An appraisal of the sediment yield in western Mediterranean river basins

Science of the Total Environment 572 (2016) 538–553

Contents lists available at ScienceDirect

Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv

An appraisal of the sediment yield in western Mediterranean river basins C. Buendia a,b,⁎, A. Herrero a,b, S. Sabater a,c, R.J. Batalla a,b a b c

Catalan Institute for Water Research—ICRA, 17003, Girona, Catalonia, Spain Fluvial Dynamics Research Group—RIUS, University of Lleida, 25198 Lleida, Catalonia, Spain Institute of Aquatic Ecology, University of Girona, 17003 Girona, Catalonia, Spain

H I G H L I G H T S

G R A P H I C A L

A B S T R A C T

• We used data on sediment yield and basin characteristics on the western Mediterranean. • Area proved to be a limiting factor in the upper range of sediment yield values. • Multiple regression using basin-scale variables analysis indicated model instability. • Uncertainties prevent the use of the model in other regions.

a r t i c l e

i n f o

Article history: Received 15 June 2016 Received in revised form 8 August 2016 Accepted 9 August 2016 Available online xxxx Editor: D. Barcelo Keywords: Mediterranean Sediment yield Basin area Reservoir Scale-dependence Quantile regression Multiple regression

a b s t r a c t The number of studies assessing soil erosion and sediment transport has increased with the aim of achieving sustainable land and water management. Mediterranean rivers have been the object of many of these studies due to their naturally high values of sediment fluxes and a higher vulnerability under future climate scenarios. In this context, we attempt to use empirical relationships to (i) further assess the relation between sediment yield and basin scale and (ii) provide an update on the main drivers controlling sediment yield in these particular river systems. For this purpose, sediment yield data (from reservoir sedimentation surveys and sediment transport records) was collected from N100 locations distributed across the western Mediterranean area, with basin areas ranging from 1 to 100,000 km2. Quantile Regression analysis was used to assess the correlation between basin area and sediment yield, while additional basin-scale descriptors were related to sediment yield by means of multiple regression analysis. Results showed the complexity in the relationship between basin scale and sediment yield, with changes in supply conditions with increasing area introducing uncertainties in the correlation. Despite the large scatter, analysis pointed towards the same direction and area appeared to be the main constrain for the maximum value of sediment yield that can be found at a specific basin scale. Results from the multiple regression indicated that variables representing basin's physiography, climate and land use were highly correlated with the basins' sediment yield. Also, a better model performance was obtained when using total sediment yield instead of specific values (per unit area). Validation showed model instability, potentially due to data limitations and the use of catchments with varying characteristics. Overall, despite providing some insights on the correlation between sediment yield and basin-scale characteristics, validation prevented direct extrapolation of the model to other catchments. © 2016 Published by Elsevier B.V.

⁎ Corresponding author at: Catalan Institute for Water Research—ICRA, 17003, Girona, Catalonia, Spain. E-mail address: [email protected] (C. Buendia).

http://dx.doi.org/10.1016/j.scitotenv.2016.08.065 0048-9697/© 2016 Published by Elsevier B.V.

C. Buendia et al. / Science of the Total Environment 572 (2016) 538–553

1. Introduction Soil is a fundamental natural system that carries out numerous ecosystem functions and services such as carbon sequestration, water regulation, habitat preservation/functioning, food provision, and fuel production (e.g. Constanza et al., 1997; Swinton et al., 2007). Erosion affects these functions through on-site and off-site impacts (Segarra et al., 1991). On-site impacts include soil degradation and subsequent loss of productivity (Pimentel, 2006); whereas off-site impacts carry an increase in drought susceptibility and flood risk (Robinson and Blackman, 1990), surplus of sedimentation in stream channels with associated effects on aquatic biota (Buendia et al., 2013), and loss of reservoir capacity due to siltation (e.g. Verstraeten et al., 2003; Navas et al., 2007; Palazón and Navas, 2014, Batalla and Vericat, 2011). In addition, sediment may transport soil-absorbed contaminants (e.g. Stone, 2000; Quesada et al., 2014). Hence, the transport of sediment throughout the watershed can cause environmental hazards to floodplains and water bodies (rivers, lakes and estuaries) and ultimately to coastal and marine environments (Westrich and Förstner, 2007). As a result, increasing attention has been paid to understand the main drivers of soil erosion and the subsequent sediment transport processes. However, soil erosion is not a recent environmental concern, but it was already reported in Antiquity. For example, Plato wrote about soil erosion at the Acropolis of Athens: “... the Acropolis was not as now [before Plato]. For the fact is that a single night of excessive rain washed away the earth and laid bare the rock...” Over time, numerous ancient ports at the mouths of rivers deteriorated economically or were even abandoned due to the accumulation of eroded material from the hinterland and the subsequent growth of deltas, as it was the case of Miletus and Ephesus on the coast of Asia Minor (Migón, 2013). Despite the long-term concern of soil erosion and sediment production and transport, it has not been until relatively recently that studies are moving towards a sustainable management of land and water, for instance by considering the impact of land use changes on soil losses (Pacheco et al., 2014; Valle-Junior et al., 2014). One of the consequences of this new vision is the development of a myriad of models to quantify sediment loads in river basins (see de Vente et al., 2013 for a complete review). Advances in computer technology, remote sensing and geographic information systems (GIS) have contributed to significant progress of spatially distributed, process-based models. These models use grid structured information to describe soil erosion, considering their spatial variability within the basin (e.g. Schoorl and Veldkamp, 2001). They have proved useful in hydrology impact assessments (e.g. Praskievicz and Chang 2009, Bussi et al., 2014), but their ultimately application to sediment dynamics face three main drawbacks: first, the scarcity of measurements of erosion rates and, especially, sediment transport; while direct measurements are preferable, these are often difficult to obtain and data are scarce and disperse; second, deterministic models are usually data intensive and require long calibration and simulation times; and third, erosion processes depend on the selected hydrological model, and so inherent errors associated to their predictions (e.g. failing in reproducing, for instance, rainfall-runoff relationships), further propagate through the estimation of erosion and sediment transport processes (Rojas and Woolhiser, 2000; Wainwright and Parsons, 1998; Wainwright et al., 2008). Hence the use of these models at the catchment scale might be problematic, especially in large basins, where sediment deposition and transfer between landscape compartments operate at scales that models can hardly encompass (e.g. hundreds and thousands of years), and where human activities exert a large degree of control (e.g. reservoirs). Because of these constraints, the use of regressions built upon empirical information from river basins is an alternative to identify the major drivers in sediment yield at the catchment scale. As such, empirical regressions may bring further insights on the relationships between sediment yield and catchment characteristics. These relationships can be also used to back model equations, to validate model results and to

539

build more complex models. Regressions are the product of statistical analysis (e.g. based on stepwise linear regression) after the compilation of sediment yield data together with the associated catchment characteristics (e.g. morphology, topography, lithology, climate, discharge). Examples of such type of regressions can be found in Grauso et al., (2008) for rivers in Sicily, Haregeweyn et al. (2008) in Ethiopia, Verstraeten and Poesen (2001) in Belgium, and de Vente et al., (2011) in Spain. Mediterranean basins are of particular interest, probably because they have been defined as highly vulnerable to global change (e.g. UNEP, 2006). They are characterised by displaying above average sediment yields (i.e. rivers may reach sediment yields above 200 t km−2 y− 1; Inbar 1992; Milliman, 2001). These high values are typically interpreted as a function of the strong seasonality of climate and hydrological regimes, the dominance of softer rocks, the presence of elevated mountain ranges close to the coast line (i.e. high gradients, short distances between catchment headwaters and the sea level), and the long history of human presence in the region (Conacher and Sala 1998). Also, the extensive anthropogenic intervention in the fluvial territory (i.e. damming, extensive agriculture and extension of urban areas) may result in a more intense disruption of the natural water and sediment fluxes than in other less arid regions of the world (Milly et al., 2005). Understanding the response of sediment yields to global change in the Mediterranean region requires comprehending the variables controlling land erosion and subsequent sediment production and transport. Following previous studies (e.g. de Vente et al., 2011), we aim at further determining the main factors that drive sediment fluxes in western Mediterranean basins. We have therefore collected data on sediment yield at different locations across the region, including the 61 bathymetric surveys reviewed in de Vente et al., (2011), together with continuous sediment transport records found in the literature, adding up to more than one hundred locations. We have included a wide range of basin sizes (orders ranging from 1 to 100,000 km2) given that basin area is an important factor controlling sediment yield (Walling, 1983; Dedkov, 2004). This data set provides a range of climatological, hydrological and morphological conditions to assess the sediment yield drivers in Mediterranean basins. Thus, the specific objectives of the paper are to: (i) Explore the relation between sediment yield and basin area in this climatic region; and (ii) Assess the relations between climate, basin characteristics, and sediment yield.

2. Methods 2.1. The database The database comprises information on 116 locations across the western Mediterranean region, from the north of the Pyrenean Range (south of France) to the south of the Iberian Peninsula (Fig. 1). Sediment yield values were obtained from: (i) Sedimentation rates in 61 reservoirs in Spain, estimated from the volume of sediments retained in reservoirs using bathymetric surveys (Avendaño Salas et al., 1997a,b; Batalla and Vericat, 2011); and (ii) Sediment transport records available in the literature mostly for the Ebro basin, the Catalan Ranges, and the Pyrenean and Andalusian ranges (55 locations). The database included basin sizes from 1 to 85,000 km2, with most of them (ca. 70%) ranging between 1000 and 10,000 km2. Measurement periods varied from 1 to 89 years depending on the source of the data (i.e. sedimentation rates or sediment transport records). The longest periods are normally associated to sediment yields computed in reservoirs. Overall, Sediment Yield (hereafter SY) in the dataset varied between 3 and ca. 7,000,000 t y−1, and the area-specific sediment yield (SSY) varied between 0.1 and 2800 t km−2 y− 1. Table 1 gives a description of the basic characteristics of each location, the source of information, as well as the time span of the available data.

540

C. Buendia et al. / Science of the Total Environment 572 (2016) 538–553

Fig. 1. Location of the sites included in this study. Red dots indicate data from bathymetric surveys while green triangles correspond to sediment transport records. Basin areas of each individual point are shown in grey lines (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.).

2.2. Correction of sediment yields by the trapping efficiency of reservoirs Some of the sites included in this study (26% of the database; Fig. 1) are located downstream from one or more reservoirs. These reservoirs retain part of the sediment eroded from the catchment, therefore modifying the SY measured at any downstream site. In order to equally use data from sites with different upstream conditions, the data have to be corrected to take into account the trapping efficiency of reservoirs. Trapping efficiency (TE) refers to the percentage of incoming sediments that are retained in a reservoir (Brune, 1953). To take into account the variability in runoff, TE is expressed as a function of the water residence time (Δτ, years), which is the ratio of the total storage capacity (V; hm3) and the mean annual runoff for the available period (Qyr; hm3 y−1). Following Maneux et al. (2001), who also corrected sediment yield estimations for dam-trapping efficiency, we estimated the TE using the equation originally developed by Brune (1953) and modified by Ward (1980): 0:05 TE ¼ 1− pffiffiffiffiffiffi Δτ

SSY m

  SSY 0 ANR þ SSY 0 A1 1−TER1 ¼ A1 þ ANR

ð3Þ

Reordering the terms we obtain: SSY 0 ¼

SSY m ðA1 þ ANR Þ   ANR þ A1 1−TER1

ð4Þ

Since SY′ = SSY′(A1 + A2); then we can calculate the actual sediment yield as: SY 0 ¼

SY m AT   ANR þ A1 1−TER1

ð5Þ

This equation can be generalized so that the actual sediment yield (SY’) is estimated as: 2

ð1Þ

SY data are corrected based on the assumption that sediment produced per unit area (SSY; t km−2 y−1) is the same for the entire basin. Such assumption is actually an intrinsic property of the term SSY, as it is already providing an average value of the sediment produced by a basin as a whole, per unit of area and time, regardless of the intrabasin variability. Fig. 2a shows the particular case of a site downstream a single reservoir. We denote by SYm (t y−1) the measured value of sediment yield. SYm can be expressed as a function of the actual Specific Sediment Yield’ (i.e. sediment yield per unit area that would be measured in the absence of the reservoir, SSY’):   SY m ¼ SSY 0 ANR þ SSY 0 A1 1−TER1

equation:

ð2Þ

SSYm (t km−2 y−1) can be calculated by simply dividing by the total basin area of the site (AT = A1 + A2), hence obtaining the following

3 AT 4 SY X ¼ SY m Xn h i5 k  ANR þ A ∏ 1−TE i j j¼1 i¼1 0

ð6Þ

where SY'x is the corrected SY at a location x; SYm is the measured SY at this location; AT is the total basin area of x (km2); ANR corresponds to the non-regulated area of x (i.e. area downstream to the closest reservoir); Ai is the area of the basin draining the reservoir i; n is the number of reservoirs in the basin, and k refers to the number of reservoirs that are retaining the sediment eroded from the basin i. On the other hand, in those sites where SY is obtained from reservoir bathymetries the measured value has to be corrected to consider the amount of incoming sediment yield that is not trapped by the reservoir (i.e. fraction of sediment that is transported downstream during water releases from the dam). Thus: 0

SY X ¼ SY m

1 TEx

ð7Þ

It should be noted that in those cases where SY data are obtained from bathymetries in a reservoir affected by other upstream reservoirs, both corrections are applied consecutively.

C. Buendia et al. / Science of the Total Environment 572 (2016) 538–553

541

Table 1 Main characteristics and references of the locations included in the study. The column “Method” indicates whether the data has been obtained from river monitoring (Riv) or from reservoir bathymetric surveys (Bat). The Specific Sediment Yield data (SSYm) refers to the original value reported in the study. Study Site

River basin

Area (km2)

SSYm (t km−2y−1)

Study period

Method

Reference

Agly Aixola Alfonso XIII Algars Alloz Altés Añarbe Anchuricas Arbúcies Arguis Arquillo de San Blas Forata Guadalmellato Talave Barasona Barrendiola Santa María Belsué Bembezar BeniarBat Bermejales Besós Bolera Na Borges Bornós Bosc Breña Burgomillodo Buseo Camarillas Campàs Capella Cazalegas Cenajo La Cierva Cijara Contreras Corbeira Cubillas Tranco de Beas La Viñuela Embarcaderos Valuengo Ésera Puente Nuevo Fluvià Fogars Foix Cueva Forada Francolí Fuensanta Gabriel y Galán Gaià Gallipuén Gergal Argos Guadalhorce Torre de Abraham Torre del Aguila Inglabaga Isard Jorba San Juan Latxaga Santolea

French basins Cantabrian Segura Ebro Ebro Ebro Cantabrian Segura Fluvià Ebro Júcar Júcar Guadalquivir Segura Ebro Cantabrian Ebro Guadalquivir Júcar Guadalquivir Besós Guadalquivir Balearic islands Guadalquivir Ter Guadalquivir Duero Júcar Segura Ter Ebro Tajo Segura Segura Guadiana Júcar Galician coast Guadalquivir Guadalquivir Andalusian range Júcar Guadiana Ebro Guadalquivir Catalan range Catalan rage Catalan range Ebro Catalan range Segura Tajo Catalan range Ebro Guadalquivir Júcar Sur Guadiana Guadalquivir Ebro Catalan range Catalan range Tajo Ebro Ebro

1045 5 853 332 132 221 51 215 108 17 815 1067 1205 772 1511 3 190 1664 409 282 1020 170 316 1356 2 1485 785 219 2712 3 439 3996 2600 172 24,489 3312 16 603 284 125 16,284 1243 894 989 967 784 182 674 820 1206 1833 289 149 1726 417 1304 320 444 116 1 215 1921 2 1227

192 35 290 0.9 2063 9 46.5 363 35 501 59 109 252 246 3367 15 216 97 357 280 15 1415 0.3 450 1.8 168 127 64 67 4.8 544 34 89 279 711 320 8 127 179 1946 178 73 338 291 2.4 39 0.7 171 0.7 403 345 0.4 100 184 198 397 364 296 3.4 2800 5.5 965 29 27

1991–1997 2003–2005 1916–1985 2008–2010 1930–1997 2012–2013 2003–2005 1957–1979 1990–1998 1929–1971 1960–1988 1969–1983 1965–1992 1918–1993 1932–1991 2003–2005 1931–1980 1963–1994 1971–1991 1958–1978 1995–2006 1967–1979 2004–2006 1961–1990 2005–2009 1935–1991 1953–1989 1912–1980 1960–1993 2005–2009 2005–2012 1949–1990 1960–1992 1929–1987 1956–1983 1975–1994 2004–2007 1956–1990 1945–1990 1986–1994 1952–1983 1959–1985 2011–2012 1972–1994 1995–2003 1991–1998 1995–2008 1926–1996 1995–2002 1933–1991 1961–1990 1995–2002 1937–1989 1979–1985 1970–1991 1972–1991 1974–1988 1947–1992 2005–2013 1991–1999 2001–2003 1955–1992 1996–2005 1932–1993

Riv Bat Bat Riv Bat Riv Riv Bat Riv Bat Bat Bat Bat Bat Bat Riv Bat Bat Bat Bat Riv Bat Riv Bat Riv Bat Bat Bat Bat Riv Riv Bat Bat Bat Bat Bat Riv Bat Bat Bat Bat Bat Riv Bat Riv Riv Riv Bat Riv Bat Bat Riv Bat Bat Bat Bat Bat Bat Riv Riv Riv Bat Riv Bat

Serrat, 1999 Zabaleta et al., 2007 Sanz-Montero et al., 1998 Tena and Batalla, 2013 Batalla and Vericat, 2011 Tuset et al., 2016 Zabaleta et al., 2007 Sanz-Montero et al., 1998 Batalla et al., 2005 Batalla and Vericat, 2011 Avendaño Salas et al., 1997a/b Avendaño Salas et al., 1997a/b Avendaño Salas et al., 1997a/b Sanz-Montero et al., 1998 Batalla and Vericat, 2011 Zabaleta et al., 2007 Batalla and Vericat, 2011 Avendaño Salas et al., 1997a/b Avendaño Salas et al., 1997a/b Avendaño Salas et al., 1997a/b Liquete et al., 2009 Avendaño Salas et al., 1997a/b Estrany et al., 2009 Avendaño Salas et al., 1997a/b Pacheco et al., 2011 Avendaño Salas et al., 1997a/b Avendaño Salas et al., 1997a/b Sanz-Montero et al., 1998 Sanz-Montero et al., 1998 Pacheco et al., 2011 López-Tarazón et al., 2012 Avendaño Salas et al., 1997a/b Sanz-Montero et al., 1998 Sanz-Montero et al., 1998 Avendaño Salas et al., 1997a/b Avendaño Salas et al., 1997a/b Rodriguez-Blanco et al., 2013 Avendaño Salas et al., 1997a/b Avendaño Salas et al., 1997a/b Avendaño Salas et al., 1997a/b Avendaño Salas et al., 1997a/b Avendaño Salas et al., 1997a/b Lobera et al., 2016 Avendaño Salas et al., 1997a/b Liquete et al., 2009 Batalla et al., 2005 Liquete et al., 2009 Batalla and Vericat, 2011 Liquete et al., 2009 Sanz-Montero et al., 1998 Avendaño Salas et al., 1997a/b Liquete et al., 2009 Batalla and Vericat, 2011 Avendaño Salas et al., 1997a/b Avendaño Salas et al., 1997a/b Avendaño Salas et al., 1997a/b Avendaño Salas et al., 1997a/b Avendaño Salas et al., 1997a/b Tuset et al., 2016 Regués et al., 2000 Farguell and Sala, 2005 Avendaño Salas et al., 1997a/b Casalí et al., 2008 Batalla and Vericat, 2011

Study site Guadalen Guadalest Linabat del Arroyo Llavina Llobregat Maria Cristina Masquefa Matarraña

River basin Guadalquivir Júcar Duero Catalan range Llobregat Júcar Catalan range Ebro

Area (Km2) 1358 49 765 34 4952 1443 22 1035

SSYm (t km−2a−1) 359 2703 127 11 20 37 2.8 0.5

Study period 1954–1977 1965–1989 1951–1980 1990–1998 1995–2002 1920–1991 2001–2010 2008–2010

Data Bat Bat Bat Riv Riv Bat Riv Riv

Reference Avendaño Salas et al., 1997a/b Avendaño Salas et al., 1997a/b Avendaño Salas et al., 1997a/b Batalla et al., 2005 Liquete et al., 2009 Avendaño Salas et al., 1997a/b Martínez-Casasnovas et al., 2012 Tena and Batalla, 2013 (continued on next page)

542

C. Buendia et al. / Science of the Total Environment 572 (2016) 538–553

Table 1 (continued) Study Site

River basin

Area (km2)

SSYm (t km−2y−1)

Study period

Method

Reference

Mediano Guadalmena Mequinenza Minilla Moneva Mora d'Ebre Muga Sa Vall de la Nou Oliana Doña Aldonza Conde de Guadalhorce Oskotz Palmaces Pas de l'Ase Pedro Marin Pena Son Pou Puebla Puentes Guaiaraz Can Revull Ribarroja (Riv) Ribarroja (Batervoir) Riudecañas Rodó El Salado Santomera Sastago Segre Siurana Sotonera Sant Sadorní Santa TeBata Taibilla Talarn La Tejería Têt Ter Torrente Las Torcas La Tranquera El Vado Valdeinfierno Xerta

Ebro Guadalquivir Ebro Guadalquivir Ebro Ebro Catalan range Balearic islands Catalan range Guadalquivir Andalusian Ebro Tajo Ebro Guadalquivir Ebro Balearic islands Ebro Segura Andalusian range Balearic islands Ebro Ebro Catalan range Catalan range Andalusian range Segura Ebro Ebro Ebro Ebro Catalan range Duero Segura Ebro Ebro French basins Catalan range Ebro Ebro Ebro Tajo Segura Ebro

1568 1220 57,906 985 471 83,983 1833 264 2717 8683 222 17 274 83,280 9141 63 142 192 1417 368 1 82,936 82,768 30 4 7 148 49,777 12,975 73 327 729 1854 317 2062 2 1380 3287 9626 465 1474 377 426 84,743

184 46 120 150 18 0.8 0.4 0.7 162 131 613 64 120 0.7 94. 1239 0.1 356 105 146 3.1 0.5 18 112 710 184 27 13 3 0.4 1121 75 338 386 534 183 39 3 13 250 2.7 72 480 13

1959–1996 1969–1989 1966–1982 1956–1984 1929–1999 2008–2010 1995–2002 2004–2006 1959–2001 1955–1977 1921–1991 1999–2009 1954–1984 1998–2010 1954–1977 1930–1989 2004–2006 2010–2011 1884–1985 1971–1982 2004–2007 2008–2010 1969–1982 1918–1981 1991–1999 2007–2010 1965–1993 2002–2010 2008–2011 2011–2012 1963–1986 2001–2003 1960–1989 1973–1981 1916–1990 1996–2005 1980–1999 1995–2002 2008–2011 1946–1979 1960–1994 1972–1979 1897–1984 1998–2010

Bat Bat Bat Bat Bat Riv Riv Riv Bat Bat Bat Riv Bat Riv Bat Bat Riv Riv Bat Bat Riv Riv Bat Bat Riv Riv Bat Riv Riv Riv Bat Riv Bat Bat Bat Riv Riv Riv Riv Bat Bat Bat Bat Riv

Batalla and Vericat, 2011 Avendaño Salas et al., 1997a/b Batalla and Vericat, 2011 Avendaño Salas et al., 1997a/b Batalla and Vericat, 2011 Tena and Batalla, 2013 Liquete et al., 2009 Estrany et al., 2009 Batalla and Vericat, 2011 Avendaño Salas et al., 1997a/b Avendaño Salas et al., 1997a/b Casalí et al., 2008 Avendaño Salas et al., 1997a/b Tena and Batalla, 2013 Avendaño Salas et al., 1997a/b Batalla and Vericat, 2011 Estrany et al., 2009 Buendia et al., 2016 Sanz-Montero et al., 1998 Avendaño Salas et al., 1997a/b Estrany et al., 2009 Tena and Batalla, 2013 Batalla and Vericat, 2011 Avendaño Salas et al., 1997a/b Regués et al., 2000 Durán Zuazo et al., 2012 Sanz-Montero et al., 1998 Tena and Batalla, 2013 Tena and Batalla, 2013 Lobera et al., 2016 Batalla and Vericat, 2011 Farguell and Sala, 2005. Avendaño Salas et al., 1997a/b Sanz-Montero et al., 1998 Batalla and Vericat, 2011 Casalí et al., 2008 Serrat et al., 2001 Liquete et al., 2009 Tena and Batalla, 2013 Batalla and Vericat, 2011 Batalla and Vericat, 2011 Avendaño Salas et al., 1997a/b Sanz-Montero et al., 1998 Tena and Batalla, 2013

For clarification, an example of the application of this equation is shown in Fig. 2b, where the case of a site affected by two consecutive dams is considered. In this site, the measured sediment yield (SYm) has been computed from bathymetric surveys in the reservoir R3. Thus, apart from considering the sediment trapped in the upstream reservoirs, the sediment released from the reservoir itself has also been accounted for in SY′. Once TE has been computed for the three reservoirs following Eq.1, the corrected sediment yield (SY′) is computed as follows: 0

SY ¼ SY m



 AT 1 A3 þ ½A1 ð1−TER1 Þð1−TER2 Þ þ ½A2 ð1−TER2 Þ TEx

ð8Þ

For further interpretation of the equations, an additional example with quantitative data is shown in the Appendix. 2.3. Characterisation of the sites Variables selected for the analysis have been widely reported to be related to soil erosion and subsequent sediment transport (e.g. Ludwig and Probst, 1998; de Vente et al., 2011). Fig. 3 shows the potential influence of each group of variables in the processes involved in soil erosion and transport. Briefly, according to this conceptual model, sediment yield is the result of two main processes: (i) soil detachment and (ii) transport along hillslopes and the river network. Soil particle

detachment is mainly influenced by lithology (which will determine the erodibility of the materials), vegetation (presence of a vegetation cover will reduce soil erodibility) and climate variables; climate variables affect both processes mainly through rainfall erosivity. Rainfall causes soil particles detachment as a consequence of the impact of rain drops and further dissipation of kinetic energy, as well as control the hydrological response of the basin (runoff) and, ultimately, the downstream fluvial transport of sediment. Other climate variables, such as aridity or seasonality, influences moisture conditions in the basin, soil infiltration properties and runoff generation. A part from climate, the hydrological response of the basin is affected by relief parameters, such as elevation and slope, which ultimately influence flow velocity and transport capacity, and so sediment transport throughout the river network. Individual variables used in the analysis are specified in Table 2.

2.3.1. Basin morphology, topography and drainage network Relief variables were extracted using the SRTM Digital Elevation Database v4.1, which provides Digital Elevation Models (DEMs) for the entire world at a resolution of 90 m (Reuter et al., 2007). This database is available at the CGIAR-CSI GeoPortal (http://srtm.csi.cgiar.org). DEMs were processed in ArcGIS® v.10.2 to calculate a range of morphological and topographical variables for each location: basin area, catchment perimeter, mean basin slope, mean elevation, among others. It was also used to extract the river network and drainage density for all the basins,

C. Buendia et al. / Science of the Total Environment 572 (2016) 538–553

543

Fig. 2. Examples showing two sites (red dots), where sediment yield measurements (SYm) are affected by an uptream reservoir (a) or a sequence of reservoirs (b). Reservoirs are denoted in blue by R (R1, R2 and R3), while their corresponding basin area is indicated by A (A1, A2 and A3) and it is also highlighted in different patterns. In (a) SYm has been computed from sediment transport records and hence has to be corrected by the trapping efficiency of R1. In (b) SYm has been obtained from bathymetric surveys in the reservoir R3 and thus it has to be corrected by the trapping efficiency of the upstream reservoirs (R1 and R2) as well as for the trapping efficiency of the reservoir R3 in order to account for the sediment releases. Maps are just examples and do not represent real locations, so they are presented without graphical scale (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.).

which can be used as a surrogate for gully presence. For this, a channel was defined when the upstream contributing area is larger than 0.2 km2, a threshold that according to de Vente et al., (2011) provides a reasonable drainage network without deformations. 2.3.2. Climate Monthly and annual climate data (precipitation, temperature and potential evapotranspiration) were used to calculate a variety of descriptors of the climate properties for each study site. Mean annual data for the period of record was usually found in almost all of the studies. For those locations for which such data was not available, monthly climate data were obtained from the Integrated Water Information System (SIA) of the Spanish Ministry of Agriculture and Environment. This dataset provides gridded monthly climate data (at 1 km resolution) measured at the weather stations of the Spanish Meteorological Agency (AEMET). Since data from SIA only covered the Iberian Peninsula, monthly meteorological data for the two French sites (Agly and Têt) was provided by Météo-France for a suit of meteorological stations (i.e. Carcassone, Torreilles, Perpignan, Ste. Locaide, Formigueres, Narbonne and Le). Climate data was constrained to the period for which sediment data was measured to keep consistency in the relationships between

Fig. 3. Conceptual model showing the relationships between selected variables in the study.

sediment yield and climate factors. As in de Vente et al., (2011), monthly and annual climate data were used to calculate the Seasonality Index (SI; Walsh and Lawler, 1981); the Precipitation Concentration Index, (PCI; Oliver, 1980), the Modified Fournier Index, (MF; Arnoldus, 1977), and the rainfall erosivity R factor (Rf; Renard and Freimund, 1994) from the Revised Universal Soil Loss Equation. These variables are used as a measure of both seasonality (i.e. temporal distribution of the precipitation throughout the year) and rainfall erosivity. 2.3.3. Land use and geological data Land use information was obtained from the CORINE Land Cover map at 100 m resolution developed by the European Environment Agency. Since soil erosion and sediment yield are related to land use, we tried to match each sediment measure to the closest land use map available. For this, land use maps for the years 1990, 2000 or 2006 were used, always considering the version nearest to the measuring period of each site. Geological information was downloaded from the OneGeology Europe portal (Janssen and Kuczerawy, 2012) and had a scale of 1:100,000. Categories used for land use and geological data are indicated in Table 2. 2.4. Statistical analysis 2.4.1. Area-sediment yield correlation The relation between basin area (A) and SSY was assessed by means of Ordinary Least Squares (hereafter OLS), a usual approach in the literature (de Vente et al., 2011; Batalla and Vericat, 2011; Verstraeten et al., 2003). Besides, we also applied Quantile Regression Analysis (hereafter QR), which addresses conditional relationships between an explanatory variable and the quantiles of a response variable. QR is more robust to non-normal errors and outliers and provides a more complete picture of the correlation between two variables than OLS (which assumes that all quantiles of a response variable follow the central tendency). In QR, for a pre-defined quantile τ ranging from 0 to 1, an algorithm finds the function for which τ of the data points fall at or below the SSY value predicted. Therefore, for high values of τ(e.g. 0.90 or 0.95), only a small fraction of the points exceed the predicted value of SSY, and the

544

C. Buendia et al. / Science of the Total Environment 572 (2016) 538–553

Table 2 Variables used to characterise basins properties and further used in the statistical analysis. Type

Code

Units

Description

Morphology and river network

A Per Dist SF

km2 km km –

SL DD

km km km−2

SO S Hmin Hmax HD

– (%) m.a.s.l m.a.s.l m

Hmean HI

m.a.s.l –

RR

m km−2

P T ETP AI

mm °C mm –

MF

mm

Catchment Area Catchment Perimeter Distance between the location and the highest point Shape factor S F = A/Dist2 Stream Length Drainage density DD = SL/A Stream Order (Strahler) Mean basin slope Minimum Elevation Maximum Elevation Height Difference; HD = Hmax − Hmin Mean Elevation Hypsometric integral; HI = (Hmean − Hmin)/(HmaxHmin) Relief Ratio RR = (Hmax − Hmin)/A Mean annual precipitation Mean annual temperature Mean annual Potential Evapotranspiration Aridity Index AI = ETP/P Modified Fournier Index

Topography

Climate

12

MF ¼ ð∑p2i Þ=P i¼1

PCI

%

Precipitation Concentration Index 12

PCI ¼ ½ð∑; p2i Þ=P 2 100 Rf

MJ mm m−2 h−1

SI

–

i¼1

RUSLE erosivity factor R Rf = 0.7397MF1.847, for MF b 55 mm Rf = 95.77 − 6.081MF + 0.4770MF2 , for MF N 55mm Seasonality Index 12

SI ¼ ð1=PÞ∑jpi −ðP=12Þj i¼1

Hydrology

Land use

Lithology

R SR

hm3y−1 mm y−1

RfR

–

NP Agr For Gr BS Ig Mt Sed Unc

% % % % % % % % %

regression function estimates the upper limit of the point cloud. Conversely, for low values of τ(e.g. 0.05 or 0.1), the lower limit of the cloud of data points is estimated. Hence, QR allows the assessment of the minimum and maximum value of SSY that could be expected for a given value of A. In this study, the 10th and 90th percentiles were considered as these were the most extreme limits of the data for which quantiles could be validly calculated, given the sample size (as per Lancaster and Beylea, 2006). 2.4.2. Basin scale drivers of sediment yield We first produced an ordination to identify underlying gradients and potential groupings of the studied sites. For this, we used Nonmetric Multidimensional Scaling (NMDS, Kruskal, 1964), which is able to arrange the data in a multidimensional space (usually a 2D space). This unconstrained ordination has the advantage that it makes few assumptions about the nature of the data and allows the use of any distance measure (McCune and Mefford, 1995). Goodness of fit (i.e. how well

Where: P—Mean annual Precipitation pi—Mean monthly precipitation for month i Mean annual runoff Specific runoff SR = R/A Runoff Ratio SR = SR/Prec Non-Productive land (urban and bare soil) Agricultural land Forest Grassland Bare soil Igneous Metamorphic Sedimentary Unconsolidated material

the ordination summarises the observed distances among the samples) was expressed by the Stress value, which indicate the degree of correspondence between the distances among the points in the ordination plot and the original distances measured. Lower stress values mean that the data fit the final arrangement better. Last, two vectors representing SY and SSY were fitted to the final plot to help interpretation of the ordination. Squared correlation coefficients (R2) for the vector and the significance of these values were assessed by 1000 permutations of the sediment values. All statistical analyses were performed using the following packages within the R environment v.3.2.5: quantreg (Koenker, 2016) for Quantile Regression; and vegan (Oksanen et al., 2012) for NMDS. Since practically any of the variables was normally distributed, Spearman correlation was used to derive the correlation matrix between basin properties and sediment yield (SSY and SY). This analysis compiles the correlation coefficients between all the possible pairs of variables considered.

C. Buendia et al. / Science of the Total Environment 572 (2016) 538–553

Step-wise multiple regression methods are based on the step-bystep improvement of the model performance by adding or deleting one variable or a group of them, until the amelioration of the model falls under a specific threshold. Because of the lognormal distribution of SY and SSY, both variables were log-transformed and used as dependent variables in the analysis. Not transforming these two variables resulted in the violation of the residual normality assumption in regression analysis.

545

3. Results

(p-value = 0.03), suggesting a more consistent SSY-A correlation at the higher end of the distribution of SSY. Results from QR using the corrected SSY data, showed varying trends depending on the quantile used. When the lower quantile (i.e. 10th percentile) was considered, results indicated positive but not significant SSY-A correlations (p-values ~ 0.6). Such trend was reversed when higher quantiles were used. In the case of the 50th percentile, a slightly negative but not significant trend is observed (p-value ~0.7). Trends become clear when using the 90th percentile, in which case the SSY-A correlation became markedly negative and significant (p-value = 0.007).

3.1. Basin area and sediment yield

3.2. Correlation between sediment yield and catchment scale variables

OLS yielded a positive correlation between SY and A, regardless of the type of data used (i.e. measured or corrected; Fig. 4). However, regression results showed a better fit when considering the data corrected by the trapping efficiency of the upstream reservoirs (i.e. R2 increases from 0.45 to 0.56). Divergent results were observed when assessing the correlation between SSY and A. OLS regression showed an inverse correlation between both variables when the measured data was considered; however, the large scatter in the data produced non-significant results (p-value = 0.56). The use of the corrected data reversed the negative trend between SSY and A; however, the scatter in the data was still large and results were not statistically significant either (p-value = 0.22). QR provided a more complete picture of the conditional distribution of SSY given A (Fig. 5 and Table 3). Following the trend observed in the OLS, when the measured SSY data were considered, a negative SSY-A correlation throughout all the range of quantiles was obtained. However, only the regression for the percentile 90th showed significant results

3.2.1. Site characterization NMDS produced a 2-axis ordination (Fig. 6) with a final Stress value of 0.15, which indicates a satisfactory solution (Clarke, 1993). The SY vector fitted to the final plot showed a significant correlation with the first axis of the ordination (p-value b 0.01; R2 = 0.31), but the vector for SSY did not prove significant (p-value = 0.2; R2 = 0.03). Two main groups were distinguished by the NMDS along axis 1. The group located on the left hand side of the diagram was mainly formed by the basins located in the northern part of the Iberian Peninsula (e.g. Oskotz, Barrendiola, Corbeira, Añarbe). They are characterised by having larger values of basin relief ratio, shape factor and basin slope. Forest and grassland are the predominant land use in these sites, which also showed wetter settings, with larger values of precipitation as well as specific runoff and runoff coefficient. In the right hand side of the plot sites show larger values of SY. Sites belonging to this group were mainly located in the eastern coast of the Iberian Peninsula (e.g. Xerta, Forata), the Andalusian Range (e.g. La

Fig. 4. Regression between basin area (A; km2), Sediment Yield (SY, t y−1) and Specific Sediment Yield (SSY, t km−2y−1) according to the original (measured) data and the corrected data (i.e. using Eq. 6).

546

C. Buendia et al. / Science of the Total Environment 572 (2016) 538–553

Fig. 5. Relationship between basin area (A, km2) and Specific Sediment Yield (SSY; t km−2 y−1) using Quantile Regression Analysis (QR) for measured data (a) and corrected data (b). Lines indicate quantiles 0.1 (lower limit), 0.5 and 0.9 (upper limit). Regression coefficients for each analysis are shown in Table 3.

Viñuela, Puentes) and the Balearic Islands (e.g. Son Pou, Na Borges). These sites showed larger values of basin area, basin perimeter, stream length or drainage density, as well as a large proportion of sedimentary, unconsolidated rocks and agricultural areas. These sites characteristically have a drier climate, with larger seasonality index values and mean annual temperatures.

runoff for higher water velocities (Fig. 7b). Concerning climate variables, seasonality index showed a positive correlation with SSY, result of the influence of extreme events on sediment mobilization and transport (Fig. 7c). Regarding land use characteristics, urban area and SSY were inversely correlated (Fig. 7d), due the lower availability of sediment in highly urbanized catchments.

3.2.2. Correlation between sediment yield and basin-scale variables The Spearman correlation matrix is presented in the Table 4. Results showed that SSY correlated relatively well with topographic variables, such as slope and maximum height. Concerning climate variables, evapotranspiration had the highest correlation. Regarding land use, the percentage of urban soil was the most correlated. High significant correlations with total SY occurred with the morphological variables of the catchment (area, perimeter and distance from the site to the highest point of the catchment). Drainage network characteristics such as stream length and stream order were also strongly related to SY, whereas height difference and relief ratio were the most influential amongst the topographic variables. Climate variables presented a comparatively lower correlation with SY, and evapotranspiration and aridity index were the ones more correlated with SY. Nevertheless, dependences occur amongst variables. Some examples are: catchment area was positively correlated with the basin perimeter and other drainage characteristics (distance from the site to the highest point, stream length, stream order); stream length and stream order also presented a high correlation; morphological variables showed a strong correlations with orographic characteristics; basin slope showed a negative correlation with the percentage of agricultural area, which in turn had a positive correlation with forest cover (probably as a consequence of the selection of easily workable surfaces for cultivation); last, some climate variables were also correlated, mainly precipitation and aridity index, Fournier index and rainfall erosivity. Fig. 7 shows some examples of the correlation between SSY and morphological, topographical, climate and land use variables. Slope and SSY were positively correlated due to the higher erosion power of

3.3. Multiple regression analysis Only those variables with a relatively high correlation coefficient (N0.2) were considered in the regression analysis. The transformed version of SSY and SY (lnSSY and lnSY) and of the environmental variables was used for the analysis (except land use variables which were expressed in percentages). The resulting models were (Table 5): ln ðSSY Þ ¼ −3:58−0:029  S þ 1:83  10−3  H min þ 2:39  10−4  HD þ 6:37  10−3  ETP−0:115  U þ 1:21  10−3  SR þ 2:19 ð8Þ  10−2  RR

Table 3 Regression coefficients for Quantile Regression analysis (QR) and Ordinary Least Squares Regression (OLS) using measured and corrected specific sediment yield data. Results for the 10th, 50th, and 90th percentiles are shown (τ = 0.1, 0.5 and 0.9, respectively). Numbers in bold indicate significant at the 95% confidence level (i.e. τ = 0.05). Measured

Corrected

Analysis

Intercept

slope

p-value

Intercept

slope

p-value

OLS QR τ = 0.1 τ = 0.5 τ = 0.9

2.07

−0.11

0.52

1.79

0.09

0.22

0.43 2.17 3.58

−0.04 −0.05 −0.31

0.63 0.78 0.04

0.21 2.19 3.54

0.18 −0.04 −0.24

0.22 0.72 0.007

Fig. 6. NMDS ordination plot for studied sites. Lines are drawn connecting each site and the centroid of the group they belong to. Ellipses are also drawn to show 95% confidence ellipses around the class centroid.

Table 4 Correlation matrix between the catchment variables and sediment yield (total and specific). Variables with low correlation coefficient were removed for clarity. SSY

SY

A

Per

Dist

SF

SL

DD

SO

S

Hmean

Hmax

HD

HI

RR

P

T

ETP

AI

SI

Rf

Urb

Agr

For

0.73 0.03 0.01 0.00 0.00 −0.02 −0.19 0.06 0.24 0.41 0.47 0.21 0.14 −0.08 0.12 −0.15 0.23 0.03 0.19 0.10 −0.26 −0.14 0.08 0.11 −0.01 0.24 0.31

0.39 0.49 0.47 −0.25 0.40 0.16 0.68 0.06 0.11 0.36 0.54 −0.17 −0.46 −0.15 −0.09 0.23 0.25 0.14 −0.11 −0.17 −0.02 −0.09 0.12 0.40 0.00 0.12

0.97 0.97 −0.19 1.00 0.17 0.68 −0.06 −0.31 −0.15 0.31 −0.20 −0.11 −0.24 0.04 −0.12 0.13 −0.10 −0.13 −0.01 0.12 −0.20 0.09 0.99 −0.10 0.01

0.99 −0.26 0.98 0.22 0.78 −0.10 −0.33 −0.12 0.40 −0.25 −0.17 −0.26 0.03 −0.08 0.17 −0.10 −0.17 −0.01 0.14 −0.23 0.10 0.96 −0.14 −0.03

−0.28 0.98 0.21 0.77 −0.10 −0.33 −0.13 0.38 −0.23 −0.17 −0.25 0.03 −0.09 0.15 −0.09 −0.15 −0.01 0.13 −0.22 0.11 0.96 −0.13 −0.02

−0.19 −0.25 −0.33 0.14 0.21 0.02 −0.26 0.12 0.09 0.15 −0.06 −0.09 −0.05 −0.12 0.04 0.10 −0.09 0.22 −0.17 −0.19 0.10 0.06

0.18 0.69 −0.07 −0.30 −0.15 0.31 −0.20 −0.11 −0.24 0.04 −0.12 0.14 −0.10 −0.13 −0.01 0.13 −0.21 0.09 0.98 −0.10 0.00

0.43 −0.50 −0.28 −0.26 0.08 −0.41 −0.41 −0.46 0.34 0.17 0.35 0.15 −0.33 0.16 0.42 −0.39 −0.18 0.16 −0.32 −0.18

−0.21 −0.25 0.01 0.48 −0.35 −0.55 −0.36 0.04 0.10 0.34 −0.02 −0.27 0.05 0.17 −0.24 0.01 0.67 −0.26 −0.16

0.27 0.61 0.45 0.29 0.26 0.56 −0.65 −0.25 −0.48 −0.16 0.44 −0.08 −0.68 0.48 0.39 −0.01 0.58 0.47

0.78 −0.05 0.21 0.06 0.12 −0.55 −0.17 −0.19 −0.19 0.04 −0.32 −0.33 0.30 0.12 −0.29 0.21 0.22

0.49 0.26 −0.03 0.17 −0.76 −0.17 −0.21 −0.21 0.08 −0.29 −0.49 0.33 0.32 −0.12 0.27 0.29

−0.32 −0.20 0.05 −0.46 −0.04 −0.04 −0.20 0.01 −0.06 −0.23 0.03 0.27 0.35 0.11 0.16

0.16 0.13 −0.21 −0.13 −0.18 0.07 0.19 −0.12 −0.30 0.30 0.15 −0.21 0.15 0.10

0.29 0.00 −0.15 −0.20 0.00 0.09 −0.06 −0.13 0.12 0.09 −0.11 0.20 0.11

−0.44 −0.31 −0.80 −0.10 0.79 −0.05 −0.50 0.46 0.11 −0.20 0.74 0.47

0.54 0.60 0.62 −0.30 0.14 0.62 −0.44 −0.37 −0.01 −0.46 −0.42

0.59 0.65 −0.23 0.04 0.33 −0.31 −0.14 −0.14 −0.32 −0.22

0.36 −0.67 −0.01 0.48 −0.39 −0.23 0.08 −0.59 −0.41

0.01 −0.10 0.26 −0.22 −0.09 −0.12 −0.15 −0.11

−0.10 −0.40 0.37 0.14 −0.10 0.74 0.42

0.01 −0.02 −0.15 −0.01 −0.07 −0.05

−0.82 −0.53 0.09 −0.44 −0.33

0.04 −0.18 0.11 0.38 0.12 −0.05 0.24 0.13 0.07 0.86

Gr

R

SR C. Buendia et al. / Science of the Total Environment 572 (2016) 538–553

SY A Per Dist SF SL DD SO S Hmean Hmax HD HI RR P T ETP AI SI Rf Urb Agr For Gr R SR RR

547

548

C. Buendia et al. / Science of the Total Environment 572 (2016) 538–553

Fig. 7. Relationship between some of the environmental variables and sediment yield.

ln ðSY Þ ¼ −1:43−9:14  10−5  A þ 7:60  10−3 Per−2:15  103  D−3:20  10−2  SF−2:24  10−4  SL−0:40  SO þ 3:78  10−3  Hmean −2:41  10−4  HD−1:52  10−2  RR þ 7:02 ð9Þ  10−3  ETP−1:77  10−2  AI þ 6:59  10−4  R The performance of these equations is shown in Fig. 8. Results showed a markedly higher dispersion for SSY than for SY, as indicated by the correlation coefficient (R2 = 0.39 and 0.73, respectively). AIC criteria (Akaike Information Coefficient) informed on the quality of the models obtained (Table 5). It gives an estimate of the information lost when each of the variables is excluded from the model. Based on these results, Hmean, HD

and ETP explained most of the variance in SSY. In the case of SY, Hmean and ETP were the most influential, showing AIC values significantly higher than those obtained for the rest of variables. Accordingly, these variables showed the highest RSS (Residual Sum of Squares values). Last, the performance of the model was tested following the “Jackknife” validation method (as per Shao & Tu, 1995). This method consists in deleting one site and carrying out the multiple regression analysis with the same dependent variables and the rest of sites. The SSY of the deleted site is calculated with the equation resulting from the multiple regression associated with the rest of the sites. This process is repeated

Table 5 Results of the multiple regression analysis associated to the equations of SSY and total SY (RSS: Residual Sum of Squares; AIC: Akaike Information Coefficient; F: F-test statistic). Sediment measure

Sum of squares

RSS

AIC

F value

Pr (NF)

SSY S H Hmin HD ETP Urban SR RR

2.5 1.5 4.4 1.1 65.5 7.7 1.3 5.5

336.5 335.6 338.5 335.2 399.6 341.8 335.3 339.5

139.6 139.2 140.2 139.1 159.5 141.3 139.1 140.6

0.8 0.5 1.4 0.4 21.0 2.5 0.4 1.8

0.37 0.48 0.24 0.55 0.00 0.12 0.52 0.19

SY A Per D SF SL SO Hmean HD RR ETP AI R

1.61 11.55 0.18 0.04 4.09 6.40 140.17 1.29 19.09 51.28 0.003 4.47

419.42 429.36 417.99 417.85 421.90 424.21 557.98 419.10 436.90 469.09 417.81 422.28

173.09 175.81 172.70 172.66 173.78 174.41 206.21 173.00 177.83 186.08 172.65 173.88

0.40 2.85 0.04 0.01 1.01 1.58 34.55 0.32 4.71 12.64 b0.01 1.10

0.53 0.09 0.83 0.92 0.32 0.21 b0.01 0.57 0.03 b0.01 0.98 0.30

Fig. 8. Comparison between the measured and corrected sediment yield (total and specific) and the values calculated with Eq.8 and Eq.9.

C. Buendia et al. / Science of the Total Environment 572 (2016) 538–553

deleting one site each time. Fig. 9a shows the comparison between SSY calculated through this validation method and the corresponding observed value. The scatter in the data is quite important and R2 is similar to that obtained in the calibration results (0.39 for calibration and 0.41 for validation), suggesting that the performance of the model is not completely satisfactory. The uncertainty in the predictions is relatively high when applied to basins not included in the statistical analysis performed herein. Fig. 8b shows the results of the validation for the SY data. The scatter in the data was slightly lower (R2 = 0.65 and 0.73 in calibration). This result can be associated to the higher number of dependent variables involved in the regression analysis. 4. Discussion 4.1. Evaluating the SSY-A correlation Geomorphic processes that control erosion, and sediment transport and deposition are known to be highly scale-dependent and so catchment area has been long considered one of the most important variables in the prediction of sediment yields. Its importance is such that in many cases it has been the only variable used for prediction purposes in ungauged catchments at the regional or even global scales (e.g. Strand, 1975; Lahlou, 1988; Renwick et al., 2005a; Syvitski et al., 2003). Total Sediment Yield (t y− 1) has been reported to increase with increasing basin area (i.e. the more extensive the draining area, the greater the yield; Romero Díaz et al., 1992; Milliman and Syvitski, 1992). Such trend is also corroborated by this study. In contrast to SY, the Specific Sediment Yield (t km− 2 y− 1) usually decreases with increasing catchment area, as reported in the theory of sediment sources and sinks by Walling (1983). Such inverse relationship between SSY and A is the result of an increased presence of flat areas (e.g. floodplains, concavities, alluvial planes) that promote sediment deposition and hence act as sediment traps. Connectivity between sources and

Fig. 9. Comparison between the observed sediment yield (total and specific) and the values calculated through the “jackknife” validation (equations resulting from the regression analysis over the rest of the catchments).

549

transport paths is also larger in smaller basins, therefore facilitating the export of great proportions of the sediment produced in the catchments (this phenomenon is best explained by the classic concept of Sediment Delivery Ratio). Results obtained from OLS when using the measured data were in accordance with this general pattern and showed an inverse correlation between SSY and A. These results agreed to others elsewhere. Batalla and Vericat (2011) reported a decreasing trend in SSY with increasing A using data from bathymetric surveys for the entire Ebro basin (NE Spain). These authors found a notably higher statistical dependence (i.e. regression slope − 0.55), a value in the range with those previously estimated by Verstraeten et al., (2003) and de Vente et al., (2011) for 60 basins in Spain (i.e. ~−0.4). In our study the declining SSY-A trend was weak (slope of − 0.07) and even positive when considering the corrected values of SSY. Such differences might be the result of two factors: First, SSY-A relationships in the Iberian Peninsula have been usually assessed using bathymetric data from reservoirs reported in Avendaño and Salas (1997a,b). Here, we have also included data from sediment transport records and, especially, for a wider range of basin scales. The main drawback of using such type of data is that usually sediment transport records are short (in this study record lengths rarely include N10 years of data, Table 1). Given the large inter annual variability in sediment transport, the inclusion of short-term data (e.g. 1-year) may enhance scatter and hence may bias the long-term SSY-A correlations. DeBoer and Crosby (1996) also pinpointed the potential distortion in the SSY-A resulting from the lack of temporal overlap between sediment records. These authors suggested care when analysing this type of data to ensure that observed trends are not artefacts resulting from the use of different periods with contrasting precipitation and runoff characteristics. Second, despite using data from bathymetric surveys, previous studies rarely accounted for the trapping efficiency of the upstream reservoirs. Only Batalla and Vericat (2011) corrected SSY values from bathymetric surveys to obtain a more accurate value of SSY. These authors used Brune's (1953) trap efficiency curve to estimate the amount of sediment released by each reservoir and then added this value to the SSY estimated from the bathymetric surveys. In order to account for the potential effects of the upstream reservoirs, other authors have extracted the area of these reservoirs from the total catchment area of the study site prior to the computation of SSY (e.g. de Vente, 2009). The use of different methods to correct the data, or the absence of such correction, likely has an effect in the final SSY values used in the analysis, and ultimately in the obtained results and trends. In our study, when the corrected data were applied, a positive trend was observed. This was mainly due to the fact that the larger basin areas are located at the outlet of the Ebro basin (where basin scale is ~ 80,000 km2). These locations have a large number of reservoirs upstream. For example, Xerta (basin area of ca 85,000 km2) has ca. 190 large reservoirs upstream, with a capacity ranging from 1 to 1500 hm3 (i.e. 1 hm3 = 1 × 106 m3). Applying the correction moves data points upwards in the y-axis (i.e. SSY), forcing the slope to change in the computed regressions. Such change does not match the results found in Batalla and Vericat (2011), who reported a negative trend for the entire Ebro basin. However, when comparing the results, it should be taken into account that they did not consider the sediment trapped in the upstream reservoirs, only the sediment released by the studied reservoir. The use of the correction noticeably leads to contrasting results. In addition, while the SSY-A correlation has always been addressed using OLS, in this study QR analysis provided a new insight from which two main results can be drawn. First of all, results showed that SSY may vary several orders of magnitude for a given basin scale. This indicates the existence of multiple factors and interactions that affect SSY, which results in the existence of a non-linear correlation between SSY and A. Secondly, in both cases (i.e. QR using measured and corrected data), a strong linear negative correlation was found on the upper quantiles of SSY (percentile 90th, which represents the maximum response). This fact might indicate that A is posing an upper limit or

550

C. Buendia et al. / Science of the Total Environment 572 (2016) 538–553

constrain on SSY, and hence SSY might not change above the upper limit set by A (but may change to less when other unmeasured factors act as limiting factors (Cade and Noon, 2003). The complexity observed in the results is backed by the range of positive and negative SSY-A correlations found in the literature. On one hand, a decreasing average slope with increasing area leads to lower SSY due to increased chances of sediment deposition. On the other hand, other properties or processes can lead to positive correlations between the area and SSY. For example, Dedkov (2004) indicated that where there is a dense vegetation cover, surface erosion (sheet and gully) is limited, and the main supply of sediment is provided by erosion of river channel and banks. The remobilization of unconsolidated sediments (Slaymaker et al. 2003) or the absence of depositional areas due to the shape of the valley (Kirchner et al. 2001) can also lead to positive SSY-A correlations. Another example of the complexity in the SSY-A correlation was introduced by de Vente and Poesen (2005), who reported cases in which SSY first increased with A, and then decreased. These authors argued that for small spatial scales, SSY was found to increase or decrease depending on a range of factors such as the basins' runoff, vegetation, and soil moisture conditions (e.g. Cammeraat, 2002; Arnau-Rosalén et al., 2008). As the basin area increased, rill and gully erosion and mass movements gain relevance, connectivity between sediment sources and the fluvial network increased, and hence SSY may increase with increasing A. For even larger values of A, when the basin's slope gradient decreased, SSY was limited by fluvial dynamics (i.e. magnitude and frequency of competent floods, affecting the rate of downstream sediment conveyance), greater opportunities for sediment deposition existed, and so SSY generally decreased with further increasing A. Also, Dedkov and Moszherin (1992), and later Church et al., (1999) and Slaymaker et al., (2003) suggested that the SSY-A correlation depended on the main source of sediment in the basin. They suggested that an inverse SSY-A correlation could be attributed to the dominance of hillslope erosion processes, later deposition of sediments during transport and channel aggradation. A positive SSY-A relation was attributed to a gain in sediment along the channel due to dominance of channel degradation and also to the delivery of sediments to the mainstem river by floodplain sediment reworking, landsliding and other discrete sources unevenly distributed in the basins. These authors found this inverse relation in areas heavily affected by glaciation where huge amounts of sediment stored in the basin are still to be reworked likely during centuries and even millennia. Their findings and interpretation of results were mostly related to the sediment source and the further downstream cycles of channel aggradation and degradation, rather than to general basin characteristics as we tried to examine here. None of our catchments are controlled by such processes at least in the magnitude that Canadian river basins described by them are still affected. A number of studies have indicated that processes and factors controlling sediment transfer can change dramatically between scales so that the SSY-A correlation may not be straightforward. For instance, according to de Boer and Crosby (1996), sediment yield (SY) and specific sediment yield (SSY) are controlled by supply conditions, with transport capacity playing a secondary role. This fact has been also reported in catchments characterised by extremely active sediments sources patches through the basin, as in the case of badland stripes (e.g. Lopez-Tarazón et al., 2010). Consequently, the relation between SSY and A is extraordinary complex, since supply conditions may not change proportionally with basin scale, as a result of changes in both sediment sources and sinks. Although it may sound trivial and easy to anticipate this is not an insignificant outcome from this work, in which a combination of statistical analysis has been applied to a wide range of basin scales, allowing to draw more general conclusions than those obtained in previous works (despite finding a large scatter in the results). Sediment loads affects fundamental ecosystems processes, such as the transport of nutrients and pollutants and the in-stream habitat functioning; while its variability and somewhat unpredictable behaviour

poses fundamental challenges for water and sediment yield modelling exercises (i.e. model calibration and validation) that are more and more used to assess the hydromorphic and ecological effects of global change. In addition, sediment yield data for short periods may provide a misleading picture of larger scale patterns, and vice versa, owing to the effect of local sediment sources and sinks controlled by factors such as channel slope and valley morphology. 4.2. Effects of basin-scale variables on SSY and SY The model presented in Eq.8 describes SSY as a function of topographical, climate, and land use characteristics of the catchments. In relation to the topographic characteristics, there is a logical positive correlation between slope and SSY associated to the higher erosive potential of the runoff (i.e. overland flow). The positive correlation of SSY with the minimum height of the basin is probably related to the correlation between altitude and slope. Moreover, sites located in mountain areas are often far from large deposition areas characteristic of the lowlands (i.e. floodplains), therefore increasing SSY. The positive correlation between SSY and ETP is more complex to interpret. A possible explanation lays on the positive correlation (Spearman coefficient: 0.54) between temperature and ETP. Previous studies consider temperature as a representative climate factor, as it is related to the occurrence of extreme convective storms and also to soil formation, soil erosion, runoff from snow melting and freeze-thaw cycles (Syvitsky et al., 2003). Moreover, the slightly negative correlation with the proportion of urban area could be mainly related to the lower sediment availability. Nevertheless, urban areas can increase runoff water velocities, generating timely higher erosion at the local scale. Finally, SSY is positively correlated with specific runoff (runoff per unit area) and with the runoff coefficient (percentage of precipitation that becomes runoff), as a consequence of the runoff erosive potential. Some of the correlations discussed above apply to the SY equation (dependence on Hmean, HD and ETP). The main difference in Eq. 9 is the influence of scale variables as basin area, perimeter and distance across the basin. Total stream length (SL) showed a weak negative correlation with SY, probably as a result of the correlation of SL with catchment size. The negative correlation of the stream order may be associated to the increment of potential depositional areas as the river progresses downstream. Relief ratio (RR; calculated as HD/A) showed a negative correlation that is more difficult to interpret and ETP is positively correlated, probably as a consequence of the positive correlation with temperature. A negative correlation was observed between the aridity index (AI) and the SY. This relation encompasses complex interactions between counteracting effects, that could be summarized with less precipitation (thus lower overland flow) reducing erosivity regardless vegetation abundance (Langbein and Schumm 1958, Douglas 1967). Last, there is a logical positive correlation between SY and runoff. In view of this set of complex and sometimes conflicting results we conclude that factors that mostly drive sediment yield at the catchment scale (SSY) are the average catchment slope, the vegetation coverage and the abundance of extreme events in the rainfall regime. Such variables control the amount of rain that reaches the ground, its intensity, and ultimately the velocity of runoff water, which in turn determines the erosion of the land and the sediment transport capacity of the flow. If we assess total sediment yield, basin scale overcomes previous factors and becomes the main controlling parameter. 4.3. Uncertainty of results A large uncertainty was observed during models validation (Eq. 8 and Eq. 9), potentially due to following factors: Sediment yield has been related to variables describing topographical, geometric, geologic, climate and land use characteristics of the associated catchment. While the first three are constant in time, climate and specially land use patterns may have suffered significant changes during

C. Buendia et al. / Science of the Total Environment 572 (2016) 538–553

the recording period. This situation may be especially important in sites where SY has been obtained from reservoir bathymetric surveys. The type and density of vegetation also has a crucial effect on soil protection against rainfall induced erosion (Casermeiro et al. 2004; Gyssels et al. 2005). Important reforestation processes have occurred in Mediterranean catchments in the last fifty years (García-Ruiz, 2010; Poesen et al. 1997; Rojo-Serrano, 2003). Therefore, associating a specific percentage of urban, agricultural or forest soil to the whole period is probably a rough estimate that likely brings uncertainty into the results. Measurement of sediment yields themselves are also subject to uncertainty. Reservoir bathymetric surveys are usually obtained by measurements which do not cover the whole reservoir surface, and do not consider irregularities in reservoir bed topography (Morris and Fan, 2008). Moreover, limitations in GPS coverage may cause the effective surface of the survey to be significantly smaller than the total reservoir surface, making it necessary to extrapolate deposited volumes along the reservoir contour. Concerning sediment yield obtained from sediment transport records, they are affected by the measurement errors associated with turbidity meter malfunctioning and cross-sectional variability of sediment concentrations (e.g. Lopez-Tarazón et al., 2009). The number and the specific choice of the dependent variables used in the multiple regression analysis is an additional source of uncertainty. Previous studies highlighted the fact that using a higher number of variables in the regression analysis does not necessarily lead to a better performance of the model. Syvitsky et al. (2003) developed a sediment delivery model (Area Relief Temperature model) that was calibrated with only three variables for global rivers. In addition, several factors that have not been taken into account in this study may have produced an impact on the sediment production at the catchment scale. Despite the consideration of land use characteristics in the analysis, other human activities such as the application of conservation strategies have not been taken into account. From a geological point of view, factors such as local lithological features, tectonics, or the rate of rock uplift can have an impact on sediment yield (de Vente et al. 2005; Hovius, 1998; Woodward, 1995). The variability within the dataset may also be a source of scatter. Catchment of significantly different size, climate conditions and land use characteristics were considered herein. The different catchment properties are evidenced in the ordination analysis (NMDS), where two main groups of basins are distinguished following a climate and morphological gradient. Clearly, forested basins in upland and wetter locations showed lower values of sediment yield, in contrast to those located in the lowlands, with drier climates and higher seasonality in rainfall and higher proportion of agricultural areas. The introduction of basins with such contrasting climate and morphological characteristics may influence the results, as different erosional and sedimentary processes govern erosion and sediment dynamics in each of them. The Mediterranean region encompasses highly diverse and contrasted environments, and the inclusion of such range of conditions can weaken the correlation between variables. The divergence in their characteristics probably makes it difficult to fit a single group of calibration parameters that leads to a single equation that succeeds in predicting the sediment yield for the whole group of catchments within an acceptable range of confidence. Regarding climate characteristics, precipitation has been included in the analysis using the mean annual rainfall of the catchment. Previous studies highlight the fact that sediment yield is often related to the contribution of a small number of extreme events due to the nonlinear correlation between water and sediment discharge (Abell et al. 2013; Gonzalez-Hidalgo et al., 2009; Nearing et al. 2004). As we have mentioned above, average temperature is related to the occurrence of convective storms and the timing and importance of snow-melt annual cycles (Syvitsky et al. 2003). Nevertheless, this relation is not direct. The introduction of a parameter that would directly describe high rainfall events (e.g. annual maximum rainfall in 24 h) could lead to an improvement of the results in the regression analysis.

551

5. Final remarks This paper relates river sediment yields to basin-scale characteristics in the western Mediterranean area. Large scatter and statistical divergences were observed in the relation between catchment area and specific sediment yield, showing that using the derived statistical correlation for prediction purposes would be inexact. Also, a range of responses were observed depending on the data and the method used to do the analysis. Despite this, results showed a general pattern where basin size was the limiting factor for the upper range of the specific sediment yield values. This might suggest that even though varying response can be found for a given basin size as a response of the particular characteristics of such basin (i.e. location of sediment sources and sinks and dominance of erosional or depositional processes throughout the basin), the values of specific sediment yield might not exceed a certain threshold for a particular size. Overall, trends in this relationship should be interpreted with care as it does not provide a reliable basis for predicting quantitative changes or trends in sediment yield resulting from changes in the environment. However, it might be useful for providing information on the main sediment controls, the predominant sources and the effectiveness of the sinks, which may aid in the application and development of more complex models. Multiple regression analysis revealed interesting patterns in the relation between sediment yield and catchment properties; for example, there is a positive correlation with basin's slope and seasonality index, and a negative correlation with the percentage of urban areas. Moreover, results indicated a better statistical performance in all the analysis when using total sediment yield instead of specific sediment yield. Model performance was, however, not completely satisfactory, which can be probably explained by a number of factors such as not considering the impact of human strategies to control erosion; the use of a set of catchments with significantly diverse characteristics; changes in catchment characteristics (mainly land use) during the period to which sediment yield measures took place; and errors associated to sediment transport measurements. Such instability of the model prevents a direct unsupervised application to other catchments without further investigation. Acknowledgements This work has been supported by the European Communities 7th Framework Programme Funding under Grant agreement no. 603629ENV-2013-6.2.1-Globaqua. The authors also acknowledge the support from the Economy and Knowledge Department of the Catalan Government through the Consolidated Research Groups: 2014 SGR 645 (RIUS-Fluvial Dynamics Research Group) and 2014 SGR 291 (Catalan Institute for Water Research). Appendix A. Quatitative example of the Sediment Yield correction Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.scitotenv.2016.08.065. References Abell, J.M., Hamilton, D.P., Rutherford, J.C., 2013. Quantifying temporal and spatial variations in sediment, nitrogen and phosphorus transport in stream inflows to a large eutrophic lake. Env. Sci. Process Impacts. 15, 1137–1152. Arnau-Rosalen, E., Calvo-Cases, A., Boix-Fayos, C., Lavee, H., Sarah, P., 2008. Analysis of soil surface component patterns affecting runoff generation. An example of methods applied to Mediterranean hillslopes in Alicante (Spain). Geomorphology 101, 595–606. Arnoldus, H.M.J., 1977. Methodology used to determine the maximum potential average annual soil loss due to sheet and rill erosion in Morocco. FAO Soils Bulletin. Avendaño-Salas, C., Cobo Rayán, R., Sanz Montero, E., Gómez Montaña, J.L., 1997a. Capacity situation in Spanish reservoirs. Dixneuvième Congrès des Grands Barrages. Commission Internationale Des Grands Barrages, Florence, pp. 849–862. Avendaño-Salas, C., Sanz-Montero, E., Gómez Montaña, J.L., 1997b. Sediment yield at Spanish reservoirs and its relationship with the drainage basin area. Dix-neuvième

552

C. Buendia et al. / Science of the Total Environment 572 (2016) 538–553

Congrès des Grands Barrages. Commission Internationale De Grands Barrages, Florence, pp. 863–874. Batalla, R.J., Vericat, D., 2011. An appraisal of the contemporary sediment yield in the Ebro basin. J. Soils Sediments 11, 1070–1081. Batalla, R.J., Garcia, C., Rovira, A., 2005. A decade of sediments transport measurements in a large Mediterranean river (the Tordera, Catalan Ranges, NE Spain). In: Garcia, C., Batalla, R.J. (Eds.), Catchment Dynamics and River Processes: Mediterranean and Other Climate Regions. Developments in Earth surface sciences vol.7. Elsevier, Amsterdam, pp. 17–29. Brune, R.C., 1953. Trap efficiency of reservoirs. Trans. Geophys. Union 34, 407–418 1955). Buendia, C., Gibbins, C., Vericat, D., Batalla, R.J., 2013. Effects of flow and fine sediment dynamics on the turnover of stream invertebrates. Ecohydrology 7, 1105–1123. Buendia, C., Vericat, D., Batalla, R.J., Gibbins, C.N., 2016. Temporal dynamics of sediment transport and transient in-channel storage in a highly erodible catchment. Land Degrad. Dev. 27, 1045–1063. Bussi, G., Francés, F., Horel, E., López-Tarazón, J.A., Batalla, R.J., 2014. Modelling the impact of climate change on sediment yield in a highly erodible Mediterranean catchment. J. Soils Sediments 14, 1921–1937. Cade, B.S., Noon, B.R., 2003. A gentle introduction to quantile regression for ecologists. Reviews. Front. Ecol. Environ. 1, 412–420. Cammeraat, L.H., 2002. A review of two strongly contrasting geomorphological systems within the context of scale. Earth Surf. Process. Landf. 27, 1201–1222. Casalí, J., Gastesi, R., Álvarez-Mozos, J., De Santisteban, L.M., Del Valle de Lersundi, J., Giménez, R., Larrañaga, A., Goñi, M., Agirre, U., Campo, M.A., López, J.J., Donézar, M., 2008. Runoff, erosion, and water quality og agricultural watersheds in central Navarre (Spain). Agric. Water Manag. 95, 1111–1128. Casermeiro, M.A., Molina, J.A., Caravaca, M.T.D.L., Costa, J.H., Massanet, M.I.H., Moreno, P.S., 2004. Influence of scrubs on runoff and sediment loss in soils of Mediterranean climate. Catena 57, 91–107. Church, M., Ham, D., Hassan, M., Slaymaker, O., 1999. Fluvial clastic sediment yield in Canada: scaled analysis. Can. J. Earth Sci. 36, 1267–1280. Clarke, K.R., 1993. Non-parametric multivariate analysis of changes in community structure. Aust. J. Ecol. 18, 117–143. Conacher, A.J., Sala, M. (Eds.), 1998. Land Degradation in Mediterranean Environments of the World: Nature and Extent, Causes and Solutions. John Wiley & Sons, Chichester. Constanza, R., d'Arge, R., de Groot, R., Farberk, S., Grasso, M., Hannon, B., Limburg, K., Naeem, S., O'Neill, R.V., Paruelo, J., Raskin, R.G., Suttonkk, P., van der Belt, M., 1997. The value of the world's ecosystem services and natural capital. Nature 387, 253–260. De Boer, D.H., Crosby, G., 1996. In: Walling, D.E., Webb, B.W. (Eds.), Suspended sediment yield and drainage basin scale, in Erosion and Sediment Yield: Global and Regional Perspective 236. IAHS Publ., pp. 333–338. de Vente, J., 2009. Soil Erosion and Sediment Yield in Mediterranean Geoecosystems. Scale Issues, Modelling and Understanding (PhD thesis) K.U.Leuven, Belgium. de Vente, J., Poesen, J., 2005. Predicting soil erosion and sediment yield at the basin scale: scale issues and semi-quantitative models. Earth Sci. Rev. 71, 95–125. de Vente, J., Poesen, J., Verstraeten, G., 2005. The application of semiquantitative methods and reservoir sedimentation rates for the prediction of basin sediment yield in Spain. J. Hydrol. 305, 63–86. de Vente, J., Verduyn, R., Verstraeten, G., Vanmaercke, M., Poesen, J., 2011. Factors controlling sediment yield at the catchment scale in NW Mediterranean geoecosystems. J. Soils Sediments 11, 690–707. de Vente, J., Poesen, J., Verstraeten, G., Covers, G., Vanmaercke, M., Van Rompaey, A., Arabkhedri, M., Boix-Fayos, C., 2013. Predicting soil erosion and sediment yield at regional scales: where do we stand? Earth-Sciences reviews 127, 16–29. Dedkov, A., 2004. The relationship between sediment yield and drainage basin area. Sediment Transfer through the Fluvial System 288. IAHS Publi. Dedkov, A., Moszherin, V.I., 1992. Erosion and sediment yield in mountain regions of the world. Erosion, Debris Flows and Environment in Mountain Regions, International Association of Hydrological Sciences 209. UK, Publication, Wallingford, Oxon, pp. 29–36. Douglas, I., 1967. Man, vegetation and the sediment yield of rivers. Nature 215, 925–928. Duran-Zuazo, V.H., Francia Martínez, J.R., García Tejero, I., Rodríguez Pleguezuelo, C.R., Martínez Raya, A., Cuadro Tavira, S., 2012. Runoff and sediment yield from a small watershed in southeastern Spain (Lanjarón): implications for water quality. Hydrol. Sci. J. 57, 1610–1625. Estrany, J., Garcia, C., Batalla, R.J., 2009. Groundwater control on the suspended sediment load in the Na Borges river, Mallorca, Spain. Geomorphology 106, 292–303. Farguell, Q., Sala, M., 2005. Seasonal suspended sediment concentrations and their relationship with discharge in the Anoia River basin, NE Spain. Geomorphological Processes and Human Impacts in River basins (Proceedings of the International Conference held in Solsona, Catalonia, Spain, May 2004). IAHS Publications, p. 299. García-Ruiz, J.M., 2010. The effects of land uses on soil erosion in Spain: a review. Catena 81, 1–11. González-Hidalgo, J.C., de Luis, M., Batalla, R.J., 2009. Effects of largest daily events on total soil erosion by rainwater. An analysis of the USLE database. Earth Surf. Process. Landf. 34 (15), 2070–2077. Grauso, S., Pagano, A., Fattoruso, G., De Bonis, P., Onori, F., Regina, P., Tebano, C., 2008. Relations between climate–geomorphological parameters and sediment yield in a Mediterranean semi-arid area (Sicily, southern Italy). Environ. Geol. 5, 219–234. Gyssels, G., Poesen, J., Bochet, E., Li, Y., 2005. Impact of plant roots on the resistance of soils to erosion by water: a review. Prog. Phys. Geogr. 29, 189–217. Haregeweyn, N., Poesen, J., Nyssen, J., Govers, G., Verstraeten, G., de Vente, J., Dekers, S., Moeyersons, J., Haile, M., 2008. Sediment yield variability in Northern Ethiopia: a quantitative analysis of this controlling factors. Catena 75, 65–76. Hovius, N., 1998. Controls on sediment supply by large rivers. In: Shanley, K.W., McCabe, P.J. (Eds.), Relative Role of Eutasy, Climate and Tectonism in Continental Rocks. SEPM Special Publication No. 59, Tulsa, OK, pp. 4–15.

Janssen, K., 2012. Kuczerawy, A. 2012. Increasing the availability of spatial data held by public sector bodies: some experiences and guidelines from the OneGeology-Europe Project. International Journal of Spatial Data Infrastructures Research 7, 249–276. Kirchner, J.W., Finkel, R.C., Riebe, C.S., Granger, D.E., Clayton, J.L., King, J.G., Megahan, W.F., 2001. Mountain erosion over 10 yr, 10 ky, and 10 my time scales. Geology 29 (7), 591–594. Koenker, R., 2016. Quantreg: Quantile Regression. R Package version. 5 p. 24 Available at) https://cran.r-project.org/web/packages/quantreg/index.html. Kruskal, J.B., 1964. Nonmetric multidimensional scaling: a numerical method. Psychometrika 29, 115–129. Lahlou, A., 1988. The silting of Moroccan dams. Sediment Budgets (Proceedings of the Porte Alegre Symposium, December). vol. 174. IAHS Publ., Wallingford, UK, pp. 71–77. Lancaster, J., Belyea, L., 2006. Defining limits to local density: alternative views of abundance-environment relationships. Freshw. Biol. 51, 783–796. Langbein, W.B., Schumm, S.A., 1958. Yield of sediment in relation to mean annual precipitation. Trans. Am. Geophys. Union 39, 1076–1084. Liquete, C., Canals, M., Ludwig, W., Arnau, P., 2009. Sediment discharge of the rivers of Catalonia, NE Spain, and the influence of human impacts. J. Hydrol. 366, 76–88. Lobera, G., Batalla, R.J., Vericat, D., López-Tarazón, J.A., Tena, A., 2016. Sediment transport in two Mediterranean regulated rivers. Sci. Total Environ. 540, 101–113. López-Tarazón, J.A., Batalla, R.J., Vericat, D., Francke, T., 2009. Suspended sediment transport in a highly erodible catchment: the River Isábena (Southern Pyrenees). Geomorphology 109 (3), 210–221. 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. López-Tarazón, J.A., Batalla, R.J., Vericat, D., Francke, T., 2012. The sediment budget of a highly dynamic mesoscale catchment: the river Isábena. Geomorphology 138, 15–28. Ludwig, W., Probst, J.L., 1998. River sediment discharge to the oceans: present-day controls and global budgets. Am. J. Sci. 298, 265–295. Maneux, E., Probst, J.L., Veyssy, E., Etcheber, H., 2001. Assessment of dam trapping efficiency from water residence time: application to fluvial sediment transport in the Adour, Dordogne, and Garonne River basins (France). Water Resour. Res. 37, 801–811. Martínez-Casasnovas, J.A., Guevara Bonilla, M., Ramos, M.C., 2012. Multi-temporal analysis of sediment yield caused by hydric erosion in a basin of the Anoia–Penedès Vineyard region (NE Spain). Cuadernos de investigación científica 38, 95–114. McCune, B., Mefford, M.J., 1995. PCORD. Multivariate Analysis of Ecological Data, Version 2.0. MjM Software Design, Gleneden Beach, Oregon. Migoń, P., 2013. Cultural Heritage and Natural Hazards. Encyclopedia of Natural Hazards, Dordrecht. Milliman, J.D., 2001. Delivery and fate of fluvial water and sediment to the sea: a marine geologist's view of European rivers. Sci. Mar. 65 (Suppl. 2), 121–132. Milliman, J.D., Syvitski, J.P.M., 1992. Geomorphic/Tectonic control of Sediment Discharge to the Ocean: the Importance of Small Mountainous Rivers. J. Geol. 100, 525–544. Milly, P.C.D., Dunne, K.A., Vecchia, A.V., 2005. Global pattern of trends in streamflow and water availability in a changing climate. Nature (438/17), 347–350. Morris, G.L., Fan, J., 2008. Reservoir Sedimentation Handbook. 1.04 ed. McGraw-Hill Book Co., New York. Navas, A., Machin, J., Begueria, S., Lopez-Vicente, M., Gaspar, L., 2007. Soil properties and physiographic factors controlling the natural vegetation re-growth in a disturbed catchment of the Central Spanish Pyrenees. Agrofor. Syst. 72, 173–185. Nearing, M.A., Pruski, F.F., O'Neal, M.R., 2004. Expected climate change impacts on soil erosion rates: a review. J. Soil Water Conserv. 59, 43–50. Oksanen, J., Blanchett, F.G., Kindt, R., Legendre, P., Minchin, P.R., O'Hara, R.B., Simpson, G.L., Solymos, P., Stevens, M.H.M., Wagner, H., 2012. Vegan: community Ecology Package. Oliver, J.E., 1980. Monthly precipitation distribution: a comparative index. Prof. Geogr. 32, 300–309. Pacheco, E., Farguell, J., Úbeda, X., Outeiro, L., Miguel, A., 2011. Runoff and sediment production in a Mediterranean basin under two different land uses. Cuaternario y Geomorfología 25, 103–114. Pacheco, F.A.L., Varandas, S.G.P., Sanches Fernandes, L.F., Valle Junior, R.F., 2014. Soil losses in rural watersheds with environmental land use conflicts. Sci. Total Environ. 485486C, 110–120. Palazón, L., Navas, A., 2014. Modeling sediment sources and yields in a Pyrenean catchment draining to a large reservoir (Ésera River, Ebro Basin). J. Soils Sediments 14, 1612–1625. Pimentel, D., 2006. Soil erosion: a food and environmental threat. Environ. Dev. Sustain. 8, 119–137. Poesen, J., van Wesemael, B., Govers, G., Martinez-Fernandez, J., Desmet, P., Vandaele, K., Quine, T., Degraer, G., 1997. Patterns of rock fragment cover generated by tillage erosion. Geomorphology 18, 183–197. Praskievicz, S., Chang, H.J., 2009. A review of hydrological modelling of basin-scale climate change and urban development impacts. Prog. Phys. Geogr. 33, 650–671. Quesada, S., Tena, A., Guillén, D., Ginebreda, A., Vericat, D., Martínez, E., Navarro-Ortega, A., Batalla, R.J., Barceló, D., 2014. Dynamics of suspended sediment borne persistent organic pollutants in a large regulated Mediterranean river (Ebro, NE Spain). Sci. Total Environ. 473-474, 384–390. Regués, D., Balasch, C., Castelltort, X., Soler, M., Gallart, F., 2000. Relación entre las tendencias temporales de producción y transporte de sedimentos y las condiciones climáticas en una pequeña cuenca de montaña mediterránea (Vallcebre, Pirineos Orientales). Cuadernos de investigación científica 26, 41–65. Renard, K.G., Freimund, J.R., 1994. Using monthly precipitation data to estimate the R-factor in the revised USLE. J. Hydrol. 157, 287–306. Renwick, W.H., Carlson, K., Hayes-Bohanan, J., 2005a. Trends in recent reservoir sedimentation rates in southwestern Ohio. J. Soil Water Conserv. 60, 72–79.

C. Buendia et al. / Science of the Total Environment 572 (2016) 538–553 Reuter, H.I., Nelson, A., Jarvis, A., 2007. An evaluation of void filling interpolation methods for SRTM data. Int. J. Geogr. Inf. Sci. 21, 983–1008. Robinson, D.A., Blackman, J.D., 1990. Soil erosion and floding: consequences on land use policy and agricultural practice on the South Downs, East Sussex, UK. Land Use Policy 7, 41–52. Rodríguez-Blanco, M.L., Taboada-Castro, M.M., Taboada-Castro, M.T., 2013. Linking the field to the stream: soil erosion and sediment yield in a rural catchment, NW Spain. Catena 103, 74–81. Rojas, R., Woolhiser, D.A., 2000. Erosion parameter identifiability in the KINEROS model. Proceedings of the Debris Flow and Disaster of December 1999 in Venezuela, Dynamics of Debris Flow Section. Universidad Central de Venezuela, Cara cas, Venezuela (CD Rom). Rojo-Serrano, L., 2003. Description of the Land Use Changes in Relation to Type of Crops, Landscapes and Physiographic Units Involved in the Guadalen Basin. Deliverable 3.3a. DESERTLINKS, EVK2-CT-2001-00109. Romero Díaz, M.A., Cabezas, F., López Bermúdez, F., 1992. Erosion and fluvial sedimentation in the River Segura basin (Spain). Catena 19, 379–392. Sanz-Montero, M.E., Avendaño, C., Cobo, R., Gómez, J.L., 1998. Determinación de la tasa de erosión en la cuenca del Segura a partir de los sedimentos acumulados en sus embalses. Geogaceta 23, 135–138. Schoorl, J.M., Veldkamp, A., 2001. Linking land use and landscape process modelling: a case study for the Alora region (SE Spain). Agric. Ecosyst. Environ. 85, 281–292. Segarra, E., Ervin, T.R., Dicks, M.R., Taylor, D.B., 1991. On-site and off-site impacts of soil erosion: their implications for soil conservation policy. Resour. Conserv. Recycl. 5, 1–19. Serrat, P., 1999. Dynamique sédimentaire actuelle d'un système fluvial méditerranéen: l'Agly (France). Earth Planet. Sci. 329, 189–196. Serrat, P., Ludwig, W., Navarro, B., Blazi, J.L., 2001. Spatial and Temporal Variability of Sediment Fluxes From a Coastal Mediterranean River: The Tet (France). Comptes Rendus De L Academie Des Sciences, Serie II, Fascicule a-Sciences De La Terre Et Des Planetes. 333 pp. 389–397. Shao, J.X., Tu, D., 1995. The Jakknife and Bootstrap. Springer, New York. Slaymaker, O., Souch, C., Menounos, B., Filippelli, G.M., 2003. Advances in Holocene mountain geomorphology inspired by sediment budget methodology. Geomorphology 55, 305–316. Stone, M., 2000. The role of erosion and sediment transport in nutrient and contaminant transfer. Proceedings of the Waterloo Symposium, Wallingford, UK, International Association for Hydrological Sciences. IAHS. Publication No. 263 307 pp.). Strand, R.I., 1975. Bureau of Reclamation procedures for predicting sediment yield. Present and Prospective Technology for Predicting Sediment Yield and Sources. Proceedings of the Sediment Yield Workshop, USDA Sedimentation Laboratory, Oxford, Mississippi, November 28–30, 1972, ARS-S-40. USDA-Agricultural Research Service, Washington, DC, pp. 10–15.

553

Swinton, S.M., Lupi, F., Robertson, G.P., Hamilton, S.K., 2007. Ecosystem services and agriculture: cultivating agricultural ecosystems for diverse benefits. Ecol. Econ. 64, 245–252. Syvitski, J.P.M., Peckham, S.D., Hilberman, R., Mulder, T., 2003. Predicting the terrestrial flux of sediment to the global ocean: a planetary perspective. Sediment. Geol. 162, 5–24. Tena, A., Batalla, R.J., 2013. The sediment budget of a large river regulated by dams (The lower Ebro river, NE Spain). J. Soils Sediments 13, 966–980. Tuset, J., Vericat, D., Batalla, R.J., 2016. A water and sediment budget for a Mediterranean mountainous catchment (Southern Pyrenees). Geophysical Research Abstracts. Vol, 18. EGU2016–8373-1. EGU General Assembly, 2016 (Vienna). UNEP: United Nations Environment Programme, 2006. Mediterranean Action Plan. Riverine Transport of Water, Sediments and Pollutants to the Mediterranean Sea. MAP Technical Reports Series No.141. Valle Junior, R.F., Varandas, S.G.P., Sanches Fernandes, L.F., Pacheco, F.A.L., 2014. Environmental land use conflicts: a threat to soil conservation. Land Use Policy 41, 172–185. Verstraeten, G., Poesen, J., 2001. Factors controlling sediment yield from small intensively cultivated catchments in a temperate humid climate. Geomorphology 40, 123–144. Verstraeten, G., Poesen, J., De Vente, J., Koninckx, X., 2003. Sediment variability in Spain: a quantitative and semi-quantitative analysis using reservoir sedimention rates. Geomorphology 50, 327–348. Wainwright, J., Parsons, A., 1998. Sensitivity of sediment-transport equations to errors in hydraulic models of overland flow. In: Boardman, J., Favis-Mortlock, D. (Eds.), Modelling Soil Erosion by Water. Volume 55 of the series NATO ASI series, pp. 271–284. Wainwright, J., Parsons, A.J., Müller, E.N., Brazier, R.E., Powell, D.M., Fenti, B., 2008. A transport-distance approach to scaling erosion rates: 1. Background and model development. Earth Surf. Process. Landf. 33 (5), 813–826. Walling, D.E., 1983. The sediment delivery problem. J. Hydrol. 65, 209–237. Walsh, R.P.D., Lawler, D.M., 1981. Rainfall seasonality: description, spatial patterns and change through time. Weather 36, 201–208. Ward, P.R.B., 1980. Sediment transport and a reservoir siltation formula for ZimbaweRhodesia, Die Siviele Ingenier in Suid-Afrika, pp. 9–15, January 1980 (2007) In: Westrich, B., Förstner, U. (Eds.), Sediment Dynamics and Pollutant Mobility in River—An Interdisciplinary Approach. Springer, New York 430pp.). Westrich, B., Förstner, U., 2007. (Eds). Sediment dynamics and pollutant mobility in rivers—an interdisciplinary approach. J. Soils Sediments 8, 151–153. Woodward, J.C., 1995. Patterns of erosion and suspended sediment yield in Mediterranean river basins. In: Foster, I.D.L., Gurnell, A.M., Webb, B.W. (Eds.), Sediment and Water Quality in River Catchments. Wiley, Chichester, pp. 365–389. Zabaleta, A., Martínez, M., Uriarte, J.A., Antigüedad, I., 2007. Factors controlling suspended sediment yield during runoff events in small headwater catchments of the Basque country. Catena 71, 179–190.