The first flush effect of different urban underlying surfaces through artificial simulated rainfall

Three kinds of underlying surfaces: asphalt road, impervious tiles and grassland at Chang’an University in south second ring road of Xi’an were chosen as testing sites. Self-made device was used to moderate rainfall with different raining conditions: uniform rainfall intensity of 0.3mm/min, 0.5mm/min, 1.0mm/min; front peak type rainfall and back peak type rainfall. The pollutants in the runoff were measured, that included TN, TP, NO3--N, NO2--N, NH4+-N, PO43--P and PN M(V) curves and MFF30 values were analysed to drew these conclusions: under uniform rainfall intensity of 0.3mm/min, 0.5mm/min, 1.0mm/min, the mean values of MFF30 from asphalt road runoff were 0.47, 0.42, 0.41, the mean values of MFF30 from watertight tiles runoff were 0.40, 0.39, 0.38, the mean values of MFF30 from grassland runoff were 0.33, 0.34, 0.32. The first flush magnitude decreased with the increasing of rainfall intensity on asphalt road and watertight tiles, while the first flush intensity first increased and then decreased with the increasing of rainfall intensity on grassland. The mean values of MFF30 from the front peak type rainfall were larger than that from the back peak type rainfall on both asphalt road and watertight tile. The mean values of MFF30 from the back peak type rainfall were larger than that from the front peak type rainfall on grass. Grassland should be built around asphalt road and watertight tiles to attenuate first flush effect.


Introduction
Rainfall runoff through the leaching, erosion, coffin to carry pollutants and it means the main driving force of various pollutants entering the water on the surface sink surface [1]. With the continuous development of urbanization, impervious area of impervious surface gradually increased, making the urban runoff pollution more serious and increasing the city surrounding rivers and lakes sewage discharge pressures [2]. Therefore, the urban surface runoff pollution has become the main form of urban non-point source pollution, and it is the second largest urban water pollution besides domestic sewage and industrial wastewater and gradually deteriorating urban water environmental quality [3][4][5]. It is found that the concentration of pollutants in the early stage of rainfall runoff is significantly higher than the later stage, and a large amount of pollutants are discharged into the receiving water in a short period of time, causing serious damage to the water environment. This is the FFE (first flush effect), that is, the initial runoff is disproportionately carrying the most of the pollutants in whole runoff [6][7][8][9]. It has important economic significance and environmental significance for the management and governance of urban storm runoff [10]. Many domestic and foreign researchers have studied the FFE from various aspects. And the results show that the main factors are rainfall characteristics (rainfall, rainfall duration, maximum rainfall intensity), number of dry days, the type of land use, etc. [11][12][13][14][15][16]. At present, the research of the FFE is based on natural rainfall, and the monitoring of the FFE under natural rainfall is complex, random and extensive. It has high monitoring time, long cycle and large consumption of human and material resources. So the data obtained is less effective, and the results are also similarities and differences [17] . Such as Li Chunli found that the number of dry days and the degree of FFE has a positive correlation [15], but J.H. Lee found that there is no correlation between the two [18].
The study used self-made artificial rainfall devices to overcome the unfavorable factors of natural rainfall, and to discuss the different types of rainfall on the FFE of common urban surfaces. So as to provide the basis for the evaluation and command of urban rainfall runoff pollution.

Study Sites
The study area is Chang'an University campus, which is located in south second ring road of Xi'an. The proportion of land use types in Chang'an University campus is Construction land (66.2%), road (19.9%), Greenland (11.7%), and Watertight tiles (2.2%). Three kinds of typical underlying surfaces, asphalt road, grassland and watertight tiles were selected for artificial rainfall study.

Experimental device.
We used the self-made experimental device to study runoff pollution, and the device consists of simulated rainfall and runoff collection. Simulated rainfall device consists of water pump, motor, flow meter, pressure gauge (0-0.6MPa), detachable bracket (3m high, 2m wide), and control box. Nozzle has a large (aperture 4mm), medium (aperture 2.5mm), small (diameter 2mm), and each nozzle is controlled by a separate solenoid valve. The device has a high similarity to natural rainfall in terms of rainfall characteristics. Before the rain, the runoff collection device is placed on the appropriate surface and sealed between the baffle and the underlying surface. So that all the runoff within the baffle from the collection port. The complete set of experimental device diagram shown in Figure 1.  [20], and the most commonly used is the cumulative pollutant-runoff curve [21,22], that is according the ratio of cumulative runoff to total runoff (cumulative runoff ratio), the ratio of cumulative pollution load to total rainfall load (cumulative pollution load ratio) and create a dimensionless curve -M(V) curve. According to the curve deviation degree of diagonal above to determine the strength of the FFE. The more the curve deviates from the diagonal, the stronger the FFE. FFE analysis of M(V) curve is more intuitive, but the lack of quantitative expression, which is not conducive to the application of non-point source pollution assessment and control. Saget conducted a quantitative analysis of the FFE [23], which using a power function to fit the pollutant M(V) curve: F (X) = X b. . X is the ratio of the cumulative runoff to the total runoff, F(X) is the ratio of the cumulative pollution load to the total rainfall load, b is the FFE coefficient. The b value can be determined by software. When X is 30%, the corresponding pollutant accumulation is called MFF30, and the calculated is:  Greenland underlying surfaces, which under three uniform rain intensity, such as 0.3mm/min, 0.5mm/min, 1.0mm/min.

Asphalt road FFE.
It can be seen from Figure 2 that NH4 + -N, NO3 --N and PO4 3--P in the asphalt road occurred in the FFE when rainfall is 0.3mm/min, in which the M(V) curve of NH4 + -N deviates from the diagonal, and the remaining pollutants has not obvious flush effect. When the rainfall intensity increased to 0.5mm/min, the M(V) curve of NH4 + -N and NO3 --N is more obvious deviates from the diagonal, and the M(V) curve of PN is significantly increased in the later stage of rainfall, indicating that the flushing has occurred in the later stage. In addition, PO4 3--P flush effect was significantly reduced, the remaining pollutants without significant flush amplitude changes. When the rainfall intensity increased to 1.0mm/min, the flush effect of NH4 + -N decreased, and TP appears more obvious flush effect.
In summary, NH4 + -N and NO3 --N on asphalt road are more prone to FFE, and PN is less prone to FFE in the small rainfall intensity. When the rainfall intensity increases, the FFE gradually appears in the late rain, and when the rain continues to increase, the FFE has weakened. Rainfall has washed, leaching, dilution and other multiple effects for pollutants on the underlying surfaces. The stronger the rain, the greater the erosion effect, so it can carry more pollutants, which may be the cause of the PN FFE increased with rain increased. At the same time, when the rainfall intensity increases, the runoff on the pollutant dilution effect also will be strengthened and the pollutant concentration decreases, it will weaken the pollutant flush effect [24]. In addition, PN has two clear flush in the rain which the M(V) curve appears more substantial rise in the late-rain. Asphalt road pavement has strong adhesion to particulate contaminants because of the certain sticky, and particles are not easy to run with the flow of flush. Since then,, the soaking and flush by the early runoff, the particulate gradually released in the late rain, which it may be caused by the above reasons.

Watertight tiles FFE.
It can be seen from Figure 2 that NO2 --N, NH4 + -N and NO3 --N in the watertight tiles have a certain degree of flush effect when rainfall is 0.3mm/min, and the remaining pollutants has not obvious flush effect. When the rainfall intensity increased to 0.5mm/min, the flush degree of NO2 --N also increased, but flush effect of NH4 + -N and NO3 --N were weakened. When the rainfall intensity is 1.0mm/min, all of the pollutants showed a weak flush effect. In the study area, the pollution source mainly comes from the sedimentation of pollutants in the atmosphere and the diffusion and transfer of pollutants on the road. Compared with the road, the accumulation of contaminants on the watertight tiles is less, and the dilution effect of the runoff on the pollutants is more obvious. So the bigger the rain is, the harder it is to observe the FFE. In addition, asphalt road water permeability is stronger, which compared with watertight tile. Therefore, it may be the one of the reasons why the FFE of the watertight tiles is weaker than the asphalt road.

Greenland FFE.
It can be seen from Figure 2 that NH4 + -N and NO2 --N in the grassland have a certain degree of flush effect when rainfall is 0.3mm/min. When the rainfall intensity increased to 0.5mm/min, the flush degree of NH4 + -N and NO2 --N also increased slightly, and PN showing late flush effect that the M(V) curve increased at the later stage. When the rainfall intensity is 1.0mm/min, the M(V) curves of the pollutants are close to the diagonal, indicating that no significant flush effect.
Grassland is one of the few in the city and occupies a certain percentage of the underlying surface, so we understand that the FFE on grasslands is important for estimating and controlling urban runoff pollution. Grassland permeability is extremely strong and particle retention is very effective, which is different from poor permeability surface of asphalt road and watertight tiles. The results show that the concentration of runoff pollutants in grassland is always lower than other cities underlying surface, and the flush effect of grassland runoff is small, which has a certain interception effect on pollutants in runoff [25]. Compared with the other two underlying surfaces, the flush degree of the grassland is low, which is the same as the above research results.

M(V) Curve of in unimodal rainfall
In order to be closer to the actual rainfall intensity, we use the Chicago rain-type to set the front and back peaks of two single peak rainfall situation. The Chicago rain-type was a kind of uneven design rain that based on intensity-duration-frequency relationship in 1957. After a lot of simulation and comparison, most domestic and foreign scholars believe that the effect is better and easier to determine the rainfall intensity process, and it is widely used in many research fields at home and abroad [26]. In this study, the uniform rainfall intensity changes every 5mins, and the specific changes is shown in Table 1 and 2.
The artificial rainfall is carried out on each of the underlying surfaces which according to two types in the table, and the resulting of pollutant M(V) curve is shown in Figure 3. Intuitive analysis shows that the contaminants of two peak-type rainfall have occurred in varying degrees of FFE besides NO2 --N in the asphalt road, and the FFE of the front peak type rainfall is significantly higher back peak type rainfall. The FFE of the front peak type rainfall on watertight tiles is higher than the back peak type, and the output of TP fluctuates obviously in the back peak rainfall. TP on grassland in front type rains showed significant FFE, and TN and NO3 --N did not occur FFE. While in the back peak rainfall, only the FFE occurred in TP, and TN and NH4 + -N did not occur FFE, and the output rate of the remaining pollutants was similar to the runoff rate. In general, front peak rainfall is more likely to cause FFE. Therefore, it can be judged that the rain intensity distribution, especially the maximum rain intensity, is the main factor affecting the FFE.  It can be seen from Figure 4 that most of the rainfall pollutants MFF30 are between 0.3 and 0.6. Compared with the underlying surfaces, the FFE of the grassland was weak, and some scenes of rain pollutants MFF30 value was less than 0.3. The relation between the strength of FFE on different underlying surfaces is asphalt road> watertight tiles> grassland. However, for the NO2 --N, the MFF30 value on the watertight tiles is significantly larger than the MFF30 value on the asphalt road. It can be seen that we should give priority to controlling NO2 --N pollution of the watertight tiles. It is found that the MFF30 values of NH4 + -N and TP are the highest in each pollutant, indicating that the two pollutants are prone to FFE. In addition, although TP and PO4 3--P similar in origin, but the latter FFE to be significantly weaker than TP. The MFF30 value of TN and NO3 --N on the asphalt road increased slightly with the increase of rainfall intensity, indicating that the greater the rainfall intensity, the more obvious the flush effect of the two pollutants. And the MFF30 of NH4 + -N, TP and PO4 3--P decreases with the increase of rainfall intensity, which indicates that these pollutants are not easily washed by runoff, and the runoff dilution effect is stronger. The MFF30 of NO2 --N and TP decreases with the increase of rainfall intensity, but PN increases slightly, and the variation of other pollutants is not obvious. Finally, there was no significant change in the remaining pollutants except that the MFF30 of the grassland increased with the increase of the rainfall intensity.   Table 3 shows the MFF30 average values of contaminants for different rainfall intensity on each underlying surface. In the table, the FFE of asphalt road and watertight tiles decreases with the increase of rainfall intensity, indicating that the dilution of pollutants on the two underlying surfaces is more pronounced. While the FFE of the grassland increased first and then decreased with the increase of rain intensity, which indicated that the grassland was less prone to dilution.  Figure 5, and the average values of the pollutants MFF30 is shown in Table 4. In conjunction with Figure and Table, it can be concluded that the top 30% of the runoff in the pre-peak rainfall is significantly higher than back-type rainfall that on the surface of the asphalt road. Similarly, in addition to NO2 --N, the contaminants MMF30 values of pre-peak rainfall were higher than those the post-peak in the watertight tiles. The reason is that the runoff is often formed at the beginning in the impervious surface, and can carry a large number of pollutants. The results showed that the post-peak rainfall MMF30 value of TN, NO3 --N and TP was higher than the former peak on the grassland, indicating that the FFE of the post-peak rainfall on the grassland was stronger. Because the permeability of the grassland is strong, and the higher descent rate of early rainfall is difficult to form a large number of runoff, and the flush of the pollutants is also less. Therefore, for the impervious surface, we should pay attention to controlling the early pollutants when the pre-peak rainfall. While for the permeable surface, we should focus on pollutant control that late post-peak rainfall.  4.2 Intuitive analysis of M(V) curve, all the underlying surface of the FFE is more obvious, when the rainfall is maintained at 0.5mm/min. With the MFF30 value analysis, the asphalt road and watertight tiles, all show the rain increased and the FFE intensity decreased.
4.3 In the unimodal rainfall, the MFF30 value of each pollutant is generally lower than the MFF30 value of uniform rainfall. The FFE of back peak type rainfall of asphalt road and watertight tiles is weaker than that of the former peak type rainfall, and the results on the grassland are the opposite. Asphalt roads and watertight tiles on MFF30 of the pollutants is generally higher than the grassland. So grassland should be built around asphalt road and watertight tiles to attenuate FFE.