Limits to inferring population-environment interactions with coarse time series data
Time series data for many animal populations is collected yearly. Standard statistical methods require that environmental covariates are summarized at the same sampling timescale, despite the fact that most populations interact with environmental factors continuously throughout the year. Here we use theory combined with empirical models to determine how well yearly summary statistics perform as predictors when the population interacts with the environment continuously.
Our results predict that average yearly environmental conditions can yield accurate population growth predictions when the lifetime of individuals is long relative to the timescale of environmental fluctuations. When life history timescales are fast, average environmental predictors may not be useful and in some cases appear statistically indistinguishable from noise. We show that under conditions where average environmental conditions are not good predictors of the population both the scale (variance) and ordering (autocorrelation) of the environment are important to dynamics. This work highlights the limitations of current approaches and highlights a way forwards to improve temporal predictions of coarsely sampled populations using high-resolution environmental data.