Impact of climate on landscape form, sediment transfer and the sedimentary record

The relationship between climate, landscape connectivity and sediment export from mountain ranges is key to understanding the propagation of erosion signals downstream into sedimentary basins. We explore the role of connectivity in modulating the composition of sediment exported from the Frontal Cordillera of the south‐central Argentine Andes by comparing three adjacent and apparently similar semi‐glaciated catchment‐fan systems within the context of an along‐strike precipitation gradient. We first identify that the bedrock exposed in the upper, previously glaciated reaches of the cordillera is under‐represented in the lithological composition of gravels on each of three alluvial fans. There is little evidence for abrasion or preferential weathering of sediment sourced from the upper cordillera, suggesting that the observed bias can only be explained by sediment storage in these glacially widened and flattened valleys of the upper cordillera (as revealed by channel steepness mapping). A detailed analysis of the morphology of sedimentary deposits within the catchments reveals catchment‐wide trends in either main valley incision or aggradation, linked to differences in hillslope–channel connectivity and precipitation. We observe that drier catchments have poor hillslope–channel connectivity and that gravels exported from dry catchments have a lithological composition depleted in clasts sourced from the upper cordillera. Conversely, the catchment with the highest maximum precipitation rate exhibits a high degree of connectivity between its sediment sources and the main river network, leading to the export of a greater proportion of upper cordillera gravel as well as a greater volume of sand.

glaciated catchment-fan systems within the context of an along-strike precipitation gradient. We first identify that the bedrock exposed in the upper, previously glaciated reaches of the cordillera is under-represented in the lithological composition of gravels on each of three alluvial fans. There is little evidence for abrasion or preferential weathering of sediment sourced from the upper cordillera, suggesting that the observed bias can only be explained by sediment storage in these glacially widened and flattened valleys of the upper cordillera (as revealed by channel steepness mapping). A detailed analysis of the morphology of sedimentary deposits within the catchments reveals catchment-wide trends in either main valley incision or aggradation, linked to differences in hillslope-channel connectivity and precipitation. We observe that drier catchments have poor hillslope-channel connectivity and that gravels exported from dry catchments have a lithological composition depleted in clasts sourced from the upper cordillera. Conversely, the catchment with the highest maximum precipitation rate exhibits a high degree of connectivity between its sediment sources and the main river network, leading to the export of a greater proportion of upper cordillera gravel as well as a greater volume of sand.
Finally, given a clear spatial correlation between the resistance of bedrock to erosion, mountain range elevation and its covariant, precipitation, we highlight how connectivity in these semi-glaciated landscapes can be preconditioned by the spatial distribution of bedrock lithology.
These findings give insight into the extent to which sedimentary archives record source erosion patterns through time.

K E Y W O R D S
channel steepness, climate, connectivity, glaciation, landscape evolution, sediment storage, sedimentary record 1 | INTRODUCTION Sediment export from mountain ranges controls the morphodynamic behaviour of lowland rivers (Baynes et al., 2020;Pfeiffer et al., 2019) as well as the physical characteristics of their sedimentary records (Quick et al., 2019;Watkins et al., 2020). An understanding of the controls on sediment export is therefore essential for constraining how river systems, and their associated hazards, have and will respond to changes in their climatic or tectonic boundary conditions (Micheletti & Lane, 2016).
It has often been assumed that sediment export occurs relatively instantaneously and is therefore proportional to erosion and sediment production rates in upstream catchments (Allen, 2008;Castelltort & Van Den Driessche, 2003;Molnar, 2004). In source regions, erosion rates are modulated by bedrock lithology and precipitation rates, which are expected to be recorded in the gradient of hillslopes and the steepness of river channels (Hurst et al., 2013;Johnstone & Hilley, 2015;Roda-Boluda et al., 2018). All else being equal, a landscape experiencing uniform erosion is expected to display steeper slopes where rocks are more resistant to erosion or where precipitation is lower, with steeper slopes providing the increase in erosion potential needed to counterbalance the effect of greater rock strength or deficit in discharge, respectively (D'Arcy & Whittaker, 2014;Kirby & Whipple, 2012;Zondervan et al., 2020). On this basis, one might assume that the stratigraphic record preserved downstream of such mountain ranges will have a lithological composition that reflects upstream bedrock erosion rates and thus can be used to reconstruct the lithological and tectonic evolution of the sediment routing system (Amidon et al., 2005a(Amidon et al., , 2005b. Indeed this assumption is at the root of cosmogenic and thermochronology techniques for estimating catchment-averaged rates of bedrock erosion (10 2-5 years) and bedrock exhumation (> 10 6 years), respectively (Delunel et al., 2020;Kirstein et al., 2010;Reinhardt et al., 2007;Riesner et al., 2019;von Blanckenburg, 2005).
While it is expected that sediment production is dependent on the tectonic, lithologic and climatic drivers controlling landscape form, modelling (Tofelde et al., 2019), experimental (Jerolmack & Paola, 2010) and field studies (Clapp et al., 2000) have shown that sediment export has a more complex relationship with the geomorphological evolution of the sediment routing system. This is particularly evident in landscapes that have experienced glaciation. In Such a mechanism can reduce the concavity of longitudinal river profiles (Brocklehurst & Whipple, 2002). It can also widen valleys and increase hillslope erosion through the de-buttressing of steepened slopes upon glacier retreat (Ballantyne, 2002). A number of studies have reported that sediment yields from post-glacial catchments are heavily influenced by sediment storage in wide, low gradient reaches over a range of timescales (Brardinoni et al., 2018;Church & Ryder, 1972;Harbor & Warburton, 2006;.  recognised that these reaches act as sediment flux capacitors, which modulate the release of sediment downstream as a function of water and sediment discharge. A number of field studies have recognised that, while climate wetness is a critical driver of sediment export, differences in the timing of sediment delivery and terrace aggradation on piedmonts is fundamentally linked to temporal variability in landscape connectivity Norton et al., 2016;Schildgen et al., 2016). Here, landscape connectivity is defined by how easily sediment can be mobilised both longitudinally, along the length of a river network and between hillslopes and channels, thereby influencing the efficiency of landscapes in storing sediment (Hooke, 2003;Jerolmack & Paola, 2010;Li et al., 2016;Molnar, 2001;Rainato et al., 2018). Longitudinal connectivity can be modulated by, for example, bed slope and grain size thresholds in sediment transport, which can differ between river reaches (Whitbread et al., 2015). Connectivity between hillslopes and the river network is modulated by catchment geometry, river planform, hillslope morphology and hillslope processes (Mishra et al., 2019;Norton et al., 2016). While connectivity may be established in one part of the sedimentary system, it can have negative feedbacks elsewhere along its length (Korup et al., 2010;Lane et al., 2017;Rainato et al., 2018). For example, in the Humahuaca basin of NW Argentina, Schildgen et al. (2016) documented differences in the timing of terrace aggradation in response to an increase in precipitation-induced landsliding. They suggested that the propagation of the precipitation-induced landslide flux to the downstream alluvial system was dependant on whether or not the increased hillslope activity, as well as the basin's shape, promoted temporary damming of the river network. In the Tian Shan,  documented differences in the amplitude of piedmont aggradation and incision cycles in response to a period of increased precipitation. Cycle amplitudes were found to be higher downstream of catchments that had been partially glaciated in their upper reaches, when compared with catchments that were entirely glaciated or strictly fluvial. They related this trend to longitudinal dis-connectivity along the river network of semi-glaciated catchments.
It is known that both hillslope-channel and longitudinal connectivity are driven by water discharge but limited by sediment supply  (Tofelde et al., 2019). The evolution of connectivity along the length of coupled mountain-basin river systems and how it impacts sediment export, however, is still largely undocumented. Filling this research gap is critical if we are to have a better understanding of how sedimentary systems and their associated hazards will evolve in response to future climatic change.
In this contribution, we examine to what extent the lithologic and climatic controls on landscape form are translated downstream into the composition of sediments deposited on three adjacent alluvial fans. In comparing three apparently similar catchment-fan systems, we observe the role of connectivity in modulating sediment export from semi-glaciated catchments, within the context of an along-strike precipitation gradient.

| STUDY AREA
The Frontal Cordillera of the Andes between 30 S and 31 S was uplifted during the Neogene as one of a number of east-verging structures activated by the progressive shallowing of plate subduction ( Figure 2) (Gonzalez et al., 2020;Jordan et al., 1993;Martinod et al., 2020;Ramos et al., 2002). Thick-skinned deformation of Palaeozoic basement created the high topography of the Frontal Cordillera during a main stage of uplift in the late Miocene between~9.5 and 4.5 Ma (Jordan et al., 1993), which generated 2 ± 1 km of relief (Hoke et al., 2014). Since~5 Ma, active shortening has migrated to the eastern Precordillera and to the Pampean range, leaving the Frontal Cordillera and western Precordillera influenced primarily by strikeslip fault activity (Siame et al., 1997). The current tectonic stability of this region makes the Andean cordillera at 30-31 S a good location to investigate the impact of lithologic and climatic forcing on landscape evolution.
In this study, we focus on five adjacent catchments, east of the Frontal Cordillera drainage divide, which feed three alluvial fans in the Iglesia basin (Harries et al., 2018(Harries et al., , 2019. The Palaeozoic basement exposed in these catchments is composed of two dominant lithological units of similar age ( Figure 2a). The upper cordillera is composed of a sequence of volcaniclastic units from the Early Permian-Triassic, comprising tuff, ignimbrites and agglomerates interbedded with limestone. The lower cordillera is composed of a predominantly fluvial sequence of sedimentary units including sandstones, conglomerates, limestone and marine shales making up the Silurian-Early Permian Agua Negra Formation. This formation is intruded by granites emplaced during the Permian-Triassic (Jones et al., 2016). The highest peaks in the cordillera are capped with the Late Pliocene-early Pleistocene Olivares basalt in the south, and elsewhere by Oligocene-Miocene volcanic rocks, predominantly andesite and rhyolite, from the Cerro de las Tortolas and Dona Ana formations, which are associated with the El Indio-Pascua belt exposed to the west of the drainage divide (Jones et al., 2016). The clear differentiation between the volcanic/volcaniclastic rocks exposed in the upper cordillera and sedimentary rocks in the lower cordillera allows sediment sourced from these two regions to be distinguished. This is useful as the boundary between these two units is exposed at the approximate elevation of the Pleistocene Equilibrium Line Altitude (ELA),~4,000 m (Clapperton, 1994), and therefore, broadly speaking, identifies volcanic reaches as formerly glaciated, while sedimentary reaches of the lower cordillera were not. There are no terminal moraines preserved at this latitude to verify glacial extent (D'Arcy et al., 2019), however the altitude of the Pleistocene ELA can be extrapolated from the altitudes of terminal moraines preserved in nearby valleys, notably the Rio Aconcagua, Rio Mendoza and Elqui valleys. This ELA reconstruction is supported by observations of the relationship between elevation and precipitation gradients, snowline altitudes and the distribution of glaciers, in this region (Clapperton, 1994).
The climate in this mountain range is semi-arid and therefore lends itself to the study of physical surface processes given the absence of extensive vegetation (cf. Jeffery et al., 2014). While rivers here are ephemeral and are predominantly supplied by glacier and snow melt throughout the spring, extreme precipitation events, often associated with the El Niño Southern Oscillation, have an important role in sediment transport (Perucca & Martos, 2012), as observed in both flash flood events and in the Quaternary sedimentary record of this region (Colombo et al., 2009). The present ELA in the studied catchments is~5,000 m ( Figure 2b), peak elevations increase from 5,000 to 6,000 m in the south, where glaciers are larger and more F I G U R E 2 (a, b): Study location in the Argentine Andes, 30-31 o S, South America, including delineation of the three catchment-fan systems studied (fans labelled 1-3 with contributing catchments labelled 1-5). (a) Geological map. (b) Spatial distribution of annual precipitation averaged for the period 1970-2000. Here we include the Pleistocene (pink) equilibrium line altitude (ELA), at 4000 m asl (Clapperton, 1994) and plot the catchment area in km 2 above 4,000 m for each catchment. (c) Location of study site with respect to main tectonic units, including isobaths of the subducted oceanic slab, adapted from Ramos et al., (2002) [Colour figure can be viewed at wileyonlinelibrary.com] abundant (Harries et al., 2018). This trend in elevation is mirrored in precipitation rates which are on average 100 mm/year higher in the south than in the north. This latitudinal gradient in precipitation and elevation provides the ideal setup for exploring how spatially variable climate and topography can influence the geomorphic evolution of adjacent catchments.
Beyond the mountain front, a bajada of at least four generations of fan terraces extends into the Iglesia basin (Perucca & Martos, 2012). The development of these terraces lacks quantitative constraint, though elsewhere in the Andes, such cut and fill sequences have been linked to alternations between wet and dry periods on a variety of orbital timescales (Bekaddour et al., 2014;Litty et al., 2018;Norton et al., 2016;Siame et al., 1997;Steffen et al., 2009;Tofelde et al., 2017Tofelde et al., , 2019. We isolate three adjacent alluvial fans and focus on their modern rivers, which are currently incised~2 m into a surface interpreted to be Holocene in age (Harries et al., 2018(Harries et al., , 2019. The largest alluvial fan, fan 1, extends~40 km into the centre of the basin, and is fed by two large catchments (2 and 3 in Figure 2). Fans 2 and 3, situated south and north of fan 1, respectively, extend~25 km into the basin and are fed by catchments 1, and 4 and 5, respectively ( Figure 2). The base level of these river systems is modulated by the Rodeo lake, which today is controlled by a man-made dam. The opening and closure of the basin through the Holocene, by episodic damming of the Jachal River, is evident from lake deposits preserved in the Jachal valley (Colombo et al., 2009).

| APPROACH
To explore how the glacial conditioning of a landscape influences connectivity, sediment export and the construction of stratigraphic records, we draw on three lines of evidence; (i) river network morphology and its relationship with spatial variability in bedrock lithology and precipitation, (ii) channel-hillslope morphology and (iii) alluvial fan gravel composition. Combined, these methodologies allow us to test a hypothesis that the lithological and climatic controls on landscape form are directly translated downstream into their sedimentary record. In this idealised case, we would expect source regions that are experiencing the greatest erosion rates to be overrepresented in their sedimentary record (Litty et al., 2017). Where there is a deviation from this case, we can test the extent to which sediment storage and connectivity modulates sediment transfer through this postglacial landscape.

| Alluvial fan gravel analysis
As strong contrasts in rock resistance to abrasion could lead to significant, systematic downstream changes in relative abundance of clast types over distances of tens of kilometres (e.g. Attal & Lavé, 2006Lavarini et al., 2018), we first test to what extent abrasion might have influenced the composition of sediment exported from each catchment. We do this by analysing the relationship between a clast's lithology and its size along the length of each fan. This is an ideal place to test an abrasion hypothesis as transport distances are in the order of tens of kilometres and the supply of fresh clasts from hillslopes is limited along the fans.
To characterise the composition of sediment exported to each of three alluvial fans, we recorded the lithology and size of a statistically significant number of gravels, n~200 (Harries et al., 2018) at sites spaced at regular intervals along their fan length ( Figure 2). On these dry riverbeds we also documented the proportion of the bed covered by sand. Sampling began~3 km upstream of each bedrock canyon mouth and continued downstream at~3 km intervals up to the fan's toe. These datasets were collected from the same localities as grain size datasets previously published in Harries et al. (2018Harries et al. ( , 2019. At each locality, approximately 50% of the clasts were sampled on the surface of a gravel bar and 50% from the channel, using the Wolman point counting technique, that is, clasts were selected randomly from within a predefined area of~4 m 2 (Whittaker et al., 2010;Whittaker et al., 2011). The size of each clast was defined by its intermediate axis and was measured with a ruler. The lithology of each clast was categorised as intrusive, extrusive, sedimentary, metamorphic or quartzite. Whilst most clasts could easily be placed into these categories, the metamorphic overprint in some sedimentary clasts may have gone unrecognised, particularly for smaller pebbles; the metamorphic proportion of lithologies may therefore be underestimated.
We then isolated the lithologies that make up the coarse tail of the size distributions from those in the bulk of the distribution for all sites along the downstream length of each fan. The coarse tail was defined as clasts coarser than 84% of the sample (>D 84 ) where the bulk of the distribution was defined as all clasts in the finer 84% of the distribution (<D 84 ). This allowed us to determine, firstly, if clasts derived from certain lithologies were systematically coarser grained than others and, secondly, if clasts of certain lithologies disappeared from the coarse tails with increasing distance from the fan apex. The latter would be expected if grains from a particular rock type reduced in size faster than others through abrasion. We subsequently evaluated the possibility that some bedrock lithologies preferentially weathered to grain sizes smaller than gravel on hillslopes. This can be the case if sediments spend a greater amount of time weathering on hillslopes or, given certain lithological traits (Riebe et al., 2015;Roda-Boluda et al., 2018). To analyse the potential for preferential sand production, we therefore observe the relative proportions of sand and gravel deposited on each fan in the context of their respective catchment traits, where we would expect catchments with steeper slopes and greater areal exposure of certain lithologies to produce relatively less sand.

| Catchment analysis
As the cordillera is currently tectonically stable, spatial variations in slope and channel steepness are expected to be dependent on the spatial variability in erosion rates, modulated by lithology and precipitation rates (D'Arcy & Whittaker, 2014;Hurst et al., 2013). We analyse the relationship between channel steepness and the spatial variability in bedrock lithology and precipitation rates across the cordillera to determine whether erosion, and therefore sediment production, is spatially variable. We then analyse the impact of glacial erosion on channel steepness and valley morphometry by testing the conceptual model outlined in Figure 1: main valleys influenced by glacial erosion are expected to be wider and less steep than those that are strictly fluvial, with tributaries notably steeper than their main valley (MacGregor et al., 2000). Finally, we use these observations to make predictions of what the composition of sediment ought to be given the erosional trends observed.

| Bedrock lithology
The lithology of bedrock exposed in each source catchment was derived from a geological map produced by the Servicio Geol ogico Minero Argentio, SegemAR. The units mapped ( Figure 2) were not specific lithologies but formations or groups of rock units that were comparable to our pebble classification scheme, with the exception of quartzite and other metamorphosed lithologies which were found within the sedimentary Agua Negra Formation. These data were used in two ways: firstly, to quantify the relationship between channel steepness and bedrock lithology and, secondly, to compare the proportions of different pebble lithologies on the alluvial fans with their proportional areal exposure in each source catchment.

| Topographic analysis
We downloaded a 1 arc-second (approximately 30 m) resolution SRTM digital elevation model (DEM) for the study area from the USGS EarthExplorer database. Topographic analysis was performed using the LSDTopoTools algorithms (Mudd et al., 2014, detailed below. The topographic gradient was calculated by fitting a polynomial surface to a local window of elevations with a leastsquares regression (Zevenbergen & Thorne, 1987), the details of which are outlined in Hurst et al. (2012).
We calculated the relative steepness of channels. Channels in headwaters tend to be steeper than channels further downstream, so some normalization is required to compare the steepness of channels at different points in the channel network (e.g. Wobus et al., 2006).
Many authors have noted the power-law relationship between slope and discharge or its proxy drainage area (e.g. Morisawa, 1962), and this has been used to define a channel steepness index (e.g. (Wobus et al., 2006;Kirby & Whipple, 2012). The normalised channel steepness index, k sn , at any point along the river network, i, can be defined as the product of local channel slope, S, and effective discharge, Q: This requires the use of a reference concavity index, ref (Wobus et al., 2006;Kirby & Whipple, 2012).
Calculation of k sn directly from Equation 1 is problematic because topographic slope can vary both due to reach-scale variations as well as errors in topographic data. As a result, some degree of smoothing is required (e.g. Wobus et al., 2006).
An alternative to using slopes is to use the elevation of the channel directly to calculate steepness through the derivation of a longitudinal coordinate that integrates discharge or drainage area, as first proposed by Royden et al. (2000).
The integrated coordinate can be illustrated by reference to an ideal channel where the elevation is perfectly defined by Equation (1).
We can rearrange Equation (1) to solve for channel slope and, because slope is the derivative of elevation with respect to distance along the channel, we can solve this equation for elevation via integration (e.g. Royden et al., 2000). This results in an equation: where Q 0 is a reference discharge that ensures the integrand in Equation 2 is dimensionless and x b is the distance of an arbitrary base level (e.g. Royden & Taylor Perron, 2013). We then define a coordinate, χ, which is of dimension length, as We calculate χ for the channel network using Equation (3). We define Q 0 = 1 (for any chosen unit of discharge), so that the slope of the channel profile in χ-elevation space (elevation is derived from the DEM) is equal to the channel steepness index. If the concavity index is fixed to the reference value, ref , then the slope in χ-elevation space is equal to the normalised channel steepness index.
As we do not have direct measurements of discharge, it must be approximated. In this study, we do this in two ways. The first method, which is used in most studies, is to set discharge as proportional to drainage area (e.g. Kirby & Whipple, 2012;Wobus et al., 2006): We can then define the χ coordinate as Where we use χ A to denote that this coordinate is calculated using drainage area as a proxy for discharge.
In areas with strong precipitation gradients such as ours, however, discharge is not linearly proportional to drainage area. To overcome this limitation, we calculate the χ coordinate by accumulating precipitation (Babault et al., 2018): Where χ P denotes the chi coordinate calculated using precipitation. The precipitation data used to compute Equation (5)  We denote the normalised steepness indices calculated using drainage area and precipitation as k sn,A and k sn,P , respectively. Comparison of normalised steepness indices calculated using the area-based approach to those calculated using the precipitation approach can be a challenge since these naturally take different values. We therefore calculate the z-score Z ksn for each dataset and compare the two datasets using the difference between their z-scores. At each point i along the river network, the difference between the two z-scores is expressed as where k sn,P,i and k sn,A,i are the values of k sn calculated at pixel i for k sn,P and k sn,A , respectively, k sn,P and k sn,A are the mean values for each dataset, and σ kP and σ kA are their standard deviations.
The relationship between k sn,P and bedrock lithology is then analysed by grouping channels into the lithological formations they flow over and observing the distributions of z-scores of k sn,P .

| Landscape connectivity and sediment storage
Based on the trends mapped out by channel steepness indices, and accounting for the spatial variation in lithology and precipitation, the gravel lithology datasets are used to determine whether the areas of the catchments that are expected to experience greater erosion contribute a larger proportion of sediment to their alluvial fans. In doing so, we test to what extent the sediment exported from these catchments represents a spatially integrated sample of upstream erosion.
We then combine this analysis with morphometric mapping of hillslopes and channel deposits to determine whether sediment storage and landscape connectivity, both longitudinally and between hillslopes and channels, could have had a role in modulating sediment export from the cordillera.
Estimates of longitudinal connectivity would typically require detailed channel geometry measurements to formally calculate sediment transport capacities along the downstream length of the river networks. While we do not have these constraints, we can approximate the potential for sediment transport along the river network using the channel steepness indices, as bed slope and river discharge are dominant controls on sediment transport . Such an approximation of longitudinal connectivity utilising k sn,P is possible where slope is not dependent on the precipitation rate.
We explore hillslope-channel connectivity, using Google Earth to categorise catchment sediment deposits into three distinct groups based on morphological evidence for fluvial incision. We analyse all deposits with width > 100 m located within the main valley and > 5 km downstream of the headwaters, so to delimit the area of the catchment dominated by fluvial and not debris flow processes . Based on their morphological features, deposits are grouped as (1) floodplain deposits, (2) alluvial and colluvial fans that feed sediment directly to the main river channel, and (3) slope deposits and moraine that form terraces and ridges along valley sides. In combining these analyses, the role of climate in modulating connectivity and sediment export is explored within the context of the strong, along-strike precipitation gradient that prevails across the study area.

| RESULTS
The impact of past glaciation on the cordillera is evident from the cirque morphology of their headwaters, wide low-gradient valleys with steep tributaries, and the lateral and medial moraines that are preserved in their upper reaches (Figures 1, 2a).

| Slope and channel steepness
A southward increase in range elevation is correlated with steeper slopes (Figure 4a). There is a systematic trend across all catchments for tributaries in the upper cordillera to be over-steepened relative to those in the lower cordillera (Figure 4b). Selecting a single concavity that satisfies the entire river network is therefore not possible: the distribution of best-fit values appears to be bimodal, producing large uncertainties that are evident in Figure 4d. Channel steepness is highly variable in the upper cordillera; however, a best fit is achieved with a concavity index of 0.45 ± 0.05. The lower cordillera, delineated as the area <10 km upstream of the catchment mouth, on the other hand, has a lower best fit concavity index of~0.25 ( Figure 4e).
As the best fit concavities and uncertainty ranges determined for k sn,A and k sn,P are the same, we use a median value of 0.47 to calculate both ( Figure 4b). While recognising that the rainfall gradient does not appear to significantly influence profile concavities, the z-score distribution presented in Figure 4c does indicate that channels have a greater steepness (positive ΔZ i ) in the upper reaches when precipitation rates are accounted for.
In Figure 4e, we present chi-elevation profiles calculated using the best fit concavity derived for the lower cordillera, 0.25. Although this low value produces a weak relationship between slope and discharge at the catchment scale, it produces the best fit along the reaches of the river network that have the lowest steepness variability

| Hillslope-channel morphology
The mapping of sediment deposits within the cordillera ( Figure 6) shows that large hillslope and valley deposits are limited to the upper cordillera, in contrast with the deposits in the lower cordillera, which are fewer in number and smaller in area (beyond our mapping criteria).
Valley-side terraces dominate the catchments that supply sediment to fans 1 and 3 ( Figure 6). In some places, terrace scarps are more than

| Longitudinal connectivity
As we have demonstrated that channel steepness varies independently of precipitation rates and is not systematically affected by lithology, we can use k sn,P as a proxy for the sediment transport capacity (see methodology). We note that k sn,P , and therefore the transport capacity, is lower in the main channel of the upper cordillera ( Figure 4b), despite the elevated precipitation rates in these reaches ( Figure 2). In comparing k sn,P and k sn,A , we observe that the steepness index is higher in the upper cordillera of the southern catchments when the spatial variability in precipitation rate is accounted for (higher and lower ΔZ i , respectively). The steepness indices of the northern catchments, however, are comparable between methods.

| Sediment sourcing
The composition of gravels on three alluvial fans in the Iglesia basin are dominated by lithologies exposed in the lower reaches of their source areas. While the greater transport distance of upper cordillera clasts or contrasts in rock resistance to abrasion have been evoked to explain this trend in other river systems (e.g. Attal & Lavé, 2006Dingle et al., 2017;Lavarini et al., 2018), here we observe that all major lithologies are represented in each fan, with no evidence of systematic downstream changes in the relative abundance of clasts over distances of tens of kilometres (Harries et al., 2019). These observations, along with the fact that all major lithologies are represented in each fan, suggests abrasion cannot be a major control on sediment composition in this study area. Where we do observe an abrupt compositional change along the length of the largest fan, the distance at which this occurs corresponds with a major confluence, as F I G U R E 4 (a) Slope, (b) spatial distribution of k sn,P , (c) ΔZ i , the difference between the z-scores of k sn,P and k sn,A (Equation 6), d) best-fit concavity indices calculated using disorder metric for each catchment and (e) χ plots of all five catchments coloured according to lithology. Observe the disequilibrium between upstream and downstream reaches. With a concavity index of 0.25 (best fit for the lower cordillera), we observe tributaries upstream, above the Pleistocene ELA, that are variably over-steepened with respect to the main channel [Colour figure can be viewed at wileyonlinelibrary.com] such, we attribute the compositional change to the mixing and dilution of sediment sources on the fan and not abrasion.
We acknowledge that we cannot rule out the disappearance of particularly weak sub-lithologies through abrasion within a given lithological category before they reach the fan (e.g. some of the slightly higher in the sedimentary units of the lower cordillera. As the distributions of channel steepness, k sn,P are highly variable across all lithological classes, it suggests that bedrock lithology does not determine erosion rate (z scores, Figure 5). Indeed, the mapping of deposits within each catchment (Figure 2, 6) combined with the topographic analysis (Figures 1, 2, 4) reveal that erosion and sediment production may be enhanced in the upper cordillera. Consequently, we reject the hypothesis that an over-representation of sedimentary clasts in this setting is due to a higher rate of sedimentary bedrock erosion. An can be well explained by glacial carving, the low concavity of the lower cordillera is unexpected, though could well be attributed to the role of glacial sediment in formerly blanketing these reaches or downstream changes in channel width. The potential for valleys to both produce and store sediment following glaciation has long been recognised (Ballantyne, 2002). We find that the main transition between upstream reaches with low steepness indices and downstream reaches with high steepness indices occurs at an elevation of 3,800 m on all main river profiles, which approximates the Pleistocene ELA previously reported for the dry Andes,~4,000 m (Clapperton, 1994). While there are no terminal moraines to verify the limits of past glaciation in these catchments, we interpret the transition between upstream segments of low k sn and downstream segments of high k sn as a marker of the downstream extent of glacial erosion, cumulated over several Quaternary glaciations (Korup & Montgomery, 2008;Montgomery & Korup, 2011).

| Climate control on sediment export
An inverse correlation between the spatial distribution of channel steepness, which is lower in the upper cordillera (Figure 4b), and precipitation rate, which is higher (Figure 2), suggests that slope could be a function of precipitation pattern (D'Arcy & Whittaker, 2014). The morphology of the upper cordillera, however, indicates that glacial erosion, and not precipitation rate, has been the dominant control on channel gradient here (Anderson et al., 2006;MacGregor et al., 2009).
While accounting for the spatial distribution of precipitation rates in calculations of k sn results in higher channel steepness indices (Figure 4c), this increasing effect does not compensate for the low channel gradients in this region. Interpreting k sn,P as a proxy for the sediment transport capacity along the river network, we observe that there is a low potential for sediment transport, where k sn,P is lowest, in the upper cordillera, despite the enhanced precipitation focused on this region. The upper cordillera may be interpreted as being, to some extent, decoupled from the lower cordillera by its low gradient reaches. Such a decoupling is further evidenced by the presence of a knick-zone between the upper and lower cordillera.
While the decoupling of upstream reaches from those downstream can explain the under-representation of gravel sourced from upper cordillera bedrock on the fans, we observe that both the degree to which the upper cordillera is under-represented on the fans and the hillslope-channel morphology varies between the three catchment-fan systems. We first observe that the largest catchments feeding fan 1 has the lowest k sn,P , and therefore the lowest potential for sediment mobilisation, in its upper cordillera (Figure 4). These low-mobility reaches comprise large valley fills and terraces, which are indeed indicative of sediment storage ( Figure 6). Correspondingly, fan 1 has the lowest representation of upper cordillera clasts in its gravels.
We observe discrete differences between the two catchment-fan systems of similar size, fans 2 and 3. Fan 3's contributing catchments have a higher k sn,P in their upper reaches than the other catchments and therefore a higher potential for sediment mobilisation, yet the presence of extensive terraces in their upper cordillera indicates a high potential for sediment storage ( Figure 6). Furthermore, the alluvial gravels of fan 3 are dominated by sediments sourced from the lower cordillera. Conversely, the upper reaches of the catchment feeding fan 2 has a lower k sn,P and therefore lower potential for mobilisation, yet terraces are few in these reaches, and clasts from the upper cordillera are better represented on fan 2's fan ( Figure 6).
The transfer of upper cordillera clasts downstream does not therefore appear to be directly linked to channel gradient and discharge, described more generally as the longitudinal connectivity between the upstream and downstream reaches. Discrete differences in how sediment is being stored within the different catchments are recognised, which elude to hillslope-channel connectivity being an important modulator of long-term sediment export at a basin scale.

| Climate and hillslope-channel connectivity
The catchment sourcing fan 2 is unique in having few terraces and an abundance of large colluvial/alluvial fans that feed sediment from tributaries into the braided river channels of the main valley ( Figure 6). Fans can act as buffers of hillslope-channel connectivity when aggrading under high sediment supply conditions (Bowman, 2018;Mather et al., 2017). On the other hand, they can be effective hillslope-channel couplers if the sediment supply from upstream is limited and incision is focused on the fan itself, thereby promoting the remobilisation of previously stored sediment (Bowman, 2018;Mather et al., 2017). The majority of the alluvial/col- In contrast, there is an abundance of terraces with large scarps in the catchments feeding fans 1 and 3. While main valley incision in fluvial settings is generally thought to promote hillslope-channel coupling Whittaker et al., 2010), the creation of large terraces here suggest that valley incision has had an adverse effect on coupling in these widened, post-glacial valleys. Such large terraces not only store older hillslope sediments but also limit younger hillslope sediments in reaching the active channel. The evacuation of such terraced sediments ought to be dependent on the ability of the main channel to migrate laterally across the widened valley (Blum et al., 2013;Carretier et al., 2020). Given that the incision and aggradation dynamics of these sediment systems are dependent on the ratios of sediment and water fluxes delivered to them (Tofelde et al., 2019), we can explore these relationships in the context of spatially variable precipitation ( Figure 7). Where both precipitation rates and the abundance of glaciers are highest in the south, we observe the highest degree of hillslope-channel connectivity and the greatest relative abundance of upper cordillera sediment deposited downstream. We hypothesise that higher precipitation rates in the headwaters of the cordillera can promote the transfer of sediment from tributaries to the main valleys (Meigs et al., 2006). In the main valley, an ample supply of sediment limits river incision, thereby ensuring greater connectivity between hillslopes, tributaries and the main channel. This behaviour is observed in numerical models of drainage basin evolution where increased runoff intensities initiate expansion of the channel network and aggradation along the main channel while sediment supply remains high (Tucker & Slingerland, 1997

| Controls on glacial imprint and connectivity
The rate of bedrock erosion by warm-based glaciers, which co-varies with their downstream length, is known to be controlled by both climate and catchment morphology (Brocklehurst & Whipple, 2004;Cook et al., 2020;Koppes et al., 2015).  however that glacial erosion has not had a dominant role in limiting the relief of the mountain range (Brozovi c et al., 1997;Egholm et al., 2009;Pedersen et al., 2010). The fingerprint of glaciation in these catchments must therefore have been modulated by another factor controlling relief (Brocklehurst & Whipple, 2004).
While the steepness of channels draining different bedrock lithologies varies between catchments ( Figure 5), the distribution of relief across the cordillera does correlate with outcrops of hard bedrock, specifically the exposure of relatively young basalts in the southern cordillera, which create a plateau at high elevation. Given that the width of the mountain range does not vary along strike, the observed increase in hillslope gradient with range elevation suggests that the increased bedrock resistance to erosion in the southern catchments could contribute to the along strike variability in elevation (Stutenbecker et al., 2016). As both elevation and its co-variant precipitation strongly influence the elevation of the ELA, there is evidence to suggest that the fingerprint of glaciation is modulated by the erodibility of the bedrock sequence. More locally, outcrops of hard granite, which coincide with the transitions from fluvially to glacially influenced reaches in some catchments, may also play a role in modulating the rate of bedrock incision and the re-equilibration of the postglacial landscape upstream.
The maintenance of elevation in these catchments ensures that they also experience higher precipitation rates in post glacial periods, which promotes both continued alpine glacial activity and the delivery of a greater water flux to the river network. This greater discharge may support the evacuation of sediment from tributaries to the main valley, as inferred from fan 2's catchment morphology. In this way, the wider lithological controls on topography may also precondition the effectiveness of river networks in evacuating their long-term sediment supply and the translation of erosion signals to their basin sedimentary record.

| Wider implications
Numerous models of river network evolution assume that the size and volume of sediment exported from a catchment is a function of slope, drainage area and some consideration of how sediment discharge diverges or reduces in size with increasing transport distance (Gasparini et al., 2004;Hobley et al., 2011). A downstream reduction in coarse, gravel supply is most readily considered a function of clast abrasion during transport (Dingle et al., 2017;Lukens et al., 2016;. This study, based in an area with no significant contrasts in abrasion rates between the different lithologies exposed, demonstrates that erosion signals are also heavily modulated by the degree of connectivity between hillslopes and river channels. As we demonstrate that connectivity and sediment export from catchments is sensitive to climate through its influence on sediment delivery to the river network, this has important implications for our understanding of how climate influences the evolution of river networks (Gasparini et al., 2007;Hodge et al., 2011), particularly following glacial perturbation.
While we have demonstrated qualitatively that sediment storage and release is a critical modulator of sediment flux from these mountain catchments, there is an outstanding question regarding the timescales over which these processes operate. In these glacial-widened valleys, river terraces have been abandoned as the main river has incised, and in a number of catchments, terrace deposits have been sealed by modern valley-fill sediments, indicating long-term (>10 3 years) sediment storage. The presence of at least four large alluvial terraces on the Iglesia basin fans (Harries et al., 2019), however, provides evidence for a step-wise increase in sediment export once deposits are mobilised. These findings highlight the potential for a disproportionate increase in sediment export under future climate scenarios if sediments previously stored can be rapidly mobilised downstream (Clapp et al., 2000). As the mobilisation of sediment stores in mountain regions impacts the morphology of rivers from mountain front to coast, an understanding of their sensitivity to climatic change is essential for predicting future societal impacts.
Finally, given our primary observation that source areas along glaciated reaches are under-represented in the sediments deposited downstream of the mountain front (Blöthe & Korup, 2013), we highlight a potential bias in the application of cosmogenic radionuclide methods and thermochronology to sand and, more recently, riverbed gravels, to reconstructing erosion or exhumation in mountain catchments (Lavarini et al., 2018;Lukens et al., 2020). Our results underline the fact that a clear understanding of source area dynamics is critical for constraining the sensitivity of the sedimentary record to external forcing.

| CONCLUSIONS
Sediment deposited on three alluvial fans in the Andes is preferentially sourced from the lower reaches of their mountain catchments.
By examining the relationship between clast lithology and size, we demonstrate that this lithological bias cannot be explained by abrasion. Similarly, through an analysis of the steepness of longitudinal river profiles, we infer that the bias cannot be explained by the preferential erosion of the lower cordillera bedrock. Instead we conclude that the differences between the abundances of lithological classes on the alluvial fans and the distribution of bedrock types in the source catchments is explained by sediment storage in the upper cordillera. The segmentation of the river profiles fits well with a glacial erosion hypothesis where main channels in the upper cordillera have a lower gradient and their tributaries are systematically steeper.
Erosion focused in the upper cordillera is not communicated downstream to the alluvial fans, and it is likely that this sediment is being stored upstream in over-deepened and over-widened glacial valleys. cordillera clasts and sand. We therefore recognise hillslope-channel connectivity as a critical modulator of sediment export from these Andean catchments. By placing these trends within the context of an along-strike variation in topographic elevation and its co-variant precipitation, the importance of precipitation in promoting hillslope-channel connectivity and sediment export is recognised. As a result, we highlight the importance of considering the geomorphological evolution of source areas when reconstructing tectonic and climatic histories from stratigraphy.