Characterization of the evolutionary trajectories of the world's rivers and floodplains is crucial for assessment of space–time changes in biogeomorphic processes, the quantification of anthropogenic impacts, and the planning of sustainable management based on conservation and restoration options (Grabowski et al., 2014; Mould and Fryirs, 2018). The analysis of long-term (> 150 year) channel changes provides a holistic understanding of the trajectory of fluvial hydrosystems and their drivers, which helps in defining process-based restoration objectives, rather than determining static reference states (Dufour and Piégay, 2009; Eschbach et al., 2018). It is necessary to extend the timescale of past trajectories to characterize possible shifts in baseline conditions related to climate variability and different degrees of human pressures to help in the assessment of future river changes in the Anthropocene (Pont et al., 2015; Słowik, 2015; Piégay et al., 2020).

The reconstruction of river trajectories requires substantial historical datasets. Series of maps, aerial images, topographic, and bathymetric surveys are generally selected at regular time steps, if available, or before and after major events (floods, river engineering phases) to assess changes in the river hydromorphological conditions. This set of archival materials is used to extract channel features such as active channel width, bar-island areas (Hohensinner et al., 2004, 2021; Michalková et al., 2011; Scorpio et al., 2015; Lestel et al., 2018), bed degradation and floodplain aggradation rates (Steiger et al., 2001; Downs et al., 2013; Arnaud et al., 2015), riparian vegetation patterns (Cadol et al., 2011; Belletti et al., 2013; Kui et al., 2017; Safran et al., 2017), and landscape unit characteristics (Dufour et al., 2015; Solins et al., 2018; Piégay et al., 2020). Following Roux et al. (1989) and Habersack and Piégay (2008), comparative analyses of river changes, notably in Europe, have become a critical issue for addressing a series of questions of interest to geographers, historians, engineers, ecologists, geologists, and geochemists. For example, historical maps can help to better contextualize sediment core samples and geophysical surveys when investigating floodplain evolution (Salvador and Berger, 2014) and delta evolution during the Holocene (Vella et al., 2005). Historical cartographic and topo-bathymetric datasets are also of interest to hydrologists analysing flood frequency based on historical geometries (Billy et al., 2018). They are also key resources in historical ecology, allowing the formulation of hypothesis on fish habitat–species relationships since the industrial revolution, especially given the lack of old quantitative data in terms of species number, abundance, and distributions (Carrel, 2002; Frimpong et al., 2016; Belliard et al., 2018).

A number of historical analyses have been conducted on the Rhône River in Western Europe. Whereas the Rhône has been a braided and anabranching river in multiples reaches along its 530 km-long French course since the 17th century, it has been deeply regulated for 150 years and is now targeted by large-scale restoration measures aimed at improving morpho-ecological processes (Riquier and Cottet, 2019). The mid-19th century period is usually considered the benchmark for fluvial patterns before extensive river training. A detailed map series was plotted by the Ponts et Chaussées (Bridges and Roads) public administration in the 1860s at a scale of 1 : 10 000. It provides data on (i) the morphological pattern of the Rhône and its former branches, (ii) the type of land cover on the floodplain, and (iii) topo-bathymetric information (Bravard, 2010). Most diachronic studies have used this 1860s map series as a starting point for detecting geomorphic changes by making comparisons with later maps surveyed at the end of the 19th century and mid-20th century and aerial images taken since the 1930s (Bravard et al., 1986; Bravard, 2010; Provansal et al., 2014; Parrot, 2015; Dépret et al., 2017; Räpple, 2018; Thorel et al., 2018; Tena et al., 2020). Therefore, the post-1860 planform adjustments are fairly well understood and are combined with vertical adjustments measured on reaches where topo-bathymetric information has been available since the late 19th century. However, little is known about the morphodynamics of the Rhône in the early 19th century and times before this.

This paper presents a historical database of 350 cartographic and topo-bathymetric resources covering the French Rhône River, collected from 14 national, regional, and departmental archive services. The timescale extends from the 17th to the mid-20th century, with a pre-1860 focus for cartographic data and a pre-1900 focus for topo-bathymetric data. The aim of this paper is to provide substantial new material to extend and detail historical trajectories of the Rhône River for a wide range of research applications including historical hydrology and historical biogeomorphology, within flood management and river rehabilitation contexts. The compiled database will enable further comparisons with other large braided and/or anabranching multi-impacted rivers worldwide.

The Rhône is an 812 km long river connecting the French–Swiss Alps with the Mediterranean Sea. Originating at the Furka Glacier, the alpine Rhône drains the Valais region up to Lake Geneva. The lower 534 km of the river course are in France. As a whole, the catchment size is 98 500 km2, with 90 500 km km2 being in France (Olivier et al., 2021) (Fig. 1). The river delimits or crosses 11 French administrative departments. It drains areas with a wide range of geological, altitudinal, hydroclimatic, and land-use conditions, which results in a heterogeneous channel reach characterized by alternating fixed and mobile channel beds, incised valleys, and wide floodplains, and distinct sediment and flow inputs from tributaries (Bravard, 2010; Riquier and Cottet, 2019). The mean bed slopes are 0.9 % on the Swiss river course, 0.1 % on the upper Rhône section (Lake Geneva to Lyon-Saône confluence), 0.05 % on the middle Rhône section (Lyon to Valence-Isère confluence), 0.06 % on the lower Rhône section (Valence to Arles-Gard confluence), and 0.004 % at the delta (Parrot, 2015).

The mean annual discharge is 1720 m3 s-1 at the most downstream gauging station (Beaucaire), which makes the Rhône the most important freshwater input into the northwestern Mediterranean Sea (Van Den Broeck and Moutin, 2002). The 100-year flood is 10 300 m3 s-1. The highest floods since the beginning of daily hydrological measurements, from 1816 at Beaucaire, have occurred in October 1840 (13 000 m3 s-1) and May 1856 (12 500 m3 s-1). These floods are considered as a “reference” for the contemporary period. The mid-19th century corresponds to the end of the Little Ice Age, a period during which flood activity was high: the 1701–1710 period has especially exhibited the highest flood frequency of the Rhône history since the 14th century (22 months per decade at Arles against 13 months per decade for the 1851–1860 period; Pichard, 1995; Arnaud-Fassetta, 2003). The floods of November 1694 and December 1755 are the highest of their centuries (Pichard and Roucaute, 2014). The flow regime of the Swiss alpine Rhône is snowmelt-dominated, with low flows from November to April and high flows in late spring and summer. The French upper Rhône and the middle Rhône sections are influenced by a wide diversity of tributaries showing several flow patterns like ice and snowmelt (Isère, Arve rivers) and rainfall (Saône, Ain rivers). On the lower Rhône, the major tributaries (Drôme, Ardèche, and Durance rivers) have a Mediterranean regime with high discharges mostly during spring and autumn and frequent flash floods (Bravard, 2010). River training was conducted along both the Swiss and French parts of the Rhône. In France, human interventions were developed during three main periods (Thorel et al., 2018; Tena et al., 2020): (1) in the second half of the 19th century, with the construction of longitudinal “unsubmersible” dykes to locally protect the population from flooding; (2) from the late 19th to the early 20th century with the systematic construction of groyne fields connected to submersible dykes (infrastructures named “Casiers Girardon” from the engineer who designed them), created almost continuously from Lyon to Arles, and aimed at concentrating flows into a narrow single-bed channel to improve navigation; and (3) from the 1920s to the 1980s, with the construction of 19 hydroelectric dams and bypassing schemes between Lake Geneva and the Mediterranean Sea by the National Company of the Rhône River (CNR).

Channel correction and flow regulation induced major hydromorphological modifications such as incision and armouring, disconnection of secondary channels, floodplain terrestrialization, and subsequent loss of ecological functionalities (Roux et al., 1989; Amoros and Bornette, 2002; Parrot, 2015; Dépret et al., 2017; Tena et al., 2020). A large-scale restoration plan for the French Rhône has been conducted since 1998 to improve ecological conditions in the main channel and floodplain. This plan is based on the combination of two main measures: (i) increasing the minimum flow in bypassed reaches and (ii) reconnecting abandoned channels (Lamouroux et al., 2015). A second phase of restoration was launched in 2009 and currently consists of enhancing sediment transport processes into bypassed and main channel sections. The measures include riprap removal to reinitiate bank erosion and reintroduction of gravels dredged from reservoirs and navigable channels or excavated from restored side channels. Monitoring programmes are co-constructed between researchers and practitioners to assess restoration success and provide insights to improve restoration strategies (Lamouroux et al., 2015; Thorel et al., 2018).

The type of archive data, either manually photographed or scanned, and in some cases georeferenced, allows the user a qualitative or quantitative analysis, including illustration of the fluvial landscape of a large river at the beginning of the 19th century, evaluation of the main transformations following river regulation, and diachronic GIS extraction of metrics such as channel widths, secondary channel numbers, and gravel bar areas. The scanning and georeferencing of the sources are not exhaustive. The archive references are accurately described in the table of data properties to facilitate easy consultation and scanning of data of interest.

The dataset depicts heterogeneous information, which depends on the survey date and map objectives. A critical analysis should therefore be carried out before any use of the dataset on a given river reach. As an example, Fig. 10 compiles maps surveyed in the upstream section of the Donzère-Mondragon (DZM) reach, which is currently involved in side channel reconnections, riprap removal, and gravel augmentation. The 5 km long section (KP 171 to 176) embraces the eastern Bayard secondary channel, which was the main course of the river in the late 18th century. Among the 26 maps available prior to 1860, nine have been selected at scales of 1 : 5000–1 : 10 000. Maps from 1810 to 1842 do not specify the nature of islands. Partial information is given on the 1843 map, with gravel bars drawn in light orange. Symbols are then used on the 1846 map to differentiate gravel bars and vegetated areas. These symbols are also used on the 1852 and 1856 maps; their base map is the 1846 edition superimposed with the new channel pattern and engineering infrastructure, which is a common practice in engineering cartography (Lestel et al., 2018). The 1856 map was edited to report damage to “unsubmersible” dykes after the flood of May 1856. New islands and gravel bar areas formed after the flood were drawn in light orange. Note that the titles of the maps are equivocal: 71 maps of the database have the word “dyke” in their title and were surveyed to report damages after large floods, rebuild dykes, or design new ones. The challenge of map-based historical studies is to use individual sources over time, with sometimes little information on the context in which the map was created and therefore on its certainty level. The comparison of multiple maps available at close dates (some of them surveyed in the same year by different engineers) allows inconsistencies between individual sources to be resolved while revealing persistent landscape features and patterns (Safran et al., 2017). For example, the coloured map of 1842 from Switzerland to the Mediterranean Sea surveyed by the engineer Arnaud is fairly aesthetic; however, a comparison with other maps highlights the lack of accuracy in the depiction of geomorphic features, notably on the downstream part of the Bayard channel, as well as the non-exhaustive mapping of roads and villages, which limits the number of ground control points available for georeferencing. In addition, when analysing long-term vegetation changes, it is important to carefully examine vegetated area representation (when available) because vegetation could be represented in many different ways, either by colours, symbols, or annotations (Figs. 10 and 11). In most cases, only vegetated islands and the riparian vegetation close to the channel were mapped (Fig. 11a and b). The depiction of vegetation in the entire floodplain was less frequent (Fig. 11c), especially regarding vegetation types (grassland, shrubs, forest, agricultural areas; Fig 11d). Note that the more recent maps did not have systematically more information than the older ones.

Keeping in mind the preliminary examination of the data accuracy, the large number of historical data items collected in this study should enable deeper analyses of channel shifting, island dynamics, and landscape changes over a longer time period and at a finer time step than those performed in previous studies and also on many reaches of the Rhône River and tributary confluences for which no historical maps had been collected until now. The archives prior to 1860 provide us with historical benchmarks for process-based river restoration and permit comparisons with other rivers with available archives. A potential application could be the detailing of past trajectories of floodplain units, such as that performed by Tena et al. (2020) on three bypassed reaches of the middle and lower Rhône (Fig. 12). Similarly, Dépret et al. (2017) combined historical maps and aerial images (1860–2009) to date a series of channel disconnections and study terrestrialization processes. They showed that terrestrialization of naturally cut-off channels was much quicker than that observed on artificially cut-off channels, probably because of much higher bedload transport, which promoted active sediment infilling. The historical dataset is also valuable to explore longer-term ecological succession, such as initiated by Bravard et al. (1986) and Roux et al. (1989) on the upper Rhône section. During the 18th and 19th centuries, the backswamps and the gravel bars of the braided belt were devoted to extensive grazing, and the woody islands were cut every five years for supplying the riverine populations and the town of Lyon with fuel. On the margins of the floodplain, a few hardwood forests of Fraxinus excelsior, Quercus robur, and Acer pseudoplatanus escaped land clearances for pasture, subsistence agriculture, and firewood cutting (Roux et al., 1989). More recently, Janssen et al. (2020) showed from aerial images that the critical changes undergone by the riparian forests of the Rhône River under persistent multiple stressors, strongly diverged from near-natural systems during the 20th century.

Historical maps and associated physical data have been used to assess fish-faunistic reference conditions in the framework of European biomonitoring improvement (Wolter et al., 2005; Ramos-Merchante et al., 2021). An historical analysis (from the end of the 19th century) of fish community structure in the Seine River catchment highlighted the major role of river continuity alteration in the decline of migratory species, the importance of introductions of non-native species, and changes in the proportion of rheophilic and eurytopic species in relation to the increase of human population and water pollution (Belliard et al., 2018). Historical data on fish communities along the Rhône River were discussed by Carrel (2002). In addition to fish data, Léger's maps available in the Arnaud et al. (2020a) dataset also provide morphological information on each river section (mean channel width, mean water depth). Combining physical data describing habitat heterogeneity (channel width, river bank length, diversity, and accessibility of floodplain channels; see Hohensinner et al. 2004, 2011, for examples on the Danube River) at different periods with data on fish habitat preferences (Lamouroux et al., 1999; Rifflart et al., 2009) may be useful to achieve a better understanding of the influence of fluvial landscape transformation over the last few centuries on fish abundance and spatial patterns. More generally, the combination of biotic and abiotic archival data is important to assess how changes in hydromorphological river conditions influenced biodiversity.

Finally, another potential use of the dataset is the reconstruction of extreme flood events based on historical channel geometries. The Beaucaire-Tarascon station (KP 269.5) records flows of the whole basin and is the oldest hydrometric station on the Rhône. Daily water level measurements started in 1816 and daily estimates of discharge started in 1920 (Billy et al., 2018). These chronicles are part of the HISTRHONE database, which contains a large volume of archives related to hydroclimatic events in the lower Rhône over a period between the 14th and 20th century (Pichard and Roucaute, 2014). Further work by hydrologists will involve exploiting historical topographic data prior to river training to reduce uncertainties in the extrapolation of rating curves of hydraulic modelling based on original channel geometries. The topo-bathymetric data highlighted in this study will be of particular interest (Figs. 6 and 8).

The dataset is freely available in Pangaea at 10.1594/PANGAEA.922437 (Arnaud et al., 2020a). It contains the iconographic database (12.7 GB), the .csv table describing the properties of the 350 data items (109 KB), the .csv table listing the 281 archive box references (13 KB), and the GIS layer of the longitudinal coverage of the 350 data items (10.8 MB). The dataset is available under the terms of the Creative Commons Attribution-NonCommercial (CC BY-NC) license, which permits use, distribution, and reproduction in any medium, provided that the use is not for commercial purposes and that the original work is properly cited. This license was chosen to meet the data reuse policies of the 14 French archive services investigated in this study, four of which have a reuse policy equivalent to the CC BY license (National Archives, AD Ain, AD Vaucluse, DREAL, Univ. Grenoble) and five of which have a reuse policy equivalent to the CC BY-NC license (BnF, AD Rhône, AD Drôme, AD Ardèche, AD Gard). Ten items among the 350 items in the dataset were collected from AD Isère, AD Savoie, CNR, and IGN and they are not furnished in the iconographic database. They are indicated as “available on request” because of specific reuse policies of these institutions. The URLs of each institution's reuse policy are given in Pangaea metadata.

Extracts of the 350 data items are available in the public photo gallery of the Rhône Valley Human-Environment Observatory (http://photo.driihm.fr/index.php/category/52, last access: 2 May 2021) to enhance data visibility. This photo gallery uses the open source Piwigo software, with images being geolocated on an interactive map. Map extracts have been oriented according to the flow direction of the Rhône, as long as this did not disturb the toponym reading.

In addition to Pangaea, technical information on the dataset is available in the public metadata catalogue of the Rhône Valley Human-Environment Observatory (http://elvis.ens-lyon.fr/geonetwork/srv/eng/catalog.search#/metadata/6b60389a-92da-4353-b4d9-9280d0638f5f, last access: 2 May 2021). Based on the open source GeoNetwork software, this metadata catalogue is harvested by the LabEx DRIIHM (international consortium of Human-Environment Observatories) geocatalogue (http://ccwbvps01.in2p3.fr:8080/geonetwork/srv/eng/catalog.search#/search?from=1&to=20&sortBy=changeDate&fast=index&_content_type=json, last access: 2 May 2021), which is then harvested by the French geocatalogue (http://www.geocatalogue.fr/, last access: 2 May 2021) and the European INSPIRE Geoportal (https://inspire-geoportal.ec.europa.eu/, last access: 2 May 2021), so that data discovery is possible at several hierarchical levels (Arnaud et al., 2020b).

This study presents a historical database of 350 maps, topo-bathymetric cross sections, and long profiles compiled on the French Rhône River over the 1640–1950 period. Such an inventory of historical data over such a long time period and long river course (530 km) has never been created for the Rhône and has rarely been created for other large rivers. This database will enable further comparisons with other multi-impacted large rivers in Europe, such as the Danube, the Rhine, the Ebro, and the Po, or even from one continent to another when data are available. The densification of archived time series is useful to detail and extend past trajectories of geomorphic, hydraulic, and ecological river features and for conducting comparative analyses to detect differences and similarities between reaches and between rivers located in different geographical settings. Recently, rehabilitation measures have been extended to other bypassed reaches of the lower Rhône River (Baix-le-Logis-Neuf, Saint-Vallier, Montélimar) and these will also require analyses of channel changes over the last two centuries to design restoration objectives. Some archive services were consulted in entirety for the time period and theme of this data article, notably the National Archives and AD Rhône, whereas additional investigations could be conducted in CNR or other departmental archives in coming years. Possible updates or extensions to the dataset might therefore happen in the future.