Analysis of Seasonal and Long-Term Population Dynamics for Modeling Populations at Low Density: Experience with Light Traps

Methods for estimating and modeling the long-term and short-term adult flight dynamics of the conifer silk moth Dendrolimus superans (Lepidoptera: Lasiocampidae) are examined. The analysis uses light trap adult catch data collected over 21 years, from 2005 to 2025. Three models of adult flight are considered: a flight-initiation model driven by weather factors, an autoregressive model of long-term catch dynamics, and a binary model of seasonal catch. For the flight-initiation model, we propose estimating the accumulated temperature sum ST from the date when the first derivative of the remote sensing vegetation index NDVI becomes positive until the date of the first adult capture of the season. ST is shown to be sufficiently stable across all years of observation, with flight each year beginning after this temperature sum is reached. The second model demonstrates that the long-term light trap catch time series is well described by a second-order autoregressive model AR(2), in which the catch of the current year depends on catches from the two preceding years. This long-term series is compared with a previously studied larval population density series of the Siberian silk moth; both are shown to be AR(2) series with similar coefficient values, which suggesting that adult catch data may serve as a proxy for absolute larval population density. In the third model, we describe the transition from absolute-scale seasonal catch dynamics (number of adults per day) to a binary scale (0, 1), where 0 denotes days on which no adults were attracted to the trap, and 1 denotes days on which at least one individual was captured. The seasonal absolute catch series is thereby transformed into a binary series of zeros and ones, and relationships between adjacent values in such a binary series are examined. A linear relationship between the absolute and binary seasonal dynamics series is demonstrated, making it possible to estimate absolute catches from binary catch values and to analyze seasonal flight in sparse pest populations. This potentially opens new avenues for understanding how outbreak populations function at chronically low density. Author summaryForest pests can cause catastrophic damage, yet predicting their outbreaks remains challenging. During periods of low population density, standard monitoring methods become labor-intensive and uninformative, while the transition to an outbreak often occurs unexpectedly. Using a 21-year dataset of adult Siberian silk moth (Dendrolimus superans) captures from light traps, we developed an approach combining three complementary models. First, we showed that moth flight begins upon reaching a specific temperature sum, with the starting point determined by NDVI vegetation index dynamics rather than a calendar date--making the forecast more ecologically relevant. Second, long-term adult population dynamics follow a second-order autoregressive model AR(2), matching the dynamics previously observed for larval populations. This establishes light trap data as a reliable proxy for absolute population density when ground surveys are impractical. Third, we introduced a method to analyze seasonal flight using binary data (presence/absence of moths per day), which we showed is linearly related to absolute abundance. This enables studying population dynamics during periods of extremely low density, when traditional methods fail. Our approach opens new possibilities for early warning systems to detect when a population risks transitioning from a latent state to an outbreak phase.