Developing Multisite Dynamic Models of Mixed Species Plant Communities

Abstract

Data on the development of two white clover cultivars (AberHerald and Huia) in mixed clover/ryegrass swards were available at 12 sites in Europe from experiments conducted for several years under a common protocol. Swards were measured up to seven times over winter and up to seven times over the growing season. In the overwintering period, detailed morphological measurements were taken for clover at each sampling time and, during the growing season, the clover contribution to total available biomass was recorded. Detailed meteorological data were available at all sites. The development of these clover/ryegrass communities over time and sites was modelled. The modelling strategy had three main elements: ( a) division of the annual growth cycle of the clover/ryegrass community into a number of functional periods; ( b) development of relationships within each functional period using models incorporating plot-level biotic variables characterizing each community at the start of the period and site-level climatic and management variables measured during the period; and ( c) introduction of a dynamic element by linking the models across functional periods. The response variable(s) for a functional period was the biotic independent variable(s) of the succeeding period. The object was to produce a dynamic series of models in which community development within and across sites was described as a resultant of the initial state of the community and climatic and other forces acting on it. The analysis used a mixed models technique in recognition of the complex error structure of the data. Various statistical aspects of the modelling are discussed including the models and fitting strategy used, the complexity of the error structure in an experiment that includes sites and years, and the desirability of transforming certain variables before modelling. The issues in presenting the results of a series of complex models are discussed and a graphical/tabular approach is outlined.