volume 42 number 3 july/august/september 2000csbe-scgab.ca/docs/journal/42/42_3_all_ocr.pdf ·...
TRANSCRIPT
cSAE
sCGR
The Journal of the CSAE (The Canadian society for engineering in agricultural, food, and biological systems)La Revue de la SCGR (La SOCiele canadienne de genie agroalimentaire et biologique)
CAE 42(2) 111-156 (2000)CN ISSN 0045-432X
Soil and IVater
EFFECTS OF WATERTABLE DEPTH, IRRIGATION WATER SALINITY, AND FERTILIZERAPPLICATION ON ROOT ZONE SALT BUILDUP
R.M. Palcl, S.D. Prasher and R.B. Bonnell III
Power GIUI J\
CANADIAN AGRICULTURAL ENGINEERING2000
JulylAugust/SeptemberVolume 42, Number 3
EDITOR
D.I.NORUMAgricultural and Bioresource Engineering
University of Saskatchewan57 Campus Drive
Saskatoon, Saskatchewan S7N 5A9
ASSOCIATE EDITORS
N.B. McLAUGHLIN(Power and Machinery)Agriculture and Agri-Food CanadaECORC - K.W. Neatby Bldg - CEFOttawa, Ontario KIA OC6
J.-A.LANDRY(Information & Computer Technologies)Agricultural and Biosystems EngineeringMacdonald College of McGill UniversitySte. Anne de Bellevue, Quebec H9X 3V9
S. CENKOWSKI(Bioprocessing Engineering)Department of Biosystems EngineeringUniversity of ManitobaWinnipeg, Manitoba R3T 5V6
C.P.MAULE(Waste Management)Agricultural and Bioresource EngineeringUniversity of Saskatchewan57 Campus DriveSaskatoon, Saskatchewan S7N 5A9
S.CHIENG(Soil and Water)Chemical and Biological EngineeringUniversity of British Columbia2357 Main MallVancouver, British Columbia V6T 1Z4
J.JOFRIET(Structures)School of EngineeringUniversity of GuelphGuelph, Ontario NIG 2W I
CSAElSCGR COUNCIL 2000-2001
REGIONAL DIRECTORS
T,J. NYVALL British ColumbiaBC Ministry of Agriculture, Food & Fisheries1767 Angus Campbell RoadAbbotsford, British Columbia V3G 2M3
R. BORG AlbertaAlberta Agriculture Food and Rural Development4920 - 51 st StreetRed Deer, Alberta T4N 6K8
Ontario
T. LUCYSHYN SaskatchewanPrairie Agricultural Machinery InstituteBox 961Humboldt, Saskatchewan SOK 2AO
J.K. WEEDEN613 Durham CrescentWoodstock, Ontario N4S 5X5
A. MADANI AtlanticNova Scotia Agricultural CollegeAgricultural Engineering DepartmentP.O. Box 550Truro, Nova Scotia B2N 5E3
Q. ZHANG ManitobaDepartment of Biosystems EngineeringUniversity of ManitobaWinnipeg, Manitoba R3T 5V6
S. GODBOUT QuebecCDPQ Quebec Hog Development Centre2795 Boulevard LaurierSainte-Foy, Quebec G IV 4M7
Secretary
Treasurer
P.JUTRASP.O. Box, C.P. 316Mansonville, Quebec JOE IXO
R.D. MacDONALDAGVIRO, Inc.367 Gordon StreetGuelph, Ontario NIG IX8
D.I. NORUM EditorAgricultural and Bioresource EngineeringUniversity of SaskatchewanSaskatoon, Saskatchewan S7N 5A9
D. SMALL Vice-President (Technical)DGH Engineering Ltd12 Aviation BoulevardS1. Andrews, Manitoba RIA 3N5
A. GHALY PresidentBiological Engineering DepartmentDaiTech, Dalhousie UniversityHalifax, Nova Scotia B3J 2X4
E. NYBORG Past PresidentDNL Technologies1410 Avenue KingsleyDorval, Quebec H9S IG I
R.L. KUSHWAHA President-ElectAgricultural and Bioresource EngineeringUniversity of SaskatchewanSaskatoon, Saskatchewan S7N 5A9
M. NGADI Vice-President (Regional)Agricultural and Biosystems EngineeringMacdonald College of McGill UniversitySte. Anne de Bellevue, Quebec H9X 3V9
Canadian Agricultural Engineering publishes papers covering the general fields of Agricultural, Food, and Biosystems Engineering that fit into one of the followingclassifications: (l) a scientific paper based on original research; (2) a technical paper based on design, development, testing or analysis of machines, equipment,structures, processes, or practice; (3) a general paper on education relative to curricula and philosophy or trends in science, on a surveyor investigation of some phaseof research or research methods, or on extension or extension methods.
Manuscripts for publication should be submitted to the Editor. The papers must be original and must have not been published elsewhere in a refereed publicationor copyrighted. Authors are required to suggest three or more potential referees. Before publication, all authors must complete an Assignment of Copyright toCSAElSCGR. The author(s), not CSAElSCGR, is (are) responsible for opinions expressed. Information published in Canadian Agricultural Engineering may be quotedin whole or part provided that credit is given to the author(s) and to the journal. Publication charges are $75/page plus cost of translation, if required. Reprint chargesare $16/page for 100 copies.
All claims for missing issues must be made to the address below.
Central Office Address: Box 381, RPO University, Saskatoon, SK S7N 4J8Published Quarterly
Canadian Publications Mail Product Sales Agreement No. 0466247Return Postage Guaranteed
Subscription rate: Canada $50.00 per annumOutside Canada US$50.00 per annum
Effects of watertable depth, irrigation watersalinity, and fertilizer application
on root zone salt buildupR.M. PATEL, S.O. PRASHER and R.B. BONNELL
Agricultural and Biosystems Engineering. Macdonald Campus ofMcGill University, 21111 Lakeshore Road, Ste-Anne de Bellevue,QC, Canada H9X 3V9. Received 31 August 1999; accepted 19 June 2000.
Patel, R.M., Prasher, S.D. and Bonnell, R.B. 2000. Effects ofwatertable depth, irrigation water salinity, and fertilizerapplication on root zone salt buildup. Can. Agric. Eng. 42: 111-115.Salt buildup due to irrigation water salinity and fertilizer applicationwas studied in field Iysimeters planted with green peppers (Capsicumannuum).Water was applied by subirrigation, and the fertilizers wereincorporated at the soil surface. Three subirrigation water salinities, I,5, and 9 dS/m and two watertable depths, 0.4 and 0.8 m, were used.The soil salinity was determined by first measuring the bulk soilsalinity by time domain reflectometry (TDR) and then converting it tosoil solution salinity (ECsw)' It was found that the salinity of thesubirrigation water affected ECsw in the upper soil profile when thewatertable was maintained at 0.4 m depth. The subirrigation water alsoaffected the lower half of the soil profile when the watertable wasmaintained at 0.8 m depth; however, it did not affect any salt buildupin the upper half. Also, the addition of N, P, and K fertilizers did notcontribute to the salt buildup in the soil. Although watertable depth andsubirrigation water salinity affected ECsw' they did not affect the greenpepper yield. The experiment was conducted using field Iysimetersfilled with a sandy soil and covered with a plastic sheet to simulatearid conditions. Therefore, caution should be exercised in extrapolatingthe results of this study to field conditions and other soils.
L'accumulation de sel dans Ie sol due al'irrigation avec de l'eausaline et aI'utilisation de fertilisants, a ete etudie au champs dans desIysimetres OU ont ete semes des plants de piment vert (Capsicumannuum). Un systeme d'irrigation souterraine a ete utilize, et lesfertilisants ont ete incorpores ala surface du sol. Nous avons teste troisconcentrations de sel (I, 5 et 9 ds/m) adeux profondeurs de la nappephreatique (0.4 et 0.8 m). La concentration de sel dans Ie sol a etedeterminee en mesurant sa salinite globale par reflectometrietemporelle (TDR). Cette valeur a ete ensuite convertie en sel solublepresent dans Ie sol: ECsw (salinite de la solution du sol). Les resultatsobtenus, ont montre que quand la nappe phreatique est maintenue a0.4m, la salinite de l'eau d'irrigation affecte la ECsw, uniquement a lasurface du sol. Cependant, quand la nappe phreatique est maintenue a0.8 m, toute la moitie inferieure du profil du sol est affectee alors quela moitie superieure est epargnee. L'addition de fertilisants (N, Pet K)n'a pas eu d'effet sur la salinite du sol etudie. Meme si la profondeurde la nappe phreatique et la salinite de I'eau d'irrigation ont affecte lasalinite du sol, ils n'ont eu aucun effet sur Ie rendement des plantes depiment vert. L'experience a ete conduite en utilisant des Iysimetresremplis de sol sableux et couverts de plastique pour simuler desconditions de climat aride.Il faudrait donc, etre prudent quant al'extrapolation de ces resultats ad'autres conditions et ad'autres typesde sols.
INTRODUCTION
Fresh water resources are limited in many regions of the worldthat require irrigation for crop production. Although brackish
waters may be used to supply crop water needs (Rhoades et a1.1992), salt-sensitive crops do not perform well when salinity isabove a certain limit (Maas and Hoffman 1977; Grattan et a1.1987). For green peppers, the threshold salinity (salinity ofsaturated soil paste extract, ECe) has been reported to be 1.5dS/m (Maas and Hoffman 1977). Under field conditions, theelectrical conductivity of soil solutions (ECsw) is 2 to 4 times theECe (Smedema and Rycroft 1983) and so the threshold salinityin terms of soil water can be as high as 3 to 6 dS/m. In general,surface methods of irrigation can lead to a rapid increase in saltcontent in the upper soil profile if evaporation rates are high andthe water contains dissolved salts. This is extremely deleteriousto salt-sensitive crops, particularly in early stages of growth.Although surface irrigated brackish water can be alternated withfresh water after the plants reach a less sensitive growth stage(Rhoades 1984), extensive salt buildup in the upper profile mustoften be dealt with before the next crop is sown. In regions withhigh evaporation and limited fresh water, even the applicationof nitrogen fertilizer can increase soil solution salinity and resultin salt buildup (Lunin and Gallatin 1965a; Jurinak and Wagenet1981).
There is little quantitative information about salt buildupwith subirrigation using brackish water or from the surfaceapplication of fertilizers under subirrigation. Von HoyningenHuene (1994) suggested that subirrigation can supply brackishwater for crop growth without injury to plant shoots and,possibly, with little or no soil salt buildup problems. Theobjectives of this study were to: determine if application offertilizer and subirrigation with brackish water would increasesoil salinity; and evaluate the effect of fertilizer andsubirrigation treatments on the yield of green pepper (Capsicumannum). Because green peppers are one of the world's mostpopular vegetables, their growth responses to increasingsalinity, water level, and fertilizer applications were alsostudied.
MATERIALS and METHODS
Field lysimeters
Seventy-two field lysimeters located at Macdonald Campus ofMcGill University, Quebec, were used in this experiment. Thelysimeters were constructed from 12.5 mm thick, 1.0 m longpolyvinyl chloride (PVC) pipe with an inside diameter of0.45 m. A 10 mm thick, 0.6 m x 0.6 m PVC sheet was welded
CANADIAN AGRICULTURAL ENGINEERING Vol. 42, No.3 JulylAugust/September 2000 III
The soil solution salinity (ECsw) was calculated from bulksoil salinity by using (Rhoades et al. 1976):
mgIL, calcium 375 mgIL, magnesium 307 mgIL, chlorides 6548mgIL, and carbonates 457 mgIL (the major ions were Na+ andCn. The saline water was diluted using tap water to obtain thedesired salinity levels for subirrigation. The fertilizerapplications were: KP, NIK, NIPK and N2PK, where K=230kg Klha at planting, P= 200 kg Plha at planting, N1= 35 kg Nlhaat planting and 35 kg Nlha at flowering, N2=70 kg Nlha atplanting and 70 kg Nlha at flowering stage. The fertilizers wereapplied as solids in the form of ammonium nitrate (N), triplesuper phosphate (P), and muriate of potash (K). The fertilizerper plant (one plantllysimeter) was calculated on the basis of24,691 plantslha (0.9 m x 0.45 m plant spacing). Each treatmentwas randomly assigned to the 72 Iysimeters, giving threereplicates per treatment.
Green pepper (Capsicum annum cv. BellBoy) seedlingswere planted in plastic pots (0.1 m diameter and 0.1 m deep) onJune 18. Water was sprinkled every third day until one plantwas transferred to each Iysimeter on July 15, at which time oneliter of water was applied to each Iysimeter. Thereafter, 1.5liters of water, divided in five equal volumes, was applied everythird day to each plant. The basal fertilizer was applied in theIysimeters on July 14. Subirrigation with brackish water beganon August 3. The fertilizer for the flowering stage was appliedon August 12.
There were 74.8, 94.6, 57.2, and 119.2 mm of rainfall inJune, July, August, and September, respectively. However, eachIysimeter was covered with plastic to keep out rain and simulatearid conditions, and no surface irrigation water was applied inthe Iysimeters. The green pepper plant in each lysimeter grewthrough a small cut in the plastic cover. Some evaporation fromthe soil surface could have occurred through this slit and aroundthe edge of the cover, because its circumference was fixed to thelysimeter only at five places using small pieces of duct tape.The monthly means of daily average temperatures were 17.2,21.0, 20.1, and 13.9°C in June, July, August, and September,respectively.
The bulk soil salinity (ECl!) and water content (8) weremeasured at weekly intervals at different depths with the timedomain reflectometry (TOR) method, using a Tektronix 1502Bcable tester. The salinity probes were inserted horizontally at 0.1and 0.3 m depths from the soil surface in those lysimeterswithin which the watertable (WT) was held at a 0.4 m-depth. Inthe Iysimeters wherein the watertables were maintained at a 0.8m-depth, the salinity probes were at 0.1, 0.3, 0.5, and 0.7 mbelow the soil surfaces. For converting ECl! to ECsw, the TORwas calibrated. First, the surface conductance (ECs) of the soilwas determined in the laboratory, and it was found to benegligible. Next, ceramic cups were inserted at the same depthsat which the TOR probes were inserted (0.1, 0.3, 0.5, and 0.7m). The lysimeters were then flushed with water having salinitylevels of 1, 3, 5, 7, and 9 dS/m. Soil water was extracted byapplying vacuum to the ceramic cups, and ECsw was determinedfor each depth. At the same time, ECl! and 8 were determinedwith TOR. The transmission coefficient (or) values weredetermined by dividing ECl! by ECsw * 8 from which therelationship between or and 8 was derived:
Mariotte bottlewith brackish water
All dimensions arc in mm
Fig. 1. Schematic of Iysimeter and subirrigation system.
over one end of the pipe. A 40 mm diameter perforated PVCpipe was welded inside the lysimeter across the bottom tosupply subirrigation water. A fabric filter was wrapped aroundthe supply pipe to prevent soil particles from blocking thesystem. The supply pipe was connected to a riser pipe of equaldiameter on the outside of the lysimeter, through which waterwas supplied. A Mariotte bottle was used to maintain thedesired watertable depth. The lysimeters were kept aboveground to facilitate measurement of salinity (Fig. 1).
The lysimeters were filled with sand-textured soil of the St.Amable complex found on Macdonald Campus of McGillUniversity. The St. Amable complex has deep sandy depositsoverlying a clay layer (Lajoie 1960). The upper 0.6 m sand layerwas used and consisted of91.2% sand, 4.2% silt, 1.1 % clay, and3.5% organic matter and had a bulk density of 1.4 Mglm3
•
The soil was uniformly mixed before filling the Iysimeters.It was weighed and packed into the lysimeters in layers thatwere 0.1 m thick. Bulk density was maintained at 1.4 Mglm3 byusing the same mass of soil for each 0.1 m soil layer. The uppersurface ofeach layer was scratched before packing another layerto ensure good contact between layers.
Experimental procedure and salinity measurement
The 72 lysimeters were laid out on the north-south axis in 6rows of 12 lysimeters each. There were 24 treatment~,
comprised of two watertable depths (0.4 and 0.8 m from the SOlI
surface), three irrigation water salinities (ECiw of 1, 5, and 9dS/m) and four types of fertilizer applications. Naturally salinewell water was used for subirrigation. The chemistry of thewater was: total dissolved solids 12,713 mgIL, sodium 4100
"t = 0.0575 +0.95198 (1)
112 PATEL, PRASHER and BONNELL
2..,__--------.......1.......------,
....._ ..._d..~.f2=~~-8020 40 60
Days after planting
0+------,,.----_----,.----4o8020 40 60
Days after planting
O+---.....,~--__r---"""'r"-----I
o
Subirrigation water salinity
IdSlm SdSlm 9dS1mSoil solution~ty - ~ -Standard aevuwon I [) I
Each data point is mean of twelve values. Each bar represents one standard deviation.
Fig. 2. Soil solution salinity at different depths in the Iysimeters under 0.4 m watertable.
WT: watertable depth; IR: salinity levels of subirrigation water;D: salinity measurement depths; T: time of salinity measurement** significant at P<O.O I
The ECsw at 0.1 and 0.3 m depths (D), which were commonto both watertable treatments, were analyzed with repeatedmeasures using SAS Version 6.12 for Windows (SAS Institute1996) to determine salt buildup over time in the soil profile.
Table I. Repeated measures ANOVA of watertable depth,subirrigation, water salinity, measurement depth,time, and interaction effects on soil solutionsalinity.
Whenever the green peppers were ready to harvest, the fruitswere handpicked and weighed. At the end of the season, thetotal mass of green pepper harvested per plant was calculated.Also, at the end of the season, the plants were cut and theirfresh masses were recorded. The total of green pepper massharvested per plant and the shoot masses were analyzed withthe General Linear Model (GLM) using SAS.
RESULTS and DISCUSSION
Analysis of ECsw dataThe electrical conductivity of the soil water (ECsw)' derivedfrom time domain reflectometry measurements (ECa), wasdetermined eight times (T), after the initiation of subirrigation.In the first step of a repeated measures ANOVA analysis ofthese data, it was found that the fertilizers did not influence theECsw' This may have been due to the low doses of applied N,most of which was likely taken up by the plants. Since fertilizerand its interactions with the other factors (WT, ECiw, D, T) werenot significant, an ANOVA analysis was run that did notconsider fertilizer in the statistical model (Table I). It is evidentfrom the Table I that ECsw values in space and time aredependent on watertable depth and subirrigation water salinity.
The ECsw data from the lysimeters with the watertable heldat 0.4 m are plotted in Fig. 2a-b, while those from lysimeterswith a 0.8 m watertable are shown in Fig. 3a-d. The effect ofirrigation water salinity at a given depth is understood to beindicated by the separation of the ECsw curves (Figs. 2a-b and3c-d). When the curves are concurrent, there is no measurabledifference due to the irrigation water salinity. Separation of thecurves tends to increase with time and depth. This indicates thatthe saline front is mixing with the initial fresh water, first atlower depths and, subsequently, at higher depths in the profile.Thus, at the initial stage of mixing of the saline irrigation waterwith the original soil water, the increase in ECsw caused by the9 dS/m irrigation water is not distinguishable from that causedby the waters of lower salinity.
With time, the original fresh water continues to be removedby evapotranspiration from the top of the Iysimeter, the dilutioneffect diminishes, and the soil solution salinity (ECsw) tends to
(3)
(2)
0.05758+0.9519(/
= soil solution salinity (dS/m),= bulk soil salinity (dS/m),= surface conductance (dS/m),= volumetric moisture content, and= transmission coefficient.
ECsw
Source Source
WT ** T **IR ** TxWT **WTxIR ** TxIR **D ** TxWTxIR **DxWT ** DxT **DxIR ** DxTxWT **DxWTxIR ** DxTxlR **
DxTxWTxIR **
ECSI\'
By substituting Eq. 1 for t, Eq' 2 was rewritten as:
where:ECsw
ECa
ECs
et
CANADIAN AGRICULTURAL ENGINEERING Vol. 42. No.3 JulylAugust/September 2000 113
2 2
~Oe
Mean '!S 0 Mean~ ~'a
~616
g 4
~a 4
'::1 .§=''§ 2 '§ 2
~o =a1'1) 0
0 20 40 60 80 0Days after planting
2 2e .@~ 0 rao~ ~'a 'a1 6 :a 6
fn
.§ 4
.~c:.8 4
.a .ai 2 =2::: ..-- .~o
'S1'1)0
0 20 40 60 80 0Days after planting
b. At 0.3 mde th
-~.........--........~• •
20 40 60 80Days after planting
20 40 60 80Days after planting
Subinigation water salinity1 dS/m 5 dS/m 9 dS/m
Soil solution salinity - -.- -Standard aevfliffon. 0 I
Bach data point is mean of twelve values. Bach bar represents one standard deviation.
Fig. 3. Soil solution salinity at different depths in the Iysimeters under 0.8 m watertable.
increase. In such a process, one would expect the separation ofcurves to occur fIrst at the measurement depths closest to thewatertable depth. This is corroborated by the series of figurespresented (Figs. 2a-b, 3a-d).
It is interesting to note the time at which the salinityincreased at different depths in the profile; this can have a directimpact on plant growth. Figures 2a and 3a show that the upperroot system was not affected by the salinity of the irrigationwater until about 19 days or longer after the start ofsubirrigation. In practice, this would give salt-sensitive crops anexcellent head start, particularly if they are transplanted ratherthan direct-seeded. One might also expect that crops withshallow root systems will perform better when subirrigated withbrackish water than when surface irrigated with water of thesame salinity.
There are several arid and semiarid regions in the worldwhere there are shortages of fresh water and where groundwater is shallow and saline. In such areas, fresh water may beavailable in very limited quantities or be obtainable only duringa certain time of the year, e.g., during the monsoons. Not onlycan this fresh water be used to flush out accumulated salts in theroot zone, but it can effectively be stored as soil moisture forcrops in dry periods; as the stored fresh water becomes depleted,subirrigation with brackish water can bring the crops tomaturity.
The experiment was conducted in sandy soil, and the effectof subirrigation with brackish water may differ in other soils,
114
given different physio-chemical properties. Arid conditionswere simulated by covering the soil surface with plastic. Theplastic cover was not an airtight lid and the atmospheric aircirculation over the lysimeter soil surface took place. Therefore,some evaporation from the soil surface would have occurred.Under field conditions, values of the ambient variables can varyconsiderably, but we believe that the general trend of saltbuildup observed in the experiment would be similar to whatmight be expected to occur under field conditions. Morever,standard deviation also increased with time, as shown in Figs.2 and 3, which may also indicate some intrusion of saline waterinto the crop root zone with subirrigation.
Green pepper yield
There was a significant difference in the yield of green peppersbetween the PK fertilizer treatment, and the N IPK and N2PKtreatments; but, the yields for the N1K, N IPK, and N2PKtreatments were not significantly different (Table II). The yieldwas highest for the Iysimeters supplied with all three nutrients(Table II). Lunin and Gallatin (1965b) found that yielddecreased due to an increase in salinity from the fertilizer, butGerg et al. (1993) reported that such was not always true. Datafrom our experiment suggest that salinity from the fertilizer didnot reduce the yield of green peppers. It is to note that thresholdECsw for green pepper is about 4.5 ds/m, a figure obtained bymultiplying the threshold ECe of 1.5 dS/m (Maas and Hoffman1977) by 3, as suggested by Smedema and Rycroft (1983). It isevident that ECsw in the upper portion of the root zone is below
PATEL, PRASHER and BONNELL
Table II. Wet masses for green peppers and plant shoots,mean and standard deviation.
combined with crop growth and climatic models, could be usedto plan crop production in arid and semi-arid regions wherefresh water supplies are limited.
CONCLUSIONS
this limit (Figs. 2a-b, 3a-b). This might explain why there wasno detectable effect on yield due to salinity of irrigation water,while the yield was fertilizer dependent (Table II).
REFERENCES
Gerg, B.K., S.P. Vyas, A.N. Lahiri, P.C. Mali and Y.C.Sharma. 1993. Salinity-fertility interaction on growth,mineral composition and nitrogen metabolism of Indianmustard. Journal ofPlant Nutrition 16(9): 1637-1650.
Grattan, S.R., C. Shennan, D.M. May, J.P. Mitchell and R.G.Burau. 1987. Use of drainage water for irrigation of melonsand tomatoes. California Agriculture 41(9/10): 27-28.
Jurinak, JJ. and R.I. Wagenet. 1981. Fertilization and salinity.In Salinity in Irrigation and Water Resources, ed. D. Yaron,103-119. New York, NY: Marcel and Dekker Inc.
Lajoie, P.G. 1960. Soil Survey of Argenteuil, two mountainsand Terrebonne Counties, Quebec. Report. ResearchBranch, Canada Department of Agriculture in Co-operationwith Quebec Department of Agriculture, and MacdonaldCampus, McGill University.
Lunin, J. and M.H. Gallatin. 1965a. Salinity fertility interactionin relation to the growth and composition of beans. I. Effectof N, P and K. Agronomy Journal 57(4): 339-341.
Lunin, J. and M.H. Gallatin. 1965b. Salinity fertility interactionin relation to the growth and composition of beans. II.Varying levels of N, and P. Agronomy Journal 57(4): 342345.
Maas, E.V. and GJ. Hoffman. 1977. Crop salt tolerance.Journal ofIrrigation and Drainage Division 103(IR2): 115134.
Rhoades, J.D. 1984. Use of saline water for irrigation.California Agriculture 38( 10): 42-43.
Rhoades, J.D., A. Kandiah and A.M. Mashali. 1992. The use ofsaline waters for crop production. Irrigation and DrainagePapers 48. Rome, Italy: Food and Agriculture Organizationof the United Nations.
Rhoades, J.D., P.A.C. Raats and R.I. Prather. 1976. Effect ofliquid-phase electrical conductivity, water content, andsurface conductivity on bulk soil electrical conductivity.Soil Science Society ofAmerica Journal 40(4): 651-655.
SAS Institute. 1996. Users guide. Cary, NC: SAS Institute Inc.
Smedema L.K. and Rycroft D.W. 1983. Land Drainage:Planning and Design of Agricultural Drainage System.London, UK: Batsford.
Von Hoyningen Huene, B. 1994. Subirrigation of Maize UsingSaline-sodic Water. Unpublished Ph.D. thesis. Montreal,QC: Department of Agricultural and BiosystemsEngineering, McGill University.
ACKNOWLEDGMENTS
The authors thank NSERC for providing financial support forthis project.
Peppers Shoots(g1plant) (g/plant)
494±35a 468±23a450±30a 452±19a
515±4la 475±30a429±42a 462±28a473±43a 443±25a
367±32b 380±23b441±43a,b 416±29b570±41a 504±25a510±63a 539±34a
472 460
K: 230 kg Klha; P: 200 kg Plha; N1: 35 kg Nlha when plantingand 35 kg Nlha when flowering; N2: 70 kg Nlha when plantingand 70 kg Nlha when flowering .Mean masses with same letters under watertable depth, salinitylevel of subirrigation water, and fertilizers are not significantlydifferent (P<0.05).
Average mass
Salinity level of subirrigationwater (dS/m)I59
Watertable depth (m)0.40.8
FertilizersPKNIKNIPKN2PK
Variable
This study indicated that the salt buildup in the soil profileoccurred from the bottom upwards when brackish water wasused to subirrigate green peppers plants that were initially raisedin pots with non-saline water. The resulting dynamics of saltbuildup in a sandy soil permitted early plant developmentwithout hindrance to the activity of the upper root system. Theresults indicated that, under the conditions of this experiment,water with a salinity as high as 9 dS/m had been used forsubirrigation for up to three weeks before the upper root zone(0-0.3 m) was influenced, even when the watertable depth wasas shallow as 0.4 m. However, the higher the salinity of theirrigation water, the higher would be the residual salinity at theend of the growing season. Soil flushing with suitable waterwould be needed in such cases to prevent salt accumulation. Iffresh water is available during other periods in the year, onecould use it for flushing the salts. It could also be noted that ifthe soil salinity is very high at the end of the cropping season,one could use slightly brackish water (as an alternative to freshwater) to flush out the excessive salts present in the soil.
The significance of this work is that supplementalsubirrigation of sandy soils in arid regions may be possibleusing brackish waters to produce green peppers and perhapsother salt-sensitive crops. Field experiments in arid conditionscould establish the wider applicability of this method. Further,work can be oriented towards developing a model of saltbuildup in subirrigated soil which, when appropriately
CANADIAN AGRICULTURAL ENGINEERING Vol. 42. No.3 JulylAugust/September 2000 115
Reduced drift from air-assisted sprayingM. PICHE', B. PANNETON1 and R. THERIAULT2
IHorticultural Research and Development Centre, Agriculture and Agri-Food Canada, Saint-Jean-sur-Richelieu, QC, Canada J3B3£6; and 2Faculty ofAgriculture and Food Sciences, Universite Laval, Sainte-Foy, QC, Canada G1K 7P4. Received5 October 1999;accepted 21 June 2000.
Piche, M., Panneton, B. and Theriault, R. Reduced drift from airassisted spraying. 2000. Can. Agric. Eng. 42: I 17-122. Field trialswere conducted in 1994 and 1995 to compare spray drift produced bytwo spraying techniques. A hydraulic boom sprayer providing bothair-assisted and conventional spraying was used for two reducedvolume applications. For each treatment, 14 trials were conductedover short grass under wind conditions varying from I to 5 mis,measured 2 m above the ground. Drift was measured on a 8 m tower,10 m away from the boom sprayer. Results showed that air-assistedspraying combined with use of coarser spray provided significantlyless drift under wind conditions of 5 mls compared to conventionalspraying under a wind of 1 mls.
Des essais au champ ont ete effectues en 1994 et 1995 pourcomparer la derive produite par deux systemes de pulverisation. Vnerampe de pulverisation produisant ala fois une pulverisation assisteepar de I'air et une pulverisation conventionnelle a ete utilisee pourdeux applications avolume reduil. Pour chaque traitement, 14 essaisont ete effectues sur une vegetation rase sous des vitesses de ventvariant entre I et 5 mis, a2 m au-dessus du sol. La derive a etemesuree sur une tour de 8 m de haut, Ie pulverisateur appliquant lestraitements a 10m de celte tour. Les resultats obtenus demontrentque la pulverisation assistee par de I'air combinee al'utilisation d'unepulverisation plus grossiere produit significativement moins dederive sous des vitesses de vent de 5 mls que I' applicationconventionnelle sous des vents de I mls.
INTRODUCTION
Drift generated by the application of pesticides has been thesubject of many studies over the years. Drift creates sideeffects to society in general by increasing the deposition ofpesticides on undesirable targets and increasing waterpollution. Bystander exposure is also ofconcern because it canresult in acute toxic effects.
The balance between the requirements for drift reductionand the conditions needed to achieve the optimumperformance of pesticide is not obvious. With a conventionalhydraulic boom sprayer, finer sprays can improve coverage butcoarser sprays will decrease drift (Quanquin 1992). However,the trend towards the use of reduced spray volumes requiresthe use of smaller droplets in order to maintain a sufficientlyhigh number of droplets to cover the target (Western andHislop 1991). Reducing drift and maintaining adequate targetcoverage are often conflicting requirements.
Watson et al. (1984) adapted air-assisted spraying, asystem already used in tree crops, for row crop chemicalapplication. The technique showed improved spraypenetration, reduced drift, and more uniform coverage andallowed lower carrier volume.
The difference in drift production between conventionalhydraulic boom sprayers and air-assisted hydraulic boom
sprayers has been tested many times in chambers and in thefield (Nordbo and Taylor 1991; Quanquin 1992; Reed et aI.1993; Taylor et al. 1989; Wolff and Smith 1989). Operationalparameters of the sprayers were varied from using coarser tofiner spray nozzles, lower to higher air speeds, and using lowvolume and high-volume applications. Drift was measured atdifferent heights and distances from the application point andwind speeds varied from I to 8.5 mls. In general, the resultsshowed that air-assisted systems reduced drift from 50 to 90%compared to conventional application. The natural wind speedand the generated air speed both influenced the amount of drift.Higher wind speed provided the highest amount of drift forconventional and air-assisted systems. Air-assistance usinghigher air speeds increased drift control.
Western and Hislop (1991) tested electrostatic sprayers fortheir capacity to reduce drift using fine spray droplets in a windtunnel. They compared a conventional sprayer, an air-assistedsprayer, and an electrostatic sprayer, with and without airassistance. All tests showed that spray drift increased with windspeed and decreased with air-assistance. Drift fromconventional application was reduced by 71 % using minimumair-assistance and by 88.3% using maximum air-assistance, at2 mls wind speed. Electrostatically charged droplets did notreduce drift in wind tunnel experiments. This conclusion is inagreement with the results obtained by Miller (1989) but theresults of Sharp (1984) showed that using electrostatic and airassisted spraying together increased drift. This increased driftwas most likely due to the mutual repulsion of small, highlycharged droplets driving spray upwards (Western and Hislop1991). In all cases, electrostatic charging of droplets appearedto decrease the drift control effectiveness of air-assistance.
May (1991) compared a conventional hydraulic boomapplication to the Hardi Twin™ (Hardi International, Taastrop,Denmark) system. This system forces air downwards through aslot extending parallel to the boom and covering its wholelength. The orientation of the air duct can be adjusted such thatthe air velocity vector at the exit is varied from -300 to +300
from the vertical. It was possible to conclude that keeping theair duct vertical or angling it forward reduced drift compared toa conventional application. Angling the air 300 backwardprovided increased drift. Taylor et al. (1989) obtained similarresults.
This previous work suggests that the droplet size and airspeed and orientation might be of great importance when settingup a machine for spraying under different conditions. Theoperational parameters of the machine are not well determinedas for the exact air speed and orientation and the nozzles to beused. The possible interactions between drop size and air
CANADIAN AGRICULTURAL ENGINEERING Vol. 42, No.3 July!August/September 2000 117
EXPERIMENTAL DESIGN
MATERIALS and METHODS
time as feasible. Typically, a sel of two trials required less Ihan30 minutes. Environmental condilions are relatively steady oversuch a period (Panofsky el aJ. 1984).
Instrumentation
A lower was set up to support the micro-meleorologicnlinstruments. Anemometers (Young Model 0570 I-I 0, R.M.Young, Traverse City, MI) were placed at 2, 4, and 8 m aboveground. These provided wind speed and orientation al the threeheights. A sonic anemometer-thermomeler (Campbell Scienti fieCA27, Campbell Scientific Inc, Edmonton, AS) was placed al3.7 m above ground to measure Ihe turbulcnce intensily on thevenical axis as well as vertical momentum and heat Ouxes. ACampbell Scienlific Micrologger Model 21 X was fixed on thetower to monilor sensor OlitPUIS 10 times per second for Ihesonic anemometer-thermometer and once pCI' second for theanemometers. Fifteen minutes of recording was necessarybefore and after the period of treatment application to oblainuseful wind statistics.
A second tower was erected 10 support the drift samplers.The sampling unils used were rotating rod samplers (Fig. I).The stainless steel sampling rod is 130 mmlong and 1.6 mm indiameter. The rod is mounted perpendicular to the axis of a DCmotor. The tower supporting the drift samplers was designed toallow alignment such that Ihe plane defined by the rOlation orIhe sampling rod is perpendicular 10 (he wind vector. Thisalignmelll was achieved wilhin ±30 degrees. The samplers wereoperaled at an approximate constant speed of 2500 RPM,mcasured and recorded before each experiment. These samplersare described more Fully in Panneton and Theriault (1995).There were 8 sampling units located at 0.5, I, 1.5,2,3,4,6, and8 III above ground. Taking measuremenlS over 8 m in heightallowed recovery of the overall drift cloud 10 show a compleledri fl protile.
Experimental protocol
The application of Ihe 2 treatments was performed with the airassisled hydraulic boom sprayer developed ;:ll the Horticulturaland Development Center of Agricullural and Agri-Food Canada.Saint-Jean sur Richelieu, QC. The air-assistallce syslem designis a scaled down version of the Hardi Twin™ sprayer, the boomwidth being 3 m. The sprayer was adjusted 10 obtain an airvelocity of 31 mis, air-now rate of I m3/s per meier of boomand an angle of application 15~ forward, in line with therecommendalion in Pannelon et aJ. (2000a and 2000b). Theangle could nol be sci to Ihe recommended 20° because ofmechanical conslrainls.
All applications were performed al a volume rate of 100Llha and al a fixed ground speed oF7.8 km/h. This volume ratewas determined after some preliminary tests. Working wilhhigher volume rates is likely to resuh in drift sampler salUrationand should be avoided. Pannelon el al. (2000a, 2000b) workedat a volume rate of 250 L/ha. Therefore, it was not possible 10
work with the same nozzle. A substitute nozzle was selectedkeeping Ihe volume mcdian diameter (YMD) nearly conslalll.The selecled nozzles are Delavan (Delavan Inc, Lexington, TN)80-2R operated at 195 kPa. For water, the resulting VMD is 185flm compared 10 175 flm for Ihe 80-3R m 385 kPa of Pannelonel al. (2000a, 2000b). These VMD were oblaincd from Ihe
(21 27.5 mm
,201m
---'-f
143 mm
L
r-Sample43 mm
L~-----.-
130 rnm
12mm J
2mrn Lt
1.6 mm
The experiment was conducted over short grass in a fieldlocated on the experimental farm of the Horticultural Researchand Development Center (HRDC) of Agriculture and AgriFood Canada, Saint-Jean sur Richelicu. QC. Experiments wererepeated over IWO years, 1995 and 1996. The experimentationconsisted of quanlifying drift using a nuoresccnt tracer withtwo different application techniques: air-assisted spraying andconventional spraying. The experiment was repeated as oftenas possible under different wind speeds varying from I to5 m/s. To simplify the experimenlal procedure and the dataanalysis. experiments with the air-assisted and theconventional sprayers were paired. This was achieved byperforming trials Wilh both sprayers within as short a period of
OBJECTIVES
Fig. 1. Drift sampler design.
The objectives or this slUdy are to compare the amount of driftproduced using two sprayers, (I) a hydraulic air-assisted boomsprayer adjusted using paramelers recommended in Pannetonel aJ. (2000a, 2000b); and (2) a conventional hydraulic bool11sprayer configured for fungicide or insecticide applications onvegetable crops. The comparison is limited 1O the near-driftmeasured lOin away from Ihe spray swath.
volume and their effect on drift are not known. Followingspray chamber tests on broccoli, Pannelon ct al. (2000a)recollllllended an air speed greater than 25 mis, an air flow rategreater than 0.9 1113/s per meter of boom and an air jetorientation at aboul 20° towards the direction of travel with theuse of 8003 (Spraying Syslems Co.. Whemon, Il.) f1m fannozzle operaled m 385 kPa (250 L~la). This recommendalionalso applies for pOlaloes (Pannelon el al. 2000b). Oncommercial air-assisted sprayers, the maximum air flow raleis abollt 0.3-0.4 1113/s per meIer of boom. The airflow raterecommended by Pannetol1 et al. (2000a) and Panncton et al.(2000b) is much larger. Also, Theriaull el al. (1996) haveconcluded thal for both potato and broccoli the insecticide andfungicide efficacy was as good using air-assistance at theserecommended settings as when using a conventional boomsprayer equipped with cone nozzles. It was Iherefore decided1O perform some field experimellls to verify that drift controlis slill good using air assistance at higher airflow rates.
liS PICI·IE. PANNETON and THERIAULT
Data analysis
To calculate the amount of drift, themass of tracer recovered from thesampling rods must be converted to atime-integrated flux. Time integration isperformed by the sampler since it isoperated over a long enough period toensure that the drifting cloud passing bythe sampling tower is sampled from itsleading edge to its trailing edge. Themass of tracer is converted to ameasured time-integrated flux using:
Fig. 2. Field layout. Fm(h) = MC,scos(ew )
(I)
Total drift is the volume of spray that drifted per meter offorward travel of the sprayers. It was computed using:
where:D =drift (LIm) and!J..Zj =vertical distance associated with the itb sampler (m).
Finally, percentage drift can be evaluated using:
where:Fm(h) =measured time integrated flux at height h (Um2
),
M =mass of tracer extracted on sampling rods (kg),Ct = tracer concentration in the tank mix (kgIL),S =sampling surface (m2
), and6w =angle between the wind vector and the axis of
rotation of the sampler.
The measured flux differs from the actual flux because thedrift samplers do not sweep the whole volume of air that crossesits sampling zone. To account for this effect, a sampling ratio,SR, must be computed using:
RD (5)
(2)
(3)
(4)
MU(h)
2C,dSro
2dro
U(h)cos(ew )
8
D="2.F;{h;)&;;=1
F(h) = Fm(h), SR
SR
where:RD = relative drift,V = tractor speed (m/s),nb =number of nozzles,Qb =nozzle flow rate (Us), andnp =number of tractor passes.
where:d =diameter of the sampling rod (0.0016 m),w =rotational speed of the sampler (rps), andU(h) = windspeed at the sampler height, h.
The actual time-integrated flux is then:Laboratory analysis
Fluorometric analysis of the dye collected on the driftsampling rods was performed on a Perkin-Elmer LS-5 (PerkinElmer, Oak Brook, IL). The wavelengths of excitation andemission were adjusted to 365 and 445 nm, respectively. Bothexcitation and emission slots were adjusted to 10 mm. A graphwas drawn providing a calibration curve and was rechecked afew times over the summer and fall periods.
A visual evaluation of the concentration of tracer on therods was performed under a UV light. Preliminary dosageallowed determination of the sufficient amount ofmethanol forextraction to place the values within the valid region of thecalibration curve.
The sampling rods were prepared for analysis by firstcutting and discarding 2 mm from each end of the rod (Fig. I).A sample was made of two 43 mm long pieces collected ateach end of the remaining rod. These pieces were cut directlyinto a glass tube of 10 mL or 5 mL depending on the volumeof methanol to be added. Methanol was added to the tubes anda 30 minute period was necessary to ensure a completeextraction of Blankophor. After extraction, the samples wereanalyzed in the fluorometer. The measured intensities werecompared to the calibration curve and the tracer concentrationwas determined.
consitent data set presented in Womac et al. (1997). Thenozzles used for the conventional treatment were Delavan D223 at 785 kPa for a VMD of 165 J.lm (data for tap water fromDelavan Inc.). Disc-core nozzles are recommended for theapplication of wettable powders, contact fungicides, andinsecticides (Spraying Systems 1995). The water-solublefluorescent tracer was Blankophor™ BA liquid at 80% (BayerCanada Inc, Pointe Claire, QC), at a concentration of 5% VNin tap water.
For the application of the treatments, the mixture wassprayed over about 30 m (spraying was allowed for about15 s), making 6 consecutive passes, 10 m away from thesampling tower (Fig. 2). The orientation of the tractor coursewith respect to the sampling tower was determined in order toalways have the wind direction ±30° from the perpendicular ofthe sprayer course. After the first treatment was applied, thesampling units were detached and locked in a light proof boxto be brought back to the laboratory and another set ofsamplers was installed for the application of the secondtreatment.
CANADIAN AGRICULTURAL ENGINEERING Vol. 42. No.3 JulylAugust/September 2000 119
Table I. Micro-meteorological data and relative drift for both treatments, air-assisted and conventional spraying.
Micro-meteorological data observed Relative drift
Date U2 (m1s) U4 (m1s) U8 (m1s) Wrms (m1s) U· (m1s) Zo(m) L(m) Air(%) No air(%)
1995-08-30 1.07 1.15 1.32 0.34 0.15 0.05 -3.29 0.26 2.811995-09-05 am 3.11 3.43 3.68 0.44 0.25 0.01 -13.53 0.51 8.421995-09-05 pm 4.89 5.43 5.94 0.63 0.36 0.02 -69.20 1.53 9.031995-09-08 2.56 2.76 3.00 0.43 0.20 0.01 -11.15 0.33 1.631995-09-20 am 3.54 3.82 4.17 0.42 0.20 0.01 -44.01 0.38 2.421995-09-20 pm 2.28 2.48 2.67 0.31 0.18 0.01 -11.17 0.66 2.911995-09-25 3.67 4.08 4.44 0.48 0.26 0.02 -70.87 0.34 4.121995-09-28 2.97 3.27 3.61 0.43 0.25 0.01 -25.24 0.29 2.201995-10-02 am 4.23 4.71 5.12 0.54 0.31 0.01 -57.12 0.27 2.921995-10-02 pm 4.91 5.51 5.97 0.59 0.37 0.01 -50.59 1.11 8.191995-10-03 3.11 3.41 3.71 0.46 0.26 0.01 -14.55 0.24 2.811995-10-05 2.18 2.40 2.57 0.31 0.15 0.01 -27.66 0.25 1.891995-10-11 4.36 4.84 5.27 0.55 0.27 0.00 -972.23 0.54 4.391995-10-12 4.97 5.55 6.05 0.56 0.36 0.01 -75.86 0.39 9.44
U2, U4, and U8 =, respectively, wind speed at 2, 4, and 8 m above ground. Wrms = vertical turbulent velocity. U· = friction velocity.Zo =roughness length. L = Monin-Obukhov length scale (Panofsky and Dutton 1984).
The drift percentages calculated and the micrometeorological data are summarized in Table I.
Paired comparisons of the drift are presented in Fig. 4. Thewind speed corresponding to each pair of experiments is shown
0.006
-o-NoaJr__ AIr
Total driftAIr:O.51%No air: 8.42%
O+---........---I----+----+---~--___to 0.001 0.002 0.003 0.004 0.005
Time Integrated volumetric flux (lJrn2)
7
2
8
8
8
Fig. 3b. Drift profile observed under a 3.11 mls windspeed at 2 m above ground.
RESULTS and DISCUSSION
The experimentation was performed under wind varying from1 to 5 mls and with atmospheric stability varying from neutralto moderately unstable. Typical drift profiles are shown for alow wind speed case (Fig. 3a), a medium wind speed case(Fig. 3b), and a high wind speed case (Fig. 3c). For the airassisted cases, the value of the maximum flux is almost stablewhile it increased by a factor of 5 in the absence of airassistance as wind speed increased. Furthermore, themaximum flux is located closer to the ground with airassistance, this effect being more pronounced at the lowerwind speed. With no air-assistance, the maximum flux occursbetween 1 and 2 m above the ground, this range correspondingto the breathing height. In that case, bystander exposure couldbe much more severe. These three graphs show that driftprofiles are much more stable under the effect of increasingwind speed for the air-assisted case.
0.001 0.002 0.003 0.004 0.005TIme fntegrated vclurne1rtc flux (1..JmI)
Tune Integrated volumetric flux ( LJm2 )
0.0060.005
-a-Noalr__ AIr
0.004
TotaIdrfftAlr:O.390kNo air:9.44%
0.0030.0020.001
6
2
7
5E~ 4'iiiJ: 3
0.006
-o-Noalr.......Nt
Total drlftAIr: 0.26%No alr:2.81%
2
6
7
8
Fig. 3a. Drift profile observed under a 1.07 mls windspeed at 2 m above ground.
Fig. 3c. Drift profile observed under a 4.97 mls windspeed at 2 m above ground.
120 PICHE, PANNETON and THERIAULT
Table II. Correlation matrix for drift and the available meteorological data.
U2 (m/s) U4 (m/s) U8 (m/s) Wnns (m/s) U' (m/s) z" (01) L(m) Air (% drifl) No air (% drift)
U2 (m/s) 1000' 0.999 0.999 0.926 0.924 -0.5t3 -0.312 0.535 0.676U4 (m/s) 1.000 0.999 0.928 0.929 -0.504 ·0.312 0.537 0.683U8 (m/s) 1.000 0.932 0.931 -0.490 -0.313 0.537 0.681Wrllls (m/s) l.OOO 0.940 -0.268 -0.317 0.574 0.678U' (m/s) 1.000 -0.306 -0.115 0.564 0.765z" (m) 1.000 ·0.271 -0.112 -0.116L (m) 1.000 -0.062 -0.040Air (% drifl) l.000 0.658No air (% drift) l.000
'Bold indicnlcs a significant correlation (p<O.05)
1.a
9 aa a8
oAk y'=' 1.672x - 1.20197 a Noair A2,=, 0.457
6~
5(§ a a•
3 a a a a
a a2 a y,=,0.171x-0.On2a
Fr '=' 0.2866 :0
0• 0 0 0
• 2 3 • 5 6
Wind speed at 2m (mls)
Fig. 5. I~ehltive drift 1'01' both treatmenls as a function ofwind speed at 2 III above ground.
CONCLUSION
The results obtaincd by this experimentation clearly show thatair-assiSl<ince using high air volume coupled with the use ofcoarser spray decreases drift significantly compared 10
conventional application. On average. the amount of drift isdecreascd by a factor of 9.9. At a downwind distance of 10m,only wind speed innucnced drift significantly for both airassisted and conventional boom spraying. The effect of windspeed on drift is reduced using air-assistance and a coarser
friction velocity is very high because the range of a1mosphericstability is narrow. For both the air-assisted and the non airassisted cases. the relative drift is positively correlated with thewind speed at all levels. the venica! turbulent velocity. and thefriction vclocity. The relative drift for the air-assisted case ispositively cOITelatcd 10 the relative drift for the non air-assistedcase. These con'elation resuhs suggest that a simple linearregression betwecn relative drift and the wind speed at any levelis the besl model that can be filted to the data. This wasconrirmcd using a forward and stepwise analysis (Drnper andSmith 1981). The regression of drift 3S u function of the windspeed at 2 m above ground is shown in Fig. 5. There is asignificant differcnce bctween both slopes at thc 5% confidencelevel. There is a factor 01' 9.8 between the slopes indicatingagain. that drift when using air-assistance coupled to a coarserspray is much less sensitive to meteorological conditions.
0.'
23 4.36 4.89 4.91 497Wind speed al 2m (mls)
.. --- I -I J-I,---
.v t-- -=i,1----
./ - I
,/ .i - r,V- I
'-.1---- rI;J. -
,IV i--,
~ - .- - .
1111.072.182.28 - N2.56 2.97 3.11 3 11
• 3.!>4 3.67 4 Ai<
Fig. 4. Paired comparisons of relative drift.
on the horizontal axis. Clearly the combination of airassistance wilh the lise of a coarser spray decreases drindramatically. The raLio of drift from the convcl1lional case tothe one for the air-assisted case is 9.9 on average. A linearregression between this ratio and the wind speed al 2 111 aboveground has been computed and it is not SlaLiSlically significant(p = 0.47). This indicates that the benefit of air-assistance andcoarser spray is of the same magnitude over the whole rangcof wind speed covered during the experimcnts. The design ofthe expcrimcnt has confounded the effect of air·assiswnce withthe effect of droplet size. According to the BCPC nozzlec1assi rication (Matthcws 1992), the nat fan nozzles used withthe air-assisted treatments givc a medium spray. The sprayquality associated with the hollow cone nozzles used fortreatments without air-assistance is more difricult to estimatcbut typically. hollow cone nozzles producc a fincr spray in thevery fine or line catcgories. Miller (1998) presellls datacomp;'lring drift from a conventional boom equipped withnozzles giving a medium or a fine spray. Typically. theamount 01' drift associated with the fine spray is twice as largeas the alllount produced using a medium spray. In our C'ISC.
confounding (he effect of droplet size and air-assistance. atenfold reduction in drift was measured. Therefore. it cun besafely inferred that the reduction in drift associated with airassistance alone is very significant.
A correlation matrix was computed using all the data fromTable I (Table II). As expectcd. the correlation between windspeed at various heights. the venicalturbulent velocity. and the
CANADIAN AGRICULTURAL ENGINEERING Vol. -12. No.3 July/Augll~tJScplclllbcr2000 121
spray. This is of practical interest because it implies that usingair-assistance and a coarser spray, pesticide application can beperfonned under meteorological conditions where applicationwith a conventional boom sprayer would not be recommended.For example, BCPC (1986) recommends to avoid sprayingwhen the wind speed at boom height is higher than 2.7 mls. At2 m above the ground, this translates to a wind speed of about4 mls for conditions similar to the ones over our experimentalsite. For the non air-assisted case, the regression model yields5.5% relative drift. For the air-assisted case, the regressiongives 0.6%.
The results showed an improvement in drift control usingair-assistance combined with a coarser spray and it has beenargued that a significant portion of the drift reduction can beattributed to air-assistance alone. On this basis, it might be ofinterest to perfonn experiments using smaller droplet sizes toverify if expected gains in pesticide efficacy associated withsmaller drops can lead to a reduction in the amount ofpesticide required. In this context, the role of air-assistance incontrolling drift is essential.
REFERENCES
BCPC. 1986. Nozzle Selection Handbook. Thornton Heath,UK: British Crop Protection Council.
Draper, N.R. and H. Smith. 1981. Applied RegressionAnalysis, 2nd ed. New York, NY: John Wiley and Sons,Inc.
Matthews G.A. 1992. Pesticide Application Methods, 2nd ed.Harlow, UK: Longman Scientific and Technical.
May, MJ. 1991. Early studies on spray drift, depositmanipulation and weed control in sugar beet with two airassisted boom sprayers. In Air-Assisted Spraying in CropProtection, eds. A. Lavers, P. Herrington, and E.S.E.Southcombe, BCPC Monograph No.46:89-97.
Miller, P.C.H. 1989. The field perfonnance ofelectrostaticallycharged hydraulic nozzle sprayers. In Proceedings of the4th EWRS Mediterranean Symposium, 324-333. ValencaPalau de la Musica, Spain, April 17-19.
Miller, P.C.H. 1998. The measurement and prediction of spraydrift - Work at the Silsoe Research Institute. InProceedings of the North American Conference onPesticide Spray Drift Management, ed. D. Buckley, 229244, Portland, MN, March 29 - April 1.
Nordbo, E. and W.A. Taylor. 1991. The effect of airassistance and spray quality (drop size) on the availability,unifonnity and deposition of spray on contrasting targets.In Air-Assisted Spraying in Crop Protection, eds. A.Lavers, P. Herrington, and E.S.E. Southcombe, BCPCMonograph No. 46:113-125.
122
Panneton, B., H. Philion, R. Theriault and M. Khelifi. 20ooa.Spray chamber evaluation of air-assisted spraying onbroccoli. Crop Science 40(2):444-448.
Panneton, B., H. Philion, R. Theriault and M. Khelifi. 2000b.Spray chamber evaluation of air-assisted spraying on potatoplants. Transactions ofthe ASAE 43(3):529-534.
Panneton, B. and R. Theriault. 1995. Nouveau laboratoire depulverisation agricole. In Comptes rendusdu 11~m~Colloquede genie rural, 19-34. Universite Laval, Quebec, QC, March1995.
Panofsky, H.A. and J.A. Dutton. 1984. AtmosphericTurbulence. New York, NY: John Wiley & Sons.
Quanquin, B. 1992. Less drift, more on target with the Twinsystem. ASAE Paper No. 92-1564. St-Joseph, MI: ASAE.
Reed, J.P, F.R. Hall and R.M. Riedel. 1993. Biologicalimplications of drift from sprayers in tomato fungicide fieldtrials. Plant Disease 77(2): 186-189.
Sharp, R.B. 1984. Comparison of drift from charged anduncharged sprays. In Proceeding of the British CropProtection Conference - Pests and Disease, 3:1027-1031.Brighton, England, November 19-22.
Spraying Systems Co. 1995. Teejet, Agricultural SprayProducts Catalog 45. Wheaton, IL: Spraying Systems Co.
Taylor, W.A., P.G. Andersen and S. Cooper. 1989. The use ofair-assistance in a field crop sprayer to reduce drift andmodify drop trajectories. In Proceedings of the BrightonCrop Protection Conference - Weeds, 2:631-640. Brighton,England, November 20-23.
Theriault, R., B. Panneton, H. Philion, M. Piche, and M.Khelifi. 1996. Optimisation de la pulverisation arampe avecjet d'air en vue de reduire les quantites de bouillie et depesticides, ]a derive et ]a puissance demandee. ReportNo. 14-41696815-007, Canada/Quebec Agreement for aSustainable Environment in Agriculture.
Watson, D.B., R.L. Wolff and G. Kapusta. 1984. Developmentof improved row crop spraying. Transactions of the ASAE28(5): 1445-1448.
Western, N.M. and E.C. Hislop. 1991. Drift of charged anduncharged spray droplets from an experimental air-assistedsprayer. In Air-Assisted Spraying in Crop Protection, eds.A. Lavers, P. Herrington and E.S.E. Southcombe, BCPCMonograph No. 46:69-76.
Wolff, R.L. and O.R. Smith. 1989. Field results of air-assistedspraying - Drift and dessication. ASAE Paper No. 89-1520.St-Joseph, MI: ASAE.
Womac, A.R., J.C. Goodwin and W.E. Hart. 1997.Comprehensive evaluation of droplet spectra from driftreduction nozzles. ASAE Paper No. 971069. St-Joseph, MI:ASAE.
PICHE, PANNETON and THERIAULT
Sensors to measure mass-flow-ratethrough a forage harvester
H. MARTELl and P. SAVOIE2
/Departement des sols et de genie agroalimentaire, Universite Laval, Sainte-Foy, QC, Canada G1K 7P4; and 2Agriculture and AgriFood Canada, Soils and Crops Research and Development Centre, 2560 Hochelaga Boulevard, Sainte-Foy, QC, Canada GIV 2J3.Contribution no. 669, Agriculture and Agri-Food Canada, Soils and Crops Research and Development Centre. Presented as ASAECSAE Paper No. 99-1050. Received 30 November 1999; accepted 31 May 2000.
Martel, H. and Savoie, P. 2000. Sensors to measure mass-flow-ratethrough a forage harvester. Can. Agric. Eng. 42: 123-129. Fourdifferent sensors were used to estimate mass-flow-rate and moisture ona pull-type forage harvester. The sensors measured feedrolldisplacement, crop impact force against a hinged plate located abovethe blower, the frequency drop of a capacitance controlled oscillatornear the end of the spout, and the number of light beam interruptionsby forage particles in the spout. Tests were conducted in a com fieldwith a commercial forage harvester modified with the first two sensors(feedroll displacement, impact force), and in the laboratory using aforage blower adapted to a forage harvester spout for the last twosensors (capacitance controlled oscillator, light beam interruption).The capacitance controlled oscillator was also characterized in a staticmode in the laboratory with alfalfa and timothy particles. When testingin a com field, good correlations were obtained between estimatedmass-flow-rate and either the feedroll displacement (R2 = 94%) or thecrop impact force (R2 =95%). When testing in the laboratory, thecorrelation between mass-flow-rate and the oscillator drop was verygood (R2 =96%) after a correction procedure. The number of lightbeam interruptions was not well correlated with mass flow (R2 = 43%for LEOs placed after the capacitor and R2 = 6% for LEOs placedbefore the capacitor). During static measures with alfalfa and timothy,the oscillator frequency drop was also related to crop moisture butcalibration corrections were required to consider differences betweencrop species and chop lengths.
On a utilise quatre capteurs pour mesurer Ie debit de masse et lateneur en eau des fourrages durant la recolte avec une fourrageretrainee. Les capteurs mesuraient Ie deplacement des rouleauxd'alimentation, la force d'impact des fourrages dans Ie coude de lagoulotte, la chute de frequence d'un oscillateur integre ades plaquesde capacitance et Ie nombre de faisceaux de lumiere interrompus parIe passage des particules hachees. On a evalue les deux premierscapteurs (deplacement des rouleaux, force d'impact) dans un champ demars fourrager tandis qu'on a evalue les deux autres capteurs(oscillateur integre a des plaques de capacitance, interruption defaisceaux) au laboratoire a I'aide d'un souffleur surmonte d'unegoulotte de fourragere. L'oscillateur a aussi fait l'objet d'uneevaluation statique avec deux fourrages haches (fleole, luzerne).Durant les essais dynamiques dans Ie champ de mai's, on a obtenu debonnes correlations entre Ie debit de masse et Ie deplacement desrouleaux d'alimentation (R2 = 94%) ou la force d'impact (R2 = 95%).Durant les essais dynamiques au laboratoire, on a aussi obtenu unebonne relation entre la chute de frequence de I'oscillateur et Ie debit demasse (R2 =96%) apres une procedure de correction. Le nombre defaisceaux de lumiece interrompus n'etait pas bien correle au debit(R2 =43% pour les diodes placees apres Ie condensateur et R2 =6%pour les diodes placees avant Ie condensateur). Durant les mesures
statiques avec des particules de fleole et de luzerne, on a observe unecertaine relation entre la chute de frequence de I'oscillateur etla teneuren eau, mais une calibration importante etait necessaire pour tenircompte de differences entre les especes et les longueurs de hachage.
INTRODUCTION
Precision agriculture may be applied to forage crops to measurelocal yield variations in the field, to estimate total crop yieldharvested, and to plan the feeding program on a livestock farm.Forages present a particular challenge compared to most othercrops because their moisture content can vary at the time ofharvest from as low as 10% for hay to up to 80% for silage.Therefore, moisture and mass of forage crops must be measuredsimultaneously and for a relatively wide range of moisturecontents.
If mass and moisture are to be measured at the time forageis removed from the field (usually after a wilting period),appropriate sensors are likely to be different for hay and silagecrops. Hay is typically baled and stored at moisture contentsbetween 10 and 20% while crops for silage are typicallyharvested and stored at moisture contents between 40 and 70%.Chopped forages present an additional constraint because theyare not accumulated in the harvester as opposed to baled forageswhich are accumulated until a bale is formed and allowweighing within the baler.
Estimation of material flow in a forage harvester willtherefore have to rely on indirect measurements that arecorrelated with mass and moisture. Several studies haveexamined a number of sensors in forage harvesters such as afeedroll displacement transducer (Mains et a1. 1983; Barnett andShinners 1998), an infrared emitter-receptor system (Bull 1993),a radioactive isotope emitter-receptor system (Auernhammer eta1. 1995), impact force measurement, or a combination of othersensors (Barnett and Shinners 1998). Marcotte et a1. (1999)provided a critical review of various sensors which have beenconsidered in forage harvesters. Signals measured by suchsensors may be more or less influenced by total mass flow andmoisture flow.
The objective of this research was to assess a selectednumber of sensors installed within a forage harvester and therelationship between the observed signals and mass andmoisture flow variations.
CANADIAN AGRICULTURAL ENGINEERING Vol. 42. No.3 July/August/September 2000 123
a?d a volume sensor composed of light emittingdiodes (LEDs) on one vertical side of the spoutand photo transistors on the other side. Theposition of these four sensors within the forageharvester is illustrated in Fig. I.
The load cell was an "S" type with acapacity of 216 N. It measured the impact forcecaused by the moving forage particles againstthe hinged plate with a surface area equivalentto the chute width (235 nun) by a length of 500mm. The displacement transducer was a waterresistant LT model sensor (Cymatix,Burlingame, CA) with a 0-150 mmdisplacement measuring range. These twosensors were fitted to the forage harvester forthe field experiment. A portable computer witha data acquisition card (PCMCIA DAQ-700,National Instruments, Austin, TX) was used tocollect signals at a frequency of 37 Hz. Thiswas the maximum frequency at which allsensors could be scanned individually andsequentially without interference.
The capacitance controlled oscillator wasfabricated in-house with a model TS555CNtimer operating in the astable mode at about 880
kHz (SGS Thomson Microelectronics, Marlow, England). Thisfrequency was near the maximum operating frequency (!!! IMHz) of the timer and was chosen to minimize the displacementof free charges naturally present in the forage during ameasurement period. The electronic circuit of the oscillator isshown in Fig. 2. Similarly the LED emitter-receptor was builtin-house with 16 LEDs (880 nm, model KIE 7304, KnightLites~ placed vertically on one side of the spout and 16 phototransistors (model KID 7404, Knight Lites) on the other side ofthe spout. The 16 parallel light beams could be interrupted asforage particles flowed in the spout. The LED beam receptors(photo transistors) were scanned at a frequency of 37 Hz. Asmore particles flowed in the chute, more beams were expectedto be interrupted and the signal attenuated to the receptors. Theelectronic design of the LED emitter-receptor is shown in Fig.3. The capacitance controlled oscillator and the LED emitterreceptor were evaluated in the dynamic laboratory experimentdescribed below. The oscillator was also characterized in astatic laboratory experiment.
Field experiment
Field measurements were carried out by harvesting whole-plantcom silage during the week of September 28, 1997 inLennoxville, QC. A section of the field was separated into 24plots measuring 1.5 m wide (the width of two com rows) by 26m in length. Six trials were carried out using two pairs of slowand high speed as illustrated in Fig. 4. The transmission wasshifted rapidly to high or low gear to create a rapid change (±10to 20%) in mass-flow-rate (actual forward speeds were 4.7 and5.8 kmIh at low speed, and 6.1 and 8.3 kmIh at high speed). Therapid change in forward speed was done to evaluate thesensitivity of sensors to a rapid change in the mass-flow-rate.
Yield samples were taken in each of the 24 plots by cuttinga I-m length over two rows at a random location along each ofthe 26 m long plot prior to harvesting. An average yield wasestimated by grouping three plots in each section (trials 1,2, and3 grouped; trials 4, 5, and 6 grouped). The mass-flow-rate (wet
22nF
IkO
Fig. 2. Electronics diagram of oscillator (880 kHz).
Fig. 1. Schematic view of sensors evaluated on the forage harvester.The capacitor controlled oscillator and LED emitter-receptorwere evaluated in the laboratory only.
MATERIALS and METHODS
Instrumented forage harvesterSensors were placed either on a commercial forage harvester forfield experiments or on the forage harvester's spout adapted toa blower for laboratory experiments. A pull-type forageharvester (Dion model 1224, 12 knives) was used for the fieldexpe.rim~nt. This ~arvest~r is ~esigned to cut and blow forageuDl?lfectlOnaily Without Side displacement between the cuttingcylinder and the blower. In the case of laboratory experiments,pre~chopped forage was fed from a horizontal belt conveyor toa DlOn model 1660 flywheel type blower above which the spoutof a 1224 forage harvester was fitted. The technicalcharacteristics of the fomge harvester and the blower aredescribed in more detail elsewhere (Dion 1999).
Four types of sensors were selected: a load cell measuringt~e impact force against a hinged plate in the spout, adisplacement transducer placed at one end of the two upperfeedrolls to measure the vertical movement of the rolls'centerline, a capacitance controlled oscillator placed at the endof the spout to measure changes induced by the forage particles,
124 MARTEL and SAVOlE
Fig. 3. Electronics diagram of LED emitter-receptor system.
matter per unit time) for each of the 24 sections was thenestimated as:
Q = wVYa\·,.
exp 3.6
Receptors
(1)
according to ASAE Standard S358.2(ASAE 1999a).
dOn Signal filtering for the fieldexperimentData collected in the field were
600 Emitters processed either with a low-passfilter or with one of two band-passfilters (Doebelin 1975). In the case ofthe low-pass filter, the cutofffrequency was 1 Hz for the load cellmeasuring the impact force in thespout and 3 Hz for the positiontransducer measuring feedroll displacement. The two band-pass filtersisolated the signal within a frequencyrange of 0.5 Hz, in one case aroundthe natural frequency (i.e. theoccurrence of maximum amplitude)and in the other case between 14.0and 14.5 Hz. The natural frequencyvaried from 3 to 10Hz, depending onthe sensor and the trial. The choice ofthe range for the second band-passfilter (14.0 to 14.5 Hz) was based ona frequency analysis.
The data retained for analysis in each plot started 3 s aftereach section change. This delay allowed for completion of thetractor gear and speed change after a manual sign given to theoperator and acceleration or deceleration of the tractor.
4
Dynamic laboratory experimentThe dynamic laboratory experiment was carried out during theweek of March 9, 1998. Whole-plant com was previouslyharvested and chopped at 13 mm theoretical length at theDeschambault Experimental Farm (Deschambault, QC) andfrozen immediately (October, 1997). The material was thaweda few hours prior to experimentation. Three feeding rates of 3.6,7.2, and 10.7 kg/s were repeated three times and the nine testswere repeated on three different days (blocks). A measuredquantity of chopped forage was placed on a horizontal, 6.4 mlong conveyor that moved at 1.3 mls.
The forage was fed into the blower and conveyedpneumatically through the spout past the sensors. The LEDemitter-receptor system was placed after the capacitor during
two blocks (6 replications) and before thecapacitor during the other block (3 replications).The LED signals were analyzed on the basis ofthe number of light beams interrupted by forageparticles at every scanning (37 Hz). Theoscillator was calibrated with the blowerrunning at zero particle flow. The variation inoscillation frequency was compared tovariations in the wet mass flow. Due tohardware problems, data from the capacitancecontrolled oscillator were recorded only on thesecond day.
Static laboratory experimentThe static laboratory experiment was carried outwith fresh alfalfa and timothy after a wilting
I:::: I 8.3 kmIh
//1111//11////////11
1111111111//111111//
3
111111////1111111111
DS.8km1h
.. ' ::::::::::::::::::::2
////////////1//////1
////11////1111111111
1111116.1 kmIh
Direction ofharvest
........................................//111111111111111111
~ 26m ~
~~x~~~~: ~:4£'~ji:.;;:~i;~'~: :::::::::::::::::::: '~1~~~~;~g:'" '.'~:,::.;.:~~: :::::::::::::::::::::::::::::::::::::::: h;. ;-.~~~-:,: :~;~;.';! :::::: :::::::::::::: .~!;~~~.c~-::~~~,~..~~..~:.~~:~)
Yav\:
Trial:123456
Section:
Fig. 4. Experimental subdivisions of corn field.
where:Qexp
wV
= experimental mass-flow-rate (kg/s),=com harvester width (m),=forward tractor speed (kmJh), and=yield, on a wet matter basis, averaged over three
grouped plots (kg/m2).
The sensor outputs were averaged over each 26 m lengthafter filtering as described below. The average sensor outputswere then correlated with the experimental mass-flow-rate foreach section in each trial. Moisture content was estimated fromfour samples collected after forage harvesting four consecutive26-m long sections in a single trial. Samples were oven dried at103°C for 24 h to estimate dry matter and moisture content
CANADIAN AGRICULTURAL ENGINEERING Vol. 42, No.3 JulylAugust/September 2000 125
(2)Q1 = - 0.611 +2.46F
The static laboratory experiment was a four-factor split-splitblock design. Two factors (crop species and quantity of drymatter in the capacitor) were completely randomized whilematurity was the first split factor and moisture content was thesecond split factor. There were two replications for eachtreatment combination.
RESULTS and DISCUSSION
Field experiment
Wet matter yield of com ranged from 6.7 to 10.6 kg/m2 andaveraged 8.6 kg/m2. Average moisture content was 79% and drymatter yield 18.3 t1ha. Speed varied from 4.5 to 9.5 kmIh whilemass-flow-rate ranged from 15.2 to 32.2 kg/so The low-passfilter provided a considerably better relationship (R2=94.8%)than the two band-pass filters (R2= 8.4%, based on the naturalfrequency; R2 = 1.4%, based on the frequency range of 14.0 to14.5 Hz). The average low-pass filtered signals of the load cellin the spout per test run versus estimated mass-flow-rate areillustrated in Fig. 5. The linear relationship (R2=94.8%) was:
where:QI =predicted mass-flow-rate (kg/s) as a function of
impact force, andF =impact force (N) measured in the spout.
The feedroll displacement transducer signals were alsoprocessed with a low-pass filter. The average feedrolldisplacement per test run versus estimated mass-flow-rate isshown in Fig. 6. The coefficient of determination R2was 93.7%and the linear relationship:
3.5
3530
30
10 15 20 25
Mass-flow-rate (kgIs)
10 15 20 25Mass-flow-rate (legis)
5
5
26
-22
! 18
Ii 148§ 10~ 6is
2
-20
14
12
g 10
§ 8
& 64
2
Fig. S. Measured force from the load cell versus massDow-rate for chopped corn.
where:Q2 = predicted mass-flow-rate (kg/s) as a function of
feedroll displacement, andD = feedroll displacement (mm).
The low-pass filters provided a good relationship betweenmass-flow-rate and either the load cell in the spout or thedisplacement transducer at the feedroll. These observations arelimited for a mass-flow-rate of 15 to 32 kg/s, where the averagemoisture of com was 79%. This mass-flow-rate rangecorresponds to a harvest rate of 54 to 115 tIh which isapproximately between half and full capacity of the Dion 1224harvester. The relationship between feedroll displacement orforce against a plate in the spout might change for mass-flowrates outside this range and for a forage harvester with ageometry different from the one used in this experiment.
Dynamic laboratory experiment
Chopped com was fed in the laboratory through the blowerspout system at three mass-flow-rates of 3.6, 7.2, and 10.7 kg/soThe average moisture content of this frozen-thawed com was60%. The LED emitter-receptor system was placed either beforeor after the capacitor in the spout. Figure 7 shows a typicalhistogram of the number of beams interrupted by particlesduring two trials. Each trial lasted on average 4.9 s with ascanning frequency of 37 Hz. As the amount of forage used ina trial increased, one would expect the number of interrupted
Fig. 6. Measured displacement from the position transducer versus mass-Dow-rate for chopped corn.
period. Crops were mowed at three dates: June 1, 15, and 29,1998, corresponding to three stages of maturity: bud, earlybloom (5-10%) and half bloom for alfalfa, and boot, earlyheading, and full heading for timothy. Crops were left to wiltuntil they reached one of four levels of moisture content:approximately 80, 70, 60, and 45% on a wet basis. Afterwilting, crops were chopped with a laboratory rotary cutter(Forano, Plessisville, QC), sieved according to Standard 8424.1(ASAE 1999b) and separated into three groups according tolength. Short particles were defined as those collected from thepan « 1.7 mm) and the fifth screen (1.7 to 5.6 mm). Mediumparticles were collected from the third screen (9.0 to 18.0 mm).Long particles were collected from the second screen (18 to 27mm).
Samples were prepared to obtain fixed quantities of drymatter (0, 25, 50, and 75 g) so the quantity of water woulddecrease with wilting time. Each sample was spread in aplexiglass box (300 mm by 230 mm area, 100 mm high) whichwas placed between the two plates of the capacitor. The bottomplate of the capacitor was just under the bottom of the box whilethe upper plate was fixed 110 mm above, just over the box. Theoscillator's frequency change reported was the average value often scans of the forage sample in the box.
Q2 =-0.359+ 1.37D (3)
126 MARTEL and SAVOlE
(7)
(5)
(6)
16
R I _3 =1.70- 0.0694n
R4-9 = 1.19 -0.0529n
(l>f=). =(l>f",).(::J
14
the various densities of particles. Ifa low mass-flow-rate wassufficient to hide the LEDemissions from the receptors, thena high mass-flow-rate alsointerrupted the same number oflight beams. A higher scanningfrequency than the one used (37Hz) might result in betterdiscrimination between differentmass-flow-rates.
The other sensor used in thedynamic laboratory experimentwas the capacitance controlledoscillator. When forage particlesflowed between the two parallel
plates, the frequency of the oscillator decreased. The differencebetween the control frequency (no flow) and the actualfrequency was summed over each trial. This summation wasexpected to be proportional with accumulated mass flow.Therefore, the ratio of summed frequency drop to mass-flowrate should be relatively constant. The original experimentaldata showed however a gradual decrease from trial 1 to 3 andfrom trial 4 to 9, and a sudden change between trial 3 and 4(Table I). This drift in the ratio of frequency drop to mass-flowrate reflected an experimental constraint whereby the samechopped forage particles were recycled through the blowerspout system because of limited quantities. The chopped comtended to become finer and drier as the trial number increased.The sudden change from trial 3 to trial 4 reflected a suddenreduction in particle size after a blower jam caused grinding ofan important proportion of the chopped forage in the third trial.Two linear estimators of this ratio (frequency drop over massflow-rate) were used to correct the frequency drop:
(4)
2 4 6 8 10 12
No. of interrupted ligbtbeams
(b)~ ----
- ,...-l-
I-
. .
14
12
~ 10
~ 8c::g 6
1 42
oo144 6 8 10 12
No. of interrupted Iightbeams
2
(a)
Q3 =-36.1 +4.40L
3S
30
~ 2S
~ 20
I IS
110SOIl:....--"_......_-IooI........I-Lo&.............-.I-&..&...~--
o
Fig. 7. Examples of signals from the LED emitter-receptor placed:(a) after the capacitor; (b) before the capacitor.
where:Q3 = predicted mass.,.flow-rate (kg/s) as a function of
median number of light beam interruptions, andL = median number of light beam interruptions.
When the LED system was placed before the capacitor, therelationship between mass flow and the number of interruptedlight beams was very poor (R2 = 6.4%; data not shown). Thelow response of the LED emitter-receptor to change in mass-flow-rate was probably due to the low discrimination between
Table I. Summed experimental (dfexp) and corrected (dfcor) frequency dropof capacitor controlled oscillator with chopped corn in the dynamiclaboratory experiment.
beams to increase, causing the normal curve associated to thehistogram to shift upward or to the right, or both.
Although the mass-flow-rate ranged from 3.6 to 10.7 kg/s,the average number of beam interruptions ranged only from9.05 to 10.70 (out of 16 light beams). There was a slighttendency for the average median value to increase with massflow-rate (9.33, 10.02, and 10.09 for low, medium, and highmass-flow-rate, respectively). The linear relationship with R2 =43% was:
Trial mass-flow- ~fexp Frequency drop per unit mass- ~fcurnumber rate (MHz) flow-rate (MHz/[kg/s]) (MHz)
(kg WMls)Experimental Estimated
7.2 12.00 1.67 1.63 12.53
2 7.2 10.74 1.49 1.56 11.71
3 3.6 5.50 1.53 1.49 6.28- - - - - - - -- - - - - - - - -- - - - - - - -4 10.7 9.78 0.91 0.97 17.15
5 10.7 10.22 0.96 0.92 18.90
6 3.6 3.53 0.98 0.87 6.90
7 7.2 5.95 0.83 0.82 12.34
8 3.6 2.02 0.56 0.76 4.52
9 10.7 8.68 0.81 0.71 20.80
CANADIAN AGRICULTURAL ENGINEERING Vol. 42. No.3 July/August/September 2000
where:RI_3
~)
= estimate of frequency drop perunit mass-flow-rate(MHz/[kg/s]) for trials I to 3,
= trial number,= estimate of frequency drop per
unit mass-flow-rate(MHz/[kgls]) for trials 4 to 9,
= corrected frequency drop(MHz),experimental frequency drop(MHz) from the capacitorcontrolled oscillator, and
= linear estimate of initialexperimental frequency dropper unit mass-flow-rate (1.70MHz/[kg/sD.
127
Fig. 8. Frequency drop of oscillator for static quantities of forage placed betweenthe capacitor plates: (a) early maturity timothy, medium chop at 83%moisture (R2 =98%); (b) late maturity timothy, long chop at 68% moisture(R2 =87%).
• frequency drop due to foragemass and the tray (Hz),basic frequency drop when anempty tray is placed betweenthe two plates of the capacitor(Hz),
P(i) = function specific to each crop,moisture content and lengthof cut (Hzlkg), andactual forage dry matter massbetween plates (kg).
d =s
where:~f =
The function P(i) was approximated asa linear function of moisture content by:
•(b)_ ISO
£125
!IOOt' 75
g SO
! 2S
0~""""'~""""1.I.I. .....~u.a.1.I.I.""""",.I."I.,IL.I..
0.00 0.02 0.04 0.06 0.08 0.10 0.12
Dry mass (kg)
0.0I 0.02 0.03 0.04 O.OS 0.06
Dry mass (kg)
(a)
(10)
1344
1576
where:bo = parameter to estimate frequency drop for a specific
crop and chop length (Hzlkg),b l = parameter to estimate frequency drop for a specific
crop and chop length (Hzlkg), andM = moisture content on a decimal wet basis (water over
total wet forage mass).
Figure 8 shows two examples of the relationship betweenthe frequency drop and the dry matter mass placed between theplates. Parameters bo and b l vary with crop species and lengthofcut. Table II shows parameters estimated by linear regressionfor alfalfa and timothy at three lengths of cut.
With the above parameters, it is possible to estimate themass-flow-rate or the water content when the other variable isknown. For example, the timothy illustrated in Fig. 8(a) had amoisture content of 83% and medium length. The frequencydrop factor Plil was estimated as 706 kHzlkg from parameters inTable II. Therefore, a quantity of 0.050 kg between the twocapacitor plates would cause a net frequency drop of about 35kHz. For a measured frequency drop, it would therefore bepossible to estimate the moisture content or the mass-flow-rateif one of these two variables is measured by anotherindependent sensor.
These parameters can also be used to estimate the precisionobtained with the capacitance controlled oscillator bycomparing estimations with measured values. This was done to
estimate moisture content when dry matter mass andthe frequency drop are known. For alfalfa, thecoefficient ofdetermination (R2
) ranged between 53and 59% and the absolute root mean square errorbetween 9.3 and 9.9%. For timothy, R2 ranged from74 to 78% and RMSE between 9.4 and 10.9%.
It is likely parameters would need to beestimated for each crop species. Length of cut wasalso an important factor as shown by the estimatedparameters in Table II. Data from different stages ofmaturity were used to generate the parameters inorder to make them independent of specific growthconditions and harvesting time. A higher initialfrequency than the one used for the oscillator (880kHz) might also improve the sensitivity of thefrequency drop as a function of moisture content by
(8)
(9)
-109
-146
Long (18-27 mm)
Q4 =0.708+0523(8/cor)
Particle length range
Short (0-6 mm) Medium (9-18 mm)
bo bl bo bl
Alfalfa 46.7 354 -31.4 823
Timothy 30.5 372 -6.4 858
Table II. Parameters boand b l (kHzIkg) to estimate the frequencydrop factor Pm for the static capacitance controlledoscillator system.
When the corrected frequency drops from Table I wereanalyzed against mass-flow-rate, a very good correlation (R2 =96.1 %) was obtained. The relationship was:
where Q4 =predicted mass-flow-rate (kg/s) as a function of thecorrected frequency drop.
The mass-flow-rate is seen to be well correlated with the"corrected" summed frequency differences. Such a correctionis valid only for the specific experimental conditions. A moregeneral correction might be developed for harvest with thecapacitance controlled oscillator under a wide range ofconditions. This laboratory experiment indicates nonetheless thepotential for oscillation differences to be well correlated withmass particle flow.
Static laboratory experiment
As in the previous experiment, the oscillator's frequencydecreased with an increase of material between the capacitorplates. Since the experiment was static, only an average valueof the frequency drop based on ten successive readings wasused. The relationship observed was:
128 MARTEL and SAVOlE
further minimizing the displacement of free charges present inthe forage within a measurement period. This would reduce therelative influence of other factors such as species, maturity andchop length. Further study is required to understand therelationship between oscillation frequency drop, dry matterflow, and water flow in a capacitor.
CONCLUSIONS
I. In a field experiment with chopped whole-plant corn, theimpact force measured against a hinged plate in the spout ofa forage harvester was well correlated with the mass-flowrate (R2 = 94.8%). A vertical displacement transducerlocated at the feedrolls was also well correlated (R2 =93.7%) with mass-flow-rate after signal filtering.
2. In a laboratory experiment with chopped whole-plant corn,the median number of interrupted beams from an LEDemitter-receptor was not sensitive to changes in the massflow-rate. A two-plate capacitor with an oscillator (880 kHz)showed a linear drop of the oscillator's frequency as themass-flow-rate increased (R2 = 96.1 % after a correctionprocedure).
3. The drop in frequency of a capacitance controlled oscillatorincreased linearly with mass and the slope was related tomoisture content. A number ofcalibration parameters wouldbe required to cover a broad range of crop species,maturities and chop lengths.
ACKNOWLEDGEMENTS
The authors express their appreciation for the financialcontribution of Agriculture and Agri-Food Canada (AAC)through its Matching Investment Initiative of the Sainte-FoyResearch Centre and of Dion Machineries of Boisbriand andInnotag of Beloeil, QC. Financial support also was provided bythe Conseil des recherches en pecheries et agroalimentaire duQuebec. Long term support from the Natural Science andEngineering Research Council of Canada, through its researchgrant program, is also acknowledged. Authors thank ProfessorRoger Theriault for advice, Martin Roberge, Eric Morel,Sebastien Descoteaux, and Dominic Marcotte for technicalassistance with the equipment and experimentation, and PatrickLemire and Anne-Marie Tremblay for assistance with editing.
REFERENCES
ASAE. I999a. Moisture measurement - forages. StandardASAE S358.2. In ASAE Standards 1999, 46th ed., 569. St.Joseph, MI: ASAE.
ASAE. 1999b. Method of determining and expressing particlesize of chopped forage materials by screening. StandardANSIIASAE S424.1. In ASAE Standards /999, 46th ed.,578-580. S1. Joseph, MI: ASAE.
Auernhammer, H., M. Demmel and P.J.M. Pirro. 1995. Yieldmeasurement on self propelled forage harvesters. ASAEPaper No. 951757. S1. Joseph, MI: ASAE.
Barnett, N.G. and K.J. Shinners. 1998. Analysis of systems tomeasure mass-flow-rate and moisture on a forage harvester.ASAE Paper No. 981118. S1. Joseph, MI: ASAE.
Bull, C.R. 1993. Calibration of a near infrared reflectancemoisture meter for grass. Journal of AgriculturalEngineering Research 54(3): 177-185.
Dion. 1999. Model 1224 forage harvester with corn cracker.Available at http://www.dionmachineries.com/. AccessedApril 6, 1999.
Doebelin, E.G. 1975. Measurement Systems, Application andDesign, revised edition. New York, NY: McGraw Hill.
Mains, W.H., H.P. Harrison and R. Hironaka. 1983. Feedrolldisplacement of a forage harvester as a measurement of thethroughput of harvested crops. Agricultural Electronics 1983 and Beyond 1:52-59. St. Joseph, MI: ASAE.
Marcotte, D., P. Savoie, H. Martel and R. Theriault. 1999.Precision agriculture for hay and forage crops: A review ofsensors and potential applications. ASAE Paper No. 991049.S1. Joseph, MI: ASAE.
CANADIAN AGRICULTURAL ENGINEERING Vol. 42. No.3 JulylAugust/September 2000 129
Treatment of slaughterhouse wastewaterin anaerobic sequencing batch reactors
OJ. MASSE and L. MASSE
Agriculture and Agri-Food Canada, P.O. Box 90,2000 Route 108 East, Lennoxville, QC, Canada J1M 123. Agriculture and AgriFood Canada contribution No. 659. Received 5 July 1999; accepted 6 June 2000.
Masse, D.I. and Masse, L. 2000. Treatment of slaughterhousewastewater in anaerobic sequencing batch reactors. Can. Agric.Eng. 42: 131-137. Slaughterhouse wastewater was treated in four 42-Lanaerobic sequencing batch reactors (ASBRs) operated at 30·C. TwoASBRs were seeded with anaerobic granular sludge from a milkprocessing plant (MPP) reactor and two ASBRs received anaerobicnon-granulated sludge from a municipal wastewater treatment plant.Influent total chemical oxygen demand (TCOD) ranged from 6908 toJJ 500 mgIL, of which approximately 50% were in the form ofsuspended solids (SS). Total COD was reduced by 90% to 96% atorganic loading rates (OLRs) ranging from 2.07 to 4.93 kg m') d· 1 anda hydraulic retention time of 2 days. Soluble COD was reduced byover 95% in most samples. During the start-up period, highconcentrations ofsolids were lost in the effluent, but under steady stateoperation, at OLRs above 3 kg m-3 d- I
, biomass retention was adequateand effluent SS averaged 364 mgIL. Reactors seeded with municipalsludge performed slightly better than those containing the MPP sludge,especially during start-up, but differences between the two sludgesdecreased with time. The biogas contained 75% methane. About90.5% of the COD removed was methanized and volatile suspendedsolid (VSS) accumulation (apparent biomass yield plus undegradedsolids from the influent) was evaluated at 0.068 kg VSS per kg CODremoved. This high degree of methanization indicated that mostsoluble and suspended organics were degraded during treatment inASBRs operated at 30·C.
Quatre bioreacteurs it operations sequentielles (BOS) d'unecapacite de 42 L ont ete utilises pour Ie traitement anaerobie d'eauxusees d'abattoir a30·C. Deux des BOS ont ete ensemences de bouesgranulaires anaerobies provenant d'une usine de transformation du lait(UTL), tandis que les deux autres BOS ont ete ensemences de bouesnon granulaires provenant d'une usine municipale de traitement deseaux usees. La demande chimique en oxygene totale (DCOT) dusubstrat variait de 6 908mgIL it 11 500 mgIL, dont la moitie provenaitde matieres en suspension (MES). La DCOT a ete reduite de 90 % it96% it des taux de charge organique (TCO) variant de 2,07 it 4,93 kgm·3j"1 et it un temps de retention hydraulique de deux jours. La DCOsoluble a ete reduite de plus de 95% dans la plupart des echantillons.On a observe d'importantes pertes de boues pendant la periode dedemarrage, mais en regime stationnaire, ades TCO de plus de 3 kg m·)j"1 , la retention de la biomasse dans les reacteurs etait adequate, et lesconcentrations de MES de l'effluent s'elevaient en moyenne a364mgIL. Le rendement des reacteurs ensemences de boues provenant deI'usine de traitement municipale s'est revele legerement superieur acelui des reacteurs ensemences des boues provenant de I' UTL,notamment pendant la periode de demarrage. Cependant, les ecartsentre les deux types de boues se sont estompes avec Ie temps. Lebiogaz comprenait 75 % de methane. La methanisation de la DCOeliminee s'estelevee a90.5%. L'accumulation de matieres volatiles ensuspension (MVS) (soit Ie rendement apparent en biomasse plusI'accumulation des solides non decomposes du substral) a ete evaluee
it 0,068 kg de MVS par kg de DCO eliminee. Ce niveau eleve demethanisation indique que la plus grande partie de la matiereorganique soluble et en suspension a ete decomposee dans les BOS aucours du traitement anaerobie a30°C.
INTRODUCTION
Slaughterhouses produce a wastewater highly charged in solubleand insoluble organics. In Quebec and Ontario, hogslaughterhouses generally discharge their wastewater inmunicipal sewers after some level of primary and/or chemicalpretreatment at the plant (Masse and Masse 2000). Thesepreliminary treatments, however, are not sufficient to lowerpollutant levels below municipal standards. Slaughterhousesmust therefore pay a surcharge to have their wastewater furthertreated at the municipal treatment plant. Existing in-plantwastewater treatment systems also produce large quantities ofputrefactive and bulky sludge, which requires special handlingand/or further treatment.
Anaerobic digestion in high-rate reactors represents anattractive alternative for wastewater treatment at theslaughterhouse plant. First, slaughterhouse wastewater isparticularly well suited for anaerobic treatment. It contains highconcentrations of biodegradable organics, mostly from fats andproteins, sufficient alkalinity, and adequate phosphorous,nitrogen, and micronutrient concentrations for bacterial growth.It does not include toxic compounds and has a relatively warmtemperature between 20 and 30·C. Secondly, anaerobicdigestion provides high COD and suspended solid (SS) removalwhile producing a recoverable source of energy in the form ofmethane. It generates very low quantity of sludge and does notrequire aeration or chemical pretreatment. Finally, anaerobicbacteria can survive unfed for long periods of time, an importantfeature for smaller slaughterhouses that operate just a few daysa week or close down during slow or holiday periods.
The anaerobic sequencing batch reactor (ASBR) developedby Agriculture and Agri-Food Canada would be especiallyappropriate for slaughterhouses, because it can operate withlimited capital costs, energy, and manpower. This newtechnology has been successfully applied on laboratory andsemi-commercial scales for the treatment of swine manureslurry (Masse 1995; Masse and Croteau 1998; Masse andDroste 1997; Masse et al. 1996, 1997). The objective of thisproject was to demonstrate the feasibility of using ASBRsoperated at 30·C to treat slaughterhouse wastewater. Data froma five-month experiment are presented and discussed.
CANADIAN AGRICULTURAL ENGINEERING Vol. 42, No.3 July/August/September 2000 131
Fig. I. Opel"ation of the anaerobic sequcncing hatchreactor.
close contact betwecn organics und bacteria. Mixing should be asgentle as possible to uvoid disrupting the fonnation of bacterialnoes. Intennittent mixing is also preferred to continuous mixingbccuuse it improves biollluss settling and rcactor perfonnance(Sung and Dague 1995). The food to micro-org,anism (F/M) ratiois high at thc beginning of the react phasc and organic conversioninto biogas, as predicled by Monod kinetics, is at its maximum(Sung and Dague 1995). The length of the reaCI period will dependon substrntc characteristics and cfnucnt quality requiremcnts. Forwastewaters containing high 55 concentrations, more contact timebetween bactelia and substratc will bc required for the completehydrolysis of particulatcs. \Vhen gas production rat.e has decreasedto a minimum, reactor contcnt is allowed to settle. The low F/Mratio at the end of the react phm;c favours biomass nocculation andsCllling (Sung and Dague 1995). During thc settling phase, thepartial pressure of CO2 above the liquid zone is constant and inequilibrium with dissolved CO2, As a result, no signi ticant quantityof CO., is transferred to thc heudspace. a situation that contributesto the·establishmcnt of quiescent settling conditions. When thebiomass fonns a compact layer at the bottom of the reactor, thesupernatant is drawn to a predetcnnined level. usually at somedistance above the biomass bed. During efflucnt drawdown_microorganisms with poor scttling characteristics are also removedfrom the reactor, leaving behind the heavier bactclial noes (Sungand Dague 1995).
Thc advantages of the ASBR technology include low capitaland operating costs as well as minimum daily maintcnancc.
Removetreatedwastewater
Reactionperiod
Addwastewater
Clarify
Fill
Read
Settle
I I
Influent
----, I
Idle
k:;;:.;;J ~;:-" -. - "-=::.!-': -
Wastewater
LITERATURE REVIEW
Most laboratory studies on the anaerobic lreallncnl ofslaughterhouse wastewater have been conducted with anaerobicfilter reactors (AFRs) or upnow anaerobic sludge blanket(UASB) reactors. The AFR can sustain high organic loadingrates (OLRs) when the wastewater contains mostly solublechemical oxygen demand (SCaD) (Aror:! and ROUlh 1980:Borja et aJ. 1993, 1995a). Borja et aJ. (1994a) reponed a 94.5%COD reduction at an operating temperature of 35·C. an OLR of10.1 kg m-) dol and a hydraulic retention time (HRT) of 12 h.However, raw slaughterhouse wastewater contains highconcentrations of insoluble, slowly biodegradable solids, ortenrepresenting over 50% of the polluting ch:lrge (Masse c[ Masse2000; Sayed et de Zeeuw 1988). Ruiz et aJ. (1997) used an AFRoperated at 3rC to treat slaughterhouse wastewater cOI1lainingbelween 15 and 30% of ils COD as SS. When the OLR wasincreased above 3 kg m-l dol, COD reduction dropped below65% and effluent 55 concentration reached 1000 mgIL. Trill(1992) used an AFR to trcat a slaughterhouse Wilstcw3ter thatcontained 46% of its COD as SS. Reduction in COD rangedfrom 80% at OLRs below 2.5 kg m-l d-I to 30% al 18 kg m-3 dol.Treating the same wastewater after a two-hour scnling periodimproved COD reduction by an additional 10 to 15%. Saxenael al. (1986) treated a slaughterhouse wastewater that containedaboul 1000 mg SS/L in an AFR operaled at 25"C. At OLRsexceeding 2 kg COD 11,-3 d- l , fat and 55 deposition in thebioreaclOr caused a rapid deterioration of the biomass.
upnow anaerobic sludge blanket reactors arc also efficient intreating low 55 slaughterhouse wasteWaicr. At 35°C, COD wasreduced by morc than 90% at OLRs up (0 6.4 kg 111'3 dol (BOIja elal. I994b). AI 25"C, COD was reduced by 78% at an OLR of 6.1kg m'] dol (Zheng and Wu 1985). Lowering the operatingtemperature to 13
Q
C still pennitlcd a 75% COD removal at all
OLR of 3.3 kg m-3 d- l, as long as the HRT was maintained abovc
10 h. Howevcr, at 55 concclllrmions ranging from 15 to 30% ofinfluenl COD, total COD (TCOD) reduction decreased to 70% andSS losses reached 2000 IllgIL, at an OLR above 5 kg Ill" d·1and anoperating temperature of 37"C (Ruiz et al. 1997). Sayed et al.(1984. 1987, 1988) investigated the effect of SS concentration onthe anaerobic digestion of slaughterhouse wastewater in UA5Breaclors operated at 20 and 30"C. Results indicated that a ponionof thc removed COD was not converted into methanc bur waseliminated by other means such as nocculation and adsorption ofcolloidals on sludge panicles and cmrapment of coarse 55 in thesludge blanket. Weekend feed intemtptions pennitlcd a partialdigestion of retained solids (Sayed el al. 1984, 1987). However.when the OLR was increased from 0.5 to 10-20 kg COD m-3 dolover 57 days of continuous loading at 30
Q
C, a suddcn and heavysludge llotation led to the completc loss of biomass from thereaclor (Sayed el de Zeeuw 1988). Process failure lVas allribuledto an excessive accumulation of substratc materinl within thebiomass bed. Sayed el al. (1988) concluded that the controllingfactor in the digestion of unsettled slaughterhouse wastewater wasthe liquefaction "Ite of the adsorbed and elllmpped solids. and Ihusoperating temperature was critical.
Anacrobic sequential batch reactor
The ASBR represents a fairly new design of high-rJte anaerobicsystcm. In ASBRs. four treatment phases (feed. react, scttle. anddraw) are accomplished sequentially in one vessel (Fig. I). Duringthe feed and rcact phases. the rei.1CtQr content is mixed to allow
132 MASSE and tvlA$SE
MATERIALS and METHODS
Fig. 2. Plan of laboratory scale sequencing batch reactor.
Table I. Characteristics of the experimental slaughter-house wastewater.
Parameters Days after ASBR inoculation(mgIL except pH)
0-24 25-58 59-94 95-157
Total COD 6908 9665 11500 9445Soluble COD 3449 4714 5490 4505
Total solids 4892 6098 7121 6119Volatile solids 3647 4864 5724 4779Suspended solids nd· 2135 2658 2900Volatile suspended nd 1936 2458 2546
solids
Total Kjeldahl nitrogen 534 619 735 617Protein 246 89 221 172
1800 3313 3213 2781
pH 7.0 7.4 7.1 6.7Alkalinity (as CaC03) 1056 667 972 889
not determined
Operating conditionsThe four 42-L plexiglas ASBRs used in the experiment areillustrated in Fig. 2. The reactors were located in a temperaturecontrolled room maintained at 30°C. They were batch-fed everytwo days. The OLR was progressively increased by augmentingthe volume of wastewater fed to the reactors from 2 to 13 L.During the react phase, the digester content was mixed for oneminute every five minutes by recirculating the biogas with dualhead air pumps with a maximum capacity of 22.5 Umin. Nitrogenwas injected in the headspace during drawdown. After effluentdrawdown, the sludge bed was 14 L.
Seed sludgeTwo reactors received 13 L of anaerobic granular sludge from amilk processing plant (MPP) UASB reactor and two reactorsreceived 13 L of anaerobic non-granulated biological sludge froma municipal wastewater treatment plant. The MPP sludge treateda substrate consisting mainly of proteins and fats. Its total andvolatile solid content was 7.4 and 3.2%, respectively. Themunicipal sludge treated the raw sludge from primary andsecondary clarifiers. It had a total and volatile solid contents of4.8and 2.7%, respectively. Although the MPP sludge had 50% moretotal solids than the municipal sludge, volatile content, which givesa better indication of active biomass, was only 19% higher in theMPP than municipal sludge.
Sampling and analysisAt each feeding, all four digesters received the same volume ofwastewater. Biogas production was monitored daily with wet cupgas meters. Biogas composition (methane, carbon dioxide,hydrogen sulphide, and nitrogen) was determined weekly by gaschromatography. Methane content was corrected for nitrogen,since N2 gas was used as a filler during drawdown. Effluentsamples were collected at a port situated 80 mm (5 L) above thesurface of the 14.L sludge bed. Approximately 250 to 500 mL wasallowed to flow out of the reactor before the sample was collected.The effluent was analysed for soluble COD (SCOD) and TCOD,
9 SkIdge sampang part. cillo usedfor WdgBwastage
10 t.tmoIQucr orQ)OInOtant~port
11 GaaClUb12 Gaameter
13 FeedortuOO
1 300 11m dam81Br pIaldgIas cIgaster
2 SlJdge bocIzcne
3 \QtaI:)l8 Wlkm8 zcne4 Head spcx:e zane6 Gall8Cb:UaIIDnme6 BIDgaa I8CKUa!Icn puT1)
7 tdUJntlnG8 B'IUJnt InG
Another important feature is that ASBRs allow batch as well assemi-continuous or intermittent feeding. At the slaughterhouse, thewastewater could be fed to the reactor as it is produced during the8-h working shift, thus eliminating the need for an equalising tankor recycling line. The main disadvantage is that biogas productionis not uniform, making it difficult to plan a biogas-use strategy(Masse et at. 1997). Another drawback is that limited controlstrategies and experimental data are available, especiallyconcerning the use of ASBRs for slaughterhouse wastewatertreatment. Morris et aI. (1998) treated slaughterhouse wastewaterin two 11.5 L ASBRs. The ASBRs were operated at 30°C andreactor content was mixed 30 severy 10 min. The HRT rangedfrom 18 to 36 h. The SCOD was reduced by over 90%, but TCODremoval declined from about 60% at an HRT of 36 h to 30% at 18h. Lower TCOD reduction probably reflected high SS losses dueto poor biomass settling, especially at low HRT. No other reportsfrom the use of ASBRs to treat slaughterhouse wastewater areavailable at this time.
SubstrateThe wastewater was collected at a hog slaughterhouse in StValerien, Quebec, approximately once a month, in 200-L barrels.At the slaughterhouse, the wastewater was screened to remove hairand solids larger than 1 mm. In the laboratory, the wastewater wasmixed, transferred to IQ-L jugs, and stored at -15°C. It was heatedto approximately 20°C before feeding the reactor. Wastewaterquality over the study period is given in Table I.
CANADIAN AGRICULTURAL ENGINEERING Vol. 42, No.3 July/August/September 2000 133
85
135 137 139days
83 days 84
d) OLA =1.46 kg m4 0'HRT=6days
b) OLA '" 4.93 kg m4 0'HRT= 2 days
-- municipal sludge
50
2 40
~ 30
j2010
021 82
.-60do
i 45.ep30
:::E16
0109 133107 108
days
19 20days
c) OLR ... 4.39 kg m4 d"HRT=2days
a) OLA =0.74 kg m4 0 1
HRT=2days
-- milk processlng plant sludge
Fig. 4. Methane production during four cycles atdifferent times over the experimental period.
was generally higher in reactors seeded with MPP than municipalsludge. Over the whole start-up period, SS concentration averaged1400 mg/L in the former and 900 mg/L in the latter. The MPPsludge consisted of large but light granules, which were of thefilamentous type described by Lettinga et al. (1987). Thesespherical granules can have a diameter up to 5 mm in size. Theyare mainly composed of intertwined filamentous organisms andthey fall apart easily. Under mixing in the ASBRs, the granulesslowly disintegrated into small bacterial flocs and free floatingfilamentous microorganisms left the reactors during drawdown.Biomass loss thus contributed to high effluent SS concentrationsin the reactors seeded with MPP sludge. Total COD, which is thesum ofsoluble and particulate COD, also tended to be higher in theeffluent from reactors inoculated with MPP than municipal sludge(Fig. 3). It ranged from 3692 to 576 mgIL in the former and from2437 to 341 mgIL in the latter.
Start-up is the period during which the anaerobic bacteria arebeing acclimatized to new environmental conditions and substrate.A new equilibrium is slowly established between the variouspopulations of microorganisms, until the biomass can stably andefficiently degrade the substrate at maximum or target OLR. A 40day start-up was reported for a mesophilic (30 to 35°C) anaerobiccontact reactor treating slaughterhouse wastewater (Kostyshyn etala 1988). Borja et ala (1994a) also reported a 40-day start-up for anAFR treating slaughterhouse wastewatercontaining mainly solubleorganics. Methanol was mixed with the substrate during the firstmonth of start-up to encourage the proliferation of methanogens.With the ASBRs, the start-up period may possibly have beenshortened. Figure 4a shows methane production during a two-daycycle, 18 days after the beginning of start-up. Average methaneproduction rate declined from 8.5 Ud in the first 16 hours of thecycle to 2.4 Ud during the second day. The sharp decrease inbiogas production after 24 hours indicated that most organics weredegraded within one day and the OLR could have been increasedmore rapidly. After 25 days of operation, the OLR was graduallyincreased from 1.04 to 3.29 kg m-3 d-I over a 20-day period withoutdetrimental effect on effluent SCOD (Fig. 3). However, higher gasproduction at the end of the react phase created turbulentconditions during settling, which resulted in a temporary increase
50
2 40
1=10
160 0106
160
••••••munlclpaJ sludge
operation ~I
50
402CD 30
Organic loadlng rate 12010
080 100 120 140 160 18
days
60 80 100 120 140 160days
4020o
--milk proc. plant sludge
20 40 60 80. 100 120 140I days
2260j\)'!~~!Total COD". : ~ !'& 1600 ..... : ". iE • ~ I ••~,•••
760 ~~~••••••• : ~
~~ ----o , ~-,--..,....-.....,....-.....,
o 20 40 160 80 100 120 140I days
1000
760
~600E
250
6
\, 4'?E 3
8 2
~
1\J:. I Suspended solids1600 :~ I'& 1000 : \ I
E 600 •••••• .... 1
, -.o , , ,
RESULTS and DISCUSSION
Start-upVolumetric OLR (kg TCOD/m3 of sludge per day of cycle) andeffluent SCOD, TCOD, and SS concentrations during theexperimental period are presented in Fig. 3. During start-up, theOLR was progressively increased from 0.49 to 3.29 kg m-3 d-I overa 55-day period. Effluent SCOD decreased from over 1000 mgILin the first few days of start-up to 281 and 155 mgIL on day 28, inreactors seeded with MPP and municipal sludge, respectively.Soluble COD remained well below 250 mgIL in all reactor effluentduring the rest of the start-up period. Effluent SS concentration
Fig. 3. Volumetric organic loading rate and quality ofemuent from four anaerobic sequencing batchreactors treating slaughterhouse wastewater at30°C.
solid content, ammonia-nitrogen (NH4-N), total Kjeldahl nitrogen(TKN), volatile fatty acid (VFA) concentration, pH, and alkalinity.Protein concentration was calculated by multiplying the differencebetween TKN and NH4-N by 6.25 (AOAC 1984). Analyses weredone according to methods outlined in APHA (1992). Volatilefatty acid concentration was determined by gas chromatography.
134 MASSE and MASSE
On day 92, the OLR was reduced from 4.93 to 4.39 kg m·3 d".However, the wastewater contained a higher fraction ofundissolved solids (Table n, and the amount of SS fed to thedigesters was increased from 16.0 to 18.8 gld. Over the next twoweeks, effluent SCOD fluctuated between 331 and 645 mgIL (Fig.3). Figure 4c presents methane production during one cycle at thehigher SS loading rate. During the last six hours of the cycle,methane was still produced at a rate of lOUd. High biogasproduction maintained turbulent conditions in the reactors duringthe settling phase, and unsettled biomass was removed with theeffluent. Average effluent SS concentration increased from 215 to460 mgIL. After three weeks of feeding with the high SSwastewater, effluent SCOD concentration had decreased below300 mg/L in every reactor, but SS concentration was stillfluctuating between 150 and 500 mgIL.
After 133 days of continuous operation, the HRT wasincreased to six days for one cycle. Methane was still produced ata rate of 5.5 Ud on day 4 and 2.2 Ud on day 6 (Fig. 4d). Thereactors produced approximately 60 L of methane over the six-daycycle, for an average of0.49 L of methanelg ofCOD fed, while thetheoretical maximum methane yield at 30°C is 0.39 Ug of CODremoved. Continued methane production above the theoreticalmaximum yield was probably caused by the degradation of the SSthat had settled with the microorganisms at the end of each cycleand accumulated within the biomass bed. This slow accumulationof SS in the reactors suggests that the operational strategy shouldinclude occasional feed interruptions to allow the completedegradation of the undissolved solids, even though the ASBRprocess can lower COD concentrations to acceptable values atOLR of 4.93 kg m·3 dol.
Reduction in soluble, particulate, and total COD Figure 5presents average reduction in SCOD, TCOD, and SSconcentration at OLRs ranging from 1.73 to 4.93 kg m-3 d".Soluble COD reduction decreased slightly with increased OLR,but it remained between 94 and 98% over the whole period.Reduction in S5 increased with OLR. As more 55 was loaded intothe reactors, a greater proportion of the 55 was probably removedby biological degradation as well as physical separation duringsettling. However, low OLRs also coincided with earlier periods,during which biomass losses were the largest. Total COD wasreduced by 90 to 96% with no obvious trend with respect to OLRrate. A TCOD reduction of 82% at an OLR of 1.73 kg m-3 dol wascaused by high 55 losses during start-up (day 26 to 32).
Methane and biomass yield The biogas averaged 75.0%methane, 24.6% carbon dioxide and 0.4% hydrogen sulphide.Figure 6 presents the COD removed via methane production (2.57kg COD per m3 of methane at 30°C) and total COD removed(TCODinllucnt- 5CODeffiucn,) during the experiment. The differencebetween the two curves is a measure of volatile suspended solid(V5S) accumulation, i.e. apparent bacterial yield (true biomassyield minus decay) as well as undegraded V55 from the substrate.Over the experimental period, 90.5% of the COD removed wastransformed into methane. This high degree of methanizationindicated that most soluble and particulate COD fed to the A5BRswas effectively degraded within the HRT. Solid accumulationrepresented 0.095 g of COD/g of COD removed, or 0.068 g ofVSS/g of COD removed (based on 1.41 g COD per g V55). BOIjaet al. (1995b) calculated a true biomass yield of 0.07 g V55/gCOD removed in AFRs treating slaughterhouse wastewater at35°C. Under similar conditions, Metzner and Temper (1990)
Soluble COD
c 98 •~ •;:,
i 95~0
•92
1 2 3 4 5OlA (kg COD m-3 d'1)
100 Total COD
c 95 • • • •0 •15 •;:, 90 •'C!(/. 85
801 2 3 4 5
OlA (kg COD m-3 cf1)
100 Suspended solids
c •0 8013::I'Ce 60(/.
•401 2 3 4 5
alA (kg COD m-3 cf1)
Fig. S. Reduction in soluble COD, total COD, and sus-pended solids with respect to organic loading rates.
in effluent SS concentration to 3498 and 1792 mgIL in reactorsseeded with MPP and municipal sludge, respectively. However, atthe end of the start-up period, SS concentration had decreased to370 and 191 mgIL in effluent from both types of reactors,respectively.
OperationEmuent SCOD, TCOD, and SS concentration Between day 55and 90, the OLR was increased from 3.29 to 4.93 kg m-3 dol(Fig. 3). Effluent SCOD remained below 300 mgIL in mosteffluent samples. Average SS concentration was 236 and 162mgIL and TCOD concentration averaged 703 and 481 mg/L inreactors seeded with MPP and municipal sludge, respectively. TheASBRs inoculated with municipal sludge were still performingslightly better than the ones seeded with MPP sludge, butdifferences between reactor progressively decreased.
Figure 4b shows methane production during a two-day cycleat an OLR of4.93 kg m-3 dol. Methane production rate averaged 32Ud in the first 18 h of the cycle and 8.5 Ud in the last 5 h.Although effluent SCOD was 167 mg/L at the end of the cycle,biogas production remained important, indicating that particulateswere still being hydrolysed and methanized. Low VFAsconcentration in reactor effluent throughout the experimentalperiod, ranging from trace amounts to 100 mg/L, also suggestedthat methanization of the SCOD was rapid compared to particulatehydrolysis.
CANADIAN AGRICULTURAL ENGINEERING Vol. 42. No.3 July/August/September 2000 135
CONCLUSION
buffer (Fanin 1987). Effluent pH also increasedslightly during treatment, and ranged from 7.4 to7.9.
REFERENCES
AOAC. 1984. Official Methods ofAnalysis ofthe Association ofOfficial Analytical Chemists. Arlington, VA: Association ofOfficial Analytical Chemists.
APHA.1992.StandardMethodsfortheExaminationofWaterandWastewater. Washington, DC: American Public HealthAssociation.
Arora, H.C. and T. Routh. 1980. Treatment of slaughterhouseeffluents by anaerobic contact filter.1AWPC Technical AnnualVI & VII: 67-78.
BOlja, R., C.J. Banks and Z. Wang. 1994a. Performance andkinetics ofan upflow anaerobic sludge blanket (UASB) reactortreating slaughterhouse wastewater. Journal ofEnvironmentalScience and Health A29:2063-2085.
BOlja, R., C.J. Banks and Z. Wang. 1994b. Stability andperformance of an anaerobic downflow filter treatingslaughterhouse wastewater under transient changes in processparameters. Biotechnology and Applied Biochemistry 20:371383.
BOlja, R., C.J. Banks and Z. Wang. 1995a. Effect of organicloading rate on anaerobic treatment of slaughterhousewastewater in a fluidised-bed reactor. Bioresource Technology52: 157-162.
BOlja, R., C.J. Banks and Z. Wang. 1995b. Performance of ahybrid anaerobic reactor, combining a sludge blanket and afilter, treating slaughterhouse wastewater. AppliedMicrobiology and Biotechnology 43:351-357.
Borja, R., M.M. Duran and A. Martin. 1993. Influence of thesupport on the kinetics of anaerobic purification ofslaughterhouse wastewater. Bioresource Technology 44:57-60.
Dague, R.R., C.E. Habben and S.R. Pipaparti. 1992. Initial studieson the anaerobic sequencing batch reactor. Water Science andTechnology 26:2429-2432.
days
_·_-CH4-COO
-coo removed
6000
5000
S4000
83000
2000 Slaughterhouse wastewater containing between1000 6908 and 11 500 mgIL of TCOD was treated in
o+-"'-::;:~...---__----r---.........----r--~-----, four 42-L ASBRs operated at 30°C. Total COD20 40 60 80 100 120 140 160 was reduced by 90 to 96% at OLRs ranging from
2.07 to 4.93 kg mo3 d-I and an HRT of two days.Soluble COD was reduced by over 95%. Resultsindicated that at a steady state operation of 4.93kg m-3 d- I, the reactors will be stable in terms ofeffluent quality and biomass retention. The two
reactors seeded with anaerobic sludge from a municipal treatmentplant originally performed better than the two reactors inoculatedwith a sludge from a milk processing plant, but difference betweensludges decreased with time. The reactors produced a biogascontaining 75% methane. About 90.5% of the COD removed weremethanized. Solid accumulation, including apparent biomass yieldand undegraded SS, was evaluated at 0.068 g VSS per g CODremoved. The high degree of methanization suggested that mostsoluble and suspended organics in slaughterhouse wastewaterweredegraded during treatment in the ASBRs.
Fig. 6. Cumulative gas production (CH4-COD) and COD removed duringthe experimental period.
obtained a net accumulation ranging between 0.05 and 0.07 gVSS/g COD removed.
Over the experimental period, however, there was a net loss ofbiomass from the reactors. At the beginning of the experiment,total solid content in the sludge bed averaged 75 400 and 48 900mgIL in the reactors seeded with MPP and municipal sludge,respectively. After 139 days of operation, total solid content in thesludge bed averaged 18 700 mgIL in all reactors. Biomass loss wasmostly observed during start-up when large amounts of sludge leftthe reactors with the effluent. However, sludge quality in terms ofVSS content improved. Volatile solids represented 69% of totalsolids at the end of the experiment, while it originally accountedfor 43 and 55% of total solids in the MPP and municipal sludge,respectively.Nitrogen and protein In the raw slaughterhouse wastewater,TKN ranged from 534 to 735 mg/L (Table I). In effluent, it variedbetween 473 and 808 mgIL. During the first month of start-up,total nitrogen was higher in effluent than influent, probablybecause of the large losses of bacteria, which contain 10 to 15% oftheir weight as nitrogen (Grady and Lim 1980). Afterwards,nitrogen reduction during treatment averaged 13% in all reactors.Bacterial synthesis is the main sink for influent nitrogen. The lowgrowth rate and low yield ofanaerobic bacteria translates into lowoverall nitrogen removal.
Ammonia-N represented from 14 to 46% of total nitrogen ininfluent, and from 75 to 98% of total nitrogen in effluent. Theproportion ofammonia-N in effluent was lower at the beginning ofthe experimental period, due to low degradation rate of influentCOD and large biomass losses in effluent. During that period,effluent protein content reached 1100 and 620 mgIL in reactorsseeded with MPP and municipal sludge, respectively. Thereafter,protein concentration in effluent averaged 241 and 139 mgIL forboth types of sludge, respectively, corresponding to a proteinreduction of 92 and 95%, respectively, due to mineralization oforganic nitrogen.Alkalinity and pH Influent alkalinity ranged from 667 to 1056mgIL as CaC03, and pH from 6.7 to 7.4 (Table I). During the firsttwo weeks ofthe experiment, effluentalkalinity was approximately2100 and 3200 mgIL in reactors seeded with MPP and municipalsludge, respectively. During the remainder of the experimentalperiod, alkalinity averaged 2550 mgIL in all reactors,corresponding to an increase of over 200% in alkalinity duringtreatment. The increase in alkalinity was mainly caused by themineralization of protein into ammonia. The latter combines withthe carbonic acid in solution to form an ammonium bicarbonate
136 MASSE and MASSE
Fannin, K.F. 1987. Start-up, operation, stability, and control. InAnaerobic Digestion ofBiomass, eds. D.P. Chynoweth and RIsaacson, 171-196. London, England: Elsevier AppliedSciences.
Grady C.P. and H.C. Lim. 1980. Biological WastewaterTreatment. New York, NY: Marcel Dekker Inc.
Kostyshyn, C.R, W.A. Bonkoski and J.E. Sointio. 1988.Anaerobic treatment of a beef processing plant wastewater: Acase history. In Proceedings of the 42'u' Industrial WasteConference, 673-692. Ann Arbor, MI: Ann Arbor Science.
Lettinga, G., W. de Zeeuw, W. Wiegant and L. Holshoff Pol.1987. High-rate anaerobic granular sludge UASB-reactors forwastewater treatment. In Bioenvironmental Systems, I, ed. D.L.Wise, 132-159.
Masse, DJ. 1995. Psychrophilic anaerobic digestion of swinemanure slurry in intermittently fed sequencing batch reactors.Ph.D. thesis. Ottawa, ON: University of Ottawa.
Masse, DJ. and F. Croteau. 1998. La digestion anaerobiepsychrophile du lisier de porc a I' interieur de bioreacteurs aoperations sequentielles. Final report presented to theFederation des producteurs de porcs du Quebec. Agricultureand Agri-Food Canada, Lennoxville, QC.
Masse, 0.1. and R.L. Droste. 1997. Microbial interaction duringanaerobic treatment of swine manure slurry in a sequencingbatch reactor. Canadian Agricultural Engineering 39:35-41.
Masse, 0.1., RL. Droste, KJ. Kennedy, N.K. Patni and J.A.Munroe. 1997. Potential for the psychrophilic anaerobictreatment of swine manure using a sequencing batch reactor.Canadian Agricultural Engineering 39:25-33.
Masse, 0.1. and L. Masse. 2000. Characterization of wastewaterfrom hog slaughterhouses in Eastern Canada and evaluation oftheir in-plant wastewater treatment system. CanadianAgricultural Engineering 42: 139-146.
Masse, 0.1., N.K. Patni, R.L. Droste and KJ. Kennedy. 1996.Operation strategies for psychrophilic anaerobic digestion ofswine manure slurry in sequencing batch reactors. CanadianJournal ofCivil Engineering 23: 1285-1294.
Metzner, G. and U. Temper. 1990. Operation and optimization ofa full-scale fixed-bed reactor for anaerobic digestion of animalrendering wastewater. Water Science and Technology 22:373384.
Morris D., S. Sung and R.R Dague. 1998. ASBR treatment ofbeef slaughterhouse wastewater. Downloaded from theinternet (http://ce.ecn.purdue. edu.l-alleman/w3piwc/paperslsung.html).
Ruiz, 1, M.C. Veiga, P. de Santiago and R. Blazquez. 1997.Treatment of slaughterhouse wastewater in a UASB 'reactorand an anaerobic filter. Bioresource Technology 60:251-258.
Saxena, K.L., S.N. Kaul, M.Z. Hasan, S.K. Gadkari and S.D.Badrinath. 1986. Packed bed anaerobic reactor for treatment ofmeat wastes. Asian Environment 8:20-24.
Sayed, S.K.l, L. van Campen and G. Lettinga. 1987. Anaerobictreatment of slaughterhouse waste using a granular sludgeUASB reactor. Biological Wastes 21:11-28.
Sayed, S.K.l, 1. van der Zanden, R Wijffels and G. Lettinga.1988. Anaerobic degradation of the various fractions ofslaughterhouse wastewater. Biological Wastes 23: 117-142.
Sayed, S.KJ. and W. de Zeeuw. 1988. The performance of acontinuously operated flocculent sludge UASB reactor withslaughterhouse wastewater. Biological Wastes 24:199-212.
Sayed, S.K.l, W. de Zeeuw and G. Lettinga. 1984. Anaerobictreatment of slaughterhouse waste using a flocculent sludgeUASB reactor. Agricultural Wastes 11:197-226.
Sung, S. and RR. Dague. 1995. Laboratory studies on theanaerobic sequencing batch reactor. Water EnvironmentResearch 67:294-301.
Tritt, W.P. 1992. The anaerobic treatment of slaughterhousewastewater in fixed-bed reactors. Bioresource Technology41:201-207.
Zheng, Y. and W. Wu. 1985. A study of meat packing plantwastewater treatment with upflow anaerobic sludge blanketprocess. In Anaerobic Digestion 1985, 327-337. Proceedingsof the 4th International Symposium on Anaerobic Digestion.Guangshou, China.
CANADIAN AGRICULTURAL ENGINEERING Vol. 42, No.3 July/August/September 2000 137
Characterization of wastewaterfrom hog slaughterhouses in Eastern Canada
and evaluation of their in-plantwastewater treatment systems
DJ. MASSE and L. MASSE
Agriculture and Agri-Food Canada, P.O. Box 90,2000 Route 108 East, Lennoxville, QC, Canada J1M 1Z3. Agriculture and AgriFood Canada contribution No. 660. Received 8 July 1999; accepted 6 June 2000.
Masse, DJ. and Masse, L. 2000. Characterization of wastewaterfrom hog slaughterhouses in Eastern Canada and evaluation oftheir in-plant wastewater treatment systems. Can. Agric. Eng.42: 139-146. Literature on existing systems for slaughterhousewastewater treatment was reviewed and discussed in terms oftechnology usefulness and relevance under Canadian conditions. Thewastewater from six hog slaughterhouses in Quebec and Ontario wasalso characterised before and after treatment at the plant. In rawwastewater, total chemical oxygen demand (TCOD) ranged from 2333to 8627 mgIL and suspended solids (SS) varied between 736 and 2099mg/L. Slaughterhouse wastewater composition in terms of organicstrength, inorganic elements, alkalinity, and pH is adequate forbiological treatment. Two slaughterhouses only settled theirwastewater before discharging it to the municipal sewer. Three plantsused primary treatment to precipitate blood and remove floating fat,while one further treated its wastewater using an aerobic tricklingfilter. Although preliminary treatment at the slaughterhouse reducedthe level of pollutants, TCOD and SS concentrations were still toohigh for sewer discharge without being imposed a municipalsurcharge. In addition, all treatments produced large amounts ofputrefactive and bulky sludge, which required special handling and/orfurther treatment.
Cet article presente d'abord une recension des ecrits sur lessystemes de traitement des eaux usees d'abattoir. Les technologiesd'epuration sont decrites en fonction de leur utilite dans Ie contextecanadien. Les eaux usees provenant de six abattoirs de pores auQuebec et en Ontario ont ensuite ete caractt~risees avant et apres Ietraitement effectue aux abattoirs. Dans les eaux usees brutes, lademande chimique en oxygene totale (DCOT) varie de 2 333 a8 627mg/l, et les matieres en suspension (MES), de 736 a2 099 mg/1. Leseaux usees d'abattoir contiennent une charge organique, desconcentrations d'azote, de phosphore et d'oligo-etements, unealcalinite et un pH adequats pour un traitement biologique. Deux desabattoirs visites se contentent de transvider les eaux usees dans desbassins de decantation avant de les deverser dans Ie reseau municipald'egouts. Trois abattoirs effectuent une epuration primaire dans Ie butde precipiter Ie sang et de favoriser la flottaison des gras, tandis qu' unabattoir traite aussi ses eaux usees a I'aide d'un filtre bacterienaerobie. Les traitements d'epuration utilises dans les abattoirsreduisent Ie niveau des matieres polluantes mais se revelentinsuffisants pour ramener la DCOT et les MES it des concentrationspermettant aux etablissements de deverser leurs eaux dans les egoutsmunicipaux sans se voir imposer une surtaxe. En outre, tous lestraitements recenses produisent une quantite considerable de bouesputrides et volumineuses qui necessitent une manutention speciale oudes traitements ulterieurs.
INTRODUCTION
Slaughterhouse wastewater is very harmful to the environment.Effluent discharge from slaughterhouses has caused thedeoxygenation of rivers (Quinn and Farlane 1989) and thecontamination of groundwater (Sangodoyin and Agbawhe1992). The pollution potential of meat-processing andslaughterhouse plants has been estimated at over 1 millionpopulation equivalent in the Netherlands (Sayed 1987), and 3million in France (Festino and Aubart 1986). Blood, one of themajor dissolved pollutants in slaughterhouse wastewater, has achemical oxygen demand (COD) of 375 000 mglL (Tritt andSchuchardt 1992). Slaughterhouse wastewater also containshigh concentrations of suspended solids (SS), including piecesof fat, grease, hair, feathers, flesh, manure, grit, and undigestedfeed (Bull et al. 1982). These insoluble and slowlybiodegradable SS represented 50% of the pollution charge inscreened (1 mm) slaughterhouse wastewater, while another 25%originated from colloidal solids (Sayed et al. 1988).
Slaughterhouse wastewater quality depends on a number offactors, namely:I. Blood capture: the efficiency in blood retention during
animal bleeding is considered to be the most importantmeasure for reducing biological oxygen demand (BOD)(Tritt and Schuchardt 1992);
2. Water usage: water economy usually translates intoincreased pollutant concentration, although total BOD masswill remain constant;
3. Type of animal slaughtered: BOD is higher in wastewaterfrom beef than hog slaughterhouses (Tritt and Schuchardt1992);
4. Amount of rendering or meat processing activities: plantsthat only slaughter animals produce a stronger wastewaterthan those also involve in rendering or meat processingactivities (Johns 1995).
Most slaughterhouse wastewater quality data have beengenerated in Europe (Bull et al. 1982; Sachon 1982, 1986;Sayed 1987; Tritt and Schuchardt 1992), Australia (Johns 1995)and the USA (Camin 1970) and little information exists onquality and treatment of slaughterhouse wastewater in Canada.The objective of this project was thus to characterise wastewater
CANADIAN AGRICULTURAL ENGINEERING Vol. 42, No.3 JulylAugust/September 2000 139
from hog slaughterhouses in Quebec and Ontario. In 1995 and1996, six hog slaughterhouses were visited. Data on wastewatertreatment and wastewater quality before and after in-planttreatment are presented. Literature on existing technologies forslaughterhouse wastewater treatment will first be brieflyreviewed and discussed in terms of the usefulness and relevanceof these technologies to Canadian conditions.
SLAUGHTERHOUSE WASTEWATER TREATMENT:A REVIEW
Sewer dischargeSewer discharge of wastewater without preliminary treatmentis mostly used by smaller slaughterhouses located close tomunicipal treatment plants. Domestic wastewater, generallymuch lower in BOD and inorganic nutrient concentration,dilutes the slaughterhouse wastewater and makes it moreamenable to biological treatment. The main disadvantage ofsewer discharge is the surcharge imposed by municipalities totreat the wastewater. In addition, few municipal treatment plantswill accept large quantities of untreated slaughterhousewastewater.
Land applicationLand application of slaughterhouse wastewater by sprayirrigation has been mainly used in the USA (Bull et al. 1982).The advantages of the system are its simplicity and low cost.The disadvantages include possible surface and ground watercontamination, odour problems, greenhouse gas emission, andsoil pore clogging from excessive fat loads. Application onconstructed wetlands could also be used as a polishingtreatment for biologically treated wastewater (John 1995). Landapplication, however, is not practical in subfreezingtemperatures, and in most parts of Canada large volumes ofwastewater would have to be stored during the winter months.
Physico-chemical treatmentsGrit chambers, screens, settling tanks, and dissolved airflotation (DAF) units are widely used for the removal of SS,colloidals, and fats from slaughterhouse wastewater. In DAFunits, air bubbles injected at the bottom of the tank transportlight solids and hydrophobic material, such as fat and grease, tothe surface where scum is periodically skimmed off. Camin(1970) surveyed wastewater treatment in over 200 meat packingplants in the USA and concluded that, compared to aerobic andanaerobic systems, air flotation was the least efficient treatmentin terms of dollars per weight of BOD removed.
Blood coagulants (e.g. aluminium sulphate and ferricchloride) and/or flocculents (e.g. polymers) are sometimesadded to the wastewater in the DAF unit to increase proteinflocculation and precipitation as well as fat flotation. ChemicalOAF units can achieve COD reduction ranging from 32 to 90%,and are capable of removing large amounts of nutrients (Johns1995). However, operational problems have been reported, andthe system produces large volumes of putrefactive and bulkysludge that requires special handling and further treatment(Johns 1995).
Aerobic treatmentIn aerobic digestion, microorganisms degrade organics in thepresence of oxygen. Belanger et aI. (1986) described theoperation of a 1000 m3 aerobic lagoon treating slaughterhouse
140
wastewater in southwestern Quebec. Twenty-four submergedemitters transferred 850 Htres of oxygen per minute. InfluentBODs ranged from 1500 to 3000 mgIL and hydraulic retentiontime (HRT) averaged II days. Effluent BODs concentration wasgenerally below 50 mglL, except in winter when it reached 645mgIL and remained high for almost two months due to coldconditions in the lagoon. The system required daily maintenanceby a trained technician and daily drainage of accumulatedsludge. One disadvantage of aerobic systems is the generationof large quantities of biological sludge that must be treatedbefore disposal.
Besides lagoons, extended aeration systems and tricklingfilters have been the most popular aerobic processes for thetreatment of meat packing and slaughterhouse wastewater (Bullet al. 1982). High BOD removals are reported but effluent SSconcentrations are often elevated due to poor sludge settleability(Johns 1995). In addition, oxygen requirements and treatmenttime increase steeply with wastewater strength. For this reason,aerobic digestion is considered less economical than anaerobictreatment for wastewaters with COD concentrations above 4000mgIL, and with the development of high-rate anaerobic reactorsthe cut-off level may be lower than 4000 mgIL (Rudd 1985).Aerobic systems, however, could be used for final purificationand nutrient removal, following physico-chemical or anaerobictreatment, wherever slaughterhouses must treat their wastewaterto river discharge standards.
Anaerobic treatmentDuring anaerobic digestion, organics are degraded by a diversityof bacteria into methane in the absence of oxygen. Anaerobicsystems are not used in Canada but they represent an interestingalternative for treatment at the plant. Their advantages are:I. a high efficiency in reducing COD in soluble and insoluble
form;2. a low sludge production of only 5% to 20% of that
generated by aerobic systems (Speece 1996);3. the recovery of usable energy in the form of methane;4. no aeration energy requirement;5. no chemical handling;6. the biomass can remain unfed for long periods without
deteriorating.
Anaerobic treatment can be divided into two maincategories, low-rate (lagoons) and high-rate systems.
Anaerobic lagoons Anaerobic lagoons have been one of themost extensively used systems for treating slaughterhousewastewater in the USA and Australia, where climatic conditionsand land availability permitt the construction of large lagoons(Johns 1995; Rollag and Dornbuh 1966). Low capital,operational, and maintenance costs combined with a highefficiency in reducing polluting charges have all contributed tothe popularity of lagoons. The disadvantages of lagoons includethe large area requirement, odor problems, and the emission ofmethane, one of the major contributors to greenhouse gas, witha heat-trapping capacity 20 to 30 times that of carbon dioxide.
Odor and gas emissions can be contained by coveringlagoons. Dague et al. (1990) described the operation of a largecovered anaerobic lagoon treating hog slaughterhousewastewater. Influent BODs ranged from 1600 to 4800 mgIL andthe HRT was 13 days. Reduction in BODs and SS averaged87% and 81 %, respectively. A methane production of 0.51 m3
MASSE and MASSE
per kg of BODs removed was obtained. However, coveredlagoons require high BOD loading to generate economicquantities of biogas (Safley and Westerman 1988, 1992). InCanada, the construction of a lagoon cover with enoughdurability and strength to resist large unbalanced forces due towind, ice, and snow accumulation would be very costly. Inaddition, liquid temperature would be excessively low in winter.In southwestern Quebec, the average temperature of an aerobiclagoon varied between 0 and 8.5°C during the winter months(Belanger et a1. 1986). The efficiency of anaerobic lagoons isgreatly reduced below 21°C (Hammer and Jacobson 1970). Inaddition, anaerobic bacteria are sensitive to rapid changes intemperature, and it would be almost impossible to restart a largeanaerobic lagoon if it failed during a cold period.
High-rate anaerobic reactor More sophisticated anaerobicsystems were developed to accelerate treatment and reduce arearequirement, especially in places where land is expensive andscarce, such as Europe and Asia. Area requirement is also animportant factor in cold climatic conditions where wastewatertreatment has to be applied indoors. The first high-rateanaerobic design, the anaerobic contact reactor (ACR),basically consisted ofa stirred tank reactor followed by a sludgeseparator. The first report of a full-scale ACR treatingslaughterhouse wastewater was from the United Kingdom(Black et al. 1974). The reactor was operated at 32.5°C andreceived pre-settled wastewater at organic loading rates (OLRs)ranging from 0.12 to 0.28 kg m-3d -I. Reduction in BODs wasapproximately 90%. However, because of technical problemswith the clarifiers, the effluent contained high concentrations ofbiomass, and volatile solids reduction ranged between 41 and67%. A mesophilic ACR (30 to 35°C) was also built in a meatpacking plant in the USA (Kostyshyn et al. 1988). Wastewaterwas first pre-treated in a OAF unit and average COD and SSconcentrations in influent were 6320 and 2342 mgIL,respectively. During the first six months of operation, COD andSS reduction averaged 85 and 75%, respectively, at OLRsbetween 2 and 3 kg m-3d -I and HRTs between 1.7 and 2.5 days.However, the plant operator reported malfunctionning of theclarifiers (Personal communication: Packerland Packing Co.1996). Poor biomass settleability appears to be a recurrentproblem with ACRs.
Most modern high-rate anaerobic reactors have built-indevices to retain bacteria. In anaerobic filter reactors (AFRs),retention is achieved through biomass adhesion to a fixed orfloating inert material called filter. Anaerobic filter reactors aregenerally robust and resistant to shock loads, but carriermaterial is expensive and some designs require intensesupervision (Defour et al. 1986). Filter material can alsobecome clogged by high concentrations of undissolvedorganics. Campos et a1. (1986) described the operation of anindustrial anaerobic filter reactor (AFR) treating meatprocessing wastewater at 25°C over a 6-year period. At an OLRof 1.4 kg m-3d -I, COD reduction was 76 and 85% at an HRT of13 and 24 h, respectively. Influent SS concentration (889 mgIL)was reduced by 88%. A full-scale AFR was also constructed ina German rendering plant (Metzner et al. 1990). Primarytreatment included a grease separator, a mud-trap, and a 0.6-mmdrum screen. The AFR was operated at 36°C at OLRs between3 and 10 kg m-3d'( and HRTs between 21 and 27 h. Reductionin COD ranged from 70 to 90%.
In upflow anaerobic sludge blanket (VASB) reactors,influent enters at the bottom of the digester, flows across acompact layer of bacteria (the sludge blanket) and exits at thetop of the reactor. Successful operation depends on theformation of bacterial flocs or granules that accumulate andeasily settle at the bottom of the digester. Reactor operationrequires close supervision. Liquid velocity must be low enoughto prevent excessive lifting of the sludge blanket andequalisation tanks must be used to prevent strong variations inorganic loading (Defour et al. 1994). Full-scale VASB reactorsfor slaughterhouse wastewater treatment were installed in theNetherlands, Belgium, and New Zealand, but operating data areunavailable except that in the Netherlands, granules could notbe obtained, and the UASB had to be operated as a flocculentsystem (Johns 1995). The VASB can operate with flocculentsludge, although granules tend to settle better and thus allowincreased flow rate.
The anaerobic sequencing batch reactor (ASBR), asdeveloped by Agriculture and Agri-Food Canada, couldrepresent an economical, stable, efficient, easy-to-use, and easyto-operate process to treat and recover usable energy fromslaughterhouse wastewater. This new technology, which hasbeen successfully applied at laboratory and semi-commercialscales for the treatment of swine manure slurry, can operatewith limited capital costs, energy, and manpower (Masse 1995;Masse and Croteau 1998; Masse and Droste 1997; Masse et al.1996, 1997). It has only been tested on a laboratory-scale for thetreatment of slaughterhouse wastewater but preliminary resultsare encouraging (Masse and Masse 2000).
MATERIALS and METHODS
Between June 1995 and January 1996, three Quebec and threeOntario slaughterhouses dealing exclusively in pork meat werevisited. The slaughterhouses varied in size and included smallfamily plants as well as large corporations. In this paper, theslaughterhouses are referred to by a number from 1 to 6.Slaughterhouse I had the lowest capacity, while Slaughterhouse6 had the largest (Table I). Three of the plants only dealt inslaughtering, while the other three were also involved in somerendering or meat processing activities. Their wastewatertreatment systems were examined and information was collectedon plant capacity, waste disposal, water usage, and treatmentsystem operation and cost. Wastewater volume per animal killedwas estimated by dividing wastewater production, as evaluatedby the slaughterhouse, by the average number of animals killedduring the same period.
At all slaughterhouses, samples were collected in themorning or early afternoon. The untreated (raw) wastewater wassampled after the screening or settling of coarser solids. Screensand primary settling tanks are usually located at the inlet ofwastewater treatment areas, and it is difficult to sample beforethat point. Raw wastewater samples did not include washwateror wastewater from the scalding tank. At Slaughterhouses 1 and2, one sample was collected from the holding tank, where thewastewater is sent before being discharge to the municipalsewer. At Slaughterhouses 4 and 6, one sample was collectedfrom the settling tank and one at the outlet of the OAF unit. AtSlaughterhouse 3, one sample was collected at each of fourplaces in the treatment room: (1) after the I-mm screen, (2) atthe outlet of the DAF unit, (3) at the outlet of the aerobictrickling filter, and (4) at the final discharge pipe. At
CANADIAN AGRICULTURAL ENGINEERING Vol. 42. No.3 July/August/September 2000 141
Table I. Capacity, water usage, and wastewater treatment at six pork slaughterhouses(1995-1996).
(I)
9000030000
NA650000100000
NA
Treatment cost($/yr)#
=coefficient based on quantity of wastewaterproduced and treatment cost at the municipal plant,
=(total suspended solids - 350)/350, where 350mglL is the maximum level allowed,
Surcharge = K[05TSS +05BOD+0.2FOG +
0.2N +O.lP]
SS
where:K
a 10 to 30% loss betweenwater input and wastewateroutput (data supplied by theplant). However, theseslaughterhouses remainfairly efficient waterconsumers, since waterconsumption varyingbetween 200 and 600 L perhog slaughtered wasreported for Germanslaughterhouses (Tritt andSchuchardt 1992). The threeslaughterhouses combiningslaughtering and meatprocessing activities utiliseda higher volume of water,especially Slaughterhouse 4which included a large
processing plant and thus averaged 1250 L per head killed andprocessed.
The two smaller slaughterhouses (1 and 2) did not treat theirwastewater but stored it for a short period of time in a holdingtank, where coarse SS were allowed to settle and floating fatwas periodically skimmed off. All other plants had screens orsettling tanks for the removal of coarser particles and DAF unitsfor fat and light particle recovery. Slaughterhouse 4 also addedferric chloride and sulfuric acid to the wastewater as it flowedinto the DAF unit to induce blood precipitation and flocculation.The pH of the effluent from the chemical-DAF unit wasadjusted from 2.7 to 5.5 by mixing it with the wastewater fromthe meat processing plant. Slaughterhouse 5 used ferric chlorideto precipitate proteins and a polymer to flocculate colloidals andincrease sedimentation. The sludge captured at the top andbottom of the DAF unit was mixed with lime to control pH andsmell and shipped out daily for storage or land application,depending on the time of the year. Slaughterhouse 3 usedbiological aerobic trickling filters to further treat the effluentfrom the DAF unit. To increase treatment time, microorganismswere added to the holding tank, but the bacteria were killed bylack of oxygen. The plant operator was planning to installemitters to diffuse oxygen in the holding tank. The spraynozzles above the tower had to be washed every morning toremove the scum covering them. Unfortunately, long-term dataon the use of trickling towers could not be obtained because theslaughterhouse closed down.
Slaughterhouse 1 had its holding tank periodically pumpedout by a private contractor who charged $90 000 a year for theservice. All other slaughterhouses discharged their partiallytreated wastewater to the city sewer. The municipalitiesimposed a treatment charge to the slaughterhouses based onvarious criteria. For example, one municipality calculatedsurcharge fees based on:
In-plant wastewater treatment
scum removal in holding tankscum removal in holding tankdrum screen; OAF unit; trickling filtersettling tank; chemical-OAF unitdrum screen; chemical-OAF unitdrum screen; OAF unit
RESULTS and DISCUSSION
# One plant provided very precise figures including municipal surcharge as well as manpower,maintenance and operating costs of in-plant treatment, but most slaughterhouses supplied approximatecosts.• Included some rendering and/or meat processing activities.
Slaughter Capacity Wastewaterhouse (head/wk) production
(m3/d)
I 1900 572· 2800 763 11500 2464· 15000 38005 17500 3036· 45000 3600
Wastewater production and treatmentJohns (1995) suggested that slaughterhouses were fairlyefficient in terms of waste recovery and management. Thisstatement is also true of the Canadian meat industry. Mostdiscarded animal parts are sent for further transformation.Screened solids, skin, hair, and unusable interiors (e.g. badlivers, lungs, spleens) are used in cosmetic production; pancreasare kept for penicillin production; blood is dried andtransformed into an animal protein feed. Slaughterhouses,however, consume significant amounts of water and thusproduce large volumes of wastewater.
Table I presents plant capacity, wastewater production, andin-plant wastewater treatment at the six Quebec and Ontarioslaughterhouses that were visited between 1995 and 1996. Thethree plants dealing exclusively in slaughtering activities (1, 3,and 5) produced between 90 and 140 L of wastewater per hogkilled. The actual volume of fresh water used per animal killedwill be slightly higher since Slaughterhouses 3 and 5 measured
Slaughterhouse 5, raw wastewater was sampled six times ineight months. On one occasion, the effluent from the DAF unitwas also sampled.
All samples were analysed in duplicate for soluble and totalCOD (SCOD and TCOD), solid content, ammonia-nitrogen(NH4-N), total Kjeldahl nitrogen (TKN), volatile fatty acids(VFAs), pH, and alkalinity. Analyses were done according tomethods outlined in APHA (1992). Protein concentration wascalculated by multiplying the difference between TKN andNH4-H by 6.25 (AOAC 1984). Volatile fatty acid concentrationwas determined by gas chromatography. Samples from fourslaughterhouses were also analysed for metal and micronutrientcontent by the inductively coupled plasma (lCP) method at theUniversity of Ottawa, Ottawa, ON.
Data on effluent following in-plant treatment were alsomade available by Slaughterhouses 3, 4, and 5. AtSlaughterhouses 3 and 5, five and ten samples, respectively,were analysed at the plant laboratory over a one-year period.Slaughterhouse 4 collected samples on a bi-monthly basis, andthe data covered an I8-month period. Analyses were performedat the plant.
142 MASSE and MASSE
Table II. Raw wastewater quality at six hog slaughterhouses in Quebec andOntario (1995·1996).
t Slaughterhouse 5 was sampled six times in eight months. Except for metals which wereanalysed in only one sample, the average and standard deviation of the parameters aregiven in the table.
6
175
6.5250
805654
20914212o
15820
856
2481184614311149
34171250
fat. However, if laws governing landapplication of waste become stricter, asthey have for the pork producers, moreadvanced sludge treatment may berequired
Raw slaughterhouse wastewaterqualityTable II presents slaughterhousewastewater quality prior to anytreatment, except for the screening orsettling of coarser solids. Samples didnot include water from the afternoonwashing, which would probably dilutethe wastewater. However, it alsoexcluded the water from the scaldingtank, which is highly charged. Sachon(1986) estimated that the COD contentof scalding water ranged from 5000 to8000 mgIL and represented 20% of thedaily polluting load fromslaughterhouses.
Total COD in the raw wastewaterfrom the six slaughter-houses variedbetween 2333 and 8627 mgIL (TableIT). At Slaughterhouse 5, where the rawwastewater was sampled six times overan eight-month period, TCODconcentration ranged from 6908 to11530 mgIL. These TCOD values aresimilar to those measured in wastewaterfrom hog slaughterhouses in France(4118 ± 1409 mgIL; Sachon 1986) andGermany (between 500 and 10 5000mgIL; Tritt and Schuchardt 1992). Even
though all samples were collected after screening or settling ofthe coarser solids, SS accounted for 27 to 67% of TCOD.Volatile suspended solids (VSS) represented 80% of the SS andhad a COD content of .01± 0.36 g per g of VSS. The volatilefraction of the dissolved solids was 65 ± 18%. Protein contentranged from 444 to 2775 mgIL and made up between 37 and58% of the total volatile solids and 30 ± 7% of TCOD (1.15 gCOD per g protein; Sayed et al. 1988). Aside from Slaughterhouse 4, the wastewaters were neutral to slightly acid. Theycontained between 83 and 900 mgIL of alkalinity as CaC03•
Alkalinity tended to increase with wastewater strength. In allsamples, heavy metal concentration (cadmium, cobalt, nickel,copper, chromium) was below the detection limit. Nitrogen andphosphorous concentrations generally exceeded the limitsimposed by municipalities for discharging industrial wastewaterto municipal sewers without surcharge (see Eq. I). However,nutrients and micronutrients (calcium, sodium, magnesium,sulphur, and iron) were in adequate concentrations for abiological treatment ofslaughterhouse wastewater. Nitrogen andphosphorous averaged 6.0 and 2.3 g per 100 g of TCOD,respectively, while requirements for biological treatment areestimated at 3 and 0.7% of COD removed, respectively (Gradyand Lim 1980). Ammonia-nitrogen and sulfide concentrationswere well below the 3000 and 100 mgIL toxicity level,respectively, for the anaerobic biomass (Table II). High levelsof light metal cations can be inhibitory to bacteria, but in low
5t
6112215
23812367o
311 ± 34
593 ± 95169 ± 66
2648 ±66
6.9 ± 0.2906 ± 157
5748 ± 8234458 ± 7512099 ± 6221887 ± 550
8627 ± 16694753 ± 883
4.983
6.5333
7.2333
6.7333
p
N
pHAlkalinity as CaC03
Parameters Slaughterhouse(mg/L except pH) 1 2 3 4
Total COD 2941 3589 4976 2333Soluble COD 1510 2605 2817 778
Total solids 2244 2727 3862 2747
Volatile solids 1722 1966 3153 1204
Suspended solids 957 736 1348 877
Volatile SS 770 576 1192 594
Volatile fatty acids 197 166 221 164
Total Kjeldahl N 174 271 372 90
Ammonia-N 41 154 99 19
Protein 831 731 1700 444
Phosphorous 20 - - 28
Potassium 27 - - 60
Calcium 56 - - 54
Sodium 54 - - 369
Magnesium 25 - - 17
Sulphur 54 - - 49
Iron 2 - - 25
Manganese 0 - - 2
BOD =(BOD - 300)/300, where 300 mglL is themaximum level allowed,
FOG = (animal and/or vegetable grease - 100)/100, where100 mgIL is the maximum level allowed,
=(Kjeldahl nitrogen - 100)/100, where 100 mglL isthe maximum level allowed, and
=(total phosphorous - 10)/10, where 10 mgIL is themaximum level allowed.
A second municipality estimated that the localslaughterhouse utilised 80% of the public wastewater treatmentfacility and had to pay 80% of the cost. In a third case, themunicipality evaluated surcharge based on wastewater quantityand quality during three of the most productive days of the year.
Municipal surcharge represented between 10 and 100% oftotal treatment cost depending on the level and cost of the inplant pretreatment. Based on the data supplied by theslaughterhouses, overall treatment cost in 1995-1996 rangedfrom $0.70 to $1.60 per m3 of wastewater, except for Slaughterhouse 1, which paid over $5 per m3
• Wastewater treatment atthe slaughterhouses also produced a large quantity of sludgerequiring special handling and/or further treatment. At the timeof the survey, the sludge was hauled to private fanns and mixedwith manure or compost before it was spread on agriculturalfields. Barnett et al. (1997) reported that land application ofsludge from meat processing plants had a positive effect onyield, except when the sludge contained high concentrations of
CANADIAN AGRICULTURAL ENGINEERING Vol. 42, No.3 JuIyIAugust/September 2000 143
Slaughterhouse and type of wastewater treatment
Table TIl. Emuent quality following preliminary treatment at fourslaughterhouses in Quebec and Ontario (1995·1996). Thevalues in parentheses were supplied by the slaughterhouse.
Raw wastewater from hog slaughterhouses in Quebec andOntario contained high concentrations of degradable organics.Existing wastewater treatment at all surveyed slaughterhouseswas not sufficient to produce an effluent that complied withmunicipal treatment plant criteria and usually generated largevolumes of sludge. Consequently, slaughterhouses had to paysurcharge fees to municipalities to further treat their wastewaterand had to dispose of the sludge. Wastewater treatment costscan only increase in the future since they depend on municipaltaxation level and on the price of non-renewable chemical input.
CONCLUSION
44
unit
110
41174
831
6.6167
4242
14212154o
treatment in the trickling towers decreased SCODby an additional 27% but did not further reduceTCOD, probably because bacteria were sloughedfrom the filters and left the reactors with theeffluent, thus increasing particulate COD. Thehigher SS concentration in the final effluent,compared to the raw wastewater (Tables II and III),also suggested some biomass losses from theaerobic filters. However, in five effluent samplesanalysed by Slaughterhouse 3 over a one-yearperiod, average SS concentration was less than halfthe concentration measured in the effluent samplecollected for this survey (Table III). Nevertheless,TCOD in the biologically treated effluent remainedhigh both in the sample collected for this project andin samples analysed by the slaughterhouselaboratory.
At Slaughterhouse 4, the chemical-DAF unitreduced TCOD and SCOD by 58 and 26%,respectively. Over 50% of the SS and 35% of thenitrogen were removed. However, effluent TCODand SS concentrations were still slightly above themaximum allowable levels for industrial wastewaterdischarge into municipal sewer without surcharge(see Eq. I; COD:BODs ratio in slaughterhousewastewater varies between 1.3 and 2.0 (Sachon1986; Temper et a1. 1988». Data supplied bySlaughterhouse 4 showed that, over an 18-monthperiod, the effluent had a BODs concentrationvarying between 400 and 500 mglL and an SScontent between 200 and 300 mgIL. Slaughterhouse4 had the most dilute raw wastewater and theweakest effluent. However, it had the highest waterconsumption, probably because of the meatprocessing activities. The municipal surcharge forwastewater treatment was kept low, but the costassociated with water input was high.
Slaughterhouse 5 had the most efficienttreatment system; the chemical-DAF unit removed67% of TCOD and SCOD. However, effluentTCOD and SS concentrations remained high at 3121and 1974 mgIL, respectively. The effluent TCODdata supplied by the slaughterhouse laboratory were
similar to those measured in this survey (Table III). Effluentfrom Slaughterhouse 5 was among the most polluted, but its rawwastewater was also 2 to 3 times stronger than that from otherslaughterhouses.
6OAF
19691347682893
12902325
pH 7.1 5.7 7.0Alkalinity as CaCOJ 667 167 542
Parameters3 4 5(mgIL except pH)
DAFunit Chemical- Chemical-trickling filter OAF unit OAF unit
Soluble COD 1598 576 1435Total COD 3921 986 3121
(2569 ± 512)t (3229 ± 1231) t
Total solids 2197 3460Volatile solids 1676 633 2157Volatile SS 1792 265 1646Suspended solids 1956 422 1974
(836 ± 770)t (226 ± 85)* (754 ± 47l)t
Volatile fatty acids 673 273 279
Ammonia-N 228 19 100Total Kjeldahl N 295 59 269
(183 ± 113)t (268 ± 245)t
Protein 419 250 1061Fat and grease (291 ± 316)t (65 ± 35)* (22 ± 15)t
Phosphorous 22 78(28 ±19)t (l0±4)* (22 ± 15)t
Potassium 38 214Calcium 54 44Sodium 404 453Magnesium 17 17Sulphur 48 63Iron 19 43Manganese 2 0
t Based on five to ten samples collected over a one-year period*Based on bi-monthly samples collected over an 18-month period
concentrations of approximately 75 to 200 mgIL, calcium,sodium (up to 400 mgIL), potassium, and magnesium havestimulatory effects on anaerobic microorganisms (McCarty1964). Low concentrations of iron also have a beneficial effecton biomass activity (Speece 1996).
Slaughterhouse quality after preliminary treatmentThe quality of wastewater as it leaves the slaughterhouse for themunicipal treatment plant is presented in Table III.Slaughterhouses 1 and 2 were not included because theirwastewater was not treated at the plant. Slaughterhouse 6 hadthe lowest level of in-plant wastewater treatment. The OAF unitremoved approximately 35% of the TCOD and SS, but it didnot reduce SCOD, nitrogen, and protein concentrations. Therewere no other data supplied by the slaughterhouse.
At Slaughterhouse 3, the DAF unit reduced the TCOD andSCOD concentrations by 22 and 16%, respectively. Aerobic
144 MASSE and MASSE
Hog slaughterhouses produce a wastewater that is suitablefor anaerobic treatment in terms of BOD, nutrient, andmicronutrient concentration. However, hydraulic retention timewould have to be sufficient to allow for the degradation of theSS, which represented between 27 and 67% of TCOD.
REFERENCES
AOAC. 1984. Official Methods ofAnalysis of the AssociationofOfficial Analytical Chemists. Arlington VA: Associationof Official Analytical Chemists.
APHA. 1992. Standard Methods for the Examination ofWaterand Wastewater. Washington. DC: American Public HealthAssociation.
Barnett, G.M., K. Kouassi, T. Attiave and R. Lewandowski.1997. Valorisation des boues gem!rees a partir des eauxresiduelles de l'industrie de transformation de la viandepour lafertilisation des sols. Dossier no 12-39448872-084.Lennoxville, QC: Agriculture and Agri-Food Canada.
Belanger, D., P. Bergevin, J. Laperriere and R Zaloum. 1986.Conception, contrOie et efficacite d'un reacteur biologiquesequentiel pour I'epuration des eaux usees d'un abattoir.Sciences et techniques de I'eau 19: 142-156.
Black, M.G., J.M. Brown and E. Kaye. 1974. Operationalexperiences with an abattoir waste digestion plant at Leeds.Journal ofthe Water Pollution Control Federation 73:532537.
Bull, M.A., R.M. Sterritt and J.N. Lester. 1982. The treatmentof wastewaters from the meat industry: A review.Environmental Technology Letters 3: 117-126.
Camin, K.Q. 1970. Cost of waste treatment in the meat packingindustry. In Proceedings of the 25th Industrial WasteConference, 193-202. Ann Arbor, MI: Ann Arbor Science.
Campos, J.R., E. Foresti and R.D.P. Camacho. 1986. Anaerobictreatment in the food processing industry: Two case studies.Water Science Technology 18:87-97.
Dague, RR, R.F. Urell and E.R. Krieger. 1990. Treatment ofpork processing wastewater in a covered anaerobic lagoonwith gas recovery. In Proceedings of the 44th IndustrialWaste Conference, 815-823. Ann Arbor, MI: Ann ArborScience.
Defour, D., D. Derycke, J. Liessens and P. Pipyn. 1994. Fieldexperience with different systems for biomass accumulationin anaerobic reactor technology. Water Science Technology30: 181-191.
Festino, C. and Ch. Aubart. 1986. Epuration d' effluents liquideset valorisation energetique de dechets solides d'abattoir parvoie anaerobie. Entropie 130-131 :57-67.
Grady C.P. and H.C. Lim. 1980. Biological WastewaterTreatment. New York, NY: Marcel Dekker Inc.
Hammer, MJ. and C.O. Jacobson. 1970. Anaerobic lagoontreatment of packing house wastewater. In Proceedings ofthe 2nd International Symposium for Waste TreatmentLagoons, 347-354. Kansas City, MO.
Johns, M.R. 1995. Developments in wastewater treatment inthe meat processing industry: A review. BioresourceTechnology 54:203-216.
Kostyshyn, C.R., W.A. Bonkoski and J.E. Sointio. 1988.Anaerobic treatment of a beef processing plant wastewater:a case history. In Proceedings of the 42"d Industrial WasteConference, 673-692. Ann Arbor, MI.: Ann Arbor Science.
Masse, 0.1. 1995. Psychrophilic anaerobic digestion of swinemanure slurry in imermittently fed sequencing batchreactors. Ph.D. thesis. Ottawa, ON: University of Ottawa.
Masse, D.I. and F. Croteau. 1998. La digestion anaerobiepsychrophile du lisier de porc a l'interieur de bioreacteursa operations sequentielles. Final report presented to theFederation des producteurs de porcs du Quebec.Lennoxville, QC: Agriculture and Agri-Food Canada.
Masse, 0.1. and R.L. Droste. 1997. Microbial interaction duringanaerobic treatment of swine manure slurry in a sequencingbatch reactor. Canadian Agricultural Engineering 39:35-41.
Masse, 0.1., R.L. Droste, K.J. Kennedy, N.K. Patni and J.A.Munroe. 1997. Potential for the psychrophilic anaerobictreatment of swine manure using a sequencing batch reactor.Canadian Agricultural Engineering 39:25-33.
Masse, 0.1., and L. Masse. 2000. Treatment of slaughterhousewastewater in anaerobic sequencing batch reactors.Canadian Agricultural Engineering 42: 131-137.
Masse, 0.1., N.K. Patni, R.L. Droste and KJ. Kennedy. 1996.Operation strategies for psychrophilic anaerobic digestion ofswine manure slurry in sequencing batch reactors. CanadianJournal ofCivil Engineering 23: 1285-1294.
McCarty, P.L. 1964. Anaerobic waste treatment fundamentals,Part 3. Public Works 95:91-94.
Metzner, G. and U. Temper. 1990. Operation and optimizationof a full-scale fixed-bed reactor for anaerobic digestion ofanimal rendering wastewater. Water Science Technology22:373-384.
Quinn J.M. and P.N. McFarlane. 1989. Effects ofslaughterhouse and dairy factory wastewaters on epilithon:A comparison in laboratory streams. Water Research23: 1267-1273.
Rollag, D.A. and J.N. Dornbuh. 1966. Anaerobic stabilizationpond treatment of meat packing wastes. In Proceedings ofthe 2JS' Industrial Waste Conference, 768-782. Ann Arbor,MI: Ann Arbor Science
Rudd, T., SJ. Hicks and J.N. Lester. 1985. Comparison of thetreatment of a synthetic meat waste by mesophilic andthermophilic anaerobic fluidized bed reactors.Environmental Technology Letters 6: 209-224.
Sachon, G. 1982. Traitement des eaux residuaires des abattoirsde betail. In Proceedings ofthe International Symposium onFood Industries and The Environment, 293-305. Budapest,Hungary.
Sachon, G. 1986. Les eaux residuaires des abattoirs de betail.Gestion et traitement. Tribune Cebedeau 515: 27-45.
Safley, L.M. and P.W. Westerman. 1988. Biogas productionfrom anaerobic lagoons. Biological Wastes 23: 181-93.
Safley, L.M. and P.W. Westerman. 1992. Performance ofa lowtemperature lagoon digester. Bioresource Technology 41:167-75.
Sangodoyin, A.Y. and O.M. Agbawhe. 1992. Environmentalstudy on surface and groundwater pollutants from abattoireffluents. Bioresource Technology 41 :193-200.
CANADIAN AGRICULTURAL ENGINEERING Vol. 42, No.3 July/August/Seplember 2000 145
Sayed, S.K.I. 1987. Anaerobic Treatment of SlaughterhouseWastewater Using the UASB Process. PhD thesis.Wageningen, The Netherlands: Agricultural University ofWageningen.
Sayed, S.K.I., J. van der Zanden, R. Wijffels and G. Lettinga.1988. Anaerobic degradation of the various fractions ofslaughterhouse wastewater. Biological Wastes 23:117-142.
Speece, R.E. 1996. Anaerobic Biotechnology for IndustrialWastewaters. Nashville, TN: Archae Press.
146
Temper, U., R. Muller and G. Metzner. 1988. Anaerobicdigestion of rendering wastewater. In Proceedings ofthe 5th
International Symposium on Anaerobic Digestion, 591-594.Bologna, Italy.
Tritt, W.P. and F. Schuchardt. 1992. Materials flow andpossibilities of treating liquid and solid wastes fromslaughterhouses in Germany. Bioresource Technology41:235-245.
MASSE and MASSE
Application of artificial neural networks in imagerecognition and classification of crop and weeds
C.-C. YANGl, S.O. PRASHERl, J.-A. LANDRyl, H.S. RAMASWAMy2 and A. DITOMMAS03
IDepartment of Agricultural and Biosystems Engineering, McGill University, Ste-Anne-de-Bellevue, QC, Canada H9X 3V9;2Department of Food Science and Agricultural Chemistry, McGill University, Ste-Anne-de-Bellevue, QC, Canada H9X 3V9; andJDepartment ojCrop and Soil Sciences, Cornell University, Ithaca, NY, USA /4853-280/. Received 5 May /999; accepted 30 June2000.
Yang, C.-c., Prasher, 5.0., Landry, J.-A., Ramaswamy, H.S. andDiTommaso. 2000. Application of artificial neural networks inimage recognition and classification ofcrop and weeds. Can. Agric.Eng. 42:147-152. The objective of this study was to develop a backpropagation artificial neural network (ANN) model that coulddistinguish young com plants from weeds. Although only the colourindices associated with image pixels were used as inputs, it wasassumed that the ANN model could develop the ability to use otherinformation, such as shapes, implicit in these data. The 756x504 pixelimages were taken in the field and were then cropped to 100x 100-pixelimages depicting only one plant, either a com plant or weeds. Therewere 40 images of com and 40 of weeds. The ability of the ANNs todiscriminate weeds from com was then tested on 20 other images. Atotal of 80 images of com plants and weeds were used for trainingpurposes. For some ANNs, the success rate for classifying com plantswas as high as 100%, whereas the highest success rate for weedrecognition was 80%. This is considered satisfactory, given the limitedamount of training data and the computer hardware limitations.Therefore, it is concluded that an ANN-based weed recognition systemcan potentially be used in the precision spraying of herbicides inagricultural fields. Keywords: artificial neural networks, machinevision, precision farming, weeds, herbicide application, pollution.
L'objectif de l'etude est de developper un modele qui pourraitdistinguer les jeunes plants de mats des mauvaises herbes. Bien queseul l'indice de couleur calcule a partir d'images numerisees ait eteutilise comme donnee, it a ete suppose que Ie modele ANN ainsideveloppe, pourrait etre employe pourd'autres types d'informations quiseraient en correlation directe avec les donnees premieres, comme parexemple la forme. Des images de 756*504 pixels ont ete prises dansles parcelles. A partir de celle-ci, d'autres de 100* 100 pixels ont eteselectionnees avec comme critere qu'elles ne contiennent qu'un seulplan, soit de mats, soit de mauvaises herbes. On a pris 40 images demats et 40 de mauvaises herbes. La capacite de ANN adistinguer lesmauvaises herbes du mats, a ete ainsi teste sur 20 autres images. Untotal de 80 images de plants de ma'is et de mauvaises herbes a eteutilise pour arriver anos fins. Pour certains modeles ANN developpes,Ie taux de reussite areconnaitre les plants de mats, a ete aussi haut que100%. Alors que pour les mauvaises herbes Ie meilleur resultat n'aatteint que 80%. Ce resultat peut etre considere satisfaisant etant donneles limitations d'une part de l'acquisition de donnees et d'autre part dela capacite du materiel infonnatique. En conclusion, un systeme dedetection des mauvaises herbes base sur ANN peut potentiellementetre utilise dans l'epandage de precision des herbicides dans leschamps.
INTRODUCTIONSignificant progress in the development of machine vision andimage processing technology has been made in the past few
years in conjunction with improvements in computer technology(Baxes 1994). Equipment for machine vision and imageprocessing has been reduced in cost, size, and weight, can beinstalled in most vehicles (e.g., tractors), and is accessible forcivilian use. Machine vision and image processing are usedincreasingly in biology, materials science, photography, andother fields (Baxes 1994). Many experiments have suggestedthat machine vision can be used to recognize and localize weedsin agricultural fields (Anonymous 1994a, 1994b; Blackmer andSchepers 1996; Meyer et a1. 1997; Schmoldt et al. 1997; Staffand Benlloch 1997). It might therefore be used to control sitespecific spraying herbicide application, thus reducing bothenvironmental pollution from the overuse of agrochemicals, aswell as the cost of weed control.
It is presently quite difficult to use machine vision todistinguish weeds from the main crop in real time, due to thesubstantial computational resources and the complicatedalgorithms required. Artificial neural networks (ANNs) canovercome some of these difficulties by interpreting imagesquickly and effectively. ANNs are composed of numerousprocessing elements (PEs) arranged in various layers, withinterconnections between pairs of PEs (Haykin 1994;Kartalopoulos 1996; Kasabov 1996). They are designed toemulate the structure of natural neural networks such as thoseof a human brain. For most ANNs, PEs in each layer are fullyconnected with PEs in the adjacent layer or layers, but are notconnected to other PEs in the same layer. The PEs simulate thefunction of the neurons in natural neural networks, while theinterconnections between them mimic the functions ofdendritesand axons.
There have been many applications of ANNs reported forthe interpretation of images in the agri-food industry. Studieshave shown that for the interpretation of images ANNs can beas accurate as procedural models (Deck et a1. 1995;Timmermans and Hulzebosch 1996). For example, the accuracyof classification of potted plants can be greater than 99%(Timmermans and Hulzebosch 1996), apples can be graded bycolour with an accuracy of 95% (Nakano 1997), theclassification of logs for defects using computed tomographyimagery can be 95% accurate (Schmoldt et a1. 1997), and theaccuracy for the classification of wheat kernels by colour can be98% or more (Wang et al. 1999). Generally, ANNs canefficiently model various input/output relationships with theadvantage of requiring less execution time than a proceduralmodel (Yang et a1. 1997a, I997b). These features make theANN approach very appealing for real-time image processing.
CANADIAN AGRICULTURAL ENGINEERING Vol. 42. No.3 JulylAugust/September 2000 147
(A) (D)The most cOl11mon weeds in the fields were Agropyron
repens. Cyperlls esclllellllts. Plalllogo major, Stellaria media.Chenopodium a/buill. Abuti/oll theophrasti and Taraxacumofficina/e. Together. these weed spccies constitutcd onecategory of objects to be diffcrentiated from com plants. Tosimplify the experimcnt, the study did not involve the trainingof ANNs 10 distinguish between differcnt species of weeds. Thedifferentiation betwecn various weed species will be carried oulin a futurc study.
The digital images wcrc downloaded to a personal compulcrhaving a Pcntium 200MMX microprocessor <lnd 96 MB ofRAM and wcrecol1vcrted from the nativc Kodak digital camcraformat (KDC) 10 Ihe 8-bit colour bitm"p fonnal (BMP). Thesize of the images W<lS 756x504 pixels. These images wereviewed and further cropped to a sizc of IOOx 100 pixels so thateach includcd either a com plaI1l or a group of weeds. It wouldhave been impractical to use the 756x504 pixel images, sinccthe PC mcmory would have been inadequate. Care was takcnnot to include both corn and weeds in a given image, so as tosimplify the ANN lmining proccss. Mixed images. containingboth corn plants and weeds, will be used in future studies. Someexamplcs of cropped images are shown in Fig. I.
After the BMP images wcre obtained. lhey wcre prcprocessed with thc Image Processing Toolbox v2.0 forMATLAB v5.0 (MalhWorks 1997a. I997c). The BMP imageswcre convcned to indexed images based all a red-green-blue(RGB) colour system. Each pixel of an image was classifiedinto one of 256 categories, reprcsented by an intcger in therange from a(black) to 255 (white). Each assigncd colour indexnumber served as an ANN inpul and, therefore, there were 10000 (I OOx 100) inputs for cach image. Although the colourindices wcre the only inputs used in this study. othcr featurcs.such as shapes. were expected to be taken into account by thcANNs since information aboul them is implicit in therelationships betwcen the pixel colours.
The 'eural Network Toolbox v2.0 for MATLAB v5.0(MalhWorks 1997b. I997c) was used to build the A N models.During training. the ANNs wcre presentcd with binary outputdata. Two classification schemes were tried to rcprescilt theoutput data. In one scheme (Type I). as shown in Fig. 2. thercwas only one output variablc in the training data set. An outputvariable valuc of zcro was assigned to weeds and a value or oncto corn. In thc other scheme (Type 2), as shown in Fig. 3. therewere two output variables: thc first was an cstimate of thcpossibility that the object was a corn plant. and the second wasan estimate of the possibility that it was a weed. For a cornplant. the first oULput should be one and the second should bczero, and vice-versa for a wecd. Forty images of corn and fonyimages of weeds were used to train the ANI s. Testing was donewith 10 other imagcs of com and 10 of weeds.
While crisp values of one and zcro wcre uscd duringtraining, values between zero and aile could rcsult during [hetesting of the ANNs. and some sLrategy was therefore requiredto deal with such values. To address the problem of uncertainclassification, two schemcs were tcstcd for the Type Iclassification and four were tested for thc Typc 2 classification.These are named Type I-A and Type I-B. "nd Type 2-A toType 2-D. The conditions undcr which an objcct wasrecognized as a wced or a corn plant arc summarized in TableI. Both output methods would ideally Icad to the same results if
MATERIALS and METHODS
Fig. 1. Examples of 100xlO0 cropped images. (A) and (B)for corn in training data sct; (e) for corn in tcstdata sct; (D) and (E) for weeds in training dahl sct;and (F) for weeds in test data set.
(B) (E)
(C) (F)
A Kodak DC50 zoom camera was Llsed to acquire digitalimages of corn plums and weeds in [wa experimental fields (# 18and #24) on Ihe Macdonald Campus Farm of McGill University.Ste-Annc-de-Bellcvuc. QC. Canada, in 1997. Images were takcnfrom Ficld # 18 on June I 1, 12. and 20 and from Field #24 onJune 10. 13. and 20, the time at which post-emcrgent herbicideapplication is usually carricd out. The pictures were taken atsevcral randomly-chosen locations in the fields. During imagecollection, the camera was always held at the same height (600111m) to capture a bird's-eye view of objects on the ground.Since the size of the plants varied significantly from onc spot toanother, it was ncccssary to zoom in or out to obtain c1carimages. Although the actual area covered varicd slightly fromimage {Q image. it was approximately 300 mm x 200 mm.
Given the aforementioned considerations. it was decided touse ANNs for weed recognition in this study. Specifically. thepossibility of using ANNs to distinguish between images ofcrop plants and weeds, captured in real-lime by a digilal camera,was investigated. In this study. the use of ANNs was confined10 lhe differentiation of corn (Zea mays. L.) and seven weedspecies cOl11l11only encountered in the experimental fields.
148 YANG. PRASI-IER. LANDRY. RAMASWAMY and DiTOMI\otASO
Fig. 2. The ANN structure for Type I output.
was normalized so that it ranged from zero to one(in~lead of from zero to 255) before beingpresented to the ANN. A log sigmoid transferfunction was implemented in each PE.
During the training procedure of the ANNs,the maximum acceptable sum-squared error wasempirically set at I x IO·s. The training process wascarried out with fast back-propagation, until amaximum of 2000 epochs (cycles) or themaximum acceptable sum-squared error wasreached. Some initial nllls showed that thesesettings appeared to be sufficient for this study.
A trial-and-error method was used to set thenumber of PEs in the hidden layer. For the TypeI output. the number of PEs in the hidden layerwas arbitrarily selected and varied from 70 to 300.For the Type 2 output. the number of PEs in thehidden layer was varied from 120 to 300.Computer memory in this study was insufficientfor training an ANN with more than 300 PEs inthe hidden layer. The success classification ratefor each object (corn or weeds) was determinedafter training in order to evaluate the ANN modelperformance.
The McNemar test for the significance of changes (Daniel1990: Fleiss 1981) was used to compare each pair of outputtypes for the A Ns with the same number of PEs in the hiddenlayer. It is a chi-square test based on the binomial distributionror matched pairs 01" data. The critical value of the test statisticwas chosen to be 1.96 (signifying a confidence level or 95%).When the computed valuc of the test statistic is less than 1.96,IwO outplH types compared by the McNemar test arc notsignificantly different. In addition, the Brier score (I-land 1994)was used to evaluate the predictive ability of different ANNmodels containing different numbers of PEs. It is a unitlessindex of predictive accuracy and the range of the Brier score isfrom zero to one. The smaller the score is. the better is thepredictive abilily of rhe model.
•••••••••••
•••••••••••
oO~LO~l
0"'---------+
the same classification thresholds arc used and if there areenough images to ensure proper lraining of the ANNs.However, the slightly different ANN architectures that are usedin each method could lead to differences in effectiveness, and.moreover. the Type 2 method allows a more nexibleinterpretation of lhe results.
Various types of ANNs can be created with MATLAB.Back-propagation networks were selected for this projectbecause they have been successfully used in various imageprocessing applications in agriculture (Deck et al. 1995:Schmoldt el al. 1997; Timmennans and Hulzebosch 1996:Wang et :11. 1999). Each PE in the input layer received thecolour index value of one of the pixels in the input images. Onehidden layer was used between the input and output layers. Dueto the choice of the back-propagation network, the input data
Fig. 3. The ANN structure for Type 2 output.
RES LTS and DISCUSSION
•••••••••••
•••••••••••
--+[1,0] for com[0, I] for weeds
--+
The A N's ability to properly classify imagesusing the Type I output and the two evaluationschemes is shown in Table II for various numbersof PEs in the hidden layer. As shown in Table II.the Brier scores varied from 0.153 to 0.271. Thelowest Brier score was obtained by the A 'with 110 PEs in the hidden layer. The generallylow Brier scores indicate the potential capabilityof A's to classify and recognize images.However. the ANN model with the lowest Brierscore does not necessarily mean the best modelin this case since it is not expected for a modelthat would give the best prediction efficiency forboth corn and weed ima!!e recoe.nitions. Rather.il is very impol1ant for o~r appli~ation that weedimages do not get classified as corn since itwould lead to poor weed control and yield losses.Thus, it is necessary to further analyse the ANNmodel performance in detail and carefullyconsider the success recognition r<.lles forclassifying corn and weeds.
CANADIAN AGRICULTURAL ENGINEERING Vol. ~2. No.3 July/AllgU~I/ScplCl11hcr2000 t49
Table I. The output types of ANNs and the threshold of classification.
The results shown in Table II indicate that ANNs can, ingeneral, classify and distinguish images of com plants fromimages of weeds with a success rate of 80 to 100%. The successrate for distinguishing images of weeds from images of corn
Table II. Success classification rate for Type 1 ANNs.
PEs in the Type I-A Type I-B
hidden BrierScore Com Weeds Com Weeds
layer (%) (%) (%) (%)
70 0.23 80 60 70 70
80 0.17 90 60 80 70
90 0.23 80 40 60 80
100 0.17 80 80 70 80
110 0.15 100 60 80 70
120 0.19 80 50 70 80
130 0.21 70 70 60 70
140 0.25 60 50 60 70
150 0.20 90 60 70 60
160 0.27 80 50 80 50
180 0.19 90 60 90 70
200 0.24 90 40 90 50
220 0.21 90 60 90 70
230 0.17 90 60 90 70
240 0.24 90 50 90 60
260 0.22 90 50 80 60
280 0.21 90 60 80 60
300 0.19 90 70 90 70
When the output did not match thecondition in the left column.
plants often lay between 70 to 80%, althoughsometimes it was as low as 50%. Results indicatethat, after training, ANNs seemed to misclassifymore weed images as images of com plants thanvice versa, Le., the success recognition rate forcom is always higher than the success recognitionrate for weeds. When the threshold value forclassification was set at 0.50 (Type I-A),indicating an equal possibility of either com orweeds, the success rate for recognizing com was100%, which was much higher than thecorresponding success rate for recognizing weeds(60%). The lower success rate for weed detectionindicates it is more likely for an ANN tomisclassify weeds as com. In the real world, thissituation is highly undesirable and it could lead toinadequate herbicide application and weedcontrol.
To reduce the possibility of missing weeds, the thresholdwas set at the average value of 20 outputs, 0.62, for Type I-Boutput, meaning that when images are difficult to classify, moreimages would be classified as weeds. In some cases, the successrecognition rate for weed classification with this scheme wasincreased by 10% (Le., for an ANN with 220 PEs in the hiddenlayer) to 40% (Le., for an ANN with 90 PEs in the hiddenlayer). At the same time, however, the success recognition ratefor com classification decreased by 10% (i.e., for an ANN with100 PEs in the hidden layer) to 20% (Le., for an ANN with 90PEs in the hidden layer).
To compare and evaluate the two Type I output schemes,the McNemar test was used to perform pairwise comparisons.Although the results from Type I-A and Type I-B appear to bedifferent (Table II), the results of the test in Table ill show thatthe difference between these two output schemes is notsignificant (P~0.05).
As shown in Table II, single-output ANNs did notaccurately classify some images. The actual values of ANNsoutputs for these images were around 0.50 or 0.62, instead ofapproaching zero or one. To attempt to improve this situation,additional ANN models were created that produced two outputs.The first output represented the possibility of the image beingof a com plant while the second output represented thepossibility of it being of a weed. Four schemes for this type ofclassification were investigated in this study. Due to a greaternumber of PEs in the output layer, no ANNs were investigatedthat had less than 120 PEs in the hidden layer. The Brier scoresand the success recognition rates for the four Type 2 schemesare given in Table IV. Some of the comparisons by theMcNemar test between the Type I-A and each of the Type 2 arealso presented in Table III.
The Brier scores in Table IV range from 0.241 to 0.358. Thelowest Brier score was obtained by an ANN with 160 PEs in thehidden layer. Although this model is among the best for cornrecognition, the success recognition rate for weeds isunsatisfactory. As stated earlier, this is contrary to the resultsexpected in this study since weed misrecognition would lead topoor weed control and significant yield losses. Therefore, it alsorequires further analysis and interpretation of the results.
Generally, the success classification rates shown in Table IVranges from 60 to 90% for corn and from 40 to 80% for weeds.
Weeds
(Second output> First output) or (Secondoutput> 0.5)
Second output> 0.5
(Second output> First output) or (Secondoutput> Average value of first outputs)
Second output> First output
When the output did notmatch the condition inthe right column.
Output> Average valueof all outputs
Com
Output> 0.5
I-B
I-A
2-D
2-C
2-B
Type
2-A
150 YANG. PRASHER. LANDRY. RAMASWAMY and DiTOMMASO
number of PEs in the hidden layer, the highestsuccess rate for weed classification was obtainedusing the Type 2-C scheme. However, the successrate for com recognition was generally lower withthe Type 2-C scheme than the others. With theType 2-C scheme, only the ANN with 200 PEs inthe hidden layer could obtain the successrecognition rates as high as 80% for both com andweeds.
Although ANNs using different Type 2 analysisschemes differed from one another in performance,as shown in Table III, the McNemar test indicatedthat these differences were not statisticallysignificant (P~0.05). Neither did the McNemar testshow any significant differences between the resultsof the Type I and Type 2 methods. However,particular analysis schemes might still be bettersuited to certain image classification andrecognition problems since they do providedifferent ways of looking at the output.
To further improve ANN performance in imagerecognition and classification, other methods maybe investigated in the future. Firstly, due tocomputer memory limitations, the number of PEs inthe hidden layer was limited to 300 in this study,
i.e., 3% of the number of input PEs. Although there is nomethod for determining the best number of PEs to include in thehidden layer based on the number of inputs, the number of PEsused in this work may have been insufficient for such a largeamount of input data. More PEs in the hidden layer would resultin better performance. However, more computer memory wouldbe required to generate such ANNs, and a faster processorwould also be needed to save time during training. Secondly,the training data set contained only 80 images, which might nothave been sufficient. It would be necessary to collect more data
1.414
1.000
0.000
1.342
1.000
0.577
1.000
0.000
0.447
0.447
Type I-Avs
Type 2-D
PEs in the Type I-A Type I-A Type I-A Type I-Ahidden vs vs vs vslayer Type 1-8 Type 2-A Type 2-8 Type 2-C
120 1.732 1.000 0.447 1.000
140 1.000 1.342 1.342 1.342
160 0.000 1.000 1.000 1.000
180 1.000 0.000 0.000 0.000
200 1.000 1.000 1.000 1.000
220 1.000 1.414 1.414 1.414
240 1.000 0.000 1.000 1.000
260 0.000 0.447 0.000 0.447
280 0.000 0.577 0.577 0.577
300 0.000 0.000 0.000 0.000
Table III. Some results of the McNemar test between different outouttypes of ANNs.
Comparing Tables II and IV, the success classification rate forweeds is greater in some cases for the Type 2 method. Forexample, an ANN with 200 PEs in the hidden layer had asuccess recognition rate for weeds increased from 40 to 50%(Table II) to 70 to 80% (Table IV). An ANN with 260 PEs inthe hidden layer had a success recognition rate for weeds thatincreased from 50 to 60% (Table II) to 70 to 80% (Table IV).The success rate of com classification, however, decreased inall these cases. Such a dilemma was also observed with resultsobtained using the Type 2-A to 2-D schemes. For the same
Table IV. Success classification rate for Type 2 ANNs.
PEs in the Type 2-A Type 2-B Type 2-C Type 2-D
hiddenBrierScore Com Weeds Com Weeds Com Weeds Com Weeds
layer (%) (%) (%) (%) (%) (%) (%) (%)
120 0.270 90 50 80 50 80 60 80 50
140 0.245 90 50 90 50 90 50 90 50
160 0.241 90 50 90 50 90 50 90 40
180 0.250 90 60 90 60 90 60 90 60
200 0.325 80 70 80 70 80 80 80 70
220 0.305 80 50 80 50 70 60 80 50
240 0.257 90 50 90 60 90 70 90 60
260 0.358 60 70 60 80 60 80 70 80
280 0.305 80 60 80 60 80 60 80 60
300 0.301 90 70 90 70 80 70 90 70
CANADIAN AGRICULTURAL ENGINEERING Vol. 42. No.3 July/August/September 2000 151
from the field to increase the size of the training data set andensure proper training. Although the results in this study do notshow statistically significant differences among the variousoutput analysis methods used, by the McNemar test, the notableresults, that were observed, point to a need to further investigatedifferent methods for image classification using ANNs. Togenerally evaluate the ANN performance by different numbersof PEs in the hidden layer, the Brier score can be used. Toinvestigate the detail of ANN performance for com or weedrecognition, the success recognition rate for each plant objectshould be discussed respectively. The McNemar test can beused to examine the differences between different ANNmethods.
Another noteworthy result from this study is that while thetime needed to train an ANN model was approximately 20hours, the testing time was usually less than one second perimage. This information is of paramount importance for realtime weed recognition and herbicide application problemswhere one may have only a few seconds to do the recognitionand communicate with the controller of the herbicide sprayer tomake informed decisions about site-specific spraying.
SUMMARY
This study was undertaken to develop an ANN to classifyimages taken from the field and detect the presence of weeds.The images were taken from cornfields in Ste-Anne-deBellevue, southwestern Quebec, Canada, in June 1997. Colourindex values were assigned to the pixels of the indexed imageand used as ANN inputs. There were 80 images, I OOx I00pixels, for training, and 20 images for testing. Many backpropagation ANN models were developed with differentnumbers of PEs in their hidden and various output layers. Sixdifferent evaluation schemes for two ANN output strategieswere used. The performance of the ANNs was compared andthe success rate for the identification ofcom was observed to beas high as 80 to 100%, while the success rate for weedclassification was as high as 60 to 80%. The Brier score and theMcNemar test were used to statistically evaluate the resultsobtained in this study. Although the study was limited by theavailable computational resources and training data, the resultsindicate the potential of ANNs for fast image recognition andclassification. Fast image recognition and classification can beuseful in the control of real-world, site-specific herbicideapplication. Challenges still remain, however, in the analysis ofreal-world images where com, weeds, and other plants mayappear together.
REFERENCES
Anonymous. 1994a. The forces fueling precision farming. FarmIndustry News 27: 4-9.
Anonymous. 1994b. Aiming for smart weed control. FarmIndustry News 27: 28-29.
Baxes, G.A. 1994. Digital Image Processing: Principles andApplications. New York, NY: John Wiley & Sons, Inc.
Blackmer, T.M. and J.S. Schepers. 1996. Using DGPS toimprove com production and water quality. GPS World 7:44-52.
Daniel, W. W. 1990. Applied Nonparametric Statistics, 2ndedition. Boston, MA: PWS-KENT Publishing Company.
152
Deck, S.H., C.T. Morrow, P.H. Heinemann and H.J. SommerIn. 1995. Comparison of a neural network and traditionalclassifier for machine vision inspection of potatoes. AppliedEngineering in Agriculture 11: 319-326.
Fleiss, J.L. 1981. Statistical Methods for Rates and Proportions.New York, NY: John Wiley & Sons, Inc.
Hand, D.J. 1994. Assessing classification rules. Journal ofApplied Statistics 21: 3-17.
Haykin, S. 1994. Neural Networks: A ComprehensiveFoundation. New York, NY: Macmillan College PublishingCompany, Inc.
Kartalopoulos, S.V. 1996. Understanding Neural Networks andFuzzy Logic. Basic Concepts and Applications. New York,NY: the Institute of Electrical and Electronics Engineers,Inc.
Kasabov, N.K. 1996. Foundations of Neural Networks, FuzzySystems, and Knowledge Engineering. Cambridge, MA: TheMIT Press.
MathWorks. 1997a. MATLAB Image Processing ToolboxUser's Guide. Natick, MA: The MathWorks, Inc.
MathWorks. I997b. MATLAB Neural Network Toolbox User'sGuide. Natick, MA: The MathWorks, Inc.
MathWorks. 1997c. Using MATLAB. Natick, MA: TheMathWorks, Inc.
Meyer, G.E., T.G. Franti and D.A. Mortensen. 1997. Seek anddestroy. Resource 4: 13-14.
Nakano, K. 1997. Application of neural networks to the colorgrading ofapples. Computers and Electronics in Agriculture18: 105-116.
Schmoldt, D.L., P. Li and A.L. Abbott. 1997. Machine visionusing artificial neural networks with local 3Dneighbourhoods. Computers and Electronics in Agriculture16: 255-271.
Staff, J.V. and J.V. Benlloch. 1997. Machine-assisted detectionof weeds and weed patches. In Precision Agriculture '97.Volume II. Technology, IT and Management, ed. 1. V.Stafford, 511-518. Herndon, VA: SCI Bios ScientificPublishers.
Timmermans, A.J.M. and A.A. Hulzebosch. 1996. Computervision system for on-line sorting of pot plants using anartificial neural network classifier. Computers andElectronics in Agriculture 15: 41-55.
Wang, D., F.E. Dowell and R.E. Lacey. 1999. Single wheatkernel color classification using neural networks.Transactions ofthe ASAE 42: 233-240.
Yang, C.-c., S.O. Prasher, R. Lacroix, S. Sreekanth, A. Madaniand L. Masse. 1997a. Artificial neural network model forsubsurface-drained farmlands. Journal of Irrigation andDrainage Engineering 123: 285-292.
Yang, C.-C., S.O. Prasher and G.R. Mehuys. 1997b. Anartificial neural network to estimate soil temperature.Canadian Journal ofSoil Science 77: 421-429.
YANG. PRASHER. LANDRY. RAMASWAMY and DiTOMMASO
TECHNICAL NOTE
Mechanical dewatering of chopped alfalfa usingan experimental piston-cylinder assembly
S. SINHA 1, S. SOKHANSANJ', W. 1. CRERAR', W. YANG', L.G. TABIL1
, M.H. KHOSHTAGHAZA2 and R.T. PATIL3
IDepartment ofAgricultural and Bioresource Engineering, University ofSaskatchewan, Saskatoon, SK, Canada S7N 5A9; 2TarbiatModarress, Tehran, Iran; and 3CentralInstitute ofAgricultural Engineering, Bhopal, India. Received 10 December 1999; accepted31 May 2000.
Sinha, S., Sokhansanj, S., Crerar, W.J., Yang, W., Tabil, L.G.,Khoshtaghaza, M.H. and Patil, R.T. 2000. Mechanical dewatering ofchopped alfalfa using an experimental piston-cylinder assembly.Can. Agric. Eng. 42: 153-156. Experiments were conducted toinvestigate the quantity and quality of the chopped alfalfa pulp andjuice extracted in a piston-cylinder assembly. Single- and doublemacerated alfalfa chops at an initial moisture content of 75.7% (wetmass basis) were pressed under constant or increasing pressures up to20 MPa. The pressed pulp and the juice were analyzed for the contentsofdry matter, protein, and p-carotene. Approximately 52% of the totalwater content was removed at pressures higher than 4 MPa. Thepressure required to extract extra juice from already pressed pulpincreased exponentially. The juice had on average solids content of 1012% (dry mass basis). On a dry matter basis, there were no reductionsin either protein content or p-carotene in the pulp. Keywords: alfalfa,lucerne, drying, dewatering, cylinder-piston, pressure, moisturecontent, p-carotene, protein.
Des experiences ont ete faites pour etudier la quantite et la qualitede la pulpe et du jus extraits de luzerne hachee grace a un assemblagepiston-cylindre. De la luzerne hachee, ayant subi une macerationsimple ou une maceration double, et dont la teneur en eau initiale etaitde 75.7% (base humide), a ete pressee. La pression appliquee etaitconstante ou allait en augmentant jusqu'a 20 MPa. On a analyse lateneur en matieres seches, en proteine et en p-carotene de la pulpepressee et du jus. A des pressions superieure a 4 Mpa, on a reussi itextraire 52% de la teneur en eau totale. La pression necessaire itI'extraction de jus additionnel de la pulpe deja pressee augmente defa~on exponentielle. La quantite de solides contenus dans Ie jus etaiten moyenne de 10-12% (base seche). En terme de pourcentage dematiere seche, les teneurs en proteine et en p-carotene de la pulpen'ont pas diminue.
INTRODUCTION
Fresh alfalfa has a moisture content (m.c.) ranging from 70% to85% wet mass basis (moisture contents in this paper are givenin percent wet mass basis, unless otherwise specified) at thetime of harvest. The optimum moisture content for pelleting isabout 10%. The excess moisture is removed by partial sundrying in the field followed by artificial drying in a drying plant.Complete artificial drying at a high temperature (typically800°C) tends to preserve vitamins thus it is preferred to sundrying. Artificial drying, however, is more expensive than predrying in the field. Bad weather may damage the crop during
field-drying. Savoie et a1. (1995) showed that alfalfa macerationenhances drying rates in the field resulting in reduced dryingcosts.
Mechanical dewatering as a pre-drying alternative deservesinvestigation. Hibbs et a1. (1968) showed that mechanicaldewatering eliminated an average of 60% of water in thestanding crop. The percent protein in the dry matter of thedewatered alfalfa averaged 3.2% lower than that in the standingcrop. When the crop was left in the field to wilt, the percentprotein reduction averaged 1.8%. The loss of protein in thewilted material was more variable than mechanically dewateredalfalfa due to a variable leaf loss.
Holdren et a1. (1972) determined that the requireddewatering pressures ranged from 1.4 to 13.8 MPa dependingon the degree of crop conditioning. The duration of dwell timevaried from 0.5 to 10 min. The dry matter content of the juiceremained constant irrespective of the quantity of juiceexpressed. The juice from minced alfalfa contained about 50%more dry matter than juice from chopped alfalfa. Koegel et a1.(1973) performed three sets of experiments to study the effectof uniaxial force on the rupture of plant cells in alfalfa stemsand leaves. Tests done at 0.35-27.6 MPa showed that pressurealone was not responsible for cell rupture in stems and leaves ofalfalfa. Savoie and Beauregard (1990) studied the effects ofinitial moisture content, density, and number of macerations onthe juice quality of alfalfa. The proportion of juice extractedincreased from 10 to 21 % when the alfalfa moisture contentincreased from 80 to 84%. Double macerations increased theextracted juice by 13-16%.
Past research shows that compressive stress, dwell time,initial moisture content, and the degree of crop macerationaffects the amount of juice extracted from alfalfa. The reportedexperiments were generally performed by applying constantpressures on a fixed amount of material. For commercialapplication such as in a screw press, the material experiencescontinuously increasing compressive stress (Sinha 1995).Therefore, engineering data relating the varying pressures andthe quantity and quality of the pressed pulp are important.
The objective of this research was to determine the effect ofpressure on quantity and quality of fresh and pressed alfalfachops and juice.
CANADIAN AGRICULTURAL ENGINEERING Vol. 42. No.3 July/August/September 2000 153
RESULTS
The data from the series one experiment showed a poorcorrelation coefficient (-0.45) between the pressures applied andthe moisture content in the pulp. The mass of single-maceratedchops charged into the cylinder was 60.4 g (standard deviation,sd=0.2 g, n=14). The pressed pulp yield was 38.3 g (sd=1.8 g,n=14) and the juice yield was 19.1 g (sd=2.4 g, n=14). Themoisture content of the pulp was 70.7% with a large variation(sd = 10.0%, n=14). The reason for the large variation inmoisture content was the loss of juice in the hose and cylinderand a possible re-mixing of juice with the pulp at the end of apressure test.
Constant pressure testsTable I presents the data from the second series of tests in whichthe chopped fresh alfalfa was macerated twice. The quantity ofmaterial charged into the cylinder increased to 65 g. The yieldof juice increased progressively with increasing pressure. Thecorresponding moisture content of the pressed pulp decreasedto less than 63.3% when pressures exceeded 4 MPa.Approximately 52% of the total water content was removed.
Koegel and Bruhn (1972) used Eq. I to relate the finalmoisture content of pressed alfalfa to the applied pressure.
(1)Me =48.3P-{).l746
where:P = applied pressure (MPa), andMe = final moisture content (%wb).
alfalfa placed into the cylinder was increased to 65 g. Testswere conducted using nine target pressures: 2,4,6, 8, 10, 12,14, 16.5, and 18.6 MPa.
Series three
Approximately 65 g of double-macerated alfalfa chops wereplaced in the cylinder. The pressure on the charge sample wasincreased progressively. For example, the chops were pressedfrom 0 to 2 MPa and maintained at 2 MPa until all of theexpressed juice was collected. The pressed pulp was removed,placed in plastic bags, and stored in a freezer. The cylinder wasrefilled with the new material. The pressure was increased fromoto 2 MPa. Once the flow of juice stopped, without removingthe pulp, the pressure on the piston was increased to 4 MPa andmaintained at this pressure to collect the juice in a separate cup.Since a smaller amount of juice was expressed in the 2-4 MParange, the experiment was repeated five times in order toaccumulate adequate juice for analysis. This procedure wasrepeated for higher pressure ranges of 4-6, 6-8, and 8-10 MPaand the juice was stored. The collected juice from each pressurerange was frozen at -20°C and the pressed pulp was kept in asealed plastic container in cold storage at 4°C until furtheranalysis.
The pressed pulp, juice, and fresh alfalfa were analyzed forprotein, p-carotene, and moisture content. The SaskatchewanFeed Testing Laboratory, University of Saskatchewan,Saskatoon, conducted the protein and p-carotene analyses. Themoisture content of the fresh and pressed pulp was measuredfollowing the ASAE S358.3 oven method (ASAE 1997). Thedry matter content of the juice was measured by evaporating thejuice in the oven (104°C for 4 h) and weighing the remainingsolids.
Porous stone
Cylinder
Discharge opening
Plunger
50.00
Fig. 1. Schematic diagram of piston cylinder used forpressing alfalfa chops.
Base
MATERIALS and METHODS
Figure 1 is a schematic diagram of the piston-cylinder and theremovable bottom flange used in these experiments. Theinternal diameter and the height of the cylinder were 50 mm and114.3 mm, respectively. The piston had a clearance of 0.05 mmwith the cylinder wall. A 10 mm orifice at the bottom of thecylinder provided a drain. Three 50 mm diameter porous bronzediscs were placed on top of each other at the bottom of thecylinder to prevent the pressed pulp from draining out. A steelscreen was then placed on the top of these porous discs toprevent blockage.
Freshly cut alfalfa variety Beaver (fltSt cut) was placed inplastic bags and stored in a freezer (-20°C). To conduct theexperiments, the frozen alfalfa was thawed at room temperatureand hand cut into 50 mm pieces. The chops were macerated bypassing them once or twice through a pair of grooved rollers.The cylinder was filled to the top. The piston was pressed intothe filled cylinder using a motorized hydraulic press. Theadvancement of the piston was stopped when the flow of juicestopped or a target pressure was reached. Three series of testswere conducted.
Series oneApproximately 60 g of single-macerated alfalfa pieces wereplaced in the cylinder. Fourteen target pressures ranging from1 to 30.8 MPa (1.0,2.1,3.1,4.0,5.1,5.9,8.3, 10.2, 13.2, 14.5,16.5, 18.6, 20.7, 30.8 MPa) were tested. A material balanceshowed that some of the extracted juice was lost in the cylinderand inside the drain hose.
Series twoAlfalfa was macerated twice. A sponge was placed above thebase of the piston to absorb residual juice passing through thecylinder-piston clearance. A larger drain hose with a pinch cockwas used to prevent the back flow of juice. The quantity of
154 SINHA etaJ.
Table I. Mass balance and moisture content of pulp after pressingalfalfa at 75.7% m.c. in the second series of tests.
Table II. Average yield of pressed pulp and juice expressed fromalfalfa chops at 75.7% m.c. for various pressure ranges inseries three tests (standard deviations are in parenthesis).
Run Pressure Initial mass Pulp yield Juice yield Pulp m.c.no. (MPa) (g) (g) (g) (%wb)
I 2.1 65.9 39.4 18.9 66.12 4.1 66.2 39.9 23.9 65.53 6.2 65.6 34.5 27.2 61.84 8.3 65.7 37.9 23.7 63.35 10.3 66.0 32.4 31.0 60.36 12.4 65.6 31.9 30.7 60.57 14.5 65.3 32.3 31.1 61.58 16.5 65.5 31.3 29.8 60.09 18.6 66.0 32.4 31.6 61.1
Increasing pressure testsTable II lists the pressure ranges and the number ofsamples for each pressure treatment. For freshalfalfa, two samples (65 g each) yielded 20.8 g ofjuice each. For subsequent pressures, the yield ofjuice from pre-dewatered pressed pulp was small,ranging from 1.2 to 2.0 g. Four to six repetitionswere needed to collect a quantity of juice adequatefor analysis.
An exponential relation for the ratio ofpressure/moisture gradient (L\P/L\M) vs the initialmoisture content of the pulp was fitted to the datain Table II:
~M1 =7592 exp(-0.2279 Mo) (3)
where Mo =initial moisture content of the pulp(%wb). The units for L\P and dM are MPa and%wb, respectively. The regression coefficient forestimation (logarithmic transformation) is 0.94.Figure 3 shows the plot of Eq. 3 along with theexperimental data. The equation is extrapolatedand plotted to the left of the data. The curve showsan exponential increase in pressures required toreduce moisture content at lower levels ofremaining moisture in the plant material.
Table III lists the data for moisture content,protein content, and ~-carotene content from seriesthree. Fresh alfalfa at 75.7% moisture had 20.6%
crude protein and 18.9 mglkg (31473 international units, IUlkg)of ~-carotene. The first compression stage (0-2 MPa) reducedthe moisture content of the fresh alfalfa to 68.3% m.c. Thesecond compression range (2-4 MPa) reduced the moisturecontent of the pressed pulp to 65.9% m.c. Finally, the fifthcompression range (8-10 MPa) reduced the moisture content ofthe pressed pulp to 63.8%. The total drop in moisture contentduring five stages of compression was 11.9% points.
Crude protein content of the pulp varied from approximately22.2 to 24.3% on a dry mass basis (db). The protein content inthe juice varied from 14.2 to 18.1 % (db) indicating that much ofthe protein remained in the pulp. The ~-carotene content in thepulp (26.4-33.6%) was also higher than that in the juice. Thehighest ~-carotene content in the juice was extracted at thecompression range of 0 to 2 MPa. ~-carotene content of thejuice extracted at the compression range of 8-10 MPa was 9.8mglkg, indicating that most of the ~-carotene had been leachedout of the plant material at a lower pressure.
20.8 (5.3)2.0 (0.4)1.7 (0.2)1.2 (0.7)1.6 (0.2)
Juice mass(g)
20
•
Eq.2
•
5 10 15Pressure (MPa)
o
68
60
58+--------------~
Pressure range No. of Fresh charge Pulp mass(MPa) samples (g) (g)
0-2 3 65.5 (0.3) 36.4 (1.5)2-4 5 65.6 (0.3) 32.0 (0.4)4-6 6 65.5 (0.3) 31.9 (0.7)6-8 5 65.6 (0.3) 31.3 (1.0)8-10 4 65.7 (0.2) 31.7 (0.4)
66
if 64........CD~62
Fig. 2. Equilibrium moisture content of the pressed pulpvs the applied pressure.
Equation I predicted a lower moisture content for the pressedpulp than the experimental value of the present study. Thismight have been due to excessive maceration of alfalfa samplesby Koegel and Bruhn (1972). Data in Table I yielded:
Me =68.4 p-{).0442 (2)
The regression coefficient for the linearized form of Eq. 2(logarithmic transformation) was R2 =0.79. Figure 2 shows theplot of Eq. 2 and the experimental data.
DISCUSSION
Previously published research, as well as the results of thepresent work on alfalfa juice extraction, revealed severalimportant points. The degree of maceration prior to pressing hasa large effect on the quantity and quality of juice extracted.Double-maceration increased the amount of charged material inthe cylinder from 60 to 65 g. The fluctuations in the moisturecontent of the pressed pulp were higher in single-maceratedchops than the double-macerated chops. Pressures up to 4 to 6MPa reduced the moisture content of the fresh alfalfa from 75.7to 60-65%. Higher pressures did not increase the yield of juice.Dry matter content of the juice was in the range of 10 to 12%.
CANADIAN AGRICULTURAL ENGINEERING Vol. 42, No.3 JulylAugust/September 2000 155
Table III. Analysis of pressed pulp and juice expressed at various pressure ranges inseries three tests. Fresh alfalfa moisture content, 75.7%; crude proteincontent, 20.6%; ~-carotene, 18.9 mglkg (31500IUlkg).
ACKNOWLEDGMENTS
The research was supported by the Natural Sciences andEngineering Research Council of Canada and by theSaskatchewan Agriculture and Food ADF program. TisdaleAlfalfa Dehy Ltd., of Tisdale, SK provided the test material.
CONCLUSIONS
27.621.521.79.79.8
Jiuce~-carotene
(mglkg)
18.115.614.214.915.3
Juiceprotein
(%)
REFERENCES
ASAE. 1997. ASAE S358.2, Moisture measurement - Forages.In ASAE Standards 44th Edition, 557. St. Joseph, MI: ASAE.
Hibbs, I.W., R.H. Conrad and W.H. Johnson. 1968. Macerated,dewatered vs. wilted alfalfa-grass silage for dairy cows.Research Bulletin 1013:3-9. Wooster, OH: OhioAgricultural and Development Center.
Holdren, R.D., W.L. Harris and GJ. Burkhardt. 1972.Squeezing juice from forage. Transactions of the ASAE15(6): 1044-1048.
Koegel, R.G. and H.D. Bruhn. 1972. Pressure fractionationcharacteristics of alfalfa Transactions of the ASAE 15(5):856-860.
Koegel, R.G., V.I. Fomin and H.D. Bruhn. 1973. Cell ruptureproperties of alfalfa. Transactions of the ASAE 16 (4):712716.
Kraus, TJ., K.G. Koegel, RJ. Straub and KJ. Shinner. 1997.Leachate conductivity as an index for quantifying level offorage conditioning. ASAE Paper No. 97-1100. St. Joseph,MI:ASAE.
Savoie, P. and S. Beauregard. 1990. Potential in the field foragejuice extraction. ASAE Paper No. 90-1055. St. Joseph, MI:ASAE.
Savoie, P., M. Pouliot and S. Sokhansanj. 1995. Potentialimpact of mowing-maceration on an alfalfa dehydratingplant. Canadian Agricultural Engineering 37(4):295-304.
Sinha, S. 1995. Mechanical dewatering of chopped alfalfa.M.Sc. thesis. Department of Agricultural and BioresourceEngineering, University of Saskatchewan, Saskatoon, SK,Canada.
I. Double maceration ofalfalfa chops increased theamount of charged freshalfalfa in the cylinderhousing; it also reducedfluctuations in the yield ofpressed pulp and juice.
2. About 52% of the juicemass was extracted byapplying pressures greaterthan 4 MPa. The amount ofpressure required to extracta unit of moisture increasedexponentially with decreasing available moisture inthe pre-dewatered pulp.
3. The juice had a minimum of 10% dry matter content.4. On a dry matter basis, the contents of protein or ~-carotene
in the pulp did not decrease as compared to these contentsin fresh alfalfa.
10
- 8~
~ ,/ Eq.3Q. 6:E........:E 4~
it2~
Stage Pressure Pulp m.c. Pulp crude Pulp Juice dryno. range protein ~-carotene matter
(MPa) (%wb) (%) (mglkg) (%)
I 0-2 68.3 23.1 32.1 12.02 2-4 65.9 24.3 30.6 10.53 4-6 65.0 22.2 33.7 10.54 6-8 64.4 22.4 26.4 10.45 8-10 63.8 23.7 29.4 10.4
o-L-- ~---,:===::!r===;
60 64 68 72 76 80Pressed pulp m.c. (% wb)
Fig. 3. Pressure/moisture content gradient vs moisturecontent of the pressed pulp.
The present work, however, has several shortcomings thatshould be considered in the planning of future experiments.I. The alfalfa chops were frozen at -20°C for storage. The
samples were then thawed prior to a pressure test. Thefreezing and thawing cycle might have affected the state ofthe water in the cellular structure of the plant material. It isimportant to repeat the tests with fresh chops withoutfreezing.
2. All tests were done on one sample of field-cut alfalfa chopsat a moisture content of75.7% (wb). Future tests should beperformed on larger samples of wider initial moisturecontents.
3. The degree of maceration prior to pressing was difficult toquantify, yet maceration had a large effect on mechanicaldewatering. A method should be developed to quantify thedegree of maceration.' Kraus et al. (1997) proposed anelectrical conductivity probe for quantifying level of forageconditioning.
4. Unfortunately, we did not record either the rate of travel forthe piston or compression ratios. These data are importantfor scale up designs.This research showed that maceration is an important factor
to the mechanical dewatering of alfalfa. It also showed that withpressures just greater than 4 MPa, more than 50% of the watercontent of the alfalfa can be removed.
156 SINHA et at.
NOTES TO CONTRIBUTORSThe ~ditorial Boar~ w~ll a~sess suitability and essential detail ofpaperssubnutted for pubbcatlon 10 Canadian Agricultural Engineering. Oneor more ~eviewers will.be used. Their comments and suggestions willbe compiled and submitted to the author. The review will ensure that:
1. A research paper presents a piece of research carried to awell-defined stage ofadvancement and the conclusions are adequatelysupported by the experimental results.
2. A technical paper presents a clear, concise, and factual outlineand ~nterp~tation of the.d~velopment, design, test, or analysis underconSIderation and that It IS a contribution in one of the fields ofagricultural, biosystems, or food engineering.
3. A general paper on education, research, or extension is pertinentto major changes in curriculum, research, or extension or to forwardlooking developments in these areas.
4. A technical note on equipment development, technique of measurement, or method of analysis will have an application for otherworkers in the fields of agricultural, biosystems, or food engineering.
MANUSCRIPT
The manuscript should be typed double-spaced on paper 216 x 279 mm(8.5" x II") with margins not less than 30 mm. The first page shouldcontain only the title, authors' names, addresses (including postalcodes), and contribution number where applicable. The telephonenumber, FAX number, and E-mail address of the corresponding authorshould also be included. Tables and captions for illustrations should beon separate pages, placed after the text. Manuscript paper withnumbered lines is preferred. Five copies are required. After a paper hasbeen accepted for publication, the author will be expected to providea copy of the paper in electronic form in a format compatible withMS-DOS systems.
The title of the paper should give an accurate description of thearticle, using key words that can be used for computer-indexing.
ORGANIZATION
The paper should be organized to conform with present Journal practice. See Norum and Jayas (1995). All papers must include a shortabstract section ofabout 200 words. Authors are encouraged to submitthe abstract in both English and French. There will be a charge fortranslation services that must be provided by the Journal.
Major headings - Center on page with all words in capital letters.
Subheadings - Start at left-hand margin, capitalize first letter.
Sub-subheadings - Same as subheadings but underline.
Technical and detailed information should be included only in theform of description, table, graph, chart or photograph. In general,follow the style given in Norum and Jayas (1995).
References
List references alphabetically by authors at the end. Follow the formatset by Norum and Jayas (1995). Material in press, with the name of thejournal, may be used as a reference. Private communications andunpublished reports should be referred to in parentheses in the text.Private communications should include the person's title and address.Avoid the use of footnotes. Use the author-date system in the manuscript when referring to articles in the Reference section.
Tables
Designate tables at the top by table number (Roman numerals) andtitle, in upper and lower case letters. All headings and other information in tables are to be in lower case except first letter of first word.Keep the table compact and place it across the page wherever possible.Do not use vertical lines.
Measurements
Only metric system (SI) units are to be used.
Equations
Equations and formulas must be set up clearly. Use capitals for symbols ~s much as possible and lower case for superscripts andsubsc,:pts. Greek and other characters should be identified clearly.Equations should be numbered on the right-hand margin and in linewith the center of the equation.
Abbreviations
Abbreviate units of measure only when used with numerals. Usecorrect SI unit abbreviations. Do not use abbreviations in the title.
ILLUSTRATIONS
Either original drawings or glossy photographs are acceptable forillustrations. An illustration should be planned to fit, after reduction,into a space 90 mm wide (one column) or 183 wide (two columns).The original should be not more than three times the size of the finalfigure. For identification, the figure number and author's name shouldbe written on the lower left corner with soft pencil.
Lin~ drawings should be machine produced on white drawing paperor tracmg paper. Authors are encouraged to produce drawings usingone of the commonly used computer packages. Letters, numerals,labels and axis captions should have only the first word capitalized.Axis captions should be followed by a comma, the symbol in italics,and the units in parentheses [Le. Acceleration of particle, Ap (m/s2
)].
If a symbol is not used, omit the comma. Letters and numerals must beat least 1.5 mm high and preferably 2 mm high in final form. Curveson graphs must be 0.3 mm wide after reduction. Axes lines should beclearly visible but inconspicuous; a width of 0.2 mm after reduction issuggested. Figure numbers and captions should be typed on a separatepage, not on the original illustrations. When a paper is submitted forpublication, the original illustrations need not be provided so long asthe copies are of such quality that reviewers can understand them.Original drawings must be provided when the paper is accepted forpublication. If a drawing has been produced by a computer package,a copy of the file should be submitted on disk at the time that themanuscript is submitted on disk (JPEG format preferred).
DISCUSSIONS
Discussions may be submitted on any paper or technical note publishedin the Journal for a period of not more than four months followingpublication. Discussion of a paper or technical note is open to anyonewho has significant comments or questions about the content of thepaper/technical note. A discussion will not be accepted for publicationif it contains material readily found elsewhere, is purely speculative,introduces personalities, or otherwise falls below the standards of atechnical paper in a professional journal. Authors will be given anopportunity to reply to discussions.
The format for discussions differs from those of papers in thatfigures are to be identified by capital letters to avoid confusion withthe original paper. The discusser should refer to himlherself as "thewriter" or "I" and to the author of the original paper as "the author."The first page shows the title of the original paper with a footnote toidentify the author, volume, page, and date. Name and address of thewriter of the discussion follow the title.
Discussions will be reviewed by the Editorial Board and possiblythe reviewers of the original paper. The length of a discussion isrestricted to one journal page. Lengthy discussions will be returned forshortening, or the writer may be encouraged to submit a paper ortechnical note.
REFERENCE
Norum, DJ. and D.S. Jayas. 1995. Instructions for preparing a paperfor Canadian Agricultural Engineering. Canadian AgriculturalEngineering 37(3):239-243.