.I. stored Prod. Res. Vol. 33. No. 2, pp.

99-114, 1997
CC 1997 Ekvier Science Ltd
Pergamon All rights we&i. Printed in Great Britain
PII: SOO22_474X(%)OOO34-3 0022-474X/97 $17.00 + 0.00

Review
Decision Tools for Grain Storage Pest
Management
BARRY C. LONGSTAFF

CSIRO Division of Entomology, GPO Box 1700, Canberra, ACT 2601, Australia

(Accepted 22 July 1996)

Abstract-The application of a range of decision tools to the management of pests in stored grain
is discussed. Included in this review are: Knowledge Acquisition and Surveys, Decision Analysis,
Modelling and Expert Systems. A project integrating a number of these techniques, to provide
a coherent training system, is also described. Improving understanding of underlying issues by
all levels within a management hierarchy is seen as a key to enhancing the effectiveness of pest
management. Improvements in decision-making procedures available to managers and pest
control staff should lead to greater application of effective storage procedures and thus to
reductions in overall storage losses and improved efficiency of pesticide usage. 0 1997 Elsevier
Science Ltd

Key words---decision analysis, modelling, expert systems, decision-support

INTRODUCTION

The solution of pest management problems is frequently complex and usually involves agronomic,
pest control, logistic, economic and sometimes even political factors. To fully appreciate and
accurately diagnose pest problems and make appropriate recommendations for management, all
of these factors need to be taken into account (Wilkin and Mumford, 1994). Another impediment
to effective pest management is the so-called ‘institutional gap’, where research scientists, exension
workers, and managers often operate independently, because of institutional barriers, despite the
fact that each has an important role to play in the process. Decision tools of various sorts can help
to resolve these difficulties by providing improved problem specification to identify the key issues,
explicit and rigorous analyses of the problems at hand and facilitating interactions between the
various management levels through improved training and dissemination of appropriate
information (Fig. 1).
In this review, I will consider a range of decision tools and discuss their application to the
management of pests in grain storage systems. These will include: Knowledge Acquisition and
Surveys, Decision Analysis, Modelling and Expert Systems. The paper will conclude with an
example of a project which is trying to integrate a number of these techniques into a single package
of broad scope. This is a personal view of the subject and also necessarily brief. There are a number
of textbooks dealing with this subject in much more detail and Norton and Mumford (1993) is
strongly recommended for further reading.

KNOWLEDGE ACQUISITION AND SURVEYS

Before it is possible to improve the decision-making process in pest management in grain storage
systems, it is essential to understand current practice and how decisions are made. This means
talking to all appropriate ‘players’, i.e. head office managers, regional managers, store managers,
pest control staff and farmers. This knowledge acquisition process is commonly achieved using
99
 B. C. Longstaff

Central policymaker ti + Scientist

\
A A

I-‘-
DECISION TOOLS

t
+ +

Regional Manager Pestcontrol operator

Fig. 1. Decision tools, and the process of constructing them, can overcome the ‘institutional gap’ and
facilitate interaction between decision-makers within a grain-handling system (Adapted from Norton
and Mumford, 1993).

some sort of survey technique, which should be designed to document current practice and identify
areas for improvement of management practice. Appropriately designed and implemented, surveys
have the potential to get important information on behaviour, attitudes and unpublished facts that
could have a major influence on the effectiveness of any management initiative. It is important to
tailor the survey document to target particular groups of workers so that the information derived
has the greatest value.
The mechanism used to conduct the survey needs careful consideration. One can use a personal
approach, interviewing either face-to-face or over the telephone. These methods allow follow-up
questions of clarification to be asked and the information derived is likely to be verifiable.
Unfortunately, this method is labour-intensive and thus limiting in terms of the potential sample
size achievable. The alternative is to distribute questionnaires by hand or through the mail. This
method has the potential to achieve a much bigger sample size but this benefit is usually more than
offset by a low response rate, often below 30%. There is also no way of checking the validity of
the responses. In addition, this method has biases in favour of those with strong opinions and
against those having difficulties with reading and writing, sometimes an important factor in
developing countries.
The ideal survey protocol would:
l set objectives and sample size
l identify the population to be surveyed
l design questionnaire
l conduct pilot survey
0 revise questionnaire
l conduct main survey
0 analyse results.
The design of survey questionnaires and the analysis of survey results are complex issues and
will not be described here. They are discussed at length in Norton and Mumford (1993).
The Delphi technique is a variant of the traditional survey in which a number of experts are
asked to respond to the questionnaire, at first individually and then collectively, with feedback
between the phases. Such an approach, which was used by Jones et al. (1993) in the early stages
of the development of an expert system for pest management in tropical grain stores, allows a
consensus to develop through the feedback process and is thus likely to give a more reliable result.
It goes without saying that the success of such an approach is highly dependent upon the level of
expertise of the participants and their willingness to be objective.
 Decision-supporttools for grainpest management 101

DECISION ANALYSIS

Once knowledge has been acquired, an analysis of the decision-making processes involved in
current pest management practices is essential if optimum efficiency is to be achieved. This will
require a comparison of the objectives of pest management with the available options, identification
of the various levels of management involved and their interrelationships, and an assessment of
the major factors likely to influence the outcome (Table 1). An analysis of the way decisions are
made within an organisation can lead to changes that will improve efficiency. Thus in the
Indonesian centralised grain-handling system, BULOG, the major decision-making for pest
management occurs at the Head Office Stock Bureau and regional Dolog levels. There is an
argument to support the decentralisation of some of the decision-making to the district Sub-Dolog
level, where the important local factors are more fully understood. Circumstances may require a
revision of the overall management objectives (Fig. 2).
Taking the storage of barley as an example of decision analysis, the factors that need to be
considered are:

l end-use: e.g. malting, feed

l owner: domestic mill, brewer, export agency
l estimated storage time: weeks, months
l temperature: high, medium, low
l moisture content: high, medium, low
l weather damage: high, medium, low
l control options: fumigation, cooling, drying, protectants
0 grain variety: many options.

These may, in turn, be organised into a decision tree, such as that shown in Fig. 3. If the
commodity to be stored is malting barley, for example, the major management objective is the
maintenance of a germination level of 95% or greater and all pest management decisions must be
contingent on this. Failure to achieve this objective will result in significant commercial losses. As
germination is affected by temperature, moisture content and degree of weather damage of the
grain, it is essential for the manager to be aware of these critical parameters before a decision is
made on the choice of storage and quality maintenance procedures to be used. Other factors that
will have to be considered include technically feasible treatments (e.g. is the store sealed or sealable
so that fumigation with phosphine is possible), commodity-specific registration issues (methoprene
is not registered for use on malting barley although it would otherwise be very suitable as a
protectant on malting barley which is often stored for long periods), and acceptability of potential
treatments to markets (a number of importing countries will not accept grain treated with any
residual grain protectants).
The decision to follow a particular course of action to manage pests usually needs to include
the cost efficiency of the alternatives, in addition to the efficacy issues. The determination of
economic thresholds and of costs of alternative management strategies is vital to the process.
Hodges et al. (1992) investigated the cost efficiency of respraying store surfaces in Indonesia with
insecticide to protect milled rice from insect attack and prolong the period between fumigations.
Their null hypothesis was that respraying store surfaces would produce a delay between required
fumigations that would result in reduced overall costs. The natural refumigation period in the
BULOG system they were describing is 120 d and the routine insecticide spraying occurs every
28 d. If these spray treatments were sufficiently effective to slow down reinfestation for an entire
refumigation period, there would be one fumigation and eight spraying episodes in 240 d. Since
an insecticide spray treatment typically costs about 8% of a fumigation, the cost of the spraying
strategy would be 64% of the second fumigation. In other words, if the pest were controlled for
an additional 64% of 120 d (77 d) the strategy would be seen to be cost-efficient. The experimental
data collected by Hodges et al. (1992) showed not only that this was not the case, but that the
spraying of store surfaces produced no measureable increase in the period between refumigations.
These authors concluded that the practice of respraying between fumigations was not cost-efficient
and that it would be more prudent to invest in improving the effectiveness of fumigation and store
hygiene.
 Table 1. An example of the hierarchy of decision-making within a grain storage system
Organisation Person Current decision Future decision
Head office Stock bureau National movement
Regional movement
Stock/personnel/training Training
Stock bureau Select contractors
Decide on fumigants/insecticide
Financial decision on reprocessing
Regional office Regional manager Repairs that influence hygiene
Head of Pest Quality Control (PQC) Assessment of state of sheets and other equipment District level
Assessment of pesticide stocks District level
Supervisor of fumigation team Whether sheets are installed correctly District level
Head of distribution section Decision to distribute based on monitoring
Head of PQC section Decision on fumigation based on pest monitoring
Regional manager and district manager Which source should be long term (based on experience)
District office Head PQC subsection Surface spraying (calendar-based) Spraying based on population
Choice of fumigants
Frequency of fumigation
Distribution section Choose complex
Procurement section Contract with warehouse complex/private miller for stock
PQC subsection CO, pressure test, CO, level
Monitoring quality and pest population
Warehouse complex Warehouse manager Hygiene
Warehouse rotation not feasible
Choose unit and stack to distribute
Claims on failed nrivate fumiaations
CO2 contractor Approves the CO2 sheet installation
 Decision-support tools for grain pest management 103

Management
Management options:
objectives effectiveness and
costs

Assess extent
Knowledge acquisition

Assess potential outcomes from

management options in terms of Decision analysis
objectives
-I

Revise
management
objectives

Management decision Decision-making

-I
Fig. 2. The decision-making process in pest management

MODELS

Models are very important decision-support tools at both the strategic and tactical level,
providing an understanding of the way in which complex systems function and thus giving insights
into the consequences of many managerial actions.

Pest management models

Longstaff (1991) and Throne (1994) have provided overviews of the role of modelling in
evaluating alternative pest management strategies for stored products. The Sitophilus demographic
model (Longstaff, 1991) is a complex model combining all of the known features of the biology
of the species, including its response to grain temperature and moisture content. However, because
it is not yet possible to adequately describe the temperature changes in the grain brought about
by the metabolic activities of the insects, use of the model has been limited to situations where this
is relatively unimportant, i.e. where the control measure involves temperature manipulation. This
model has been used to show the beneficial effects of grain-cooling in suppressing insect population
growth (Longstaff and Cuff, 1984), reducing the rate of protectant breakdown or following
fumigation (Longstaff, 1988a) and particularly upon the rate of spread of a resistance gene in a
population (Longstaff, 1988b) through the prolongation of the developmental rate (Table 2).
‘Phosplan’ is a model developed to predict phosphine concentrations during fumigation (Annis and
Banks, 1993). Recently, it has been further developed to combine the physical and chemical models
of phosphine generation with a mathematical description of the large set of biological data on the
effect of phosphine dose and exposure period on the mortality of adult T. castaneum. Phosplan
collects information from the user and predicts the time course of phosphine concentration during
fumigation, as well as the stage-specific insect mortality levels. The model incorporates significantly
more realism than the fairly simplistic fumigation module used by Longstaff (1988a) and allows
104 B. C. Longstaff

Short High
storage temperature

Weather
Malting
damage
Barley
LI

Low
Long temperature
storage

No weather
damage
Feed
Barley

“, ’ =
Fig. 3. Part of a decision-tree depicting the hierarchy of factors to be considered in the storage of barley.

the user to explore the effects of changing particular variables, such as fumigation time, dose rate
or sealing level (Fig. 4).
A number of other population dynamics models have been used to consider management issues
in stored grain systems. Hagstrum and Heid (1988) developed a simple model to predict
Rhyzopertfza numbers in grain stores in the USA. Flinn and Hagstrum (1990a) built upon the basic
population dynamics model of Hagstrum and Throne (1989) to consider the effects of aeration,
fumigation and protectants upon Rhyzopertha and Hagstrum and Flinn (1990), in turn, expanded
this to consider another four species. Kawamoto et al. (1989, 1991a) developed population
dynamics models for Cryptolestes ferrugineus and Acarus siro and these are being used as the
biological basis for a composite granary-ecosystem model, into which is also built a physical model
(Metzger and Muir, 1983; Flinn et al., 1992) for changes in grain temperature and moisture content
with and without aeration. The model is able to simulate population dynamics under changing
physical conditions (Kawamoto et al., 1991b, 1992). Sinclair and Alder (1985) developed a model
to simulate the management of stored grain pests on farms and the implications of this for the
central grain-handling system. Factors considered included the type, timing and frequency of
hygiene practices and residual chemical treatments and insecticide resistance. The study highlighted
the importance of cleaning the harvesting machinery, particularly if residual chemicals were not
used.

Detection of pest threat

The ability to detect insect pests before economic damage occurs is fundamental to integrated
pest management strategies. Early warning of pest presence can lead to reductions in both grain

Table 2. The proportional increase in the population of Sitophilus oryzae after one year of storage where insecticide was
applied at a rate sufficient to achieve control in the absence of resistance. The initial frequency of the resistance gene
is 0.01 and the resistance factor is 64 x (from Longstaff, 1988b)
Temperature regime Population increase Frequency of resistance gene
No cooling 15,000 x 0.94
Slow cooling 200 x 0.73
Rapid cooling 2x 0.16
 Decision-support tools for grain pest management 105

I 0 2 I

Leak rate X/dcq

Nomfnal conceotretion -724 da3

Effectivs exposure time 121.00 hn

Cafculated moftafity of ?--cH&#auw

99.999999992 X mortalityatkdh
99.997942767 x martalitypupae
0 turvivo~rif 1 .[Link] beded
25 rurvivarill .[Link] pupae treated

1 .[Link]
individu& in this fmligation wdd ba 3.33 psr kg

Fig. 4. An output screen from the Phosplan model, which is designed to show users the interaction
between applied dose of phosphine, leakage, exposure time and insect mortality. The model comprises
two complex sub-models. The fumigation sub-model describes the effect of store type and leakiness,
grain type, temperature and moisture content upon phosphine gas concentrations over time. The
mortality sub-model describes the effect of gas concentration, period of exposure, insect life-stage,
temperature and moisture content upon insect mortality. The slide bars allow for rapid changing of
key variables.

losses and costs incurred for control. The traditional method of detecting infestations in both
bag-stacked storage and bulk-stored grain has involved taking grain samples, usually with a spatial
sampling program, at regular intervals of between 2 and 4 weeks. This approach has many
drawbacks but the costs of making the strategy more effective by more frequent or more intensive
sampling are prohibitive. This problem led to development of trapping procedures that take
advantage of insect behaviour to detect them in both bulk grain and bag stacks, and in empty
storage structures. Traps tend to capture more insects than probe sampling methods because they
are left in place for a period of time. Because trapping procedures depend upon insect movement,
factors such as temperature, trap placement, and grain type may influence numbers and species
of insects caught. The effects of such influences and interpretation of trap catches have been
discussed by Cuperus et al. (1990), Hagstrum et al. (1990) and Hagstrum (1994). There are also
problems in relating numbers of insects caught in traps to the actual size of the parent population.
Traps provide only relative numbers, rather than absolute estimates, and require interpretation
based on trap ‘efficiency’. This is generally very difficult to assess, as it is usually dependent on
environment, species and other variables. Recently, dynamic models of insect movement have been
used to generate probability models of trapping data, which are then used to provide absolute
estimates of population size (Wileyto, 1994; Wileyto et al., 1994).
Kenkel et al. (1991) considered the decision-making problems of a commercial grain store
manager receiving grain at harvest time which will be sold in the following spring. The major
sources of uncertainty for the manager were considered to be the probability of detection of insects
and collateral damage at the time of sale and the potential rate of increase of the insects present
in the grain during the storage period. They combined a model predicting the probability of
detecting insects through normal sampling procedures (Hagstrum et al., 1990) with an insect
population growth model (Flinn and Hagstrum, 1990a; Hagstrum and Flinn, 1990) and subjected
106 B. C. Longstaff

the predictions to a risk analysis. A risk-neutral strategy would seek to maximize the expected
returns on the grain in storage and a manager would be expected to employ strategies as long as
costs remained less than the expected benefits. By contrast, a risk-averse strategy would seek to
use pest control options so as to reduce the variability of returns, even if this resulted in a slightly
lower average return. The results suggested that some of the insect control strategies used in
commercial grain stores in the USA are not risk efficient, due in large part to the manager’s lack
of knowledge of insect population growth patterns in particular circumstances. The study illustrates
the potential of models in providing significant support to the pest management decision-making
process.

EXPERT SYSTEMS

Grain storage staff responsible for pest management face a number of problems:
0 scarcity of appropriate expertise;
l need for more rapid and consistent decision-making;
l need for reducing costs;
l need to actively conserve existing treatment measures;
l need to make use of the most appropriate scientific information, often in the form of
mathematical models;
l need to access corporate databases containing information about the store and its contents,
including potential market, treatment, and infestation history.
Such complexities make the decision-making process very difficult and the costs associated with
a less than perfect decision could be considerable. Decision-making depends, to varying degrees,
upon information derived from both experience and experiment, and expert systems seek to
combine such information to allow users to make optimal pest management decisions. Such systems
offer the potential to offset many of the problems identified above, although it must be realised
that they can only make recommendations, with users retaining the responsibility for making the
final decisions. Expert systems have been used for pest diagnosis, pest prediction, executive
decision-support and training (Heong, 1990; Flinn and Muir, 1994) but their application to
managing stored grain has been somewhat limited, with very few groups developing systems (Flinn
and Hagstrum, 1990b, c; Wilkin et al., 1990; Jones et al., 1993; Longstaff and Cornish, 1994).
The core of the expert system is the ‘knowledge base’ which is the explicit functional associations
between items of information and/or data in a particular application and is functionally equivalent
to the equations in a systems model. The knowledge is most frequently represented in the form
of simple or complex rules:
l IF facility is sealed THEN fumigate
l IF the commodity is a cereal AND there is an infestation, AND the infestation is Oryzuephilus
AND the storage period is greater than 3 weeks THEN propose Methoprene and Dichlorvos
or Methoprene and Reldan.
l IF a group 2 pest exists for a bin, AND the bin does not have Bruchus pisorum present AND
the bin does not have a group 1 pest present AND the bin’s grain temperature is greater than
20°C THEN propose Phosphine treatment for 5 d at 0.2 g/m’ or 7 d at 0.15 g/m3.
Knowledge acquisition is usually an iterative process (Fig. 5). A ‘knowledge engineer’ interviews
an expert, gains access to any appropriate documentation, and then produces a prototype
knowledge base. This is then reviewed with the expert and subsequently refined. The design of the
knowledge base is a crucial stage in the development process because decisions made at this stage
will have consequences for the subsequent maintainability of the system by third parties. This stage
is, or should be, independent of the mechanism of implementation, which may be a commercially
available ‘shell’, such as Level5 Object@ or ART-IM@, or a programming language, such as
PROLOG or C.
During the operation of the expert system, the rules are managed by a special program called
an inference engine, which accepts input from the user, fetches data from the database, performs
the logical reasoning required by the rules, and presents the system’s response to the user. In many
 Decision-support tools for grain pest management 107

Problem
spacficawrl

Knowledge
acquis%on

Revision

w
Knowledge
Validation engineering

Implementation

Fig. 5. The typical development cycle for expert systems

applications of expert systems the consequences of an incorrect or misunderstood recommendation

may be serious. It is, therefore, essential that such systems have an explanation facility so that the
user can question the system’s action. A good explanation facility will provide a controlled depth
of response, with the option to further explain elements of the earlier explanation (Fig. 6).
As a word of caution, it is worth pointing out that not all problems should be seen as candidates
for expert system development (Flinn and Muir, 1994). It is necessary that the problem be
reasonably well-defined and, because of the high cost of development, the perceived benefits should
be considerable.
The first expert system for stored grain pest management, Grain Pest Adviser, was developed
initially as a training device (Denne, 1988). In the first phase of its development,
commercially-available ‘shells’ were used for rapid prototyping but, as the project developed, the

Stocks and storage Knowledge base

database

INFERENCE ENGINE

Customer requirements

Regulatory raquiremants

Fig. 6. A conceptualized expert sytem model, illustrating the interaction between the components.
108 8. C. Longstaff

system was re-implemented in PROLOG. It was used to provide decision-support for grain store
managers in the UK (Wilkin et al., 1990). A second decision-support system, for the management
of bagged maize stored in the tropics (Compton et al., 1992; Jones et al., 1993), was implemented
in yet another package, ‘Knowledge Pro@‘, which facilitated the development of a hypertext-based
Help system. These changes in the implementation tool reflect the rapidly changing software
environment of knowledge-based programming and highlight the basic need to produce a
well-engineered knowledge base, independent of the implementation tool. The knowledge base of
the second system was created from decision trees derived from numerous interviews with domain
experts. In these discussions, the interviewer asked the domain expert(s) how they would respond
to particular pest management problems and why. By contrast, the primary methodology adopted
by Flinn and Hagstrum (1990a, 1990b) was to derive rules from the results of several hundred
simulations conducted with sophisticated computer models of insect population growth. The
system, ‘Stored Grain Advisor’, was designed to provide the user with a prediction of insect
population growth in commodities with particular properties, stored under different seasonal
conditions and the consequences for this of various management practices.

‘PestMan’
A project to develop an expert system for managing stored grain pests in the Australian central
grain-handling systems was initiated in 1991 by CSIRO. The objective was to combine principles
and methods derived from research with expertise from the grain industry. The system was seen
as an aid in the training of management and pest control staff. First, a brief study was carried out
to consider the suitability of applying knowledge-based technology to this problem. Grain storage
expertise was scarce and tended to be lost as experts were promoted or moved to other jobs. There
was also a large capability gap between expert and non-expert. There appeared to be tangible
economic benefits to be gained from reductions in operational costs and increased productivity that
could be introduced with such a system. The exercise was seen to be feasible - it was not too trivial
or too difficult - and knowledge was readily available, as the experts were both interested in, and
capable of, articulating their methods, seeing such a system as a productivity aid, rather than as
a threat.
It was decided that the first goal of the project would be to develop a simple prototype system
(PestMan), with a limited knowledge base, to demonstrate the potential of Decision-Support
Systems (DSS) to the Australian grain storage industry (Longstaff and Cornish, 1994). A prototype
was succesfully implemented in Level5 Object@ and, subsequently, an enhanced pilot version was
developed for trial by the Grain Elevators Board of Victoria (GEB) and then customised for other
state systems. This pilot system included a much enlarged knowledge base (reimplemented in
ART-IM@ and Visual Basic@) and an improved Graphical User Interface (GUI) and expanded
‘Help’ system (developed in Microsoft Multimedia Viewer@). It also provided the user with
transparent access to corporate databases and to a range of simulation models.

PestMan operation
A grain storage site/unit is chosen and data accessed from a local or remote database or provided
via the keyboard. The data presented are assessed and a set of recommendations generated, using
rules within the knowledge-base, that take into account corporate practices, storage facilities and
market requirements (Fig. 7). These are then costed and ranked and accompanied by an
explanation. A graphical option allows the user to quickly identify preferred treatment(s). The
system will allow users to create any pest management scenario they desire and get the results for
proposed treatments. Users can, having seen the list of proposed treatments, go back and alter any
of the details to see what effect that variation has on the treatments. This level of interaction makes
PestMan an ideal training tool for new pest and quality control managers.
A sophisticated on-line help system was included to provide instruction in using the package and
tutorial and reference material about insect biology (Fig. 8), fumigation practice, and pest and
quality control measures, The system also provided access to a range of simulation routines
including the temperature and moisture-content dependent loss of germination in malting barley
(Bason et al., 1993), which provides managers with a tool to enable them to assess storability of
this commodity. If the predicted maximum storage period is less than is desired, the model
calculates the temperature to which the grain must be cooled to achieve this goal.
 Decision-support tools for grain pest management 109

Agsnts: pt-mpmc Pest:

RecanmendationFunigate this bk~ with Phosphk. Funigate v&h SIROFLO. Eitha release the pak witk 22.05W&S fromti
d inbadii, 01 cml it to 24 ~CQWC~ C. GlllOrLO ~ettkgr: 2.0 +dar. Gap bdt+ 1435.440. n&d
concentrdix 43.958. Revised fur@tion days 18.32. Alow a peed of ventilkn appropriate for ti store.

Expmn: Siofb is flted to thii bii. Whorl W3 recirculation equipant a longer periodis necessary to ~RSU~thd an
adequate level of gas distribution is achieved.

FalR@eGroupl ~ds:14~ds~lratsofSgpprntogiveanri~of~.
Dosage.:
Constrdnts:

Mtirnum phas ptae: 0.01 mgkg

8
Residue
Levsl:
~ddlng:NowwIohalgperiodnecessary.
/
..-,- _II_~ --~__-__^__._~
_w...ae.-- ~--e=---~~~~~II.__L~___ ...,
L
Fig. 7. An example of the PestMan output screen, showing recommendations and associated explanatory
notes.

A Chinese PestMan
Overlapping with the development of PestMan, a collaborative project between CSIRO and
researchers from the Chinese Ministry of Commerce began the development of a DSS for use in
China. This system was developed using the prototype version of PestMan as a template. It evolved
in parallel to the Australian system and is currently undergoing trial at several grain depots in
Sichuan. This trial will enable the developers to modify those aspects of the system shown to be
at variance with the operational environment.
The Chinese researchers were faced with a problem of great magnitude-that of developing a
system that could be used in China using computing tools developed in the West. Their original
plan was to develop a character-based system using a Chinese version of the C language in DOS.
However, after viewing the CSIRO prototype, it was decided that a Windows-based system with
a graphical user-interface (GUI) would have much greater potential to display useful information
in Chinese characters than a DOS-based system. They have, during the course of the project,
adopted ideas from the original CSIRO prototype, developing a system appropriate to their needs,
using a Chinese version of Windows and have translated the Level5 Object@ user interface into
Chinese.

INTEGRATION-PEST MANAGEMENT WORKBENCH

Whilst methods for effective pest management in stored grains exist, they are not always used
in the most effective manner and one reason for this appears to be a lack of integration of technical,
economic, and management issues (Evans and van Someren Graver, 1989), the ‘institutional gap’
discussed earlier. ‘Grain Pest Adviser’, ‘Stored Grain Advisor’ and ‘PestMan’ attempt to address
part of this problem but were designed primarily to facilitate the active decision-making process,
although the inclusion of simulation models and extensive Help systems does make them useful
for training purposes. Specific Computer-Aided Learning (CAL) packages offer the potential for
much more effective communication of complex concepts to trainees. Such systems allow students
to explore alternative pest-management options and learn the consequences of their actions either
110 B. C. Longstaff

on their own or in small groups. With this in mind, a project has been initiated to develop a suite
of CAL decision tools to facilitate training of staff in rational and sustainable grain pest
management procedures in Indonesia, building upon the ideas from PestMan. The Indonesian
grain storage system currently relies on the combined use of fumigants and stack surface and
structural spraying with contact insecticides but the management philosophy is to improve
cost-efficiency and to reduce residues in the commodity. This project aims to achieve a more
effective integration of pest management strategies through an evaluation of the existing
management practices and by providing access to decision tools, thereby facilitating improved
understanding at all levels of the critical issues involved.
The various component systems under development are being integrated within a ‘Pest
Management Workbench’ (Fig. 9). The component systems will use many common databases,
including images, biological and chemical parameters, and language features, ensuring efficient use
of resources and consistent style (Halid et al., in press). The Workbench will have three
components:
l PestID-a sophisticated yet simple interactive key, derived from the LUCID system developed
by the Co-operative Research Centre for Tropical pest Management, Brisbane, for training field
staff to identify pest species. The system will be linked to the pest information database so as to
provide the user with detailed background information, where necessary;
l Teach-an interactive tutorial system covering a range of essential pest management concepts
and techniques. It includes the capacity to teach the logical processes used by grain storage experts
in the diagnosis of problems. It presents a series of pest management scenarios from which the user
selects information, applying the reasoning to make a diagnosis of the problem. The program then
offers the correct solution and describes the most efficient way of reaching this conclusion. This
feature is supplemented by tutorials on specific pest management topics and will include
high-quality graphics, video and sound. The program will be accompanied by a builder, enabling
trainers to develop applications to suit local conditions;

In brief - MbOllURI cartaneum

Order: Colaoptera
Family :Tenebrionidae
Common names : red flour beetle (rust red flour t

Flattened, reddish brow, parallel sided. Eyes

crescent shaped. Larvae elongate, light brown.

Major pest of stored products, sspsrially grain a~

milled cereal products e.g. flour and milled rice.
confueum replaces T. ceeteneum in cooler region
mills and milled products.

Lila l+l fl. -/)

Optimum: 20 days at WC, 75% r.h.
Range: 22 _ 40°C. 5unives very dry conditions
Maximum population growth rate per month: 70

Eggs : Laid at random

Larvae: Mobile, not concealed
Adult: long lived, feeds, infestations often leaving
unacceptable taint, strong flier, otten a very early
arrival. T. con?ueum not known to lly.
-1 distribution

Pm
r
Viaw ArMi inn Panal

Fig. 8. An example of PestMan on-line Help system, showing an insect and the underlined active fields
that will show additional information when clicked upon.
 Decision-support tools for grain pest management

PESTMANAGEMENTWORKBENCH
. .
I
4

Simulator PestID

Guidance

Ic

Models InformationBase

Population growth Pest handbook

lnfonnatbn

Demographic
Chemical data Text files Images
parameters
-8 L

Fig. 9. The organisation of Pest Management Workbench, illustrating the use of common resources by
different modules.

l Simulator-allows the user to explore various management strategies using simulated grain
storage complexes, based upon real sites and using real grain procurement and distribution data.
Decision-making features are being incorporated into this system, rather than having a stand-alone
expert system, to emphasize the consequences of particular actions to the user. It will include a
number of simulation models, covering insect population growth (Longstaff, 1988a), breakdown
of residual chemicals (Desmarchelier and Bengston, 1979), phosphine fumigation (Annis and
Banks, 1993) and the yellowing of rice (Bason et al., 1993).
Decision-making features of such a system will be incorporated into each of the modules. As
an integral part of the process of developing a bilingual capability for the software, the simulator
can be switched between English and Bahasa Indonesia. The use of a Windows-based GUI to the
system means that, although the underlying programming language is English, the GUI can be in
any language. In this project, a prototype interface has been developed that uses a Microsoft
ACCESS@ language database, allowing the system to have either an English or a Bahasa
Indonesian interface (Fig. 10). Conversion to any other language in the ASEAN region would be
very simple.

THE FUTURE

This paper has attempted to provide a perspective of the developing role of decision tools in
the management of pests of stored products. The term ‘decision tools’ has been interpreted in a
broad fashion to encompass knowledge acquisition, decision analysis, modelling, expert systems
and an integrated CAL approach. Improvements in decision-making procedures available to
managers and pest control staff should lead to greater application of effective storage procedures
and thus to reductions in overall storage losses and improved efficiency of pesticide usage. A
number of countries are placing increased reliance on fumigation to control insect pests and, whilst
fumigations under good conditions are still effective, the potential for the development of serious
levels of phosphine resistance has been clearly identified (Bengston, 1986). Cost-effective grain
storage management systems and practices that delay the development of resistance are thus clearly
beneficial.
 B. C. Longstaff

Fig. 10. A typical screen from the Simulator, showing the layout of a warehouse complex in Indonesia.
Each stack within each warehouse is monitored by the software, so that management processes, such
as inloading and outloading of grain, sampling, chemical treatment and fumigation are recorded. A
population growth model runs within each stack and insect densities are indicated by different colours.
Grain losses and treatment costs are shown in the lower right of the screen.

A key to enhancing the effectiveness of pest management lies in improving understanding of

underlying issues by all levels within a management hierarchy. An integrated CAL approach, such
as that of Pest Management Workbench, has the potential to make an important contribution to
this issue. By improving the ability of correctly identify insect pests, the PestID component will
facilitate use of the most appropriate control measure. Similarly, the Teach component will provide
managers and pest control staff with a much greater level of understanding of both the problems
and solutions to the problems. The Simulator will allow the user to explore a range of different
management strategies, their effects upon the quality of grain in store and the cost-effectiveness
of various strategies. It also has the potential to be used in a direct managerial context by providing
recommendations for specific courses of action. Such a system could be used as an on-line contoller
of aeration systems or as a monitoring station for temperature or acoustical sensors (Flinn and
Muir, 1994; Hagstrum et al., 1991). Gibbs (1994), for example, describes the development of a
remotely accessible PC-based system for controlling and monitoring aeration systems, using local
weather stations and arrays of temperature and humidity sensors within the grain stores. Such a
system, together with an acoustical detection system, could be incorporated into a derivative of
the Simulator and ‘intelligence could be applied to interpret the output from the aeration controller
and the acoustical sensors, and then to provide warnings of actual or potential problems and
recommendations for remedial action. By reducing the need to travel to remote sites (and in
Australia, they can be very remote) considerable cost-savings could be achieved. The need to have
trained operatives at such sites would be reduced, freeing up resources that could be used to
enhance the capability of regional centres.
Pest Management Workbench is an example of a specific implementation of a powerful generic
technology that will be of value for commodities other than rice and in many other agricultural
and horticultural industries in many countries, both as a decision-making support tool, and as an
educational tool to illustrate integrated storage pest-management principles. There is no doubt that
 Decision-support tools for grain pest management 113

the implementation of this type of technology will result in reduced commodity losses and improved
commodity quality during storage.

Acknowledgemenrs-The author would like to thank Peter Annis, CSIRO Division of Entomology, and Myron Zalucki,
Co-operative Research Centre for Tropical pest Management, Brisbane, for their comments on earlier drafts. Financial
assistance was provided by the Australian Centre for International Agricultural Research (ACIAR).

REFERENCES

Annis P. C. and Banks H. J. (1993) A predictive model for phosphine concentration in grain storage structures. In
Proceedings International Conference on Controlled Atmosphere andfumigation (Edited by Navarro, S. and Donahaye,
E.), pp. 299-312, Winnipeg, Canada.
Bason M. L., Ronalds J. A. and Wrigley C. W. (1993) Prediction of safe storage life for sound and weather-damaged malting
barley. Cereal Foods World 38, 361-363.
Bengston M. (1986) Background studies on fumigants: Session Chairman’s summary. In Pesticides and humid tropical grain
storage systems: Proceedings of an international seminar (Edited by Champ, B. R. and Highley, E.), p. 222, Manila,
Philippines.
Compton J. A. F., Tyler P. S., Mumford J. D., Norton G. A., Jones T. H. and Hindmarsh P. S. (1992) Potential for an
expert system on pest control in tropical grain stores. Tropical Science 32, 295-303.
Cuperus G. W., Fargo W. S., Flinn P. W. and Hagstrum D. W. (1990) Variables affecting capture of stored-grain insects
in probe traps. Journal of the Kansas Entomological Society 63, 486-489.
Denne T. (1988). An expert system for stored grain pest management. Ph.D. dissertation, London University.
Desmarchelier J. M. and Bengston M. (1979) Residual efficacy of grain protectants. In Proceedings of the Second
International Working Conference on Stored-Product Protection (Permanent Committee) pp. 138-151, Ibadan, Nigeria,
1978.
Evans D. E. and van Someren Graver J. (1989) Integrated pest and commodity management: putting first things first.
In Grain Postharvest Systems: Proceedings of the 10th ASEAN Technical Seminar on Grain Postharvest Technology,
(Edited by de Mesa), pp. 206217, Bangkok, Thailand, 1987.
Flinn P. W. and Hagstrum D. W. (199Oa) Simulations comparing the effectiveness of various stored-grain management
practices used to control Rhyzopertha dominica (Coloeoptera: Bostrichidae). Environmental Entomology 19, 725-729.
Flinn P. W. and Hagstrum D. W. (1990b) Stored Grain Advisor: A knowledge-based system for management of insect
pests of stored grain. AI Applications 4, 44-52.
Flinn P. W. and Hagstrum D. W. (1990~) An expert system for managing insect pests of stored grain. In Proceedings of
the Ffth International Working Conference on Stored-Product Protection, III (Edited by Fleurat-Lessard. F. and Ducom,
P.), pp. 201 I-2018, Boredeaux, France, 9-14 September 1990.
Flinn P. W. and Muir W. E. (1994) Expert System Concept. In Stored Grain Ecosystems (Edited by Jayas, D. S.,
White, N. D. G. and Muir, W. E.), pp. 33-54. Marcel Dekker, New York, USA. ISBN 0-8247-8983-O.
Flinn P. W., Hagstrum D. W., Muir W. E. and Sudayappa K. (1992) Spatial model for simulating changes in temperature
and insect population dynamics in stored grain. Environmental Entomology 21, 1351-1356.
Gibbs P. A. (1994) Development of a programmable aeration controller. In Proceedings of the 6th International Working
Conference on Stored Product Protection (Edited by Highley, E., Wright, E. J., Banks, H. J. and Champ, B. R.), pp.
286289, CAB International, Canberra, Australia, April 1994.
Hagstrum D. W. (1994) Field monitoring and prediction of stored-grain insect populations. Postharvest News and
Information 5, 39N45N.
Hagstrum D. W. and Flinn P. W. (1990) Simulations comparing insect species differences in response to wheat storage
conditions and management practices. Journal of Economic Entomology 83, 2469-2475.
Hagstrum D. W., Flinn P. W., Subramanyam B., Keever D. W. and Cuperus G. W. (1990) Interpretation of trap catch
for detection and estimation of stored-product insect populations. Journal of the Kansas Entomological Society 63,
500-505.
Hagstrum D. W. and Heid W. G. (1988) U.S. wheat marketing system: An insect ecosystem. Bulletin of the Entomological
Society of America 34, 33-36.
Hagstrum D. W. and Throne J. E. (1989) Predictability of stored-wheat insect population trends from life-history traits.
Environmental Entomology 18, 66&X4.
Hagstrum D. W., Vick K. W. and Flinn P. W. (1991) Automatic monitoring of Tribolium castaneum populations in stored
wheat with computerized acoustical detection systems. Journal of Economic Entomology 84, 16041608.
Halid H., Longstaff B. C., Annis P. C., Sidik M. and Turner J. (in press) Grain quality maintenance in the ASEAN region:
a role for decision-support tools. Biotropia.
Heong K. L. (1990) Computer expert systems for improving insect pest management. Review of Agricultural Entomology
78, l-11.
Hodges R. J., Sidik M., Halid H. and Conway J. A. (1992) Cost efficiency of respraying store surfaces with insecticide
to protect bagged milled rice from insect attack. Tropical Pest Management 38, 391-397.
Jones T. H., Mumford J. D., Compton J. A. F., Norton G. A. and Tyler P. S. (1993) Development of an expert system
for pest control in tropical grain stores. Postharvest Biology and Technology 3, 335-347.
Kawamoto H., Sinha R. N., Muir W. E. and Woods S. M. (1991a) Simulation model of Acarus siro (Atari: Acaridae)
in stored wheat. Environmental Entomology 20, 1381-1386.
Kawamoto H., Sinha R. N. and Muir W. E. (1991b) Ecosystem modelling to provide early warning of pest infestation
of stored grain. In Proceedings of the Fifth International Working Conference on Stored-Product Protection III (Edited
by Fleurat-Lessard, F. and Ducom, P.), pp. 2019-2026, Bordeaux, France, 9-14 September 1990.
Kawamoto H., Sinha R. N. and Muir W. E. (1992) Computer simulation modelling for stored-grain pest management.
Journal of Stored Products Research 28, 139-145.
Kawamoto H., Woods S. M., Sinha R. N. and Muir W. E. (1989) A simulation model of population dynamics of the rusty
grain beetle, Cryptolestes ferrugineus, in stored wheat. Ecological Modelling 48, 137-157.
114 B. C. Longstaff

Kenkel P., Adam B . D., Cuperus G. and Fargo W. S. (1991) A risk analysis of insect control strategies in stored wheat.
Oklahoma State University Agricultural Economics Publication # 9151.
Longstaff B. C. (1988a) A modelling study of the effects of temperature manipulation upon the control of Sitophilus orysae
by insecticide. Journal of Applied Ecology 25, 1633175.
Longstaff B. C. (1988b) Temperature manipulation and the management of insecticide resistance in stored grain pests: a
simulation study for the rice weevil, Sitophilus oryzae. Ecological Modelling 43, 303-313.
Longstaff B. C. (1991) The role of modelling in the management of stored-product pests. In Proceedings of the Fi#h
International Working Conference on Stored-Product Protection III (Edited by Fleurat-Lessard, F. and Ducom, P.), pp.
1995-2007, Bordeaux, France, 9-14 September 1990.
Longstaff B. C. and Cornish P. (1994) ‘PestMan’-A decision-support system for pest management in the Australian central
grain-handling industry. AI Applications 8, 1323.
Longstaff B. C. and Cuff W. R. (1984) An ecosystem model of the infestation of stored wheat by Sitophilus oryzae (L.)
(Coleoptera: Curculionidae): A reappraisal. Ecological Modelling 25, 97-l 18.
Metzger J. F. and Muir W. E. (1983) Aeration of stored wheat in the Canadian Prairies. Canadian Agricultural Engineering
25, 127-137.
Norton G. A. and Mumford J. D. (Eds.) (1993) Decision Tools for Pest Management. CAB International, Wallingford,
UK. ISBN 0851987834.
Sinclair E. R. and Alder J. (1985) Development of a computer simulation model of stored product insect populations on
grain farms. Agricultural Systems 18, 95-l 13.
Throne J. E. (1994) Computer modeling of the population dynamics of stored-product pests. In Stored Grain Ecosystems
(Edited by Jayas, D. S., White, N. D. G. and Muir, W. E.), pp. 169195. Marcel Dekker, New York, USA. ISBN
0-8247-8983-O.
Wileyto E. P. (1994) The statistical interpretation of insect self-marking and trapping. In Proceedings of the 6th International
Working Conference on Stored Product Protection (Edited by Highley, E., Wright, E. J., Banks, H. J. and Champ, B.
R.), pp. 459462. CAB International, Canberra, Australia, April 1994.
Wileyto E. P., Ewens W. J. and Mullen M. A. (1994) Markov/recapture population estimates: A tool for improving
interpretation of trapping experiments. Ecology 75, 1109-l 117.
Wilkin D. R. and Mumford J. D. (1994) Decision support systems for integrated management of stored commodities. In
Proceedings of the 6th International Working Conference on Stored Product Protection (Edited by Highley, E., Wright,
E. J., Banks, H. J. and Champ, B. R.), pp. 879-883. CAB International, Canberra, Australia, April 1994.
Wilkin D. R., Mumford J. D. and Norton G. A. (1990). The role of expert systems in current and future grain protection.
In Proceedings of the Fifth International Working Conference on Stored-Product Protection III (Edited by
Fleurat-Lessard, F. and Ducom, P.), pp. 2039-2046, Bordeaux, France, 9-14 September 1990.