Paul Mac Berthouex

Monitoring Sewage Treatment Plants: Some Quality Control Aspects

Common features of environmental quality data are serial correlation, seasonality, missing values, nonconstant variance, and nonnormal distributions. These features are found in air and water quality data, in biological and chemical data, and in data fr..

The use of stochastic models in the interpretation of historical data from sewage treatment plants

Abstract A 3-yr historical record of sewage treatment plant performance has been evaluated graphically and with time series methodology. Models relating influent biochemical oxygen demand to effluent biochemical oxygen demand, mixed liquor suspended solids, temperature, and hydraulic detention time were constructed. Four possible definitions of plant efficiency were defined and studied as well. The use of efficiency rather than effluent BOD for modeling performance was not helpful. A logarithmic transformation of BOD seems to be a better metric than BOD itself. Studying historical records should lead to an improved understanding of plant dynamics. Time series methods seem to be one useful tool for this work.

The effect of tracer diffusion in biofilm on residence time distributions

Abstract Estimating fixed-film biological reactor characteristics is often complicated by interaction of the tracer that is used in residence time distribution (RTD) studies with the biofilm in a way that causes the RTD curve to have a long tail. This tail can result from hold-up of tracer within the biofilm. A reactor model that ignores the diffusion rate at which tracer enters and leaves the biofilm may be seriously in error. Data from RTD studies in a fluidized bed reactor are presented and a model that accounts for diffusion of tracer in the biofilm is used to explain the experimental data. A material balance on inorganic nitrogen is also given to verify the organic dye tracer results. The conditions under which diffusion of tracer will cause large distortions are defined.

Dynamic behavior of an activated sludge plant

Abstract The purpose of this paper is to study how and to what extent effluent BOD 5 is related to influent BOD 5 and flow in an activated sludge process. The analysis is based on data collected hourly over a 2-week period at a Wisconsin sewage treatment plant. The methodology applied to establish a dynamic model for the system is that of Box & Jenkins (1970). With this approach, stochastic and transfer function components can be combined to form a model. The relative importance of these two components can be quantitatively assessed. A simple first-order model is able to explain the data very well. In this empirical model, which does not need flow as a predictor variable, the stochastic component is much more important than the transfer function component. Frequent return sludge control is not needed for the plant studied because it possesses remarkable inherent stability.

Environmental intervention analysis: Wisconsin's ban on phosphate detergents

Abstract Intervention analysis is a method of assessing the effect of a deliberate change to a system when the data record is a time series. A statistical analysis that assumes fixed mean levels before and after the intervention often is inappropriate for environmental time series, because typically environmental systems display some random drift in level over time, unrelated to the intervention itself. A reliable estimate of the effect of the intervention must taken into account such drift. The method presented does this by using suitably weighted averages of observations before and after the intervention; also it can allow for a gap in time over which the intervention gradually realizes its full effect. The method has been used to estimate the decrease in total phosphorus load on the two large wastewater treatment plants in Milwaukee, WI, caused by the ban on phosphate laundry detergents that went into force on July 1, 1979. The estimated effect was equivalent to 0.506 kg P per capita per year. The result of this ban being lifted in 1982 was also estimated; the increase in total P loading was equivalent to 0.187 kg P per capita per year.

Modeling and control of a phosphorus removal process by multivariate time series method

A control equation was developed for reducing the variability of effluent total soluble phosphorus in a 200 MGD (8.76 mJS−1 chemico- biological process at Milwaukee, Wisconsin, U.S.A. A statistically adequate multivariate time series model. Autoregressive Moving Average Vector model (ARMAV) was obtained using the operating data. This model describes the dynamic and stochastic characteristics of the process in relation to total soluble phosphorus removal process. The control equation derived from the model is based on the minimum mean square error (MMSE) strategy, a feedforward, feedback control scheme. The result of implemeting the control scheme will be approximately a 60% reduction in the variability (variance) of effluent total soluble phosphorus when compared to the existing feedforward control scheme at the treatment plant. Field testings have yet to be performed to validate the analysis.

Intervention Analysis of Seasonal and Nonseasonal Data to Estimate Treatment Plant Phosphorus Loading Shifts

Publisher Summary This chapter presents an intervention analysis of seasonal and nonseasonal data to estimate treatment plant phosphorus loading shifts. In the non-seasonal model, an exponentially weighted moving average (EWMA) is used to estimate the levels immediately before and after the gap. The magnitude of the effect of the intervention is the difference of the two, since the forecast for ARIMA or observation error/random walk mode. The seasonal model incorporates an annual pattern, but it is not a deterministic sine pattern; instead, it allows more flexibility. The smoothness-prior state space approach was used in a study given in the chapter to estimate the intervention effect. The calculations required for analysis of the non-seasonal and seasonal data sets were readily programmed in APL and carried out on a microcomputer. An ARIMA time series model has been used successfully on about sixty percent of the dataset analyzed to date. A seasonal model has been successfully applied to three data sets. The model may possibly be adequate for the remaining sets. The effect of a detergent phosphate ban on influent wastewater treatment plant P loads appears to be about 0.3 kg/cap. yr.

Deriving components of variance in a transfer function-noise model

Abstract The total variation in Y t can be separated into three parts: (1) measurement errors in Y t accounting for 8% of the total variability; (2) variation in X t transferred to Y t accounting for 34% of the total variability; and (3) as much as 58% of the total variability being due to unidentified sources acting through the first-order system. It is the second part, variation in X t transferred to Y t , that is the contribution of the transfer function so the transfer function is an important part of the model even though it explains less than half the total variation. The purpose of this note has been to show how the components of variance can be derived. It is not our purpose to press a claim that the models presented are the true models for the system in question, nor that they have general validity at different plants. The experiments were severely restricted with regard to being able to perturb the process and excite the dynamics of the system. Because of this limitation, which is serious, indeed, it is likely that all the dynamics of the system have not been revealed, in which case the transfer function of the model cannot include them.

Data management for land disposal of liquid sludge

Abstract Metrogro is a program for augmenting the soil on privately-owned farms by application of liquid digested sludge. The system has been operated since 1979. In 1982, more than 60 farmers had approx. 2900 acres of crop land treated. The success of the program has been due to the active marketing and service program as well as to financial savings to farmers who may save more than

Statistics for Environmental Engineers

60 per acre in fertilizer costs. A comprehensive program to monitor the soil, crop, sludge, and groundwater provides large amounts of information. The data base now includes data on the physical and chemical properties of the sludge, soil chemistry, amounts of sludge hauled and applied, and information on more than three hundred wells in the region. These data are organized into seven computer files which can be queried and manipulated using a data management program.