We studied the relationship between counts of Panonychus ulmi (Koch) winter eggs per bud and mite-days per leaf accumulated in early to mid-summer on 92-96 apple trees in an orchard in Quebec, Canada. Stepwise regression was used to compute mite-days in the 4 year (1992-1995) data set using winter egg density, cumulative rainfall (R), cumulative degree-days (D), squared values of R and of D, and the product RD as potential predictors. Degree-days were accumulated above 10.1 degrees C, the minimum threshold of development for P. ulmi in Eastern Canada. The 4-year model that gave the best fit included terms for winter eggs, rainfall, R2, and the product RD, and explained 76% of the variation in mite-days. When the 4-year model was applied to each year's data separately, winter eggs, adjusted for current year rainfall and degree-days, were always highly significant predictors of mite-days. These significant effects of weather indicate that estimation of potential economic loss should take account of rainfall and heat units, as well as the density of winter eggs. Predators, including the stigmaeid, Agistemus fleschneri Summers, and occasional low numbers of phytoseiids, did not have any evident within-season effect on mite-days but their appearance in July-September of 1994 was followed by a ten-fold decrease in the mean density of winter eggs in 1995 compared with the previous spring. This reduction contrasted with increasing densities each successive spring from 1992 to 1994, which followed summers when predators were scarce or absent.