Dynamic models for longitudinal butterfly data

There has been recent interest in devising stochastic models for seasonal insects, which

respond rapidly to climate change. Fitted to count data, these models are used to construct

indices of abundance, which guide conservation and management. We build upon Dennis et

al. (2014, under review) to produce dynamic models, which provide succinct descriptions of

data from all years simultaneously. They produce estimates of key life-history parameters

such as annual productivity and survival.

Analyses for univoltine species, with only one generation each year, extend to bivoltine

species, with two annual broods. In the latter case we estimate the productivities of each

generation separately, and also devise extended indices which indicate the contributions

made from different generations.

We demonstrate the performance of the models using count data for UK butterfly species,

and compare with current procedures which use generalized additive models. We may incor-

orate relevant covariates within the model, and illustrate using northing and measures of

temperature. Consistent patterns are demonstrated for multiple species. This generates a

variety of hypotheses for further investigation, which have the potential to illuminate features

of butterfly phenology and demography which are at present poorly understood.