the described research.) Retained materials on the separate sieves were collected and weighed to establish whether differences existed between harvesters. The size ranges used for foreign matter characterization are shown in Table 1.

Table 1. Fractionation categories

Particle size
Minimum	Maximum	Category
16 mm	—	Nuts and twigs
9.5 mm	16 mm	Leaves, small nuts, and twigs
8 mm	9.5 mm	Leaves and grass
1 mm	8 mm	Grass
—	1 mm	Soil

Samples from 2011 were irreparably damaged before analysis during shipment from the orchard to the laboratory. No visibly noticeable differences in sample composition (e.g., large differences in visible dirt, leaf, or sticks) were observed between harvest methods during collection of samples in 2011.

An analysis of variance (ANOVA) test was conducted using the General Linear Model function in SPSS (SPSS v. 14.0; SPSS, Inc.; Chicago, IL) to determine whether differences existed in the composition of products delivered to the huller between harvesters. The null hypothesis tested was that the mean percentages of sieved samples in each size range from each harvester were equal. Means were compared with the least squares difference (LSD) pairwise multiple comparison test.

Results and Discussion

Exact E4000 harvester

Meteorological conditions during sampling of the Exact E4000 harvester are shown in Table 2. Average characteristics of PSDs measured from TSP samples are shown in Table 3. In order to achieve equal variances during analysis, an inverse square root transformation was applied to sample MMDs and a power transformation (λ = –1.25) was applied to sample GSDs. Neither the MMD (p = 0.699) nor GSD (p = 0.671) of PM emitted during harvest operations varied by harvester type.

Table 2. Meteorological parameters measured onsite during sampling (Exact harvester)

	Exact harvester			Conventional harvester
	Minimum	Maximum	Average	Minimum	Maximum	Average
Albedo	0.21	0.35	0.23	0.20	0.24	0.22
Relative humidity (%)	20	55	33	21	47	31
Temperature (°C)	24	30	27	24	31	27
Solar radiation (W/m^2)	5	843	512	54	1,008	550
Wind speed (m/sec)	0.5	2.6	1.7	0.5	2.7	1.5

Table 3. Particle size distribution parameters from TSP filters (Exact harvester)

	Exact harvester	Conventional harvester
MMD (µm AED)^[	15.3 x	16.2 x
GSD	2.11 x	2.20 x
Average % PM10	30.6	29.1
Average % PM2.5	0.8	1.0
Note: No statistical differences were detected in means in the same row followed by the same letter (α = 0.05). MMD = mass median diameter. AED = aerodynamic equivalent diameter. GSD = geometric standard deviation. Average % PM10 obtained utilizing eq 2 with d50 = 10 µm and FEC slope = 1.5.

Emissions calculations

Before statistical analysis, data were analyzed for outliers and to ensure that the assumptions of normal distribution and equal variance between treatments that are required for standard analysis of variance (ANOVA) techniques were satisfied. The following samples were omitted from analysis:

•	Test 3 (Exact harvester) was omitted due to low filter masses (not statistically different from upwind). Soon after this test was started, the wind shifted from the southeast to the west such that the downwind samplers were no longer near the center of the plume.
•	Test 18 (conventional harvester) was omitted due to extremely high calculated emission rates. Shortly after this test was started, extremely stable atmospheric conditions set in, causing little plume movement. Faulkner et al. (2008) demonstrated that AERMOD is increasingly sensitive to changes in wind speed at values below 2 m/sec. As wind speeds approach these low ranges, the Monin–Obukhov length calculated by AERMOD's meteorological preprocessor (AERMET) decreases below 1 m, indicating an extremely stable atmosphere that rarely occurs in most locations. These stable atmospheric conditions lead to unusually high estimates of emissions for a given downwind concentration.
•	Samples from location S3 during test 13 (Exact harvester) were omitted because erroneously high sampler flow rates led to poor sampler performance.
•	Samples from location S1 during test 9 (Exact harvester) were omitted because emission rates from this location lay more than three standard deviations from the mean emission rate for this test.

After these samples were omitted, a log transformation was applied to TSP, true PM₁₀, true PM_2.5, and FRM PM₁₀ emission rates, while an inverse square root transformation was applied to FRM PM_2.5 emission rates to produce normally distributed transformed data for analysis. Average emissions derived from ambient concentration measurements, inverse dispersion modeling, and particle size analysis are shown in Table 4.

Table 4. Emissions (kg/km²) from harvest treatments (Exact harvester; n = 8 per treatment)

	TSP	True PM10	True PM2.5	FRM PM10	FRM PM2.5
Emissions
Exact E4000	1,149	369	12	561	264
Conventional	4,835	1,300	39	1,981	401
Reduction	76%*	72%*	69%*	72%*	34%
Significance
Within-subject	0.373	0.524	0.275	0.930	0.840
Between-subject	0.002	0.013	0.006	0.002	0.931
Note: Reductions marked with asterisks were significant at p = 0.10.

Within-subject differences were not significant for emission rates of any of the particle size classes analyzed, indicating that emission rates calculated from each sampler location (S1, S2, S3, and S4) did not vary significantly. However, differences in emissions between harvesters were significant for all sizes except FRM PM_2.5. The Exact harvester produced 76% fewer TSP emissions and 72% fewer PM₁₀ emissions. Differences in emission rates determined using FRM PM_2.5 samplers were not significant, but utilizing TSP samples and particle size analysis, the Exact harvester reduced emissions of particles smaller than 2.5 µm AED by 69%.

Size fractionation

Multivariate ANOVA results indicated no significant differences in foreign matter content by harvest treatment (p = 0.474 using Wilks' lambda). Similarly, within each size range, no pairwise differences were significant with the exception of soil content, for which the Exact harvester resulted in a lower soil content (Table 5; p = 0.040). However, because multivariate analysis of variance (MANOVA) results were not significant, the chances of a type I error in which the null hypothesis is rejected even though it is true are increased, and thus differences in soil content should be viewed with caution. Furthermore, the observed differences in soil content are functionally insignificant in that most processors would not penalize growers for the soil content observed in samples harvested using the conventional harvester.

Table 5. Size separation results (Exact harvester)

	Mass fraction (%)
Size separation range	Exact harvester	Conventional harvester
Nuts and twigs	75.7 (2.6) x	72.2 (2.6) x
Leaves, small nuts, and twigs	19.6 (2.0) x	20.5 (2.0) x
Leaves and grass	0.8 (0.1) x	1.1 (0.1) x
Grass	1.5 (0.4) x	2.5 (0.4) x
Soil	2.4 (0.4) x	3.7 (0.4) y
Note: No significant differences were detected in means in the same row followed by the same letter (α = 0.05). Standard errors are shown in parentheses.

Flory Industries 8550 harvester

Meteorological conditions during sampling of the Flory Industries 8550 harvester are shown in Table 6. Average characteristics of PSDs measured from TSP samples are shown in Table 7. Neither the MMD (p = 0.500) nor GSD (p = 0.269) of PM emitted during harvest operations varied by harvester type.

Table 6. Meteorological parameters measured onsite during sampling (Flory harvester)

	Flory 8550 harvester			Conventional harvester
	Minimum	Maximum	Average	Minimum	Maximum	Average
Albedo	0.23	0.30	0.25	0.21	0.27	0.24
Relative humidity (%)	13	24	18	13	23	18
Temperature (°C)	32	35	34	30	35	34
Solar radiation (W/m^2)	331	846	681	469	826	678
Wind speed (m/sec)	1.1	2.3	1.7	0.6	2.0	1.4

Table 7. Particle size distribution parameters from TSP filters (Flory harvester)

	Flory 8550 harvester	Conventional harvester
MMD (µm AED)^[b,c]	19.3 x	18.2 x
GSD^[d]	1.86 x	1.90 x
Average % PM10 ^[e]	18.6	21.3
Average % PM2.5	0.1	0.1
Notes: No statistical differences were detected in means in the same row followed by the same letter (α = 0.05). MMD = mass median diameter. AED = aerodynamic equivalent diameter. GSD = geometric standard deviation. Average % PM10 obtained utilizing eq 2 with d50 = 10 µm and FEC slope = 1.5.

Emissions calculations

Before statistical analysis, data were analyzed for outliers and to ensure that the assumptions of normal distribution and equal variance between treatments that are required for standard analysis of variance (ANOVA) techniques were satisfied. Samples collected from location S1 during tests 4, 5, and 9 (Flory 8550 harvester) were omitted from analysis due to extremely high calculated emission rates. During these tests the wind direction was such that these samplers were located on the edge of the plume, where uncertainties in back-calculated emission rates are much higher than for samplers located near the center of the plume (Faulkner et al., 2007).

After these samples were omitted, a square-root transformation was applied to true PM_2.5 emission rates to produce normally distributed, transformed data for analysis. No transformations were applied to other emission rates. Average emissions derived from ambient concentration measurements, inverse dispersion modeling, and particle size analysis are shown in Table 8.

Table 8. Emissions (kg/km²) from harvest treatments (Flory harvester; n = 5 per treatment)

	TSP	True PM 10	True PM 2.5	FRM PM 10	FRM PM 2.5
Emissions
Flory 8550	2,281	436	2	1,034	359
Conventional	3,891	802	5	1,599	280
Reduction	41%	46%*	69%*	35%*	−28%
Significance
Within-subject	0.787	0.201	0.945	0.771	0.813
Between-subject	0.178	0.021	0.050	0.061	0.502
Note: Reductions marked with asterisks were significant at p = 0.10.

Within-subject differences were not significant for emission rates of any of the particle size classes analyzed, indicating that emission rates calculated from each sampler location (S1, S2, S3, and S4) did not vary significantly. However, differences in emissions between harvesters were significant for PM₁₀ and PM_2.5 calculated from TSP samples as well as FRM PM₁₀ samples, with the Flory 8550 harvester producing 35% fewer PM₁₀ emissions than the conventional harvester. Differences in emission rates determined using FRM PM_2.5 samplers were not significant, but utilizing TSP samples and particle size analysis, the Flory 8550 harvester reduced emissions of particles smaller than 2.5 µm AED by 69%.

Weiss-McNair 9800 California Special harvester

Meteorological conditions during sampling of the Weiss-McNair 9800 California Special harvester are shown in Table 9. Average characteristics of PSDs measured from TSP samples are shown in Table 10. Neither the MMD (p = 0.886) nor GSD (p = 0.253) of PM emitted during harvest operations varied by harvester type.

Table 9. Meteorological parameters measured onsite during sampling (Weiss-McNair [WM] harvester)

	WM 9800 CA Special harvester			Conventional harvester
	Minimum	Maximum	Average	Minimum	Maximum	Average
Albedo	0.20	0.27	0.24	0.20	0.31	0.24
Relative humidity (%)	12	20	16	14	23	19
Temperature (°C)	30	36	34	28	35	33
Solar radiation (W/m^2)	479	824	690	323	838	692
Wind speed (m/sec)	0.4	2.7	1.7	0.6	2.3	1.7

Table 10. Particle size distribution parameters from TSP filters (Weiss-McNair harvester)

	Weiss-McNair harvester	Conventional harvester
MMD (µm AED)	20.9 x	20.8 x
GSD	1.87 x	1.90 x
Average % PM10	16.0	16.6
Average % PM2.5	0.1	0.1
Notes: No statistical differences were detected in means in the same row followed by the same letter (α = 0.05). MMD = mass median diameter. AED = aerodynamic equivalent diameter. GSD = geometric standard deviation. Average % PM10 obtained utilizing eq 2 with d50 = 10 µm and FEC slope = 1.5.

Emissions calculations

Before statistical analysis, data were analyzed for outliers and to ensure that the assumptions of normal distribution and equal variance between treatments that are required for standard analysis of variance (ANOVA) techniques were satisfied. Samples collected from location S1 during test 20 (conventional harvester) were omitted from analysis due to extremely high calculated emission rates. During this test the wind direction was such that these samplers were located on the edge of the plume, where uncertainties in back-calculated emission rates are much higher than for samplers located near the center of the plume (Faulkner et al., 2007). All results from tests 11 (Weiss-McNair harvester) and 12 (control) were also omitted. Soon after each of these tests commenced, the wind shifted such that samplers were only downwind for a portion of the test duration.

After these samples were omitted, a square-root transformation was applied to TSP emission rates while an inverse square-root transformation was applied to FRM PM_2.5 emission rates to produce normally distributed transformed data for analysis. No transformations were applied to other emission rates. Average emissions derived from ambient concentration measurements, inverse dispersion modeling, and particle size analysis are shown in Table 11.

Table 11. Emissions (kg/km²) from harvest treatments (Weiss-McNair harvester; n = 4 per treatment)

	TSP	True PM 10	True PM 2.5	FRM PM 10	FRM PM 2.5
Emissions
WM 9800 CA Special	5,095	1,036	3	1,453	121
Conventional	6,865	1,137	5	2,628	313
Reduction	26%	9%	29%	45%	61%*
Significance
Within-subject	0.107	0.032	0.226	0.524	0.105
Between-subject	0.394	0.614	0.418	0.127	0.003
Note: Reductions marked with asterisks were significant at p = 0.10.

Within-subject differences were significant only for “true” PM₁₀ emission rates. Emissions from the new harvester trended lower for all constituents but were significant only for FRM PM_2.5, where emissions were reduced by 61%.

DiAnna's Clean Air Concept cyclone

Meteorological conditions during sampling of the harvester equipped with DiAnna's Clean Air Concept are shown in Table 12. Average characteristics of PSDs measured from TSP samples are shown in Table 13. The MMD (p = 0.040) and GSD (p = 0.031) of PM emitted during harvest operations were lower for the cyclonic abatement system. The efficiency of cyclonic abatement systems increases with particle size, so a reduction in median size of emitted particles could be expected from such a system.

Table 12. Meteorological parameters measured onsite during sampling (DiAnna cyclone)

	Harvester with cyclone			Conventional harvester
	Minimum	Maximum	Average	Minimum	Maximum	Average
Albedo	0.23	0.28	0.25	0.18	0.40	0.24
Relative humidity (%)	13	18	15	13	38	19
Temperature (°C)	33	37	35	26	36	32
Solar radiation (W/m^2)	448	830	701	187	819	640
Wind speed (m/sec)	1.5	2.4	2.0	0.4	2.6	1.8

Table 13. Particle size distribution parameters from TSP filters (DiAnna cyclone)

	Cyclone retrofit	Conventional harvester
MMD (µm AED)	22.1 x	18.7 y
GSD	1.96 x	1.86 y
Average % PM10	15.5	19.8
Average % PM2.5	0.1	0.1
Notes: No statistical differences were detected in means in the same row followed by the same letter (α = 0.05). MMD = mass median diameter. AED = aerodynamic equivalent diameter. GSD = geometric standard deviation. Average % PM10 obtained utilizing eq 2 with d50 = 10 µm and FEC slope = 1.5.

Emissions calculations

Before statistical analysis, data were analyzed for outliers and to ensure that the assumptions of normal distribution and equal variance between treatments that are required for standard analysis of variance (ANOVA) techniques were satisfied. Samples collected from location S1 during tests 24 and 30 (harvester with cyclone) were omitted from analysis due to extremely high calculated emission rates. During these tests the wind direction was such that these samplers were located on the edge of the plume, where uncertainties in back-calculated emission rates are much higher than for samplers located near the center of the plume (Faulkner et al., 2007).

After these samples were omitted, a log transformation was applied to TSP, true PM₁₀, FRM PM₁₀, and FRM PM_2.5 emission rates while a square-root transformation was applied to true PM_2.5 emission rates to produce normally distributed transformed data for analysis. Average emissions derived from ambient concentration measurements, inverse dispersion modeling, and particle size analysis are shown in Table 14.

Table 14. Emissions (kg/km²) from harvest treatments (DiAnna cyclone; n = 5 per treatment)

	TSP	True PM 10	True PM 2.5	FRM PM 10	FRM PM 2.5
Emissions
With cyclone	2,292	337	1	771	14
Conventional	7,871	1,600 y	8	3,156	279
Reduction	71%	79%	88%	76%*	95%*
Significance
Within-subject	0.152	0.352	0.431	0.916	0.879
Between-subject	0.145	0.138	0.357	0.031	0.005
Note: Reductions marked with asterisks were significant at p = 0.10.

Within-subject differences were not significant for emission rates of any of the particle size classes analyzed, indicating that emission rates calculated from each sampler location (S1, S2, S3, and S4) did not vary significantly. However, differences in emissions between harvesters were significant for FRM PM₁₀ and PM_2.5. The Clean Air Concept cyclone reduced emissions of PM₁₀ and PM_2.5 by 76 and 95%, respectively.

Conclusions

Emissions from three new, “low-dust” almond harvesters and one exhaust abatement device were compared to those from a conventional harvester operating in the same orchard. No differences were detected in the PSD characteristics of PM emitted from each harvester, with the exception of the DiAnna Harvesters Clean Air Concept cyclone, where large particles were efficiently captured by the cyclone. Emissions of TSP and PM₁₀ trended lower for all new harvesters and were significantly lower for most harvesters. Significant reductions in PM_2.5 emissions were observed from the Weiss-McNair 9800 California Special harvester and the DiAnna Harvesting Clean Air Concept. Although fractionation analysis was not conducted on nut samples collected in 2011, few differences were detected in the composition of material that would be delivered to the huller between the Exact E4000 and conventional harvesters.

Presented results were derived from measurements conducted in a single orchard in Kern County. Emissions reductions in other production areas and under different orchard management practices may differ somewhat between machines. However, the results of these tests imply that new harvest technologies are able to reduce PM₁₀ emissions from one of the largest sources in the SJVAPCD without affecting product quality. As such, use of these new harvesters should be considered a conservation measure that would help the SJVAPCD meet the requirements of their PM₁₀ maintenance plan.

Acknowledgments

The researchers acknowledge support from the USDA Natural Resources Conservation Service Conservation Innovation Grant program (grant 68-9104-0-127) and the San Joaquin Valley Air Pollution Control District for funding this project, along with participating equipment manufacturers and Paramount Farming for in-kind contributions of equipment and personnel.