MODELING THE EPIZOOTIOLOGY OF GYPSY-MOTH NUCLEAR POLYHEDROSIS-VIRUS

Qualitative understanding of the dynamics of epizootics of the nuclear polyhedrosis virus of gypsy moth has become complete enough to justify attempts to quantitatively predict the timing and intensity of epizootics within a season. In earlier work (Dwyer and Elkinton, 1993), we compared the predictions of a simple differential equation model derived from Anderson and May (1981) to time series of virus mortality in each of eight gypsy moth populations (Woods and Elkinton, 1987). The model's predictions were very accurate for high density populations, but seriously under-estimated virus mortality in low density populations. Here we compare the predictions of the simple model to those of the gypsy moth life system model (GMLSM, Sheehan, 1985; Colbert and Racin, 1991), a highly complex computer simulation of gypsy moth population dynamics and forest stand growth that incorporates much of the existing knowledge of the many factors influencing gypsy moth populations. In particular we looked at two different versions of the GMLSM that incorporate two different models of virus transmission. One was identical to that of the simple model (Anderson and May, 1981; Dwyer and Elkinton, 1993), in which the rate of transmission was a constant (the transmission coefficient) times the product of the densities of healthy larvae and the densities infectious virus particles on foliage. The other approach, developed by Valentine and Podgwaite (1982) was a detailed model of production and consumption of infective particles on foliage. The model took account of age-related changes in the amount of foliage consumed and the variation in susceptibility of larvae to virus (LD(50)). The Anderson and May version of the GMLSM performed about as well as the simple differential equation model, but only for values of the transmission coefficient about 250 times higher that those we had determined experimentally (Dwyer and Elkinton, 1993). The Valentine and Podgwaite (1982) version of the GMLSM gave a much better fit, but only far values of LD(50) that were 100 times higher than those determined experimentally. Future research will focus on efforts to refine our understanding of virus transmission in order to explain and reduce the discrepancies between model predictions and observed mortality from virus in fieldpopulations.