How to analyze long-term insect population dynamics under climate change: 50-year data of three insect pests in paddy fields

We can precisely predict the future dynamics of populations only if we know the underlying mechanism of population dynamics. Long-term data are important for the elucidation of such mechanisms. In this article we analyze the 50-year dynamics of annual light-trap catches of three insect pest species living in paddy fields in Japan: the rice stem borer, Chilo suppressalis (Walker) (Lepidoptera: Pyralidae); the green rice leafhopper, Nephotettix cincticeps (Uhler) (Hemiptera: Deltocephalidae); and the small brown planthopper, Laodelphax striatellus (Fallen) (Hemiptera: Delphacidae). We separate the long-term dynamics into two components by using locally weighted scatterplot smoothing: (1) the underlying dynamics of populations, and (2) the influence of the past changes in the environment. The former component is analyzed by response surface analysis and vector autoregression to evaluate the nonlinearity of density-dependence and the inter-specific influence of density, respectively. On the basis of these analyses, we perform the state-space model analyses. The state-space model selected by Akaike's information criterion indicates that the observed number of light-trap catches of C. suppressalis and N. cincticeps in summer increases with increasing temperatures in the previous winter. It also indicates that the influence of temperature is not carried over to the next year. We utilize the selected model to predict the impact of global warming on these species, by substituting the temperature predicted by a general circulation model.