What remains today of pre‐industrial Alpine rivers? Census of historical and current channel patterns in the Alps

To date, no survey on the diverse channel patterns existing prior to the major phase of river regulation in the mid‐19th–early 20th century has been elaborated at the scale of the whole European Alps. The present paper fills this knowledge gap. The historical channel forms of the 143 largest Alpine rivers with catchments larger than 500 km2 (total length 11,870 km) were reconstructed based on maps dating from the 1750s to 1900. In the early 19th century, one‐third of the large Alpine rivers were multi‐channel rivers. Single‐bed channels oscillating between close valley sides were also frequent in the Alps (28%). Sinuous and even more so meandering channels were much rarer. Historical river patterns generally followed an upstream–downstream gradient according to slope condition, floodplain width and distance from the sources. The local occurrence of certain channel patterns, however, primarily reflected the tectonic/orographic conditions. Multi‐channel reaches were widespread within the whole Alpine area, alternating with confined and oscillating reaches. This demonstrates that most areas were mainly transport‐limited rather than supply limited. Sinuous and meandering reaches were more frequent in the north‐eastern Alps and were characterized by lower denudation rates and less sediment delivery.


| INTRODUCTION
The typology of channel patterns or river styles (the planform geometry) is commonly used to classify rivers because it is very integrative in terms of geomorphic functioning (Brierley & Fryirs, 2005;Church, 2002;Kondolf, Montgomery, Piégay, & Schmitt, 2016). Such typologies are usually based on the number of flow channels within the floodplain and on channel sinuosity. Channel shifts, lateral valley confinements or un-vegetated bar extents are additional parameters considered. River pattern types can also be used as a proxy for specific ecological functioning because each type is characterized by specific habitat templates and dynamics (Amoros & Petts, 1996;Thorp, Thoms, & Delong, 2006). Characterizing channel patterns and their evolution at a regional scale is, therefore, meaningful to assess potential human pressures on river systems. Historical maps facilitate tracing channel planform changes through longer time periods (Bravard & Bethemont, 1989;Hohensinner, Jungwirth, Muhar, & Schmutz, 2011).
Although historical sources often show great inaccuracies in geographical position and depicted riverine structures, they provide the only possibility to reveal past channel forms at larger scales.
Numerous studies on the historical morphological state of Alpine rivers prior to river regulation programmes have been published in recent decades (Habersack & Piégay, 2007). Most of them focus on individual river sections or on the human-induced transformation of specific river systems (Arnaud, Schmitt, Johnstone, Rollet, & Piégay, 2019;Bertrand, Piégay, Pont, Liébault, & Sauquet, 2013). To date, investigations that cover several Alpine catchments on a regional scale have been conducted for various countries (e.g., Muhar, Schwarz, Schmutz, & Jungwirth, 2000;Müller, 1995;Piégay, Alber, Slater, & Bourdin, 2009). Surprisingly, no overview study is available on historical and current channel patterns covering the entire Alpine sphere. Moreover, a large-scale study on the physical controlling factors associated with different Alpine channel patterns is also missing.
The present study aims to fill this knowledge gap.
One of the most comprehensive investigations on Alpine channel styles was completed by Muhar, Kainz, Kaufmann, and Schwarz (1996) and Muhar, Kainz, and Schwarz (1998) for rivers with catchments larger than 500 km 2 in whole Austria. Accordingly, braided channel reaches made up 23% and meandering reaches even 25% in the early 19th century. Braided rivers in the French Alps were analysed in detail by Piégay et al. (2009), who distinguished seven types of channel braiding. Around 1800, at least 1,214 km of the river courses showed braided patterns. Controlling factors for the evolution of different braiding types were discussed, but other channel forms were not included in the study. Similar studies were conducted by Surian et al. (2009) for 12 alluvial rivers in northern Italy as well as by Comiti (2012) and Marchese, Scorpio, Fuller, McColl, and Comiti (2017) for mountain rivers in the Italian Alps. For Swiss Alpine rivers, the "Hydrological Atlas of Switzerland" (HADES) provides the best overview of the fluvial morphology at the mid-19th century and around 1990 (Koblet, 1995). It does not, however, provide quantitative data on the distribution of the former channel styles. Comparing the available studies reveals a focus on braided and to a minor extent transitional, that is, "wandering," river sections. This hinders drawing conclusions on the importance of such rivers with respect to the whole spectrum of Alpine channel forms.
This lack of basic data on Alpine fluvial morphology prior to industrialization means that our knowledge about anthropogenic impacts also remains fragmentary. Martinet and Dubost (1992) reported that only approximately 10% of the reaches of large rivers in the Alpine region can be classified as "near-natural," a conclusion also supported by vegetation studies on Alpine braided rivers in Germany (Müller, 1991(Müller, , 1995. Muhar et al. (2019) recently highlighted that at least 41% of the Alpine rivers with catchments larger than 10 km 2 have been hydrologically and/or morphologically human transformed.
At minimum, 23% have experienced some type of severe morphological alteration. Despite the diverse basic data and methods used for the previous studies, a common conclusion is that larger rivers generally suffer more intensive human transformations than smaller rivers or upstream headwater sections (Gurnell, Surian, & Zanoni, 2009;Muhar et al., 2019;Piégay et al., 2009;Surian et al., 2009).
The different methodological approaches used in the past hinder a direct comparison of the data or an Alpine-wide census of historical and current channel patterns. Against this background, the present study is the first methodologically coherent evaluation of channel pattern types prior to industrialization, that is, the onset of systematic river regulation programmes in the first half of the 19th century (Brown et al., 2018). We analysed the 143 largest rivers with catchment areas exceeding 500 km 2 in the whole Alpine arc in Europe based on numerous historical sources. Our study addresses both the pre-industrial channel patterns and the regional and longitudinal (upstreamdownstream) differences of human-induced channel changes. In differentiating distinct channel pattern types, we follow a hypothetical continuum of channel patterns along a river's course (based on Church, 1992, andStanford, 1995; Figure 1). In particular, the resulting GIS dataset helps answer the following research questions: (1) What channel patterns were characteristic of large Alpine rivers prior to major human modification?
(2) Which physical factors promoted the evolution of the observed channel forms?
The geographical location of a river section within the Alpine complex might play a role in the spatial distribution of channel patterns. For example, rivers draining to the north and east must follow F I G U R E 1 Simplified scheme of channel pattern types from the steep headwater to the lowland or delta in an Alpine lake (*potentially anabranching in wide Alpine valleys at sufficient river flow; compare Section 3) much longer distances to the sea (Figure 2). Beyond the regionally varying orographic setting (e.g., altitude, valley slope, valley width), the magnitude of the sediment supply governs channel evolution (e.g., Brierley & Fryirs, 2005;Buffington & Montgomery, 2013;Pont et al., 2009). Different lithological zones show varying areal denudation rates and sediment availability (Hinderer, Kastowski, Kamelger, Bartolini, & Schlunegger, 2013). We, therefore, also investigated the local geological conditions and those in the upstream catchments of the individual river reaches. For example, one might expect that historically braided rivers were more frequent in bedload-abundant Alpine areas formed by sedimentary rocks (Mueller & Pitlick, 2013. In this context, karst mountain ranges such as the Northern and Southern Limestone Alps are commonly held to favour multichannel rivers. Today, this assumption is supported by some of the last remaining large braided rivers such as the Fiume Tagliamento in north-east Italy (Ward et al., 1999).
(3) How have the Alpine channel patterns been human modified over time?
Historically, land use management and river engineering practices have varied regionally (Haidvogl, Pont, & Zwitter, 2019). Rivers in the Alpine countries today, therefore, reflect different forms and intensities of human pressure. The underlying question here is whether patterns in modification trajectory can be identified by country, channel type or longitudinal location along the river courses. This is tested against the generally supposed upstream-downstream gradient of increasing intensities of human interventions (Gurnell et al., 2009;Muhar et al., 2019).
The present study provides a solid basis for more detailed future investigations on the history of Alpine rivers. It highlights the importance of specific river types in different countries or geographical regions of the Alps. It contributes to our current understanding of the role of physical channel controls for the genesis of specific channel planforms. Finally, it helps to identify river reaches that were most intensively transformed along the river profiles. This is an important step forward in discussing the impacts of human interventions on Alpine river systems on a larger scale.

| Study site
The study includes the 143 largest rivers with catchment sizes over 500 km 2 within the area defined by the Alpine Convention. This includes rivers whose catchments are partly-or in some cases largely-located outside of the Alpine area ( Figure 2). The area specified by the international Alpine Convention measures approx.
190,000 km 2 . Beyond the main Alpine complex, it also includes some smaller areas of the Alpine foreland at its fringes. Limestone Alps and regions predominantly formed by dolomite and marble make up the largest part (36%) of the overall area (based on the "International Geological Map of Europe"; Asch, 2003). The Central Alps comprising crystalline and Palaeozoic rocks cover the second largest share (31%), followed by the sedimentary rock zone formed by Flysch, Faltenmolasse and Bünden schist (20%). The Molasse zone, marine sediments, together with inner-Alpine basins cover only 11%, and the smallest areas are formed by volcanic material in northern Italy (2%).  towards the Danube River in the north show the longest distance to the delta (i.e., Black Sea). This is reflected by the highest median altitudes of the river sections at the boundary of the study site (493 m a. s.l.). Analogously, rivers flowing to the west and discharging via the Rhone River into the Mediterranean Sea as well as those flowing to the Po River in the south and finally to the Adriatic Sea show the lowest altitudes (205 m a.s.l.). Today, the total length of all Alpine rivers with catchments larger than 10 km 2 is almost 60,000 km; Austria's Alps host the largest share (32%; Muhar et al., 2019). The largest catchment (13,400 km 2 ) is that of the Drava River, followed by the Rhone River with approx. 13,000 km 2 .

| Human drivers of channel changes
Notable human impacts on Alpine river systems started with the establishment of first human settlements and associated forest clearings for arable land during the late Neolithic and Bronze Age between 5,000 and 2,500 BP (Comiti, 2012;Jarman, Bailey, & Jarman, 1982).
These induced erosion processes, re-deposition of sediments in larger valley floors and consequently affected channel patterns (Goudie & Viles, 2016;Jacob et al., 2009). These processes intensified predominantly in the Italian and French Alps during Roman Times (Tinner et al., 2003). Humans again became a major geomorphological agent during the Middle Ages, when land reclamation, deforestation and salt and ore mining were significantly promoted also north of the Alpine crest (Comiti, 2012;Haidvogl et al., 2019). The significantly increased sediment supply caused a phase of intensified aggradation and braiding in several Alpine river systems (Bravard, 1989;Gurnell et al., 2009). Smaller streams were rerouted or water was abstracted from larger rivers to mills for energy extraction. In early Modern Times, direct human interventions in form of embankments, flood protection dikes and channel cut-offs increased, but larger channelization measures were only sparsely implemented before the 18th century (Girel, 2008;Vischer, 1989). Systematic channelization programmes started along several Alpine rivers in the early or mid-19th century, although most were first channelized in the late 19th century. The main purposes for the ambitious hydraulic projects were flood protection, land reclamation and the enhancement of the waterways for log driving and rafting Hauer et al., 2019). Such river training measures substantially reduced braided river sections throughout the Alpine sphere Surian et al., 2009).
Further river straightening programmes and the construction of dams and reservoirs for hydropower production during the 20th century additionally impaired the Alpine river systems (Comiti, 2012).

| Differentiation of channel pattern types
In analysing the historical and current channel patterns, we largely applied the classification scheme defined by Muhar et al. (1996Muhar et al. ( , 1998 for Austrian rivers. It proved to be a useful typological framework for identifying channel types based on historical maps (Table 1). Because the historical sources often do not enable a detailed differentiation of certain channel patterns, the category "multi-channel rivers" includes all its various morphological sub-types such as bar-braided, islandbraided, anabranching and anastomosing. Thus, it also comprises very slow-flowing, delta-type river reaches with multiple branches and riverbeds characterized by sand, silt and clay. These latter anastomosing channels made up only very small reaches in the Alps. Braided, bedrock-confined reaches without floodplain pockets are also included in the "multi-channel" category. Muhar et al. (1996Muhar et al. ( , 1998 identified a new "oscillating" river type that can be best described as "moderately confined single-channel river" (compare "partly confined rivers" in Brierley & Fryirs, 2005;Buffington & Montgomery, 2013).
At first view, such rivers are similar to sinuous ones, which typically develop distinct river bends in wide valleys or alluvial plains. Moderately confined single-channel rivers, however, are constrained by close valley sides, older and higher river terraces, or alluvial fans of confluent tributaries. Most of them, therefore, oscillate between the valley sides and cannot freely form larger river bends. For better readability of the text, tables and figures, we apply the short term "oscillating" because it best describes the channel configuration shown by the historical sources.
Based on the historical maps, we could not clearly identify the socalled "wandering gravel-bed rivers," which show a transitional channel form between fully braided and oscillating/sinuous patterns (Desloges & Church, 1989;Nanson & Knighton, 1996). Depending on map accuracy, they were most likely grouped with multi-channel rivers when a gravel corridor was mapped, in other cases also with oscillating rivers. Compared to typical braided rivers, they have fewer flow branches. Usually, one main flow channel is clearly dominant and the active channel, that is, water-covered area and un-vegetated sediment bars, is narrower.
Evaluating the current human-modified river courses called for defining additional channel patterns because they deviate substantially from the natural channel forms. Arch-shaped or arched regulated channels are largely bank protected and show forms similar to naturally oscillating rivers. The river bends, however, are mostly flatter and less pronounced to facilitate regulation. This applies even more so to river stretches that were subject to linear straightening. These have often been given what is known as a standard cross-section during regulation. They are mostly devoid of bars and, therefore, appear very monotonous today, resembling canals. Another human-caused channel form refers to dammed-up river reaches, including both impoundments of run-of-river power plants and large lake-like reservoirs.

| Data sources
We analysed the past channel types in the Alpine arc based on numerous historical maps ranging from 1750 to ca. 1900 (Table 2). They mostly stem from before the onset of systematic river regulation programmes in the 19th century. The scales of the cartographic sources differ considerably between 1:2,500 and 1:100,000. Fortunately, the large Alpine regions of the former Habsburg Monarchy  Similar map series are available for the German Alps, where the accurate cadastral maps ("Flurkarte") and the less detailed "Positionsblätter" were produced between 1808 and 1817 ( Table 2).
The French regions of the Alps are also well covered by two cartographic sources that were created consecutively between 1750 and 1866. The "Carte de Cassini" from the 18th century lacks information for small rivers, but provides a valuable supplement to the "Carte de l'État-Major" 1818-1866 created by military surveyors at an original scale of 1:40,000 (Bravard & Bethemont, 1989;Lefort, 2004;Piégay et al., 2009).
Because the available reproduction shows a scale of only 1:100,000, we additionally used the more detailed but younger "Siegfriedkarte" (also known as "Topographischer Atlas der Schweiz"). This map series is based on the original "Dufourkarte" surveys, but was considerably updated. Because both map series lack details in some areas, we con- basis and provide additional information on potential earlier channel changes. In rare cases, we also consulted historical topographical literature and views for additional information on the former channel forms.
The current state of the Alpine rivers is assessed based on numerous orthophotos from recent years and Google Earth, which facili-

| GIS and data analyses
Historical and current channel patterns of the 143 study rivers were identified according to the morphological attributes specified in Table 1 using ArcGIS 10.6. To determine the relevant rivers and their associated catchments, we used a GIS polyline-dataset comprising the current channel axes that were compiled in the project "Strategic Planning for Alpine River Ecosystems" (Muhar, Grüner, Böck, Scheikl, & Becsi, 2018). Already geo-referenced historical maps were available only for most of Austria, Slovenia and north-east Italy (WMTS-Services for the First and Second Military Surveys; see Table 2). Covering the remainder of the Alpine area would have required geo-referencing more than 1,000 map sheets, an unfeasible task due to the high costs and long processing time. The historical channel patterns were, therefore, mostly identified with nongeoreferenced maps using orthophotos as basemaps for proper geographical positioning. Historical channel types were added to the attributes of the GIS polylines. Afterwards, the lengths of the historical river sections were calculated with MS Excel based on the current section lengths, which were corrected for changes in channel lengths due to river straightening measures. Mean correction factors were calculated based on the defined ranges in sinuosity for the individual channel types (Table 1). A factor of 1.1 was applied for multichannel sections, 1.2 for sinuous and 1.65 for meandering river stretches. Confined channels and incised meanders were not corrected because they were generally not affected by straightening measures. The same holds true for oscillating channels, which, in most cases, were arch-shaped regulated and showed almost the same sinuosity. For the thereby resulting potential error range, see Section 5. The spatial arrangement of the 143 largest rivers highlights the principal configuration of the complex historical-and also current-Alpine channel network (Figure 3a). Several long river systems that run almost parallel to the Alpine arc are evident. Such running waters in longitudinal valley furrows are primarily located in the Austrian Eastern Alps, but also in the Swiss Alps at the Alpine Rhine and the Rhone.
These river systems are joined by numerous other, largely shorter ones, which flow more or less transversally, starting from the Alpine divide or the longitudinal valleys towards the edge of the Alps. In the early 19th century, the total lengths of all investigated rivers were approximately 11,870 km.
Within the network of the large Alpine rivers, the quantitative importance of the individual channel patterns differed from region to region. Almost a quarter of these featured confined channels. They are still present in the upper regions of the Alps but also in lower-lying narrow canyon sections. About a third of the studied network was occupied by multi-channel rivers (defined here as a fairly large class including different sub-types) in the early 19th century (Figure 4). This observation concurs with the widespread notion that this planform type was significantly representative of Alpine rivers prior to industrialization (Comiti, 2012;Liébault & Piégay, 2002). Oscillating river sections were comparably frequent in the Alps, that is, 28% of the total channel length. These were neither braided nor freely meandering.
They characterized the landscape of many Alpine valleys because the run of the river constantly altered its course between the sides of the valley. Sinuous and even more so meandering waterbodies were much rarer in the Alps, making up 7 and 5%, respectively, of the studied river courses. The side valleys also once featured sinuous or even meandering flowing waters. Nonetheless, examining only the larger main rivers shows that they were less significant. Only 2% of the total channel length were incised meanders, which can also be described as geologically fixed meander bends.
Importantly, the different channel patterns were not equally distributed ( Figure 3a). The Austrian Eastern Alpine region featured significantly more sinuous and, in particular, meandering river stretches than the rest of the Alps (each type 11% related to all large rivers in the Austrian Alps; Table 3). The German Alpine area also featured high shares of meandering rivers, but even more outstanding were the numerous sinuous river sections (16%) as well as the comparably high proportion of incised meanders. The situation in the Swiss Alps differed entirely: here, only 6% of the larger rivers were sinuous, and none were meandering. In France, Italy and Slovenia, these river types were below the Alpine average.

The situation of multi-channel rivers was different. The French
Alps showed the highest percentage, that is, 44% of the length of the large rivers. In Switzerland, Germany and Italy, they also made up between 35% and 39%. In Slovenia, in contrast, only 20% of the investigated river continuum showed forms of multiple channels.
Finally, oscillating channel patterns were relatively equally distributed in Austria, France, Germany and Switzerland (between 23 and 26%), but showed much higher values in Italy (35%) and Slovenia (38%).
F I G U R E 4 Distribution of channel planform types in the Alps in the early 19th century and in 2017 (left bars: historical, right bars: 2017; per cent values indicate the relative shares of the individual river types in total river lengths at the respective time) T A B L E 3 Lengths of historical and current channel patterns per Alpine country (in km and % of total river length in the respective country) 4.2 | Which physical factors promoted the evolution of the observed channel forms?
The prerequisite for answering this research question is to consider the physical factors behind the evolution of the different channel patterns. This helps highlight notable differences between the historical channel types with respect to the individual attribute. Subsequent multivariate analysis shows which of these factors significantly contributed to the occurrence of certain channel forms.
Only a partial differentiation of the historical river types can be made based solely on the altitudes of the river sectors ( Figure 5a). As  Thus, formerly braided rivers in France were reduced by 47% (Figure 9), but new, largely less braided ones were identified for 2017 in other river sections (see Section 5.3).
a Percentage related to current river length.
reflect the idealized sequence of channel patterns shown in Figure 1.
This pattern also inversely reflects the pattern of altitudes (Figure 5a).
The mean width of the valley bottom showed extraordinary ranges between a few metres in confined sections up to 8,600 m at the Rhine River. Except for multi-channel and sinuous rivers, the median values differed significantly (Figure 5d). By definition, confined sections were located in the narrowest valleys (median width 43 m).
Incised meanders that were also largely confined featured the second narrowest valley bottoms (median 127 m). Generally, meandering sections were located in the broadest valleys (777 m).
As commonly supposed, confined channels showed the highest median channel slope (4.0%; Figure 5e). Oscillating river stretches fea- In combination with the PCA, the sequence relation diagram in

| How have the Alpine channel patterns been human modified over time?
Since the early 19th century, the large Alpine rivers lost a stretch of about 510 km, namely 4.3% of the original extent, due to channel straightening. Two hundred years ago, 86% of the larger Alpine river courses were formed by confined, oscillating and multiple channels.
Today, these types contribute only 50% (Figures 3b and 4). Multichannel sections, historically 34% of the river courses, were reduced to 15%. In absolute numbers, these particularly "space-occupying" waterbodies-which include braided, anabranching and anastomosing river sections-showed the strongest decline. In total, 2,370 km (58%) of the formerly branched rivers have been straightened and regulated.
Meandering rivers suffered the greatest percentage loss of previous course length: 80% or 430 km. In contrast, regarding planform geometry, confined waters were the least frequently affected by human interference. Only 15% of these were turned into reservoirs. This value does not consider human impacts due to smaller consolidation and retention check-dams.
The analysis of the current river courses reveals three new forms of channel patterns that did not exist almost 200 years ago (see Section 3). To date, with respect to rivers with catchments exceeding 500 km 2 , about 3,730 km were arched or linear regulated (33% of the current river course length). Moreover, at least 1,330-km-long reaches (12%) were transformed into reservoirs. Human modifications were not equally distributed along the river courses, but generally increased the further downstream a river sector is located. For example, confined channels historically showed median distances from the source of 27 km. The human interventions along these channels, however, were generally implemented much further downstream (median distance 87 km; compare arrows in Figure 5c). In the case of confined channels, the predominant form of human intervention was weir and impoundment construction. This is even more the case for incised meanders, where the historical median downstream distance was 74 km and that of the intensively modified sections now is 136 km.
Multi-channel sections were also affected by humans, primarily in more downstream locations (historically 54 km vs. currently 67 km).
Here, channel patterns were transformed by reservoir construction as well as by arch-shaped and linear straightening. Along oscillating and meandering rivers, the transformation to one of the three humancreated channel types was more uniformly distributed (Figure 5c). Across the Alps, the current river courses show an irregular distribution of intensively channelized and dammed-up river sections ( Figure 3b). Human interventions took on differing forms and intensities in the various Alpine countries. Thus, the country-wise analysis helps to reveal nationally varying approaches in Alpine river management (Table 3 and Figure 9). Accordingly, in Slovenia, Germany and Austria, large Alpine rivers are subject to particularly intensive energy use. In these countries, 15-18% of the larger river courses were dammed up. This is clearly reflected by the series of reservoirs along several rivers in the Eastern Alps, but also on the Rhone downstream of Lake Geneva. In contrast, in Switzerland, very straight (linear) channelized rivers are most frequent. Today, they amount to 32% of the large Alpine Swiss river continuums. In Austria, all three most intensively human-modified channel forms together total of 53%. Interestingly, the same value of 53% is valid for Switzerland and Germany, although the proportions of linear, arched and dammed stretches differ. France has the greatest proportion of natural and near-natural channel forms. Here, two-thirds of the network length is still confined, oscillating, braided, sinuous or meandering, although most of these have been significantly narrowed by river training or are affected by water abstractions and channel dredging.
Focusing on multi-channel rivers, Austria and Switzerland lost an astounding 90 and 83%, respectively, of former extents (Table 3). Italy and France, in contrast, lost only between 35 and 40% of such sections. Importantly, however, these numbers hide the fact that between 44 and 47% of the formerly braided sections have been lost in the latter two countries (Figure 9). The different numbers point to a partial compensation by the emergence of new, less braided reaches in other river sections. The differences can be also attributed to channel classification problems, that is, regarding historical and current transitional wandering gravel-bed rivers.
Linear straightening of multi-channel rivers was fairly frequent in Switzerland and Austria, whereas arch-shaped regulation was more frequent in Slovenia and Germany (Figure 9). Multi-channel sections were most frequently transformed into reservoirs in Germany, Austria and France. Note that none of the multi-channel sections became meandering, confined or incised meanders. Some, however, did become oscillating (maximum value of 10% in Slovenia and Italy).

| Inaccuracies of the basic data and potential errors
Using so many different data sources entails manifold potential pitfalls. Firstly, the used historical maps originate from various sources and partly different time periods. Secondly, the current datasets, that is, the used DEM, can also show geographical errors. And thirdly, the GIS methods we applied may produce additional inaccuracies that affect the results. Most of these potential sources of error proved to be minimal.
Most of the historical map series were produced by military cartographers working with very elaborate survey instructions (Fuhrmann, 2007;Gugerli & Speich, 2002;Lefort, 2004). The creation of hundreds or even thousands of sheets of maps necessitated numerous actors who-despite detailed instructions-no doubt deployed different levels of accurateness in their field work. The resulting map series are, therefore, not as homogeneous in some sections as originally intended. In some of the Austrian cadastral maps, for example, in-stream structures such as gravel bars are depicted, while they are missing in the adjacent map sheet. This, however, was not a major problem because we focused on the geometry of the active channel rather than on in-stream structures. Another illustrative example is the River Bléone in southern France. Single map sheets of the "Carte de l'État-Major" do not indicate a braiding character, but the up-and downstream, subsequent sheets do so. Instead, a wide water-covered channel is depicted. In such cases, when we were unable to find a logical reason for the deviating channel pattern and that stretch is still braiding today, we also assumed a historically multi-channel pattern. Another question arises whether the very high proportions of meandering and sinuous sections in the Austrian Alps F I G U R E 9 Transformation of channel patterns between the early 19th century (x-axis) and 2017 (y-axis) per Alpine country (based on the number of river sectors) might be due to varying accuracies of the historical maps ( Figure 3a and Table 3). The same map collections, however, were used both for the Austrian and also Italian and Slovenian Southern Alps, where far fewer such channel patterns were recorded. Nevertheless, we cannot exclude that some historical channel patterns are falsely classified. In particular, the classification of transitional wandering gravel-bed rivers as multi-or oscillating channel was not always clear. This problem was also evident for current transitional channel forms.
Another source of inaccuracies is the partly differing time periods covered by the maps. Some sources, such as the two mentioned Swiss maps, stem from the mid-19th or late 19th century, respectively, when several reaches had already been channelized. Nevertheless, The mean correction factors applied to calculate the historical lengths of river sectors based on current channel lengths are also a source of potential inaccuracy (see Section 3). These factors differ for individual channel types and were defined based on the historical maps. Typical sinuosity ranges were also considered (Table 1). In order to validate the applied correction factors, 320 sample sections were selected in regions where geo-referenced maps were already available and GIS vectorizations of the historical river courses were possible.
This approach yielded a factor of 1.06 for multi-channel sections (instead of 1.10 as applied in this study), 1.23 for sinuous sections (compared to 1.20) and 1.62 for meandering ones (compared to 1.65).
Because two-thirds of the samples are located at Austrian rivers and almost all others in Italy and Slovenia, they do not reflect the whole range of Alpine channel forms. Thus, the originally applied mean correction factors were maintained. In addition, the potential error ranges were evaluated using minimum or maximum correction factors for channel sinuosity. Accordingly, the total historical channel length would yield 228 km (−1.9%) smaller or 269 km (+2.3%) greater values than presented in this study. Similarly, minimum correction factors for historical channel lengths would result in slightly steeper channel slopes for multi-channel (+0.02%), sinuous (+0.07%) and meandering river reaches (+0.04%). In contrast, maximum correction factors would yield slightly lower slopes (−0.04% for each of the three channel types). Thus, the potential errors do not significantly affect the proportions of the channel lengths associated with different channel types. The pattern of channel slopes per river type is also very robust.
The GIS analyses revealed that the used 25 m DEM was less accurate in some areas than originally thought. In order to check the precision in elevation, we, therefore, additionally calculated the altitudes and channel slopes for the Austrian river sectors with a more accurate 10 m DEM (almost 4,900 sectors out of 16,590). The newly calculated slopes were significantly lower for all channel types (ranging between −1.00% for confined and −0.17% for multi-channel and meandering reaches; for details see supporting information). Nevertheless, the relative differences of slopes between the individual channel types remained largely stable. The resulting median altitudes were between 3 and 9 m lower for oscillating, multi-channel, sinuous and meandering rivers. The corresponding altitudes for confined channels and incised meanders, which are generally located in narrow valleys, were between 17 and 22 m lower. In light of the Austrian samples, the slope values presented in Section 4 and in Figure 5e generally seem to be too high, and the relative differences between the analysed channel types are probably smaller.

| Physical channel controls and regional organization
Alpine-wide, the data analysis showed that confined, oscillating and all other channel types combined together generally followed an upstream-downstream gradient defined by increasing distance from the source, channel slope and floodplain width. As the valley bottom widened, more space was available for the (temporal) deposition of sediments. In larger, generally further downstream located Alpine valleys, the relationship between up-/downstream gradient and channel patterns was less significant. Here, multi-channel, sinuous or meandering reaches developed prior to human modifications. Our data show that the geographical position (i.e., NE-position) and sediment yield significantly affected which type of channel pattern originally evolved. Including channel slope, the historical channel types can be clearly differentiated within four domains defined by sediment supply and available space ( Figure S5).
A closer examination of the numerous sinuous and meandering rivers in the NE-located Austrian Alps reveals that most were in longitudinal valley furrows ( Figure 2). That specific type of Alpine valley is typically associated with major tectonic faults that fostered glacial excavation during the Ice ages .
Post-Ice age lakes in glacially over-deepened valley floors were gradually filled with sediments. These long and wide valleys, with their low gradients, provided better geomorphic conditions for the development of meandering or sinuous reaches compared to rivers perpendicular to the Alpine divide, which are mostly located in steeper terrain.
Thus, the longitudinal valleys represent "landscape templates" that provide enough space and adequate gradients of the valley bottom for the formation of extensive braiding or anabranching river landscapes. In the Eastern Alps, however, the multivariate analysis reveals that limited sediment supply at least partly prevented the evolution of such fluvial systems. Instead, it significantly favoured the development of sinuous and meandering river sections. Most of these were located in the north-eastern Alps, that is, in the Eastern Central Alps or directly at the boundary to the Northern Limestone Alps (Figure 2).
According to Hinderer et al. (2013), both areas today feature the lowest median denudation rates in the entire Alps (only 80 and 64 mm per 1,000 years, respectively). In this respect, the high proportions of sinuous and meandering channels in Austria's Central Alps can be interpreted as a product of lower sediment supply and adequate tectonic/geomorphic framework conditions.
In Germany and Austria, the Northern Limestone Alps feature the lowest denudation rates and thus, supposedly, the lowest sediment supply. Rivers there, however, showed more braided sections and significantly fewer meandering and sinuous courses ( Figure S3). Because those rivers run towards north, they generally show shorter and partly steeper courses than in the Eastern Central Alps. They also feature shorter downstream distances from the sediment sources.

Long sections of the rivers Rhone and Rhine in Switzerland and
Inn in western Austria also run in longitudinal valley furrows but did not form longer sinuous or meandering reaches. Here, the main differences to those in eastern Austria are the generally higher denudation rates (sediment delivery) of up to 212 mm/1,000 years, the closer location to sediment-rich (formerly) glaciated areas, and higher and steeper terrain (Ehlers, Gibbard, & Hughes, 2011;Hinderer et al., 2013). Chiese River valleys. It showed significantly more oscillating reaches than multiple ones, reflecting a difference in valley morphology (confinement) rather than a difference in sediment supply.
Interestingly, multi-channel and sinuous reaches were very similar in several physical attributes. Sinuous reaches generally showed slightly lower channel slopes (0.75 vs. 0.90%) and narrower floodplain widths. The data also indicate that reduced sediment supply promoted the genesis of that channel pattern also outside of the sedimentlimited NE-Alps ( Figure S5b).
The above-described examples show that the observed distribution of historical channel patterns in the Alps cannot be explained primarily by differing lithological conditions and sediment availability (compare Notebaert & Piégay, 2013). Other controls related to Alpine tectonic structure, such as valley confinement, topography (altitude, steepness), (former) occurrence of glaciated areas or vegetation cover, probably had more influence on the morphological configuration of the Alpine rivers in the early 19th century. In this respect, discontinuum concepts are more meaningful than continuum concepts to explain the regional occurrence of channel patterns (Poole, 2002).
Evaluating all 16,590 river sectors independently for whether subsequent reaches show the same channel pattern or not yields an almost textbook-like longitudinal sequence (compare Figure 1 and At the next lower valley floor, they often returned to the channel pattern they featured upstream of the steep valley (compare Figure 8).
Confluences of large tributaries can be another cause for discontinuities in channel patterns (Bravard, 2010;Brierley & Fryirs, 2005;Liébault, Piégay, Frey, & Landon, 2008). The historical sources reveal that several bedload-abundant tributaries amplified the braiding intensity of the main stem. With increasing downstream distance to the confluence, the deposition processes and, therefore, the braiding intensity were gradually reduced.
The scheme of channel pattern types from the steep headwater to the lowland illustrated in Figure 1 provides a generalized perspective, one that was rarely met in nature. It potentially applies for rivers in valleys that are not affected by major tectonic faults with uplift/ subsidence processes, do not stretch across several lithological zones, and that were largely homogeneously shaped by glaciers. In most cases, such conditions apply only for individual river sections but not for whole courses. In particular, longer Alpine rivers run through "polygenetic valleys"-several valley fragments that were pieced together as a consequence of Alpine orogeny and glacial reshaping . Such "inherited landscapes" commonly feature discontinuities related to transitions between various physiographic zones rather than smoothly declining valley floors (Notebaert & Piégay, 2013). Thus, longitudinal profiles of Alpine valleys typically show several stepwise declining valley floors . The channel pattern in each of these valley floors integrates both the upstream physical controlling factors and the local geomorphic setting.

| Regional and longitudinal differences in human modifications
From a modern viewpoint, it appears appropriate to consider waterbodies before the era of industrialization as "near-natural" or fairly natural in term of riverscape features-but by no means pristine.
The channel geometry and bedload delivery of the rivers of the Alpine  Figure 5c). In contrast, historically oscillating, braiding or sinuous river reaches were primarily human transformed in more downstream sections. This is even more valid for incised meanders: they were largely modified downstream for hydro-energetic use. These findings support the notion of increasing downstream intensities of human pressures on river systems (Gurnell et al., 2009;Muhar et al., 2019;Schinegger, Trautwein, Melcher, & Schmutz, 2012). In downstream reaches, adjoining land can be more intensively used and the local and regional socio-economic demands gain weight Tockner, Pusch, Borchardt, & Lorang, 2010 (Ward & Stanford, 1995). For example, the entire Swiss stretch of the Rhone is seriously affected by hydropower plants, even in its undammed reaches (Costa, 2018).
According to Muhar et al. (2019), Alpine-wide at least 23% of all rivers with catchments larger than 10 km 2 are severely modified today. In comparison, the present study shows that arched regulated, linear straightened and dammed-up sections amount to 45% of the large Alpine rivers. Importantly, this includes only morphological modifications, such as river straightening, damming and artificial channel narrowing. Accordingly, large rivers suffered a much higher intervention intensity than smaller catchments. Considering also human interventions into river flow, 72% of the Alpine rivers with catchments between 500 and 1,000 km 2 , and 86% of those larger than 1,000 km 2 are severely affected today. The present study, however, did not examine hydrological alterations.
The different countries historically followed different paths of land use management and river training. This is reflected in varying forms and intensities of human pressure across the Alpine sphere . Within the six countries examined, the large Alpine rivers in Austria, Germany and Switzerland are most severely affected (see Section 4). The hydraulic engineering practices differed substantially in France and Italy, where the rivers were less intensively channelized. Those two countries, therefore, show the highest proportions of rivers remaining in a more natural state. Nevertheless, most of them are more or less severely affected by artificial channel narrowing, water abstraction, gravel dredging or other interventions contributing to channel incision.
These country-wise differences reflect varying socio-economic pressures and political decisions regarding the use of the Alpine rivers.
Since the late 19th century, great proportions of the rural population in the French and Italian Alps have moved to larger cities (Bätzing, Perlik, & Dekleva, 1996;Comiti, 2012). This might be a reason why France and Italy invested less in local river training. The strategy in the other countries was different: systematic regulation programmes were implemented beginning in the 19th century in order to secure arable land or to reclaim new land in the valley bottoms. This required large-scale corrections and stabilization of the river courses (Baumann, 1960;K. Oberste Baubehörde, 1888).
The major reduction of braided rivers in the French Alps (by 53%; Piégay et al., 2009) is largely reflected by the present study, which yielded a loss of 47%. Taking potential channel classification errors along with newly originated braided reaches at other river sections into account yields a net loss of 40%. One potential explanation for the different values is that our study omitted smaller rivers and differently classified wandering gravel-bed rivers. The GIS analysis revealed a significant narrowing of braided channels at larger rivers in the French and Italian Alps, as also reported by Piégay et al. (2009) and Surian et al. (2009). In some stretches, channel narrowing was caused by hydraulic constructions (embankments or levees). In others, no constraining constructions were visible but can be assumed at least locally. Upstream sediment retention, land cover changes or gravel dredging were contributing or even dominant causes behind that narrowing process (Surian & Cisotto, 2007). According to Surian et al. (2009), channel narrowing and incision drastically altered channel patterns, leading to wandering or single-channel rivers in several originally braided reaches. Channel incision may also have led to selfconstraining within the formerly broader river bed. Based on the channel pattern types distinguished here, this would mean that formerly multi-channel sections have turned into oscillating or sinuous single-channel ones. Our data show approx. 210 km of such transformations in the French and Italian Alps, corresponding to 9% of the formerly large braided river length there. Accordingly, multiple channel decline is linked not only to sectional training of river courses but partly also (up to between 5 and 10% depending on the country) to significant reductions in bedload supply inducing channel planform changes from multi-to single-bed channels.

| CONCLUSION
This study provides a first methodologically coherent census of channel pattern types in the entire Alpine sphere before the Industrial era, that is, the onset of systematic river regulation programmes. Based on the compiled dataset, we draw new conclusions about the large-scale characteristics of Alpine rivers: (1) In the early 19th century, one-third of the 143 largest Alpine rivers (total length of 11,870 km) featured multi-channel (braided) sections. Oscillating rivers, a channel type typically not explicitly addressed in river classification schemes, made up the second highest share (28%). The Eastern Alps featured significantly higher proportions of sinuous and meandering river reaches than the rest of the Alps. These reflect a regionally lower sediment supply and certain tectonically glacially shaped valley forms. This demonstrates that the complex geological and tectonic configuration of the Alpine arc fosters a regionally and country-wise varying pattern of channel forms.
(2) Textbook-like sequences of channel patterns along the river courses ( Figure 1) were rarely met. Channel planforms were significantly influenced by the physiographic conditions in the basins, that is, by their orogenetic and glacial background and the associated valley width. Nevertheless, confined, oscillating and all unconfined alluvial channel types combined generally followed a distinct upstream-downstream gradient (i.e., distance from source, channel slope and valley width). Most parts of the Alps provide abundant sediments but show certain transport-limited river sections characterized by wider valley bottoms and lower channel slopes. Such conditions favoured multi-channel patterns. Sinuous and, in particular, meandering reaches showed lower channel slopes and prevailed in areas with less sediment supply.
The erodibility of the geological basement and thus sediment supply were a basic factor for the evolution of distinct channel forms.
Nonetheless, other factors such as the terrain relief of the catchment, (former) glaciation, land cover or valley morphology no doubt affected channel patterns more significantly.
(3) Human interventions tremendously modified the Alpine river systems. Until today, approximately 510-km-long river sections or 4.3% of their historical extent have been lost due to channel straightening. River types that once occupied large areas of the valley floors, such as braided, sinuous or meandering ones, experienced the strongest reduction. As a consequence, only 15% of the large Alpine rivers still show multiple channels. Today, 45% of the rivers are linear or arch-shaped straightened, or were transformed into reservoirs. Rivers in Austria, Germany, Switzerland, the northern French Alps and Trentino-Alto Adige (Northern Italy) suffered most from human transformation. Moreover, most of the still braided or oscillating waterbodies are, in fact, regulated. River channels have been severely narrowed for land reclamation, water is being abstracted for energy production, irrigation or water supply, and bed material is being dredged. Multi-thread channels also evolved to single-thread channels due to a modification of upstream controls (flood regime and bedload delivery).
The presented data provide a first overview over the past and current state of running waters covering the entire Alpine sphere. Further research will be necessary in order to integrate smaller river systems with catchments between 100 and 500 km 2 according to the available historical sources. Moreover, the analysis can be improved by incorporating additional controlling factors and hydromorphological attributes.