Versão online: http://www.lneg.pt/iedt/unidades/16/paginas/26/30/125 Comunicações Geológicas (2012) 99, 1, 27-32 ISSN: 0873-948X; e-ISSN: 1647-581X

Statistical modelling of suspended sediment transport in the Cherf drainage basin,

Modelação estatística do transporte de sedimentos em suspensão na bacia de drenagem de Cherf, Argélia

K. Khanchoul1*, Z.A. Boukhrissa1 & H. Majour2

Recebido em 31/07/2011 / Aceite em 02/11/2011 Artigo original Disponível online em Janeiro de 2012 / Publicado em Junho de 2012 Original article © 2012 LNEG – Laboratório Nacional de Geologia e Energia IP

Abstract: The effect of suspended sediment transport on the drainage Achite, 2007; Cherif et al., 2009), in their review of suspended basin and Cherf Reservoir was evaluated through measurement and sediment transported by wadis in the Maghreb, underlined the lack modeling of inflowing fine-sediment loads. The daily suspended- of available data for such river types. Data from these systems is sediment concentration and daily water discharge data (period 1975- still fragmentary and further study of both the quantification of 1994) and the monthly outflows for the post-period of the dam construction are analyzed using regression methods. The methodology sediment transport in semiarid basins and its variability is clearly involved is developed by a conventional sediment rating curve and a required. multiple regression model. Over the 19-year period, the average annual Many river systems in the Algerian regions continue to sediment yield is equal to 350 T km-2. The annual sediment volume experience severe and uncontrolled environmental degradation. conducted at the Cherf Reservoir for the period 1995-2005 is 6.74 Mm3. This has resulted into enhanced soil erosion in the catchments, The results provided information for an improved understanding of soil thereby causing a range of problems from considerable loss of soil erosion and reservoir sedimentation in the study drainage basin. fertility and eventual desertification to accelerated river, reservoir Keywords: Cherf River, modelling, sediment load, drainage basin, sedimentation and flooding (Haregeweyn et al., 2006; Tamene et reservoir. al., 2006). In the case of Cherf River, the declining watershed resources have put considerable pressure on the agricultural land Resumo: O efeito do transporte de sedimentos suspensos na bacia de and reservoir to support households. drenagem e no reservatório de Cherf foi avaliado através da medição e modelação das cargas de sedimentos finos. Os dados de concentração The prevailing geomorphic conditions such as sparse forest diária de sedimentos em suspensão e da descarga diária de água (período cover and grassland leave soils vulnerable to erosion. However, as 1975-1994) e os fluxos mensais para o período posterior à construção da observed, where and how much is eroded and transported barragem são analisados usando modelos de regressão. A metodologia downstream largely depends also on topography, lithology, and envolvida é desenvolvida por uma curva convencional de seriação de soil hydrological properties. Once eroded, the sediment’s journey sedimentos e por um modelo de regressão múltiplo. Durante o período de -2 down the catchment depends significantly on runoff which, over 19 anos, a média anual de sedimentos é de 350 T km . O volume anual time, reduces in speed resulting in deposition (Fried et al., 2000). de sedimentos conduzido ao reservatório de Cherf para o período de 3 Considering the principles of river material extraction and 1995-2005 é 6.74 Mm . Os resultados fornecem informação para um conhecimento melhorado da erosão do solo e sedimentação do transported sediments by river flow in design of river structures, reservatório na bacia de drenagem estudada. study of various methods to predict river sediment transport rate Palavras-chave: Rio Cherf, modelação, carga sedimentar, bacia de seems to be necessary. Also, according to Leahy et al. (2008), drenagem, reservatório. river study is of great importance for reliable forecasting, but it is a difficult task due to the complexity and inherent nonlinearity of 1Department of Geology, Badji Mokhtar University-Annaba, B.P.12, Sidi Amar, hydrological systems. Consequently, various sediment yield 23000 Annaba, Algeria. models have been reviewed (e.g. Walling, 1988; Singh, 1989; 2 Laboratory of Geology, Badji Mokhtar University-Annaba, B.P.12, Sidi Amar, 23000 Annaba, Algeria. Mathur et al., 1992; Jansson, 1997; Khanchoul et al., 2007). These *Autor correspondente / Corresponding author: [email protected] models may be grouped as (a) empirical prediction models, (b) process-based prediction models, (c) dynamic simulation models, and (d) stochastic models. 1. Introduction The lack of information about the suspended sediment concentration can result in substantial errors in estimates of the In the Maghreb (North Africa), problems essentially relate to the total load. If the sampling is too infrequent, the storm events may silting up of the dams, which reduces the water storage capacity be missed, and, even with fairly intense sampling, it is difficult to and the hydraulic potential by 2–5% per year (Kassoul et al., sample a representative number of events. For this reason, a 1997). A correct estimation of the sediment volume carried by a longer record of sediment loading is needed to better define the river is therefore necessary for an adapted water resources natural condition and the response of sediment yield to human management strategy (Achite, 2007). Sediment transport is well disturbances. The historical hydrometric records of the Moulin documented in perennial rivers in temperate or humid climates but Rochefort gauging station might provide temporal and spatial is far less studied in semiarid zones. Some authors (Probst & records of sediment load, and offer revealing information about Amiotte-Suchet, 1992; Fox et al., 2001; Touaibia et al., 2001; sediment yielding processes and their controlling factors. 28 K. Khanchoul et al. / Comunicações Geológicas (2012) 99, 1, 27-32

Assessments of methods for estimating loads have been carried Because its rainfall is scanty and discontinuous, the mode of out using data from different sampled rivers, or using simulated the flows of the Cherf on an interannual, seasonal, and daily basis, data. One of these methods, widely used when and where is very irregular. The low water levels occur in general in June measurements are not available (Horowitz, 2003; Khanchoul & until September, where the streams diminish to trickles or go dry Jansson, 2008), uses empirical models. They are referred to rating in summer. Extreme streamflows are less frequent, which occur curves and are usually based on the linear relationship between usually in late autumn or early winter caused by rainfall events. suspended sediment concentration and water discharge. The mean annual water discharge (1975 to 1994) is equal to 0.59 Although some studies have focused on the water quality and m3/s and the highest monthly values of out-flows are recorded in river discharges in Cherf River, the sediment transport and its January and February, with mean values not exceeding 1.20 m3/s. impact on the Cherf dam has received scant attention. It is our During storm events, the hydraulic regime of the river becomes objective to explore the variability of water discharge and torrential, leading to high peak flows and massive sediment loads sediment load of the Cherf River, the upper fluvial system of either suspended or bed load. The aggressive streamflows are Seybouse River, and relate the interannual supply of sediment represented by the high drainage densities in two subbasins with yield to the geomorphic conditions of the drainage basin. The values of 4 and 11 km-1. assessment of the Cherf sediment yields is based on available The lithology of the Cherf drainage basin shown in Figure 2 water discharge and suspended sediment concentration and the age description have been realized from hardcopy observations (period of 1975/1976-1993/1994), and on the geologic maps in the scale of 1:50000 edited by the National application of a multivariate model to predict sediment load in the Geographic Institut and SONATRACH (geologic maps of reservoir that will serve to protect this structure from sediment (1977), Ain Babouche (1956), Jean Rigal (1977), and Ain Beida flux deposition. (1977)) and lithologic maps realized by B.N.E.D.E.R (maps of M’daourouch and Oued Kebarit (1998) in the scale of 1:50000). 2. Study Area For the most part, the basin is composed of quaternary deposits and calcareous crusts that represent as much as 73% of the total The maximum Cherf stream length is 56 km and drains a 1710 basin area. The rest of the lithologic formations are Aptian 2 km basin. It is one of the largest fluvial system in Seybouse and limestone, Mio-Pliocene conglomerate, Oligocene sandstone, originates from headwaters in the Ain Babouche Aptian Cretaceous marl and marly limestone, and some Triassic outcrops limestone at an elevation of 1635 m. From 1994, the Moulin composed of marl and clay (Fig. 2). Rochefort gauging station of this basin has disappeared because there has been a construction of a reservoir (Cherf Dam), with a capacity of 155 million cubic meters and an area of 1735 km2 (Fig. 1). This station, located upstream of the Cherf River and considered as an outlet of the study Cherf drainage basin, was controlled by the National Agency of Hydraulic Resources (ANRH) from where instantaneous water discharge and suspended sediment concentration data were collected. The climate over the basin is semi arid to arid, with dry summers and rainfalls concentrated mainly in fall and spring periods. The average annual rainfall is equal to 290 mm with an abundance rainfall occurring from March to May (mean values ranging from 30 to 41 mm), conversely to the dry season from June to August.

Fig.2 Lithology of the Cherf drainage basin.

Fig.2. Litologia da bacia de drenagem do Cherf.

In this basin, the basins and ranges form a succession of anticlines and syncline depressions. Several wadis drain wide open areas of the basin. To the east, the syncline which forms the plain of is crossed by the Crab River and its tributaries. The later river and the Hamimine River meet at south-west of Sedrata and give the Cherf River, which flows at the foot of the Hill to reach the gorge of Moulin Rochefort. The large plain of Sedrata is drained by two main rivers, first the Snob River that takes its source from the limestone outcrops of Tereguelt Hill and then the highly dissected Trouch River that starts from the Galaa, Bir El Akhal, and Tereguelt hills. Moreover, the plain of Tamlouka is drained by Settara River. This plain is a synclinal Fig.1. Location map of the Cherf drainage basin and drainage network presentation with: a - Settara River, b - Trouch River, c - Snob River, d - Hamimine River, e - basin developed between the altitude of 800 and 900 m. Crab River. In this study area most of the rivers, mainly Crab, Trouch, and Settara rivers, are adapted to the geologic structure with the Fig.1. Mapa de localização da bacia de drenagem do Cherf e rede de drenagem com: a – Rio Settara, b – Rio Trouch, c – Rio Snob, d – Rio Hamimine, e – Rio Crab. main arteries that are consistent with the synclinal axis and tributaries. Suspended sediment transport in Algeria 29

Slopes steeper than 3% represent only 24% of the basin area. about the suspended sediment transport can be extracted from the The steep slopes (over 10%) occupy the resistant rocks such as the following technique between sediment discharges and their water Reghia Mountain and the El Meida Mountain, and the moderately discharges. resistant rocks distributed mainly in the north-west and north-east and along the conglomerate formations of the upper part of Cherf River (Fig. 2). Land use is characterized by forest (Aleppo pine) associated with sparse shrubs and grasslands as steppe of Alfa (Stipatenacissima) that occupy 16% of the basin area. The forest is developed mainly on the limestone formations of Reghia Mountain and Galaa Hill, and conglomerate surfaces located east of Sedrata. Cultures and associated grasslands are the dominant land use cover of the basin. The frequent use of cultivation techniques such as up and down tillage triggers rill/ gully erosion and mass wasting processes that, in turn, can generate severe erosion forms such as badlands on sandy-clay slopes.

3. Materials and Methods

3.1. Data set Meteorological and hydrological data have been monitored routinely for years by the ANRH (National Agency of Hydraulic Resources), which operates the network of monitoring stations in Algeria. The Moulin Rochefort hydrometric dataset includes: (1) Fig.3 Suspended sediment discharge versus water discharge taken from the Moulin discrete data of suspended sediment concentration (C) and Rochefort gauging station in the Cherf drainage basin (period 1975/76-1993/94). corresponding water discharge (Q), (2) records of daily water Fig.3. Descarga de sedimentos suspensos contra descarga de água tira da estação de discharge, and (3) instantaneous water discharge (based on river Moulin Rouchefort na bacia de drenagem do Cherf (período 1975/76-1993/94). stage) during floods. Water samples were taken for measurement of suspended An attempt has also been made by subdividing the data set sediment concentration several times a year by ANRH, which into winter, spring, summer, and fall periods and by determining make 801 hourly data sets in 19-year period (1975-1994). The whether or not the plotted Qs-Q indicated a change in inclination gauging station performs the water sampling of surface water of an imagined line through the plotted values to make allowance using one-liter bottles, and the analysis of the samples. The water for influences on rating curve scatter. samples, which are taken in various conditions of stream flows, Optimization of rating curve method is validated by are filtered using a filter of Laurent type (= 32 cm). The filter comparing the predicted against observed values on scatter plots. and the mud contained in the bottle is weighed after drying in a When the least squares logarithmic regression procedure results special oven for 30 minutes at a temperature of 110°C, following in underestimation because of the bias given to the values below a standard procedure applied by many agencies throughout the the fitted line by the regression (Ferguson, 1986; Jansson, 1985, world (Khanchoul et al., 2007). Since there have been no 1997), a non parametric correction for the transformation bias measurements or studies of bed load in this basin, only suspended may be used. sediment is treated in this study. Data on monthly streamflow for To test the validation of the sediment rating curve technique, the Cherf reservoir are obtained from the National Agency of a possible mathematical association is tried with the independent Dams, but information is not available for that period on the variables. The model efficiency factor EF of observed and sediment transport component. predicted values are estimated for different predictions on validation datasets. The best model is selected based on the EF 3.2. Application of the sediment rating curve value approaching one. The model efficiency factor is estimated In the absence of actual suspended sediment concentration (SSC) for all the validation sets using the equation: measurements, hydrologists have used sediment rating curves to n ˆ predict suspended sediment concentrations for subsequent flux (Ci  Ci )² calculations, determining long-term suspended sediment loads. A EF = 1 - i1 (2) regression analysis is made between the daily suspended n sediment discharge (Qs) and the daily water discharge (Q). The (Ci  C )² most commonly used sediment rating curve is a power function: i1 Qs = aQb (1) where n is the total number of observations, Ci the ith observed value, the mean of observed values, and ˆ the ith predicted where a and b are regression coefficients. It has been C Ci demonstrated that improvement of the log transformation value. regression formulation can significantly reduce the bias introduced in the calculation (Cohn, 1995). 3.3. Multiple regression model This regression analysis has provided a moderate coefficient Available data on suspended sediment transported by rivers in the of correlation (r = 0.78) and the Qs-Q data show a scatter Maghreb, which have been summarized by Sogreah (1983), were distribution mainly at low and medium values, with an reviewed for 130 drainage basins (Probst & Amiotte Suchet, underestimation of sediment loads at high flows (Fig. 3). The 1992). The influences of several factors (precipitation (P) in mm, regression line has underestimated the true sediment load by 37% runoff (R) in mm, drainage area size (A) in km² and rock type when developed on all 149 daily data. Due to this (KER) on mechanical erosion and fluvial sediment transport were underestimation of sediment load, a further interesting prediction analysed. A multiple regression model was proposed to estimate 30 K. Khanchoul et al. / Comunicações Geológicas (2012) 99, 1, 27-32 the river sediment yields in the Maghreb. Rock type is a The performance of equation (3) in the prediction of Cherf qualitative variable, and each lithological formation has been sediment yield is used for the period (1975/76-1993/1994) prior ascribed a coefficient (KER) which represents rock erodibility to the reservoir construction, and a comparison is done between (Table 1). KER is the ratio between the mechanical denudation the multiple regression and the rating curve models. rate of a given rock and the denudation rate of granite which is considered to be the most resistant to mechanical erosion (Probst 4. Results and Discussion & Amiotte Suchet, 1992). 4.1. Estimating the suspended sediment yield through

Table 1. Rock erodibility coefficient (KER) calculated for the different lithological sediment rating curves formations. The sediment rating curves for all available and season Tabela 1. Coeficiente de erosão (KER) calculado para as diferentes litologias. suspended sediment discharge data are presented in Figure 4. The best-fit power function line through the data for the stratified single rating has given a low and moderate coefficients of correlation (r = 0.59 and 0.78), but the suspended sediment estimates at the high water discharges are accurately represented because the vast majority of the sediment is transported during these high discharge events (Fig. 4). The regression functions for the season ratings have provided coefficients of correlation ranging from 0.57 to 0.99 (Fig. 5).

By use of the lithologic map and field observations of the susceptibility of the rocks and soils to mechanical erosion, three groups are chosen for the study basin: strongly cohesive rocks (limestones and sandstones); moderate erodibility rocks that should get a subjective Ker equal to 27 (sandstone and clay, marly limestone, conglomerate and clay, Triassic rocks, Quaternary formations); and poorly cohesive rocks (marls). In Table 2, the lithologic formation and KER of each rock type for the study basin are represented. In this case, the mean weighted KER is equal to 21. Despite the fragility to erosion of the alluvial plain deposits, they have been assigned a coefficient of 10 because their surfaces are not really disturbed by the morphogenesis due to the permeable nature of soils and very gentle slopes.

Table 2. Rock erodibility coefficient (KER) of the Cherf drainage basin.

Tabela 2. Coeficiente de erosão (KER) da bacia de drenagem do Cherf. Fig.4. Stratified regression lines of sediment discharge versus water discharge taken from the Moulin Rochefort gauging station in the Cherf drainage basin (period 1975/76-1993/94).

Fig.4. Linhas de regressão estratificadas da descarga de sedimentos contra a descarga de água tirada da estação de Moulin Rochefort na bacia de drenagem do Cherf (período 1975/76-1993/94).

The error in the sum of loads by using the relationship between Qs and Q from 149 datasets and seasons with measurements, demonstrate that sediment loads calculated by using a stratified single rating curve and seasonal ratings have underestimated the load by only 3% and 1.28% after correction for bias (Table 3). Moreover, the computation of the EF has also provided information about the predictive capabilities of the model. The EF value is equal to 0.78 for the seasonal regression model and is lower for the stratified single ratings (EF= 0.64). From the different calculations, the use of the seasonal sediment The multiple regression equation for the predicted suspended rating curve has given satisfactory results. sediment yield (SY), in T km-2 y-1, has the following terms: Based on the seasonal sediment rating curve, the calculated -3 -3 -5 ln (SY) = 4.79 + (54 x10 KER) + (4 x 10 R) - (5.6 x 10 A) mean annual specific suspended sediment yield in the Cherf (3) drainage basin during the period 1975/1976-1993/1994 is equal The observed suspended sediment yields for the 130 basins to 350 T km-2 year-1, which corresponds to an annual sediment are compared with the predicted values obtained. The previous load of 11.36 x 106 tonnes. Suspended sediment loads for authors have stated that the cluster of points is relatively well 1978/1979 and 1989/1990 indicate that sediment production and centred around the line of regression, even if it is scattered. The transport are considerably higher with 3.41x106 tonnes, reflecting model therefore provides a reasonable explanation of the a much larger sediment supply which generated from somehow observed values. more surface runoff. Suspended sediment transport in Algeria 31

Table 3. Sediment load computation for the stratified single ratings and seasonal sediment rating curve method.

Tabela 3. Computação da carga sedimentar para as variações estratificadas e para o método das variações sazonais.

Furthermore, for this river basin that includes areas of highly erosive materials, sediment yield is controlled mostly by the magnitude and frequency of heavy rainfall events. The bulk of suspended sediment transport in the Cherf River took place mainly in the two first fall months and the two last spring months (Fig. 6). The monthly evolution of stream flows and yields do not coincide, where the highest sediment yields in the fall season correspond to low runoffs and the yield peak in April has a moderate flow value (Qmean = 0.87 m3/s). The four cited months represent almost 60% of the total annual sediment value. To these situations, the suspended sediment concentrations should be high because of the existence of favourable geomorphic conditions such as high rainfall intensity, less permanent vegetation, and more erodible lithologic formations.

Fig.6. Evolution of monthly mean water discharge and monthly sediment load according to data sets sampled at the Moulin Rochefort gauging station in the Cherf drainage basin (period 1975/76-1993/94).

Fig.6. Evolução da média mensal de descarga de água e da carga de sedimentos de acordo com os dados na estação de Moulin Rochefort na bacia de drenagem do Cherf (período 1975/76-1993/94).

4.2. Estimating the suspended sediment yield through a multiple regression model Before using the multiple regression model in the estimation of the sediment load entering the reservoir, a calculation is made with this model for the period 1975-1994 to assess the quality of the estimated sediment load. With a mean annual runoff of 2 Fig.5. Qs-Q relationships according to seasonal ratings based on data sets of the 11 mm, KER and basin area of respectively 21 and 1710 km , Moulin Rochefort gauging station in the Cherf drainage basin (period 1975/76- the estimated sediment yield (SY) is equal to 347.23 T km-2 1993/94). -1 year . In comparison with the rating curve method, equation Fig.5. Relação Qs-Q de acordo com as variações sazonais baseadas nos dados da (3) has underestimated the sediment yield by only 0.79%. The estação de Moulin Rochefort na bacia de drenagem do Cherf (período 1975/76- multiple regression equation has shown a low degree of error, 1993/94). and thereby the model is judged to be sufficiently reliable for the estimation of the sediment yield at the Cherf reservoir. 32 K. Khanchoul et al. / Comunicações Geológicas (2012) 99, 1, 27-32

Corresponding to the Cherf reservoir basin (1735 km2) for the Cherif, A., Errih, M., Cherif, H. M., 2009. Modélisation statistique du period 1994/1995 to 2004/2005, we have used a rock type transport solide du basin versant de l’Oued Mekerra (Algérie) en zone erodibility of 21 and a mean annual runoff equal to 19 mm. The semi-aride méditerranéenne. Hydrological Science Journal, 54, 2, mean annual suspended sediment yield (SY) that entered the 338–348. Cohn, T. A., 1995. Recent advances in statistical methods for the estimation of reservoir is estimated to 357.81 T km-2 year-1; and the total 6 sediment and nutrient transport in rivers. Reviews of Geophysics, 33, 1, suspended sediment load shows 6.83x 10 tonnes. 1117–1123. Based on some studies undertaken in Algerian reservoirs, we Ferguson, R. I., 1986. River loads underestimated by rating curves. Water estimated that 20% of the sediment flux passed through the Cherf Resources Research, 22, 1, 74–76. dam. Taking the same rate, the overall sediments deposited in the Fox, H.R., Moore, H.M., Newell Price, J. P., 1997. Soil erosion and reservoir reservoir during the 11-year period would be 5.46x106 tonnes. This sedimentation in the High Atlas Mountains, northern Morocco. IAHS value is converted to a mass using measurements of sediment Publications, 245, 233-240. density and according to the Agency of Algerian Dams (AAD), the Fried, J. S., Brown, D. G., Zweifler, M. O. & Gold, M. A., 2000. Mapping 3 contributing areas of stormwater discharge to streams using terrain dry bulk density is equal to 0.81 g/cm of the accumulation. The analysis. In Terrain Analysis, Principles and Applications, 1st ed.; Wilson, annual rate of sedimentation in the reservoir is therefore equal to J.P., Gallant, J.C., Eds. John Wiley & Sons, New York, USA, 183–203. 3 6.74 Mm and according to the initial storage capacity of the Haregeweyn, N., Poesen, J., Nyssen, J., De Wit, J., Haile, M., Govers, G., reservoir of 155 Mm3 and the total amount of accumulated Deckers, S., 2006. Reservoirs in Tigray (Northern Ethiopia): sediment in the reservoir, 4.35% of the total designed capacity is Characteristics and sediment deposition problems. Land Degradation lost (an average annual loss in storage capacity of 0.40% y-1). Development, 17, 211–230. This study has shown that the Cherf Reservoir has affected Horowitz, A. J., 2003. An evaluation of sediment rating curve for estimating more or less considerable sediment budget of the stream by suspended sediment concentration for subsequent flux calculations. Hydrological Processes, 17, 3387–3409. interrupting the normal flux of coarse and fine sediments supplied Jansson, M. B., 1985. A comparison of detransformed logarithmic regressions to the channel network by the upstream drainage basin. Sediments and power function regressions. Geografiska Annaler, 67A, 61–70. are stored during specific runoff periods, when inflowing loads are Jansson, M. B., 1997. Comparison of sediment rating curves developed on relatively large, and then gradually released when the reservoir is load and on concentration. Nordic Hydrology, 28, 3, 189–200. drawn down for irrigation water supply. Kassoul, M., Abdelgader, A., Belorgey, M., 1997. Caractérisation de la From the standpoint of long-term use of the reservoir for water sédimentation des barrages en Algérie. Revue des Sciences de l’Eau, 3, supply, continuous water release from the dam is beneficial 339–358. because it increases reservoir longevity, avoids sediment Kazi, M. N., Oussedik, M. N., 1977. Carte géologique de l’Algérie, Ain Beida à l’échelle 1/50.000, feuille 149. Institut Géographique National (IGN), concentrations and turbidity to be stored continuously for long Paris. periods, and helps to ensure a dependable future supply of water. Kazi, M. N., Oussedik, M. N., 1977. Carte géologique de l’Algérie, Jean Rigal à l’échelle 1/50.000, feuille 123. Institut Géographique National 4. Conclusions (IGN), Paris. Khanchoul, K., Jansson, M. B., Lange, Y., 2007. Comparison of suspended Suspended sediment concentrations and flow data for Cherf sediment yield in two catchments, northeast Algeria. Zeitschrift für drainage basin are used to develop sediment rating curves to Geomorphologie, 51, 1, 63–94. estimate sediment yield for a 19-year period from 1975 to 1994. Khanchoul, K., Jansson, M. B., 2008. Sediment rating curves developed on Based on the results obtained in this study, the sediment rating stage and seasonal means in discharge classes for the Mellah wadi, Algeria. Geografiska Annaler, 90A, 3, 227–236. curve method has given a reliable prediction of suspended Kothyari, U. C., Jain, M. K., Raju, K. G. R., 2002. Estimation of temporal sediment yield. As the reservoir sedimentation survey data are not variation of sediment yield using GIS. Hydrological Sciences Journal, 47, available, the multiple regression model is used to provide a 5, 693–706. reasonable approximation of suspended sediment yield estimates. Leahy, P., Kiely, G., Corcoran, G., 2008. Structural optimisation and input Results derived from the two empirical models not only have selection of an artificial neural network for river level prediction. Journal of saved time in dealing with the large historical database but also Hydrology, 355, 192–201. have allowed reconstruction by extrapolation from the history of Mathur, B. S., Singh, R., Goel, N. K., 1992. Development of Hydrological the sediment load, thereby permitting quantification of drainage Watershed Simulation Model. Department of Hydrology, University of Roorkee, Roorkee, India. basin responses to human disturbances (e.g. dam and agriculture) Probst, J. L., Amiotte Suchet, P., 1992. Fluvial suspended sediment transport as well as natural events. Sediment loads have shown fairly and mechanical erosion in the Maghreb (North Africa). Hydrological increasing trends from the late 1995s. The prevailing climatic Sciences Journal, 37, 621–637. conditions have certainly significant control over the water flux and Singh, V. P., 1989. Hydrology Systems, vol. II, Watershed Modeling. Prentice suspended sediment load patterns in the Cherf River. Hall, Englewood Cliffs, New Jersey, USA. The 30-year specific sediment transport in the study drainage Sogreah, 1983. Erosion et transport solide au Maghreb. Analyse basin is 18.19x106 tonnes. From the point of view of reservoir bibliographique. Rapport de projet RAB/80/011, PNUD. sedimentation, soil erosion appears to be a major problem in the Tamene, L., Park, S. J., Dikau, R., Vlek, P. L. G., 2006. Reservoir siltation in the semi-arid highlands of Northern Ethiopia: Sediment yield–catchment study reservoir. Hence, an integrated approach needs to be studied area relationship and a semi-quantitative approach for predicting sediment to control the erosion and reservoir sedimentation. yield. Earth Surface Processes Landforms, 31, 1364–1383. Touaibia, B., Aidaoui A., Gomez, D., 2001. Quantification et variabilité References temporelles de l’écoulement solide en zone semi-aride, de l’Algérie du Nord. Hydrological Sciences Journal, 46, 1, 41 53. Achite, M., Ouillon, S., 2007. Suspended sediment transport in a semiarid Vila, J. M., 1977. Carte géologique de Sedrata à l’échelle 1/50.000, feuille 99. watershed, Wadi Abd, Algeria (1973–1995). Journal of Hydrology, 343, Carte éditée par la Société Nationale SONATRACH, Algérie. 187–202. Voûte, C., Bétier, M. G., 1956. Carte géologique de l’Algérie, Ain Babouche à B.N.E.D.E.R. (Bureau National d’études pour le Développement Rural), 1998. l’échelle 1/50.000, feuille 122. Institut Géographique National (IGN), Paris. Etude agro-pédologique de la wilaya de . Rapport et cartes Walling, D. E., 1988. Erosion and sediment yield research; some recent lithologiques de M’daourouch et Oued Kebarit à l’échelle 1/50.000, perspectives. Journal of Hydrology, 100, 1-4, 113-141. feuilles 100 et 124. Direction Générale des Forêts, Tipaza, Algérie.