The inadequacy of first-order treatment wetland models

Abstract

First-order models are in common use for design of treatment wetlands. These are frequently based on input/output (I/O) data; less frequently, on internal longitudinal transect data. The presumption is often made that the parameters of these models, i.e. the rate constants, are true constants and do not depend on factors such as hydraulic loading rate and inlet concentration. Another common presumption is that plug flow is a reasonable approximation to the hydraulic conditions in the wetland. This paper assembles a test wetland simulation, based on known information about vegetation resistance, treatment effects of vegetation, and residence time distributions. The test wetland is then used to provide simulations of different experimental and design protocols, such as transect measurements and I/O data from parallel and sequential detention time studies. Those simulations are shown to be consistent with real system data, thus confirming the data foundation of the test wetland behavior. The test wetland results, in common with the observations for nearly all treatment wetlands, show declining concentrations that approach a plateau at long detention times. The synthetic ‘data’ so produced is free from the stochastic vagaries of real wetlands, and thus provides a platform for understanding the deterministic component of behavior. The several variations of the plug flow model were then fit to the ‘data’, with generally excellent correlation coefficients ( R 2>0.95). However, the parameters (rate constants and apparent background concentrations) were found to be very strong functions of hydraulic loading and inlet concentration. This variability renders the models incapable of acceptable performance in design. Addition of a third parameter, such as a dispersion number, does not solve the inherent problems; nor does the retreat to loading regressions. It is suggested that new paradigms are needed that incorporate the ability to describe short-circuiting and spatial distributions of vegetation.