Many ecological and environmental time series are irregularly sampled in time. This is especially so for long data series (for example due to varying funding levels over time) and palaeoecological time series from sedimentary archives. In the case of palaeoecological data, the sampling is often truly irregular owing to unknown variation in sediment accumulation rates at the time of collection. Additive models have been used successfully to model temporal trends in the mean of ecological and environmental time series. However, often other parameters related to the variance or the shape of the response distribution are of particular interest (e.g. increasing variance as an early warning indicator of critical transitions ). Here, I describe the use of location-scale additive models to model temporal change in the mean and variance of lake sediment time series records of systems experiencing environmental change; specifically lakes affected by eutrophication. Using annually resolved sediment time series I compare results from the location-scale models with dynamic linear model and stochastic volatility model fits.
Results/Conclusions
I identify statistically significant increase in variance prior to the onset of environmental, indicating altered community dynamics following nutrient enrichment. The loction-scale method is a powerful approach to modelling other moments of the distributions of species and for addressing questions such as altered community dynamics.