Drinking water biofiltration, assessing key factors and improving process evaluation

Abstract

Biodegradable organic matter (BOM) removal in drinking water biofilters can be affected by several factors. The factors investigated in this research were non-biological particles/coagulant in the influent, chlorine in the backwash water, air scour during backwashing, anthracite/sand vs. GAC/sand media, and low (5*C)/high (20*C) temperature operation. Other investigations included: impact of biomass accumulation within a filter run; impact of empty contact time (EBCT); and impact of step increases in BOM concentration and hydraulic loading rate, and filter shutdown on biofilter performance. The research also included the evaluation of biomass respiration potential (BRP) as an alternative biomass measurement, the development of a simple experimentally-based approach for estimation of key bio-kinetic parameters, and the modeling of BOM removal in drinking water biofilters. Laboratory scale biofilters fed a cocktail of easily biodegradable compounds were used for this research.

Fractional factorial design experiments in block I and II (phase I and II) showed that the three main factors (chlorine in backwash, temperature and media type) and their two or three-factor interactions were significant in most cases. The temperature effect was more significant when chlorine was present in the backwash water. AC filters were much more resistant to chlorinated water backwash than anthracite filters, as expected.

Further investigation of factors affecting drinking water biofiltration in phase III suggested that air scour, particle and coagulant effects were generally negligible although an air scour effect should be considered under unfavourable operating conditions (low temperature, chlorine, anthracite). Longer-term operation in phase III indicated that for easily biodegradable compounds (acetate, formate and formaldehyde), good removals were obtained after several months in the low temperature filter backwashed with chlorinated water (0.25 mg/L Cl2). Glyoxal removal was more sensitive to unfavourable biofiltration conditions than acetate, formate and formaldehyde. EBCT was not as important in biofilters run at the high temperature as at the low temperature. BOM removal was not sensitive to the biomass accumulation during a filter run.

The concept of bed utilization (the ratio of the bed depth required for substantial BOM removal to the entire bed depth) was introduced in the study. It could allow for better evaluation of the factors affecting biofiltration.

Based on the experimental results in phases I, II and III, multiple linear regression models were developed to roughly predict BOM removal for various conditions at contact times similar to those used in this research.

Mature biofilm in biofilters was able to maintain or quickly recover good BOM removal when exposed to a sudden increase either in BOM concentration or hydraulic loading rate. The impact of filter shutdown (24-h period, filter drained) was minor, at least for removal of easily degradable BOM components.

The amount of biomass on the filter media was evaluated by BRP (biomass respiration potential) and the phospholipid method in parallel. A good linear relationship was found between BRP and phospholipid biomass. The BRP method might be appropriate for routine use by treatment plant personnel due to its simpler and faster nature compared to other approaches.

The key bio-kinetic parameters (k and KS) were estimated in a dedicated experiment in which the biomass from the filter media and the BOM from the filter influent were used. The experimental conditions here closely simulated biofilters. The estimated kinetic parameters were expected to be more robust than previous estimates by using other approaches.

Three biofilter models, based on the simplified biofilm model, Suidan and Wang's semi-empirical equation, and Saez and Rittmann's revised solution, respectively, were applied to the modeling of acetate removal performance in biofilters in this study, by using the estimated bio-kinetic parameters (k and KS) from the experimentally-based approach in this research. The modeling results indicated that these three models are all applicable to the modeling of drinking water biofilter performance and they all provided a good prediction for acetate removal in biofilters for the conditions evaluated. The very close performance of these three models was also verified by a general comparison of the three models based on the estimated parameters from this study. The simplified biofilter model is recommended for its simplicity. Typical modeling results of biofilm thickness distribution, acetate profiles along the depth of biofilters, and acetate concentration in the biofilm could give some insights into the biofilter performance. A series of generalized curves was established to compare the percent removal of different substrates under various operating conditions among studies.

A revised Smin (minimum substrate concentration that can maintain a single layer biofilm in biofilters) and critical dimensionless contact time X*critical (beyond X*critical, little further removal is obtained) were proposed and evaluated in this study.