Additive Mixed Modelling Applied on Phytoplankton Time Series Data

Abstract

This chapter looks at a data set where our first reaction was: ‘How in heavens name are we going to analyse these data?’ The data consist of a large number of phytoplankton species measured at 31 stations in Dutch estuarine and marine waters. Measurements took place 0–4 times per month from 1990 until present (2005). Environmental data (e.g. temperature, salinity, etc.) were also measured, albeit sometimes at different sampling times! The statistical analysis of these data is complicated for several reasons:

• 1. Environmental variables and phytoplankton variables were not always measured at the same time.

• 2. There may be temporal correlation, there may be spatial correlation, and both correlation structures may be complicated.

• 3. The data contain a large number of species.

• 4. The data are irregularly spaced.

• 5. There may be heterogeneity over time (e.g. more variation in summer than in winter).

• 6. Trends over time and in space may be non-linear.

• 7. The phytoplankton data were counted by different laboratories.