Keywords
data-based mechanistic modelling, algae concentrations, mechanistic model emulation, stochastic transfer function, uncertainty analysis
Start Date
1-7-2004 12:00 AM
Abstract
Water quality predictions in an ungauged catchment require the development of a model that is able to capture the basic physical features of the process and depends only on variables that are easily available. From this point of view, the model has similar requirements to those used in future climate scenario analysis. The mechanistic water quality model, developed in GKSS, Germany, for the purpose of climate change analysis, uses only climatic variables, such as temperature, radiation and discharge, to predict the time variability of algae concentrations. This paper presents the development of a statistical analogue to this mechanistic model. The goal of this research is the derivation of a data-based model that has the minimum number of parameters required to explain the data and, at the same time, is able to represent the physical features of the process (a Data-Based Mechanistic or DBM model). The approximation of the mechanistic model is obtained by a statistical analysis of the relations between the model input and output variables, as well as the linearisation of the mechanistic algae equations, leading to the development of a statistically tractable model. The result of this analysis is a nonlinear, Multi Input Single Output (MISO) transfer function model that provides a statistical counterpart of the mechanistic algae model. The model is used to reconstruct hourly chlorophyll-a concentrations (a measure of algae concentrations) during the “pre-unification of Germany” period (before 1990) in the River Elbe, Germany. The uncertainty of the predictions is assessed and the results are validated against available monthly chlorophyll-a measurements.
Water Quality Modelling in Rivers with Limited Observational Data: River Elbe Case Study
Water quality predictions in an ungauged catchment require the development of a model that is able to capture the basic physical features of the process and depends only on variables that are easily available. From this point of view, the model has similar requirements to those used in future climate scenario analysis. The mechanistic water quality model, developed in GKSS, Germany, for the purpose of climate change analysis, uses only climatic variables, such as temperature, radiation and discharge, to predict the time variability of algae concentrations. This paper presents the development of a statistical analogue to this mechanistic model. The goal of this research is the derivation of a data-based model that has the minimum number of parameters required to explain the data and, at the same time, is able to represent the physical features of the process (a Data-Based Mechanistic or DBM model). The approximation of the mechanistic model is obtained by a statistical analysis of the relations between the model input and output variables, as well as the linearisation of the mechanistic algae equations, leading to the development of a statistically tractable model. The result of this analysis is a nonlinear, Multi Input Single Output (MISO) transfer function model that provides a statistical counterpart of the mechanistic algae model. The model is used to reconstruct hourly chlorophyll-a concentrations (a measure of algae concentrations) during the “pre-unification of Germany” period (before 1990) in the River Elbe, Germany. The uncertainty of the predictions is assessed and the results are validated against available monthly chlorophyll-a measurements.
