Redefining the stormwater first flush phenomenon

Abstract

The first flush in urban runoff has been an important, yet disputed phenomenon amongst many researchers. The vast differences in the evidence could be solely due to limitations of the first flush current definition and the approach used for its assessment. There is a need for revisiting the first flush theory in the light of its practical applications to urban drainage management practices. We propose that a catchment's first flush behaviour is to be quantified by the runoff volume required to reduce a catchment's stormwater pollutant concentrations to background levels. The proposed method for assessment of this runoff volume starts by finding the average catchment pollutant concentrations for a given increment of discharged volume using a number of event pollutographs. Non-parametric statistics are then used to establish the characteristic pollutograph by pooling statistically indifferent runoff increments (known as slices) together. This allows the identification of the catchment's initial and background pollutant concentrations and for quantification of the first flush volume and its strength. The novel technique was used on seven catchments around Melbourne, Australia, with promising results. Sensitivity to the chosen increment of runoff (for which mean concentrations are calculated) indicated that when dealing with discrete flow-weighted water quality data, a suitable slice size should closely match the flow-weighting of samples. The overall sensitivity to runoff increment and level of significance was found to be negligible. Further research is needed to fully develop this method.

Introduction

It has become a reasonable assumption that the initial volumes of runoff in urban catchments during rainfall events contain the highest pollutant levels, a phenomenon known as the first flush (Bertrand-Krajewski et al., 1998). Identifying the nature of this phenomenon has significant benefits for the implementation of stormwater quality management practices (e.g. sizing of treatment systems). However, watershed managers and researchers have questioned the existence of a first flush and the applicability of the half-inch rule (which assumes that 90% of an event's total pollutant load is transported in the first half inch of runoff) as a relevant volume for treatment (Bertrand-Krajewski et al., 1998, CWP, 2005, Kaspersen, 2008). The main cause for this lies within the method traditionally used to assess and quantify a first flush. This traditional approach, adopted in almost all first flush investigations (e.g. Saget et al., 1996, Deletic, 1998, Deletic and Maksimovic, 1998, Lee and Bang, 2000, Barco et al., 2008, McCarthy, 2009), involves the use of dimensionless cumulative pollutant load vs. cumulative runoff volume curves. An arbitrary definition for detecting the presence of a first flush is then applied to these curves. Some definitions are as strict as 70–80% of total pollutant mass transported in the first 25–30% of runoff volume (Wanielista and Yousef, 1993, Sansalone and Buchberger, 1997, Bertrand-Krajewski et al., 1998, Deletic, 1998, Deletic and Maksimovic, 1998).

An alternative methodology for a first flush detection is the Mass First Flush Ratio (MFFR), which is the division of the proportion of mass by the cumulative runoff volume at a defined point (e.g. if 80% of pollutant mass occurred at 30% runoff volume, MFFR = 2.67). However, this still relies on the same dimensionless curves and an arbitrary definition (Lee and Bang, 2000, Han et al., 2006, Barco et al., 2008).

Many researchers have found that a first flush was often only present in a small proportion of the analysed events (Saget et al., 1996, Deletic, 1998). Less than half of the events observed by Deletic (1998) showed greater than 40% cumulative pollutant mass in the first 20% cumulative runoff volume (referred to as a 40/20 first flush definition). Saget et al. (1996) only observed the first flush once in 197 events using an 80/30 definition, which was also adopted by Bertrand-Krajewski et al. (1998) who did not detect a first flush according to this definition. If any of these authors chose another definition, then the resultant conclusions may have changed. Multiple regression analyses have also been applied to help explain why events, and catchments, experience different degrees of a first flush (i.e. why one event had 80% mass in 30% volume, whilst another had 20% mass in the same volume). Most studies which report these regressions showed little consistency in variables which could explain the first flush variation (Deletic, 1998, Bertrand-Krajewski et al., 1998, McCarthy, 2009). Bertrand-Krajewski et al. (1998) highlighted the difficulty of establishing clear statistical relationships to explain results from the traditional approach. It is clear that there is the underlying problem in the way the first flush has traditionally been assessed, partly because of the lack of standardisation in the methodologies. This creates subsequent problems in understanding water quality behaviour between catchments and reduces the reliability of further analyses due to the arbitrary definition of the first flush adopted.

Another deficiency of the traditional approach is the dimensionless nature which therefore disregards the impact of storm volume. The total volume of a small event may only be as much as the first 20% of the larger event, and therefore could be considered to be entirely within a first flush volume. Kang et al., 2006, Kang et al., 2008 identified two types of pollutant source in an urban catchment: (1) short-term, that is the accumulated pollution during the dry weather period, which can be depleted if the storm is sufficiently long and intense enough and (2) long-term, that is the catchment's background pollutant levels, which cannot be depleted. If a small event cannot deplete the short-term pollutant source, pollutant concentrations will be consistently high. This explains why several studies found that smaller storms did not exhibit a strong first flush (if at all), as opposed to larger storms (Deletic and Maksimovic, 1998, Kang et al., 2006, Kang et al., 2008, Barco et al., 2008).

Another problem with the traditional method is that it characterises the first flush based solely on changes in pollutant loading within the event and not with the catchment's overall background levels. For example, 80% of the total load in a short event may be diluted in a longer event and thus represent only 60% at the same cumulative runoff. If the short-term pollutant source is depleted early in a very long event, this initial fraction may further decrease and possibly mask a first flush effect due to the influence of pollutants in the latter part of the event.

When sizing treatment systems in practice it is necessary to understand what volume of water should be treated, not what percentage of an event (since this will vary between events). Therefore, the non-dimensional nature of the traditional approach, as discussed above, is not very helpful to practice. A rare attempt to relate the first flush to an absolute volume of runoff and to observe changes in pollutant loads throughout several events over longer periods is presented by Lee et al. (2004).

Kaspersen (2008) suggests that issues of first flush detection can arise with different pollutants. Assessments on different pollutants have indicated problems with the traditional approach. Investigations by Taylor (2006) and McCarthy (2009) found that Escherichia coli (E. coli) and Total Nitrogen (TN) (respectively) exhibit so-called “end flushes” (i.e. high pollutant concentrations towards end of the event). McCarthy (2009) concluded that the comparison of initial and end pollutant concentrations, as is carried out in the traditional approach because of its dimensionless nature, is ill-posed for detecting and quantifying a first flush. If pollutant concentrations were both high at the start and at the end of the event then a first flush would not be detected.

It can be concluded that past research has shown frequent deficiencies with the traditional approach, mainly due to its dimensionless nature. These warrant the need for a revision of this method or a new technique, which should address the following issues:

• A more sound quantitative method should exist to justify the occurrence and character of the first flush as opposed to an arbitrary definition selected by the researcher,

• The first flush should be defined by actual runoff volumes and actual pollutant concentrations instead of proportions, as these quantities may serve as a guideline for management strategies and prevent over/under design of treatment systems (including treatment of ‘end flushes’).

• The method should be applicable to an array of different pollutants and not falter due to the behaviour of certain ones.

The objective of this paper is to demonstrate a new assessment technique to detect the presence and assess the characteristics of a first flush. The method aims to provide quantitative results that will assist in management practices, a more practical approach as compared to dimensionless analysis. This novel technique was tested on discrete water quality data of three pollutant types taken from seven catchments around Melbourne, Australia. A comparison with the traditional approach and a robustness assessment confirmed that the method has the potential to improve current management practices.