- Email: [email protected]
Temporal variability of suspended sediment transport and rating curves in a Mediterranean river basin: The Celone (SE Italy)
Catena 128 (2015) 135–143
Contents lists available at ScienceDirect
Catena journal homepage: www.elsevier.com/locate/catena
Temporal variability of suspended sediment transport and rating curves in a Mediterranean river basin: The Celone (SE Italy) A.M. De Girolamo ⁎, G. Pappagallo, A. Lo Porto Water Research Institute, National Research Council, 70132 Bari, Italy
a r t i c l e
i n f o
Article history: Received 3 December 2013 Received in revised form 11 July 2014 Accepted 22 September 2014 Available online xxxx Keywords: Suspended sediment load Hysteresis Sediment rating curves Temporary river Monitoring
a b s t r a c t In the Mediterranean region suspended sediment transport is the predominant process in sediment export in most river basins. The aim of this paper is to analyze suspended sediment variability over a period of 12 months in the Celone river, a temporary river located in the Puglia region (SE-Italy), and to evaluate sediment rating curves for estimating suspended sediment concentrations for subsequent load calculations. Similarly to most temporary rivers, the Celone river shows relevant differences among mean daily flows and the extreme instantaneous flows during floods. To take into account these peculiarities, the rating curves were developed as a function of hydrological conditions: high, normal and low flows. Continuous measures of streamflow and frequent samplings of suspended solid concentrations (SSCs) during flood events, normal flow and low flow were used. The plot of the SSC against discharge takes the form of a hysteresis loop. Clockwise, anticlockwise and mixedshaped loops were observed. Suspended sediment yield was found to be in the range of 250–384 t km−2 y−1. The results show that about 94% of the total suspended materials were transported during the high flow regime, while less than 0.1% were under low flow conditions. Moreover, it was observed that 90% of the total annual suspended loads were moved between November to May. Flash floods that occur in summer exhibit the highest values of SSC. The proposed method, which was based on sediment rating curves, has proved to be valuable to generate SSC data for high and normal flows although it tends to underestimate the highest values. It can represent a useful tool for water resource managers who need a quick and inexpensive method, specific for temporary rivers, to evaluate suspended sediment yield. © 2015 Elsevier B.V. All rights reserved.
1. Introduction In the Mediterranean region, most river basins are affected by erosion and soil degradation (Jones et al., 2012). The geomorphologic and climatic factors that characterize these basins and the agricultural practices such as frequent tillage, which are quite common in these areas, exacerbate soil losses (Gomez et al., 2009). In recent decades, erosion and its impacts on soil and surface waters have received an increasing interest from local, national, European, or international policy makers. Many studies have focused on soil erosion processes, sediment dynamics, sediment yield evaluation, reservoir sedimentation, and ecological aspects related to the suspended sediment transport and on specific measures to reduce soil erosion (e.g., Kirkby et al., 2000; Lenzi and Marchi, 2000; Morgan, 2005; Rodriguez-Blanco et al., 2010; Soler et al., 2008; Van Rompaey et al., 2005; Verstraeten et al., 2003). Most of the studies on the dynamics of suspended sediment carried out in the Mediterranean region have analyzed small semi-arid catchments. Few studies of suspended sediment transport have been carried out in medium Apennine basins with high seasonal differences in streamflow ⁎ Corresponding author. E-mail address: [email protected] (A.M. De Girolamo).
http://dx.doi.org/10.1016/j.catena.2014.09.020 0341-8162/© 2015 Elsevier B.V. All rights reserved.
(Pavanelli and Cavazza, 2010). In these basins, soils are characterized by a high percentage of silt and clay particles, which have a great erodibility. Soil erosion and river suspended solids (SSCs) are strongly related and suspended sediment transport can constitute a large part of the total sediment load (Pavanelli and Cavazza, 2010). Hence, the quantification of suspended sediment yield at the basin outlet provides an order of magnitude estimate of the erosion and depositional processes occurring within the catchment. The hydrologic regime of the Mediterranean rivers is an important factor in influencing erosion and sediment delivery processes. Due to the high variability in time and space of rainfall events, these rivers are often characterized by extreme variations in flow (Nikolaidis et al., 2013) and flash floods with high suspended sediment transport (Alexandrov and Laronne, 2003). This aspect makes it more difficult to make accurate and continuous measurements of SSC (Navratil et al., 2011) and at the same time it implies that suspended sediment load computation is quite difficult (Phillips et al., 1999). Several methods have been developed to predict suspended sediment yield in medium and large catchments (Moatar and Meybeck, 2005; Letcher et al., 1999). These include the use of empirical relationships (rating curves) between SSC and streamflow, and more processbased generation and transport models (Arnold et al., 1998). Data
136
A.M. De Girolamo et al. / Catena 128 (2015) 135–143
availability, size of watershed and streamflow regime play an important role in the selection of an estimator method, although all of them have their limitations (Letcher et al., 2002). Models require hydrological and environmental data for quantifying soil erosion and for validating the results (De Girolamo and Lo Porto, 2012; De Girolamo et al., 2014) and usually are not able to simulate in-stream process like bank erosion or remobilization of sediment previously deposited on the river bed (Duvert et al., 2012). Process-based methods need a huge amount of SSC and streamflow measurements but field measurement of sediment is a very demanding and expensive task which is difficult to sustain over a long time. Horowitz (2003) pointed out that SSC–Q relationship shows a high degree of scatter and consequently a sediment rating curve could under or over-estimate concentrations and loads. On the other hand, most of the suspended sediment rating curves require detailed information on the flow and sediment characteristics and generally do not agree with each other, making it difficult to choose the best equation for a given stream. In addition, these curves have been developed for perennial rivers, characterized by differences among the mean daily flow and the maximum and minimum streamflow during the day not very large. Because of these problems, there is a gap in the use of rating curves in temporary rivers where the differences among mean daily flows and the extreme instantaneous flows during the day are important. Nevertheless, a quantification of sediment loads, as well as an understanding of the dynamic of suspended sediment transfer from lands to water courses, is necessary for an integrated water and soil management, as required by the EU Water Framework Directive (EC Directive 2000/60/EC, 2000) and the EU Soil Thematic Strategy (EC, 2006). The objectives of the present work are to i) analyze the variability of the suspended sediment transport at different temporal scales in the Celone river basin, a representative example of Mediterranean temporary river located in the Puglia region (Subappennino Dauno), ii) analyze and process the most common suspended sediment rating curves usually used for evaluating suspended material: power curve, linear and polynomial Log transformed equations and iii) distinguish different hysteresis loops for flood events. The purpose is on the one hand to improve the understanding of suspended sediment dynamics and on the other hand provide an analysis of rating curves useful for water resource managers who need a quick and inexpensive tool, specific for temporary rivers to evaluate the suspended sediment yield. In the upper Celone river basin, a quantification of the suspended sediment is particularly important because the river is the predominant inflow of the Capaccio reservoir. Suspended sediment can act as a vector for a wide variety of organic and inorganic chemical constituents (Horowitz, 2008), therefore suspended sediment load can contribute substantially to water quality and habitat impairments (Larsen et al., 2010). Hence, a reliable quantification of suspended loads is fundamental for defining a Program of Measures to reduce erosion and pollutant transport. The hysteresis loops in some flood events will be useful for interpreting the spatial distribution of catchment sediment sources within the catchment. 2. Study area The study area (72 km2) is the upper Celone river basin, a watershed located in the Puglia region in southern Italy (Fig. 1). The Celone river, which is a tributary of the Candelaro river, has its origin at Monte S. Vito, one of the major summits of the “Monti Dauni” in the Apennines. It flows northeast and enters the Capaccio reservoir (capacity at full supply 25.82 Mm3; current volume 17.56 Mm3). Upstream of the Capaccio reservoir, the main course of the river is 24 km long and the whole drainage network is 81 km. The drainage system has a dendritic pattern which is the result of the constraint imposed on sediment erosion, transport and deposition by the geology and topography of the drainage
Fig. 1. Study area: Celone catchment.
area. The channel, which is incised in the upper basin, assumes a braided form in an alluvial plain downstream of the steeper reaches. Here, there is deposition of the coarser material, before the river resumes a sinuous course. To reduce the flow rate and to control bank erosion, a number of permanent check dams were built where the river is steeply sloped. Major bedrock lithological units are flyschoid formations (flysch della Daunia) and green-blue clays in the upper Celone basin and alluvial deposits in the lowlands. The main soil types are classified as Typic Haploxeroll, Vertic Haploxerept and Typic Calcixeroll according to the USDA Soil Taxonomy. Soil texture is sandy-clay-loam, clay-loam or clay. The mean elevation of the area is 386 m above sea level, ranging from 218 m to 1142 m a.s.l. Most of the basin falls between 200 and 500 m (43%), with 29% in the range of 501–800 m, and the remaining 28% of the basin in the range of 800–1142 m. The major land use is rainfed agriculture: durum wheat (46%), sunflowers (9%), olive trees (8%), and pasture (6%) are the main crops. Natural deciduous forest (mainly Quercus spp., Fagus sylvatica L.) and coniferous plantations (Pinus spp.) are also present (29%). In autumn and winter, much of the cultivated land is unprotected from erosion (drilled wheat fields and fields tilled for spring crops). Agriculture is not intensive and tillage operations are conventional. The main active erosion processes are sheet wash, and concentrated water erosion (gullies and rills). Erosion is encouraged by the up and down plowing, which is quite common in this basin where the land is divided into long strips running down to the river valleys. Moreover, bank erosion is also an active process. From 1960 to 2000 the mean annual precipitation is 795 mm at Faeto gauge and 653 mm at Troia gauge. The rainfall season is from November to May, it provides approximately 70% of total annual precipitation, while during the dry season, from June to September, rainfall is concentrated in few events of short duration and high intensity. Consequently, the streamflow regime has a typical Mediterranean semi-arid feature with a seasonal pattern of a drought period and flash floods. The mean monthly temperature ranges from 3.4 °C (January) to 20.3 °C (August), in the upper part of the basin, and from 7.2 °C to 25.5 °C in the lowland area. 3. Material and methods 3.1. Measurement of suspended sediment The surface water flow and suspended sediment concentrations (SSCs) were measured over a period of 12 months from July 2010 to July 2011 at a monitoring point (Masseria Pirro) located 8 km upstream
A.M. De Girolamo et al. / Catena 128 (2015) 135–143
of the Capaccio dam (Fig. 1). This site constitutes the only gauging station in the Candelaro basin providing data suited to evaluate a rating curve. An ISCO automatic sampler (model 6712FS; 24 bottles; pumped volume 1 L) with an internal data logger was installed at this station. The sampler was connected to a flow module to measure the channel flow (ISCO 750 Area Velocity Flow Module). The sensor provides continuous (5 min) velocity and stream water stage measurements, which were converted to discharge using a predefined stage-discharge rating curve. As Fig. 2 shows, the river section where the sensor was installed shows a regular and permanent shape. Using a template (Excel worksheet), we calculate the cross-sectional area corresponding to each water stage. Discharge is calculated by multiplying water velocity with the corresponding cross sectional area of the river. Several tests were done to verify the sensor measurements. A large number of samples was collected (N = 210) by using a different frequency for water sampling in the diverse flow conditions. The sampler has two sets of programming. The first one lets us set up time-paced sampling programs. With this standard program, periodic samples were taken at fortnightly or monthly time intervals during summer and autumn periods, and once or twice a week from November to June. The second set lets us create complex programs for flood sampling applications. We adopted a sampling strategy triggered by water level changes during the rising limb of hydrograph and flow rates during the flood recession. By using this program during floods, with some exceptions, the time intervals varied from 15 min to 2 h over the rising limb of hydrograph and from 2 h to 1 day over the flood recession. The total number of samples therefore varied between flood events. Twenty-one flood events occurred in the study period; the majority of the peak flows were sampled for SSC except for four events during which some technical problems caused the interruption of samplings. All water samples were collected by the automatic sampler. The samples were taken at random times during the day in normal and low flow conditions. The sampler intake nozzle was positioned vertically to the flow at the center of the cross section (Fig. 2) and was submerged following the US Geological Service manual (Edwards and Glysson, 1999). We verified that within the river the lateral diffusion was sufficient to ensure homogeneous lateral mixing of suspended sediment. Total suspended solid concentrations (SSCs) were determined using the APAT-IRSA analytical standard methods (APAT-IRSA-CNR, 2003). Standard 0.45 μm pore diameter cellulose filters were used to filter the samples. 3.2. Data analysis The relationship between SSC and streamflow was investigated and then loads under different flow conditions were evaluated. We also
Fig. 2. Cross section of the Celone river, Masseria Pirro gauging station. Schematic view from downstream to upstream.
137
analyzed hydrological and rainfall data recorded at two gauging stations located in the mountainous and in the plain area, respectively (Fig. 1). Annual rainfall data recorded over the study period were 43% and 2.4% higher than available historical data1 (Table 1). While, the most relevant events in terms of peak discharge are comparable to the major events recorded in the past at the S. Vincenzo gauge (recurrence interval of 30 years), an old station which was located 8 km downstream the new one (Masseria Pirro). Hence, we can say that the study period is a representative of historical hydrologic conditions in the basin. The load passing through a river cross-section during a time interval can be mathematically expressed by: Zt2 L¼
Q t C t dt
ð1Þ
t1
where Qt is the streamflow (L s− 1) at time t, Ct is the total suspended solid (TSS) concentration (mg L − 1) at time t (s), and L is the load (mg). The most common method used to calculate flux is a regression equation that defines an empirical relationship, which relates SSC to streamflow at the time of sampling (suspended sediment curve). Generally, this is a power function derived by a log-transformed leastsquares regression (linear, Eq. (2); or second-order polynomial) where the discharge (Q) is the independent variable (Phillips et al., 1999) and SSC (C) is the dependent variable. logC ¼ loga þ b logQ:
ð2Þ
The final result of the log-transformed regression method is often required in original units and the retransformation leads to a bias correction problem (Ferguson, 1986). Koch and Smillie (1986) pointed out that the mean value on the original scale is underestimated if the log-result is simply back-transformed to the original scale by exponentiation. This observation derived from the fact that the backtransformed mean value from the log scale is equal to the geometric mean2 of the values on the original scale. The geometric mean is always less than the arithmetic mean and, thus, the back-transformed mean always underestimates the arithmetic mean from the original scale. Several attempts were made to improve the C–Q relationships by using a bias correction factor (Bradu and Mundlak, 1970; Duan, 1983; Gilroy et al., 1990; Koch and Smille, 1986) or using subsets of data (seasonal, or on specific stage of flow) instead of whole available data set (Asselman, 2000; Walling and Webb, 1996). In this study, we have determined three regression equations (power curve, Log transformed linear and polynomial function) for the entire data set (210 samples) and for three subsets of data defined as a function of hydrological conditions: high, normal and low flows. The flow duration curve (FDC) was used to select the thresholds between three different flow classes: high, normal and low flows. Streamflow values higher than 0.930 m3 s− 1 are defined as high flow, corresponding to the interval of flow frequency exceedance of 0–20%, discharges ranged from 0.930 to 0.140 m3 s− 1 are defined as normal flow (20–70% frequency exceedance) and streamflow values less than 0.140 m3 s− 1 are defined as low flow. In addition, for the regression equations in logarithmic space (linear and polynomial), we also defined the bias correction of the retransformation proposed by Duan (1983) which does not require any assumption about the distribution of residuals. The latter, also called
1 Many gaps were found in the historical data especially in correspondence of high and intense events. 2 The geometric mean is defined as the nth root of the product of the n values.
138
A.M. De Girolamo et al. / Catena 128 (2015) 135–143
4. Results
Table 1 Basic hydrological data recorded in the catchment. Time period
2010–2011 1960–1996 a b c d e
Rainfall (mm y−1)
Rainfall (mm y−1)
Dischargee (mm day−1)
Faeto
Troia
Min
Max
Mean
1140 792a ± 135b
638 623a ± 110b
0.005c 0.000d
19.810 19.230
0.960 0.417
Mean value. Standard deviation. Masseria Pirro gauge (drainage area 72 km2). S. Vincenzo gauge (drainage area 86 km2). Mean annual flows normalized to the corresponding drainage area.
Smearing Estimator (CF), was applied to the back-log transformed data using the following equation: N X
CF1 ¼
εi
10
1
N
ð3Þ
where: ′ εi ¼ logðC i Þ− log C i is the residual and N is the number of observations. The correlation between contemporaneous SSC and streamflow values was determined for the entire data set, and for the three subsets of data defined above. For each type of curve, the mean error (E) was evaluated as percentage of the differences between measured concentration (Ci) and predicted values from regression (C′i) (Horowitz, 2003; Eq. (4)): 13 2 0 ′ N C i −C i X A7 6 @ 7 6 Ci 7 6 1 7 100: 6 Eð%Þ ¼ 6 7 N 7 6 5 4
ð4Þ
The C–Q relationships (rating curves) were used to evaluate SSC at a 15-minute time interval. Integration on daily time scale of the 15minute discharge and the SSC provided estimated daily loads, which were then aggregated on different time scales: year, season, and month. For each day, the load was calculated using the following expression (Eq. (5)):
Ldaily ¼
96 X
900 Q C
ð5Þ
1
where Ldaily is the daily load (kg), Q is the measured streamflow (m3 s−1), and C is the predicted SSC (g L−1) and 900 is the number of seconds in a time interval (15 min).
4.1. Suspended sediment concentration variability The hydrological regime of the Celone river plays an important role in the suspended sediment transport. Fig. 3 illustrates the variation in water discharge and SSC, from July 2010 to July 2011. The graph shows that equal peaks of discharge may have different magnitudes of SSC, for example, the peaks on 30 November (10.83 m3 s−1; 3.920 g L−1) and 16 May (10.44 m3 s−1; 0.810 g L−1). This behavior is due to the effect of vegetation in addition to a possible variation in rainfall pattern in combination with the different antecedent soil moisture contents. In November, wheat fields are unprotected while in May crops evenly cover the soil. The maximum sediment concentration recorded in the wet season was 7.13 g L−1, associated with a streamflow value of 23.5 m3 s− 1 (5th March 2011). During the summer period (June and July), two flash flood events were recorded (Fig. 3, events: 1 and 2). Each of these occurred over a period of a few hours during which water carried a huge amount of suspended material and debris. The first recorded flash flood occurred in July, after a dry period of approximately 50 days, was significant in terms of concentration (4.77 g L− 1) although the contemporaneous streamflow value (0.955 m3 s−1) is not of great magnitude. A more relevant event took place on 21st July 2010, when particularly heavy rainfall was recorded (73.6 mm in an hour at Faeto gauging station), this value being much higher than the monthly average recorded in July from 1960 to 2000 (32 mm), corresponding to the monthly rainfall of a winter month (February, March). Consequently, in 15 min the streamflow passed from 0.1 m3 s− 1 to 4.35 m3 s− 1 and in the same time interval the SSC increased from 0.069 g L−1 to 16.87 g L−1. The maximum value of the measured SSC was 37.60 g L−1 (recorded on 21st July 2010), associated with a streamflow value of 5.95 m3 s−1, corresponding to the highest value of the ratio SSC/Q in the study period. The ability of the basin to provide flows of various magnitudes and the percentage of time river flow exceeds a specified value was expressed by the flow duration curve (FDC). Fig. 4 shows SSC and the FDC for the study period; the outliers of SSC represent the values recorded during the flash flood events. The shape of FDC in its upper and lower regions is a very steep curve, indicating rain-caused floods and an intermittent regime. The river, at the Masseria Pirro gauging station, shows a very rapid rising stage and a short lag time (time between peak rainfall and peak discharge). Fig. 5 illustrates hydrographs of the recorded floods denoted by an asterisk in Fig. 3. Flood duration is typically only a few hours during which the SSC, as well as streamflow, increases and decreases rapidly. From November to March, the peak of the SSC generally precedes the flow peak (8 events) and the delay time was estimated as an hour. While, in late spring and in summer months, peaks of discharge and SSC were nearly synchronous. After the discharge reaches its flow peak, it falls off gradually resulting in a very slow recession with a duration depending on the antecedent soil moisture, and on the intensity and time interval of the rainfall that produced it. As Fig. 5 shows, in a succession of flood events there is no evident reduction of the sediment concentrations in the successive peak (24–25 December, 5–6 March, 30–31 March).
Fig. 3. Suspended sediment concentration and streamflow, plotted as a function of time. Recorded floods are denoted by a number. Floods denoted by an asterisk are illustrated in Figs. 5 and 6.
A.M. De Girolamo et al. / Catena 128 (2015) 135–143
139
0.11 g L−1 (2.33 m3 s−1). Many kinds of hysteresis loop were found in the wet season when multiple-peaks are quite common. 4.2. Fitted rating curve
Fig. 4. Flow duration curve.
The plot of the SSC against discharge often takes the form of a hysteresis loop (Fig. 6). Clockwise, anticlockwise and mixed-shaped loops were observed in the Celone river in the study period. A clockwise response was found in the first peak of multiple-peak events and when the events are caused by high intensity rainfalls (18 December; 15 mm in 3 h). When rainfall intensity is high runoff generation is mainly due to infiltration excess (Kirkby et al., 2011) and flow energy can increase rapidly causing the flushing of sediment previously accumulated on the river bed and a dilution or the exhaustion of sediment supply towards the end (Alexandrov and Laronne, 2003). Consequently, during the rising limb the SSC can be higher than that recorded at equivalent discharge in the recession limb with differences that can be relevant. For instance, on 18th December, SSC is 1.92 g L−1 in the rising limb (corresponding discharge 2.15 m3 s− 1), while in the falling limb it is only
After excluding 7 outlier points, which were the SSC values recorded during two flash floods, the correlation between the streamflow and SSC was found to be 0.81. The F-test showed this to be statistically significant (p b 0.01). The corresponding coefficient of determination (R2 = 0.66) indicates that 66% of variance of SSC may be explained by streamflow while 34% is justified by other factors such as rainfall intensity and its spatial distribution, vegetation cover and soil management. This result can be considered quite good for a regression method to be carefully applied (Quilbé et al., 2006). After the subdivision into discharge classes, the correlation coefficient decreases, taking the values: 0.78, 0.22 and 0.10, for “high”, “normal” and “low” flows, respectively. Fig. 7 shows values for SSC and Q for the all data sets; in the graph outlier points were removed. A high variability in SSC was recorded: SSC ranges from 0.005 to 4.77 g L−1 in normal flow, and from 0.005 to 0.360 g L−1 in low flow conditions. It should be noted that each class of flow includes both the steady flow and the recession limb of floods, which can record very different SSC values. Fig. 8 represents the recorded data and the second order polynomial recession curve fitted to the whole data set. Table 2 summarizes the equations and their characteristics and the correction factors calculated as stated before. 4.3. Assessment of suspended sediment loads at different time scales Among the rating curves defined in Table 2, we selected the equation with minimum mean error between the measured and estimated concentrations (58% in Table 3) which is a Log transformed polynomial
Fig. 5. Measured streamflow and suspended sediment concentrations during some flood events at the Celone Masseria Pirro gauging station.
140
A.M. De Girolamo et al. / Catena 128 (2015) 135–143
Fig. 6. Typical hysteresis effect observed in suspended sediment during some flood events at the Celone river (Masseria Pirro gauging station). Flood recorded on 3rd May follows an event occurred on 1st May (not monitored). Arrows indicate direction through time.
curve for all data sets (Eq. (8), in Table 2). By using this equation suspended sediment yields were evaluated at 31,265 and 19,064 t, with and without the smearing correction, respectively. The corresponding specific suspended sediment yields are 434 and
Fig. 7. Suspended sediment concentration versus discharge for the study area (7 outlier points were excluded).
Fig. 8. Sediment rating curve. Second-order polynomial recession curve for all data sets.
A.M. De Girolamo et al. / Catena 128 (2015) 135–143
141
Table 2 Sediment rating curves. Regression coefficients (a, b, c), number of data points (N), correction factor (CF), streamflow (Q). Log transformed data (linear) (Eq. (7)) logC = alogQ + b
Power curve (Eq. (6)) C = aQb
All (N = 210) High flow (N = 153) Normal flow (N = 63) Low flow (N = 10)
Log transformed data (polyn.) (Eq. (8)) logC = a(logQ)2 + blogQ + c
a
b
R2
a
b
R2
CF
a
b
c
R2
CF
0.1994 0.1533 0.3093 0.744
0.938 1.141 1.255 1.106
0.61 0.52 0.16 0.19
0.938 1.141 1.051 0.845
−0.700 −0.815 −0.643 −0.420
0.61 0.52 0.13 0.07
1.69 1.37 3.68 2.60
0.223 0.118 −1.344 −16.63
0.905 1.000 −0.060 −34.960
−0.779 −0.785 −0.792 −19.278
0.63 0.52 0.14 0.36
1.64 1.37 3.54 2.60
265 t km−2 y−1. However, when Eq. (8) for all data sets was used with the smearing factor, the mean error was found 158%. As expected, most of the suspended sediment yield is transported during the high flow status. In particular, the high flow is estimated to account for 94% of the total amount suspended material, while less than 0.1% of the suspended material is transported at low flow. It is interesting to point out that the high flow status has been recorded as occurring in 67 days, while normal flow and low flow in 186 and 112 days, respectively. On a monthly basis, more than 90% of the total annual suspended load is transported from November to May while in June the suspended sediment yield is less than 100 t per month, and decreases from August to October when it reaches the minimum at less than 10 t per month. The seasonal and monthly distributions of SS load follow the general pattern of the water yield (Fig. 9). A calculation of the annual load was also done by using a different equation for each class of flow. Even in this case, we selected the rating curves on the basis of the minimum mean error between measured and estimated sediment concentrations (Table 3). In details, we used for the three classes of flow (high, normal and low) the equations Eq. (8), Eq. (7), Eq. (7) (Table 2), respectively. The corresponding annual loads were estimated at 27,675 and 18,019 t y−1, with and without correction factor (CF), respectively. The corresponding specific suspended sediment yields are 384 and 250 t km−2 y−1. The mean error between measured and estimated loads is 63% and when CF is applied the results indicate an increase in mean error (263%). A comparative analysis of measured and generated suspended loads for flood events shows that actual loads are underestimated. The difference between the measured and estimated values is relevant for the major floods. For instance, in March when floods with high magnitude occurred, measured sediment load was 2898 t for event 9* (Fig. 3) and 8904 t for event 10*, while estimated values were 2423 t and 8002 t, respectively. In contrast, the loads are overestimated if the CF is used. In particular, estimated suspended sediment loads were found 3320 t and 10,962 t, respectively. As most of the suspended material is transported during floods, we can say that the actual annual loads are underestimated using suspended sediment rating curves. 5. Discussion
varying between 0.007 and 7249.792 t (i.e., 9.9 ∗ 10− 7 and 1.00 t km−2, respectively). The variability of concentrations during floods shows a hysteresis behavior similar to that observed in other Mediterranean catchments (Oeurng et al., 2010). Clockwise, anticlockwise and mixed-shaped loops were observed in the Celone river in the study period. A clockwise hysteresis is due to a progressive decline in suspended sediment availability (Gellis, 2012) that was found in events caused by high intensity rainfalls and in flood events away from previous others. This behavior reflects an exhaustion of suspended sediment supply, which occurs when one of the sources of eroded material is the channel itself. We consider sediment accumulated along the channel during the normal flow and low flow conditions likely to be remobilized during the rising limb of hydrograph with a decreasing sediment availability in the recession limb. Anticlockwise loop take place when suspended sediment sources are spread all over the catchment and it derives from processes whose dynamics are slower than the streamflow, i.e. contribution of sediment from distant areas and bank collapse may increase SSC in the recession limb. In the wet season, when multiple-peaks are quite common, many kinds of hysteresis loop were found. This behavior reveals the complexity of the suspended sediment dynamics which are controlled by several factors related to sediment sources, storm intensity, soil conditions (antecedent soil water content), vegetation density and tillage works. In contrast with other studies conducted in the Mediterranean region (Rovira and Batalla, 2006), that have recorded a progressive reduction of SSC within multiple-peak events, the analysis of SSC in the study area shows that the magnitude of peak discharges exerts a primary control on suspended sediment concentration. In the light of these results, we can say that there is no exhaustion of suspended sediment availability in the basin because different sediment sources contribute to suspended sediment material during floods: sheet wash, concentrated water erosion (gullies and rills) and channel bank erosion. The highest values of the SSC/Q ratio were recorded during the flash floods, which occur in summer after a dry period. This data reveals the potential threat for waters due to an excessive discharge of suspended sediment, and of a wide variety of organic and inorganic chemical constituents associated with sediment, which can degrade river habitat (Horowitz, 2008). Ribarova et al. (2008) pointed out that a high concentration of suspended sediment and nutrients during flash floods is very common in temporary rivers where a high level of pollutant accumulated
In common with results already reported by other studies conducted in the Mediterranean region (Gentile et al., 2010; López-Tarazón et al., 2010), the Celone suspended sediment transport response is highly variable over time. SSC vary from b0.005 to 37.60 g L−1 and daily loads Table 3 Mean error between measured and estimated sediment concentrations for the different types of rating curves. Type of rating curve
All data High flow Normal flow Low flow
Mean error (%) Eq. (6)
Eq. (7)
Eq. (8)
60 38 156 104
60 34 128 104
58 34 131 123
Fig. 9. Monthly suspended sediment load and water yield estimated from July 2010 to June 2011.
142
A.M. De Girolamo et al. / Catena 128 (2015) 135–143
along the river network is remobilized during these events (Chu et al., 2008). A previous research conducted in the study area confirmed a very high level of nutrients during flash flood events (De Girolamo et al., 2012). Alexandrov et al. (2003) observed that in the Mediterranean region data concerning flash floods are few and, generally, information about suspended sediment concentrations has been obtained mainly by manual sampling and often random. In this context, the present investigation gives a contribution to improve the knowledge about the sediment transport during flash floods in temporary rivers. With regard to the sediment rating curves, it must be noted that all the relationships presented here underestimate SSC in extreme high flow conditions while overestimate SSC in extreme low flow conditions. The resulting loads estimated using these equations underestimate the actual load. This result was expected, as the log-results were simply back-transformed to the original scale by exponentiation. In this case, the bias correction procedure developed by Duan (CF), applied to reduce error in the use of log-transformed regression to derive rating relation, fails to provide a more reliable estimate of suspended sediment yield. Moreover, the actual mean annual sediment yield is always significantly overestimated with the correction factor. These results confirm the studies by Kellerhals Engineering Services (1985) and Walling and Webb (1988). They reported that the bias associated with the logarithmic transformation is not the major cause of error in sediment rating curve estimations. Important factors are the large scatter associated with the plot of data and the hysteretic and exhaustion effects during storm events. However, we can assume that the actual value of specific yield is included in the range from 250 to 384 t km−2 y−1 (which are the values evaluated without and with CF, respectively). These results are in good agreement with the findings of Van Rompaey et al. (2005) which apply to mountain catchments with a Mediterranean climate in southern Italy (200–400 t km−2 y−1). It is well known that soil formation rates are variable but generally very low. Alexander (1988) measured a rate of soil formation ranging from 0.02 to 1.90 t km−2 y−1 from small watersheds under forest and grassland. Doran (1996) observed that 100–400 years are needed for developing a centimeter of topsoil, while for conditions which are prevalent in Europe, Verheijen et al. (2009) found a rate of soil formation of 140 t km−2 y− 1 (0.056 mm y− 1). Based on our calculations of suspended sediment load, we estimated that in the Celone basin the soil loss per year is N 0.1 mm, this means that soil is being lost 2–3 times faster than the rate of renewal and sustainability. Hence, we can say that soil erosion is effectively irreversible with potential high economic and environmental impacts. 6. Conclusions In unraveling the complexity of the sediment issue, the present work gives a contribution in increasing the knowledge of the processes concerned, which in turn is fundamental to integrating sediment and water quality into the river basin management. In fact, only a holistic approach can guarantee a high level of environmental quality, as required by the EU Environmental Policy. The temporal dynamic of suspended sediment transport in the Celone catchment showed a high variability: intra-annual, monthly and within-events. Suspended sediment load is mainly transported in winter and spring, when floods with high magnitude occurred, vegetation cover is absent or very low and tillage operations are performed. Over time, the variability of discharge and SSC resulted in different hysteric loops. Clockwise shapes mainly occurred in isolated floods when the sediments accumulated in the channel are removed with runoff; mixed shaped loops were found in multiple-peaks. All the methodologies for sediment yield evaluation show some limits due to the limitations in the model representations of natural processes characterized by heterogeneity. These difficulties increase in the Mediterranean basins where rivers have a temporary character and
where rainfall events are highly variable. The rating curves proposed here, developed as a function of hydrological conditions, suggest that this approach is good enough to generate SS concentration data for high flows, although it tends to underestimate the highest values. This method is not appropriate for low flow because of high variability in the data. However, it is important to note that low flow has a limited influence on the annual load of less than 0.1%. Hence, a better solution to estimate the SS flux could be to use a suspended sediment rating curve for high and normal flows and an averaging method for the low flow. Despite their limits, these relationships can be a very useful tool when a suspended sediment flux determination is required at a high temporal resolution (i.e. Total Maximum Daily Load methodology which needs SS load at a daily time scale). In addition, suspended sediment curves offer the possibility of reconstructing long term sediment transport data, which are fundamental to evaluate reservoir sedimentation. Hence, this technique may have important implications for the Capaccio reservoir management, providing a quick and inexpensive tool to assess the reduction of its storage capacity due to the sedimentation processes. Sediment rating curves, in fact, once defined can be used to evaluate suspended material transported by the river in different flow conditions by using only streamflow measurements that are more easier to measure than the sediment concentrations. Further sampling campaigns are needed in order to validate the sediment rating curves defined here, while an analysis of the data already collected can be useful for defining a new monitoring program (i.e. identifying the time period during which only sporadic samplings are sufficient to determine the SSC and loads and point out the critical time period during which continuous samplings are needed). Finally, further studies should investigate and quantify the uncertainty linked to load estimation. Acknowledgments This research was supported by the European Community's Seventh Framework Programme, FP7/2007–2011, MIRAGE Project (Contract no.: 211732). The authors gratefully acknowledge Giuseppe Amoruso from the Apulia Civil Protection Service for providing the data. The authors are indebted to Angelantonio Calabrese who collaborates in the field activities and analytical determinations of SSC. We thank David Cooper for his helpful comment to improve the quality of the paper. Thanks are also due to Stefano Pirro for helping us in identifying the best location for installing the equipment, to Giuseppe and Nicola Pirro for controlling it. Lastly, we would like to thank the reviewers for their valuable scientific comments and recommendations. Appendix A. Supplementary data Supplementary data associated with this article can be found in the online version, at http://dx.doi.org/10.1016/j.catena.2014.09.020. These data include Google maps of the most important areas described in this article. References Alexander, E.B., 1988. Rates of soil formation: implications for soil loss tolerance. Soil Sci. 145 (1), 37–45. Alexandrov, Y., Laronne, J.B., 2003. Suspended sediment transport in flash floods of the semiarid northern Negev, Israel. Proc. Int. Symp. Hydrology of the Mediterranean and Semiarid Regions. IAHS Publ. 278, Montpellier, pp. 346–352. Alexandrov, Y., Laronne, J.B., Reid, L., 2003. Suspended sediment concentration and its variation with water discharge in dryland ephemeral channel, northern Negev, Israel. J. Arid Environ. 53, 73–84. APAT-IRSA-CNR, 2003. Metodi analitici per le acque, APAT Manuali e Linee Guida 29/ 2003. pp. 1–1153. Arnold, J.G., Srinivasan, R., Muttiah, R.S., Williams, J.R., 1998. Large area hydrologic modelling and assessment part I: model development. J. Am. Water Resour. Assoc. 34 (1), 73–89. Asselman, N., 2000. Fitting and interpretation of sediment rating curves. J. Hydrol. 234, 228–248.
A.M. De Girolamo et al. / Catena 128 (2015) 135–143 Bradu, D., Mundlak, Y., 1970. Estimation in lognormal models. J. Am. Stat. Assoc. 65, 198–211. Chu, Y., Salle, C., Cernesson, F., Perrin, J.L., Tournoud, M.G., 2008. Nutrient load modelling during floods in intermittent rivers: an operational approach. Environ. Model Softw. 23, 768–781. De Girolamo, A.M., Lo Porto, A., 2012. Land use scenario development as a tool for watershed management within the Rio Mannu Basin. Land Use Policy 29, 691–701. De Girolamo, A.M., Calabrese, A., Pappagallo, G., Santese, G., Lo Porto, A., 2012. Impact of anthropogenic activities on a temporary river. Fresenius Environ. Bull. 21 (11), 3278–3286. De Girolamo, A.M., Lo Porto, A., Pappagallo, G., Tzoraki, O., Gallart, F., 2014. The hydrological status concept. application at a temporary river (Candelaro, Italy). River Res. Appl. http://dx.doi.org/10.1002/rra.2786. Doran, J., 1996. The international situation and criteria for indicators, chap 7. In: Cameron, K.C., Comforth, I.S., McLaren, R.G., Beare, M.H., Basher, L.R., Metherell, A.K., Kerr, L.E. (Eds.), Soil Quality Indicators for Sustainable Agriculture in New Zealand: Proceedings of a Workshop. Lincoln Soil Quality Research Centre, Lincoln University, New Zealand. Duan, N., 1983. Smearing estimate: a nonparametric retransformation method. J. Am. Stat. Soc. 78, 605–610. Duvert, C., Nord, G., Gratiot, N., Navratil, O., Nadal-Romero, E., Mathys, N., Némery, J., Regüés, D., García-Ruiz, J.M., Gallart, F., Esteves, M., 2012. Towards prediction of suspended sediment yield from peak discharge in small erodible mountainous catchments (0.45–22 km2) of France, Mexico and Spain. J. Hydrol. 454–455, 42–55. EC, 2006. A strategy to keep Europe's soils robust and healthy. http://ec.europa.eu/ environment/soil/index.htm. EC Directive 2000/60/EC of the European Parliament and the Council Directive establishing a framework for Community action in the field of water policy, Official Journal of the European Communities, Brussels, 22/12/2000. Edwards, T.K., Glysson, G.D., 1999. Field methods for measurement of fluvial sediment. USGS Open-file Report Book 3 (Chapter 2), pp. 1–97. Kellerhals Engineering Services, 1985. Sediment in the Pacific and Yukon region: review and assessment. Water Resources Branch, Inland Waters Directorate Environment Canada, Report IWD-HQWRB-SS-85-8. Ferguson, R.I., 1986. River loads underestimated by rating curves. Water Resour. Res. 21, 74–76. Gellis, A.C., 2012. Factors influencing storm-generated suspended-sediment concentrations and loads in four basins of contrasting land use, humid-tropical Puerto Rico. Catena 104, 35–57. Gentile, F., Bisantino, T., Corbino, R., Milillo, F., Romano, G., Trisorio, Liuzzi G., 2010. Monitoring and analysis of suspended sediment transport dynamics in the Carapelle torrent (Southern Italy). Catena 80, 1–8. Gilroy, E.J., Hirsh, R.M., Cohn, T.A., 1990. Mean square error of regression-based constituent transport estimates. Water Resour. Res. 26 (9), 2069–2077. Gomez, J.A., Sobrinho, T.A., Giraldez, J.V., Fereres, E., 2009. Soil management effects on runoff, erosion and soil properties in an olive grove of Southern Spain. Soil Tillage Res. 102, 5–13. Horowitz, A.J., 2003. An evaluation of sediment rating curves for estimating suspended sediment concentrations for subsequent flux calculations. Hydrol. Process. 17, 3387–3409. Horowitz, A.J., 2008. Determining annual suspended sediment and sediment-associated trace element and nutrient fluxes. Sci. Total Environ. 400 (1–3), 315–343. Jones, A., Panagos, P., Barcelo, S., Bouraoui, F., Bosco, C., Dewitte, O., Gardi, C., Erhard, M., Hervás, J., Hiederer, R., Jeffery, S., Lükewille, A., Marmo, L., Montanarella, L., Olazábal, C., Petersen, J.E., Penizek, V., Strassburger, T., Tóth, G., Van Den Eeckhaut, M., Van Liedekerke, M., Verheijen, F., Viestova, E., Yigini, Y., 2012. The state of soil in Europe. JRC Reference Report 68418. 1831-9424 http://dx.doi.org/10.2788/77361. Kirkby, M.J., Le Bissonais, Y., Coulthard, T.J., Daroussin, J., McMahon, M.D., 2000. The development of land quality indicators for soil degradation by water erosion. Agric. Ecosyst. Environ. 81, 125–136. Kirkby, M.J., Gallart, F., Kjeldsen, T.R., Irvine, B.J., Froebrich, J., Lo Porto, A., De Girolamo, A., the MIRAGE team, 2011. Classifying low flow hydrological regimes at a regional scale. Hydrol. Earth Syst. Sci. 15, 3741–3750. Koch, R.W., Smille, G.M., 1986. Bias in hydrologic prediction using log-transformed regression models. Water Resour. Bull. 22 (5), 717–723.
143
Koch, R.W., Smillie, G.M., 1986. Comment on “river loads underestimated by rating curves” by R.I. Ferguson. Water Resour. Res. 22 (13), 2121–2122. Larsen, M.C., Gellis, A.C., Glysson, G.D., Gray, J.R., Horowitz, A.J., 2010. Fluvial sediment in the environment: a national problem. Proceedings, 2nd Joint Federal Interagency Conference, Las Vegas, NV. June 27–July 1 (15 pp.). Lenzi, M., Marchi, L., 2000. Suspended sediment load during floods in a small stream of the Dolomites (northeastern Italy). Catena 39, 267–282. Letcher, R.A., Jakeman, A.J., Merritt, W.S., McKee, L.J., Eyre, B.D., Baginska, B., 1999. Review of techniques to estimate catchment exports. Report EPA 99/73, Environment Protection Authority, Sydney, Australia. Letcher, R.A., Jakeman, A.J., Calfas, M., Linforth, S., Baginska, B., Lawrence, I., 2002. A comparison of catchment water quality models and direct estimation techniques. Environ. Model Softw. 17, 77–85. López-Tarazón, J.A., Batalla, R.J., Vericat, D., Balasch, J.C., 2010. Rainfall, runoff and sediment transport relations in a mesoscale mountainous catchment: the River Isábena (Ebro basin). Catena 82, 23–34. Moatar, F., Meybeck, M., 2005. Compared performances of different algorithms for estimating annual nutrient loads discharged by the eutrophic River Loire. Hydrol. Process. 19 (2), 429–444. Morgan, R.P.C., 2005. Soil Erosion and Conservation, 3rd. Blackwell, Oxford, p. 314. Navratil, O., Esteves, M., Legout, C., Gratiot, N., Némery, J., Willmore, S., Grangeon, T., 2011. Global uncertainty analysis of suspended sediment monitoring using turbidimeter in a small mountainous river catchment. J. Hydrol. 398, 246–259. Nikolaidis, N.P., Demetropoulou, L., Froebrich, J., Jacobs, C., Gallart, F., Prat, N., Lo Porto, A., Campana, C., Papadoulakis, V., Skoulikidis, N., Davy, T., Bidoglio, G., Bouraoui, F., Kirby, M., Tournoud, M.G., Polesello, S., Barberá, G.G., Cooper, D., Gomez, R., SanchezMontoya, M.M., Latron, J., De Girolamo, A.M., Perrin, J.L., 2013. Towards a sustainable management of Mediterranean river basins. Policy recommendations on management aspects of temporary river basins. Water Policy 15, 830–849. Oeurng, C., Sauvage, S., Sánchez-Pérez, J.M., 2010. Dynamics of suspended sediment transport and yield in a large agricultural catchment, southwest France. Earth Surf. Process. Landf. 35, 1289–1301. Pavanelli, D., Cavazza, C., 2010. River suspended sediment control through riparian vegetation: a method to detect the functionality of riparian vegetation. Clean Soil Air Water 38 (11), 1039–1046. Phillips, J.M., Webb, B.W., Walling, D.E., Leeks, G.J.L., 1999. Estimating the suspended sediment loads of rivers in the LOIS study area using infrequent samples. Hydrol. Process. 13, 1035–1050. Quilbé, R., Rousseau, A.N., Duchemin, M., Poulin, A., Gangbazo, G., Villeneuve, J.P., 2006. Selecting a calculation method to estimate sediment and nutrient loads in streams: application to the Beaurivage River (Québec, Canada). J. Hydrol. 326 (1–4), 295–310. Ribarova, I., Ninov, P., Cooper, D., 2008. Modeling nutrient pollution during a first flood event using HSPF software: Iskar River case study, Bulgaria. Ecol. Model. 211, 241–246. Rodriguez-Blanco, M.L., Taboada-Castro, M.M., Taboada-Castro, M.T., 2010. Sources and sediment yield from a rural catchment in humid temperate environment, northwest Spain. Earth Surf. Process. Landf. 35, 272–277. Rovira, A., Batalla, R.J., 2006. Temporal distribution of suspended sediment transport in a Mediterranean basin: the Lower Tordera (NE SPAIN). Geomorphology 79, 58–71. Soler, M., Latron, J., Gallart, F., 2008. Relationship between suspended sediment concentrations and discharge in two small research basins in a mountainous Mediterranean area (Vallcebre, Eastern Pyrenees). Geomorphology 98, 143–152. Van Rompaey, A., Bazzoffi, P., Jones, R.J.A., Montanarella, L., 2005. Modeling sediment yields in Italian catchments. Geomorphology 65, 157–169. Verheijen, F.G.A., Rickson, R.J., Smith, C.J., 2009. Tolerable versus actual soil erosion rates in Europe. Earth Sci. Rev. 94, 23–38. Verstraeten, G., de Poesen, J., Vente, J., Koninckx, X., 2003. Sediment yield variability in Spain: a quantitative and semiqualitative analysis using reservoir sedimentation rates. Geomorphology 50 (4), 327–348. Walling, D.A., Webb, B.W., 1988. The reliability of rating curve estimates of suspended sediment yield: further comments. IAHS Publ. 174, 337–350. Walling, D.A., Webb, B.W., 1996. Erosion and sediment yield: a global overview. IAHS Publ. 236, 3–19.
Contents lists available at ScienceDirect
Catena journal homepage: www.elsevier.com/locate/catena
Temporal variability of suspended sediment transport and rating curves in a Mediterranean river basin: The Celone (SE Italy) A.M. De Girolamo ⁎, G. Pappagallo, A. Lo Porto Water Research Institute, National Research Council, 70132 Bari, Italy
a r t i c l e
i n f o
Article history: Received 3 December 2013 Received in revised form 11 July 2014 Accepted 22 September 2014 Available online xxxx Keywords: Suspended sediment load Hysteresis Sediment rating curves Temporary river Monitoring
a b s t r a c t In the Mediterranean region suspended sediment transport is the predominant process in sediment export in most river basins. The aim of this paper is to analyze suspended sediment variability over a period of 12 months in the Celone river, a temporary river located in the Puglia region (SE-Italy), and to evaluate sediment rating curves for estimating suspended sediment concentrations for subsequent load calculations. Similarly to most temporary rivers, the Celone river shows relevant differences among mean daily flows and the extreme instantaneous flows during floods. To take into account these peculiarities, the rating curves were developed as a function of hydrological conditions: high, normal and low flows. Continuous measures of streamflow and frequent samplings of suspended solid concentrations (SSCs) during flood events, normal flow and low flow were used. The plot of the SSC against discharge takes the form of a hysteresis loop. Clockwise, anticlockwise and mixedshaped loops were observed. Suspended sediment yield was found to be in the range of 250–384 t km−2 y−1. The results show that about 94% of the total suspended materials were transported during the high flow regime, while less than 0.1% were under low flow conditions. Moreover, it was observed that 90% of the total annual suspended loads were moved between November to May. Flash floods that occur in summer exhibit the highest values of SSC. The proposed method, which was based on sediment rating curves, has proved to be valuable to generate SSC data for high and normal flows although it tends to underestimate the highest values. It can represent a useful tool for water resource managers who need a quick and inexpensive method, specific for temporary rivers, to evaluate suspended sediment yield. © 2015 Elsevier B.V. All rights reserved.
1. Introduction In the Mediterranean region, most river basins are affected by erosion and soil degradation (Jones et al., 2012). The geomorphologic and climatic factors that characterize these basins and the agricultural practices such as frequent tillage, which are quite common in these areas, exacerbate soil losses (Gomez et al., 2009). In recent decades, erosion and its impacts on soil and surface waters have received an increasing interest from local, national, European, or international policy makers. Many studies have focused on soil erosion processes, sediment dynamics, sediment yield evaluation, reservoir sedimentation, and ecological aspects related to the suspended sediment transport and on specific measures to reduce soil erosion (e.g., Kirkby et al., 2000; Lenzi and Marchi, 2000; Morgan, 2005; Rodriguez-Blanco et al., 2010; Soler et al., 2008; Van Rompaey et al., 2005; Verstraeten et al., 2003). Most of the studies on the dynamics of suspended sediment carried out in the Mediterranean region have analyzed small semi-arid catchments. Few studies of suspended sediment transport have been carried out in medium Apennine basins with high seasonal differences in streamflow ⁎ Corresponding author. E-mail address: [email protected] (A.M. De Girolamo).
http://dx.doi.org/10.1016/j.catena.2014.09.020 0341-8162/© 2015 Elsevier B.V. All rights reserved.
(Pavanelli and Cavazza, 2010). In these basins, soils are characterized by a high percentage of silt and clay particles, which have a great erodibility. Soil erosion and river suspended solids (SSCs) are strongly related and suspended sediment transport can constitute a large part of the total sediment load (Pavanelli and Cavazza, 2010). Hence, the quantification of suspended sediment yield at the basin outlet provides an order of magnitude estimate of the erosion and depositional processes occurring within the catchment. The hydrologic regime of the Mediterranean rivers is an important factor in influencing erosion and sediment delivery processes. Due to the high variability in time and space of rainfall events, these rivers are often characterized by extreme variations in flow (Nikolaidis et al., 2013) and flash floods with high suspended sediment transport (Alexandrov and Laronne, 2003). This aspect makes it more difficult to make accurate and continuous measurements of SSC (Navratil et al., 2011) and at the same time it implies that suspended sediment load computation is quite difficult (Phillips et al., 1999). Several methods have been developed to predict suspended sediment yield in medium and large catchments (Moatar and Meybeck, 2005; Letcher et al., 1999). These include the use of empirical relationships (rating curves) between SSC and streamflow, and more processbased generation and transport models (Arnold et al., 1998). Data
136
A.M. De Girolamo et al. / Catena 128 (2015) 135–143
availability, size of watershed and streamflow regime play an important role in the selection of an estimator method, although all of them have their limitations (Letcher et al., 2002). Models require hydrological and environmental data for quantifying soil erosion and for validating the results (De Girolamo and Lo Porto, 2012; De Girolamo et al., 2014) and usually are not able to simulate in-stream process like bank erosion or remobilization of sediment previously deposited on the river bed (Duvert et al., 2012). Process-based methods need a huge amount of SSC and streamflow measurements but field measurement of sediment is a very demanding and expensive task which is difficult to sustain over a long time. Horowitz (2003) pointed out that SSC–Q relationship shows a high degree of scatter and consequently a sediment rating curve could under or over-estimate concentrations and loads. On the other hand, most of the suspended sediment rating curves require detailed information on the flow and sediment characteristics and generally do not agree with each other, making it difficult to choose the best equation for a given stream. In addition, these curves have been developed for perennial rivers, characterized by differences among the mean daily flow and the maximum and minimum streamflow during the day not very large. Because of these problems, there is a gap in the use of rating curves in temporary rivers where the differences among mean daily flows and the extreme instantaneous flows during the day are important. Nevertheless, a quantification of sediment loads, as well as an understanding of the dynamic of suspended sediment transfer from lands to water courses, is necessary for an integrated water and soil management, as required by the EU Water Framework Directive (EC Directive 2000/60/EC, 2000) and the EU Soil Thematic Strategy (EC, 2006). The objectives of the present work are to i) analyze the variability of the suspended sediment transport at different temporal scales in the Celone river basin, a representative example of Mediterranean temporary river located in the Puglia region (Subappennino Dauno), ii) analyze and process the most common suspended sediment rating curves usually used for evaluating suspended material: power curve, linear and polynomial Log transformed equations and iii) distinguish different hysteresis loops for flood events. The purpose is on the one hand to improve the understanding of suspended sediment dynamics and on the other hand provide an analysis of rating curves useful for water resource managers who need a quick and inexpensive tool, specific for temporary rivers to evaluate the suspended sediment yield. In the upper Celone river basin, a quantification of the suspended sediment is particularly important because the river is the predominant inflow of the Capaccio reservoir. Suspended sediment can act as a vector for a wide variety of organic and inorganic chemical constituents (Horowitz, 2008), therefore suspended sediment load can contribute substantially to water quality and habitat impairments (Larsen et al., 2010). Hence, a reliable quantification of suspended loads is fundamental for defining a Program of Measures to reduce erosion and pollutant transport. The hysteresis loops in some flood events will be useful for interpreting the spatial distribution of catchment sediment sources within the catchment. 2. Study area The study area (72 km2) is the upper Celone river basin, a watershed located in the Puglia region in southern Italy (Fig. 1). The Celone river, which is a tributary of the Candelaro river, has its origin at Monte S. Vito, one of the major summits of the “Monti Dauni” in the Apennines. It flows northeast and enters the Capaccio reservoir (capacity at full supply 25.82 Mm3; current volume 17.56 Mm3). Upstream of the Capaccio reservoir, the main course of the river is 24 km long and the whole drainage network is 81 km. The drainage system has a dendritic pattern which is the result of the constraint imposed on sediment erosion, transport and deposition by the geology and topography of the drainage
Fig. 1. Study area: Celone catchment.
area. The channel, which is incised in the upper basin, assumes a braided form in an alluvial plain downstream of the steeper reaches. Here, there is deposition of the coarser material, before the river resumes a sinuous course. To reduce the flow rate and to control bank erosion, a number of permanent check dams were built where the river is steeply sloped. Major bedrock lithological units are flyschoid formations (flysch della Daunia) and green-blue clays in the upper Celone basin and alluvial deposits in the lowlands. The main soil types are classified as Typic Haploxeroll, Vertic Haploxerept and Typic Calcixeroll according to the USDA Soil Taxonomy. Soil texture is sandy-clay-loam, clay-loam or clay. The mean elevation of the area is 386 m above sea level, ranging from 218 m to 1142 m a.s.l. Most of the basin falls between 200 and 500 m (43%), with 29% in the range of 501–800 m, and the remaining 28% of the basin in the range of 800–1142 m. The major land use is rainfed agriculture: durum wheat (46%), sunflowers (9%), olive trees (8%), and pasture (6%) are the main crops. Natural deciduous forest (mainly Quercus spp., Fagus sylvatica L.) and coniferous plantations (Pinus spp.) are also present (29%). In autumn and winter, much of the cultivated land is unprotected from erosion (drilled wheat fields and fields tilled for spring crops). Agriculture is not intensive and tillage operations are conventional. The main active erosion processes are sheet wash, and concentrated water erosion (gullies and rills). Erosion is encouraged by the up and down plowing, which is quite common in this basin where the land is divided into long strips running down to the river valleys. Moreover, bank erosion is also an active process. From 1960 to 2000 the mean annual precipitation is 795 mm at Faeto gauge and 653 mm at Troia gauge. The rainfall season is from November to May, it provides approximately 70% of total annual precipitation, while during the dry season, from June to September, rainfall is concentrated in few events of short duration and high intensity. Consequently, the streamflow regime has a typical Mediterranean semi-arid feature with a seasonal pattern of a drought period and flash floods. The mean monthly temperature ranges from 3.4 °C (January) to 20.3 °C (August), in the upper part of the basin, and from 7.2 °C to 25.5 °C in the lowland area. 3. Material and methods 3.1. Measurement of suspended sediment The surface water flow and suspended sediment concentrations (SSCs) were measured over a period of 12 months from July 2010 to July 2011 at a monitoring point (Masseria Pirro) located 8 km upstream
A.M. De Girolamo et al. / Catena 128 (2015) 135–143
of the Capaccio dam (Fig. 1). This site constitutes the only gauging station in the Candelaro basin providing data suited to evaluate a rating curve. An ISCO automatic sampler (model 6712FS; 24 bottles; pumped volume 1 L) with an internal data logger was installed at this station. The sampler was connected to a flow module to measure the channel flow (ISCO 750 Area Velocity Flow Module). The sensor provides continuous (5 min) velocity and stream water stage measurements, which were converted to discharge using a predefined stage-discharge rating curve. As Fig. 2 shows, the river section where the sensor was installed shows a regular and permanent shape. Using a template (Excel worksheet), we calculate the cross-sectional area corresponding to each water stage. Discharge is calculated by multiplying water velocity with the corresponding cross sectional area of the river. Several tests were done to verify the sensor measurements. A large number of samples was collected (N = 210) by using a different frequency for water sampling in the diverse flow conditions. The sampler has two sets of programming. The first one lets us set up time-paced sampling programs. With this standard program, periodic samples were taken at fortnightly or monthly time intervals during summer and autumn periods, and once or twice a week from November to June. The second set lets us create complex programs for flood sampling applications. We adopted a sampling strategy triggered by water level changes during the rising limb of hydrograph and flow rates during the flood recession. By using this program during floods, with some exceptions, the time intervals varied from 15 min to 2 h over the rising limb of hydrograph and from 2 h to 1 day over the flood recession. The total number of samples therefore varied between flood events. Twenty-one flood events occurred in the study period; the majority of the peak flows were sampled for SSC except for four events during which some technical problems caused the interruption of samplings. All water samples were collected by the automatic sampler. The samples were taken at random times during the day in normal and low flow conditions. The sampler intake nozzle was positioned vertically to the flow at the center of the cross section (Fig. 2) and was submerged following the US Geological Service manual (Edwards and Glysson, 1999). We verified that within the river the lateral diffusion was sufficient to ensure homogeneous lateral mixing of suspended sediment. Total suspended solid concentrations (SSCs) were determined using the APAT-IRSA analytical standard methods (APAT-IRSA-CNR, 2003). Standard 0.45 μm pore diameter cellulose filters were used to filter the samples. 3.2. Data analysis The relationship between SSC and streamflow was investigated and then loads under different flow conditions were evaluated. We also
Fig. 2. Cross section of the Celone river, Masseria Pirro gauging station. Schematic view from downstream to upstream.
137
analyzed hydrological and rainfall data recorded at two gauging stations located in the mountainous and in the plain area, respectively (Fig. 1). Annual rainfall data recorded over the study period were 43% and 2.4% higher than available historical data1 (Table 1). While, the most relevant events in terms of peak discharge are comparable to the major events recorded in the past at the S. Vincenzo gauge (recurrence interval of 30 years), an old station which was located 8 km downstream the new one (Masseria Pirro). Hence, we can say that the study period is a representative of historical hydrologic conditions in the basin. The load passing through a river cross-section during a time interval can be mathematically expressed by: Zt2 L¼
Q t C t dt
ð1Þ
t1
where Qt is the streamflow (L s− 1) at time t, Ct is the total suspended solid (TSS) concentration (mg L − 1) at time t (s), and L is the load (mg). The most common method used to calculate flux is a regression equation that defines an empirical relationship, which relates SSC to streamflow at the time of sampling (suspended sediment curve). Generally, this is a power function derived by a log-transformed leastsquares regression (linear, Eq. (2); or second-order polynomial) where the discharge (Q) is the independent variable (Phillips et al., 1999) and SSC (C) is the dependent variable. logC ¼ loga þ b logQ:
ð2Þ
The final result of the log-transformed regression method is often required in original units and the retransformation leads to a bias correction problem (Ferguson, 1986). Koch and Smillie (1986) pointed out that the mean value on the original scale is underestimated if the log-result is simply back-transformed to the original scale by exponentiation. This observation derived from the fact that the backtransformed mean value from the log scale is equal to the geometric mean2 of the values on the original scale. The geometric mean is always less than the arithmetic mean and, thus, the back-transformed mean always underestimates the arithmetic mean from the original scale. Several attempts were made to improve the C–Q relationships by using a bias correction factor (Bradu and Mundlak, 1970; Duan, 1983; Gilroy et al., 1990; Koch and Smille, 1986) or using subsets of data (seasonal, or on specific stage of flow) instead of whole available data set (Asselman, 2000; Walling and Webb, 1996). In this study, we have determined three regression equations (power curve, Log transformed linear and polynomial function) for the entire data set (210 samples) and for three subsets of data defined as a function of hydrological conditions: high, normal and low flows. The flow duration curve (FDC) was used to select the thresholds between three different flow classes: high, normal and low flows. Streamflow values higher than 0.930 m3 s− 1 are defined as high flow, corresponding to the interval of flow frequency exceedance of 0–20%, discharges ranged from 0.930 to 0.140 m3 s− 1 are defined as normal flow (20–70% frequency exceedance) and streamflow values less than 0.140 m3 s− 1 are defined as low flow. In addition, for the regression equations in logarithmic space (linear and polynomial), we also defined the bias correction of the retransformation proposed by Duan (1983) which does not require any assumption about the distribution of residuals. The latter, also called
1 Many gaps were found in the historical data especially in correspondence of high and intense events. 2 The geometric mean is defined as the nth root of the product of the n values.
138
A.M. De Girolamo et al. / Catena 128 (2015) 135–143
4. Results
Table 1 Basic hydrological data recorded in the catchment. Time period
2010–2011 1960–1996 a b c d e
Rainfall (mm y−1)
Rainfall (mm y−1)
Dischargee (mm day−1)
Faeto
Troia
Min
Max
Mean
1140 792a ± 135b
638 623a ± 110b
0.005c 0.000d
19.810 19.230
0.960 0.417
Mean value. Standard deviation. Masseria Pirro gauge (drainage area 72 km2). S. Vincenzo gauge (drainage area 86 km2). Mean annual flows normalized to the corresponding drainage area.
Smearing Estimator (CF), was applied to the back-log transformed data using the following equation: N X
CF1 ¼
εi
10
1
N
ð3Þ
where: ′ εi ¼ logðC i Þ− log C i is the residual and N is the number of observations. The correlation between contemporaneous SSC and streamflow values was determined for the entire data set, and for the three subsets of data defined above. For each type of curve, the mean error (E) was evaluated as percentage of the differences between measured concentration (Ci) and predicted values from regression (C′i) (Horowitz, 2003; Eq. (4)): 13 2 0 ′ N C i −C i X A7 6 @ 7 6 Ci 7 6 1 7 100: 6 Eð%Þ ¼ 6 7 N 7 6 5 4
ð4Þ
The C–Q relationships (rating curves) were used to evaluate SSC at a 15-minute time interval. Integration on daily time scale of the 15minute discharge and the SSC provided estimated daily loads, which were then aggregated on different time scales: year, season, and month. For each day, the load was calculated using the following expression (Eq. (5)):
Ldaily ¼
96 X
900 Q C
ð5Þ
1
where Ldaily is the daily load (kg), Q is the measured streamflow (m3 s−1), and C is the predicted SSC (g L−1) and 900 is the number of seconds in a time interval (15 min).
4.1. Suspended sediment concentration variability The hydrological regime of the Celone river plays an important role in the suspended sediment transport. Fig. 3 illustrates the variation in water discharge and SSC, from July 2010 to July 2011. The graph shows that equal peaks of discharge may have different magnitudes of SSC, for example, the peaks on 30 November (10.83 m3 s−1; 3.920 g L−1) and 16 May (10.44 m3 s−1; 0.810 g L−1). This behavior is due to the effect of vegetation in addition to a possible variation in rainfall pattern in combination with the different antecedent soil moisture contents. In November, wheat fields are unprotected while in May crops evenly cover the soil. The maximum sediment concentration recorded in the wet season was 7.13 g L−1, associated with a streamflow value of 23.5 m3 s− 1 (5th March 2011). During the summer period (June and July), two flash flood events were recorded (Fig. 3, events: 1 and 2). Each of these occurred over a period of a few hours during which water carried a huge amount of suspended material and debris. The first recorded flash flood occurred in July, after a dry period of approximately 50 days, was significant in terms of concentration (4.77 g L− 1) although the contemporaneous streamflow value (0.955 m3 s−1) is not of great magnitude. A more relevant event took place on 21st July 2010, when particularly heavy rainfall was recorded (73.6 mm in an hour at Faeto gauging station), this value being much higher than the monthly average recorded in July from 1960 to 2000 (32 mm), corresponding to the monthly rainfall of a winter month (February, March). Consequently, in 15 min the streamflow passed from 0.1 m3 s− 1 to 4.35 m3 s− 1 and in the same time interval the SSC increased from 0.069 g L−1 to 16.87 g L−1. The maximum value of the measured SSC was 37.60 g L−1 (recorded on 21st July 2010), associated with a streamflow value of 5.95 m3 s−1, corresponding to the highest value of the ratio SSC/Q in the study period. The ability of the basin to provide flows of various magnitudes and the percentage of time river flow exceeds a specified value was expressed by the flow duration curve (FDC). Fig. 4 shows SSC and the FDC for the study period; the outliers of SSC represent the values recorded during the flash flood events. The shape of FDC in its upper and lower regions is a very steep curve, indicating rain-caused floods and an intermittent regime. The river, at the Masseria Pirro gauging station, shows a very rapid rising stage and a short lag time (time between peak rainfall and peak discharge). Fig. 5 illustrates hydrographs of the recorded floods denoted by an asterisk in Fig. 3. Flood duration is typically only a few hours during which the SSC, as well as streamflow, increases and decreases rapidly. From November to March, the peak of the SSC generally precedes the flow peak (8 events) and the delay time was estimated as an hour. While, in late spring and in summer months, peaks of discharge and SSC were nearly synchronous. After the discharge reaches its flow peak, it falls off gradually resulting in a very slow recession with a duration depending on the antecedent soil moisture, and on the intensity and time interval of the rainfall that produced it. As Fig. 5 shows, in a succession of flood events there is no evident reduction of the sediment concentrations in the successive peak (24–25 December, 5–6 March, 30–31 March).
Fig. 3. Suspended sediment concentration and streamflow, plotted as a function of time. Recorded floods are denoted by a number. Floods denoted by an asterisk are illustrated in Figs. 5 and 6.
A.M. De Girolamo et al. / Catena 128 (2015) 135–143
139
0.11 g L−1 (2.33 m3 s−1). Many kinds of hysteresis loop were found in the wet season when multiple-peaks are quite common. 4.2. Fitted rating curve
Fig. 4. Flow duration curve.
The plot of the SSC against discharge often takes the form of a hysteresis loop (Fig. 6). Clockwise, anticlockwise and mixed-shaped loops were observed in the Celone river in the study period. A clockwise response was found in the first peak of multiple-peak events and when the events are caused by high intensity rainfalls (18 December; 15 mm in 3 h). When rainfall intensity is high runoff generation is mainly due to infiltration excess (Kirkby et al., 2011) and flow energy can increase rapidly causing the flushing of sediment previously accumulated on the river bed and a dilution or the exhaustion of sediment supply towards the end (Alexandrov and Laronne, 2003). Consequently, during the rising limb the SSC can be higher than that recorded at equivalent discharge in the recession limb with differences that can be relevant. For instance, on 18th December, SSC is 1.92 g L−1 in the rising limb (corresponding discharge 2.15 m3 s− 1), while in the falling limb it is only
After excluding 7 outlier points, which were the SSC values recorded during two flash floods, the correlation between the streamflow and SSC was found to be 0.81. The F-test showed this to be statistically significant (p b 0.01). The corresponding coefficient of determination (R2 = 0.66) indicates that 66% of variance of SSC may be explained by streamflow while 34% is justified by other factors such as rainfall intensity and its spatial distribution, vegetation cover and soil management. This result can be considered quite good for a regression method to be carefully applied (Quilbé et al., 2006). After the subdivision into discharge classes, the correlation coefficient decreases, taking the values: 0.78, 0.22 and 0.10, for “high”, “normal” and “low” flows, respectively. Fig. 7 shows values for SSC and Q for the all data sets; in the graph outlier points were removed. A high variability in SSC was recorded: SSC ranges from 0.005 to 4.77 g L−1 in normal flow, and from 0.005 to 0.360 g L−1 in low flow conditions. It should be noted that each class of flow includes both the steady flow and the recession limb of floods, which can record very different SSC values. Fig. 8 represents the recorded data and the second order polynomial recession curve fitted to the whole data set. Table 2 summarizes the equations and their characteristics and the correction factors calculated as stated before. 4.3. Assessment of suspended sediment loads at different time scales Among the rating curves defined in Table 2, we selected the equation with minimum mean error between the measured and estimated concentrations (58% in Table 3) which is a Log transformed polynomial
Fig. 5. Measured streamflow and suspended sediment concentrations during some flood events at the Celone Masseria Pirro gauging station.
140
A.M. De Girolamo et al. / Catena 128 (2015) 135–143
Fig. 6. Typical hysteresis effect observed in suspended sediment during some flood events at the Celone river (Masseria Pirro gauging station). Flood recorded on 3rd May follows an event occurred on 1st May (not monitored). Arrows indicate direction through time.
curve for all data sets (Eq. (8), in Table 2). By using this equation suspended sediment yields were evaluated at 31,265 and 19,064 t, with and without the smearing correction, respectively. The corresponding specific suspended sediment yields are 434 and
Fig. 7. Suspended sediment concentration versus discharge for the study area (7 outlier points were excluded).
Fig. 8. Sediment rating curve. Second-order polynomial recession curve for all data sets.
A.M. De Girolamo et al. / Catena 128 (2015) 135–143
141
Table 2 Sediment rating curves. Regression coefficients (a, b, c), number of data points (N), correction factor (CF), streamflow (Q). Log transformed data (linear) (Eq. (7)) logC = alogQ + b
Power curve (Eq. (6)) C = aQb
All (N = 210) High flow (N = 153) Normal flow (N = 63) Low flow (N = 10)
Log transformed data (polyn.) (Eq. (8)) logC = a(logQ)2 + blogQ + c
a
b
R2
a
b
R2
CF
a
b
c
R2
CF
0.1994 0.1533 0.3093 0.744
0.938 1.141 1.255 1.106
0.61 0.52 0.16 0.19
0.938 1.141 1.051 0.845
−0.700 −0.815 −0.643 −0.420
0.61 0.52 0.13 0.07
1.69 1.37 3.68 2.60
0.223 0.118 −1.344 −16.63
0.905 1.000 −0.060 −34.960
−0.779 −0.785 −0.792 −19.278
0.63 0.52 0.14 0.36
1.64 1.37 3.54 2.60
265 t km−2 y−1. However, when Eq. (8) for all data sets was used with the smearing factor, the mean error was found 158%. As expected, most of the suspended sediment yield is transported during the high flow status. In particular, the high flow is estimated to account for 94% of the total amount suspended material, while less than 0.1% of the suspended material is transported at low flow. It is interesting to point out that the high flow status has been recorded as occurring in 67 days, while normal flow and low flow in 186 and 112 days, respectively. On a monthly basis, more than 90% of the total annual suspended load is transported from November to May while in June the suspended sediment yield is less than 100 t per month, and decreases from August to October when it reaches the minimum at less than 10 t per month. The seasonal and monthly distributions of SS load follow the general pattern of the water yield (Fig. 9). A calculation of the annual load was also done by using a different equation for each class of flow. Even in this case, we selected the rating curves on the basis of the minimum mean error between measured and estimated sediment concentrations (Table 3). In details, we used for the three classes of flow (high, normal and low) the equations Eq. (8), Eq. (7), Eq. (7) (Table 2), respectively. The corresponding annual loads were estimated at 27,675 and 18,019 t y−1, with and without correction factor (CF), respectively. The corresponding specific suspended sediment yields are 384 and 250 t km−2 y−1. The mean error between measured and estimated loads is 63% and when CF is applied the results indicate an increase in mean error (263%). A comparative analysis of measured and generated suspended loads for flood events shows that actual loads are underestimated. The difference between the measured and estimated values is relevant for the major floods. For instance, in March when floods with high magnitude occurred, measured sediment load was 2898 t for event 9* (Fig. 3) and 8904 t for event 10*, while estimated values were 2423 t and 8002 t, respectively. In contrast, the loads are overestimated if the CF is used. In particular, estimated suspended sediment loads were found 3320 t and 10,962 t, respectively. As most of the suspended material is transported during floods, we can say that the actual annual loads are underestimated using suspended sediment rating curves. 5. Discussion
varying between 0.007 and 7249.792 t (i.e., 9.9 ∗ 10− 7 and 1.00 t km−2, respectively). The variability of concentrations during floods shows a hysteresis behavior similar to that observed in other Mediterranean catchments (Oeurng et al., 2010). Clockwise, anticlockwise and mixed-shaped loops were observed in the Celone river in the study period. A clockwise hysteresis is due to a progressive decline in suspended sediment availability (Gellis, 2012) that was found in events caused by high intensity rainfalls and in flood events away from previous others. This behavior reflects an exhaustion of suspended sediment supply, which occurs when one of the sources of eroded material is the channel itself. We consider sediment accumulated along the channel during the normal flow and low flow conditions likely to be remobilized during the rising limb of hydrograph with a decreasing sediment availability in the recession limb. Anticlockwise loop take place when suspended sediment sources are spread all over the catchment and it derives from processes whose dynamics are slower than the streamflow, i.e. contribution of sediment from distant areas and bank collapse may increase SSC in the recession limb. In the wet season, when multiple-peaks are quite common, many kinds of hysteresis loop were found. This behavior reveals the complexity of the suspended sediment dynamics which are controlled by several factors related to sediment sources, storm intensity, soil conditions (antecedent soil water content), vegetation density and tillage works. In contrast with other studies conducted in the Mediterranean region (Rovira and Batalla, 2006), that have recorded a progressive reduction of SSC within multiple-peak events, the analysis of SSC in the study area shows that the magnitude of peak discharges exerts a primary control on suspended sediment concentration. In the light of these results, we can say that there is no exhaustion of suspended sediment availability in the basin because different sediment sources contribute to suspended sediment material during floods: sheet wash, concentrated water erosion (gullies and rills) and channel bank erosion. The highest values of the SSC/Q ratio were recorded during the flash floods, which occur in summer after a dry period. This data reveals the potential threat for waters due to an excessive discharge of suspended sediment, and of a wide variety of organic and inorganic chemical constituents associated with sediment, which can degrade river habitat (Horowitz, 2008). Ribarova et al. (2008) pointed out that a high concentration of suspended sediment and nutrients during flash floods is very common in temporary rivers where a high level of pollutant accumulated
In common with results already reported by other studies conducted in the Mediterranean region (Gentile et al., 2010; López-Tarazón et al., 2010), the Celone suspended sediment transport response is highly variable over time. SSC vary from b0.005 to 37.60 g L−1 and daily loads Table 3 Mean error between measured and estimated sediment concentrations for the different types of rating curves. Type of rating curve
All data High flow Normal flow Low flow
Mean error (%) Eq. (6)
Eq. (7)
Eq. (8)
60 38 156 104
60 34 128 104
58 34 131 123
Fig. 9. Monthly suspended sediment load and water yield estimated from July 2010 to June 2011.
142
A.M. De Girolamo et al. / Catena 128 (2015) 135–143
along the river network is remobilized during these events (Chu et al., 2008). A previous research conducted in the study area confirmed a very high level of nutrients during flash flood events (De Girolamo et al., 2012). Alexandrov et al. (2003) observed that in the Mediterranean region data concerning flash floods are few and, generally, information about suspended sediment concentrations has been obtained mainly by manual sampling and often random. In this context, the present investigation gives a contribution to improve the knowledge about the sediment transport during flash floods in temporary rivers. With regard to the sediment rating curves, it must be noted that all the relationships presented here underestimate SSC in extreme high flow conditions while overestimate SSC in extreme low flow conditions. The resulting loads estimated using these equations underestimate the actual load. This result was expected, as the log-results were simply back-transformed to the original scale by exponentiation. In this case, the bias correction procedure developed by Duan (CF), applied to reduce error in the use of log-transformed regression to derive rating relation, fails to provide a more reliable estimate of suspended sediment yield. Moreover, the actual mean annual sediment yield is always significantly overestimated with the correction factor. These results confirm the studies by Kellerhals Engineering Services (1985) and Walling and Webb (1988). They reported that the bias associated with the logarithmic transformation is not the major cause of error in sediment rating curve estimations. Important factors are the large scatter associated with the plot of data and the hysteretic and exhaustion effects during storm events. However, we can assume that the actual value of specific yield is included in the range from 250 to 384 t km−2 y−1 (which are the values evaluated without and with CF, respectively). These results are in good agreement with the findings of Van Rompaey et al. (2005) which apply to mountain catchments with a Mediterranean climate in southern Italy (200–400 t km−2 y−1). It is well known that soil formation rates are variable but generally very low. Alexander (1988) measured a rate of soil formation ranging from 0.02 to 1.90 t km−2 y−1 from small watersheds under forest and grassland. Doran (1996) observed that 100–400 years are needed for developing a centimeter of topsoil, while for conditions which are prevalent in Europe, Verheijen et al. (2009) found a rate of soil formation of 140 t km−2 y− 1 (0.056 mm y− 1). Based on our calculations of suspended sediment load, we estimated that in the Celone basin the soil loss per year is N 0.1 mm, this means that soil is being lost 2–3 times faster than the rate of renewal and sustainability. Hence, we can say that soil erosion is effectively irreversible with potential high economic and environmental impacts. 6. Conclusions In unraveling the complexity of the sediment issue, the present work gives a contribution in increasing the knowledge of the processes concerned, which in turn is fundamental to integrating sediment and water quality into the river basin management. In fact, only a holistic approach can guarantee a high level of environmental quality, as required by the EU Environmental Policy. The temporal dynamic of suspended sediment transport in the Celone catchment showed a high variability: intra-annual, monthly and within-events. Suspended sediment load is mainly transported in winter and spring, when floods with high magnitude occurred, vegetation cover is absent or very low and tillage operations are performed. Over time, the variability of discharge and SSC resulted in different hysteric loops. Clockwise shapes mainly occurred in isolated floods when the sediments accumulated in the channel are removed with runoff; mixed shaped loops were found in multiple-peaks. All the methodologies for sediment yield evaluation show some limits due to the limitations in the model representations of natural processes characterized by heterogeneity. These difficulties increase in the Mediterranean basins where rivers have a temporary character and
where rainfall events are highly variable. The rating curves proposed here, developed as a function of hydrological conditions, suggest that this approach is good enough to generate SS concentration data for high flows, although it tends to underestimate the highest values. This method is not appropriate for low flow because of high variability in the data. However, it is important to note that low flow has a limited influence on the annual load of less than 0.1%. Hence, a better solution to estimate the SS flux could be to use a suspended sediment rating curve for high and normal flows and an averaging method for the low flow. Despite their limits, these relationships can be a very useful tool when a suspended sediment flux determination is required at a high temporal resolution (i.e. Total Maximum Daily Load methodology which needs SS load at a daily time scale). In addition, suspended sediment curves offer the possibility of reconstructing long term sediment transport data, which are fundamental to evaluate reservoir sedimentation. Hence, this technique may have important implications for the Capaccio reservoir management, providing a quick and inexpensive tool to assess the reduction of its storage capacity due to the sedimentation processes. Sediment rating curves, in fact, once defined can be used to evaluate suspended material transported by the river in different flow conditions by using only streamflow measurements that are more easier to measure than the sediment concentrations. Further sampling campaigns are needed in order to validate the sediment rating curves defined here, while an analysis of the data already collected can be useful for defining a new monitoring program (i.e. identifying the time period during which only sporadic samplings are sufficient to determine the SSC and loads and point out the critical time period during which continuous samplings are needed). Finally, further studies should investigate and quantify the uncertainty linked to load estimation. Acknowledgments This research was supported by the European Community's Seventh Framework Programme, FP7/2007–2011, MIRAGE Project (Contract no.: 211732). The authors gratefully acknowledge Giuseppe Amoruso from the Apulia Civil Protection Service for providing the data. The authors are indebted to Angelantonio Calabrese who collaborates in the field activities and analytical determinations of SSC. We thank David Cooper for his helpful comment to improve the quality of the paper. Thanks are also due to Stefano Pirro for helping us in identifying the best location for installing the equipment, to Giuseppe and Nicola Pirro for controlling it. Lastly, we would like to thank the reviewers for their valuable scientific comments and recommendations. Appendix A. Supplementary data Supplementary data associated with this article can be found in the online version, at http://dx.doi.org/10.1016/j.catena.2014.09.020. These data include Google maps of the most important areas described in this article. References Alexander, E.B., 1988. Rates of soil formation: implications for soil loss tolerance. Soil Sci. 145 (1), 37–45. Alexandrov, Y., Laronne, J.B., 2003. Suspended sediment transport in flash floods of the semiarid northern Negev, Israel. Proc. Int. Symp. Hydrology of the Mediterranean and Semiarid Regions. IAHS Publ. 278, Montpellier, pp. 346–352. Alexandrov, Y., Laronne, J.B., Reid, L., 2003. Suspended sediment concentration and its variation with water discharge in dryland ephemeral channel, northern Negev, Israel. J. Arid Environ. 53, 73–84. APAT-IRSA-CNR, 2003. Metodi analitici per le acque, APAT Manuali e Linee Guida 29/ 2003. pp. 1–1153. Arnold, J.G., Srinivasan, R., Muttiah, R.S., Williams, J.R., 1998. Large area hydrologic modelling and assessment part I: model development. J. Am. Water Resour. Assoc. 34 (1), 73–89. Asselman, N., 2000. Fitting and interpretation of sediment rating curves. J. Hydrol. 234, 228–248.
A.M. De Girolamo et al. / Catena 128 (2015) 135–143 Bradu, D., Mundlak, Y., 1970. Estimation in lognormal models. J. Am. Stat. Assoc. 65, 198–211. Chu, Y., Salle, C., Cernesson, F., Perrin, J.L., Tournoud, M.G., 2008. Nutrient load modelling during floods in intermittent rivers: an operational approach. Environ. Model Softw. 23, 768–781. De Girolamo, A.M., Lo Porto, A., 2012. Land use scenario development as a tool for watershed management within the Rio Mannu Basin. Land Use Policy 29, 691–701. De Girolamo, A.M., Calabrese, A., Pappagallo, G., Santese, G., Lo Porto, A., 2012. Impact of anthropogenic activities on a temporary river. Fresenius Environ. Bull. 21 (11), 3278–3286. De Girolamo, A.M., Lo Porto, A., Pappagallo, G., Tzoraki, O., Gallart, F., 2014. The hydrological status concept. application at a temporary river (Candelaro, Italy). River Res. Appl. http://dx.doi.org/10.1002/rra.2786. Doran, J., 1996. The international situation and criteria for indicators, chap 7. In: Cameron, K.C., Comforth, I.S., McLaren, R.G., Beare, M.H., Basher, L.R., Metherell, A.K., Kerr, L.E. (Eds.), Soil Quality Indicators for Sustainable Agriculture in New Zealand: Proceedings of a Workshop. Lincoln Soil Quality Research Centre, Lincoln University, New Zealand. Duan, N., 1983. Smearing estimate: a nonparametric retransformation method. J. Am. Stat. Soc. 78, 605–610. Duvert, C., Nord, G., Gratiot, N., Navratil, O., Nadal-Romero, E., Mathys, N., Némery, J., Regüés, D., García-Ruiz, J.M., Gallart, F., Esteves, M., 2012. Towards prediction of suspended sediment yield from peak discharge in small erodible mountainous catchments (0.45–22 km2) of France, Mexico and Spain. J. Hydrol. 454–455, 42–55. EC, 2006. A strategy to keep Europe's soils robust and healthy. http://ec.europa.eu/ environment/soil/index.htm. EC Directive 2000/60/EC of the European Parliament and the Council Directive establishing a framework for Community action in the field of water policy, Official Journal of the European Communities, Brussels, 22/12/2000. Edwards, T.K., Glysson, G.D., 1999. Field methods for measurement of fluvial sediment. USGS Open-file Report Book 3 (Chapter 2), pp. 1–97. Kellerhals Engineering Services, 1985. Sediment in the Pacific and Yukon region: review and assessment. Water Resources Branch, Inland Waters Directorate Environment Canada, Report IWD-HQWRB-SS-85-8. Ferguson, R.I., 1986. River loads underestimated by rating curves. Water Resour. Res. 21, 74–76. Gellis, A.C., 2012. Factors influencing storm-generated suspended-sediment concentrations and loads in four basins of contrasting land use, humid-tropical Puerto Rico. Catena 104, 35–57. Gentile, F., Bisantino, T., Corbino, R., Milillo, F., Romano, G., Trisorio, Liuzzi G., 2010. Monitoring and analysis of suspended sediment transport dynamics in the Carapelle torrent (Southern Italy). Catena 80, 1–8. Gilroy, E.J., Hirsh, R.M., Cohn, T.A., 1990. Mean square error of regression-based constituent transport estimates. Water Resour. Res. 26 (9), 2069–2077. Gomez, J.A., Sobrinho, T.A., Giraldez, J.V., Fereres, E., 2009. Soil management effects on runoff, erosion and soil properties in an olive grove of Southern Spain. Soil Tillage Res. 102, 5–13. Horowitz, A.J., 2003. An evaluation of sediment rating curves for estimating suspended sediment concentrations for subsequent flux calculations. Hydrol. Process. 17, 3387–3409. Horowitz, A.J., 2008. Determining annual suspended sediment and sediment-associated trace element and nutrient fluxes. Sci. Total Environ. 400 (1–3), 315–343. Jones, A., Panagos, P., Barcelo, S., Bouraoui, F., Bosco, C., Dewitte, O., Gardi, C., Erhard, M., Hervás, J., Hiederer, R., Jeffery, S., Lükewille, A., Marmo, L., Montanarella, L., Olazábal, C., Petersen, J.E., Penizek, V., Strassburger, T., Tóth, G., Van Den Eeckhaut, M., Van Liedekerke, M., Verheijen, F., Viestova, E., Yigini, Y., 2012. The state of soil in Europe. JRC Reference Report 68418. 1831-9424 http://dx.doi.org/10.2788/77361. Kirkby, M.J., Le Bissonais, Y., Coulthard, T.J., Daroussin, J., McMahon, M.D., 2000. The development of land quality indicators for soil degradation by water erosion. Agric. Ecosyst. Environ. 81, 125–136. Kirkby, M.J., Gallart, F., Kjeldsen, T.R., Irvine, B.J., Froebrich, J., Lo Porto, A., De Girolamo, A., the MIRAGE team, 2011. Classifying low flow hydrological regimes at a regional scale. Hydrol. Earth Syst. Sci. 15, 3741–3750. Koch, R.W., Smille, G.M., 1986. Bias in hydrologic prediction using log-transformed regression models. Water Resour. Bull. 22 (5), 717–723.
143
Koch, R.W., Smillie, G.M., 1986. Comment on “river loads underestimated by rating curves” by R.I. Ferguson. Water Resour. Res. 22 (13), 2121–2122. Larsen, M.C., Gellis, A.C., Glysson, G.D., Gray, J.R., Horowitz, A.J., 2010. Fluvial sediment in the environment: a national problem. Proceedings, 2nd Joint Federal Interagency Conference, Las Vegas, NV. June 27–July 1 (15 pp.). Lenzi, M., Marchi, L., 2000. Suspended sediment load during floods in a small stream of the Dolomites (northeastern Italy). Catena 39, 267–282. Letcher, R.A., Jakeman, A.J., Merritt, W.S., McKee, L.J., Eyre, B.D., Baginska, B., 1999. Review of techniques to estimate catchment exports. Report EPA 99/73, Environment Protection Authority, Sydney, Australia. Letcher, R.A., Jakeman, A.J., Calfas, M., Linforth, S., Baginska, B., Lawrence, I., 2002. A comparison of catchment water quality models and direct estimation techniques. Environ. Model Softw. 17, 77–85. López-Tarazón, J.A., Batalla, R.J., Vericat, D., Balasch, J.C., 2010. Rainfall, runoff and sediment transport relations in a mesoscale mountainous catchment: the River Isábena (Ebro basin). Catena 82, 23–34. Moatar, F., Meybeck, M., 2005. Compared performances of different algorithms for estimating annual nutrient loads discharged by the eutrophic River Loire. Hydrol. Process. 19 (2), 429–444. Morgan, R.P.C., 2005. Soil Erosion and Conservation, 3rd. Blackwell, Oxford, p. 314. Navratil, O., Esteves, M., Legout, C., Gratiot, N., Némery, J., Willmore, S., Grangeon, T., 2011. Global uncertainty analysis of suspended sediment monitoring using turbidimeter in a small mountainous river catchment. J. Hydrol. 398, 246–259. Nikolaidis, N.P., Demetropoulou, L., Froebrich, J., Jacobs, C., Gallart, F., Prat, N., Lo Porto, A., Campana, C., Papadoulakis, V., Skoulikidis, N., Davy, T., Bidoglio, G., Bouraoui, F., Kirby, M., Tournoud, M.G., Polesello, S., Barberá, G.G., Cooper, D., Gomez, R., SanchezMontoya, M.M., Latron, J., De Girolamo, A.M., Perrin, J.L., 2013. Towards a sustainable management of Mediterranean river basins. Policy recommendations on management aspects of temporary river basins. Water Policy 15, 830–849. Oeurng, C., Sauvage, S., Sánchez-Pérez, J.M., 2010. Dynamics of suspended sediment transport and yield in a large agricultural catchment, southwest France. Earth Surf. Process. Landf. 35, 1289–1301. Pavanelli, D., Cavazza, C., 2010. River suspended sediment control through riparian vegetation: a method to detect the functionality of riparian vegetation. Clean Soil Air Water 38 (11), 1039–1046. Phillips, J.M., Webb, B.W., Walling, D.E., Leeks, G.J.L., 1999. Estimating the suspended sediment loads of rivers in the LOIS study area using infrequent samples. Hydrol. Process. 13, 1035–1050. Quilbé, R., Rousseau, A.N., Duchemin, M., Poulin, A., Gangbazo, G., Villeneuve, J.P., 2006. Selecting a calculation method to estimate sediment and nutrient loads in streams: application to the Beaurivage River (Québec, Canada). J. Hydrol. 326 (1–4), 295–310. Ribarova, I., Ninov, P., Cooper, D., 2008. Modeling nutrient pollution during a first flood event using HSPF software: Iskar River case study, Bulgaria. Ecol. Model. 211, 241–246. Rodriguez-Blanco, M.L., Taboada-Castro, M.M., Taboada-Castro, M.T., 2010. Sources and sediment yield from a rural catchment in humid temperate environment, northwest Spain. Earth Surf. Process. Landf. 35, 272–277. Rovira, A., Batalla, R.J., 2006. Temporal distribution of suspended sediment transport in a Mediterranean basin: the Lower Tordera (NE SPAIN). Geomorphology 79, 58–71. Soler, M., Latron, J., Gallart, F., 2008. Relationship between suspended sediment concentrations and discharge in two small research basins in a mountainous Mediterranean area (Vallcebre, Eastern Pyrenees). Geomorphology 98, 143–152. Van Rompaey, A., Bazzoffi, P., Jones, R.J.A., Montanarella, L., 2005. Modeling sediment yields in Italian catchments. Geomorphology 65, 157–169. Verheijen, F.G.A., Rickson, R.J., Smith, C.J., 2009. Tolerable versus actual soil erosion rates in Europe. Earth Sci. Rev. 94, 23–38. Verstraeten, G., de Poesen, J., Vente, J., Koninckx, X., 2003. Sediment yield variability in Spain: a quantitative and semiqualitative analysis using reservoir sedimentation rates. Geomorphology 50 (4), 327–348. Walling, D.A., Webb, B.W., 1988. The reliability of rating curve estimates of suspended sediment yield: further comments. IAHS Publ. 174, 337–350. Walling, D.A., Webb, B.W., 1996. Erosion and sediment yield: a global overview. IAHS Publ. 236, 3–19.