Evaluating an interpolation approach for modelling spatial variability in pest development

Air temperatures estimated by partial thin plate spline interpolation, or from the 'nearest station' (Voronoi polygon method), were used to model the phenology of three pests of horticultural crops throughout England and Wales. Temperatures for a particularly hot (1976) and a particularly cold (1986) year were interpolated to a grid resolution of 1 km. Estimates were made of the timing of spring emergence (Cecidophyopsis ribis (Westwood)), the maximum number of generations completed during the summer (Plutella xylostella (Linnaeus)) and the numbers of days when mating was possible (Merodon equestris (Fabricius)). The relative accuracy of the two temperature estimation methods was compared using jack-knife cross-validation. For C. ribis and P. xylostella, modelling with interpolated temperature input data was more accurate than using data from the 'nearest station'. Of the three phenology models used, the one that relied on an activity threshold (M. equestris) was the most sensitive to both types of input data. Spatial variability in the activity of M. equestrisadults was investigated in the two main areas (south-west peninsula and Lincolnshire) where its host crop (Narcissus) is grown. Modelling at cruder scales (up to 25*25 km) masked local variation, but the degree to which this was important varied from region to region and over time, as did the geography of the variability itself. The results indicate that interpolated data, computed to a resolution of 1 km using the UK synoptic network, have the potential for wider use within agricultural decision support systems for horticultural crops.