Naturally maize prices change in a relative and, to some extent, predictable fashion during the course of a year in response to demand. Prices tend to be lowest just after harvest, in August and September throughout much of Benin, and highest at the end of the dry season when grain is needed for planting, individual stocks of staple food grains are low and few alternative fresh food sources are available. Maize prices also change from year to year, often with respect to quantity and timing of rainfall - and in response to government policy on grain importation. Poor weather in a neighboring province or country may cause milder price rises in the surrounding regions. The value of maize also depends on its quality. Some varieties have higher economic value than others, and damaged maize, which is often sold alongside ‘clean’ maize, can change in its value relative to the clean maize. The definition of 'clean' maize also changes during the course of the season, as does the moisture content, which can significantly affect the weight of the maize being traded. In order to quantify these changes, we collect data on maize, including price, variety, grain moisture content and insect density and damage, every three months in at least ten markets across Benin. Quantifying factors such as the preference by farmers for certain varieties (e.g. Defoer et al., 1997) will be done with the farmers themselves.

Before any particular sampling plan can be responsibly proposed to either extension agents or farmers, two kinds of parameters need to be determined (for a review of the concepts discussed here, please see Pedigo & Buntin [1994]). One of them is the error rate of the sampling plan, that is, the probability of making a decision that the insect densities are high enough to justify an intervention when they are not, or that the densities are at a safely low level when in fact they are too high and action should be taken. Rather high error rates, such as 0.1, are often accepted for field sampling plans (Binns, 1994). However, any assumptions would need to be carefully assessed in relation to the economic circumstances of small-scale farmers storing maize and their perception of risk. Many farmers depend on their grain store for the subsistence of their household and have few alternatives or financial reserves on which to fall back, in the event of making a bad decision regarding store management.

The second kind of parameter is the economic injury level (EIL), that is, the insect density at which the cost of the pest damage exceeds the cost of the intervention strategy and a decision should be made (technically, the actual point at which a decision should be made about management, called the ‘action threshold,’ allows for the time delay between implementation and effect). The management decision itself usually involves some risk - of the unnecessary purchase and application of pesticide, for example, or of losing maize later on because of a decision to do nothing, so proper estimation of the EIL is important. The EIL and the action threshold are complex parameters, and components include the price of the maize at the time it would be sold, the cost of treatment relative to the value loss caused by the insect, the recruitment rate of the pest population in the store and other factors. The EIL is rarely known or agreed upon for pests in West African agriculture, and would be expected to vary greatly with respect to geographic location. One output of an economic analysis would be a map of EIL indices across different regions for different times of the year.

Two basic kinds of sampling plans for whole-cob grain stores are destructive and nondestructive plans (shelled maize is another matter). Nondestructive plans, such as examining the external state of the grain store, or removing some cobs for visual examination and replacing them in the store, have the advantage of being inexpensive in terms of maize cobs. Examining the appearance of the grain store is relatively easy if the cobs are exposed, as in an "awa" store, but may be difficult in enclosed structures, depending on the particular design. Another advantage of nondestructive sampling is cost in materials; since sampled maize cobs are not permanently removed, a particular store can be evaluated several times for the expense of only the

labor involved. However, there are some drawbacks. In order to make a decision based on the appearance of the grain store, the appearance must be quantified and associated with either beetle density or grain loss inside the store. The error associated with any particular appearance parameter, such as percentage of cobs showing signs of beetle attack, is difficult to estimate and the observed damage is often difficult to associate reliably with a particular insect species. Since the maize is only examined externally, where pests are only seldom found, estimates for population density and grain loss must be made indirectly, and that can be a source of a great deal of error. Emerging S. zeamais or S. cerealella, for example, might damage maize in an outwardly similar way to tunneling P. truncatus, but the threat posed by each species is quite different. In the case of a cob-by-cob inspection, the overall error can be reduced by taking more samples, but the additional number of cobs needed may be vast.

Destructive sampling plans can provide direct measures of insect density and grain loss, but only at the expense of grain being removed from the store. Destructive plans have the advantage of being much easier to evaluate and of requiring fewer cobs before a decision can be made about an economic activity. Techniques for constructing sampling plans using direct observations are well established; we have developed sequential sampling plans using destructive sampling techniques in rural grain stores (Meikle, unpublished data). Sequential sampling plans, which involve sampling cobs individually in a sequential manner, involve keeping a running total of insects along with the total number of cobs sampled. If the running insect total crosses the upper boundary (or ‘stop line’), which is itself determined using error rates and the EIL, the density is judged high enough for a management strategy to be implemented. Similarly, if, in the course of the sampling, the total falls below the lower stop line, the pest density is judged insufficient to pose a threat. Usually, sampling continues until one stop line or the other is crossed. While it is not necessary for a farmer or an extension agent to understand the various assumptions used to generate the stop lines, proper identification and counting of the insects is essential.

In practice it is very unlikely that farmers or even extension agents would adopt sequential sampling procedures using destructive techniques, at least in this form. However, perhaps the main value in developing formal sampling plans in this conventional, rigorous way, lies in the way the exercise forces us to analyze the storage system, define the management options available to the farmer and the factors that might weigh in the timing and quality of a decision. For instance, studies of P. truncatus in grain stores have usually shown that population densities even in severe infestations were seldom detectable before

The synthesis of modeling and GIS approaches provides us with a common framework within which to integrate our understanding of both biological processes within a grain store under a specified agroclimatological regime and of the economic factors at work within a particular policy environment. This holistic analytical approach will help us to develop sound decision-support tools for policy-makers, researchers and extensionists working to enhance sustainable pest control options available for management of rural grain stores. Within the next year we hope to i) complete the linking of the population models to build an integrated model of insect density and grain loss in the grain store ecosystem; ii) develop maps, using the model, of potential pest damage in storage systems in West Africa, and iii) extend analyses to include fungal pathogens and grain store economics. We are actively soliciting the cooperation of partners in different parts of Africa to collaborate in this continuing effort.

It should be emphasized, in concluding, that the modeling approach presented here is in no way intended to substitute for on-farm farmer-participatory work. On the contrary, the two approaches are complementary and indeed synergistic. Modelling can help us (especially as research and policy-makers) to understand the biological and socio-economic processes which underlie phenomena observed in specific studies. Analyzing these processes can help us to understand which factors are location-specific and which are likely to play a wider role. Detailed on-farm work, whether in research or extension, is relatively expensive in time and resources. Modelling- and GIS-based approaches can help us to target such efforts to where they are most needed and are most likely to achieve an impact.