Proceedmgs of the 7th Internaiumal Workvng Conference on Stored-product Protection - Volume 2

Computer-Assisted Learning (CAL) to improve the quality ofpest management in grain storage systems

Barry Longstaff

Abstract

Improving the understanding of underlying issues by alllevels within a management hierarchy is a key to enhancingthe effectiveness of pest management. Improvements to thedecision-making procedures available to managers and pestcontrol staff will lead to greater application of effectivestorage procedures and so to reductions m overall storagelosses, and improved efficiency of pesticide usage. Acollaborative project between Indonesia and Australia hasdeveloped a Computer-Assisted Learnmg (CAL) system toaugment conventional training methods The packagemcludes a tutonal system, a game to tram users m thelogical processes used by experts in diagnosing problems, aSimulated gram storage complex that allows the user toexplore various pest management strategies, and anmteractrve pest tdenttfication component. Extension of thisproject, to mclude development and conduct of internationaltrammg courses and the identification of mechanisms bywhich CAL can be mtegrated into the trammgmfrastructures of the participatmg organisations, isdiscussed Benefits arising from this project can spillover toother countnes in the region and adoption and mtegration ofCALby organisations should lead to substantial reductions mcosts of training pest management staff, particularly wherethey have geographically-distributed workforces.

Introduction

Techniques for managing pests in gram stores are not alwaysused in the most effective manner. Evans and van S. Graver(1989) suggest that this IS often due to a lack ofunderstandmg of the technical and economic consequences ofavailable pest management options Expert systems havebeen developed as decision-makerste g. Jones et al , 1993;Wilkin et a1., 1990; Flmn and Hagstrum, 1990; Longstaffand Cornish, 1994). However, their use sometimesrequired an act of faith rather than understanding which hasprobably contributed to their relatively low adoption rate

1CSIRO Entomology, GPO Box 1700, ACT 2601, Australra

Other types of decision tools can help by improving problemspecification to Identify the key Issues, allowmg explicit andrigorous analysis of the problems and facihtating interactionsbetween vanous management levels through improvedtraining (Norton and Mumford, 1993; Longstaff, 1997).Computer-Assisted Learnmg (CAL) has the potential to

make a significant contribution to trammg pest managementstaff. CALsystems, whichcan utihse the latest multimediatechnologies, are inherently interactive and allow self-pacedlearning. They can be structured to allow students toexplore alternative pest-management options and learn theconsequences of their actions, either on their own or insmall groupsAn ACIAR-funded project, mvolvmg Austraha (CSIRO

Entomology, Queensland Department of Pnmary Industries,and the Co-operative Research Centre for Tropical PestManagement in Bnsbane ) and Indonesia (BULOG) , hasdeveloped a SUiteof CALtools to facilitate training of staff inrational and sustainable use of grain pest managementprocedures. The focus of the project has been primanly onthe Indonesian central gram storage system, which storespnmanly milled rice and which IS heavily reliant oncombined use of fumigants and stack surface and structuralspraying With contact insecticides. However, this genencapproach is more Widely apphcable in Australia andelsewhere.

'Pest ManagementWorkbench'

The project has focussed on the areas of pest identification,mformation access, developing expertise, and pestmanagement strategies A number of discrete modules havebeen developed and mtegrated Within a framework called'Pest Management Workbench' (Hahd et a1., 1997;Sardjono et a1., in press). The system has a bilmgualgraphical user-mterface (GUI) and currently includes fourmodules, which use many common resources such asimages, biological and chemical parameters, and languagefeatures. They are:

Teach

'Teach' is an interactive multimedia tutorial systemcovenng a range of essential pest management concepts and

1862

Proceedings of the 7 th International Working Conference an Siorei-product Protection - Volume 2

techniques (Figure 1). The module includes a player and abuilder which enables trainers to develop their own tutorialsby importing text and making hyperlinks to other resources,including pictures, sound and video files, help files, andother tutorial sections (Figure 2). No particular computing

expertise is required to create a tutorial with this product.The initial tutorial deals with the carbon dioxide fumigationof bag-stacks, based on the ASEAN manual (Annis and vanS. Graver, 1990) and includes extensive video material onthe procedures involved.

Fig. 1. An example from a Teach tutorial. Words or phrases in bold or italics represent links to other resources such as a pictureor a model.

in ,las_ enc"'"ures

Th. ~bJectiv. of e.-bon d!.al1de fturusauon 1/; to nturrtlW'l toXICconcentrations ofth. gUWlW@"I\outtheenclo:wreforloog enough to kill the most tolerant species ofill:;(;!:t-; thalmay be found. This meansIti.Unaining 'l. conc-enln.konof cu!:lon dioxide of33% or ~6ht, U'l."u pasti ofth. enelo"Ul~ fer 15 day'or mere, alembient temperatures et, or above. 2SC. Carbon ltioude cannot be used ifmore rapidfuau.eltt.lO:t:lis nquiled. In 't.Wl'I cas e.otner fum.!gants wiI1 have to be cor.udued. The uu of cu'bon .:::.:-dioxide inso:ak'lba{-:.Itllo:k:o is plUticwiUly t;ui.h.ble for, :[:~~

dJy commcditie s destined for Icog-tarm :stOl~e «0 ruonths)

for.ime~t conb"ol-.n, corn.tt,,~dilie, req'-UN'lg total freedom from th el'll.ku l~~W~"'SIII Situations whoe", the", ere pW~vly cheap and ccnvement.sources OfCM'oon.dionde

CarN)ojdio;'ld~ shruIdnot.t1/ 1.I~edwtwl'''';

- tb~ ~OlC;R~l/X~faib to meet the r~qWrecipfesSUfll test standard fOI gas-tightness . where Ulerei' 6 requirement fOEa f'umigld.lOn to b 6 completed in leH than 15 days.

when lrogtHJ.lnfffl ~Fecje, are known to be Jln~mtm Ihe commodity

1.~~JLt~~d.3. F"'l'aJ'iltion'ilr~;Won4.·Pn".~-..z~m~iliI;.5, TJuduq-W.o)x15.J,bJCIg-_t'Jll\~t~f'7, ~~lUlI.g till'l ~:j-io.f:k.. . .

Fig. 2. The Teach Builder screen. The buttons along the top of the screen allow the user to make links between words or phrasesand other resources.

1863

Proceedings of the 7 th International Working Conference on Stored-prod:uct Protection - Volume 2

Simulator

'Simulator' allows the user to explore various pestmanagement strategies via a simulated grain storagecomplex, based upon real sites and grain movementstatistics data. The system includes a number of simulationmodels, covering insect population growth (Longstaff,1988), breakdown of residual chemicals (Bengston, 1986),the Fumigation Decision-Support Sytem developed by NRI(Hodges et aI. , 1997) and phosphine fumigation (Annis andBanks, 1993). The latter predicts the likely phosphine gaslevels under defined storage conditions. It allows the user toexplore the effects of changing particular variables such asfumigation time, dose rate or sealing level upon gas levels

and also predicts the level of mortality achievable underthese conditions.The site layout comprises a suite of standard bag stores,

each containing up to 8 small and 4 large stacks. The systemcreates a database to monitor the changing status of eachstack, insect population growth and pest managementactions. The amount of grain lost and the cost ofmanagement are indicated to the user (Figure 3). The useris able to manipulate a large number of variables within thesystem, including monthly grain arrivals and level ofinfestation at intake, fumigation strategy and success, andvarious costs. The user is able to see the consequences ofvarying one or more of these variables.

Fig. 3. A set of windows from Simulator, showing the layout of a warehouse complex in Indonesiaon the lower right and 3 ofthe other windows that may be calledup allowing the user to change variousaspectsof the simulation. Each stack withineach warehouse is monitored and management processes, such as inloading and outloading of grain l sampling, chemicaltreatment and fumigation are recorded. A population growth model runs within each stack and insect densities areindicated by different colours. Grain lossesand treatment costs are shown in the lower right of the screen.

Diagnosis

,Diagnosis' is a game to teach users the logical processesused by grain storage experts in problem diagnosis. Usersare presented with pest management scenarios andresources, including comments from pest control staff,access to a laboratory and to field sites. They must then use

the resources and apply reasoning to diagnose the problem.Diagnosis then offers the correct solution and describes themost efficient way of reaching this conclusion. At any pointin the process the user is able to seek clarification of mostissues through the extensive commou Help system.

1864

Proceedings of the 7 th Inierruiiumal. Working Conference on Stored-product Protection _ Volume 2

LucID

'LucID' is an interactive key for identifying insect pestspecies. At the highest level, users are presented with a setof pest taxa and a set of characters of all of the major pestspecies found in storages (Figure 4). These are used toidentify the order, family and then species in question byviewing tbe various character states and selectingappropria tel y.

The seqnential selection of character states reduces thenumber of taxa remaining until a point is reached where alower level key must be accessed and the process continued.Once the species has been identified, the user is able toaccess the pests help file, which provides detailedbackground information about the species (Figure 5).

System requirements

Incorpora tion of videos into the package is demanding III

terms of storage space but delivery in a CD-ROM formatmeans that the system requirements are relatively modest,in terms of memory and hard disk storage. Pest

Management Workbench will operate on a Windows® -based386 computer with 8 Mb of RAM but, for best results, a

Pcntiume' with 32 Mb is probably the minimum necessary toachieve adequate video playback. This is still quite modest interms of current industry standards. It is important toremember when designing such a system that not allpotential users have access to the very latest equipment andthat design compromises may have to be made to ensure thatsystem requirements are not unrealistically high.

Adoption

Pest Management Workbench is an example of a powerfulgeneric technology that has widespread relevance in theagricultural and horticultural sectors, both as a decision-makingsupport tool, and as an educational tool to illustrate integratedpest- management principles. By improving the ability tocorrectly identify insect pests, LucID will facilitate use of themost appropriate control measure whilst Teach will providemanagers and pest control staff with a much greater level ofunderstanding of both the problems and their solution. TheSimulator will allow the user to explore, in a non- threateningmanner, a range of different management strategies, theireffects upon the quality of grain in store and their cost-effectiveness. Diagnosis will give the user some insight into theways that experts interpret problem symptoms.

Fig. 4. An example of a screen from LucID, showing character states and taxa.

1865

Proceedings of the 7 th International Working Conference em Stored-product Protection - Volunw 2

w_ Jntef l)~-;CTmrl(J]i

&ww.;L"@ora;."l.C~ Simil':!.J:iW:;~15., t)C{~rW(·,;~",~"1{g~z1 Dis:te:iFI1Hlotm.Limii;;; ad nil: 4mulililli'-atlll'll.f~M. M~f~'.d~.·Jj'».~_>}Jj)~~~._~_;:;}~.I.'l.!!1m.5 )k~.'-~JN"~_l.1J.t~J.f_:M~.~.~

Cbn< .. rislic ,N ....... ,rojlln,loJrthain,eases surt aJlll patl.t::med, utelUlU5 Sec;ltlellts and the final OM whiCk anFemuroflrind It.g Just]uu tootklilu: spms in trow(ane ,it;: 'aJId 'iWClSJllall).

Pest;;; .rdriedkiheyuul JiaricOt1)eallS. Not ofltnfinool OJi otherlegwnes. Mare-cold hardy,tlwtCallosobnulms:U1ll. trequn.t m:teinperate regions.

Optimum: 27 d.1.ysa130°C. 80910 r Jt.RaNg.:, U·)J°eirldx:iJrnuft popll1dtioM gro.l11 rat" par lJIOlIt~:25tilfClS

Fig. 5. An exampleof a page from the Pests help file. providing the user with a picture of the pest. a synopsisof the biology,and links to more detailed information.

To illustrate the inherent flexibility of this approach, anAustralian version of Workbench incorporates two additionalfeatures not found in the Indonesian version. These are thePestMan expert system and GrainMan, a graphical tool thatcalculates dose parameters for fumigation of multi-celledstorage complexes with phosphine using the SIROFL()F)technique. The former provides advice on appropriatetreatments for particular infestation problems, the latter is avery user-friendly graphical tool that significantly simplifiesa fairly complex set of calculations, and provides a facility tocreate an organisation-wide database to collate pestmanagement actions.A number of countries are placing increased reliance on

fumigation to control insect pests and the potential for thedevelopment of serious levels of phosphine resistance,resulting from poor practices, has been clearly identified(Bengston, 1986). CAL systems could play an importantrole in enhancing the effectiveness of pest management byimproving understanding of underlying issues at all levelswithin a management hierarchy. Widespread adoption ofCAL systems will lead to greater application of effectivestorage procedures and thus to reductions in the cost ofstorage and in overall storage losses. It will also contributeto the sustainability of these practices and facilitate theintroduction of new ones. For example, by improving the

capacity of organisations to carry out effective fumigationsthe development of phosphine resistance could besignificantly delayed. The introduction and integration ofalternatives to methyl bromide into existing post-harvestsystems could be significantly enhanced througb theappropriate use of CAL.The next phase of this work is to validate the approach

and deploy the technology in the field. To achieve this,courses will be run in Indonesia, the Philippines, Thailandand Vietnam. CALwill be essential parts of the course butwill be integrated with more traditional methods. Thecourse would be conducted in English, but Workbench willallow self-paced instruction in the students' own language.In addition to this, new tutorials will be incorporated intothe system, encompassing grain drying and mycotoxins.An important secondary objective of this new project will

be identification of mechanisms by which tbis novel approachto training in the region can be integrated into the traininginfrastructures of the participating organisations. In the caseof Pest Management Workbench, the collaboratingorganisations will be able to refine and extend the coursecontent as they wish, and to develop new courses in theirown languages. This will contribute to improvements intechnology transfer and training in the agricultural sectorand, ultimately, to improved overall food security, through

1866

Proceedings of the 7th Internatwnal Working Conference on Stored-product Protection. - Volume 2

reduced wastage. Adoption and mtegration of CAL intocorporate training systems will lead to more effectivetrammg programs, particularly where organisations havegeographically-distributed workforces, as is the case withour collaborators m Indonesia. There Will also be substantialcost savmgs: For example, m Austraha, the Department ofEducation, Employment, Training and Youth Affairs(DEETYA) claim that the recent adoption of a CAL systemto tram staff has produced savings of greater than 50%(pers. comm. MIchael Grosser, Catalyst Interactive Pty.Ltd. ).

Acknowledgements

Financial assistance was provided by the Austrahan Centrefor International Agncultural Research (ACIAR).

References

Anrus P. C. and Banks H J 1993 A predictive model forphosphme concentration m grain storage structures. In:Proceedings International Conference on ControlledAtmosphere and fumigation, Winmpeg, Canada, (NavarroS. and Donahaye E. , Eds) pp.299. - 312.Annis P. C and van Someren Graver J. 1990 Suggestedrecommendations for the fumigation of grain in the ASEANRegion. Part 2. Carbon dioxide furrugation of bag-stackssealed in plastic enclosures: an operations manual, KualaLumpur, ASEAN Food Handhng Bureau/Canberra,ACIAR, 58pp.Bengston M. 1986 Background studies on fumigants:SeSSIon Chairman's summary In: Pesticides and humidtropical grain storage systems: Proceedmgs of aninternational semmar , Manila, Philippines, (Champ, B.R. and Highley, E. , Eds). p 222.Evans D. E and van Someren Graver J. 1989 Integratedpest and commodity management: puttmg first thingsfirst In: Grain Postharvest Systems: Proceedings of the10th ASEAN Technical Semmar on Gram PostharvestTechnology, Bangkok, Thailand, 1987. (de Mesa, ed.)pp. 206 - 217.

Fhnn P. W. and Hagstrum D. W. 1990. Stored GramAdvisor: a knowledge-based system for management of

msect pests of stored gram. AI Applications, 4:44-52.Halid H., Longstaff B. C., Annis P. C. Sidik M. andTurner J. 1997 Grain quality maintenance in the ~ASEANregion: a role for decision-support tools., J3\9troP SpecialPubhcation No. 59 Pest ~anagement for Stored Food andFeed, ISSN 0125 - 975X, pp. 83 - 93.Hodges R. J. , Smith M , Madden A. , Russell D. , GudrupsI. and Halid H 1997. Development of a decision supportsystem for the fumigation of milled-rice bag-stacks in thetropics. In: Donahaye E. J. , Navarro S. and Varnava A.(Eds ) Proceedings of the International Conference onControlled Atmosphere and Fumigation m Stored Products.(NIcosia, Cyprus, 21- 26 April, 1996) pp. 425 - 434.

Jones T. H. , Mumford J. D , Compton J. A. F., NortonG. A. , and Tyler P. S. 1993. Development of an expertsystem for pest control m tropical grain stores.Postharvest Biology and Technology, 3: 335 - 347.Longstaff B C. (1988) A modelhng study of the effects oftemperature marupulation upon the control of Sitophilusoryzae by insecticide. Journal of Applied Ecology, 25: 163-175.

Longstaff B. C. 1997 Decision tools for grain storage pestmanagement. J. stored Prod. Res., 33, 99 - 114Longstaff B. C. and Cornish P. 1994 'PestMan' - Adecision-support system for pest management in theAustrahan central grain-handling mdustry. AIApplications, 8: 13 - 23.Norton G. A. and Mumford J. D. (Eds.) 1993 DeCISIontools for pest management. CAB International,Wallingford, UK. ISBN 0851987834.Sardjono, T, Hahd, H. , Longstaff, B. C., Sidik, M. andAnnis, P. C. (in press) Improving the quahty of pestmanagement in ASEAN gram storage systems through theuse of computer-aided learning (CAL). Proceedings of the18th ASEAN Seminar on Grams Postharvest Technology,(Manila, The Philippines, March 11-13, 1997).WIlkm D. R , Mumford J. D. and Norton G. A. 1990. Therole of expert systems in current and future grainprotection. In: Fleurat-Lessard F. and Ducom P. (Eds).Proceedings of the Fifth International Working Conferenceon Stored-Product Protection, Bordeaux, France,September 9 -14, 1990 III. pp. 2039 - 2046.

1867