This work benefitted from funds provided by the INRAP, the Lascarbx (program of the Agence Nationale de la Recherche ANR-10-LABX-52), the University of Bordeaux and the UMR-5199 PACEA. C. Kasse and J. Vandenberghe presented to the authors the section illustrated in fig. 4A. We also acknowledge M. Bateman and J. Vandenberghe whose remarks have improved the manuscript.

1 - Introduction

1During the Pleistocene, periglacial landscapes have formed over large surfaces in mid-latitude Western Europe. The Mid to Late Pleistocene climate oscillations caused the land area affected by periglacial conditions to expand and contract repeatedly. Many relict periglacial features bear witness of these events and raised abundant research by geologists and geomorphologists. Aside from the engineering aspects of these discoveries, the scientific community focused on the climatic significance of the features, the reconstruction of past environments, and finally, the possible impact of the periglacial milieus on Palaeolithic populations.

2Reconstructions of the previous extent of permafrost in Western Europe have been proposed for over 60 years (Poser, 1948; Büdel, 1951; Kaiser, 1960; Maarleveld, 1976; Lautridou & Sommé, 1981; Velichko, 1982). Further improvements were later made using increasingly available observations on present-day permafrost (Péwé, 1966; Romanovskij, 1985; Mackay & Burn, 2002; Murton, 2013), and sporadic, discontinuous and continuous Pleistocene permafrost was differentiated in the proposed maps (Vandenberghe & Pissart, 1993; Van Vliet-Lanoë, 1989, 1996; Huijzer & Vandenberghe 1998; Isarin et al., 1998; Van Vliet & Hallégouët, 2001; Renssen and Vandenberghe, 2003; Vandenberghe et al., 2014). Despite the lack of agreement between these reconstructions, these studies agree that permafrost extended over large parts of France during the Pleistocene.

3Most of the datasets available have been published in French language with limited circulation, and unfortunately remains largely inaccessible to the international community. During the last decade, the launch of online databases of aerial photographs (Google earth, Géo-portail) has eased the recognition of periglacial features, adding a lot of new data. The huge development of rescue archaeology also increased significantly the number of identified features.

4This paper serves as an introduction to an online database aiming at gathering as exhaustively as possible the features related to past permafrost in France, to give easier access to a homogenised and reliable dataset. It also reports on the main limits that should be considered when using the map or the data to make interpretations. The establishment of such a database is of the outmost importance to progress towards a better general understanding of periglacial environments.

5The database was first introduced by Bertran et al. (2014), who focused on the geographic distribution of the georeferenced features and reviewed the available chronological data. Periglacial features and their characteristics were entered into a GIS to elucidate the different factors that influenced the development of periglacial features and to propose a new map of the main Pleistocene permafrost boundaries in Western Europe (Andrieux et al., 2015). We propose here an improved online version of the database illustrating and explaining identification criteria of the periglacial features gathered.

2 - Data collection, database structure

6The selected items comprise features observed from aerial photographs and classified in (i) polygonal networks and (ii) soil stripes, and features observed in cross-section and classified in (iii) sand wedges, (iv) ice-wedge pseudomorphs, and (v) composite wedge pseudomorphs. Involutions, patterned grounds other than soil stripes, pingo and lithalsa scars, which are also considered potentially testifying to past permafrost, have not been integrated into the database, since re-evaluation of the features mentioned in the literature is still under progress. They will be added to the database at the end of this process. Similarly, mountain permafrost features were not included in the database.

7Multi-user tests were performed on the photographs to be sure that only indisputable features were added to the database. When different features were present on the same site, it appears in the file of the most represented type. Search for periglacial features in aerial photographs have been made primarily in areas where some had been previously reported in the literature. Search was then extended to neighbouring regions. In this process, it became rapidly obvious that many features were preferentially associated with specific substrates, for example old (Lower Pleistocene and Tertiary) alluvial sand and gravel deposits for polygons, chalk with a thin loess cover for soil stripes… Blind tests in other regions were also made but were most of the time unsuccessful.

8We also conducted a thorough search of the published literature on periglacial features in France, including articles in journals, PhD theses, other dissertations (often unpublished) and geology reports of rescue archaeology. These data were rigorously selected and all entries not accompanied by drawings, photographs or good descriptions were rejected from the database. We were forced to exclude a significant amount of features because the sources did not document them enough (lack of description or figure) or did not give satisfying geographic information (e.g. coordinates, grid references).

9The investigated literature covers the last 60 years, and during such a time span the overall understanding and classification of the periglacial processes changed significantly. In the 50’s, the research community lead by A. Cailleux and J. Tricart (Bastin & Cailleux, 1941; Cailleux, 1948, 1956; Tricart, 1963; Tricart & Cailleux, 1967) described abundant features assumed to be of periglacial origin in France, proving that Pleistocene frost-induced processes played a major role in shaping the landscape. Many were also mentioned in the explanatory notes accompanying the sheets of the geological map of France. “Fente de gel” (i.e. frost crack) and “fente en coin” (i.e. wedge) were used as general terms for wedges and cracks regardless their filling and width. As a consequence, with few exceptions (e.g. Michel, 1975; Yvard, 1968; Nury & Roux, 1969; Arnal, 1971), the “frost crack” data remains hardly usable within the frame of the current database.

10A total of 615 sites have been identified. In aerial photographs, a site corresponds to a single land parcel or a few adjacent parcels.

11The database consists of 5 CSV (Comma-Separated Values) files each dedicated to a single type of feature. The CSV format was chosen because of its simple use and the possibility of easy transformation into an attribute table in a GIS. Each CSV file is accompanied with a folder that contains photographs and sketches of the features when available. The ID number allows linking the listed features and corresponding pictures, drawings, or aerial photographs.

12The basic fields documented are as follows:

13– ID, the identification number of the feature;

14– Longitude and Latitude in decimal degrees in the EPSG:4326 (WGS84) coordinate reference system;

15– Altitude, derived from the DEM Aster GDEM 30 m;

16– Site, the name of the place;

17– City, the municipality;

18– Region, the administrative region;

19– Geol_code, the code of the geological substrate as shown in the 1:50,000 geological map of France (infoterre.brgm.fr);

20– Substrat, the (simplified) lithology of the geological substrate as indicated in the explanatory note of the 1:50,000 geological map. “/” stands for “above”;

21– Reference, the bibliographic references when available;

22– Photo_credit, the author of the photograph(s) or drawing(s) when available.

23Additional fields are also provided for each type of features. They include:

24– Size, the average size of the polygons calculated with the software ImageJ (see Andrieux et al., 2015, for detailed explanation of the measurement protocol);

25– Spacing, the average spacing of the soil stripes calculated with the software ImageJ;

26– Slope, the average slope gradient (in degrees) of the soil stripes derived from the DEM Aster GDEM 30 m;

27– Expo, the orientation of the soil stripes (in degrees) measured on Google earth with the “ruler” tool;

28– Host, the lithology of the host material of the wedges determined from the description of the cross-section (fig. 1);

29– Filling_1, the lithology of the main filling of the wedge;

30– Filling_2, in case of different types of filling;

31– Depth and Width, the dimensions of the wedges (Depth = null if the apex of the wedge is not visible).

32Most of the features remain non dated at the moment. The chronological data have not been included in the database, they are accessible through the cited references.

3 - Features description

3.1 - Polygons

33When frozen ground cools down rapidly, it contracts and cracks if the tensile stress is not relieved by creep (Lachenbruch, 1962, 1966). Thermal contraction cracking occurs widely in areas subjected to deep freezing where snow and vegetation covers are thin (Mackay, 1993; Mackay & Burn, 2002; Fortier & Allard, 2005), with or without permafrost (Washburn et al., 1963; Friedman et al., 1971; Romanovskij, 1985). Thermal contractions cracking forms large polygonal networks of fissures, evolving into V shaped wedges as a consequence of repeated filling by ice or sediment. The 284 fossil polygon sites listed in the database and found in aerial photographs are between 8 and 20 m in diameter (mean = 15 m, Andrieux et al., 2015). Most of the polygons observed in the lower Loire and Rhône valleys have sharp edges and thin walls, which appear darker than the surrounding land in aerial photographs (fig. 2A, B). Cross-sections discovered in the vicinity of a few sites suggest that the walls correspond to sand wedges or composite wedges. In the Paris basin and the north of France, some polygons have more rounded outlines and walls of irregular width (fig. 2C, D). They are thought to correspond to former ice-wedge polygons that have undergone thermokarstic degradation. Polygons in vineyards such as those in the Médoc region, southwest France, are often hardly visible in aerial photographs. Although not very conclusive by themselves, some photographs have been kept in the database because of the discovery of sand wedges in nearby cross-sections (Lenoble et al., 2012; Bertran et al., 2014). Aerial photographs also revealed fields of irregular and rounded “polygonal” features showing more heterogeneous sizes than typical ice or sand wedge polygons (fig. 3). Although a periglacial origin seems probable, their classification as thermal contraction features rather than very large patterned grounds remains doubtful (see aerial photographs and cross-sections in Agache, 1963, 1970) and needs further confirmation by field survey. As a consequence, this kind of features was not added to the database.

34Small-scale polygons up to 1 m in diameter have been widely reported during field surveys but were not included in the database. They are assumed to reflect mostly (cryo-) desiccation fissures (Washburn, 1979; Van Vliet-Lanoë & Langohr, 1981).

3.2 - Ice-wedge pseudomorphs

35Ice-wedge pseudomorphs, i.e. the features left by the melt of an ice-wedge, have been widely reported in France. Severe selection was made among the reported features, 90 ice wedge pseudomorphs are listed in the database. According to Black (1976), Washburn (1979), Vandenberghe (1983), Gozdzik (1994) and Murton (2013), the main criteria for their identification and consideration in the database are: (1) a V shape showing collapse structures due to the replacement of ice; these consist in block faulting (common in the coversands of northern Europe, Murton, 2013; no example has been found in France so far, fig. 4C), downturned strata in the host material or folding of the strata toward the wedge, vertically aligned clasts along the walls of the pseudomorph (mainly in gravel); (2) a minimum size of 0.2 m in width and 1.5 m in height, (3) deformation of the host material (upturned strata, vertical gravel pebbles) due to wedge growth, (4) evidence for a polygonal pattern from cross-sections or aerial photographs. The filling may be massive or show steeply dipping or U-shaped layers. Blocks of sediment may be present within the filling (fig. 4B). In loess deposits ice-wedge pseudomorphs are usually associated with a cryoturbated bleached layer with small ferruginous spots, interpreted as a former gley soil over permafrost (Gullentops, 1954; Haesaerts & Van Vliet-Lanoë, 1973; Antoine et al., 2009, fig. 4A). Thermokarstic degradation of permafrost may lead to the erosion and deformation of the wedge, the formation of tunnels, ponds and gullies so that the relict ice-wedges are no longer identifiable or visible. The distribution of ice wedge pseudomorphs is therefore probably significantly underestimated in our dataset.

36Not included in the database are many wedge structures described in France in the literature. They are narrow and lack any clear indication of host material collapsed in the wedge after the melt of an ice body (fig. 5). These features may represent either incipient ice wedges (ice veins) or soil wedges in seasonally frozen ground (Friedman et al., 1971; Romanovskij, 1973; Murton, 2013).

3.3 - Sand-wedges

37The primary filling of thermal contraction cracks by wind-blown sand forms sand wedges (Péwé, 1959; Black, 1976; Washburn, 1979; Kolstrup, 1986; Gozdzik, 1986; Murton, 1996; Bockheim et al., 2009). Sand wedges do not show evidence of host material slumped in the wedge. They are usually V-shaped but may be irregular with multiple elementary sand veins extending out of the toe of the wedge into the host material (this typically occurs in sands, Romanovskij, 1976; Murton et al., 2000) (fig. 6A), or have a “bulbous” shape (fig. 6B). The origin of the bulbous shape, illustrated by Murton et al. (2000) in Poland, remains uncertain. In our dataset, they were observed only in fine-grained substrates and may reflect deformation during the thawing of an ice-rich host sediment or deformation of the seasonally frozen layer due to frost heave (Jetchick & Allard, 1990; Van Vliet-Lanoë, 2005). Deformation of the host material induced by wedge growth is usually poorly developed and upward bending of strata occurs only occasionally. Field exposures generally show multiple sand wedges a few metres apart, which testifies to a polygonal pattern. The sand wedges identified in France have generally a massive filling (fig. 6C) but show sometimes vertical laminations (fig. 6D). According to Péwé (1959) and Murton & Bateman (2007), the lack of vertical lamination does not necessarily imply a secondary perturbation caused by the thaw of ice veins. Small eolised gravels may occasionally be present in the upper part of the filling and have probably felt into the fissures during the opening of the wedges (Péwé, 1959) (fig. 6E).

38In the database, 82 sand wedges are listed. Only the wedges with a minimum width of 0.2 m and 0.5 m in height were added to the database to be sure to not consider possible desiccation or extension cracks filled by sand.

3.4 - Composite-wedge pseudomorphs

39Composite wedges are thermal contraction cracks filled both by wind-blown sand and ice (Kolstrup, 1986; Murton, 1996; Murton et al., 2000, 2007; Antoine et al., 2005). The composite wedges share therefore common features with ice-wedge pseudomorphs and (primary) sand wedges. The most diagnostic feature is the evidence of a secondary filling occupying a significant part of the wedge. This filling appears typically either as cross-stratified to U-shaped beds of sand or gravel material (fig. 7A, B) or as a mixture of well-sorted (aeolian) sand with scattered gravel in the whole mass. Only wedges with a significant part of the filling showing U-shaped stratification were considered as composite wedge pseudomorphs since simple sand wedges may exhibit such a stratification in the upper decimetres as shown by Péwé (1959). In many cases, however, the wedges show different types of filling cross-cutting each other, suggesting a succession of dominantly icy or sandy phases (fig. 7C). One example (fig. 7D) corresponds to a gully overlying a truncated sand wedge. The gully is thought to be of thermokarstic origin and, therefore, suggests that the wedge filling contained a significant amount of ice. Overall, 10 composite-wedge pseudomorphs are listed in the database. However, this number may be underestimated because of difficulties in identifying secondary sandy fillings. Potential composite wedge pseudomorphs may, therefore, be classified here as sand-wedges.

3.5 - Stripes

40Stripes appearing in aerial photographs of ploughed fields are regularly spaced stripes of different colours, some of which are branched. They reflect alternating bands of light coloured (usually chalky) debris and darker loess or clay loam material or uneven plant growth due to contrasts in water content (fig. 8A, B). A small number of aerial photographs show spotted patterned ground on flat surfaces stretching into stripes on slopes (fig. 8D, C). A total of 149 sites of relict soil stripes have been identified in France, mostly in the Paris basin. The spacing between the stripes range from 2 to 13 m, but the distribution can be adjusted to two Gaussian functions peaking at 4.6 and 7.9 m, which seem to be mostly related to the nature of the substrate (Andrieux et al., 2015). The largest stripes are found north of 49°N. Similar features were described previously in the UK amongst others by Williams (1964, 1968) and Ballantyne & Harris (1994). In modern periglacial environments, soil stripes form in the ground layer subjected to freezing and thawing cycles. They are found in seasonally frozen grounds, but are most developed over permafrost (Washburn, 1979; Ballantyne & Harris, 1994).

4 - Conclusions

41The aim of this database is to be integrated in broader studies and to be progressively completed with the implementation of new field evidence of periglacial features. The data presented here should not be viewed as fully exhaustive; because of limited time and funds all the potential areas where periglacial features could be found in cross-sections have not been prospected yet. This is particularly the case for large areas of northeast France. Updating the database will make it possible to reconstruct more reliably past permafrost and to shed new light on some questions, such as the controlling factors of the gaps in deep aquifer recharge during the Late Pleistocene (Jiraková et al., 2011), the gaps in speleothem growth (Genty et al., 2010), and fluctuations of the peopling of vast regions of northern Europe during the Palaeolithic (Miller, 2012; Tallavaara et al., 2015). Conflicts and discrepancies may remain as shown by Andrieux et al. (2015) and few data exist with regard of the age of the features. To solve these issues further research on the origin and palaeoclimatic significance of some features and more thorough dating is needed. The online database can be downloaded at https://afeqeng.hypotheses.org/​487 and used provided that this paper is cited in the references. The file of photographs and drawings is available on request from the first two authors (E. A. and P. B.).