In extensional provinces with low-angle normal faulting (such as the Aegean region), both tectonic processes and erosion induce landscape change, but their interaction during the evolution of topography and relief accompanying continental extension has rarely been addressed. Here we present local and catchment-wide 10Be erosion rates that document the spatial pattern of erosion in the central Menderes Massif, a metamorphic core complex consisting of two asymmetric mountain ranges (Bozdağ and Aydın) bound by detachment faults and active grabens. Catchment-wide erosion rates on the northern flank of the Bozdağ Range are rather low (40–110 mm/k.y.) but reach values of >300 mm/k.y. on the steep southern escarpment—a pattern that reflects both topography and bedrock lithology. In the Aydın Range, erosion rates are generally higher, with mean erosion rates of ∼190 and ∼260 mm/k.y. on the northern and southern flank, respectively, and more variable along strike. In both ranges, erosion rates of ridge crests derived from amalgamated clasts are 30–90 mm/k.y. The difference between local and catchment-wide erosion rates indicates that topographic relief increases in most parts of the massif in response to ongoing fault-related uplift and concomitant river incision. Our findings document that tectonic processes exert a significant control on landscape evolution during active continental extension and are reflected in both the topographic signature and the spatial pattern of erosion. In the Menderes Massif, rock susceptibility to weathering and erosion is a dominant factor that controls the erosional contribution to rock exhumation, which varies spatially between ∼10% and ∼50%.
In tectonically active regions, tectonic and surface processes act in concert to shape Earth’s landscape (e.g., Willett et al., 2006; Bishop, 2007). These processes include the activity of faults, as well as erosion, sediment transport, and deposition and ultimately lead to the exhumation of rocks. In this context, a fundamental difference between regions undergoing crustal shortening or extension exists. While in regions of shortening, rock exhumation occurs exclusively by erosion, in extensional environments, both tectonic denudation (i.e., normal faulting) and erosion contribute to the exhumation of rocks (England and Molnar, 1990; Ring et al., 1999). In regions where extension has been predominantly accommodated by low-angle detachment faulting, tectonic denudation has often been considered to be the dominant mechanism for rock exhumation (e.g., Foster and John, 1999; Ring et al., 1999; Brichau et al., 2008). On the other hand, erosional denudation exerts an important control on rock exhumation and landscape evolution in mountains bounded by high-angle normal faults (e.g., Harbor, 1997; Armstrong et al., 2003). In such ranges, 10Be-based erosion rates tend to be higher at the center of faults and lower near their tips, although the rates often vary considerably along strike (e.g., Densmore et al., 2009; Stock et al., 2009; Rossi et al., 2017; Roda-Boluda et al., 2019). In contrast to mountains bounded by high-angle normal faults, rates of erosion and the concomitant development of topography and relief have rarely been quantified in regions with low-angle detachment faults (Reinhardt et al., 2007a; Buscher et al., 2013).
In the Aegean extensional province, extension and detachment faulting led to the exhumation of metamorphic rocks from mid-crustal levels in several metamorphic massifs (e.g., the Rhodope massif, the Attic-Cycladic massif, and the Menderes Massif; Fig. 1) (e.g., Gautier et al., 1993; Hetzel et al., 1995a; Grasemann et al., 2012; Jolivet et al., 2013). In this connection, the Menderes Massif in western Turkey was recognized as the first example of bivergent extension, i.e., the exhumation of rocks occurred along two detachments faults with opposite sense of shear (Gessner et al., 2001). This finding and the fact that the Menderes Massif is entirely located above sea level render it ideally suited to investigate the interplay between tectonic and surface processes and their relative contribution to exhumation in an extensional environment. So far, the contribution of erosion to rock exhumation has only been investigated along a narrow transect through the central part of the Menderes Massif by combining 10Be catchment-wide erosion rates with low-temperature thermochronology (Buscher et al., 2013; Wölfler et al., 2017). However, to what extent erosion varies across the entire central Menderes Massif and how ongoing crustal extension and erosional denudation control the landscape evolution of the massif is still largely unknown.
Cosmogenic nuclides enable the quantification of erosion rates on local and catchment scales in various geological settings (e.g., Brown et al., 1995; Small et al., 1997; Hancock and Kirwan, 2007; Meyer et al., 2010; Portenga and Bierman, 2011). Furthermore, the combination of spatially averaged, catchment-wide erosion rates with local rates of erosion at the outcrop scale can place constraints on the development of topography and local relief (Meyer et al., 2010; Wolff et al., 2018). Combined with digital elevation data, cosmogenic nuclide-derived erosion rates can further be used to deduce the erosional pattern of entire mountain ranges and thus aid the identification of processes that drive landscape evolution on various spatial scales. In this context, the development of new geomorphologic tools to analyze the dynamic state of river networks, river profiles, and drainage divide motion as well as the relationships between topographic parameters and erosion have proven highly valuable in identifying tectonic processes that shape Earth’s surface (e.g., DiBiase et al., 2010; Kirby and Whipple, 2012; Perron and Royden, 2013; Willett et al., 2014).
Here, we present local 10Be erosion rates for 14 ridge crests and spatially averaged erosion rates for 31 catchments in the central Menderes Massif to resolve the spatial pattern of erosion across the massif. These data allow us to draw conclusions on the development of local relief during the ongoing extension and—in combination with the analysis of river networks—enable us to infer processes that have shaped the morphology of the central Menderes Massif during continental extension.
The Aegean region and western Turkey are located in the backarc of the Hellenic subduction zone (Fig. 1) and, during the past 25 m.y., have experienced pronounced extension caused by the progressive rollback of the subducting African plate (e.g., Jolivet and Faccenna, 2000; Brun and Sokoutis, 2010; Jolivet et al., 2013). The extensional deformation, aided by erosional denudation, led to the exhumation of mid-crustal rocks, which are exposed in several metamorphic massifs in Greece, Bulgaria, and western Turkey (Fig. 1). GPS data show that extension in this entire region is still ongoing (Reilinger et al., 2010). At present, western Turkey—with the Menderes Massif in its central part—is extending at a rate of ∼20 mm/yr in a north-south direction (Aktuğ et al., 2009).
The Menderes Massif consists of three submassifs, which are separated by two major east-west–striking grabens—the Gediz Graben and the Büyük Menderes Graben (e.g., Seyitoğlu and Scott, 1991; Yılmaz et al., 2000; Gürer et al., 2009; Çiftçi and Bozkurt, 2010) (Fig. 2A). The long-lasting history of extension of the Menderes Massif started with a first phase of extension during the late Oligocene and the early Miocene, when the rocks of the northern submassif were largely exhumed by normal faulting (e.g., Işık and Tekeli, 2001; Thomson and Ring, 2006; Ersoy et al., 2010). Subsequently, the metamorphic rocks of the central submassif were exhumed in the footwalls of two low-angle normal faults—the Gediz and Büyük Menderes detachments—which strike east-west and dip to the north and south, respectively (e.g., Hetzel et al., 1995a, 1995b; Emre and Sözbilir, 1997; Bozkurt and Oberhänsli, 2001; Gessner et al., 2001, 2013) (Figs. 2B and 2C). The Gediz detachment operated with a top-to-the-N to NNE shear sense, whereas the shear sense of the Büyük Menderes detachment is top-to-the-S to SSW (Hetzel et al., 1995b; Emre, 1996; Gessner et al., 2001; Işık et al., 2003). Both detachments are cut by the high-angle normal faults of the Gediz and Büyük Menderes grabens, which are seismically active as documented by raised fluvial terraces, fault scarps, and earthquakes (Ambraseys, 1971; Eyidoğan and Jackson, 1985; Altunel, 1999; Çiftçi and Bozkurt, 2010). Sedimentological and stratigraphic data indicate that the transition from low-angle detachment faulting to high-angle normal faulting occurred in the latest Pliocene to early Quaternary (e.g., Purvis and Robertson, 2005; Oner and Dilek, 2011).
The central Menderes Massif comprises two mountain ranges, the Bozdağ and the Aydın range, which are separated by the Küçük Menderes Graben (Fig. 2A). Both mountain ranges are characterized by a topographic asymmetry with rather steep mountain flanks facing the Küçük Menderes Graben and shallow-dipping flanks on the sides of the two detachment faults (Fig. 2B). The Gediz detachment along the northern side of the Bozdağ Range is exceptionally well preserved owing to highly resistant cataclasites with a thickness of 20–50 m and the widespread presence of quartz-rich mylonites underneath (Hetzel et al., 1995b; Emre, 1996; Buscher et al., 2013). Therefore, streams incising the Gediz detachment form narrow valleys with steep hillslopes (Figs. 3A and 3B). The Büyük Menderes detachment along the southern flank of the Aydın Range is less well preserved, because the cataclasites associated with this fault are less than 5 m thick (Emre and Sözbilir, 1997; Hetzel et al., 2013; observations made during fieldwork of this study). In addition, the dominant lithologies underneath this detachment are mica schists and phyllites, which are susceptible to weathering and erosion, while resistant quartzites are absent (Wölfler et al., 2017). As a consequence, most valleys at the Büyük Menderes detachment are broader, yet deeply incised, and exhibit planar hillslopes (Fig. 3C). Apart from the metamorphic rocks, two belts of Neogene sedimentary rocks extend along the northern and southern margin of the central Menderes Massif (Fig. 2A). These strata are well exposed due to footwall uplift and ongoing normal faulting along the main boundary faults of the Gediz and Büyük Menderes grabens.
The metamorphic rocks of the central Menderes Massif constitute a nappe pile that formed during Late Cretaceous to Eocene plate convergence and crustal shortening along the Izmir-Ankara suture zone (e.g., Şengör et al., 1984; Ring et al., 1999; van Hinsbergen et al., 2010; Gessner et al., 2013). The Menderes nappes include, from lowest to highest structural level, the Bayındır, Bozdağ, Çine, and Selimiye nappes, which in turn are overlain by the Cycladic Blueschist Unit in the west (Ring et al., 1999; Gessner et al., 2013) (Figs. 2A and 2C). The Gediz and Büyük Menderes detachments cut obliquely through the nappe pile, and during their activity, slivers of Çine nappe were emplaced as klippen against rocks of the Bayındır nappe in their footwalls (Fig. 2A) (Hetzel et al., 1995b; Buscher et al., 2013; Wölfler et al., 2017). The boundaries of the tectonometamorphic units define an east-west–trending synform, which developed during bivergent north-south extension (Gessner et al., 2001). Together with the detachment faults, this synform appears to control the present-day topography of the submassif and in particular the escarpments facing the Küçük Menderes Graben (Fig. 2C).
The rocks of the Bayındır and Bozdağ nappes constitute the dominant lithological units in the central Menderes Massif. These nappes mainly consist of greenschist- to amphibolite-facies two-mica schists, phyllites, and quartzites, with minor amounts of marble and local amphibolite layers (Dora et al., 1990; Hetzel et al., 1998; Candan et al., 2011). In contrast, the structurally higher Çine and Selimiye nappes are made up of orthogneisses, pelitic gneisses, as well as minor amounts of metapelites and granites (e.g., Hetzel and Reischmann, 1996; Hetzel et al., 1998; Ring et al., 1999; Candan et al., 2001). These nappes mainly occur in the eastern part of the submassif near the Küçük Menderes Graben or as tectonic klippen on the two detachment faults (Fig. 2A).
Recent studies that employed low-temperature thermochronology revealed a two-stage history of cooling and detachment faulting for the central Menderes Massif (Gessner et al., 2001; Ring et al., 2003; Buscher et al., 2013; Wölfler et al., 2017; Nilius et al., 2019) (Fig. 2). The first phase occurred in the middle Miocene (ca. 16 to ca. 10 Ma), whereas the second phase lasted from the latest Miocene to the late Pliocene–early Quaternary (ca. 6 to ca. 2 Ma). The second phase was most pronounced in the eastern part of the Bozdağ Range (i.e., in the region south of Salihli town; see Fig. 4) as evident from a phase of rapid slip (∼4 mm/yr) along the Gediz detachment between 4 and 2 Ma (Buscher et al., 2013) and Plio-Quaternary cooling ages from detrital apatite fission-track analyses (Asti et al., 2017). This interpretation is consistent with the fact that the highest topography of the entire Menderes Massif occurs in this region and with the observation that the Gediz detachment is particularly well preserved here.
Catchment-Wide and Local Erosion Rates from Cosmogenic 10Be
Sampling Strategy, Sample Preparation, and Calculation of Erosion Rates
For this study, we measured the 10Be concentration of 31 stream sediment samples and 14 samples from ridge crests to derive both catchment-wide and local erosion rates across the entire central Menderes Massif (Fig. 4). We also include the catchment-wide 10Be erosion rates previously published by Buscher et al. (2013), Wölfler et al. (2017), and Heineke et al. (2017) (Fig. 4, gray sample numbers). We collected our sediment samples along strike of the Bozdağ and Aydın ranges from streams that flow into the Gediz, Küçük Menderes, and Büyük Menderes grabens, respectively. Each sediment sample was taken from several points along the respective stream over a distance of 20–60 m. Most samples are from rivers that exclusively drain metamorphic rocks. Yet, four samples (14T3, 15T15, 16T1, and 16T4) were taken from streams that flow through the Neogene sedimentary rocks exposed along the northern and southern margins of the massif (Fig. 2A). Note that the erosion rate derived from sample 16T4 has been published by Heineke et al. (2017).
To quantify the erosion of ridge crests, we took samples consisting of 1500–2600 bedrock clasts with a length of 1–3.5 cm on 14 mountain crests (Fig. 4). Amalgamated samples of clasts or grus were used previously to quantify local erosion rates and to constrain changes in topographic relief due to river incision (Small et al., 1999; Meyer et al., 2010; Strobl et al., 2012; Wolff et al., 2018). This sampling approach builds on results of previous studies, which have shown that erosion rates for bedrock outcrops may differ significantly from the actual long-term erosion rate because bedrock erosion mainly occurs by the episodic removal of small blocks or chips (Small et al., 1997, 1999; Anderson, 2002; Reinhardt et al., 2007b; Muzikar, 2008). For example, Small et al. (1997) showed that the apparent bedrock erosion rate determined after removal of a 0.4-m-thick chip deviates from the mean erosion rate by up to +50% or –25%, respectively, depending on the sampling time within the chipping cycle. Model calculations by Reinhardt et al. (2007b) indicate that for a chip thickness of 0.1–0.2 m, ∼100 or more samples are needed to reduce the standard error induced by the sampling to <1%. For these reasons, amalgamated clast samples are well suited to obtain accurate long-term erosion rates at the outcrop scale (e.g., Small et al., 1999; Meyer et al., 2010).
We apply the approach of amalgamating clasts to ridge crests in the Bozdağ and Aydın mountain ranges, which generally exhibit smooth, convex profiles and vary in width from a few to several tens of meters (Fig. 5). Commonly, the crests are sparsely vegetated with grass, shrubs, or small trees, and the surface is covered by an up to 10-cm-thick soil or regolith layer containing variable amounts of bedrock clasts (Fig. 5B). Intervening patches of bedrock outcrop occur either as bedrock knobs, sticking out from the surroundings by a few decimeters (Fig. 5C), or are nearly flush with the surface (Fig. 5D). We collected the clasts for all amalgamated samples from the thin soil or regolith on the highest, subhorizontal part of the ridge crests but avoided bedrock outcrops and quartz clasts derived from resistant quartz veins (Fig. 5D). We did not sample cultivated ridges, ridge crests with fire lanes (which occur in the western Bozdağ Range), and sites where wind mills have been constructed (Fig. 5A). The clasts for each sample were taken over distances of 40–100 m to obtain representative erosion rates over a length of several tens of meters along each crest. Depending on crest width, this procedure resulted in sampling areas of ∼200 to ∼1000 m2 and 1500–2600 clasts per sample (i.e., two to seven clasts per square meter). Although fewer clasts may have been sufficient to calculate erosion rates, we argue that a larger clast number increases the probability of obtaining a representative erosion rate as explained above (cf. Small et al., 1997; Reinhardt et al., 2007b).
The chemical separation of Be from the amalgamated clast and stream sediment samples was carried out at the cosmogenic nuclide laboratory of the University of Münster. In a first step, all bedrock clasts from the ridge-crest samples were crushed. Then all samples were sieved and washed. Subsequently, the 250–500 μm grain-size fraction of all samples was split into a magnetic and a non-magnetic fraction using a Frantz magnetic separator. For two samples from ridge crests (16T36 and 16T38), we also used the 125–250 µm grain-size fraction because these samples comprised fine-grained phyllites and contained insufficient quantities of quartz in the coarser fraction. The subsequent chemical leaching procedure consisted of one etching step in 6 M HCl at 80 °C, four subsequent etching steps in dilute HF/HNO3 in a heated ultrasonic bath (Kohl and Nishiizumi, 1992), and two alternating etching steps in aqua regia and 8 M HF to obtain pure quartz (Goethals et al., 2009). Samples 16T36, 16T38, 17T7, and 17T8 consisted of phyllitic schists and were only etched once in dilute HF/HNO3 and then treated with fluorosilicic acid (H2SiF6, 30%) to enrich quartz. After chemical cleaning of quartz, ∼0.3 mg of Be carrier was added to all samples. Following complete dissolution of quartz in HF (40%), the samples were redissolved and converted into chloride form using 6 M HCl. Beryllium was separated using successive anion and cation exchange columns and precipitated as Be(OH)2 at pH 8–9. Following the transformation to BeO at 1000 °C and target preparation for accelerator mass spectrometry (AMS), 10Be was analyzed at the AMS facility “TANDY” of the ETH Zurich (Christl et al., 2013). Local and catchment-wide erosion rates were calculated from the blank-corrected 10Be concentrations with version 2.3 of the CRONUS-Earth online calculator (Balco et al., 2008; http://hess.ess.washington.edu) using the time-invariant production rate scaling model of Lal (1991) and Stone (2000). Note that we did not use a topographic shielding factor for calculating the catchment-wide erosion rates because it was shown recently that this correction is not required (DiBiase, 2018).
River Network Analysis
Normalized Channel Steepness
In order to derive a suitable reference concavity for streams in the central Menderes Massif, we analyzed chi-elevation plots of the sampled catchments and deduced the respective concavity index for the trunk channel of each catchment (cf. Perron and Royden, 2013). Subsequently, we calculated the mean from all concavity indices, which yielded a reference concavity of 0.4 ± 0.2 (2σ) for the study area. Finally, normalized steepness indices were determined for river segments of 1 km length for streams draining the Bozdağ and Aydın mountain ranges.
Mapping of Chi (χ) Values for Stream Networks
Drainage divides constitute dynamic geomorphic features that migrate (either progressively or by discrete river capture) due to different rates of channel erosion on opposite sides of the divide (Gilbert, 1877; Bishop, 1995; Clark et al., 2004; Bonnet, 2009; Prince et al., 2011; Perron et al., 2012). Hence, the assessment of drainage divide stability and channel equilibrium conditions can serve as a quantitative criterion for the evaluation of landscape (dis) equilibrium and constitutes a valuable tool to infer temporal changes in river networks on various spatial scales.
We generated maps of chi for the central Menderes Massif using TopoToolbox, a MATLAB-based program for the analysis of digital elevation models (Schwanghart and Scherler, 2014). We restricted our analysis to rivers that drain the slopes of the Bozdağ and Aydın mountain ranges, respectively, and did not consider the large streams in the three major grabens.
Erosion Rates from Cosmogenic 10Be
In the Bozdağ Range, catchment-wide erosion rates range from ∼40 to ∼370 mm/k.y., with distinct differences between the northern and southern flank of the mountain range (Table 1; Figs. 6 and 7). Catchments on the northern range flank mostly erode at rates of ∼40–110 mm/k.y., except for two catchments in the central part of the range, which yielded higher rates (Fig. 7A). On the southern flank, high erosion rates of >350 mm/k.y. occur at the steep escarpment south of Bozdağ Mountain and decrease toward the east and west (Fig. 7B). In the Aydın Range, most of the 10Be erosion rates are relatively high (∼150 to ∼430 mm/k.y.) with considerable variations along strike of the range (Figs. 6, 7C, and 7D). Two small catchments in Neogene sedimentary rocks NW of Aydın town yielded erosion rates of ∼84 and ∼108 mm/k.y., whereas two similar catchments SE of Salihli erode at higher rates of ∼206 and ∼331 mm/k.y.
The local 10Be erosion rates for ridge crests fall mainly in the range from ∼30 to ∼90 mm/k.y. and are similar in the Bozdağ and Aydın mountains (Table 2; Fig. 6). These rates are significantly lower than most catchment-wide erosion rates, with the exception of the western Bozdağ Range, where local and catchment-wide erosion rates are similar. In the Aydın Range, two samples yielded quite different rates of 13 ± 1 and 165 ± 20 mm/k.y., respectively. The low erosion rate of ∼13 mm/k.y. for sample 17T2 in the west can be attributed to the resistant bedrock lithology of the sampled crest, which is located in augen gneisses of the Çine nappe. The sample with the rate of ∼165 mm/k.y. was taken just above the Büyük Menderes detachment in an area with highly faulted and weathered mica schists, which may explain this rather high rate.
Channel Steepness and Chi (χ) Values in the Stream Network
The spatial pattern of normalized channel steepness for streams that drain the northern slope of the Bozdağ Range reveals particularly steep channels in the eastern part of the mountain range (Fig. 8). These steep channel segments, with ksn values of ≥80, typically stretch over a length of 4–7 km and can be traced along strike of the Gediz detachment, occurring at approximately the same position in adjacent stream networks (Fig. 8). In a westerly direction, the channel steepness decreases, and most trunk channels in the central and western part of the Bozdağ Range have ksn values below 80. On the southern slope of the Bozdağ Range, steep channel segments with ksn values of 80–130 are restricted to the area near Bozdağ peak, whereas farther east and west, channels are generally less steep (Fig. 8). Along the northern slope of the Aydın Range, steep channel segments occur only in the upstream part of rivers that flow into the Küçük Menderes Graben. On the southern slope of the Aydın Range moderately steep channels occur in the center of the mountain range, and ksn values generally decrease to the west and east.
The cross-divide pattern of chi values changes significantly along strike of the Bozdağ Range. In the eastern part of the range, a marked difference in chi values between the northern and southern slope of the mountain range exists (Fig. 9). This difference in chi across the drainage divide disappears in a westerly direction, and in the westernmost part of the Bozdağ Range, chi values of streams are similar on both sides of the drainage divide. In the Aydın Range, the cross-divide differences in chi values are generally less pronounced than in the Bozdağ Range, with streams flowing northward into the Küçük Menderes Graben having lower chi values than streams south of the drainage divide.
Our new catchment-wide and local 10Be erosion rates as well as river network analyses show that there are considerable differences in the spatial patterns of erosion in the central Menderes Massif. In the following, we first evaluate the spatially averaged erosion rates for catchments in the Bozdağ and Aydın ranges. In the second section, local erosion rates of ridge crests and the development of topographic relief in the central Menderes Massif are discussed. The imprint of normal faulting on landscape morphology is evaluated in the third section, which also includes a discussion on the distribution of channel steepness and on drainage divide mobility. Finally, we compare our 10Be-based erosion rates with exhumation rates derived from low-temperature thermochronology to estimate the relative importance of erosion and normal faulting on landscape evolution during continental extension.
Footwall Erosion along the Gediz and Büyük Menderes Detachments
The present-day topography of the central Menderes Massif is the result of a long history of extension, normal faulting, rock uplift, and erosion. The similar structural and topographic asymmetry of the Bozdağ and Aydın mountain ranges is related to their exhumation history in the footwall of the Gediz and Büyük Menderes detachments and the formation of the synform near the Küçük Menderes Graben (Fig. 2). Still, catchment-wide erosion rates in the footwalls of the two detachment faults are markedly different (Fig. 6). We interpret the low erosion rates on the northern slope of the Bozdağ Range to result from the presence of resistant 20–50-m-thick cataclasites and quartz-rich mylonites in the footwall of the Gediz detachment (Hetzel et al., 1995b; Emre, 1996). These rocks protect the detachment from erosion due to their high rock strength (Figs. 3A and 3B). The rather low mean hillslope angles of the respective catchments (16°–25°; Fig. 7A) are an expression of the gently dipping detachment surface. The mean erosion rate for all catchments on the northern mountain slope is 111 ± 22 mm/k.y. (Fig. 7A). Only two catchments in the central portion of the range erode at markedly higher rates of ∼223 and ∼300 mm/k.y. (Fig. 6). The high erosion rates for these catchments appear to be caused by the presence of weak phyllites and mica schists and the absence of quartz-rich mylonites. Moreover, because the detachment has a very gentle dip in this region, and the catchments are characterized by flat upstream areas, the high erosion rates may mainly reflect the incision of the trunk channels. However, to evaluate this issue in detail, an analysis of drainage basin connectivity would be required, which is beyond the scope of the present study.
In the Aydın Range, catchment-wide erosion rates are quite variable on both sides of the mountain range and—unlike in the Bozdağ Range—do not reflect its topographic asymmetry (Fig. 6). Catchments along the northern flank of the Aydın Range erode at a mean rate of 189 ± 30 mm/k.y., whereas the average erosion rate on the shallower southern slope is 262 ± 37 mm/k.y. (Figs. 7C and 7D). We attribute the rapid erosion of the footwall of the Büyük Menderes detachment (as compared to the Gediz detachment) to be related to the predominance of weak phyllites and mica schists (as opposed to quartz-rich mylonites at the Gediz detachment) and to the thin (<5 m) cataclasites of the detachment (Emre and Sözbilir, 1997; Hetzel et al., 2013; observations made during fieldwork of this study), which do not shield the underlying rocks from erosion. Owing to the susceptibility of the mica schists and phyllites to erosion, the Büyük Menderes detachment is not as well preserved as the Gediz detachment (Wölfler et al., 2017). The considerable along-strike variations in erosion rates for catchments on both sides of the Aydın Range are difficult to explain. In general, such spatial variations may be caused by variations in bedrock quartz content (e.g., Carretier et al., 2015), incomplete mixing of sediment in the catchment, non-uniform erosion through time (e.g., due to the occurrence of debris flows; Kober et al., 2012), or a violation of the steady-state erosion assumption (e.g., Heimsath, 2006; Dunai, 2010). Since the quartz content of the phyllites and two-mica schists of the Bozdağ and Bayındır nappes is similar, we argue that variations in the quartz content should have no pronounced effect on the measured 10Be erosion rates in the Aydın Range. With respect to the second assumption of sediment mixing, it should be kept in mind that sediment transport in the channel network of catchments is generally discontinuous (e.g., Benda and Dunne, 1997; Gran and Czuba, 2017). Hence, incomplete sediment mixing in the channels might to some degree explain the observed variability of erosion in the studied catchments. In particular, incomplete sediment mixing is more likely in smaller catchments than in larger ones (Niemi et al., 2005; Delunel et al., 2013), because dispersion and overlap of sediment pulses from several sediment sources is more efficient in larger catchments (Benda and Dunne, 1997). Although our approach of taking a sample from several points along each stream should increase the probability of obtaining a well-mixed sample with a 10Be concentration reflecting the average catchment erosion, we cannot rule out incomplete sediment mixing in our catchments. Regarding the assumptions of uniform erosion through time and a steady state between 10Be production in the catchment and 10Be outflux via sediment export and radioactive decay, it can be expected that these conditions are not fully met in natural settings (Dunai, 2010), and this may apply also to the Aydın Range. The steady state between 10Be influx and outflux may, for example, be violated by a long-term storage of sediments within catchments, even though the absence of fluvial terraces suggests that this process does not play a significant role in our study area (see Fig. 3). According to Heineke et al. (2017), debris flows seem to have no pronounced effect on measured erosion rates in the central Menderes Massif. Thus, it seems at least unlikely that they are the main factor causing the observed variability in catchment erosion. Although the ultimate reasons for this variability remain unclear, we consider it most likely that the observed 10Be erosion rates result from a combination of the different factors described above.
Ridge Crest Erosion and Topographic Relief
Most of our local 10Be erosion rates for ridge crests (i.e., 11 out of 14 values) fall in the narrow interval between 50 ± 4 and 87 ± 9 mm/k.y. (Table 2; Fig. 6), which is considerably smaller than the spread of the catchment-wide erosion rates (Fig. 7). The consistency of the local erosion rates across the massif shows that the amalgamation of a large number of clasts (≥1500) yields robust and representative erosion rates. Hence, this approach can overcome the problem that local erosion rates derived from bedrock outcrops may underestimate or overestimate rates of ridge-crest erosion (cf. Small et al., 1999; Reinhardt et al., 2007b; Meyer et al., 2010; Strobl et al., 2012). The different sampling sites lie at elevations between ∼800 and ∼2100 meters above sea level (masl) (Table 2); yet despite this significant spread, there is no relation between sample elevation and erosion rate. In addition, the lowering rate of ridge crest appears to be largely independent from crest width, because the broadest of all sampled ridges erodes at a rate of 51 ± 5 mm/k.y. (Fig. 5A), which is similar to the erosion rate of 70 ± 6 mm/k.y. for the narrowest of all sampled crests (Fig. 5D).
The erosion rates for ridge crests are in most cases lower than the rates obtained for neighboring catchments, indicating that topographic relief in most parts of the central Menderes Massif is growing. Under the assumption that incision rates of rivers are similar to rates of basin erosion, the rate of relief growth equals the difference between local and catchment-wide erosion rates (Meyer et al., 2010). In the Aydın Range, the average erosion rate of 18 catchments is ∼230 mm/k.y., whereas rates on intervening ridges are between 13 ± 1 and 87 ± 9 mm/k.y. (except for one high value of ∼165 mm/k.y., which we neglect for reasons explained above). Although there is a significant spread in both data sets, we attempt to provide a crude estimate for the rate of relief growth by using average values for local and catchment-wide erosion. We assign an uncertainty of ±100 mm/k.y. to the average value of 230 mm/k.y. for the catchments in the Aydın Range, because this leads to a range of 130–330 mm/k.y., which encompasses the erosion rates obtained for most catchments (14 out of 18). We use a value of 50 ± 30 mm/k.y. as the average rate of ridge-crest lowering. The difference between the two rates is 180 ± 100 mm/k.y., which we interpret as a reasonable estimate for the rate of relief production in the Aydın Range.
In the Bozdağ Range, we consider the western and eastern part of the range separately. In the west, local and catchment-wide erosion rates are almost similar (i.e., 30–82 mm/k.y. versus 40–109 mm/k.y., respectively; Fig. 6). This similarity suggests that local relief in this part of the range does not significantly change, which implies that rock uplift and river incision rates match each other. On the northern flank of the eastern Bozdağ Range, the situation appears to be the same at first sight, because catchment erosion rates in the vicinity of Salihli are similar to rates of ridge-crest lowering (Fig. 6). However, the assumption that river incision rates are similar to spatially averaged erosion rates is unlikely to be correct in this region, because incision of the deep valleys is presumably faster than erosion of the relict Gediz detachment surface that is preserved between the valleys (Figs. 3A and 3B). This interpretation is supported by the observation that the north-flowing rivers incising into the detachment have high channel steepness indices (Fig. 8) and are not graded (Fig. 10A). As a consequence, we argue that—in contrast to the western Bozdağ Range—local relief in the eastern part of the range is increasing, although we do not attempt to provide an estimate for the rate at which topographic relief is growing.
Imprint of Normal Faulting on Landscape Morphology
Normal faulting and rock uplift control the current morphology of the central Menderes Massif and are reflected in the spatial pattern of normalized channel steepness and chi values in the drainage network. In the Bozdağ Range, channel segments with high ksn values of >80 and locally >130 are restricted to the northeastern part of the range (i.e., region near Salihli; Fig. 8, river profiles 2–8), where the Pliocene phase of detachment faulting was most pronounced (Buscher et al., 2013; Asti et al., 2017) and caused the development of knickpoints, which can be traced along-strike of the mountain slope (Figs. 8 and 10A). Another reason for the occurrence of knickpoints in this part of the Bozdağ Range is the formation of the southern boundary fault of the Gediz Graben in the early Quaternary and the subsequent linkage of its three different fault segments (Fig. 8), which may have caused an acceleration in fault throw rates and a relative base-level fall (Kent et al., 2016a). As a consequence, the knickpoints and steep channel segments can be interpreted as a response to sustained tectonic activity and normal faulting since ca. 6 Ma. On the southern flank of the Bozdağ Range, no comparable pattern of steep channel gradients and knickpoints exists, and stream channels are generally less steep and, in most cases, graded (Figs. 8 and 10A). This finding is in accordance with less active, high-angle normal faults in the Küçük Menderes Graben as compared to the Gediz Graben (Rojay et al., 2005).
As outlined above, the exhumation of the Bozdağ Range was accomplished by the Pliocene phase of detachment faulting, which gradually merged into high-angle normal faulting and the development of the Gediz Graben (Çiftçi and Bozkurt, 2010; Oner and Dilek, 2011; Buscher et al., 2013; Kent et al., 2016a). Hence, extension and normal faulting occurred more or less continuously in prolonged phases and were accompanied by river incision and the formation of steep valleys with knickpoints (Fig. 10A). Owing to the resistance of the cataclasites and mylonites of the Gediz detachment to weathering and erosion, the detachment surface is still quite well preserved between the steep valleys (Fig. 3). Thus, the landscape of northeastern Bozdağ Range has not yet fully responded to the base-level fall by ongoing normal faulting.
In the eastern Bozdağ Range, the drainage divide is located near the crest of the steep escarpment south of Bozdağ Mountain (i.e., rather close to the Küçük Menderes Graben; Fig. 4). This position of the divide is somewhat unexpected, because fault-related rock uplift should currently be most pronounced in the vicinity of the active boundary fault of the Gediz Graben, and therefore, one could expect the main divide to be located farther north. We argue that the present position of the drainage divide has been inherited from the pronounced second phase of detachment faulting. During this phase of top-to-the-NNE shearing, the eastern Bozdağ Range was exhumed in the footwall of the Gediz detachment, which shifted the drainage divide far to the south. The spatial pattern of chi values for the stream network reveals marked cross-divide differences in this region, suggesting that the drainage divide is currently migrating back to the north (Fig. 9), despite ongoing normal faulting on the high-angle faults of the Gediz Graben. The inference of a northward-migrating divide is in agreement with our catchment-wide erosion rates, which are high on the steep escarpment south of Bozdağ Mountain and much lower on the northern slope of the mountain range (Fig. 6). Hence, catchments in the south are growing at the expense of catchments north of the divide (Fig. 9). This process of divide migration has advanced most in the area east of Bozdağ Mountain. Here, the northward shift of the divide has removed the entire upstream part of the catchment north of the divide (Fig. 11A) and enlarged the deeply incised southern catchment (Fig. 11B).
In a westerly direction along the Bozdağ Range, the drainage divide is located in a more central position within the range, and cross-divide differences in chi decrease, suggesting that the drainage divide is stable (Fig. 9). This is supported by the similarity of catchment-wide erosion rates, which are of the same order of ∼50–100 mm/k.y. on both sides of the range (Fig. 6). Hence, the process of drainage divide migration to the north as response to faulting and rock uplift appears to be more or less completed in the western Bozdağ Range. Two reasons may be responsible for this. First, the Pliocene phase of detachment faulting was particularly rapid in the eastern part of the detachment (near Salihli, Fig. 9; Buscher et al., 2013) but less pronounced in the west, and second, the western segment of the Gediz Graben boundary fault has slipped at a lower rate than the central and eastern segments in the Quaternary (i.e., ∼0.9 mm/yr versus ∼1.4 mm/yr in the past ∼2 m.y.; Fig. 8) (Kent et al., 2016b). Further evidence for an already accomplished northward shift of the drainage divide is provided by the bifurcation of valleys formed by streams that experienced flow reversal due to divide migration (Süzen et al., 2006) (Fig. 6). Before the northward shift of the divide, the point of confluence of the streams was located north of the drainage divide, whereas it is located south of main divide today. In summary, the spatial pattern of chi values in the Bozdağ Range indicates that the eastern and western parts of the mountain range are in different dynamic states: a quasi-equilibrium state in the west and a transient state in the east. Thus, the eastern part of the range evolves toward the state that has already been reached in western portion of the range.
In the Aydın Range, knickpoints are largely absent, and the river profiles of bedrock channels are graded on both sides of the mountains (Fig. 10B). Isolated steeper channel segments with ksn values >80 are restricted to the proximity of the Büyük Menderes detachment or correlate with high-angle normal faults (Fig. 8). The absence of knickpoints and the lack of very steep channel segments may create the impression that the Aydın Range has been subjected to a lower rate of base-level fall than the Bozdağ Range. However, the similar morphology and elevation of both mountain ranges and the intervening grabens (Fig. 2), the occurrence of surface-rupturing earthquakes along both the Gediz and Büyük Menderes grabens (Altunel, 1999; Ambraseys, 1971), and the presence of uplifted fluvial terraces in their footwalls indicate that the tectonic activity in both ranges is comparable. Therefore, we interpret the river profiles in the Aydın Range to indicate that channel incision and catchment erosion are able to keep pace with the uplift caused by normal faulting. The presence of soft phyllites of the Bayındır nappe and their susceptibility to weathering and erosion can explain why bedrock channels in the Aydın Range are graded and less steep than in the Bozdağ Range.
Similar to the Gediz detachment, faulting on the Büyük Menderes detachment has shifted the drainage divide to a position near the Küçük Menderes Graben (i.e., away from the detachment fault). The pattern of chi values in the Aydın Range indicates that the drainage divide is currently mobile and tends to migrate in a southward direction (Fig. 9). However, cross-divide differences in chi values are generally less pronounced than in the Bozdağ Range and not so clearly correlated with catchment-wide erosion rates, because not all catchments having low chi values exhibit higher erosion rates—as expected if these catchments would act as aggressors and gain area (e.g., catchment 14T6 versus 14T16). In this connection, ongoing uplift of the Aydın Range and river incision in response to normal faulting along the Büyük Menderes Graben may force the drainage divide to remain close to the Küçük Menderes Graben, while the streams flowing into the Küçük Menderes Graben are unable to cause a re-migration of the divide to the south.
Landscape Evolution during Continental Extension
In regions of active continental extension, rock exhumation occurs by a combination of erosional and tectonic denudation. The relative importance of these processes can be evaluated by combining 10Be erosion rates with low-temperature thermochronological data, which provide information on the total exhumation rate. However, only few fault-bounded mountain ranges exist worldwide, where both data on exhumation and erosion rates are available (Reinhardt et al., 2007a; Densmore et al., 2009; Stock et al., 2009). Our study area in the Menderes Massif represents the first example of a metamorphic core complex, for which exhumation and erosion rates have been determined both across and along strike of the fault footwalls. In the following, we will first evaluate the relative importance of erosion and tectonic denudation in the central Menderes Massif. Although such a comparison between erosion and exhumation rates was conducted previously along a narrow transect (Buscher et al., 2013; Wölfler et al., 2017), our new 10Be data set and recently published thermochronological data (Nilius et al., 2019) allow for an improved analysis. Subsequently, we compare our results with data from other normal fault-bounded mountain ranges.
To compare our spatially averaged 10Be erosion rates (Figs. 6 and 7) with exhumation rates derived from thermochronological data in different parts of the central Menderes Massif, we summarize the range of these values in Table 3. In the northern Bozdağ Range, erosion rates in the footwall of the eastern and western Gediz detachment are 70–150 and 40–110 m/m.y., respectively, whereas thermochronological data indicate much higher exhumation rates of 600–1100 m/m.y. and 600–1000 m/m.y. (Gessner et al., 2001; Glodny and Hetzel, 2007; Buscher et al., 2013). Hence, the erosional contribution to rock exhumation at the Gediz detachment is only ∼10% (i.e., ∼13% in the east versus ∼9% in the west; Table 3). In the southern Bozdağ Range, the catchments on the steep escarpment facing the Küçük Menderes Graben erode at rates between 130 and 370 m/m.y. Here, the exhumation rates from thermochronology are lower than at the Gediz detachment (i.e., 300–700 m/m.y.), and thus the relative contribution of erosion to exhumation is ∼50% (Table 3). A similar value is obtained for the northern part of the Aydın Range, which forms the opposite margin of the Küçük Menderes Graben. In the footwall of the Büyük Menderes detachment, exhumation rates are slightly lower than at the Gediz detachment (i.e., 400–800 m/m.y.; Wölfler et al., 2017; Nilius et al., 2019). Given the rather high 10Be erosion rates in the southern Aydın Range, erosion is responsible for ∼40% of the rock exhumation, although it should be kept in mind that the data variability is large (i.e., 90–430 m/m.y.; Fig. 7D). In summary, the contribution of erosion to rock exhumation was higher at the Büyük Menderes detachment than at the Gediz detachment, where tectonic denudation was the dominant exhumation mechanism (Table 3). Along the high-angle normal faults on both sides of the Küçük Menderes Graben, erosion contributed ∼50% to rock exhumation.
A low contribution of erosion to the total denudation as observed for the Gediz detachment was reported from the Rio Torrente catchment in the Spanish Sierra Nevada (Reinhardt et al., 2007a). At this site, apatite fission-track ages indicate total denudation rates between 400 and 1500 m/m.y. over the past 3–4 m.y., whereas an average 10Be-based erosion rate of 44 ± 15 m/m.y. has been determined in the upper part of the catchment, which is characterized by low relief. In the uplands, most of the unroofing hence occurred by low-angle normal faulting (Reinhardt et al., 2007a). The lower parts of the Rio Torrente catchment are affected by high-angle normal faults, which cause footwall uplift and relative base-level fall. Hence, the lower part of the catchment experiences enhanced river incision and hillslope erosion, which results in a much higher 10Be erosion rate of 1600 ± 400 m/m.y. According to Reinhardt et al. (2007a), the difference in the erosion rates indicates that faulting-related changes in base level are not transmitted to the low-relief upland areas. In contrast to the Rio Torrente site, the north-draining catchments in the Bozdağ Range show quite consistently low erosion rates along strike of the Gediz detachment (except for two catchments with easily erodible bedrock, as explained above). This might suggest that uplift and base-level changes due to high-angle normal faulting along the Gediz Graben have so far not significantly increased the catchment-wide erosion rates in the footwall of the Gediz detachment. In contrast, faulting-related uplift and base-level changes may have a greater effect in the southern Aydın Range, where erosion rates are higher than in the northern Bozdağ Range (Table 3). However, it should be kept in mind that the presence of different bedrock lithologies also exerts a significant control on catchment erosion and its spatial variability.
In contrast to the Gediz and Büyük Menderes detachment faults, erosion tends to play a larger role in exhuming rocks in the footwall of the high-angle normal faults bounding the Küçük Menderes Graben (Table 3). This result can be compared with two sites in the Basin and Range Province, where 10Be erosion rates and exhumation rates were determined for the footwalls of high-angle normal faults. The first site is the central Wasatch Range, for which Stock et al. (2009) reported 10Be erosion rates between 70 and 170 m/m.y. for catchments at the range front. Because the exhumation rates along most of the range front are between 200 and 400 m/m.y. over the past ∼5 m.y. (Ehlers et al., 2003; Armstrong et al., 2004), this indicates that erosion provides an average contribution of ∼40% (range: 19%–61%) to the total exhumation rate. At the southern Salt Lake City segment, where exhumation rates reach values of 600–800 m/m.y. (Ehlers et al., 2003; Armstrong et al., 2004), erosion contributes less than 30% to exhumation at the range front (Stock et al., 2009). Toward the interior of the Wasatch Range, catchments are steeper and show higher erosion rates with a larger variability (170–790 m/m.y.) (Stock et al., 2009). Hence, in the core of the Wasatch Range, erosion rates may be as high as the exhumation rates. The Wassuk Range is the second site where exhumation rates based on thermochronology and 10Be erosion rates have been determined (Densmore et al., 2009). In the range center, most apatite (U-Th)/He ages yield exhumation rates between 500 and 800 m/m.y., whereas most catchments erode at rates of 100–300 m/m.y. As a result, the contribution of erosion to exhumation is ∼30% (range: 14%–48%). At the southern range tip, where erosion and exhumation rates are 86 ± 4 and 300 ± 10 m/m.y., respectively, the contribution of erosion is ∼30%.
In summary, the existing data sets on erosion and exhumation rates in regions of continental extension indicate that erosion provides a non-negligible contribution to rock exhumation along both low-angle and high-angle normal faults. Along low-angle normal faults, the erosional contribution may be as low as ∼10%, whereas it is generally higher and more variable along high-angle normal faults with values between 30%–70%. It should be kept in mind, however, that there are only a limited number of sites so far where both 10Be and thermochronological data are available on the same spatial scale and that the data show considerable spatial variability, especially the 10Be erosion rates (e.g., Densmore et al., 2009; Stock et al., 2009; this study). As outlined above, the ultimate reasons for the observed variability in erosion rates are often difficult to determine. Despite these uncertainties, comparisons between erosion and exhumation rates provide important insights into the relative contribution of the different processes that control rock exhumation and landscape evolution in regions of continental extension.
New catchment-wide erosion rates and the first local 10Be erosion rates for ridge crests in the central Menderes Massif document the spatial pattern of erosion in the entire massif during continental extension. Rates of drainage basin erosion in combination with river network analyses reveal that the geomorphic signature of normal faulting is best preserved in the Bozdağ Range. Its preservation in the landscape is related to the resistant lithologies exposed in the footwall of the Gediz detachment, where steep channel segments and knickpoints record ongoing normal faulting and footwall uplift. In the Aydın Range, the geomorphic signal of normal faulting and uplift is less well preserved, because the lithology in the footwall of the Büyük Menderes detachment is characterized by a thin layer of cataclasites and rather weak phyllites and mica schists, which have a low preservation potential due to their high erodibility. The spatial pattern of chi values in the drainage network, the course of the drainage divide, and cross-divide differences in catchment erosion in the central Menderes Massif document that uplift was most pronounced in the northeastern part of the Bozdağ Range. We infer that this part of the mountain range is still in a state of transience, while the western part of the Bozdağ Range has attained a quasi-equilibrium state. By comparing catchment-wide erosion rates with local rates of ridge-crest lowering, we deduce that the topographic relief in the Aydın Range and eastern Bozdağ Range increases due to ongoing uplift along the active normal faults of the Gediz, Küçük Menderes, and Büyük Menderes grabens.
We thank Eric Portenga, Liam Reinhardt, and Associate Editor Jeff Lee for constructive reviews, which significantly improved this paper. We are grateful to A. Niehus, K. Schoppengerd, and P. Gebbeken for their help during the separation of 10Be from quartz and target preparation. V. Rapelius is thanked for inductively coupled plasma–optical emission spectroscopy analysis. This work was funded by the Deutsche Forschungsgemeinschaft (DFG) grants HE 1704/18-1 and GL 724/7-1 provided to R. Hetzel and C. Glotzbach.