University of Plymouth PEARL https://pearl.plymouth.ac.uk Faculty of Science and Engineering School of Geography, Earth and Environmental Sciences 2020-05-15 Rock strength and structural controls on fluvial erodibility: Implications for drainage divide mobility in a collisional mountain belt Zondervan, JR http://hdl.handle.net/10026.1/16681 10.1016/j.epsl.2020.116221 Earth and Planetary Science Letters Elsevier All content in PEARL is protected by copyright law. Author manuscripts are made available in accordance with publisher policies. Please cite only the published version using the details provided on the item record or document. In the absence of an open licence (e.g. Creative Commons), permissions for further reuse of content should be sought from the publisher or author. NOTICE: this is the author’s version of a work that was accepted for publication in EARTH AND PLANETARY SCIENCE LETTERS. Changes resulting from the publishing process, such as editing, corrections, structural formatting, and other quality control mechanisms may not be reflected in this document. A definitive version was subsequently published in Earth and Planetary Science Letters, volume 538, 116221 – doi:10.1016/j.epsl.2020.116221 © (2020). This manuscript version is made available under the CC_BY_NC_ND 4.0 license. 1 Rock strength and structural controls on fluvial erodibility: 2 implications for drainage divide mobility in a collisional 3 mountain belt * 4 Jesse R. Zondervan , Martin Stokes, Sarah J. Boulton, Matt W. Telfer, Anne E. 5 Mather 6 School of Geography, Earth and Environmental Sciences, University of Plymouth, Plymouth, PL4 8AA, 7 United Kingdom 8 Correspondence to: Jesse R. Zondervan (jesse.zondervan@plymouth.ac.uk) 9 Highlights 10 • We estimate rock strength, erodibility and drainage divide mobility in the High Atlas 11 Mountains 12 • The weakest rock-type in the High Atlas is up to two orders of magnitude more erodible than 13 the strongest 14 • In gently deformed horizontal strata of the sedimentary cover the drainage divide is mobile 15 • Faulted and folded metamorphic sedimentary bedrock coincide with a stable drainage divide 16 • Exhumation of crystalline basement forces the drainage divide into the centre of exposed 17 basement 18 Abstract 19 Numerical model simulations and experiments have suggested that when migration of the main 20 drainage divide occurs in a mountain belt, it can lead to the rearrangement of river catchments, 21 rejuvenation of topography, and changes in erosion rates and sediment flux. We assess the 22 progressive mobility of the drainage divide in three lithologically and structurally distinct groups of 23 bedrock in the High Atlas (NW Africa). The geological age of bedrock and its associated tectonic 24 architecture in the mountain belt increases from east to west in the study area, allowing us to 25 track both variations in rock strength and structural configuration which influence drainage 26 mobility during erosion through an exhuming mountain belt. Collection of field derived 27 measurements of rock strength using a Schmidt hammer and computer based extraction of river 28 channel steepness permit estimations of contrasts in fluvial erodibilities of rock types. The 29 resulting difference in fluvial erodibility between the weakest and the strongest lithological unit is 30 up to two orders of magnitude. Published evidence of geomorphic mobility of the drainage divide 31 indicates that such a range in erodibilities in horizontal stratigraphy of the sedimentary cover may 32 lead to changes in erosion rates as rivers erode through strata, leading to drainage divide 33 migration. In contrast, we show that the faulted and folded metamorphic sedimentary rocks in the 34 centre of the mountain belt coincide with a stable drainage divide. Finally, where the strong 35 igneous rocks of the crystalline basement are exposed after erosion of the covering meta- 36 sediments, there is a decrease in fluvial erodibility of up to a factor of three, where the drainage 37 divide is mobile towards the centre of the exposed crystalline basement. The mobility of the 38 drainage divide in response to erosion through rock-types and their structural configuration in a 39 mountain belt has implications for the perception of autogenic dynamism of drainage networks 40 and fluvial erosion in mountain belts, and the interpretation of the geomorphology and 41 downstream stratigraphy. 42 43 Keywords: collisional mountain belt, drainage divide, rock strength, erodibility, High Atlas 44 6493 words (excl. highlights, abstract, tables, figure captions, author contributions, funding, refs) 45 1 Introduction 46 Collisional mountains form the erosional focus of the Earth’s surface. The tectonic and climatic 47 interpretation of mountain topography and depositional stratigraphy depends on understanding the 48 dynamics of eroding bedrock rivers. The central drainage divide of a mountain belt is the 49 topographic boundary between river catchments draining either flank. Any movement of the 50 drainage divide can result in the rearrangement of catchments, rejuvenation of topography, and 51 changes in erosion rates and sediment flux (Bonnet, 2009; Giachetta et al., 2014). Bedrock erodibility 52 is expected to play a significant role in drainage divide reorganisation since heterogeneous 53 exhumation of weak and strong substrates can enhance and suppress erosion respectively 54 (Giachetta et al., 2014; Gallen, 2018) and cause topographic rejuvenation, for example through the 55 exhumation of a basement palaeosurface (Strong et al., 2019). This is especially the case in post- 56 orogenic settings where erosion is dominant over crustal thickening (Gallen, 2018; Bernard et al., 57 2019). However, the magnitude of erodibility variation within a mountain belt, and the mobility of 58 the drainage divide as rivers erode through its stratigraphy are still relatively unexplored. 59 Collisional orogens are characterised by bedrock rivers eroding through variable rock strength and 60 tectonic architectures, such as very strong crystalline basement, deformed strong meta-sediments, 61 and sedimentary cover composed of weak as well as stronger strata. Recent numerical modelling 62 studies show the complexity of incision into horizontal or gently dipping strata of varying erodibility 63 (Forte et al., 2016; Perne et al., 2017), and also suggest steady-state denudation is more likely to 64 develop in tilted and /or faulted strata that have been highly deformed (Forte et al., 2016). The 65 results of these models imply drainage mobility might be prevalent during incision in sedimentary 66 cover but not in more deformed components of a collisional mountain belt. In a field study, Gallen 67 (2018) modelled the erosion of a hard horizontal rock layer in the Appalachians. This model predicts 68 that a geologically instantaneous capture of the Upper Tennessee River catchment by the Lower 69 Tennessee River occurred at 9 Ma, which has led to a shift in the drainage divide and explains 70 observed subsequent topographic rejuvenation in the landscape visible today. In the Pyrenees, 71 Bernard et al. (2019) show the drainage divide follows the position of high-strength, high-elevation 72 plutons in the crystalline basement in the centre of the belt, suggesting a direct lithological control 73 on the position of the central drainage divide. These field studies demonstrate the drainage divide 74 can be mobile in near-horizontal sedimentary stratigraphy of mountain belts and is likely to move to 75 the centre of highly resistant plutons as they get exhumed in the axis of a collisional mountain belt. 76 However, both studies also demonstrate the challenge of documenting drainage mobility, which is 77 often only recognisable in instantaneous capture events which leave pervasive topographic 78 evidence. In addition, whilst numerical simulation studies are instrumental in predicting long-term 79 processes in geomorphology which are hard to derive from observations of modern day landscapes 80 alone, these involve simplifications and it is challenging to model the full complexity of lithologies 81 and structural geology of a collisional mountain belt. Furthermore, the fluvial erodibility of rock, 82 which depends on rock mass strength as well as jointing and weathering, is challenging to measure 83 quantitatively (Bursztyn et al., 2015) and consequently the range of erodibility inputs in modelling 84 studies vary widely (Roy et al., 2015; Forte et al., 2016; Yanites et al., 2017). 85 Consequently, while theory and numerical models predict a role for bedrock erodibility in driving 86 drainage divide mobility, where and when this is expected to occur in the evolution of mountain 87 belts needs to be understood from a range of settings with varying rock strength conditions. The 88 challenges of measuring and characterising bedrock erodibility (Bursztyn et al., 2015) and drainage 89 divide dynamics in field settings (Willett et al., 2014; Forte and Whipple, 2018) remain key problems. 90 Addressing this challenge requires a dataset on rock erodibility that is combined with topographic 91 measures of drainage divide mobility in a mountain belt where erosion is dominant over tectonic 92 advection, and where rivers erode through a lithologically variable landscape. In this study, we aim 93 to derive data and present it in a form that is useful for both field geologists and numerical 94 modellers alike. We focus our study on the central High Atlas Mountains (Fig. 1), where different 95 stages of orogenic landscape evolution, from erosion through gently dipping sedimentary cover to 96 exhumed crystalline basement occur along the length of the main drainage divide, forming a natural 97 experiment. Furthermore, the continental inland setting, low weathering rates associated with an 98 arid climate and low rates of tectonic activity make this post-orogenic mountain range an ideal 99 setting to study erodibility-induced drainage divide mobility. In this study, we quantify the 100 magnitude of variation in erodibility between rock types by: (i) extracting the normalised river 101 channel steepness of rock units from a digital elevation model; (ii) collecting mechanical 102 measurements of rock strength in the field, and (iii) quantifying the mobility of the drainage divide 103 using topographic metrics. Our measures of rock erodibility and stratigraphic orientation are then 104 used to quantitatively assess when and where a drainage divide will move in response to erodibility Figure 1 – a) Geological map of the central High Atlas, showing the distribution of Mesozoic sedimentary cover consisting of limestones (blue) and continental clastic sedimentary rocks (pink), Palaeozoic meta-sediments (orange) and crystalline basement (red). The chain is flanked by sedimentary basins: the Ouarzazate Basin to the south and the Haouz Basin to the North. Three cross-sections labelled A to C show the locations of cross-sections in b). Headwater channel locations used for calculating drainage divide mobility start at a reference drainage area of 1 km2, marked as white dots (see methods 3.3). The drainage divide is marked on as a black line and is segmented into five equal lengths (Fig. 5). b) Cross-sections based on seismic sections modified from Errarhaoui (1998); Teixell et al. (2003). These cross-sections are vertically exaggerated by a factor of 1.5. The location of the drainage divide is marked on as black markers on the structural cross sections. 105 variation in the evolution of a collisional mountain belt in which rivers erode first through 106 sedimentary cover, secondly through meta-sediments and finally through crystalline basement. 107 Results from this study have implications for understanding long-term trends in the geomorphology 108 and erosion of mountain belts owing to the rock types and their structural configuration, and 109 quantitatively constrain the magnitude of contrasts in erodibility within a mountain belt which 110 enables integration of observation and modelling studies. 111 2 Study area 112 The study area is located in the centre of the High Atlas Mountains, focussed on the 250-300 km 113 length of the drainage divide bounded by the Ouarzazate Basin to the south and the Haouz Basin in 114 the north (Fig. 1). The drainage divide strikes along the structural grain of the mountain belt, 115 inherited by the configuration of a pre-existing rift (Babault et al., 2012). The age of exposed bedrock 116 and its associated tectonic architecture in the mountain belt increases from east to west in the study 117 area (Fig. 1). The along strike trends allow us to use location as a proxy for time in understanding the 118 mobility of the main drainage divide as river erosion exposes sedimentary cover to metamorphic 119 sedimentary rock, to crystalline basement. The sedimentary cover, older meta-sediments and 120 underlying crystalline basement have distinct lithologies with varying hardness and structural 121 weaknesses owing to the tectonic architecture of the mountain belt (Table 1). Thus, each 122 chronostratigraphic package has its own potential for drainage divide mobility which we investigate 123 in this study. 124 From east to west the drainage divide crosses first a landscape dominated by gently dipping beds of 125 Mesozoic carbonates and clastic sedimentary rock (Table 1) configured into low-amplitude long- 126 wavelength folds punctuated by spaced-out thrusts (Fig 1, Fig 2a) at an elevation of 2600 to 3000 m. 127 In this length of the divide, the Dades catchment lies to the south and incises into a long-wavelength 128 syncline exhibiting gently dipping strata, whereas directly north of the divide is a concentration of 129 folding and thrust offset strata dipping at higher angles to the surface (Fig. 1b). Next the drainage 130 Table 1 - chronostratigraphic packages labelled on Fig 1 with details of lithologies and structure Chronostratigraphic package Lithologies Structure Rock strength (see Table 2 for definitions) Mesozoic sedimentary cover massive marine platform horizontal stratigraphy, very weak – limestones; continental red gently deformed with moderately strong marls with gypsum; marine spaced-out faults and and continental thrust-top folds offsetting conglomerates, sandstones, strata vertically siltstones and micro- conglomerates Palaeozoic meta-sediments sandstones, shales, closely spaced faults, moderately strong siltstones, schists abundant jointing, steeply dipping and folded strata Crystalline basement granite, granodiorite, massive outcrops of strong gabbro, dolerite, migmatite igneous rock, vertical or sheet fracturing in some outcrops 131 132 divide decreases in elevation to ~2500 m, as it crosses a zone of dipping strata of faulted meta- 133 sediments (Table 1, Fig. 2b) in the centre of the study area (Fig.1). Finally, in the west, the divide 134 rises back up to elevations > 2800 m as it runs over the edge of the exhumed crystalline basement 135 consisting of igneous rocks with vertical or sub-vertical fracturing (Fig. 1, Table 1, Fig 2c-d). 136 The inland setting of the central High Atlas (800-1000 km from the coast) results in an absence of 137 eustatic base level control on drainage development. Though localised base level fluctuations are 138 likely to have occurred, studies of basin fill sediments and river profiles suggest that such 139 fluctuations were small to negligible over the Plio-Quaternary (Boulton et al., 2014; Boulton et al., 140 2019). The High Atlas is in a post-orogenic state, with long-term isostatic rock uplift rates of 0.17- 141 0.22 mm yr-1 since 15 Ma related to lithospheric thinning (Babault et al., 2008). The lack of recent Figure 2 – field photos showing the lithologies and stratigraphic configuration of units within the three lithostratigraphic packages of the Atlas. a) In the Mesozoic weathering-resistant competent limestone beds dip at sub- horizontal angles. Within the interlayered limestones and weaker clastic rocks, variations in rock strength and weathering exist. b) In the Palaeozoic meta-sediments steeply dipping, sub-vertical strata are common. Road on the flood plain for scale. c) Massive crystalline basement exposed in is sometimes found jointed in vertical plains or d) as dome-shaped or massive igneous outcrops with vertical or sheet fracturing. House for scale at the bottom. 142 tectonic deformation is apparent in the undeformed continuous Quaternary river terraces forming 143 parallel river long profiles throughout the fold-thrust belt and thrust front (Stokes et al., 2017). 144 Therefore, unlike in active mountain belts where divide migration is expected in response to changes 145 in the uplift field (Willett et al., 2001), the Plio-Quaternary divide mobility of the High Atlas is not 146 expected to reflect tectonic advection. Recorded glacial features related to Pleistocene glaciation 147 establish that the snowline was at c. 3300 m in the High Atlas (Hughes et al., 2004). Since the main 148 drainage divide varies in elevation between 2500-3000 m it has not been affected significantly by 149 glacial activity. Thus, for our purposes the main control on the Plio-Quaternary evolution of the High 150 Atlas river network is the incision through the inherited tectonic architecture of lithological units and 151 their contrasting strengths. Drainage development in the High Atlas is considered to be primarily the 152 product of the exhumation of structurally distributed lithologies with different hardness, controlling 153 where river terraces develop (Stokes et al., 2017) and affecting the occurrence of diffusive and 154 advective slopes (Mather and Stokes, 2018). The High Atlas is set in a semi-arid climate, which 155 means that the effect of weathering on rock erodibility is expected to be low. Thus, intact rock 156 strength measurements are likely to reflect the effective bedrock strength. 157 3 Methods 158 We measure two proxies of fluvial bedrock erodibility to constrain the contrast between lithological 159 units. These data allow us to compare our field observations of drainage divide mobility with 160 numerical models of river erosion through variable lithology. We systematically: 1) derive the 161 normalised river channel steepness index from a digital elevation model (DEM) as a measure of a 162 river channel’s power to erode rock; 2) record compressive rock strength using a Schmidt hammer in 163 the field as another measure of fluvial erodibility, and 3) extract topographic metrics of drainage 164 divide mobility from the DEM along the length of the main drainage divide. 165 3.1 River profile analysis, rock type and fluvial erodibility 166 Rock type influences the river network by affecting the ability of rivers to erode into bedrock, 167 determined by the fluvial erodibility. Numerical models that describe river erosion through bedrock 168 often use the stream power model, in which the erosion rate at any particular point in a bedrock 169 river channel is defined by: 170 𝐸𝐸 = 𝐾𝐾𝐴𝐴𝑚𝑚𝑆𝑆𝑛𝑛 (Eq. 1) 171 where K is an erodibility constant which depends on the rock-type over which the river channel 172 flows as well as the climatic setting, A is upstream drainage area and S is local channel gradient, and 173 m and n are constants that depend on basin hydrology, channel geometry, and erosion processes 174 (Whipple and Tucker, 1999). Any change in the erodibility of rock type exposed will force the river to 175 adjust its stream power by changing river channel slopes and thus can cause divide mobility as river 176 networks respond. The use of this formula in field studies proves problematic because of the lack of 177 data and robust methodology to determine the erodibility constant, K. However, equation 1 can be 178 written as 𝑆𝑆 �𝐾𝐾𝐸𝐸 𝑛𝑛1179 = � 𝐴𝐴−(𝑚𝑚/𝑛𝑛) (Eq. 2) 180 which may be recognised as a form of the empirical power law scaling local channel gradient (S) and 181 drainage area (A): 182 𝑆𝑆 = 𝑘𝑘𝑠𝑠𝑛𝑛𝐴𝐴−𝜃𝜃 (Eq. 3) 183 𝑘𝑘𝑠𝑠𝑠𝑠 �𝐾𝐾𝐸𝐸�𝑛𝑛1= 184 𝜃𝜃 = 𝑚𝑚/𝑠𝑠 185 where S is the local channel slope (dimensionless), ksn is normalised steepness index (m2θ) and 𝜃𝜃 is 186 the concavity index (dimensionless). We assume that erosion is proportional to specific stream 187 power and inversely proportional to bedrock erodibility, so that n > 1 (Perne et al., 2017), and ksn is 188 inversely proportional to erodibility, K: 𝐾𝐾 ∝ 𝑘𝑘𝑠𝑠1189 𝑛𝑛𝑛𝑛 (Eq. 4) 190 Unlike K, the ksn of river channels can be readily determined from digital elevation models (DEMs) 191 (e.g. Boulton et al., 2014; Gallen, 2018; Bernard et al., 2019). We therefore use ksn as a measure of 192 fluvial erodibility of geological units over which river channels flow in the High Atlas, provided that 193 spatial variability in long term rock uplift is low, which is likely, for reasons outlined above (Section 194 2). Consequently, a factor of difference in ksn between geological units will be an estimate of the 195 factor of difference in their fluvial erodibility, K. We normalise ksn values to the most erodible rock- 196 type (see section 4.1), and then use Eq. 4 to convert these normalised average ksn values to 197 normalised K values. To do this, we calculate normalised K for three feasible values of n: n=1, n=2 198 and n=4. Perne et al. (2017) show that to obtain a stream-power-model generated river profile in 199 which the slopes of river channels are steeper in low erodibility, strong rock, n must be > 1. Lague 200 (2014) demonstrate that, fully calibrated with slope information from field locations, n ~ 2 in most 201 cases, and this could be as high as n ~ 4. Calculating normalised K values from the ksn data enables a 202 quantification of contrasts in erodibility K between rock-types in the High Atlas and their control on 203 changes in river erosion rates and consequent mobility of the drainage divide. 204 We performed ksn analysis using the Topographic Analysis Kit (Forte and Whipple, 2019), a series of 205 MATLAB functions based on TopoToolbox (Schwanghart and Scherler, 2014) which uses the chi 206 approach to calculate ksn smoothed over 1000 m segments (Perron and Royden, 2013). Tests using 207 the criteria proposed by Perron and Royden (2013) show that 𝜃𝜃 is ~ 0.45 for a range of catchments 208 covering the extent of the drainage divide in the central High Atlas, which is the same value as used 209 for the High Atlas by Boulton et al. (2014). This gives ksn units of m0.9. We use the ALOS Digital 210 Elevation Model with a resolution of 30 m from the Japan Aerospace Exploration Agency 211 (http://www.eorc.jaxa.jp/ALOS/en/aw3d30/index.htm) to perform the analysis following 212 recommendations of Boulton and Stokes (2018). Finally, we extract the average and standard 213 deviation ksn of each rock type using the digitised 1: 1 000 000 geological map of Morocco (Saadi et 214 al., 1985, see Supplementary Fig. S1). Formations are grouped into lithological units that collectively 215 comprise three groups: Mesozoic sediments, Palaeozoic meta-sediments and crystalline basement 216 over which the drainage divide crosses from east to west (Fig. 1). For each of those packages we 217 group lithological units (Table 1): (i) gently folded massive marine platform limestone; (ii) 218 interbedded red beds consisting of mud and siltstone; (iii) interbedded sand-siltstones and 219 conglomerates; (iv) folded and faulted meta-sedimentary rocks such as schist and shale, and (v) 220 faulted massive outcrops of igneous units. The contrasts in ksn of river channels flowing over rock- 221 types (Supplementary Fig. S1) includes the effects of both rock strength and structural weaknesses 222 on the resulting erodibility of bedrock. 223 224 3.2 Mechanical rock strength measurements, rock type and erodibility 225 A further approach to determine the fluvial erodibility of rock-types is measuring their mechanical 226 strength. The classic experiment by Sklar and Dietrich (2001) showed that erodibility of rock, K, in 227 t𝜎𝜎he stream power model for bedrock river erosion is related to the inverse square of tensile strength 228 ( 𝜏𝜏2). Thus, a measure of tensile strength and the difference between units enables the calculation 229 of contrasts in erodibility, K. However, measurements of tensile strength cannot be achieved easily 230 in the field, and the most commonly applied method of rock strength assessment in geomorphology 231 is the Schmidt hammer because of its portability (Goudie, 2006). The Schmidt hammer records the 232 rebound distance of a weighted spring that strikes the rock surface and uses a range from 10 – 100. 233 With the Schmidt hammer, measurements of in situ uniaxial compressive rock strength in the 234 landscape can be taken in large quantities, making it a very versatile instrument in landscape studies. 235 The higher the rebound value, the higher the elastic strength of the rock, which is a measure of the 236 uniaxial compressive strength (UCS) of a rock. Tensile strength (TS) represents the resistance to 237 sediment impacts on the riverbed and its use has been advocated based on the stronger correlation 238 of fluvial metrics with tensile strength than compressive strength (Bursztyn et al., 2015). But since 239 compressive and tensile strength are correlated (UCS ~ 10*TS; Kahraman et al., 2012; Nazir et al., 1 240 2013), K is effectively also proportional to the inverse square of UCS: 𝐾𝐾 ∝ 𝑈𝑈𝑈𝑈𝑈𝑈2 (Eq. 5) 241 Thus, similar to the approach in 3.1, we normalise UCS values to the most erodible rock-type, and 242 then use Eq. 5 to convert these normalised average UCS values to normalised K values. To produce a 243 map of rock strength we digitised the 1:1 000 000 geological map of Morocco (Saadi et al., 1985) for 244 the study area and assigned an average rock rebound value to each stratigraphic unit. Typically, ten 245 to twenty Schmidt hammer measurements were taken at each location, totalling 690 readings 246 throughout the central High Atlas with up to 132 readings per geological unit (Supplementary 247 materials). Where rock strength measurements for geological units from field measurements were 248 absent, we supplemented this with an existing database of Schmidt rebound values for lithological Figure 3 – principles of the Gilbert metrics for drainage divide stability, modified after Forte and Whipple (2018). Where channels are steeper on one side of the divide compared to the other, divide motion will be in the direction of low gradients where erosion is lower. B) Similarly, erosion will progress towards those channels with higher headwater channel elevations. 249 units reported from around the world (Goudie, 2006). This database contains Schmidt rebound 250 measurements for 111 lithological units, in mostly arid environments similar to the High Atlas. For 251 each stratigraphic unit without field measurements the average of values reported in the literature 252 for the lithology of that unit is used. Units are combined in the same lithological groups as the ksn 253 data. The standard deviation of Schmidt hammer measurements and the range of values reported in 254 the literature reflect the variation of rock strengths within the lithological units (Supplementary 255 material Table 1, Figure S2). Mean and standard deviation of Schmidt hammer rebound v𝑈𝑈a𝑈𝑈lu𝑆𝑆es (SHV) 256 are t𝑒𝑒hen converted to UCS using the conversion which was derived by Katz et al. (2000) ( =257 2.21 (0.07𝑈𝑈𝑆𝑆𝑆𝑆)) for a range of carbonate rocks, sandstone, marble and igneous rocks with UCS values 258 similar to those found in the study area. In contrast to the ksn approach, the UCS of lithological units 259 only takes into account the internal rock mass strength of bedrock, and thus does not account for 260 structural weaknesses imposed by discontinuities such as bedding or jointing. Consequently, 261 estimating variation in erodibility between rock-types using this approach may be expected to 262 underestimate the erodibility of folded and faulted shales and schists and jointed igneous rock 263 (Table 1). 264 3.3 Topographic analysis of the drainage divide 265 To determine the mobility of the drainage divide in response to erodibility variation and its 266 structural configuration, we perform topographic analysis of the main drainage divide along the 267 study length. The mobility and potential direction of movement of the drainage divide depends on 268 erosion rates either side, and whilst the chi method of mapping drainage divide instabilities (Willett 269 et al., 2014) is used widely, Forte and Whipple (2018) demonstrated that this method is especially 270 problematic when integrating across multiple lithologies with different strengths. Alternatively, 271 Forte and Whipple (2018) coined the term Gilbert Metrics, based on the assumption that contrasting 272 stream gradients either side of a divide will drive divide migration. Where divides are bounded by 273 different channel gradients either side they are mobile, with higher erosion rates on the steeper 274 sides leading to migration towards the side with lower channel gradients (Fig. 3). The topographic 275 proxies for erosion rates across divides defined by Forte and Whipple (2018) are differences in 276 headwater channel elevation, local headwater hillslope gradient and local headwater relief, which 277 have proven useful for interpreting the relative mobility of catchment divides (Forte and Whipple, 278 2018). Headwater channel elevation, local headwater hillslope gradient and local headwater relief 279 were extracted from the ALOS DEM using methods outlined in Forte and Whipple (2018), at a 280 reference drainage area 1 km2. While the upstream area required to initiate channel flow varies on 281 settings (Clubb et al., 2014), this drainage area is the critical threshold area downstream of which 282 stream-flow-dominated fluvial channels are expected to dominate over debris-flow-dominated 283 colluvial channels in most settings (Wobus et al., 2006). In a numerical landscape evolution model, 284 Forte and Whipple (2018) test the applicability of the Gilbert metrics in a setting with erodibility 285 contrasts across a drainage divide, and conclude that divide migration rate approximates a linear 286 relationship with cross-divide differences in erosion rates and all three Gilbert metrics. To assess the 287 mobility of the drainage divide we define five equal lengths of about 50 km each where boundaries 288 align with those of the three lithological packages described earlier (Fig. 1), and in each the mean, 289 standard deviation, standard error and bootstrap confidence interval of headwater channel values 290 are calculated either side of the drainage divide. Figure 4 – Average and standard deviation of a) uniaxial compressive strength b) ksn for every chronolithological unit. The number of values analysed for each lithology is displayed above the plots. Colours in graphs are in accordance with Fig. 1. Sources of compressive strength data: 1) Gokceoglu and Aksoy (2000) 2) Goudie (2006) and references cited therein 3) Goktan and Gunes (2005) 4) Karakus et al. (2005) 5) Pye et al. (1986) 6) Kahraman et al. (2002). Fluvial erodibility K for each chronolithological unit normalised against the weakest rock type derived from relative UCS values (c) and relative ksn values (d). In d), values of normalised K derived from ksn depend on the value of variable n in the stream power equation (Eq. 4) e) Linear regression of 291 valu es of ksn and UCS for every geological unit (geological map 1:100 000) with a 95 % confidence envelope in dashed grey lines and points coloured by chronolithological membership. R is the correlation coefficient. Classification of rock strength is based on Schmidt hammer values and other measures of rock strength summarised in Table 2, modified from Goudie (2006). Erodibility classification is based on rock strength (Table 2). 292 Table 2 – Classification of rock-type, strength and fluvial erodibility Description Schmidt Uniaxial Characteristic rocks Fluvial hammer compressive erodibility value strength Very weak rock – 10-30 1-20 Weathered and weakly Very crumbles under sharp compacted erodible blows with geological pick sedimentary rocks – point, can be cut with rock salt, marls pocket knife Weak rock – shallow cuts 30-40 20-35 Weakly cemented Erodible or scraping with pocket sedimentary rocks – knife with difficulty, pick siltstones and point indents deeply with conglomerates firm blow Moderately strong rock – 40-50 35-75 Competent Moderately knife cannot be used to sedimentary rocks – resistant scrape or peel surface, limestone, dolomite, shallow indentation sandstone, shale, slate, under firm blow from pick schist point Strong rock – hand-held > 50 > 75 Competent igneous and Resistant sample breaks with one metamorphic rocks – firm blow from hammer granite, migmatite, end of geological pick granodiorite, basalt 293 Rock strength classification and descriptions modified from Table 2 in Goudie (2006) 294 4 Results 295 In plan view (Fig 1a), the drainage divide follows a sinuous form, with lengths mostly configured to 296 ENE-WSW and WNW-ESE orientations. The divide tends to occupy central positions in relation to the 297 topography of the range apart from a notable southerly segment in the middle of the study area. 298 Only in the eastern end is there a marked deviation in divide orientation where a small length (20 299 km) changes to NW-SE. 300 With respect to erodibility, the drainage divide from east to west runs over bedrock of differing 301 erodibilities, quantified through ksn and UCS measurements. 302 303 304 4.1 Ksn and fluvial erodibility 305 From east to west, the drainage divide runs over moderately resistant to erodible and very erodible 306 sedimentary cover, moderately resistant meta-sediments, and resistant crystalline basement (Fig. 1). 307 Red beds, other clastic sediments and limestones in the Mesozoic cover have mean ksn values of 37, 308 59 and 74 m0.9 with standard deviations of 25, 42 and 49 m0.9 respectively (Fig. 4c). Palaeozoic meta- 309 sediments average at 85 m0.9 with a standard deviation of 41 m0.9, whilst the crystalline basement 310 has the highest ksn at 146 with a standard deviation of 52 m0.9 (Fig. 4c). Fluvial erodibility K values 311 based on Eq. 4 and ksn, normalised to the Mesozoic red beds (the weakest lithology: section 3.2), 312 vary by a factor of four to fifteen (Fig. 4d). The normalised K value for red beds is 1.0, with other 313 clastics 0.40 or 0.63 and limestone in the Mesozoic cover at 0.25 or 0.50, depending whether n = 2 314 or n = 1 (Fig. 4d). Normalised K for Palaeozoic meta-sediments is 0.19 or 0.44 whilst crystalline 315 basement has the lowest normalised K at 0.066 or 0.26, about four to fifteen times less erodible 316 than the Mesozoic red beds. 317 4.1 UCS and fluvial erodibility 318 Similar to the ksn of rock units, the fluvial erodibility of bedrock along the length of the drainage 319 divide based on UCS varies from alternating very erodible to moderately resistant rock in the 320 sedimentary cover, to moderately resistant in the meta-sediments and resistant in the crystalline 321 basement. The Mesozoic sedimentary cover has the largest range in UCS of the three 322 chronostratigraphic packages, with red beds, other clastic sediments and limestones averaging 12, 323 36 and 57 MPa respectively, with standard deviations of 7, 12 and 18 MPa (Fig. 4a). UCS values for 324 the Palaeozoic meta-sediment averages at 55 MPa with a standard deviation of 19 MPa, similar to 325 the Mesozoic limestones (Fig. 4a). The igneous rocks of the crystalline basement have the highest 326 UCS values with a mean and standard deviation of 101 and 20 MPa. Grouping of rocks into five 327 classes is based on a table from Goudie (2006) which is based on Schmidt hammer values and other 328 measures of rock strength (Table 2), and classes are marked on the UCS graph (Fig. 4a). Fluvial 329 Figure 5 – Gilbert metrics for each length of the drainage divide (Fig. 1) from west to east. a-c) headwater channel mean and standard error elevation, gradient and relief respectively. d) Potential direction of divide migration indicated by the direction of mobility of the divide from the Gilbert metrics. Values are standardized to show the direction of mobility. Bars show the standard deviation and shaded boxes show bootstrap confidence intervals. 330 erodibility K values based on Eq. 5 and UCS, normalised to the weakest unit, vary by two orders of 331 magnitude (Fig. 4b). The normalised K value for red beds is 1.0, with other clastic and limestone in 332 the Mesozoic cover at 0.1 and 0.05 respectively. Normalised K for Palaeozoic meta-sediments is 0.05 333 whilst crystalline basement has the lowest normalised K at 0.015, about two orders of magnitude 334 less erodible than the Mesozoic red beds. Categories of erodibility suggested based on the 335 lithological grouping into rock strengths (Table 2) and their respective normalised K values based on 336 UCS are defined in Fig. 4b. 337 4.2 Drainage divide mobility 338 Normalised cross-divide delta values of the Gilbert metrics following Forte and Whipple (2018) show 339 the magnitude and direction of mobility varies along the divide and that the divide is stable in the 340 central length of the divide (Fig. 5). Elevation of the drainage divide is high in both the east and west, 341 around 2600-3000 m in DIV 1 and DIV 4-5, and lowest in the middle length at ~ 2500 m (Fig. 5a). This 342 low elevation length (DIV 2-3) of the divide in the centre of the study area has equal headwater 343 hillslope gradients and relief on both sides of the divide (Fig. 5b-c) suggesting a stable drainage 344 divide in the meta-sedimentary rocks (Fig. 1, 5d). From west to east, headwater hillslope gradients 345 and relief decrease from 0.6 (35°) and 500 m in the crystalline basement of DIV 1 to 0.3 (15°) and 346 220 m in the Mesozoic sedimentary cover of DIV 5. Values of headwater channel elevation, hillslope 347 gradient and relief differ either side of the drainage divide in the eastern (DIV 1) and western (DIV 4- 348 5) lengths (Fig 5b-c). In DIV 1, the Gilbert metrics indicate northward movement of the divide (Fig. 349 5d) towards the centre of the exposed crystalline basement. This mobility is based on headwater 350 values which are on average 169 m higher in elevation to the north, with lower hillslope gradients 351 and relief, which differ by 0.06 (3.4°) and 50 m of relief (10%) across the divide (Fig. 5a-c). In DIV 4 352 and 5 in the Mesozoic cover the divide is significantly mobile towards the south (Fig. 5d), with 353 headwater channel elevation contrasts of 96-113 m, a difference in hillslope gradients of 0.07-0.09 354 (4-5°) and relief 44-55 m (14-30%) (Fig. 5a-c). 355 5 Discussion 356 5.1 Rock strength, ksn and fluvial erodibility 357 The theory and empirical relationships outlined in sections 3.1 - 3.2 predict that fluvial erodibility, K 358 relates to UCS in an inverse square (Eq. 5) and to ksn in an inverse relationship in which the power 359 depends on the value of n (Eq. 4). We find a linear relationship between UCS and ksn, which suggests 360 that n = 2, such that ksn and UCS scale according to a linear relationship. However, visual inspection 361 of figures 4c and 4d suggests that normalised K values calculated from ksn data where we assume n = 362 4 are more similar to the normalised K values derived from UCS data. However, since Lague (2014) 363 found that n ~ 2 is most commonly observed in the field, and our regression of UCS and ksn data 364 yields a strong linear fit consistent with n = 2, we suggest that n ~ 2 and that any difference between 365 UCS and ksn-derived normalised K values is due to other effects. For example, UCS does not explicitly 366 include other factors influencing bedrock erodibility including the degree of weathering and 367 structural discontinuities (Table 1), which especially through zones of deformation, will lead to more 368 rapid erosion of even hard rocks (high UCS) by fluvial systems. For example, Anton et al. (2015) and 369 Baynes et al. (2015) showed that canyons can be created by extreme flood events in basalt and 370 granite respectively, where the presence of discontinuities enables rapid erosion through fluvial 371 plucking and block topple. On the other hand, whilst the stream power model of bedrock river 372 erosion only accounts for changes in river channel slope, field studies show that rock strength 373 correlates with channel width (Allen et al., 2013), as well as valley width (Schanz and Montgomery, 374 2016) and can influence the efficiency of river bed load in eroding underlying bedrock (Brocard and 375 van der Beek, 2006). Furthermore, there can be a dampening of ksn value variations across 376 lithological boundaries as sections of river with weak bedrock downstream of river reaches with hard 377 bedrock can be armoured with blocks (e.g. Thaler and Covington, 2016). Based on the lithological 378 effects on river channel and valley morphology demonstrated by these field studies, using ksn as a 379 measure of rock erodibility in the stream power model of bedrock river incision likely 380 underestimates the effect of lithology on river erosion. 381 Thus, calculated through UCS measurements, K is expected to vary by two orders of magnitude (Fig. 382 4c), whereas if using ksn, K is expected to vary by one order of magnitude only (Fig. 4d). The lack of 383 constraints of K in natural settings have led to numerical modelling studies varying widely in the 384 range of erodibilities, using from one (Forte et al., 2016), two (Yanites et al., 2017), to three orders of 385 magnitude difference between rock-types (Roy et al., 2015). Such a range of inputs is often based on 386 the experimental relationship between intact rock strength and erosion in the classic abrasion mill 387 experiment done by Sklar and Dietrich (2001), and the five orders of magnitude difference in K 388 derived through the early work of Stock and Montgomery (1999). The latter forward-modelled river 389 paleo-profiles, constrained by bedrock strath terraces and basaltic layers, to presently observed 390 profiles for a range of locations worldwide. Their values of K range from 10-2 to 10-4 m0.2 yr-1 in the 391 Figure 6: The erosion and exposure of a hard limestone layer underneath a soft red bed layer in the Mesozoic low amplitude syncline leads to a divide mobility towards the southern catchment (Fig. 5). Panel a) shows the landscape at time t0 with the location of the divide between southern catchments with underlying sub-horizontal strata and northerly catchments with steeply dipping strata. b,c) As the rivers erode through the sub-horizontal strata of the southern catchment the ksn of the surface exposed needs to adjust to the new lithological strength, steepening transiently as it incises into harder limestone bedrock. During the response, erosion rates in the catchment drop before re-establishing equilibrium with uplift rates. On the other hand the northerly catchments do not experience such a change in ksn and experience steady rates. d,e) Graphs show the change in catchment-wide erosion rates either side of the drainage divide from t0 to t1 when the lithological contact gets exposed in the southern catchment to time t2 when erosion rates have reached steady state. The graph for the southern catchments shows the results from landscape evolution modelling of the exposure of a contact between sub-horizontal soft stratigraphy on top of a hard layer (Forte et al., 2016). The resultant change in erosion rates across the divide explains the migration of the drainage divide. The response depends on the factor of difference in erodibility, K, between units. Panel f) shows how the erosion of the lithological contact in the southern catchments leads to drainage divide migration to the south over the time of the transient response. 392 mudstones of humid continental Japan, to 10-6 - 10-7 m0.2 yr-1 in the subtropical granite landscape of 393 Australia (Stock and Montgomery, 1999). An issue with the experimental approach by Sklar and 394 Dietrich (2001) is that it does not include the effects of weathering and jointing of rock in natural 395 landscapes and their influence on fluvial erodibility. In the results from Stock and Montgomery 396 (1999) the influence of rock-type is difficult to isolate from the variation in climatic setting owing to 397 the spread of study locations. Thus, whilst these early studies give some first order estimates of 398 possible absolute values of fluvial erodibility and the relationship between rock strength and erosion 399 rates, our results constrain more fully the contrasts in fluvial erodibility between rock types which 400 may be expected within a mountain belt. 401 Next to rock strength control on ksn, orographic enhancement of precipitation with greater 402 precipitation on the northern and western sides of the drainage divide may lead to in a decrease of 403 ksn values from south to north or east to west (Supplementary Fig. S3). However, there is no 404 evidence to suggest a significant difference in ksn values between north and southern or east and 405 western portions of the High Atlas (Supplementary Fig. S4). 406 5.2 Drainage reorganisation in sedimentary cover 407 The Gilbert metrics indicate drainage divide mobility where it crosses the sedimentary cover in the 408 east (Fig. 5d). Here, the Mesozoic sedimentary cover is gently deformed, resulting in gently dipping 409 strata punctuated by widely spaced thrusts and folds (Fig 1b, Fig 2a). For example, the Dades river 410 catchment south of the main drainage divide (Fig. 6a) incises into a long-wavelength syncline of 411 slightly dipping strata composed of weak continental red beds and hard limestones, whereas directly 412 north of the divide folding and thrusting is more closely spaced resulting in strata dipping at higher 413 angles to the surface (Fig. 6a). Our results show that ksn is correlated to rock strength (Fig. 4e), and in 414 the hard limestone ksn is higher than in the weak red beds (Fig. 4a,c). Whereas erosion through near- 415 vertical strata north of the divide result in near-stable ksn values through time (Fig. 6b), the 416 horizontal stratigraphy of the Dades river catchment to the south of the divide means ksn values 417 need to change through time to return to stable erosion rates (Fig. 6c) which equal rock uplift rates. 418 Consequently, there is a period of transience when ksn values adjust to the change in bedrock 419 erodibility that occurred when erosion of soft red beds exposed hard limestone along the majority of 420 the river catchment (Fig. 1a, 6f). This transition period explains the southwards movement of the 421 divide (Fig. 5d), as erosion rates stay more or less stable in the north (Fig. 6d) whilst transient ksn 422 values cause a temporary decrease in erosion rates in the southern river catchments (Fig. 6e). The 423 stratigraphic effect on river erosion rates presented in Fig. 6e was first demonstrated in a numerical 424 modelling study of river erosion through layered stratigraphy by Forte et al. (2016), who show that 425 for strata of variable erodibility dipping 5° or less the overall erosion rate of the landscape is 426 expected to change by several factors during incision. Here, we show that such an effect can lead to 427 migration of the drainage divide in the gently deformed sedimentary cover in collisional mountain 428 settings (Fig. 6f), where exhumation of hard and soft strata is isolated or offset by faults. 429 Simulations by Forte et al. (2016) show the exhumation of a stratigraphic contact with a factor of 430 two to ten difference in erodibility (Fig. 6e) could take 2 - 9 Ma to re-equilibrate in an area of 800 431 km2 with a relatively high amount of precipitation (1 m yr-1). Therefore, drainage divide migration in 432 response to the incision through soft red beds to hard limestones in the upper half of the 1500 km2 433 Dades catchment, representing a change in erodibility by a factor of 2 - 20 (Fig. 4), where rainfall is 434 on the order of 0.1 - 0.5 m yr-1, is expected to persist on a timescale of 106 – 107 yrs. 435 436 5.3 Divide migration driven by crystalline basement exhumation 437 Where crystalline basement is exposed we find the position of the drainage divide is shifting towards 438 the centre of this exposed resistant rock in the High Atlas (Fig. 5). In a numerical simulation, 439 Giachetta et al. (2014) found that when they imposed an erodibility gradient across a drainage 440 divide, which is representative of the exhumation of a crystalline basement such as in the High Atlas, 441 the drainage divide responded by moving towards the side of lower erodibility over a timescale on 442 the order of 106 yrs. Bonnet (2009) proposed that such a shift of the drainage divide is accompanied 443 by a split of catchments, creating more and smaller catchments on the side of lower erodibility. 444 Giachetta et al. (2014) also show that on the other side, a growth of larger catchments occurs. 445 Similarly, Bernard et al. (2019) found that the drainage divide in the Pyrenees follows the position of 446 strong plutons. This implies that our results show a transient stage of drainage divide migration in 447 response to exhumation of crystalline basement, where today’s drainage divide at the edge of the 448 crystalline basement is expected to be stable in the centre of the strong crystalline basement, or 449 might even continue reorganising within the basement to follow the exhumation of resistant 450 plutons. Giachetta et al. (2014) used two orders of magnitude difference in erodibility values to 451 model this effect, and here we show divide mobility can be driven by exhumation of basement that 452 is only a factor of two less erodible than the overlying meta-sedimentary rock if calculated through 453 ksn (Fig. 4d), and a factor of three less erodible if calculated through UCS (Fig. 4b). 454 455 5.4 Lithologically induced drainage divide mobility during the long term erosion of a collisional 456 mountain belt 457 The combination of estimations of contrasts in fluvial erodibility of rock types (Fig. 4), geomorphic 458 indicators of drainage mobility (Fig. 5), and considerations of their structural configuration in the Figure 7 –Conceptual model of the development of a collisional mountain belt and the behaviour of the central drainage divide in response to exhumation of lithostratigraphic units 459 460 High Atlas compared to numerical simulation studies (see sections 5.1 - 5.3) enables us to propose a 461 model of lithologically-induced drainage divide mobility during the erosion of a mountain belt. The 462 overall trend of the drainage divide, strike of faults and bedding planes follows the structural grain of 463 the mountain belt (Fig. 1), for example determined by the pre-existing structure of an extensional 464 rift such as in the High Atlas (Babault et al., 2012). We purport that changes in erosion rates as rivers 465 incise through strata of different erodibility will drive drainage reorganisation in collisional mountain 466 belts, where layers are close to horizontal and only gently deformed (Fig. 6, 7). This is because where 467 strata are deformed gently and offset by faults, local exhumation of contacts between soft and hard 468 rock leads to changes in erosion rates between catchments (Fig. 6). Consequent changes in erosion 469 rates across the drainage divide will lead to the migration of the drainage divide (Fig. 7a) as 470 illustrated by the mobility of the drainage divide in the High Atlas (Fig 5, 6), which could lead to 471 steady divide migration or instantaneous capture of catchments, such as shown in the Appalachians 472 (Gallen, 2018). The effect of rock type on drainage reorganisation will be strongest in early phases of 473 collisional mountain building, before the sedimentary cover erodes in the centre of the belt. In later 474 stages, minor reorganisation and capture could still occur closer to the thrust front on the margins of 475 the mountain belt where Mesozoic sedimentary strata are present. When deformed meta-sediments 476 become exhumed, the increase in dip and deformation of strata leads to more stable erosion as 477 rivers incise (Forte et al., 2016, Fig. 6), resulting in stable drainage divides as witnessed in the middle 478 of the study area (Fig 5, 7b). When crystalline basement gets exhumed, the drainage divide will 479 migrate into the centre of highly resistant rocks resulting in more drainage divide migration (Fig 5, 480 7c). The migration of the main drainage divide in a mountain belt has been shown to lead to 481 reorganisation of river catchments (Bonnet, 2009; Giachetta et al., 2014), imposing new boundary 482 conditions on river channels which change gradients and sediment loads (Forte et al., 2015) and the 483 ensuing response can result in a cascading effect, impacting geomorphic and stratigraphic systems 484 for millions to tens of millions of years (Beeson et al., 2017). The effects of lithologically-induced 485 drainage migration here described for a post-orogenic belt could be more complex in an active 486 mountain belt setting. 487 Conclusions 488 This study shows that in a collisional mountain belt, the drainage divide will be mobile in response to 489 changes in erosion rates of rivers incising into gently dipping and deformed strata of contrasting 490 erodibility in the sedimentary cover, and in response to the exhumation of strong crystalline 491 basement. A combination of rock mass strength measurements and ksn derived from a digital 492 elevation model constrain the contrasts in fluvial erodibility exhibited in the High Atlas to between a 493 factor of 4 calculated through ksn and two orders of magnitude calculated through UCS. In the stage 494 of a collisional mountain belt during which rivers incise through highly deformed meta-sedimentary 495 units of intermediate erodibility, rock-type induced boundary conditions affect river networks the 496 least. Based on our values of erodibility contrast and previous numerical models we estimate the 497 timescale of adjustment in response to changes in erodibility of exposed bedrock to be on the order 498 of 106-107 yrs. Our results demonstrate that the mobility of the drainage divide in a collisional 499 mountain belt can be driven by rock erodibility variation alone, which has implications for the 500 perception of autogenic dynamism of drainage networks and fluvial erosion in collisional mountain 501 belts, and the interpretation of their geomorphology and downstream stratigraphy. 502 503 Author contributions 504 J.R.Z. drafted the article with critical revision and intellectual input from co-authors according to the 505 following contribution: M.S., S.J.B., M.T., A.E.M.. J.R.Z., M.S., A.E.M. and S.J.B. performed the 506 Schmidt hammer measurements in the field, J.R.Z. performed the topographic analysis. J.R.Z., M.S. 507 and S.J.B. initiated the project. 508 509 Funding 510 Fieldwork to validate topographic analysis and take Schmidt hammer measurements during J.R.Z.’s 511 PhD was supported by a University of Plymouth PhD studentship and the Geological Remote Sensing 512 Group Student Award. 513 Acknowledgements 514 We thank Hugh Sinclair and Adam Forte for constructive comments which allowed us to improve the 515 paper quality. We thank Jean-Philippe Avouac for editorial guidance. 516 References cited 517 Allen, G.H., Barnes, J.B., Pavelsky, T.M., Kirby, E., 2013. Lithologic and tectonic controls on bedrock 518 channel form at the northwest Himalayan front. Journal of Geophysical Research: Earth Surface 118, 519 1806-1825. doi:10.1002/jgrf.20113 520 Anton, L., Mather, A.E., Stokes, M., Munoz-Martin, A., De Vicente, G., 2015. Exceptional river gorge 521 formation from unexceptional floods. Nat Commun 6, 7963. doi:10.1038/ncomms8963 522 Babault, J., Teixell, A., Arboleya, M.L., Charroud, M., 2008. A Late Cenozoic age for long-wavelength 523 surface uplift of the Atlas Mountains of Morocco. Terra Nova 20, 102-107. doi:10.1111/j.1365- 524 3121.2008.00794.x 525 Babault, J., Van Den Driessche, J., Teixell, A., 2012. Longitudinal to transverse drainage network 526 evolution in the High Atlas (Morocco): The role of tectonics. Tectonics 31, n/a-n/a. 527 doi:10.1029/2011tc003015 528 Baynes, E.R., Attal, M., Niedermann, S., Kirstein, L.A., Dugmore, A.J., Naylor, M., 2015. Erosion during 529 extreme flood events dominates Holocene canyon evolution in northeast Iceland. Proc Natl Acad Sci 530 U S A 112, 2355-2360. doi:10.1073/pnas.1415443112 531 Beeson, H.W., McCoy, S.W., Keen-Zebert, A., 2017. Geometric disequilibrium of river basins 532 produces long-lived transient landscapes. Earth and Planetary Science Letters 475, 34-43. 533 doi:10.1016/j.epsl.2017.07.010 534 Bernard, T., Sinclair, H.D., Gailleton, B., Mudd, S.M., Ford, M., 2019. Lithological control on the post- 535 orogenic topography and erosion history of the Pyrenees. Earth and Planetary Science Letters 518, 536 53-66. doi:10.1016/j.epsl.2019.04.034 537 Bonnet, S., 2009. Shrinking and splitting of drainage basins in orogenic landscapes from the 538 migration of the main drainage divide. Nature Geoscience 2, 766. doi:10.1038/ngeo666 539 Boulton, S.J., Stokes, M., 2018. Which DEM is best for analyzing fluvial landscape development in 540 mountainous terrains? Geomorphology 310, 168-187. doi:10.1016/j.geomorph.2018.03.002 541 Boulton, S.J., Stokes, M., Mather, A.E., 2014. Transient fluvial incision as an indicator of active 542 faulting and Plio-Quaternary uplift of the Moroccan High Atlas. Tectonophysics 633, 16-33. 543 doi:10.1016/j.tecto.2014.06.032 544 Boulton, S.J., VanDeVelde, J.H., Grimes, S.T., 2019. Palaeoenvironmental and tectonic significance of 545 Miocene lacustrine and palustrine carbonates (Aït Kandoula Formation) in the Ouarzazate Foreland 546 Basin, Morocco. Sedimentary Geology 383, 195-215. doi:10.1016/j.sedgeo.2019.01.009 547 Brocard, G.Y., van der Beek, P.A., 2006. Influence of incision rate, rock strength, and bedload supply 548 on bedrock river gradients and valley-flat widths: field-based evidence and calibrations from western 549 Alpine rivers (southeast France). In: Tectonics, Climate, and Landscape Evolution, vol.398, pp.101– 550 126. doi:10.1130/2006.2398(07) 551 Bursztyn, N., Pederson, J.L., Tressler, C., Mackley, R.D., Mitchell, K.J., 2015. Rock strength along a 552 fluvial transect of the Colorado Plateau – quantifying a fundamental control on geomorphology. 553 Earth and Planetary Science Letters 429, 90-100. doi:10.1016/j.epsl.2015.07.042 554 Clubb, F. J., Mudd, S. M., Milodowski, D. T., Hurst, M. D., Slater, L. J., 2014. Objective extraction of 555 channel heads from high-resolution topographic data. Water Resour. Res. 50, 4283– 4304. 556 doi:10.1002/2013WR015167. 557 Errarhaoui, K., 1998. Structure du haut-atlas : plis et chevauchements du socle et de couverture 558 (interpretations des donnees geophysiques et geologiques), p. 326 P. 559 Forte, A.M., Whipple, K.X., 2018. Criteria and tools for determining drainage divide stability. Earth 560 and Planetary Science Letters 493, 102-117. doi:10.1016/j.epsl.2018.04.026 561 Forte, A.M., Whipple, K.X., 2019. Short communication: The Topographic Analysis Kit (TAK) for 562 TopoToolbox. Earth Surf. Dynam. 7, 87-95. doi:10.5194/esurf-7-87-2019 563 Forte, A.M., Whipple, K.X., Cowgill, E., 2015. Drainage network reveals patterns and history of active 564 deformation in the eastern Greater Caucasus. Geosphere 11, 1343-1364. doi:10.1130/ges01121.1 565 Forte, A.M., Yanites, B.J., Whipple, K.X., 2016. Complexities of landscape evolution during incision 566 through layered stratigraphy with contrasts in rock strength. Earth Surface Processes and Landforms 567 41, 1736-1757. doi:10.1002/esp.3947 568 Gallen, S.F., 2018. Lithologic controls on landscape dynamics and aquatic species evolution in post- 569 orogenic mountains. Earth and Planetary Science Letters 493, 150-160. 570 doi:10.1016/j.epsl.2018.04.029 571 Giachetta, E., Refice, A., Capolongo, D., Gasparini, N.M., Pazzaglia, F.J., 2014. Orogen-scale drainage 572 network evolution and response to erodibility changes: insights from numerical experiments. Earth 573 Surface Processes and Landforms 39, 1259-1268. doi:10.1002/esp.3579 574 Gokceoglu, C., Aksoy, H., 2000. New approaches to the characterization of clay-bearing, densely 575 jointed and weak rock masses. Engineering Geology 58, 1-23. doi:10.1016/s0013-7952(00)00032-6 576 Goktan, R.M., Gunes, N., 2005. A comparative study of Schmidt hammer testing procedures with 577 reference to rock cutting machine performance prediction. International Journal of Rock Mechanics 578 and Mining Sciences 42, 466-472. doi:10.1016/j.ijrmms.2004.12.002 579 Goudie, A.S., 2006. The Schmidt Hammer in geomorphological research. Progress in Physical 580 Geography: Earth and Environment 30, 703-718. doi:10.1177/0309133306071954 581 Hughes, P.D., Gibbard, P.L., Woodward, J.C., 2004. Quaternary glaciation in the Atlas Mountains of 582 North Africa, in: Ehlers, J., Gibbard, P.L. (Eds.), Developments in Quaternary Sciences. Elsevier, pp. 583 255-260. doi:10.1016/S1571-0866(04)80131-3 584 Kahraman, S., Fener, M., Gunaydin, O., 2002. Predicting the Schmidt hammer values of in-situ intact 585 rock from core sample values. International Journal of Rock Mechanics and Mining Sciences 39, 395- 586 399. doi:10.1016/S1365-1609(02)00028-X 587 Kahraman, S., Fener, M., Kozman, E., 2012. Predicting the compressive and tensile strength of rocks 588 from indentation hardness index. J S Afr I Min Metall 112, 331-339. 589 Karakus, M., Kumral, M., Kilic, O., 2005. Predicting elastic properties of intact rocks from index tests 590 using multiple regression modelling. International Journal of Rock Mechanics and Mining Sciences 591 42, 323-330. doi:10.1016/j.ijrmms.2004.08.005 592 Katz, O., Reches, Z., Roegiers, J.-C., 2000. Evaluation of mechanical rock properties using a Schmidt 593 Hammer. International Journal of rock mechanics and mining sciences 37, 723-728. 594 doi:10.1016/S1365-1609(00)00004-6 595 Lague, D., 2014. The stream power river incision model: evidence, theory and beyond. Earth Surface 596 Processes and Landforms 39, 38-61. doi:10.1002/esp.3462 597 Mather, A.E., Stokes, M., 2018. Bedrock structural control on catchment-scale connectivity and 598 alluvial fan processes, High Atlas Mountains, Morocco. Geological Society, London, Special 599 Publications 440, 103-128. doi:10.1144/sp440.15 600 Nazir, R., Momeni, E., Armaghani, D.J., Amin, M.M., 2013. Correlation between unconfined 601 compressive strength and indirect tensile strength of limestone rock samples. Electronic Journal of 602 Geotechnical Engineering 18, 1737-1746. 603 Perne, M., Covington, M.D., Thaler, E.A., Myre, J.M., 2017. Steady state, erosional continuity, and 604 the topography of landscapes developed in layered rocks. Earth Surf. Dynam. 5, 85-100. 605 doi:10.5194/esurf-5-85-2017 606 Perron, J.T., Royden, L., 2013. An integral approach to bedrock river profile analysis. Earth Surface 607 Processes and Landforms 38, 570-576. doi:10.1002/esp.3302 608 Pye, K., Goudie, A.S., Watson, A., 1986. Petrological influence on differential weathering and 609 inselberg development in the Kora area of central Kenya. Earth Surface Processes and Landforms 11, 610 41-52. doi:10.1002/esp.3290110106 611 Roy, S.G., Koons, P.O., Upton, P., Tucker, G.E., 2015. The influence of crustal strength fields on the 612 patterns and rates of fluvial incision. Journal of Geophysical Research: Earth Surface 120, 275-299. 613 doi:10.1002/2014jf003281 614 Saadi, M., Hilali, E.A., Bensaid, M., Boudda, A., Dahmani, M., 1985. Carte Géologique du Maroc Notes 615 et mémoires (Morocco. Service Géologique). Editions du Service Géologique du Maroc. 616 Schanz, S.A., Montgomery, D.R., 2016. Lithologic controls on valley width and strath terrace 617 formation. Geomorphology 258, 58-68. doi:10.1016/j.geomorph.2016.01.015 618 Schwanghart, W., Scherler, D., 2014. Short Communication: TopoToolbox 2 – MATLAB-based 619 software for topographic analysis and modeling in Earth surface sciences. Earth Surface Dynamics 2, 620 1-7. doi:10.5194/esurf-2-1-2014 621 Sklar, L.S., Dietrich, W.E., 2001. Sediment and rock strength controls on river incision into bedrock. 622 Geology 29, 1087-1090. doi:10.1130/0091-7613(2001)029<1087:sarsco>2.0.co;2 623 Stock, J.D., Montgomery, D.R., 1999. Geologic constraints on bedrock river incision using the stream 624 power law. Journal of Geophysical Research. B 104, 4983-4993. doi:10.1029/98JB02139 625 Stokes, M., Mather, A.E., Belfoul, M., Faik, F., Bouzid, S., Geach, M.R., Cunha, P.P., Boulton, S.J., 626 Thiel, C., 2017. Controls on dryland mountain landscape development along the NW Saharan desert 627 margin: Insights from Quaternary river terrace sequences (Dadès River, south-central High Atlas, 628 Morocco). Quaternary Science Reviews 166, 363-379. doi:10.1016/j.quascirev.2017.04.017 629 Strong, C.M., Attal, M., Mudd, S.M., Sinclair, H.D., 2019. Lithological control on the geomorphic 630 evolution of the Shillong Plateau in Northeast India. Geomorphology 330, 133-150. 631 doi:10.1016/j.geomorph.2019.01.016 632 Teixell, A., Arboleya, M.-L., Julivert, M., Charroud, M., 2003. Tectonic shortening and topography in 633 the central High Atlas (Morocco). Tectonics 22, n/a-n/a. doi:10.1029/2002tc001460 634 Thaler, E.A., Covington, M.D., 2016. The influence of sandstone caprock material on bedrock channel 635 steepness within a tectonically passive setting: Buffalo National River Basin, Arkansas, USA. Journal 636 of Geophysical Research: Earth Surface 121, 1635-1650. doi:10.1002/2015jf003771 637 Whipple, K.X., Tucker, G.E., 1999. Dynamics of the stream-power river incision model: Implications 638 for height limits of mountain ranges, landscape response timescales, and research needs. Journal of 639 Geophysical Research: Solid Earth 104, 17661-17674. doi:10.1029/1999jb900120 640 Willett, S.D., McCoy, S.W., Perron, J.T., Goren, L., Chen, C.Y., 2014. Dynamic reorganization of river 641 basins. Science 343, 1248765. doi:10.1126/science.1248765 642 Willett, S.D., Slingerland, R., Hovius, N., 2001. Uplift, Shortening, and Steady State Topography in 643 Active Mountain Belts. American Journal of Science 301, 455-485. doi:10.2475/ajs.301.4-5.455 644 Wobus, C., Whipple, K.X., Kirby, E., Snyder, N., Johnson, J., Spyropolou, K., Crosby, B., Sheehan, D., 645 2006. Tectonics from topography: Procedures, promise, and pitfalls. Geological Society of America 646 Special Papers 398, 55-74. doi:10.1130/2006.2398(04) 647 Yanites, B.J., Becker, J.K., Madritsch, H., Schnellmann, M., Ehlers, T.A., 2017. Lithologic Effects on 648 Landscape Response to Base Level Changes: A Modeling Study in the Context of the Eastern Jura 649 Mountains, Switzerland. Journal of Geophysical Research: Earth Surface 122, 2196-2222. 650 doi:10.1002/2016jf004101