Comparison of models for estimating the year-round range of Spodoptera frugiperda (Lepidoptera: Noctuidae) and predictions for New Zealand under current and future climates

Spodoptera frugiperda (Lepidoptera: Noctuidae), fall armyworm (FAW), is a pest of corn, maize and other crops that has rapidly spread from its native range in the Americas to much of Africa, India, Asia and Australia, and was first detected in New Zealand (NZ) in March 2022. To estimate where FAW will occur year-round in NZ, we evaluated a range of published models of FAW’s potential distribution in tandem with various publicly available gridded climate datasets. Model predictions were compared to geocoded observations of where FAW has survived or died during winter in southeastern China and southeastern USA. All of the combinations of model and climate dataset tested performed relatively well, with an isotherm model performing best. This and a consensus of three Climex models were used to produce predictions for New Zealand in current and future climates. They predicted that in current climates FAW will occur year-round in the northern third of the North Island and in warmer years may also persist year-round in some coastal parts of the southern North Island and northern South Island. As NZ’s climate warms, FAW will occupy marginal areas year-round with increasing frequency and by about 2080 will occupy them year-round practically every year. Concurrently the areas where it occasionally survives winter during warm years will extend further south and further inland. Our study demonstrated the benefits for illustrating uncertainty in model results of evaluating a wider range of models and climate datasets than is normal practice in species distribution modelling. Moreover, using annual climate datasets in addition to 20 year means provided valuable information about annual fluctuations in FAW’s potential year-round distribution that is often lacking in species distribution estimates. The work demonstrated the high value of recording species absences (or failures to survive at specified locations) for measuring model performance. Documentation of such observations is generally rare and greater awareness of their value might encourage more recording.