Pink bollworm, Pectinophora gossypiella, is a major pest in almost all of the cot- ton growing areas of the world, causing both quantitative and qualitative losses (Taneja and Jayaswal, 1981). Control of the pink bollworm is normally achieved by enforcement of a close season (Hill, 1983) when no cotton is grown and emerging adults can find no hosts on which to oviposit – a suicide emergence (Ingram, 1980). However, in situations where control is not obtained through the use of a close season then chemical control may be necessary. Chemical control against the larvae is usually ineffective because they feed within the fruiting bod- ies where they are protected. Conventional insecticides are applied mainly to control the number of adult moths and the larvae not yet bored into the tissue (Taneja and Jayaswal, 1981).
The cost of chemical control must be more than compensated for by the increased yield obtained from controlling the pest. Hence, some measure of the level of infestation that causes damage or yield loss must be determined. With the pink bollworm there is a direct relationship between cotton boll damage and the level of infestation. The presence of a larva in a cotton boll may be taken as a direct indication of boll damage, hence the percentage of bolls infested in a sam- ple will provide an estimate of overall percentage damage in a field. The number of male adults can be easily monitored using a pheromone trap. What is needed is to establish if there is a relationship between adult numbers (represented by pheromone trap catches) and the boll damage as measured by the percentage of infested bolls.
Sampling of adult male populations was carried out using pheromone traps and related to larval boll infestations in the 1976–77 and 1977–78 seasons. A sin- gle pheromone trap (omnidirectional water trap) was set up in a field of cotton 11 weeks old and daily catches were recorded until the plants were uprooted at 28 weeks. The trap was positioned just below canopy height and was raised as the plants grew taller. The pheromone attractant septum was replaced every two weeks. Samples of 100 large green bolls were selected randomly and opened so that the percentage mined by pink bollworm could be determined. Boll sampling was carried out at weekly intervals, from week 13 until week 26. The data from the 1976–77 cotton season are shown in Fig. 2.14.
2.5.3 Predictive models
A model is a simplified representation of a system (Holt and Cheke, 1997). Within the context of forecasting, models are the rela- tionships that are quantified in some way that enable prediction of likely incidence at some future time. The temporal and spa- tial scales of such forecasts tend to be dependent on the type of monitoring that is undertaken. Where monitoring is based at the farmer’s field level, forecasts are usu-
ally short term and restricted to a particu- lar crop within a restricted area. They usu- ally employ easy to use monitoring systems, e.g. pheromone or coloured traps that provide simple population counts and a means by which a forecast can be deter- mined.
The models commonly involve regres- sion or multiple regression analyses of insect number of one stage against another (adults vs. larvae) or against damage.
The mean daily catch of the total number of moths trapped for the week end- ing on the day on which the boll samples were taken was regressed against the weekly percentage of damaged bolls (Fig. 2.15). The relationship was shown to be significant for both years and an analysis of variance showed there were no differences between slopes (F = 0.58) or elevation (F = 0.28).
Spraying is usually advised when 10% of the bolls are infested, but this level should be adjusted each season when the cost of inputs can be weighed against the expected increase in pickable cotton and the current price obtainable for the lint (Ingram, 1980). The price of a product and the cost of producing it are going to vary between seasons. Hence, the level of damage that can be tolerated before it becomes economically viable to control the pest is also going to fluctuate. This in turn will influence the threshold value of the catch data used to monitor pop- ulation change. In this context a relationship similar in form to Fig. 2.15 will be useful since then new catch thresholds can be easily estimated once a particular seasonal damage threshold has been identified. However, this does depend on there being a consistent relationship between the catch and field data, in both time and space. The existence of relationships can depend on many of the factors already discussed in this chapter, as well as the life history strategy and popula- tion phenology of each particular pest. Each pest must be assessed according to its particular circumstances, but it would be advisable to study population devel- opment in a number of fields and over a number of seasons, to validate any rela- tionship to be used in field forecasting.
Fig. 2.14. Weekly boll infestations (d) and nightly pheromone trap catches of the pink bollworm, Pectinophora gossypiella, during the 1976–77 cotton season, Barbados (after Ingram, 1980).
Fig. 2.15. Relationship between pheromone trap catches of pink bollworm and the percentage damaged bolls in the 1976–77 (s; ––––) and 1977–78 (d; – – – –) seasons and for combined data ( ) (after Ingram, 1980).
Regional forecasts tend to be based on the long term collection of population data, normally trap catch, at a specific fixed location. Traps that sample aerial populations such as light traps and suction traps are most commonly used. The catches are often collected on a daily basis and are characteristically collected for a number of years in order to establish the presence and form of seasonal population trends. Often it is hoped that the trap esti- mates may reflect regional population changes and be used to forecast levels of field infestations and provide spray warn- ing for farmers.
The advantages of such fixed position monitoring systems used for regional fore- casts are:
1. It can be used in areas where personnel for field inspections are limited.
2. Few traps may be required, hence lim- ited resources can be centralized (perhaps at agricultural research stations) where there are the trained personnel necessary to collect, sort, analyse and interpret the trap data.
The efficient dissemination of the forecast information and complementary field inspection may also be necessary.
Fig. 2.16. Regression predicting the expected level of crop infestation by Aphis fabae (log % bean plants infested). From the spring aerial migration sample (Way et al., 1981).
There are a number of disadvantages with a fixed position monitoring system for use as a regional forecasting programme.
One of the biggest disadvantages is that data in such a system need to be collected for at least 10 years in order to validate the forecasting system (Way et al., 1981), espe- cially if only one or a few traps are used, since then replication and validation is only assessed between years. This means the use of a fixed position monitoring sys- tem in order to provide a regional forecast- ing programme is a long-term project, requiring a large investment of resources but from which there is no guarantee of a successful forecasting system. The cost of the monitoring technique may be reduced and the probability of obtaining a success- ful forecasting system is enhanced if the data for a number of important pests can be collected and studied at the same time.
The important components of a fixed position monitoring and forecasting system include:
1. A thorough knowledge of the biology and seasonal cycle of each pest (Lewis, 1981).
2. An appropriate monitoring technique for both aerial and field populations.
3. A consistent relationship between catch estimates and field infestation.
4. A relationship between the level of field infestation and damage and if possible between trap estimates and crop damage.
5. Sufficient replication in time and space for validation of the forecasting pro- gramme.
A number of such regional forecasting systems exist (Way et al., 1981 (Fig. 2.16);
Pickup and Brewer, 1994; Linblad and Solbreck, 1998).