First Flush Influential Parameters

According to the research literature, there is no clear consensus on what rainfall parameters influence first flush. This is related to the selection of the first flush indicators and the rainfall parameters evaluated. The common practise in evaluating the influence of rainfall on first flush is by selecting one first flush indicator from the pollutant load distribution curve (MV) and evaluated with the event based rainfall parameters (Huang et al. 2012a; Li-Qing et al. 2007; Taebi and Droste 2004). As an example, Taebi and Droste (2004) investigated the correlation between first flush load at 20% of discharge volume (or known as L’(20) in this study) and rainfall- runoff characteristics such as total depth of rainfall, total duration of the rainfall event, the maximum rainfall intensity and average intensity. It was observed that the first flush load increases with the increase in rainfall intensity and duration (Taebi and Droste 2004).

A study conducted by Huang et al. (2012) observed that only long duration rainfall with largest runoff volume will result in a first flush event. The percentage of load washed-off at 30% of the runoff volume (FF30) was adopted as the first flush indicator and evaluated with total rainfall, flow rate, total runoff volume, rainfall duration, maximum rainfall intensity and antecedent dry period (Huang et al. 2012a) Li et al. (2007) also observed that there was no correlation between rainfall-runoff characteristics and first flush indicator except for maximum rainfall intensity. Similar to the study by Taebi and Droste (2004), Li et al. (2007) evaluated the

correlation of first flush with the average rainfall intensity. FF30 was defined as the

first flush in this study due to the fact that the maximum difference between the distribution curve and bisector for the events analysed was around 30% of the runoff volume. One important finding of this study was the role of rainfall pattern on the first flush phenomenon. It was observed that first flush is associated with intense rainfall at the beginning of the event (Li-Qing et al. 2007). This was based on the observation of the time interval between pollutant concentration peak and the runoff peak

However, the outcomes of past studies can be questionable as the rainfall parameters used in the analysis were event based rainfall parameters (such as average rainfall intensity, total rainfall) whilst first flush exists only at the initial part of the storm event. Therefore, the accuracy of the analysis is questionable as the rainfall parameters used do not represent the first flush indicator itself.

The role of the antecedent dry period in influencing first flush is also questionable due to inconsistent findings in past studies. For example, Li et al. (2007), Gupta and Saul (1996), Geiger (1987), and Nazahiyah et al. (2007) have found that the length of dry period influenced first flush. According to Gupta and Saul (1996), antecedent dry period was one of the most important rainfall parameters influencing the first flush load of suspended solids. A study conducted in a residential area in Johor, Malaysia found that the first flush magnitude increased with the increasing length of the dry period (Nazahiyah et al. 2007). In this study, the existence of first flush for SS, BOD5, COD, NO3-N, NO2-N, NH3-N and P were determined by plotting the

dimensionless cumulative pollutant load and dimensionless cumulative flow rate and the first flush load was determined at 20% -30% of discharge volume.

However, a contradictory result was observed from a study in Korea. The study was conducted in 13 catchments with different land uses. In order to assess the relationship between first flush and parameters that influence the phenomenon, the cumulative ratio of the pollutants load was plotted against rainfall parameters such as antecedent dry period, percentage of impervious area and catchment area (Lee et al. 2002). Antecedent dry period was found not to be correlated to first flush although rainfall intensity and catchment area was found to be correlated (Lee et al.

2002). Similarly, Saget et al., (1996) and Han et al. (2006) also found that antecedent dry period was not an influential parameter for first flush.

Although there are inconsistent findings in relation to the link between antecedent dry period and first flush, it is important to note that antecedent dry period is an important parameter for the build-up process (Lee et al. 2005). The pollutants build- up on a catchment surface during the dry period, and the pollutant load increase with increasing length of dry days before being washed-off by rainfall. This suggests that some correlation exists between the antecedent dry period, build-up process and first flush.

b. Catchment Characteristics

First flush existence and behaviour is also influenced by catchment characteristics. However, the role of catchment characteristics on first flush is still unclear due to inconsistent research findings. According to research literature, first flush is generally associated with small catchments with high impervious surfaces (Bertrand- Krajewski et al. 1998; Gupta and Saul 1996; Kim et al. 2003; Lee et al. 2005; Lee and Bang 2000; Lee et al. 2002; Ma et al. 2002; Vorreiter and Hickey 1994). According to Lee and Bang (2000) first flush phenomenon was observed in catchments smaller than 100 ha with more than 80% impervious area. This is based on the pollutant concentration peak occurring before the flow peak. However, Bertrand-Krawjeski et al. (1998) have questioned the hypothesis of smaller catchment will lead to a lower first flush coefficient thus resulting in strong first flush. They found that the lowest first flush coefficient value was not correlated with the smallest catchment.

Bertrand-Krawjeski et al. (1998) also observed that slope of the stormwater drainage system was not an influential parameters for first flush. They further suggested that the relationship between the first flush indicators and the catchment characteristics cannot be simplified and should be substantiated with experimental data. However, Lee et al. (2002) noted that a small catchment with steep slope produced a stronger first flush (Lee et al. 2002). A steep slope produces higher runoff rate, thus accelerating the process of pollutants detachment from the catchment surface.

Furthermore, the transport of pollutants to the catchment outlet is short due to the small size of the catchment.

In addition to the size and slope of the catchment, the percentage of impervious surface also plays an important role in the first flush phenomenon. According to Helsel et al. (1979), first flush increases with urbanisation which is related to the increase in impervious area of the catchment. Lee and Bang (2000) noted that a high fraction of impervious area produced stronger first flush (Lee and Bang 2000). This is attributed to the fact that runoff generated from rainfall has the ability to wash-off pollutants deposited on impervious surfaces due to high runoff velocity. The pollutant transport travel time is also reduced due to the high percentage of impervious surface and is further reduced due to the efficiency of the hydraulic conveyance system (Vorreiter and Hickey 1994).

c. Pollutant Characteristics

Pollutants characteristics are also important in evaluating the first flush occurrence and behaviour as these can be different from one catchment to another. Lee et al. (2002) revealed that suspended solids and phosphate had high magnitude in terms of the percentage wash-off load in residential area and industrial areas respectively compared to other pollutants. This was due to the surrounding activities in those areas. Huang et al. (2012) observed that the magnitude of total suspended solids was highest in the lawn’s runoff compared to the other pollutants such as nitrate, chemical oxygen demand and total phosphorus. As noted by Hall and Ellis (1985) only 60% to 80% of the storm events exhibited first flush phenomenon for suspended solids and chloride which is in contrast to Zn that experienced high wash- off load during the end part of the storm event. This suggests that first flush phenomenon is not always necessary happen for all the storm events and the occurrence varies within the pollutant.

Taebi and Droste (2004) found that total solids, total suspended solids and chemical oxygen demand displayed relatively weak first flush while no first flush effects were observed for heavy metals (leads, zinc) and nutrients (total nitrogen). Soller et al. (2005) also failed to detect the existence of first flush for metals and anions in their

study. In another study, the first flush phenomenon was more frequent for faecal coliforms, suspended solids and total phosphorus than dissolved phosphorus (Vorreiter and Hickey 1994).

This is because particulate pollutants such as suspended solids are removed from the surface primarily by physical processes such as wash-off (Griffin et al. 1980). However, the concentrations of dissolved pollutants in the runoff are governed by other factors such as solubility equilibrium and adsorption-desorption processes (Hall and Anderson 1986). This suggests that the first flush behaviour varies between pollutants, and proper consideration should be given in designing stormwater mitigation systems as it should be based on the pollutants characteristics.

2.9 GUIDING CONCEPTS IN STORMWATER TREATMENT DESIGN