Since the end of the Little Ice Age around 1850, the
total glacier area of the central European Alps has considerably decreased.
In order to understand the changes in glacier coverage at various scales and
to model past and future streamflow accurately, long-term and large-scale
datasets of glacier outlines are needed. To fill the gap between the
morphologically reconstructed glacier outlines from the moraine extent
corresponding to the time period around 1850 and the first complete dataset
of glacier areas in the Swiss Alps from aerial photographs in 1973, glacier
areas from 80 sheets of a historical topographic map (the Siegfried map) were
manually digitized for the publication years 1878–1918 (further called first
period, with most sheets being published around 1900) and 1917–1944 (further
called second period, with most sheets being published around 1935). The
accuracy of the digitized glacier areas was then assessed through a two-step
validation process: the data were (1) visually and (2) quantitatively
compared to glacier area datasets of the years 1850, 1973, 2003, and 2010,
which were derived from different sources, at the large scale, basin
scale, and locally. The validation
showed that at least 70 % of the digitized glaciers were comparable to the
outlines from the other datasets and were therefore plausible. Furthermore,
the inaccuracy of the manual digitization was found to be less than 5 %.
The presented datasets of glacier outlines for the first and second periods
are a valuable source of information for long-term glacier mass balance or
hydrological modelling in glacierized basins. The uncertainty of the
historical topographic maps should be considered during the interpretation of
the results. The datasets can be downloaded from the FreiDok plus data
repository (
The total glacier area of the central European Alps has considerably decreased during the last decades with differences of change in certain sub-periods (e.g. Fischer et al., 2014). Long-term glacier datasets are of great importance for understanding and assessing glacier changes (Fischer et al., 2015; Huss and Fischer, 2016) as well as for hydrological modelling of past and future streamflow (Huss, 2011; Stahl et al., 2016; Viviroli et al., 2011). Some glaciers of the central European Alps have been regularly monitored since nearly the end of the Little Ice Age (ca. 1850), but the majority were only recently or sporadically monitored and long time series of glacier data are rarely available (GLAMOS, 2015; WGMS, 2015). Remote sensing offers unique opportunities to derive glacier outlines, areas, and glacier mass balance at the large scale. Several manual and (semi-)automated algorithms have been developed in recent decades to identify from remotely sensed data the entire glacier area of the central European Alps, leading to several glacier inventories starting from 1973 (e.g. Maisch et al., 2000; Paul et al., 2011; Fischer et al., 2014; Kääb et al., 2002). Assuming that the end of the Little Ice Age represents the largest glacier extent (Collins, 2008; Ivy-Ochs et al., 2009; Vincent et al., 2005), the outlines of the moraines correspond to the glacier cover from this recent maximum glacier extension around 1850. Mapping the moraines based on historical topographic maps, field observations, and aerial photographs from the years 1973, 1988, and 1989 therefore made it possible to also create a glacier inventory for the whole of Switzerland around 1850 (Maisch, 1992; Maisch et al., 2000, 2004; Müller et al., 1976). Between 1850 and 1973 no information on glacier area can be obtained from satellite images analysis and aerial photographs are only available locally. Nevertheless, other sources exist in Switzerland, such as historical topographic maps, where glacier areas have been surveyed and drawn manually.
The first topographic surveys started in 1809 in Switzerland, leading to the publication of the first topographic map for the whole of Switzerland (the Dufour map) based on geometric measurements at a scale of 1 : 100 000. It was subsequently published between 1845 and 1864. During the second half of the 19th century cartographic techniques were improved. For example, triangulation with angles was introduced (in ca. 1870), the absolute elevation of the “Pierre du Niton” was measured (in 1879), and the depth of the major Swiss lakes was assessed for the first time (in ca. 1870). These improvements made it possible to map glaciers in remote regions more accurately (Imhof, 1927). As a result, the Siegfried map was produced between 1868 and 1949 using the Dufour map as a baseline. The aim was to create homogenous maps for the whole of Switzerland for the Topographic Atlas of Switzerland at a scale of 1 : 50 000 for the Alps and 1 : 25 000 for the rest of Switzerland. The project started under the direction of the Chief of Staff, Hermann Siegfried, but most of the mapping was done by cartographers and topographers from the private sector. To ensure homogeneity, precise mapping instructions were set from the beginning (Imhof, 1927; Swisstopo, 2017). At that time, the Siegfried map was considered the most advanced topographic map ever produced; especially impressive was the drawing in the mountainous regions and the representation of rocks e.g. in glacierized areas (Imhof, 1927). Such historical topographic maps provide unique information on large-scale glacier areas for the time period 1868–1949 and are therefore valuable to fill the data gap between 1850 and 1973. They are linked, however, to uncertainties due to the mapping methods available at the time and possible errors in geo-referencing. Such uncertainties may sometimes lead to inaccuracies when glacier areas from historical maps are compared to other products, for example glacier areas from remotely sensed data (Imhof, 1927; Hall et al., 2003; Racoviteanu et al., 2009).
The aim of our study was (1) to digitize the historical Siegfried map at two time slices between 1892 and 1944; (2) to validate the digitized glacier areas through their comparison with glacier areas of different time periods and from different data sources in order to assess their accuracy at the large scale and locally; and (3) finally to create a dataset useful for long-term studies of glacier changes or hydrological modelling.
The Siegfried map consists of a total of ca. 550 sheets that were revised at
different publication years. Each sheet covers an area of 210 km
The 80 sheets of the Siegfried map covering the glacierized area of the Swiss Alps.
The arithmetic precision requested from the topographers of the Siegfried map was 0.7 mm in the projection on the map (corresponding to 35 m in nature) for survey stations in the Alpine region (1 : 50 000). The contour lines are biased because the reference point “Pierre du Niton” is found to be 3.26 m higher than what was assumed at the time (Imhof, 1927). Errors can be up to 18 m because of this reference bias (Imhof, 1927). Furthermore, the measurement directives changed during the creation of the maps. At the beginning (around 1880) 300–500 survey points were needed for the creation of one sheet, while at the end of the 19th century, up to 6000 measurement points were prescribed (Imhof, 1927). Unfortunately, no information on the exact number of surveying stations was provided for the individual sheets (Swisstopo, Brigitte Schmied, personal communication, 24 January 2018). While the vertical accuracy of the Siegfried map has been estimated (Imhof, 1927; Rastner et al., 2016), large regional differences exist in the horizontal accuracy of the different sheets. These may relate to the number of surveying points (Caminada, 2003) and are therefore difficult to exactly estimate (Hall et al., 2003; Rastner et al., 2016).
We use four datasets of glacier areas and outlines covering the Swiss Alps for the years 1850, 1973, 2003, and 2010 (Fischer et al., 2014; Maisch et al., 2000; Müller et al., 1976; Paul et al., 2011) for the validation of the digitized glacier areas of the Siegfried map at the large scale for the first and second periods (around 1900 and around 1935). These four glacier inventories were produced with different technologies and methodologies summarized in Table 1. Furthermore, the outlines of seven glaciers (Silvretta, Oberaar, Unteraar, Limmern, Untergrindelwald, Damma, and Clariden) digitized by Andreas Bauder (ETH Zurich) from different historical maps from several years between 1864 and 1959 were available for local validation. The glacier outlines from years earlier than 1930 were digitized from the first publications of the Dufour and Siegfried maps and later than 1930 from the first publication of the National Map (e.g. Bauder et al., 2007, 2017; Huss et al., 2010). The glacier outlines used for local validation are visible in Fig. 4.
Glacier inventories used for large-scale validation.
All glacier areas of the 80 sheets from the Siegfried map were manually digitized using ArcMap 10.2.2 to create two shape files with the digitized glacier outlines of the first (around 1900) and second (around 1935) periods. Outcrops within the glaciers were removed. For the digitization, the study area was divided into two regions, the Rhine basin and the Rhone, Po, and Inn basins that were digitized by two different persons (Fig. 1) at a scale of 1 : 10 000. A third person finally controlled all digitized areas. Altogether, more than 500 000 nodes corresponding to an average of 28 nodes per kilometre of glacier outline and 250 working hours were needed to create the polygons and resulting shape files.
Frequency distribution of the publication years of the 80 sheets from the Siegfried map for the first and second periods.
To assess the quality and accuracy of the glacierized area from the Siegfried map at the large scale, the digitized glacier outlines of the first and second periods were compared with the glacier outlines of four available glacier inventories (Table 1, for the years ca. 1850, ca. 1973, 2003, and ca. 2010) in a two-step validation process. The accuracy of the Siegfried map is difficult to assess at the large scale (Hall et al., 2003; Rastner et al., 2016), as no contemporary data are available for comparison. The two-step validation process presented below, however, allowed us to assess whether the digitized glacier areas were consistent with the other available products, meaning that the digitized glacier area of the first and second periods followed a logical evolution compared with the other products.
In a first step, the shapes of the digitized glacier areas from the first and
second periods were visually compared to the glacier shapes of the four
inventories in order to ensure that they were consistent. During this
comparison, the digitized glacier outlines from the first and second
digitized periods that appeared in none of the other products were removed as
the existence of a glacier in this location could not be verified. This was
the case for ca. 0.03 % of the digitized area of the first and second
periods (61.6 and 49.8 km
To allow comparison between the glacier areas of the different data sources
and available years, the digitized glacier outlines from the Siegfried maps
and from the four inventories were divided into 957 glacier basins with a
unique identity number, based on the river basin delineations given by the
Federal Office for the Environment (FOEN). This method follows the
recommendation of the GLIMS Analysis Tutorial (Racoviteanu et al., 2009). The
total glacierized area was calculated for each of the 957 glacier basins and
is further referred to as
Assuming that all glaciers reached their maximum extent at the end of the
Little Ice Age around 1850 in the central European Alps (Collins, 2008;
Ivy-Ochs et al., 2009; Vincent et al., 2005), the glacier areas from 1850
should be the largest. In the second validation step, Highly consistent: Consistent: ( Poorly consistent: 0.1 Not consistent: ( not consistent but plausible: when it could not be decided from the glacier
shape of
As land cover classification in remotely sensed data is not unequivocal (e.g.
Racoviteanu et al., 2009) and definition and recognition of moraine partly
rely on interpretation (Clark et al., 2004), the 1850 glacier inventory also
shows uncertainties and we therefore considered
The results of the validation are presented in Table 2 and Fig. 3. As the
results were very similar for
Large-scale validation of the digitized glacier outlines
Comparison of the digitized outlines of the Siegfried map for seven glaciers with contemporary products from other sources.
Large-scale validation of the digitized glacier outlines shown, as
an example, for
Mercanton (1958) calculated the total glacier area for the main Swiss river
basins and for two time periods based on an early edition of the Siegfried
map (published between 1869 and 1895 – ca. 1876) and the first National Map
(surveyed between 1917 and 1945 – ca. 1934). We assessed the total glacier
area for the same river basins with the digitized glacier areas of the
Siegfried maps for the first and second periods (published ca. 1900 and
ca. 1935) and with the four glacier inventories (Table 3). Next, we
calculated the mean relative change in glacier area (
Comparison of the total glacier area (km
Overall, the total glacier area of the different river basins decreased
between 1850 and 2010 for the eight compared datasets and the glacierized
area of the Swiss Alps decreased by a total of ca. 53 %. The yearly
changes in glacier area are overall higher for the period after 1973 than for
the period before 1973, reflecting the observed increases in glacier area
loss in the last decades (e.g. Huss et al., 2008). However, some anomalies
can be observed between the datasets. The total glacierized area of the Swiss
Alps from the reconstructed glacierized area of ca. 1850 was 1781 km
The comparison between the glacier area of the National map (surveyed
ca. 1934) and the Siegfried map (published ca. 1935) shows the largest
differences with
In Fig. 4, the glacier outlines of the glacier areas from seven sheets of the digitized Siegfried map for the first and second periods are compared to glacier outlines digitized from the Dufour and Siegfried maps (earlier than 1930) and from the National Map (later than 1930) for local validation. For the Limmern, Clariden, Untergrindelwald, and Silvretta glaciers, only little differences were observed between the first and second digitized periods, meaning that the Siegfried map of the second period was probably only re-edited (see Sect. 3.2.1). The comparison between the datasets shows for all glaciers (with the exception of the Limmern glacier) good consistency and logical evolution in shape, especially in the ablation area. For the accumulation zone, differences can be observed between the datasets, especially for the Untergrindelwald and Unteraar glaciers. These differences are due to different delineations of the glacier area between the different products. The additional glacierized area from the accumulation zone in the digitized Siegfried map is also present in the four glacier inventories and therefore consistent. The shape of the Limmern glacier is in the ablation area different for the digitized Siegfried map and for the comparison products from the National map. The Siegfried map seems inaccurate for this glacier.
Examples of conflicts encountered during digitization of historical maps. The map shows the glacier area digitized for the end product (blue area with black outlines) and in the background the sheet of the Siegfried map for the publication year 1934 for the Wyttenwasser glacier. In cases A to C the outlines of the same glacier area digitized by five students are shown (coloured lines).
To assess the accuracy of the digitization, five hydrology masters students
(age 20–25 years) digitized all glaciers over a 23 km
We estimate the precision of digitization to be ca. 5 %. However, it is
more difficult to estimate the accuracy of the Siegfried map itself. As the
uncertainty is different for each sheet (see Sects. 2.1 and 3.2.1), large
regional differences can be found in the accuracy of the glacier outlines. On
some sheets, the inaccuracy of the Siegfried map might be much higher than
the interpretation bias of the digitization. However, the large-scale and
basin-scale validations allowed us to assess which ones of the digitized
glacier areas followed a logical evolution in shape and area and were
therefore plausible compared with the other available products of glacier
outlines for different years; 71 % of the glaciers and 88 % of the
glacier area were considered consistent through the analysis. The local
validation furthermore showed that the shape of the seven analysed glaciers
was well represented in the Siegfried map. This analysis however is only
valid for the studied glaciers and not for the entire area. While the
presented product of glacier outlines contains all digitized glacier areas
from the Siegfried map for Switzerland (Sect. 4), we recommend only using the
glacier areas that were stated as “consistent”, “highly consistent”, or
“more consistent than
During the creation of the Siegfried map, the time span from measurements to publication extended up to several years, due to the material available at that time and to the complex topography. The Siegfried sheets are unfortunately only given with the publication year and no further information can be found on the surveying year. Therefore, one should keep in mind that the year given with the digitized glacier outlines from the Siegfried map is only representative for a period of time and cannot be taken as an exact date.
The digitized glacier areas of the Rhine River basins were used to develop the glacier routine of the HBV-light model in order to implement transient changes in glacier area and volume from 1900 to date (Seibert et al., 2018). This model was then used within the ASG-Rhine project with the aim of calculating the snowmelt, glacier melt, and rainfall contribution to the Rhine discharge for the time period 1900–2006 (Stahl et al., 2017). The glacier areas and glacier mass balances of several glaciers calculated within the ASG-Rhine project for the beginning of the 20th century showed comparable results to contemporary analyses or observations from other studies (Stahl et al., 2017). These different applications show that the digitized Siegfried map brings important information on glacier area for large-scale and long-term analysis and can be successfully used to better understand and model glacier area changes.
The datasets of glacier area for the first and second
digitized periods (around 1900 and around 1935) presented in this paper are
freely available from the FreiDok plus data repository
(
We digitized glacier outlines from the Siegfried map for the Swiss Alps for two periods around 1900 and 1935. We dealt with the challenges of digitization of historic maps (e.g. uncertainties in georeferencing, time of measurement vs. time of publication) with two validation schemes at the large scale, basin scale, and locally. Comparison to four existing glacier inventories covering different time periods revealed that at least 70 % of the digitized glaciers and 88 % of the total glacier area were comparable for both digitized periods to the glacier areas and shape of the glacier inventories and therefore plausible. Further comparison at the river basin and glacier scale showed reliable glacier representation for most of the areas. The uncertainty of the digitization itself was assessed separately and was less than 5 %, which is comparable to the accuracy of deriving glacier outlines and areas from remotely sensed data. The presented datasets for a first period around 1900 and a second period around 1935 are valuable information for the glacier extent in the Swiss Alps at the beginning of the 20th century where no other data source is available covering the entire Swiss Alps. The dataset closes the gap between the reconstruction of the glacier areas at around 1850 from the moraine extent and the first complete dataset of glacier areas in the Swiss Alps from aerial photographs in 1973. Under consideration of the data uncertainty, the use of the digitized datasets in combination with other existing glacier inventories can provide important information about changes in glacier areas for the last 120 years, which is essential for long-term and accurate glacier mass balance or hydrological modelling in glacierized basins.
DF homogenized and validated the presented datasets and prepared the manuscript with contributions from all co-authors.
The authors declare that they have no conflict of interest.
The authors thank Swisstopo for providing the historical topographic Siegfried map and Matthias Huss for providing several glacier outlines for validation. We are grateful to Damaris De for the digitization of part of the glacier areas. We furthermore thank Mirko Mälicke and his students Ruben Beck, Daniela Boru, Helena Böddecker, Verena Lang, Lukas Maier, and Miranda Perrone for assessing the accuracy of the digitization. We also want to thank two anonymous referees and Rheinhard Drews for their valuable comments and suggestions that helped to improve our manuscript. The glacier areas of the Rhine basin were digitized within the ASG-Rhein project (snow and glacier melt components of the streamflow of the River Rhine and its tributaries considering the influence of climate change) funded by the International Commission for the Hydrology of the Rhine Basin (CHR). The remaining glacier areas were digitized within the Hydro-CH2018 project funded by the Federal Office for the Environment (FOEV). The first author was funded by the German Federal Environmental Foundation (DBU). Edited by: Reinhard Drews Reviewed by: two anonymous referees