Pertanika 3(2), 108-112 (1980)

Efficacy of Methyl-eugenol as Male Attractant forDacus dorsaiis Hendel (Diptera: Tephritidae).

A. GHANI IBRAHIM and A. GHAXI HASHIMDepartment of Plant Protection, Faculty of Agriculture, Universiti Pertanian Malaysia,

Serdang, Selangor, Malaysia

Keywords: Methyl eugenol; Fruitfly; Dacus dorsaiis; Carambola Orchard.

RINGKASANDacus dorsaiis ialah spesies lalat buah yang terbanyak (99.8%) di dusun belimbing besi. Tahap

populasi didapati berkorelasi (r =? 0.56) dengan kadar hujan. Buah belimbing ntula diserang apabila umurmeningkat 29 hari. Kajian di makmal tentang daya penarik methyl eugenol terhadap tiga spesis lalat buahntenunjukkan tidak terdapat perbezaan yang bererti diantara Dacus dorsaiis dan D. umbrosus tetapi per-bedzaan yang bererti (P < 0.05) didapati diantara kedua-duanya dengan D. cucurbitae. Methyl eugenoljuga didapati mempunyai daya penarik yang kuat terhadap lalat buah dewasa Dacus dorsaiis yang daraberumur lebih daripada 20 hari.

SUMMARY

Dacus dorsaiis Hendel is the predominant species of fruit-fly (99.8%) found in a carambola orchard.Population level was found to be correlated (r = 0.56) with rainfall. Infestation of fruits began 29 daysafter fruit set. Laboratory studies show that among the three species of fruit flies tested for attractiveness tomethyl eugenol, male ofD. dorsaiis and D. umbrosus showed no significant difference but both are significantlydifferent (p < 0.05) with D. cucurbitae. Virgin adult males of D. dorsaiis more than 20 days old weregreatly attracted to methyl eugenol.

INTRODUCTION

The tephritids are destructives pests of bothtropical and subtropical fruits. There are approx-imately 4000 known species within the familyTephritidae (Christenson and Foote, 1960). Themost serious pests of agricultural crops are Dacusdorsaiis Hendel which are injurious to fleshyfruits like carambola (Averrhoa carambola), guava(Psidium guava), mango (Mango indica) andpapaya (Carica papaya). Hardy (1973) reportedthat D. dorsaiis are widely distributed in thetropics. Several workers have described indetail the biology of D. dorsaiis. (Shah et al,1948; Janjuna, 1948). In Peninsular Malaya,Corbett (1928) first recognised the importanceof D. dorsaiis as a major insect pest of orchards.Biological studies of this particular pest con-ducted in Malaysia were those by Miller (1940),Ibrahim and Kudom (1978) and Ibrahim andMohamad (1978).

In the field, the abundance of D. dorsaiisis due to many factors such as climate and host-

plants. For example, temperatures influencethe reproductive rate of the adult fruit-fly (Lee,1976). In Malaysia the temperature and relativehumidity are relatively constant throughout theyear. Rainfall has its seasonal pattern but theeffect of rain on the pest population is littleknown.

Various methods have been adopted incontrolling fruit-flies, among which, are the useof specific chemical attractants. Methyl eugenol,an attractant, is being used extensively for con-trolling D. dorsaiis. Its usefulness was realisedwhen Howlett (1915) recognized methyl eugenolas one of the main constituents of citronella oilwhich then attracted D. diversus and D. zonatus.To-day, methyl eugenol has been extractedfrom diverse plants. (Kawano et al, 1968;Fletcher et al, 1975 and Shah and Patel, 1976).The use of methyl eugenol for annihilation ofD. dorsaiis was successfully adopted by Steiner(1952). Ever since then, various formulationsof methyl eugenol and insecticides have beentried under different climatic cond itions.

Key to authors' names: Ibrahim, A.G. and Hashim, A.G.

lUo

A. G. IBRAHIM AND A. G. HASHIM

(Cunningham et al, 1978). The use of methyleugenol has an advantage in that insect pests couldbe eradicated from an area with the minimumamount of insecticides (Steiner et al, 1965). InMalaysia, the attractant has been used on alimited scale.

The present study was conducted to evaluatethe effectiveness of methyl eugenol in attractingfruit-flies, especially D. dorsalis. This behavioralstudy in relation to the development of thecarambola fruits and rainfall pattern would beuseful in the integrated control of Dacus species.

MATERIALS AND METHODSField and laboratory studies were conducted

to assess the efficacy of methyl eugenol on fruitflies. A trial was conducted in a carambolaorchard ca. 1.5 ha at Serdang Baru, Selangor.The experimental site was ca. 1500 m2 situatedalmost in the centre of the orchard. The treesof ten years old were in the fruit bearing stage.Laboratory trials were conducted in room atambient temperature (28° ± 2°C).

Trial 1: Field trapping of Dacus spp.This study was initiated on 16th May 1978

for a period of nine months. Plastic traps(10cm x 10cm) with circular openings mea-suring 2.4 cm in diameter at both ends wereused for trapping the fruit-flies. This roundhole trap design was adopted for they wereeffective in trapping fruit-flies. (Ibrahim et al,1979). A total of nine traps were placed atstrategic positions in the orchard (Fig. 1). Each

Fig. 1: Placement of traps containing baits inthe experiment plot of Carambola orchard.X is the plant.

trap was baited with a mixture of 0.5 ml methyleugenol, 0.5 ml of Malathion EC 80 and 2 mlof sucrose solution soaked in cotton rolls. Thetraps were hung to the branches of trees at aheight of ca 1.2 m from the ground. At thechosen height, there was no effect on captureof fruit flies (Hooper and Drew, 1979). Collectionof the fruit flies and recharging of the poisonedbaits were made every 4th day between 4 - 5 pm.The flies were sexed and identified.

Trial II: Laboratory studiesThree different species of fruit-flies viz:

D. dorsalis, D. cucurbitae Coq and D. umbrosusFabr were reared from infested fruit of carambola,cucumber and jack-fruits. The third instarslarvae were allowed to pupate in nylon-meshedcages (82 cm x 66 cm x 66 cm) filled with sandto a depth of 5 cm. The newly emerged adultswere provided with water, sugar solution (10%)and protein hydrolysate. Thirty male fruit-fliesof the same species which had been kept incaptivity with females for ten days were testedfor their response to methyl eugenol. A totalof 90 male flies belonging to three differentspecies were released in a perspex cage(lm x lm x lm). A small trap (8 cm x 12 cm)of similar shape to the field trap was used ineach cage. The trap was baited with three dropsof methyl eugenol, one drop of Malathion and1 ml of sucrose solution. Recordings were madeat hourly intervals for four consecutive hours onall the tested species of the fruit flies.

In a further trial to evaluate the stage ofadult D. dorsalis attracted to methyl eugenol,the fruit-flies were reared using artificial diet(Tanaka et al, 1969). Twenty virgin male fliesof varying ages of 4, 8, 12, 16 and 20 days wereplaced in separate cages containing methyleugenol solution, malathion and sucrose solution.The number of flies caught in the traps wererecorded at hourly intervals for four hours.Both laboratory trials were replicated four timesusing the Completely Randomized Design. Theresults were analysed and the means wereseparated by the Duncan Multiple Range Tests.

RESULTS AND DISCUSSION

During the period of nine months (May1978 to January 1979) the total number of fruit-flies caught was 36,035. The dominant fruit flieswere the males of Dacus dorsalis. (Table 1).Though Z). pedestris (Bezzi) are known to attackcarambola fruits the dissection of the femalegenitalia failed to show their presence in thetraps. The number of female flies caught was

109

METHYL-EUGENOL AS MALE ATI R ACT ANT FOR DACVS DORS A US

extremely low. This rinding was similar tothat of Steiner et aLy (1965) who showed thatmethyl eugenol rarely attract females. BesidesD. dorsalis, the other two species caught wereD. umbrosus and D. cucurbitae, though the lasttwo species were significantly (P<0.05) few innumber.

TABLE 1

Fruit flies, Dacus spp attracted to methyl eugenolin Carambola orchards from 16.4.1978 - 19.1.1979

D.

D.

D.

Species

dorsalis

umbrosus

cucurbitae

Sex

MaleFemale

MaleFemale

MaleFemale

Total Nos

35,95976

262

55

Av. catch/trap/month

500.47

0.38

0.13

Analysis showed that there was no signi-ficant difference in response to methyl eugenolin the laboratory between ten-day-old males ofD. dorsalis and D, umbrosus; but, when com-pared to D. cucurbitae, their attraction to methyleugenol was found to be significantly different(Table 2). The percentage of D. dorsalis, D.umbrosus and D. cucurbitae caught were 43.3%,47.5% and 4.2% respectively. The higher catchof D. dorsalis in the field could be due to theirabundance in the open for the laboratory studyshowed D. dorsalis and D. umbrosus were equallyattracted to methyl eugenol. The smaller num-ber of D. umbrosus caught could possibly be dueto the plant specificity since this species is notknown to attack carambola but is specific toArtocarpus spp. Less than 50% of the totalnumber of flies from each species were attractedto methyl eugenol even after four hours. Dissec-tion of the male flies showed them to have welldeveloped testes which suggests that a consider-able number of flies are indifferent to methyl

eugenol. This suggestion conforms with thatof Umeya et at., (1973).

The fruit-flies showed marked seasonalfluctuations with the peak periods in Septemberand early December following fruit-set (Fig. 2).Bateman (1973) reported an increase of fruit flypopulation at the onset of fruit ripening. Afterall the fruits had been bagged, there was a declinein the population. This could possibly be dueto absence of fruits for oviposition which subse-quently resulted in a reduced fly population inthe field. Observation of fruits in the fieldshowed that the fly oviposited as early as 29 daysafter fruit set. The flies prefer to oviposit onripe fruits but in their absence they oviposit ongreen fruits. Lee (1976) found similar resultswith young papaya fruits.

The relationship between total catch andrainfall was found to positively correlated (r =0.56). The population was observed to be signi-ficantly high (P < 0.05) only after continuousdaily rainfall. Lee (1976) observed a similarpattern when a sunny day precedes several coldand wet days.

The placing of the traps also influences theamount of fruit-flies caught. Traps placed atthe periphery of the orchard had higher countsthan those placed in the centre. This suggeststhat peripheral traps had better access to windmovement and methyl eugenol is capable ofattracting flies at a distance of 0.6 km. (Steiner,1952).

The attraction of D. dorsalis to methyleugenol is relative to the age of the flies (Table 3).The attraction was greatest when the flies were20 days old. Since methyl eugenol is a sexattractant, it will attract flies of a specific physio-logical age. Umeya et aL, (1973) observed thatmale fruit flies were not attracted to methyleugenol until the ninth day, suggesting that

TABLE 2

Laboratory study showing cumulative number of fruit flies attracted to methyl eugenol

Dacus species

D. dorsalis

D. umbrosus

D. cucurbitae

1 hr.

29

38

1

Tot. no of flies

2 hr.

38

45

3

caught within:

3 hr.

46

49

4

4 hr.

52

57

5

o

43.3 a

47.5 ab

4.2 c

Means followed by the same letters are not significantly different at 5% level as determined by Duncan's MultipleRange Test.

no

A. G. IBRAHIM AND A. G. HASHIM

<cc

< Fruit f

• Rainf

150

MAY

A . ' . - -•

i\\ ; V V

25 30 35AUG SEPT

40 45 50OCT ' NOV DEC I JAN

YEAR;1978 YEAR;1979Fig. 2: Fluctuations of fruitfly population in relation to tree phenology and rainfall.

TABLE 3Laboratory study showing cumulative numbers of virgin D. dorsalis attracted to methyl eugenol

Age (days)

4

8

12

16

20

n

80

80

80

80

80

I hr.

4

4

14

38

67

Tot. No. of flies

2 hr.

5

8

21

44

71

caught within:

3 hr.

5

0

25

47

72

4 hr.

5

0

25

47

72

6% a

10% ab

31% c

58% d

98% e

Means followed by the same letters are not significantly different at 5% level as determined by Duncan's MultipleRange Test.

sexual maturity may play a prominent role. Theknowledge of physiological ages of flies whichare responsive to the attractant is important in acontrol programme where sex attractants areused.

CONCLUSION

Dacus dorsalis is the main pest of carambolafruits. The fruit-fly population positively corre-lated (r = 0.56) with the total amount of rainfall.The limited number of female flies in the fieldtraps using methyl eugenol failed to indicate the

presence of D. pedestris. The laboratory studiesshowed varying responses of fruit-flies to methyleugenol. Male of D. dorsalis and D. umbrosuswere equally attracted to the sex attractant butD. cucurbitae was less attracted to the chemical.Evidently, with virgin males of D. dorsalis theresponse to the attractant increased with theage of the flies.

ACKNOWLEDGEMENT

The authors wish to express their sincerethanks to Messrs Abd. Rahman Mohamad andM. Subramanian for their assistance in the field.

I l l

METHYL-EUGENOL AS MALE ATTRACTANT FOR DACUS DORSAL1S

REFERENCES

CORBETT, G.H. (1928): Division of Entomologyannual report. Malay. Agric. J. 17: 276.

BATEMAN, M.A. (1973): "Fruit Flies" in Studies ofbiological control Delucchi. V.L. (Ed.) p 11-49.London. Cambridge University Press.

CUNNINGHAM, R.T., NAKAGAWA, S., SUDA, D.Y. andURAGO, T. (1978): Tephritids fruit fly Trapping.Liquid food baits in high and low rainfall climates.J. Econ. Entomol 71: 762-763.

CHRISTENSON, L.D. and FOOTE, F.H. (1960): Biologyof fruit flies. Ann. Rev. Entomol. 5: 171-192.

FLETCHER, B.S., BATEMAN, M.A., HART, N.K. andJ.A. LAMBBRTON (1975): Identification of a fruitfly attractant in an Australian plant. Ziera smithiion methyl-eugenol. J. Econ. Entomol. 68(6): 815-6.

HARDY, D.E. (1973): The fruit flies (Dipt): Tephri-tidae, on Thailand and bordering countries.Pacific Insects. Monograph 31: l-2pp.

HOOPER, G.H.S. and DREW, R.A.I. (1979): Effect ofheight of Trap on capture of Tephritid Fruit flieswith cuelure and methyl eugenol in differentenvironments. Environmental Entomol. 8(5): 786-788.

HOWLETT, F.M. (1915): Chemical reaction of fruitflies. Bull. Entomol. Res. 6: 297-305.

IBRAHIM, A.G. and K. GUDOM, F, K. (1978): Biologyof Oriental fruit fly Dacus dorsalis Hendel withemphasis on larval development. Pertanika 1(1):55-58.

IBRAHIM, A.G., SINGH, G., KING, H.S-K. (1979):Trapping of fruit-flies, Dacus spp (Dipt: Tephri-tidae) with methyl eugenol in Orchards. Pertanika2(1), 58-61.

IBRAHIM, Y. and MOHAMAD, R. (1978): Pupal distri-bution of D. dorsalis Hendel in relation to host

plant and its pupation depth.66-69.

Pertanika 1(2),

JANJUNA, N.A. (1948): The biology of Dacus ferru-gineus (Tephritidae: Diptera) in Baluchistan.Indian J. Ent. 10(1): 55-61.

KAWANO, Y., MITCHELL, W.C. and MATSUMOTO, H.(1968): Identification of the male Oriental fruit-fly attractant in the Golden Shower blossom. J.Econ. Entomol. 61(4): 986-8.

LEE, H.S. (1976): Investigation on the oriental fruit-fly D. dorsalts Hendel domaging papaya. J. AgricRes. of China. 25(2): 156-162.

MILLER, N.CE. (1940):38(3): 112-21.

Fruit-flies. Malay. Agric. J.

SHAH, A.H-, PATEL, R.C (1976): Role of tulsi plant,Ocimun sanctum in control of mango fruit fly,D. correctus Bezzi. Current Sci. 45(8): 353-4.

SHAH, M.J., BATRA, H.N., RANJHEN, P.C. (1948):Notes on biology of Dacus ferrugineus Fabr. andother fruit-flies in north-West frontier province.Ind. J. Ento. 10: 249-66.

STEINER, L-F. (1952): Methyl eugenol as an attractantfor oriental fruit-fly. J. Econ. Entomol. 45(2):241-48.

STEINER, L.F., MITCHELL, W.C., HARRIS, E.J., KOZUMA,T.T., FUJIMOTO, M.S. (1965): Oriental fruit flyeradication by male annihilation. J. Econ. Entomol-58(5): 961-4.

TANAKA, N., STEINER, L.F., OHINATA, K., OKAMOTO, R.(1969): Low cost larval rearing medium for massproduction of Oriental fruit flies. J. Econ. Entomol.62(4): 967-68.

UMEYA, K., SEKIGUCHI, Y., USHIO, S.Z. (1973); Thereproductive ability of D. dorsalis Hendel andthe response of adults to methyl eugenol. JapaneseJ. of App. Entomol and Zoology. 17(2): 63-70.

(Received 19 June 1980)

W. H. W. SULAIMAN AND G. C VANSTEENKISTE

characterization. This study aims at applyingtheir technique in the identification of simplestructures for the soil water characteristic anddiffusivity which are necessary in solving thesoil water flow equation.

THEORETICAL FRAMEWORK

General approachThe pattern recognition approach, in simple

terms, involves a comparison of the patternemerging from certain operations on the dataset coming from the system to be modelled withan input library of patterns from known"candidate" models. A choice is then madeamong the candidates as to which structure isbest adapted to the data.

The procedure is illustrated in Fig. 1. Twostages of operation are distinguished; first is thetraining of the classification algorithm (switchesin position I) and second is the use of the classifier(switches in position II). In the first stage, anumber of candidate models are proposed. A"feature extraction'* procedure is performed onartificial data generated from these models. Theresulting "feature space" is then classified intopartitions corresponding to different modelsWhen the training is complete the classificationalgorithm or pattern recognizer is coupled to thedata being investigated (second stage) and themost suitable model selected. The next step isthen to identify the parameters of the chosenmodel.

CLASSIFICATION

IN I B

FEATURE SPACB

DECISION

SELECT

A HOnKL

Fig. 1. Structure determination by the PatternRecognition Approach of Vansteenkiste,Bens and Spriet (1978a).

Feature ExtractionFeature extraction is the most crucial part

of the method. Features or characteristic ex-pressions have to be defined, on the basis ofwhich a choice among the different models will

be made. No hard and fast rules exist in thechoice of the features. However, they should berather insensitive to noise besides having gooddiscriminating properties. Very often one featureper proposed model is used, but sometimes morefeatures can provide better discrimination. Thedifferent features form a so-called feature spacein which most of the characterization operationsare performed.

Consider the data set D^= j (ti, xt) j i = 1,2,...., n } coming from the experiment or simu-lation of the experiment. Feature extraction canbe regarded as a mapping of the set D of allpossible data sets to a feature space:

9 : D ->- R* : D ->- f

where k is the number of features and f is a pointin the feature space. The mapping 0 can besingle valued, i.e., a single point f is the imageof a set D (Simundich, 1975), or multiple valued,i.e., more than one point f is the image of D, asin the current approach of Vansteenkiste, Bensand Spriet (1978a). The advantage of themultiple-valued mapping is that a poor orerroneous measurement will not destroy theinformation present in the other measurements.

Classification of the Feature Space

During the training stage a "classifier"splits up the feature space into partitions in anoptimal way, each subset corresponding to acluster of points and hence to a candidate(proposed) model. To achieve this severalsimulation runs have to be made with eachmodel. Feature points derived from these simu-lations are used as input to the classifier, whoseparameters are then adjusted iteratively so as togive maximal correspondence between input andoutput, the latter being the various partitions orsubsets of the feature space. It is, thus, evidentthat the more simulation runs with each of thecandidate models and the more diverse theseruns are, the better would the classificationalgorithm be.

For the discussion and development of theclassifier one is referred to a number of hand-books (Nilsson, 1965; Young and Calvert, 1974).Kanal (1974) provides an excellent survey of thedifferent methods available while Vansteenkiste,Bens and Spriet (1978) discuss the variousproblems of choosing a classifier algorithm. Thelatter authors emphasized that where it is deemedfit, as in a two or three-dimensional feature space,visual classification can be the most convenientmethod.

STRUCTURE IDENTIFICATION OF SOIL HYDRAULIC PROPERTIES

TABLE 1

Some empirical formulae used to represent the water content-pressure head relationship

Formula

Model 1: S =

Model 2: S =

Model 3: S -

a + ( l n | h | > 0

8

8 + |H|

— 1 » h < h ah J

, h > ha

Source

Haverkamp et al. (1977)

Haverkamp et al. (1977)

Brooks and Corey (1964)

(1)

(2)

(3)

S is the dimensionless water content given by S = (0 — Or)/(0S — Or) where 0s and 0r are the saturation andresidual water contents respectively-

STRUCTURE CHARACTERIZATION OFTHE SOIL WATER CHARACTERISTIC

Candidate Models

In most instances, the soil water charac-teristic or the 0-h relationship yields an S-shapedcurve, although it is not uncommon to find arelationship which exhibits two or more pointsof inflection. Some of the empirical relationshipsthat have been used are listed in Table 1. Theseform the candidate models in the structureidentification.

Feature Extraction

An examination of the candidate modelsreveals little except that the 0-h curve tends tobe rather flat at the low and high ends. Althoughthe first derivative has an important physicalsignificance (this being the specific watercapacity), it offers no immediate help in thecharacterization process. We thus resort to theparameters of the models for deriving features,since in the ideal cases, these parameters areinvariant for each model. Their variabilitywould be a measure of deviation from the modelunder consideration. In the present identifi-cation process only one parameter per proposedmodel is used to derive features.

Feature 1 :

As h is negative, operations are made easierby using suction head = —h, so that Model 1now reads

aS =

a + (1

Differentiating with respect to \p we obtain

dS S2j8 B_x

— (In*/* (5)d

Elimination of a from (4) and substituting it in(5) yields

d S * I n *fi =

d

dS h l n j h lj = (6)

d * S ( l - S ) dh S ( l -S )

Thus, for any point in the data set ft may beobtained by the use of Eq. (6) and the centraldifference approximation dS/dh = (S i+ t — Si- t)/2Ah. Now, two random points, A and B withcoordinates (hA, SA ) and (hs, SB) respectively,are taken from the data set (simulated or experi-mental) and feature 1 or the first coordinate ofthe point corresponding to the chosen (A,B)-tupleis computed according to:

PA

— (7)

Hence, fj provides us with a measuK; of thevariability of parameter ft. The procedure isrepeated for different (A,B)-tuples providing aset of values corresponding to the projections onone of the axes of the feature space.

Feature 2 :The second feature is obtained by resorting

to Model 2. Again, differentiating S with respectto h or * and solving for y yields

- h dS

" S(l-S) dh(8)

J

W. H. W. SULAIMAN AND G. C. VANSTEENKJSTE

Following the same procedure as with the first, Since computation of the second set ofthe second feature, f2, is found by taking the parameters involves very little extra effort, threeratio of the y's at the same two points, A and B more features can be easily defined. These areused earlier. Thus,

« A „ 5A , h a A7A

(9)

Feature 3 :

Model 3 is now used to derive feature 3 orthe third component of the point in the featurespace. Differentiating and solving for A yield

h dSA = • , h < h a

S dh(10)

Feature 3 is then calculated using points A and B

f3 = - , h < ha (ii)

Complications arise for h > ha because now thefunction has a constant value (i.e., equal to 1)and, therefore, A is zero. One way of over-coming this dilemma is to define an alternativefeature f *3 specific to this region according to

' h (lla)

Note: If the absolute value of any feature isgreater than unity, then the ratio isinverted, i.e., its reciprocal is takeninstead. The reason for taking ratiosfor all features is thus clear, that is, toreduce them all to the same scale of —1to +1.

TrainingThe feature extraction procedure just

described is performed on each model usingsimulated data. For each model and for a simu-lation run, two random points A and B are takenand the features fu f2 and f3 or f3 computedaccording to the procedure outlined. This givesa point in the feature space belonging to thegiven model. The procedure is repeated for asmany points as desired (the more points thebetter), after which more simulations are per-formed and feature points computed likewise.In this study five simulation runs per model arecarried out. The parameter values used for thedifferent models are shown in Table 2. Tomake the training more realistic a 2% of fullscale random error (hence 0.01 cm3/cm3 watercontent) is introduced to the simulated data.

For convenience and ease of visual com-parison the features are now plotted two by two.Figures 2(a) and (b) show feature points forModel 1 and Model 2 in the ft — f2 featurespace, and for Models 1 and 3 in the fj — f3feature space, respectively. Some overlappingof feature points is observed especially inFig. 2(b). This is due to the contamination ofthe 2% error. Had no error been introducedone would expect to obtain two narrow bandsof clusters, one along the line fj = 1, consistingof points from Model 1 and the other along theline f2 = 1 or f3 = 1 for feature points fromModel 2 or Model 3, respectively. Nevertheless,

TABLE 2

Parameter values used to generate data needed in the classification of the feature spacefor the different models of soil water characteristic

Simulationrun

1

2

3

4

5

Model 1

a

50

100

250

500

1,000

2.4

2.6

3.2

3.8

3.6

Model 2

5

l x l O 2

5XI01

l x l O 3

l x l O 4

l x l O 6

y

2,2

3.0

3.4

2.8

2.5

Model 3

ha

- 5

- 1 0

- 2 5

- 5 0

-100

0.21

0.23

0.30

0.28

0.32

116

STRUCTURE IDENTIFICATION OF SOIL HYDRAULIC PROPERTIES

classification is still possible. One may choose alinear or a nonlinear splitting of the feature space.In these instances linear splitting is performedwhereby the feature space is divided into two insuch a way that there is a minimum number ofnon-conforming feature points for each model.An approximately 45° line, extending from theorigin, as shown, is found satisfactory in bothcases. Thus, in Fig. 2(a) the subspace belowthe dashed line is assigned to Model 1 and thatabove the line to Model 2, Likewise, in Fig. 2(b),the lower subspace belongs to Model 1 whilethe upper to Model 3.

1.0--

(SJ

-1 -0

i . o - -

•1.0

—I 1 h I , I 1 T

• MODEL 1

» MODEL 2

j:

V* ..

a MODEL 1

• MODEL 3

/

Identification of the structure for the soil watercharacteristic

The test cases consist of the water contentversus pressure head data for disturbed samplesfrom each of four horizons from two profiles ofthe Bungor series (Typic Paleudult) (WanSulaiman, 1979). These are shown in Fig. 3where the asymptotic model (Model 1) has beenfitted to data from each horizon.

p i H i

-: o 0-0

Fig, 3, Water content versus suction head fordifferent horizons of (a) Profile 1 and(b) Profile 2 of Bungor Series. P withsubscript refers to profile number while Hwith subscript refers to the horizon number.Solid lines are asymptotic functions (Model1) fitted through experimental points.

2. Feature space of the various candidatemodels for the soil water characteristic:(a) Model 1 and Model 2, (b) Model 1and Model 3.

Data for each soil horizon is now subjectedto the feature extraction procedure and plots off2 against ix and f3 against fj are prepared. Inall cases the choice between Model 1 and Model 2can be readily made since most of the feature

117

W. H. W. SULAIMAN AND G. C. VANSTEENKISTE

points fall in the space belonging to Model 1.However, discrimination between Model 1 andModel 3 is rather difficult in most of the testcases. Figures 4 (a) through (d) illustrate someof the features of the analysis. Figures 4 (a) &(b) clearly indicate the superiority of Model 1to the other two. Figure 4 (c), on the other hand,appears slightly to favour Model 3 to Model 1,while Fig. 4 (d) indicates the reverse; in bothcases, the distinction is not very convincing.Overall, the asymptotic model (Model 1) fitsbest data from three soil horizons, namely, PiH t,P2H! and P2H4 while the Brooks and Coreymodel (Model 3) is best for two horizons, PiH2and P2H2. For horizons P]H3, P]H4 and P2H3,both Models 1 and 3 are equally adapted. How-ever, in view of the fact that the soils are textural-ly similar, one should expect a predominance ofone particular model. Further attempts atclassification using features 4 and 6 fail to resolvethe ambiguity.

The lack of success can be attributed inpart, to the inadequacy of the chosen featuresin discriminating the suitable model; the impor-tance of finding suitable features has alreadybeen emphasized earlier. A certain degree ofexperience and insight is necessary in derivinggood features for classification. Another reasonis that the present identification has been per-formed in only two dimensions using one featureper model. By increasing the number of featuresit should be possible to obtain a more convincingresult.

Parameter IdentificationLeast squares parameter identification is

performed with all the three models so as toprovide a comparison between the least squarestechnique and the pattern recognition approach.Results of the least squares fit are presented inTable 3. The sum of squares of deviations (SSD)for Model 2 are markedly higher than those ofeither Model 1 or Model 3, thus in accord withthe pattern recognition results. As with thepattern recognition method, there is no completedominance of a particular model. One signi-ficant feature, however, emerges and that is, theleast squares method and the pattern recognitiontechnique can give conflicting results as observedin the case of horizon PiHj. The most obviousreason for this is that a few bad observations canincrease the SSD significantly, yet these sameobservations will have little impact on the overallpattern from the whole data set in the patternrecognition approach. This augurs well for thelatter technique.

STRUCTURE CHARACTERIZATION OFTHE SOIL WATER DIFFUSIVITY

Candidate ModelsEarly work with diffusivity (Childs and

Collis-George, 1950; Klute, 1952) left the func-tional form of D(0) completely general. Anexponential function of the form D(0) =a.exp(flQ) was suggested by Gardner andMayhugh (1958), which since then has beenwidely used. This can also be written as

D(9) = (13)

Ahuja and Swartzendruber (1972) suggested apower function of the form

D(9) =a9n

(14)

(es-e)n/swhich yields infinite diffusivity at saturation.

A third model is proposed herein. Fromthe commonly used relationship K = KsS

n

where K is conductivity and subscript s refers tosaturation, and the Brooks and Corey model(Eq. 3) for h < ha, from which the above con-ductivity relationship is derived, we obtain viaD(9) - K/(d0/dh), a diffusivity structure of theform

r e - Q r ) m

D(9) = b (15)

U 0jU s - 0r

K shawhere b = and m = n - 1

K (0s-O r) A

Equation (15) is also a power function but unlike(14), the diffusivity at saturation is finite andequal to b.

The three models represented by (13), (14)and (15), hereby called Model 1, Model 2 andModel 3 respectively, are now considered as thecandidate models for the diffusivity functionsof the soils under investigation.

Feature ExtractionIn contrast to the method employed for the

soil water characteristic, here both parametersof each model are used to derive a single featureper model. The feature is defined as the sumof the ratios of each parameter evaluated for arandom pair of points from the data set. Thus,

DFeature 1, f 1 —

minA

(16)

118

STRUCTURE IDENTIFICATION OF SOIL HYDRAULIC PROPERTIES

f2

11

-1-fc

1 i 1

H o r i z o n P H x

X

MODEL 2

' 'A .

MODEL 1

(a)

Horizon P H

MODEL 3

V

-1! 1

MODEL 1

(b)

0

J

1-

0 .

I-

j 1 j

Horizon P

- MODEL 3

•

/ /

1 1 !

1H2

A

V

/

j

A

K

/

/

A

(c)

A

MODEL

AA *

/

1

Horizon P^H

• • " MODEL 3

H h

MODEL 1

(d)

H 1-

Fig. 4. Feature space containing feature points for various test cases in the structure identification of thesoil water characteristic.

119

W. H. W. SULAIMAN AND G. C. VANSTEENKISTE

Feature 2, f3 = +

m A bAand Feature 3, f3 — -f-

mB bB

The various parameters for points A and Bare computed in the same manner as describedearlier. These are:

fi =1 dD

D(9-er) d9

es-e

D m i n =

n = 56

m =

— , a =

b = DI)

0n

9

(17)

es-erWhenever the ratio between parameter valuesat the two points exceeds unity, its reciprocal istaken instead. In this way, all the featuresremain within the limits of —2 and +2 .

TrainingFive simulation runs are performed with

different combinations of parameter values.Since error in the D(0) determination is rather

large, a 10% relative error is added randomlyto the simulated data. The parameter valuesused for the different simulation runs are indi-cated in Table 4.

Figures 5 (a) & (b) show the plots of featurepoints for the classification between Model 1and Model 2 and between Model 1 and Model 3.Again some overlapping is observed, neverthe-less, the splitting of the feature space into twosubsets corresponding to the respective modelsis quite evident. (Note: none of the featurestook negative values in the training operation).

Identification of the Structure for Soil WaterDiffusivity

The soil water diffusivity-water contemrelationships of various horizons of the twoprofiles of the Bungor series, whose structureare to be identified are shown in Fig. 6. Usingdata from each horizon, features are calculatedaccording to (16) and (17) and plots of featurepoints, fi~f2 and i\—U are prepared. Figures7 (a) and (b) illustrate, respectively, the plotsobtained with PjHi and P2Hi. From thesefigures it is clear that the exponential functionfits the data best. A similar trend is also observedwith all other horizons.

Exponential functions are now fitted to thedata from the various horizons and the resultspresented in Table 5. Even though high corre-lation coefficients are obtained in all cases it is

TABLE 3Results of parameter identification of Model 1, 2 and 3 by the least squares method

Soilhorizon

PROFILE

PlHj

PlH2

P1H3

P1H4

PROFILE

P2H2

P2H2

P2H3

P2H4

Gs

1

0.54

0.55

0.57

0.575

2

0.59

0.57

0.57

0.565

Or

0.01

0.01

0.02

0.025

0.02

0.03

0.03

0.03

a

75.61

321.3

289.3

667.0

45.65

236.3

394.1

551.2

Model

fi

2.764

3.429

3.342

3.718

2.389

3.242

3.477

3.596

1

SSD x 102

1.796

2.327

1.314

1.313

1.923

2.817

1.488

1.474

S

85-5

176.2

135.3

154.4

32.4

111.6

105.6

98.6

Model

y

0.932

0.986

0.925

0.895

0.700

0.898

0.843

0.790

2

SSDxl02

9.169

8.701

5.978

5.220

9.484

9.056

5.633

5.745

Model 3

h a

- 6.03

-12.05

-15.05

-22.40

- 4.87

- 9.90

-17.70

-23.35

X

0.242

0.249

0.263

0.249

0.209

0.235

0.271

0.274

SSD x 102

1.123

1.252

2.843

2.144

2.225

2.617

2.580

2.078

SSD Sum of squares of deviation = £ \Si — Si ) where superscript m refers to measured value.

STRUCTURE IDENTIFICATION OF SOIL HYDRAULIC PROPERTIES

TABLE 4Parameter values used to generate data needed in the classification of the feature space

for the different diffusivity models

Simulationrun

1

2

3

4

5

Model

Dm in

1 xlO"4

1 x 10-5

ixIO-6

1 x 10-7

1 xlO-8

1

fi

60.0

50.0

30.0

20.0

40.0

Model 2 Model 3

n m

15.0

12.0

4.0

1.0

8.0

2.0

4.0

3.5

4.0

3.0

5.0

0.0001

0.01

0.1

1.0

- 6 . 0

4.0

10.0

7.0

- 2.0

I.0--

0.0

2 . 0 - -

uk

H 1 H 1

« MODEL 1

» MODEL 2 •

vV •

/ • • •

• MODEL 1

» MODEL 3

/• t

•b)

0.0 1. 0FEATURE I

Fig. 5. Feature space of the various candidatemodels for the soil water diffusivity.(a) Model 1 and Model 2. (b) Model 1and Model 3.

found that the derived functions underpredictD(0) values at and near saturation by as muchas 30%. This definitely will affect the accuracyof the solutions to the water flow equation. Forimproved accuracy it would be preferable to use

i

\ l0° •

io 1 •

i o - 3 -

*

i

•

• • * •

p i H i

p z H :

P 2 H 3

::;S

•* * *

•!4:::i* :^A

, , F

A

• * * * * . ; " ; / '

-

Fig. 6. Soil water diffusivity versus water contentfor different horizons of (a) Profile 1 and(b) Profile 2 of Bungor Series. P withsubscript refers to profile number while Hwith subscript refers to the horizon number.

121

W. H. W. SULAIMAN AND G- C. VANSTEENKISTE

t2

1-

•

-

o.

—1—I—f—

MODEL 2

//

, //

1 1 1—

H—

1

// •>

•

•

—^_

•

/

/

%•• •

•—1—

^

V

4

•

1 1

/

•

MODEL•

••

•

•—1 1

t~H—•

•

1

--

•

4 j

1 1 1 H

' /

MODEL 3 /

MODEL 1

..

H 1 1 1 1 1 1 h2 , 0

MODEL 2

I. .

H 1 h

fb.il P2H i

MODEL 3 /

MODEL 1

-i 1 ! 1 1 1 1 1 1 1-

H 1 1 1 1 1 1 1 1

/ *

MODEL 1

1 i2 , 0

f 1

Fig. 7. Feature space containing feature points for various test cases in the structure identification of soilwater diffusivity.

122

Soil water diffusivity D(9)f expressed as a unique andlinear regression

TABLE 5

piecewise exponential function. R is the correlation coefficient for thelnD - lnD min - J3(O - 0 min)

SoilUnique Function

horizon Diffusivity function (cm2/min)

PROFILE 1

R

P1H1 D - 2.928 x 10"J exp [11.473(0-0.01)] 0.975

PlH2 D = 1.157 x 10"** exp [15.311(0-0.01)] 0.960

P tH3 D = 4.389 x 1O~4 exp [16.08 (0-0.02)] 0.940

P1H4 D = 3.074 x 10~4 exp [16.592(0-0.02)] 0.940

PROFILE 2

P2H1 D = 9.467 x 10 4 exp [14.054(0-0.02)] 0.991

P2H2 D ^ 3.277 x 10"4 exp [16.780(0-0.02)] 0.936

P2H3 D - 1.765 x 10 exp [18.235(0-0.03)] 0.963

P2H4 D - 2.730 x 10"4 exp [16.692(0-0.03)] 0.942

Piecewise Function

Diffusivity function (cm2/min)

( 2.468 x 10 "3 eD = ]

t 7.186 X lO"' e:

exp[13.345(0-0.01)], 0.01 <0<0.356

:xp[23.553(0-0.01)],0.356<0<0.54

n-32.750 x 10" exp[ 4.761(0-0.01)], 0.01 <0<0.335

1.459 x exp[25.881(0-0.01)], 0.335<0<0.55

( 1.08D -

' 6.35

D =(4.1

0~3086 x 10~3 exp[11.288(0 -0.02)], 0.02 <0<0.352

"66.354 x 10"6exp[26.310(0-0.02)], 0.352<0<0.57

7.972 x 10~4 exp[ll.747(0-0.02)], 0.02 <9<0.352P.-6

•H

D-{' -

689 x 10 exp[26.779(0-0.02)], 0.352<0<0.56

6.096 x 10"4 exp[14.012(0-0.02)], 0.02 <0<0.357

4.718 X 10"5 exp[21.600(0-0.02)],0.357<0<0.57

5.198 x 10'4exp[10.869(0-0.03)],0.03 <0<0.32

4.321 x 10~6 exp[27.408(0-0.03)], 0.32 <0<0.57

802 X 10~4exp[10.111(0-0.03)],0.03 <0<0.356

9.024 x 10~7exp[30.744(0-0.03)],0.356<0<0.57

R

0.945

0.996

0.923

0.998

0.927

0.973

0.929

0.976

0.934

0.989

0.926

0.981

0.924

0.993

';:

•

h

rrlIDE

ZH

n>0z

xo70>Crn

o"0m70Hm

W. H. W. SULAIMAN AND G. C. VANSTEENKJSTE

a piecewise exponential function instead. Asshown in Table 5, in the region of higher watercontent the correlation coefficient for the latteris much higher than for the unique exponentialfunction; soil P2Hj is an exception in which aunique function is already satisfactory, hence apartition is unnecessary.

CONCLUSIONS

The pattern recognition approach has beenapplied to the structure characterization of twosoil properties, namely, the soil moisture charac-teristic and the soil water diffusivity. While themethod was able to delineate the most suitablemodel for the diffusivity (i.e. an exponentialfunction) from three possible candidate models,classification in a two-dimensional feature spaceachieved limited success in the structure identifi-cation of the soil moisture characteristic. Twoof the candidates, an asymptotic model (Model 1)and the Brooks and Corey model (Model 3)were equally adapted to the data. Better discri-mination could probably be obtained by usingbetter and/or more features, the latter entailingclassification in multi-dimensional space.

The training stage of the procedureadmittedly entails heavy computations; however,once this stage is complete, processing of eachdata set requires minimal effort. In fact, it tookless than 0.5 hour of computation time on theCDC 1700 (32K) to process the diffusivity dataand to plot the graphs of fj — f2 and ix —f3 for allthe 8 test cases. Together with the fact that thetechnique can give different results from theleast squares method leads to the conclusionthat the pattern recognition approach providesus with a useful technique in the identificationof model structures of poorly defined systems.

ACKNOWLEDGEMENTS

The granting of study leave by the UniversitiPertanian Malaysia and the financial support ofthe Algemeen Bestuur van de Outwikkelings-samenwerking (A.B.O.S.) for the study aregratefully acknowledged.

REFERENCES

AHUJA, L.R. and SWARTZENDRUBER, D. (1972): Animproved form of soil-water diffusivity function.Soil Sci. Soc. Amer. Proc. 36, 9-14.

BROOKS, R.H. and COREY, AT- (1964): HydraulicProperties of porous media. Hydrol. Pap. 3.Colo. State Univ. Fort Collins.

CHILDS, E.C and COLLIS-GEORGE, N. (1950): Thepermeability of porous materials. Proc. Roy. Soc.201,4, 392-105.

GARDNER, W.R. and MAYHUGH, M.S- (1958): Solutionand tests of the diffusion equation for the move-ment of water in soil. Soil Set. Soc Amer. Proc.22, 197-201.

HAVERKAMP, R., VAUCLIN, M., TOUMA, J., WIERENGA,PJ . and VACHAUD, G. (1977): A comparison ofnumerical simulation models for one-dimensionalinfiltration. Soil Set. Soc. Amer. Proc. 41, 285-294.

KANAL, L. (1974): Patterns in pattern recognition1968-1974. IEEE Trans, on information theory.November, IT 20.

KARPLUS, WJ. (1972): System identification andsimulation, a pattern recognition approach. Proc.Fall Joint Compt. Conf. 385-392.

KLUTE, A. (1952): A numerical method for solvingthe flow equation for water in unsaturatedmaterials. Soil Sci. 73, 105-116.

NILSSON, N.J. (1965): Learning machines - Found-ations of trainable pattern-classifying systems.New York. McGraw Hill.

SARIDIS, G.N. and HOFSTADTER, E.F. (1974): Apattern recognition approach to the classificationof non-linear systems. IEEE Trans, on Systems,Man and Cybematics SMC-4, 362-370.

SIMUNDICH, T.M. (1975): System characterization: apattern recognition approach. Ph.D. Thesis.UCLA. School of Engineering and AppliedSciences. Los Angeles.

VANSTEENKISTE, G.C-, BENS, J. and SPRIET, J. (1978):Design of a linear classifier using simulationtechniques. Proc. United Kingdom SimulationConf., May (in press).

VANSTEENKISTE, G.C, BENS, J., and SPRIET, J. (1978a):Structure characterization for system modelingin uncertain environments. Proc. Symp. on SystemSimulation Methodology, Rehovot, Israel. Aug.1978.

WAN SULAIMAN WAN HARUN (1979): Simulation ofwater movement in the Bungor Series (TypicPaleudult) in relation to soil erosion. DoctorateThesis (unpubl.). State Univ. Gent.

YOUNG, T.Y. and CALVERT, T.W. (1974): Classifi-cation, estimation and pattern recognition. NewYork. American Elsevier.

(Received 1 August 1980)

124