Investigation of first flushes in a medium-sized mediterranean catchment

Journal of Hydrology 373 (2009) 405–415

Contents lists available at ScienceDirect

Journal of Hydrology journal homepage: www.elsevier.com/locate/jhydrol

Investigation of first flushes in a medium-sized mediterranean catchment Matthias Obermann a,*, Karl-Heinz Rosenwinkel b, Marie-George Tournoud c a

Dorsch Consult Wasser und Umwelt GmbH, Hansastr. 20, 80686 München, Germany Institute for Water Quality and Waste Management (ISAH), Leibniz Universität Hannover, Welfengarten 1, 30167 Hannover, Germany Hydrosciences Montpellier, JRU CNRS – IRD – Université Montpellier 1 – Université Montpellier 2, Maison des Sciences de l’Eau, Université Montpellier 2, 34095 Montpellier cedex 5, France b c

a r t i c l e

i n f o

Article history: Received 29 May 2008 Received in revised form 9 April 2009 Accepted 30 April 2009

This manuscript was handled by L. Charlet, Editor-in-Chief, with the assistance of Chong-yu Xu, Associate Editor Keywords: First flush Semi-arid river basins Dry weather accumulation Water quality

s u m m a r y The objective of this study is to investigate the significance and characteristic of first flushes in mediumsized agricultural catchments. Therefore, an analysis of the load distribution in the Vène River on the French mediterranean coast was done for first flood events in September 2003 and 2004. It considered total suspended solids (TSS), volatile suspended solids (VSS), total phosphorus (TP), soluble reactive phosphorus (SRP), total nitrogen (TN), organic nitrogen (org-N), nitrate and nitrite–nitrogen (NOx–N) and ammonium–nitrogen (NH4–N). Nutrient export was evaluated in terms of normalized cumulative loads (NCL) and three rating indices. Most important first flushes could be detected for ammonium (FF25 = 0.79) followed by TSS (FF25 = 0.72) and VSS (FF25 = 0.70). In situations, where first flushes can be important for downstream water bodies, a new concept for nutrient discharges should be considered taking into account the ambient pollution of the whole system. Ó 2009 Elsevier B.V. All rights reserved.

Introduction The hydrological response of catchments in arid or semiarid regions is characterized by features not seen in perennial flow. Perennially flowing rivers in humid areas typically receive water from the surrounding landscape with sufficient regularity, but rivers in semi-arid regions may receive little for longer periods of months or even years. On the contrary, they may lose a significant amount of water through the river bed or by evaporation. In some cases, these transmission losses may lead to complete drying up of the river and in low flow situation, the river may be even diverted or reduced to a sequence of separated pools. The situation is normally further intensified by anthropogenic influences as, e.g. abstractions, point sources, agricultural practices as, e.g. drainage or constructions in the stream. These features require particular attention in the analysis of nutrient and pollutant transport regime. Intensive fieldwork was carried out by HydroSciences Montpellier in the catchment of the Vène River at the French mediterranean coast during the European project tempQsim (EVK1-CT2002-00112). Seasonal load variations have already been studied in the catchment (Obermann et al., 2007). This paper will focus on load variations during flow events.

* Corresponding author. Tel.: +49 177 2767974. E-mail address: [email protected] (M. Obermann). 0022-1694/$ - see front matter Ó 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.jhydrol.2009.04.038

Historic approaches for the characterization of first flushes – which were in the past used for small catchments – are briefly reviewed, tested with artificial data and applied to data of the larger Vène catchment. First flush definitions, relevancy of concentration and load A ‘‘first flush” is normally defined as a disproportionate increase of particulate or dissolved materials in terms of concentration or load in the rising limb of a runoff event. Sansalone and Cristina (2004) specified the definition by introducing the terms ‘‘concentration-based first flush” (CBFF) and ‘‘mass-based first flush” (MBFF). In the context of dryland rivers it has to be noted that concentration based first flushes during low flow conditions often lack relevance for the overall mass transport. Nevertheless, concentration is important if the water quality evolution of the river is in the focus. If in contrast, the impact of a dryland river on a receiving standing water body (e.g. lagoon, lake or reservoir) is studied, the analysis of the mass-based first flush and loads should be favoured in order to avoid misinterpretation. As shown in Krebs et al. (1999), there are also effects which may cause an initial peak in loads for dissolved compounds, due to the fact that a flood wave travels faster than the water flow. So, from a theoretical point of view, peaks in concentration are only possible (i) for particulate constituents, which were stored in respective on the bed or (ii) for dissolved constituents, if the flushing rainwater

406

M. Obermann et al. / Journal of Hydrology 373 (2009) 405–415

picked up enough of the constituent during its overland passage, so that the concentration is higher than it was in the river. This means that a flush based on concentration is normally more likely for particulate materials than for solubles. Importance of first flushes for temporary mediterranean rivers First flushes can be of great importance for the nutrient and pollutant transport in semi-arid or arid regions. In low flow situation, the flow regime of temporary rivers is often dominated by the interaction of discharges from point sources and by transmission losses. A significant accumulation of organic matter during the summer months may, e.g. occur due to the effluents of wastewater treatment plants. Accumulation is favoured during these conditions, when precipitation or other sources are insufficient to produce uninterrupted discharge on the whole length of the river. This often happens during late spring and summer, when the river can progressively dry up until only separated pools remain. If these pools are neither interconnected nor any one is connected to downstream water bodies, there will be an excessive supply of nutrients and therefore a potential for significant growth of biomass in these sites. Hence, there is not only a temporal variation in flow, but of course also a spatial one. So even during a rain event, the flow may be discontinuous, e.g. when rainfall is highly variable in space. Occurrence of first flushes First flushes have been studied before, mostly in smaller urban catchments. The phenomenon was subject to different definitions and there have been many attempts to find a general characterization. Bertrand-Krajewski et al. (1998) analysed 197 rainfall events in 12 separate and combined sewer systems, but no general multi regression relation between first flush behaviour, site and rainfall characteristics as well as type of the sewer system could be derived. An analysis of 197 events in 14 basins in Saget et al. (1996) suggested that there is no significant relationship between the first flush parameter and basin characteristics. It was also concluded that the occurrence of a first flush is very scarce, although it must be stated that the applied criteria for identification is very strict. On the other hand, Gupta and Saul (1996a,b) succeeded in developing site specific regressional relationships in combined sewers for suspended solids with the goal to design storage tanks. For the two English catchments, maximum rainfall intensity, maximum inflow, rainfall duration and the antecedent dry weather period were identified as the most important parameters. A comparison of nine Korean urban watersheds (Lee and Bang, 2000) found that first flushes are higher with increasing rainfall intensity and percentage of paved area and decreasing watershed area. Characklis and Wiesner (1997) showed for urban runoff that elevated contaminant loadings can reach magnitudes equating to weeks or months of background flow. It was suggested that the load variation of during a storm event is related to the particle size distribution. However, the used data showed no evidence of a first flush in the investigated catchment. The authors argued that the catchment is a comparably large urban catchment of about 240 km2, which consists of many different smaller subcatchments. These might have first flushes with varying travel times to the point where measurements have been made. This superposition makes the first flush nearly invisible at this point. They therefore proposed a treatment nearer to source areas rather than for the whole basin. Cristina and Sansalone (2003) investigated suspended particulates in urban storm water runoff. It was found that there can occur a disproportional high suspended solids delivery following the

main water flow. It was suggested to treat the whole event and not only the first part, which is related to an eventual first flush. A similar result emerged from Sansalone and Cristina (2004), who found limitations in the applicability of the water quality volume (WQV) approach in the design of treatment facilities for small paved catchments. In this context, the differences between the first significant floods and following events with lower pollution seem to be important. Obermann et al. (2007) showed in a semi-arid French catchment that over 2/3 of the annual total suspended solids load can be caused by the first flood event, which strongly influences the annual fluctuation of mass transport. Deletic (1998) investigated data for one year of two small urban asphalt catchments and found slight first flushes for particulates and conductivity, but none for temperature and pH. Lee et al. (2003) analysed the fitting between normalized cumulative load and normalized cumulative flow with third polynomial equations on the basis of two storm events in a mixed urban catchment. It was concluded that the mass loading rate declines from residential watersheds (from high density to low density) over industrial ones to undeveloped urban watersheds (Lee et al., 2004). So the nature and occurrence of first flushes is very divers. In the rich history of first flush analysis, there even is a lot of dissent what a first flush is and if it actually exists. Nevertheless this study will show that the phenomenon is an important driver of nutrient and pollutant dynamics in temporary rivers. Below, approaches to visualize and identify first flushes for the Vène River on the French mediterranean coast will be reviewed: as first type the method of normalized cumulative load plots, as second type examples for three index classifications. These methods will then be applied to artificial examples which stand for typical situations. This will help in the later interpretation of results, when first flushes are identified for two flood events of the years 2003 and 2004 in the Vène River. Methods Calculation of normalized cumulative load plots In many cases, it is beneficial for downstream water bodies to capture or divert first flushes because they often carry a high load with a low amount of water. Finding the best management is then an optimisation between (i) minimising the amount of water which has to be stored or diverted and (ii) maximising the mass of nutrients or pollutants not entering downstream water bodies. Hence it is often important to know, when the transport of pollutants is most intense in comparison to the water flow, so that most of the constituents can be stored with the least effort. One common method to identify phases of increased nutrient fluxes are diagrams of normalized cumulative loads over normalized cumulative flow. The normalized cumulative flow (NCF) and the normalized cumulative load (NCL) of a parameter X can be calculated from

Rt NCF ¼

NCL ¼

t¼t0

R te

t¼t0 Rt t¼t R te 0 t¼t 0

QðtÞ dt QðtÞ dt XðtÞ  QðtÞ dt XðtÞ  QðtÞ dt

ð1aÞ

ð1bÞ

where X(t) is the parameter at a time t, water quality characteristic such as concentration [ML3], conductivity or temperature; Q(t) is the flow rate of water [L3T1] and t0, te are the defined beginning respective end of the discharge event or the regarded period (hydrological year) [T]. When the NCL is plotted against NCF, information of the temporal distribution of loadings over the event duration can be derived

407

M. Obermann et al. / Journal of Hydrology 373 (2009) 405–415

the strength of a first flush with a single value in order to compare different scenarios, catchments or events and to present strict criteria. These can be based on, e.g. the basis of differences between normalized loadings and normalized flow (Geiger, 1984), fitted exponential parameters of NCL-plots (Bertrand-Krajewski et al., 1998; Saget et al., 1996), initial slopes of NCL-curves (Bedient et al., 1978) and approximations of the percentage of mass transported at a specific percentage of flow (Saget et al., 1996; Vorreiter and Hickey, 1994). Another method would be to calculate the integral of the NCL over the event duration. In this case, an NCL-curve equal to the bisector would yield a value of 0.5, a greater value would indicate a first flush and a value below 0.5 would be a sign for a delayed reaction. Other researchers defined the significance of a first flush in terms of the load which was transported by the first x% of flow, e.g. Deletic (1998) for the first 25% (FF25) or Saget et al. (1996) for the first 30% (FF30). The maximum in these definitions would be 1 (or 100%). This could be called a ‘‘perfect first flush” as equivalent

Table 1 Boundaries of classification approaches for first flushes. Classification method

Lower limit

Centre

Upper limit

FF25 Geiger (1984) NCL-integral

0 0 (1) 0

0.25 0 0.5

1 1 1

more easily. These types of curves derived from the NCL enable a dimensionless classification of the pollutograph in terms of the temporal distribution of loadings over the event duration. Indices as classification methods for first flushes Investigation of the relationship of flow rate and pollutant transport with the help of NCL-curves is an appropriate measure as outlined above. Nevertheless it is often helpful to characterize

(a) constant concentration

(b) constant load

(c) storage depleted conc load flow

time normalized cum. loa d

1

0.5

0 0

0.5

1

0

0.5

1

0

0.5

1

normalized cum. flow (d) delay in load

(e) delay and depletion

(f) two differing storages

conc load flow

time

normalized cum. loa d

1

0.5

0 0

0.5

1

0

0.5

1

0

normalized cum. flow Fig. 1. Sample scenarios for the demonstration of rating indices and NCL-plots.

0.5

1

408

M. Obermann et al. / Journal of Hydrology 373 (2009) 405–415

to, e.g. a perfect elastic collision. It means that 100% of the regarded constituent is transported with a finite amount of water at the very beginning of the event. As it is a limiting value, it can never be reached. The minimum would be zero for no first flush. Geiger (1984) defined a criteria in terms of the maximum difference of the NCL-curve from the bisector. The highest possible value of one would again describe a perfect first flush; zero would be the lowest (cf. Table 1). A summary of the boundaries of the presented approaches is given in Table 1. Application of the methods to artificial scenarios In order to demonstrate the boundaries and the behaviour of the applied methods, they have been applied to the six sample scenarios shown in Fig. 1: (a) Constant concentration: The concentration is set constant in this example, but due to the flow rate increase the load also increases; the NCL-plot is equal to the bisector. (b) Constant load: The load is set constant during this event, a strong dilution effect for the concentration can be observed; the NCL-plot is close to the bisector. (c) Storage depleted: The concentration increases with the flow, but the storage of the parameter is exhausted before the end of the regarded event. The NCL curve is very steep and all of the total mass is transported before the end of the event, so there is a strong first flush. This is an example of a case with a limited storage, which is often also called ‘‘mass limited first flush”. (d) Delay in load: There is hardly transport in the rising limb and in the peak of the flood. The NCL-curve is similar to (c), but the transport occurs at the end of the event, so it is not a first flush. This is true, if a storage needs time to become disposable for transport. This is the case when a crusted soil has to be broken up, if a covering layer has to be removed by erosion before or if this constituent emerges from a subcatchment with a longer transfer time. (e) Delay and depletion: The beginning of the transport is delayed and the storage limited, this is a combination of (c) and (d).

(f) Two differing storages: There are two different storages of the pollutant, the first is directly available for transport, but the other one can be only transported after some time. First, the NCL-curve increases. It stagnates, during the time when no transport is happening. Then it increases again, when the second storage is resuspended. This also is a combination of (c) and (d). Fig. 2 shows an application of the three methods (NCL, FF25, Geiger) to the above mentioned scenarios. Fig. 2 shows that in all approaches, scenario (c) yields the highest rating, which would have been expected by the study of the NCL-curves. Detailed interpretation of the figure is given in the following paragraphs. FF25-criteria The FF25 criterion shows a good behaviour for (f). From a management point of view the first part of the flood could be treated separately, so this method is able to highlight this important differentiation from the other variants. The maximum of one at the case (c) does not include information about the situation before 25% of runoff, so a first flush whose total load is even more condensed would have the same value and a distinction would not be possible. Geiger’s (1984) criteria As FF25, Geiger also yields a slightly higher value for (b) than for (a). In contrast to the FF25-method, (e) and (f) give nearly equal values. A rating index for first flushes should be able to distinguish between these two cases. The criteria is usually limited from 0 (load is proportional to the flow) to 1 (perfect first flush). In order to allow additional distinctions, the method could be extended to the range from 1 to 1, if negative maxima were allowed, like in case (d). NCL-integral criteria The integral of the NCL-plot does only allow a gradation of the cases (c) and (d); the others are all in the same magnitude. Although it is often implicitly included, the aforementioned definitions are all lacking a relation to the relative time after which a peak in the pollutograph appears. There is no coupling between the intensity of the first flush and the time of its appearance.

1.00 1.00

NCL from 0 to 1

0.87 FF25

0.78 Geiger

0.51

0.50

0.51

0.49

0.50

0.52

0.33

0.30

0.29

0.25 0.25

0.11

0.09

(f) two different storages

(e) delay and depletion

(d) delay in load

(c) storage depleted

(b) constant load

0.00

(a) constant concentration

rating [-]

0.75

Fig. 2. Comparison of indices methods to rate the distinctness of first flushes based on scenarios a-f.

409

M. Obermann et al. / Journal of Hydrology 373 (2009) 405–415

Applicability of the indices Especially the method FF25 proved to be especially valuable. It shows the best performance in identifying first flushes. The integral approach and Geiger (1984) showed some limitations. The FF25 is a helpful approach to easily compare different floods and to derive trends in nutrient fluxes from one event to another. Field data and results River characteristics The above methods have been applied to the catchment of the Vène River at the French mediterranean coast (Fig. 3). With a size of 67 km2 the catchment is sparsely populated (12,400 inhabitants, only 3% of the total basin area is urban), and with 34% agricultural zones of which 21% are vineyards this is the main active land use, natural areas represent more than 60%. Details of the site and instrumentation set up are described in Tournoud et al. (2006). Flow was calculated from measurements of automatic samplers of flow depth over predefined rating curves with a time step of 5 min. All water samples were taken in the middle of the cross section and were immediately filtered, preserved and analysed within less than 4 h. The samples were analysed for total suspended solids (TSS), total phosphorus (TP), soluble reactive phosphorus (SRP), ammonium–nitrogen (NH4–N), nitrate and nitrite–nitrogen (NOx–N) and total Kjeldahl nitrogen (TKN) following the standard methods requirements (APHA et al., 1992).

of surface runoff and wash off effects will prevail. The second part of the flood directly follows. This has a duration of about 6 h with a maximum flow of 8.3 m3/s and a total runoff volume of about 142,000 m3. In the later phase the karstic spring at site K dominates the discharge. This spring is supplied by rainfall from outside the catchment (to the north) and it did not flow in the 2004-09-13 flood. Judging from data of the available weather stations, rainfall was concentrated in the north for 2003 and in the south for 2004. The fact that water originates from different sources at different times will have a significant influence on the pollutant mass transport in the course of the flood wave. It was shown that ca. 3/4 of the total phosphorus mass and approximately 2/3 of TKN originates between K and S whereas most of the nitrate comes from the karstic spring (Obermann, 2007). These events both were the first significant floods of the respective hydrological years, they vary significantly in terms of maximum flow volume and rainfall, as summarized in Table 2. The 3 months before both events were very dry and there was hardly any flow or rainfall: only four events occurred before the event of 2003 (total volume of 42 mm) and five events occurred before the event of 2004 (total volume of 54 mm), both measured at Mas Plagnol. However, an extensive antecedent dry weather period (ADWP) only occurred for the event of 2003 of 27 days for the station Mas Plagnol (see Table 2). ADWP was calculated for each day of the year as the number of days from the last rainfall event:

0

3 km

Montbazin

K WW TP

rain [mm] flow (S) [m³/s]

8

4

Event 2003-09-22

0 2003-0 9-22

mm / 5min, m³/s

Climate data were available for the stations Montbazin, Mas Plagnol and Les Clachs (see Fig. 3). Data of two significant floods of the Vène were analysed in terms of a first flush dynamic, one at 2003-09-22 (19 measurements of which 13 were taken in the first 25% of cumulative flow at S) and the second at 2004-09-13 (15 measurements of which six were taken in the first 25% of flow at S) for the two sites S and V. Point S is situated downstream the effluent of a wastewater treatment plant, and point V is near the outlet of the catchment (Fig. 3). Both were regarded for the duration of 48 h, which includes the whole rainfall duration as well as the most distinctive discharge event (see Fig. 4). Fig. 4 shows that the flow event of 2003-09-22 can be divided into three parts at site S. The first part shows only small flow rates below 4 m3/s with a total runoff volume of about 14,000 m3. It is caused by the first storm with a maximum of 7.8 mm/5 min and a rainfall amount of 33.4 mm during 1.5 h. In this period effects

mm / 5min, m³/s

Inter-event distribution of loads – magnitude of days to hours

2003-09-23

4

Event 2004-09-13

0 2004-09 -13

2004-0 9-14

MasPlagnol

S Les Clash

V

2004-09 -15

2004-09-16

Fig. 4. Rainfall at the weather station Mas Plagnol and flow at station S during two significant events in the Vène catchment.

elevation [m] 0-50 50-100 100-150 150-200 200-250 250-300 300-350

geography: • • • • •

basin area: 67 km² elevation: 2 to 323 m river length: 12 km mean slope: 0.4% mean cross sections: 3 to 5 m

• population: 12,400 inhabitants • 3% urban area • land use: 63% natural karstic zones, 34% to agricultural zones

point-sources:

WW TP

2003-09-2 5

8

landuse:

Mon tbaz in

2003-09-2 4

Gig ean

• two wineries • two sewage treatment works sized for 9800 equivalent-inhabitants

K´ Fig. 3. Sketch of the Vène catchment.

410

M. Obermann et al. / Journal of Hydrology 373 (2009) 405–415

Table 2 Basic data of the two floods. Flood

Point

Total flow volume (m3)

Peak flow (m3/s)

TSS (mg)

VSS (mg)

TN (kg)

TP (kg)

NO3–N (kg)

2003-09-22

S – Sanglier V – Outlet

1262,587 1409,769

8.56 13.27

446 782

45.0 81.9

2110 2432

482 627

1381 1260

2004-09-13

S – Sanglier V – Outlet

178,612 421,574

9.70 19.49

178 356

15.6 29.9

462 878

222 385

230 436

ADWPðdi Þ ¼ di  dj

8 dj < di ^

i1 X

rðdk Þ ¼ 0

Table 3 Basic rainfall data for the two floods.

k¼j

where ADWP is the antecedent dry weather period and di, dj the days of the year and r(dk) is the amount of rainfall day dk. It is clear that this method does not take into account the overall wetness of the system, but only the period between rainfall events. This implies that the short antecedent dry period of the events in 2004 does not necessarily mean, that there has already been enough rainfall to produce runoff in the rivers. This may explain the reason why a strong relationship between the ADWP and the occurrence of a first flush could not be found (see also Deletic, 1998). However, the event of 2004-09-13 is a very important event, because the previous rainfalls were only small (15–18 mm) and not sufficient to produce runoff in the whole length of the river. A comparison of the normalized cumulative rainfall and the normalized cumulative flow in Fig. 5 shows the higher rainfall gradients for 2004. Even though it was smaller in terms of total discharge, the event was more condensed than the one of 2003 (see Table 3). This is also reflected in the discharges, which are rising slower and continue a long time after the end of the rainfall event in 2003. An approximation of total loads is shown for both floods in Table 2. In S only the mass of nitrate seems to have a direct correlation with the total flow. Other elements are transported in a higher amount than it could have been expected on basis of the flood volume for the first flush flood for 2004. Notably is the fact that the flood of 2004-09-13 had no karstic influence.

normalized cumulative flow & rainfall

Pollutographs of the event starting at 2003-09-22 Particulate matter. At the upstream site S, the most clear first flush effect can be seen at the particulate constituents in Fig. 6, especially for volatile suspended solids. The peak concentration at S is more than four times higher than average during the event and happens long before the peak of flow. This seems to be either caused by:

1.0

2003-09-22

0.5

0.0

normalized cumulative flow & rainfall

0 1.0

1

2

1

2

2004-09-13

0.5

rainfall Mas Plagnol flow Sanglier (S) flow Poussan (V) 0.0 0

days Fig. 5. Normalized cumulative flow and rainfall for the two events.

Flood

Point

Total rainfall volume (mm)

Maximum rainfall intensity (mm/h)

ADWP (d)

2003-09-22

Mas Plagnol Les Clachs

124.60 94.60

45.10 27.70

27 13

2004-09-13

Mas Plagnol Les Clachs

50.60 190.80

38.40 70.20

1a 1b

a There has been a small did not produce significant b There has been a small did not produce significant

precedent event at 2004-09-10 of about 15 mm which runoff and had a ADWP of 8 d. precedent event at 2004-09-10 of about 18 mm which runoff and had a ADWP of 9 d.

(i) wash off from the surrounding agricultural areas or (ii) a result of a resuspension of deposited materials like partly decomposed leaves or branches and grown biomass at the bottom of the channel. The latter would mean that the first small peak would have had enough transport capacity to entrain all – or most of – the accumulated organic components. The storage of VSS seems to be exhausted after the peak at the beginning. There only seems to be transport of a base concentration, because there is – unlike in TSS – no clear reaction of the concentration to the flow rate any more. Against the former possibility (i) militates the fact that the order of magnitude of the peak is higher at the upstream point S than in V (800 mg/l vs. 175 mg/l). The clearly visible ‘‘pre-peak” at S can not be found any more in V. This could mean that the impact of the instream resuspension decreases in downstream direction compared to the wash-off effect from the landside. It could be further evidence of the influence that the accumulation at S has on the mass flux regime of the whole system. Nitrogen. Only total nitrogen shows some dilution effect caused by the flood from 8.22 mg/l at the beginning of the event at 7:29– 2.13 mg/l before the start of the second phase at 9:30. For the later stages in the event, Nitrate is nearly constant at 1 mg/l which is near the mean concentration of water coming from the karstic spring. Fig. 6 shows loads and concentrations of nitrogen parameters for point S downstream the WWTP and point V near the outlet of the catchment. The concentration imposed by the outlet of the WWTP at Montbazin (upstream point S) was measured before the event of 2003-09-22 as 2.5 mg/l (2003-09-08) and after the event as 2.3 mg/l (2003-10-20). At the beginning of 2003-09-22, the concentration in the pool downstream the effluent was only about 0.5 mg/l. Hence, a great amount of the nitrate of the point source upstream of S seems to be reduced in the pools. Although there is no proof for that, it could be possible that a lot of the nitrogen was converted to organic forms from the biomass at the riparian zone or was removed from the system through denitrification. As explained earlier, the karstic spring dominates this event in the later stage from 16:00 ongoing. From the start of the spring the concentration of nitrate–nitrogen constantly rises from 0.6 mg/l to over 1 mg/l which is the mean spring concentration.

M. Obermann et al. / Journal of Hydrology 373 (2009) 405–415

411

Fig. 6. Pollutographs of the first significant flood events at 2003-09-22 and 2004-09-13 for site S and V.

Ammonium is not, as, e.g. nitrite, transported in the first instant of the event, but there seems to be a limited storage, as indicated by the decrease as the karstic spring starts flowing. Possibly, the ammonium is originated from the agricultural surfaces, as ammonium ions can bind to soils, especially clays. At the outlet point V there is a clear peak for concentration for nitrite, ammonium and total nitrogen. Regarding the plot of loads in Fig. 6 for the event reveals an increased mass flux of nitrite and total nitrogen at station S at the beginning of the event, which could not be observed for the concentrations. The plot of ammonium in site V also reveals a common misinterpretation which happens if only the concentration is considered: the first peak of concentration does not contribute significantly to the overall mass flux, whereas the later – and much smaller – one coincides with the peak in loading. It can also be seen that, even if there is a strong dilution effect in TN at the point S, the load remains at a higher level for a long time. Phosphorus. Dilution effects prevail at S. In the beginning of the event the concentration of total phosphorus quickly decreases at S from over 4.6 mg/l to below 1 mg/l. However, there is a peak in the load at 8:29 of about 6.4 g/s caused by the first small runoff maximum. Only at the outlet at V TP shows a peak in concentration of 2.9 mg/l at 11:25 (see Fig. 6). Pollutographs of the event starting at 2004-09-13 Particulate matter. The fast decrease of total suspended solids in the first flood of 2004 could be a sign of a limitation in storage (Fig. 6). In contrast to the event of September 2003, no preceding peak appears during the rising limb of the hydrograph in the loadings of VSS. The most active period of mass flux seems to be shifted towards the beginning of the water flow, very distinctly visible in the point S. Nitrogen. During the rising limb of the flood at 2004-09-13 23:06 Nitrate shows a small peak of 0.88 mg/l in concentration at S, but

of the mass flux is with 0.12 g/s small compared to the later stages of the event (from 2004-09-14 01:36 ongoing) when loads between 7 g/s and 9.6 g/s are reached for over 4 h. Ammonium and total nitrogen show similar curves for S but with earlier peaks of 0.6 mg/l, respective, 10 mg/l on 2004-09-13 22:36 (which could not be very well documented by the measurements). In the further progression, they both are diluted to base concentrations of about 0.1 mg/l for ammonium and 2–3 mg/l for TN. Only the pollutographs of nitrite exhibit a significant flushing with a low maximum concentration of 0.21 mg/l, which happens before the maximum flow on 2004-09-13 22:36 at S. However, due to the low concentrations nitrite does not contribute significantly to the total nitrogen load of the event. Phosphorus. Soluble reactive phosphorus is diluted in site S starting from a concentration of 1.45 mg/l on 2004-09-13 22:36 to below 0.1 mg/l on 2004-09-14 00:06. Particulate phosphorus seems to have two concentration peaks of 15.9 mg/l and 17.7 mg/l in 2004 at station S (Fig. 7). Near the end of the event from 200409-14 01:36 ongoing it is also subject to dilution. This could also be a cause of an instream storage, especially if the plots of PP and VSS are compared (Fig. 7). It seems as if particulate forms of phosphorus are bound to the sediments (TSS respective VSS). Tournoud et al. (2005) found that, due to alkaline pH-values of the water column (more than 7.5) and high calcium carbonate saturation indices, co-precipitation of phosphorus and calcium carbonates (House, 2003; Plant and House, 2002) occurred and this trapped phosphorus in the bed. In both points (S and V) the course of concentration of PP and VSS is similar, so that it can be suggested that these two parameters are transported by comparable flow conditions. NCL-plots of the event 2003-09-22 Plots of normalized cumulative loads versus normalized cumulative discharge were created for the two investigated sampling points as shown in Fig. 8. Because the first water sample for an

M. Obermann et al. / Journal of Hydrology 373 (2009) 405–415

1000

100

VSS PP

500

50

0

0 150

1500 1000 500 0

75 0

2004-09-13 00:00:00

2004-09-14 00:00:00

VSS (S) [mg/l] VSS (V) [mg/l]

2004-09-15 00:00:00

Fig. 7. Comparison of concentrations of particulate phosphorus and volatile suspended solids for the flood of 2004-09-13.

1.0

normalized cumulative load

2003-09-22 0.8

(i) inflow of less polluted rain water and subsurface flow, (ii) worse data situation at V, overrepresentation of the later flood part, (iii) the inflow of an additional karstic source between S and V and (iv) possible mixing within the flood wave.

0.6

0.4

0.2 Sanglier (S)

Phosphorus. In terms of the normalized loading curves, the results are very diverse in S and V for soluble reactive phosphorus. While total phosphorus’ FF25 is above 60% for both points, transport of the soluble part is delayed based on the regarded period, as cause of a high measured concentration after the end of the event at V (compare plot at V for SRP in Fig. 6).

0.0 1.0

normalized cumulative load

Nitrogen. It is remarkable that the nitrogen derivates are very inhomogeneous in terms of their transport. The previously denoted peaks in nitrite loads find their counterpart here only at S where nearly 50% of the load is transported with only 10% of discharge. The measurements end early at V, possibly the concentration (and therefore also the load) at S was assumed too high by the interpolation and the later part got too much weight. The most extraordinary curve is the one of ammonium at S. For this flood, even the very strict criteria of Saget et al. (1995) is met, because FF30 is more than 83%. Nitrate is below the bisector in both points and because it is the dominating nitrogen species it has an important effect on total nitrogen. The direct comparison of stations Sanglier and Poussan shows, how the first flush is disappearing along the 2.8 km from the accumulation spot of the WWTP at S to the point near the outlet V. This can have different reasons:

0.8

0.6 TSS VSS TP SRP TN TKN NH4 NO2 NO3 cond.

0.4

0.2 Poussan (V) 0.0 0.0

0.2

0.4

0.6

0.8

1.0 2004-09-13

1.0

normalized cumulative discharge Fig. 8. NCL-plot of 2003-09-22 for points S and V – normalized cumulative loads against normalized cumulative discharge.

normalized cumulative load

PP (S) [µg/l] PP (V) [µg/l]

412

0.8

0.6 TSS VSS TP SRP TN TKN NH4 NO2 NO3 cond.

0.4

0.2 Sanglier (S) 0.0

Particulate matter. During the analysis of site V it must be kept in mind that the recession of the flood wave is not so well documented by measurements as for site S (see Fig. 6). So in cases, when the last measured concentration before the start of the event is relatively high, the first flush can be underestimated, if this last concentration is comparably low, the first flush might be overestimated. However, for site S, where concentration measurements cover the whole regarded duration, particulate materials show an increased transport activity for the first 30% of the event discharge.

1.0

normalized cumulative load

event is taken depending on changes in flow depth, it is possible that the record starts after the beginning of the event, and there is no sample available directly before the onset of the flood (as it happened in 2004). In order not to overestimate the first flush, the load was assumed to stay constant from the last measured point before the event up to the first measured point within the event. All other load values were then linearly interpolated.

0.8

0.6

0.4

0.2

Poussan (V) 0.0 0.0

0.2

0.4

0.6

0.8

1.0

normalized cumulative discharge Fig. 9. NCL-plot of 2004-09-13 for points S and V – normalized cumulative loads against normalized cumulative discharge.

413

M. Obermann et al. / Journal of Hydrology 373 (2009) 405–415

Phosphorus. Apart from the beginning of the event, mass flux of phosphorus seems to follow the flow intensity. Some explanations, why the first flush is less pronounced in 2004-09-13 could be:

NCL-plots of the event 2004-09-13 In comparison to 2003, for the flood of 2004-09-13 the first flushes are not so clear (Fig. 9). Particulate matter However, in terms of the normalized cumulative loadings, the particulate materials show at least a weak flushing behaviour at V compared to the other parameters (Fig. 9). Here the FF25 is about 42% for TSS which still meets the criteria of Vorreiter and Hickey (1994), which defined a first flush for a FF25 greater than 0.4–0.6. Compared to the event of 2003-09-22 the first flush is not so strong for TSS in 2004-09-13 (FF25 are 0.72 and 0.39, respectively, see Fig. 10). This can be attributed to the shape of the floods. The flood of 2004 is very condensed and the flow goes down to zero flow quickly. If it would have continued, the lower concentrations in the later stages would have shifted the weight of mass transport more to the beginning of the event, i.e. a greater share of mass would have been transported at the beginning of the event in relation to the whole event.

(i) the small rainfall events before could have displaced and diffused most of the accumulated nutrients in the river; even if the water did not reach the receiving water, the pollutants are on one hand distributed in a smaller concentration over a longer stream section and on the other hand a great part of the material could have been settled downstream of site S. (ii) the shape of the flood is more condensed and there is no small peak before the main peak, as in 2003-09-22; if this small first peak is capable of resuspending a great part of the accumulated matter, the transport activity of the flood is concentrated in the beginning. However, the maxima of particulate matter concentrations are nearly equal for both floods, so they differ mainly in their total loadings and in the inter-event distribution.

Nitrogen. The fact that nitrite exhibit a significant flushing (as said before for 2003 and indicated in Fig. 8) is also very well represented in the NCL-curves of the event in 2004 (Fig. 9). Compared to the first flood in September 2003 the magnitude of first flushes has decreased especially in S. Nitrate shows a very slow reaction to the flow condition in both sites, TN and NH4–N are a bit above the bisector in S and a bit below in V.

Application of rating methods for first flushes of each parameter According to ‘‘Indices as classification methods for first flushes”, the indices were applied to the two first significant flood events at site S as shown above. The trend for all parameters in site S in Fig. 10 is independent of the index method. The highest first flush intensity in 2003-09-22 is

site S - 2003-09-22 1.00

NCL from 0 to 1 0.79 0.72

0.75

0.81

0.79 0.74

0.72

0.70

0.65

rating [-]

0.53 0.49

Geiger (1984)

0.70

0.68 0.61

0.50

FF25

0.79

0.56

0.55

0.56

0.47

0.43 0.44

0.42

0.37

0.39

0.35 0.30

0.25

0.19

0.10 0.02

0.13 0.03

) ty

(S

(S ) iv i

O

3

2( S)

C

on

du

ct

N

N

N

H

O

4

(S )

(S )

(S )

TK N

SR P

TN

(S )

) (S TP

VS S

TS S

(S )

(S )

0.00

site S - 2004-09-13 1.00

0.74

0.75

0.72

0.71

0.68

0.68

0.67

0.59 0.49

0.50

0.39

0.40

0.36

0.38 0.33

0.33

0.39 0.30

0.50

0.48 0.40

0.38

0.30

0.29

0.20

0.25

0.29 0.21

0.19 0.13

0.11

0.03

) ity

3( S

iv

O C

on

du

ct

N

O 2 N

(S

)

) (S

) 4 H N

TK

N

(S

(S

)

) (S TN

SR P

(S

)

) (S TP

) (S VS S

S

(S

)

0.00

TS

rating [-]

0.62

Fig. 10. Rating of first flushes for the site S on September 2003 and 2004 events.

414

M. Obermann et al. / Journal of Hydrology 373 (2009) 405–415

therefore reached by ammonium, followed by the particulate properties TSS and VSS. Apart from nitrate and conductivity, all parameters are transported in excess in the first 25% of the flood in 2003. This as well accounts for the flood of the following year, but clearly less pronounced. In the flood of 2004-09-13 no first flush of ammonium is happening, so that now TSS and VSS as well as TP prevail. The comparison of the two different floods makes one major limitation of the first method obvious: due to the integral calculation, the changes in the index values are not so clear as in the other methods (FF25 and the index of Geiger (1984)).

Discussion and conclusions The method of normalized cumulative loads (NCL) was reviewed and applied for a larger watershed. It proved to be a most valuable method in this context. Three examples for index-methods (FF25, Geiger (1984), NCLintegral) to quantify first flush intensity were tested for artificial examples and discussed. All these methods showed comparable trends, but especially the FF25 method proved to be a helpful parameter. The approach classifies the extent of a first flush by the percentage of cumulative load at 25% of cumulative flow. It does not share the negative feature of the NCL-integral method, which shows only smaller differences between varying flood situations. The FF25 approach gives the most descriptive parameter and was therefore found to be the most applicable. Important first flushes for particulates have been shown with NCL-plots as well as through the application of three indices: FF25 values for TSS and VSS reach up to 0.72, respective, 0.70. The clearest first flush occurred for ammonia in the course of the flood of 2003, where 79% of the NH4–N event load was transported by the first quarter of flow. There is a strong first flush at 2003-09-22 for VSS and TSS but also for solubles as ammonium and a very early one for nitrite. However, the first flush of nitrite exists only downstream the WWTP at S. Whereas all constituents except nitrate and conductivity were transported in excess in the beginning of the event at S, the first flush diminishes in downstream direction. At station V, only for TSS, VSS, TP and NH4–N occurs a notable first flush. It can be concluded that a first flush of particulates is more likely in most cases in the Vène. This owes the fact that these constituents can be much better accumulated instream. On the other hand, a first flush of soluble compounds is not so common, but if there is the possibility of an accumulation in instream pools, the phenomena is normally more intense. Especially in fast rising floods, when particulates must follow the incoming wave of water, and there is some time lost while the sediment transport adopts to the changing flow condition, the soluble constituents are already fully transported. They can travel, at least for a while, before the front of the incoming rain water. Unfortunately, the design of measurement campaigns sometimes does not allow to capture these phenomena. The first significant flood of 2003/2004 did not show such clear flush events, which could have been a cause of the small event before. As said above, the FF25 index proved to be the most valuable and was therefore applied to the Vène. The highest value was calculated for the 2003-09-22 flood for ammonium (0.79) followed by TSS (0.72) and VSS (0.70). With regards to the implementation of the European Water Framework Directive there is a need to define a good ecological status for temporary water bodies and to provide suitable tools methods and models. Measurements in the catchment are still done in the framework of current European and national projects. Additional information, especially on the landside part and on the

reaction of the waste water treatment plants during storm events, will enable a step towards a joint investigation of the river system including downstream basins, reservoirs or lagoons. The highly varying flows cause concentrated waves of nutrients and pollutants. This means a substantial risk for downstream water bodies as lagoons, lakes or reservoirs as the substances will have a harmful potential here. This study proves the existence of first flushes in a medium sized catchment like the Vène. Further it was shown that the already available and well-tested approaches for first flush investigation in smaller catchments can be successfully applied in larger ones. This work contributes to the understanding of water quality dynamics in dryland rivers. The results clearly showed that these effects are of high relevancy to the overall mass transport within these rivers and have to be considered in any management. The derived knowledge enables advanced measurement campaigns and best management options, which can be tailored to the specific needs of comparable dryland rivers. In this context, there is a need for new approaches to an adapted integrative management following the principle of reducing ambient pollution (‘‘immission-based”). Higher loads to the receiving water bodies during the flow period could be tolerated in favour of a reduction of the accumulation during the dry summer period. This could be realised, e.g. by waste water infiltration measures, direct pumping of treated waste water into the downstream standing water bodies during summer or a real time management on basis of adopted water quality models. Acknowledgements The EC is acknowledged for financing this study within the project tempQsim (EVK1-CT2002-00112). The authors would like to thank the members of the consortium of tempQsim for the open and fruitful cooperation, especially Dr.-Ing. Jochen Froebrich for his effort as coordinator of the project, as well as Bernadette Picot for the provision and collection of main parts of the data. References APHA, AWWA, WEF, 1992. Standard Methods for the Examination of Water and Wastewater. American Public Health Association, Washington. Bedient, P.B., Harned, D.A., Characklis, W.G., 1978. Stormwater analysis and prediction in Houston. Journal of the Environmental Engineering Division – ASCE 104 (6), 1087–1100. Bertrand-Krajewski, J.L., Chebbo, G., Saget, A., 1998. Distribution of pollutant mass vs. volume in stormwater discharges and the first flush phenomenon. Water Resources 32 (8), 2341–2356. Characklis, G.W., Wiesner, M.R., 1997. Particles, metals, and water quality in runoff from large urban watershed. Journal of Environmental Engineering – ASCE 123 (8), 753–759. Cristina, C.M., Sansalone, J.J., 2003. ‘‘First flush,” power law and particle separation diagrams for urban storm-water suspended particulates. Journal of Environmental Engineering 129 (4), 298–307. Deletic, A., 1998. The first flush load of urban surface runoff. Water Research 32 (8), 2462–2470. Geiger, W.F., 1984. Characteristics of combined sewer runoff. In: Proc. 3rd International Conference on Urban Storm Drainage, Göteborg, Sweden. Gupta, K., Saul, A.J., 1996a. Specific relationships for the first flush load in combined sewer flows. Water Research 30 (5), 1244–1252. Gupta, K., Saul, A.J., 1996b. Suspended solids in combined sewer flows. Water Science and Technology 33 (9), 93–99. House, W.A., 2003. Geochemical cycling of phosphorus in rivers. Applied Geochemistry 18 (5), 739–748. Krebs, P., Holzer, P., Huisman, J.L., Rauch, W., 1999. First flush of dissolved compounds. Water Science and Technology 39 (9), 55–62. Lee, J.H., Bang, K.W., 2000. Characterization of urban stormwater runoff. Water Research 34 (6), 1773–1780. Lee, J.H., Yu, M.J., Bang, K.W., Choe, J.S., 2003. Evaluation of the methods for first flush analysis in urban watersheds. Water Science and Technology 48 (10), 167–176. Lee, H., Lau, S.L., Kayhanian, M., Stenstrom, M.K., 2004. Seasonal first flush phenomenon of urban stormwater discharges. Water Research 38 (19), 4153– 4163.

M. Obermann et al. / Journal of Hydrology 373 (2009) 405–415 Obermann, M., 2007. Nutrient Dynamics in Temporary Waters of Mediterranean Catchments, vol. 138. Veröffentlichungen des Instituts für Siedlungswasserwirtschaft und Abfalltechnik, Hannover, 228 pp. (978-3921421-68-0). Obermann, M., Froebrich, J., Perrin, J.-L., Tournoud, M.-G., 2007. Impact of significant floods on the annual load in an agricultural catchment in the mediterranean. Journal of Hydrology 334 (1–2), 99–108. Plant, L.J., House, W.A., 2002. Precipitation of calcite in the presence of inorganic phosphate. Colloids and Surfaces A: Physicochemical and Engineering Aspects 203 (1–3), 143–153. Saget, A., Chebbo, G., Bertrand-Krajewski, J., 1995. The first flush in sewer system. In: International Conference on Sewer Solids-Characteristics, Movement, Effects and Control, Dundee, UK.

415

Saget, A., Chebbo, G., Bertrand-Krajewsk, J.L., 1996. The first flush in sewer systems. Water Science Technology 33 (9), 101–108. Sansalone, J.J., Cristina, C.M., 2004. First flush concepts for suspended and dissolved solids in small impervious watersheds. Journal of Environmental Engineering – ASCE 130 (11), 1301–1314. Tournoud, M.G., Perrin, J.L., Gimbert, F., Picot, B., 2005. Spatial evolution of nitrogen and phosphorus loads along a small mediterranean river: implication of bed sediments. Hydrological Processes 19 (18), 3581–3592. Tournoud, M.-G., Payraudeau, S., Cernesson, F., Salles, C., 2006. Origins and quantification of nitrogen inputs into a coastal lagoon: application to the Thau lagoon (France). Ecological Modelling 193 (1–2), 19–33. Vorreiter, L., Hickey, C., 1994. Incidence of the first flush phenomenon in catchments of the Sydney region. In: National Conf. Publication – Institution of Engineers, vol. 3, pp. 359-364.