Environmetric time-series analysis: modelling natural systems from experimental time-series data

Using the modelling of solute transport in flowing media as an example, this paper outlines the main aspects of a systematic approach to the modelling of natural systems from experimental time-series data. The objective of the approach, which exploits sophisticated methods of recursive parameter estimation, is to produce a parametrically efficient, data-based model which is both physically meaningful and statistically well defined. Although the proposed methodology has its origins in systems and control theory and may be unfamiliar to some natural scientists, it has been developed and refined for use with natural environmental systems over the past 20 years, and has wide application potential in areas such as biology and ecology. In this sense, the paper is intended to introduce the more general reader to the topic, in the hope that the tutorial review and practical examples will stimulate interest and encourage reference to the many publications cited in the paper. The practical examples are concerned with the modelling of pollutant dispersion in stream channels: phloem translocation and carbon partitioning in plants: and rainfall-streamflow modelling in a river catchment.