First Flush Definition

A number of different definitions of first flush have been proposed by past researchers. According to research literature, first flush definitions could be categorised based on three approaches;

 Concentration-based first flush;

 Mass-based first flush; and

 Empirically based first flush.

a. Concentration-based First Flush (CBFF)

Concentration-based first flush (CBFF) is defined as a high concentration of pollutants at the early stage of the runoff event compared to the later stage (Sansalone and Cristina 2004; Thornton and Saul 1986). This definition is commonly used by researchers to define first flush, as it is easy to apply. First flush can be identified based on the concentration peak at the initial part of a pollutograph followed by a rapid decline in concentration after the peak and a relatively low and constant concentration for the remainder of the rainfall duration. Figure 2.8 illustrates the first flush based on concentration, for COD and SS. As evident in Figure 2.8, both COD and SS have high concentration at the early part of the pollutograph and the pollutant peak for both pollutants occurs followed by the runoff peak.

Figure 2.8: Illustration of first flush definition based on concentration (adapted from Thornton and Saul (1987))

Definition of first flush based on concentration has been applied in various studies (Barrett et al. 1998; Cordery 1977; Forster 1999; Larsen et al. 1998; Lee and Bang 2000). According to the study undertaken by Lee and Bang (2000), the concentration of suspended solids, n-Hexane and chemical oxygen demand rose significantly as the runoff increased and the pollutants peak concentration occurred before the runoff peak. In an investigation into water quality in three catchments in Sydney, Australia, Cordery (1977) also found that the concentration of pollutants such as suspended solids and nutrients were highest at the beginning of a storm event and reduced rapidly with the runoff.

It is important to note that the concentration peak varies between pollutants, events and method of measurement (Sansalone et al. 2006). Therefore, it is hard to determine specific parameters for treatment design. Furthermore, the ‘early’ or ‘initial’ part of runoff or rainfall referred to by the researchers is not clear as it has not been defined precisely (Bertrand-Krajewski et al. 1998; Saget et al. 1996).

b. Mass-based first flush (MBFF)

Mass-based first flush (MBFF) is also a commonly used definition used by researchers for defining the first flush phenomenon. MBFF is defined as the occurrence of a high mass during the rising limb or the early portion of the runoff hydrograph (Sansalone and Cristina 2004). Although CBFF is widely used, technically MBFF can be more acceptable. This is because the pollutant load distribution curve (MV) from multiple rainfall events or from different catchments can be compared by plotting cumulative pollutant load (M) versus cumulative runoff volume (V) graph (Batroney 2007; Tucker 2007). The MV curve represents the variation of the pollutant load removed from catchment surface in relation to the runoff volume which has left the catchment. M and V variables at a specific time can be calculated as follows:

where:

cumulative runoff volume at time t;

[2.3]

time between the initiation of runoff (t0);

time coinciding with the cessation of runoff; runoff flow rate at time t;

cumulative mass at time t; and pollutant concentration at time t.

M(t) in Equation 2.4 refers to the investigated pollutant. According to research literature, there are variations in first flush definitions provided by past researchers based on the MBFF method. According to Helsel et al. (1979), first flush exists if the cumulative percentage of load is above the cumulative percentage of flow at any time. In this definition, a cumulative percentage of load and flow against time is plotted as illustrated in Figure 2.9.

Figure 2.9: Variation of incremental load and flow with incremental time (adapted from Helsel et al. (1979))

However, Geiger (1987) postulated that first flush exists if the initial slope of the MV line is greater than 45º as shown in Figure 2.10. The cumulative percentage of pollutant mass vs. cumulative percentage of flow is plotted to define first flush. The resulting MV curve is compared with a 45º line or bisector line (referred to as B line). The B line indicates that the pollutants load removed from the catchment is directly proportional to the discharge volume leaving the catchment. It is assumed that dilution occurs when the initial slope of M is below the B line. The first flush is considered significant if the gap between the MV curve and the B line is greater than 20% (Geiger 1987). However, a gap greater than 20% between MV curve and B line could appear at any runoff volume throughout the rainfall event and there would be multiple first flush occurrences if this definition is applied which will affect the estimation of runoff volume to be treated in stormwater treatment design.

Figure 2.10: Definition of first flush by Geiger (adapted from Geiger (1987))

MV

45° line or B line

An alternative to the stipulation by Geiger (1987) has been provided by Gupta and Saul (1996), that first flush exists at the maximum divergence of the MV curve and the B line. In this approach, both cumulative percentage of suspended solids and cumulative percentage of flow are plotted against the cumulative percentage of time (refer Figure 2.11). Using this definition, the volume of runoff to be treated can be estimated as the maximum gap between cumulative percentage of pollutant and cumulative percentage of flow (Gupta and Saul 1996). Therefore, the timing of first flush occurrence can be estimated. The application of this definition can also be applied to other pollutants although most past studies adopted suspended solids as the surrogate parameter to assess first flush occurrence (Gupta and Saul 1996; Taebi and Droste 2004). This is due to the fact that other pollutants such heavy metals, hydrocarbons and nutrients are attached to suspended solids.

Figure 2.11: Definition of first flush by Gupta and Saul (adapted from Gupta and Saul (1996))

The deviation between the MV curve and B line indicates the magnitude of first flush. A relatively big gap between the MV curve and B line suggests a stronger or higher magnitude first flush. M and V can be mathematically represented by a power function as shown in Equation 2.5 (Bertrand-Krajewski et al. 1998; Lee et al. 2002; Sansalone and Cristina 2004).

[2.5]

is the first flush coefficient and the value can be computed by linear regression as follows:

[2.6]

Therefore, based on the ‘a’ value, first flush existence can be assessed. In general, if ‘a’ value is <1, it indicates that the first flush phenomenon exists while if ‘a’ is >1, then the first flush is absent (Saget et al. 1996, Bertrand-Krajewski et al. 1998). Saget et al. (1996) identified six regions which can be used to define the strength of the first flush. Table 2.3 lists the range of first flush coefficients for the six regions and Figure 2.12 illustrates these six regions.

Table 2.3: The strength of the first flush based on first flush coefficient ‘a’ (adapted from Saget et al. 1996)

Area Values of ‘a’

coefficient Deviation from diagonal

1 2 3 4 5 6 0 < a ≤ 0.185 0.185 < a ≤0.862 0.862 < a ≤1.00 1.00 < a ≤ 1.159 1.159 < a ≤ 5.395 5.365 < a ≤ infinity Positive negative

Strong mediation above the diagonal Moderate deviation above the diagonal Little deviation above the diagonal Little deviation below the diagonal Moderate deviation below the diagonal Strong deviation belowe the diagonal

Figure 2.12: First flush strength based on the first flush coefficient ‘a’ (adapted from Saget et al. (1996))

Another first flush definition based on MBFF is the comparison of a fraction of the total pollutant load to a fraction of the runoff volume (Delectic 1998). First flush is determined by simultaneously selecting the percentage of discharge volume and pollutant load at the early part of the distribution curve (MV curve).

There are numerous first flush definitions based on pollutant load and volume fractions. For example, Deletic (1998) proposed first flush as the percentage of total event pollution load transported by the first 20% of total stormwater runoff volume. Deletic (1998) proposed that for a significant first flush to have occurred, 40% of the pollutant load must be transported in the first 20% of the runoff volume (which is known as 20/40 FF). Similarly, Vorreiter and Hickey (1994) proposed that 40% - 60% of pollutants should be transported by 25% of the runoff volume. This was based on their analyses of SS, faecal coliforms and TP from six catchments in Sydney, Australia.

First flush based on fractions can be arbitrary. A researcher could choose any value without providing any substantial justification. For example, Bertrand-Krajewski et al. (1998) proposed a stringent definition for first flush known as 30/80 first flush, where at least 80% of the pollutant mass needs to be transported in the first 30% of the runoff volume. Another stringent first flush definition states that 80% of the total pollutant load must be transported by the first 20% of the flow volume (Stahre and Urbonas 1990). However, researchers have found that it is very difficult to meet the definitions proposed by Bertrand-Krajewski et al. (1998) and Stahre and Urbonas (1990) as these are too stringent (Lau et al. 2002; Lee et al. 2002; Sansalone and Cristina 2004).

c. Empirically based First Flush

The application of first flush as a design parameter is based on the water quantity. This parameter defines the quantity of the rainfall depth or runoff volume to be treated by the stormwater treatment device. The typical design parameters that have been adopted are:

 First half inch (13mm) of rainfall; and

(Deng et al. 2005)

For example, the first half inch (13mm) of rainfall means that the first half inch of rainfall is captured to be treated by the stormwater treatment devices and the rest of the rainfall depth is released directly to the receiving water. However, there is very limited empirical evidence provided to justify the adoption of a specific rainfall depth. Secondly, it is also debatable whether a standard rainfall depth can be adopted across different geographical and climatic regions.

2.8.3 First Flush Influential Parameters