A Comparative Case Study of Dynamic Models for DO-BOD-ALGAE Interaction in a Freshwater River

From recent IIASA workshops on water quality modeling a need can be identified for comparative studies of different model types against the same set of field data. Similarly, some of the motivation for a state-of-the-art survey on water quality modeling to be prepared under the auspices of IIASA stems from the desire to bring order and authenticity to a fast-developing field of technology. The problem is as follows: although models can be readily applied in management and decisionmaking, they are not always so readily subject to a prior verification against field data from the river system. One reason underlying this problem is that the relevant field data, with a sufficiently high sampling frequency and collected over a sufficiently long period, either do not exist or have not been publicized. The primary objective of this paper is the dissemination of a set of time-series field data suitable for the identification and verification of dynamic models for water quality. Here water quality is interpreted as the interaction between three variables: dissolved oxygen (DO) concentration-a broad measure of the healthy state, or otherwise, of a river; biochemical oxygen demand (BOD) concentration -- a macro-measure of typical municipal/domestic organic waste materials; and a population of algae. A secondary objective is the comparison, by means of response error statistics, of several models that have been derived by reference to the field data. And yet a third objective is to present a summarizing and concluding statement on river water quality model development exercise which spans various publications over the past four or five years. With respect to model comparison and model assessment the paper concludes with a cautionary message on the use of simple fitting error statistics; and, in any case, it is argued that judgments about the "best" model are dependent upon the intended application of the model. On the accuracy of the models as representations of the real system it is found that many questions remain unresolved, and particularly so for those aspects of the models related to the growth kinetics, death, and decay properties of floating algal populations. The hope is expressed that, given the data, others will be stimulated not only to answer these questions but also to reassess the assumptions that the paper makes concerning the mixing and transport characteristics of the case study reach of river