A statistical procedure for unsupervised classification of nutrient limitation bioassay experiments with natural phytoplankton communities

Abstract

We describe a novel method for statistical analysis of bioassay experiments with natural plankton communities. The procedure allows unsupervised classification of the type of nutrient limitation in factorial experiments with two limiting nutrients, based on objective selection between several generic limitation patterns with direct biological interpretation, using the Akaike Information Criterion (AIC), which balances the concerns of model fit and model robustness. As such, it avoids the interpretation of nuisance parameters related to time effects in classical factorial design analysis of experiments with repeated measurements over time. The proposed limitation patterns discriminate between exclusive and primary limitation, depending on whether the initially nonlimiting nutrient has an effect or not. They also discriminate between two types of colimitation, depending on whether both nutrients only have an effect in combination, or also separately. The latter response is interpreted as the result of different phytoplankton community components being simultaneously limited by different nutrients. The capabilities of the classification procedure is demonstrated on a comprehensive set of 163 bioassay experiments with N and P additions to natural phytoplankton communities from the coast of Finland (NE Baltic Sea).