Search

This study investigated the potential global distributional shifts of poikilothermic invasive crop pest species associated with climate change, aiming to understand if their overall global distributions will expand or contract, and how the species distributions will vary across different regions. An ecological niche modelling analysis was conducted for 76 species. The potential distributional changes of the species in 2050 and 2070 were scrutinized for two climate change scenarios, which were further examined across different temperature and precipitation ranges. Results showed that averages of the mean probabilities of presence of the 76 crop pest species were predicted to increase. Higher species turnovers were predicted mostly to occur in areas with increasing predicted species richness. Lower species turnovers, however, were predicted mostly to occur in areas with decreasing predicted species richness. Species richness increases were predicted to occur more often in currently lower temperature (annual mean temperature approximately

The reduction of crop yield losses caused by pests is a major challenge to productive and sustainable food production for preventing food insecurity and emergencies, and for alleviating world food crisis. Multiple regression (MR) and artificial neural network (ANN) are two widely adopted modelling approaches for the prediction of crop pest risks, which are based on empirical statistics and artificial intelligence, respectively. Each of the two alternative approaches has its advantages and disadvantages. This study evaluates the two models from two aspects: their performances on pest risk prediction, and their methodological advantages and disadvantages. Two pest species are modelled using the two approaches as case studies, which are the melon thrip Thrips palmi Karny (T. palmi) and the diamondback moth Plutella xylostella (L.) (P. xylostella). Results show that ANN has higher prediction accuracy for both species. However, ANN has some methodological demerits compared to MR modelling.