Use of models and phenological forecasting in applied entomology
Abstract
Although experienced modellers are usually aware of the dangers and benefits of the use of models in applied entomology, many scientists who may use models are not. Some remain sceptical that models serve any useful purpose, whereas others have unrealistic expectations about the information that models can provide concerning a particular problem. Here, the definition of a model is reviewed, a brief history of model use in general ecology is presented, and the limitations of uses of models in applied entomology are discussed.
Published research concerning insect development rate or in particular, phenology models, shows that the effect of fluctuating temperature and rate summation on nonlinear and linear model predictions is not well-understood. The influence of the diurnal range of the temperature regime on the rate summation effect, and the interaction between the rate summation and the mathematical structure of a particular development rate model, are clearly demonstrated. This study shows that the selection of a development rate function for prescriptive use based on closeness of fit to developmental data from constant temperature experiments is meaningless. Neither should selection be based on how well a model derived from constant temperature experiments predicts insect development measured under a narrow range of fluctuating temperatures. Theoretical reasons why a nonlinear model fitted directly to rates observed under variable temperature laboratory or field data is inappropriate for subsequent prediction, are clearly demonstrated. An appropriate technique that calculates instantaneous development rate functions from fluctuating temperature development time observations is tested with hypothetical data. The technique failed to give reasonable estimates of the parameters of the instantaneous development rate function that generated the hypothetical data.
Historical data of spring migration of the damson-hop aphid, Phorodon humuli (Shrank) recorded at Wye, Kent, and Rosemaund, Herefordshire; and phenology of overwintering host plant flowering recorded at East Malling, Kent, in England, were examined for possible associations. Relationships between mean temperature over a phenophase interval defined by flowering phenology of two overwintering host species of plum and the start of migration of P. humuli, versus the reciprocal of days over the interval, were significant for both sites.
Predictive sample reuse (PSR) methodology was used to validate the potential of derived models to predict the start of migration of P. humuli at the two sites. The predictive performance of the host plant flowering-aphid migration phenology models derived in this study was compared with that of other methods that use historical field data to predict insect life cycle events. The host plant flowering - aphid migration phenology models gave the best prediction of the start and 50% migration at Wye based on the criterion of lowest absolute mean error of the PSR residuals.
The value of using historical data of host plant and insect phenology to establish a biologically meaningful starting point for both the construction and application of predictive models of insect population events is illustrated. The desirability of comparing the performance of alternative phenology models that may be used for prescriptive use is discussed.
