The study area is located in the foreland of the Stein glacier above the tree line in the central Swiss Alps, south of the Sustenpass in the Urner Alps (approximately 47∘43′ N, 8∘25′ E). Elevations range from 1900 to 2100 m a.s.l. The area lies in the polymetamorphic Erstfeld gneiss zone, which is part of the Aar massif . The geology is defined by metamorphosed pre-Mesozoic metagranitoids, amphibolites, and gneisses ; thus the material is mainly acidic and rich in silicate. The closest official weather station is located 18 km away at Grimsel Hospiz (46∘34′ N, 8∘19′ E) at an elevation of 1980 m a.s.l. For the norm period from 1981 to 2010, the station recorded an annual mean temperature of 1.9 ∘C and an annual precipitation of 1856 mm. The precipitation distribution throughout the year is fairly uniform with a slight increase in the winter months . The glacier foreland consists of moraines with unconsolidated glacial till. The humid and cool climate together with the nutrient-poor substrate and a relative high water permeability of the glacial till favor the formation of podsolic soils and humus in this area .

The moraines of the Stein glacier were exposed due to its retreat to the south. Four moraines were selected for this study (see Fig. ). conducted a detailed dating study of the Stein glacier moraines, based on high-sensitivity beryllium-10 moraine dating and found that the ages of three moraines range between 160 to 10 000 years. The age of a fourth moraine was dated to 30 years based on maps and aerial photos.

The oldest moraine with an age of 10 000 years is facing northeast. The second oldest moraine with an age between 2000 and 3000 years is the only one facing south. The two youngest moraines were exposed in the years 1860 and 1980–1990 and thus have an age of 160 and 30 years, respectively. Both moraines are facing northeast and are located closer to the glacier tongue at a distance of approximately 1 km from the oldest moraines. Both moraines are south of the glacial lake Steinsee (1930 m a.s.l.), which is a proglacial lake that was formed by the glacier retreat in 1924 .

The vegetation of the moraines was mapped in summer 2017 . Pronounced differences in vegetation coverage and species distributions were found among the four age classes. The vegetation of the oldest moraine was dominated by a variety of prostrate shrubs together with small trees and several grasses. On the 3000-year-old moraine, a grassland cover with fern, mosses, sedges, and forbs was found. The two youngest moraines however showed a lower degree of vegetation complexity. On the 160-year-old moraine, a combination of grasses, lichen, forbs, and shrubs was present. The youngest moraine still shows only a sparse vegetation cover with mainly grass, moss, forbs, and a few shrubs.

The dye tracer experiments were conducted between mid July and mid August 2018. We used Brilliant Blue FCF as dye tracer due to its good visibility, high mobility, and non-toxicity. We used a concentration of 4 g L-1, at which the tracer shows good sorption and visibility . Three study plots were selected at each moraine, based on the degree of vegetation complexity (low, medium, and high complexity). Vegetation complexity is characterized by vegetation coverage, the number of species, and the plant functional diversity. The functional diversity is calculated based on specific leaf area, nitrogen content, leaf dry matter content, Raunkiær's life form, seed mass, clonal growth organ, root type, and growth form. The collection of the required data and calculation of the vegetation complexity was done by the geobotany group of the University of Freiburg and is described in more detail in .

The size of each study plot was 1.5 m ×1.0 m. The distances between the three study plots at each moraine ranged from 10 to 100 m. Each plot was further divided into three equal subplots of 0.5 m ×1.0 m. Figure  shows the experimental design at each moraine and illustrates the irrigation procedure.

The subplots were irrigated with three different amounts of dyed water (20, 40, and 60 mm) and an irrigation intensity of 20 mm h-1. The irrigation intensity is relatively high with a return period of 2.8 years . In preparation of the tracer application, large vegetation in the form of shrubs and bushes was cut off to a height of a few centimeters to reduce interception. The tracer was applied with a hand-operated sprayer connected to a battery-powered pump which guaranteed a constant pressure for a uniform flow rate of 60 L h-1. For a time-efficient irrigation of the three subplots with three irrigation amounts, the irrigation procedure was divided into three steps. In the first step, all three subplots were irrigated simultaneously for 60 min in a sequence of 5 min of irrigation and a 5 min break. This provides an application of 20 mm to all three subplots. After finishing the first step the first subplot was covered to avoid any additional water input. In a second step, the other two subplots were simultaneously irrigated for an additional 60 min in a sequence of 5 min of irrigation and 10 min breaks. This provides an application of an additional 20 mm to each of the two remaining subplots. In the last step, only the third subplot was irrigated for 60 min in a sequence of 2 min of irrigation and 10 min breaks, while the other two plots remained covered providing an additional 20 mm to this subplot. After the end of tracer application, the entire plot was covered to avoid any disturbance by natural rainfall.

The next day each subplot was excavated in up to five profiles of 7 to 10 cm. After the profile cuts were made with pickaxes, spades, and hand shovels, the profile walls were cleaned. Hanging roots were cut off, and rocks were not removed but made visible. The profiles of each subplot were photographed with a Panasonic Lumix DMC-FZ18 camera and at a resolution of 2248 pixels ×3264 pixels. A big umbrella was used to provide a uniform light distribution in the photographs and to avoid direct sunlight. A wooden frame for geometric correction and a grayscale (Kodak) attached to the frame (Fig. ) for later color adjustment were included in the photographs. Since dye tracer experiments only provide snapshots of flow patterns at 24 h after the irrigation, we cannot exclude the possibility that initial preferential flow paths were obliterated by a later downward movement of the infiltration front. However, as the probability of this special case is relatively low, we assume that these snapshots are a viable basis for the comparison of characteristic flow patterns along the moraine ages.

The image analysis procedure by was used to generate tricolor images of the photographs showing stained and unstained areas (Fig. ). A detailed description of the method can be found in . Instead of using the original IDL (Interactive Data Language) software package, a similar Python version was used. Basically, geometric correction, background subtraction, and color adjustment were carried out to correct differences in image illumination and changes in the spectral composition of daylight. The delineation of rocks and plants was done manually. In the resulting tricolor image, the horizontal and vertical length of a pixel correspond to 1 mm. Due to poor lighting conditions or a heterogeneous background color distribution in the soil caused by material transitions, small stones, or organic matter, the image analysis software was not able to recognize all large dye stains as coherent objects. Thus, a manual correction of the images using the photographs was necessary (see Fig. ).

For a quantitative comparison of the dye patterns, the maximum infiltration depth, the volume density, and surface area density, as well as the stained path width, were calculated. These parameters are frequently used for an objective comparison and description of the dye patterns . Volume and surface area density are originally steorological parameters which are used to relate three-dimensional structures to measured two-dimensional parameters . The volume density corresponds to the dye coverage and can be derived from one-dimensional information by calculating the fraction of stained pixels for each depth. The volume density profile is defined by the fraction of stained pixels per depth and is calculated as the average of all excavated profiles per plot. The surface area density in one dimension is calculated by using the intercept density, which describes the number of intercepts between stained and unstained pixels divided by the horizontal width of the soil profile. The profile of the surface area density describes the amount of intercepts per depth and is then also averaged over all photographed profiles per plot. Volume density provides no information about whether the stained area is the sum of many small fragments or a few large ones; thus the volume density alone should not be used to characterize flow patterns, and the surface area density should be used as a supplementary parameter. A high surface area density indicates a large number of small features.

Following the method described by , the resulting dye patterns were next classified into flow type categories based on the proportions of three selected stained path width classes (stained path width <20, 20–200 mm, and >200 mm) relative to the volume density. The stained path width is equal to the horizontal extent of a stained flow path . This classification method distinguishes between five flow types: (1) macropore flow with low interaction, (2) mixed macropore flow (low and high interaction), (3) macropore flow with high interaction, (4) heterogeneous matrix flow and finger flow, and (5) homogeneous matrix flow. Dye patterns which cannot be classified as one of these flow types are categorized as undefined. The classification method based on proportions of the classes of stained path width is illustrated in Fig. .

This method is based on the assumption that the dye patterns are mainly controlled by certain preferential flow processes and that each flow process creates a characteristic dye pattern that can be described by the extent and distribution of stained features.

This method was proven to be suitable for the investigation of and and was also used in a variety of additional studies . However, an extension of the flow type categorization was needed for our study site with soils of different ages, texture, and additionally a high stone content. In the extended classification, we avoid a clear differentiation between macropore flow and finger flow. The original classification assigns finger flow only when both classes of the medium-sized stained path width (20–200 mm) and the biggest stained path width (>200 mm) account for approximately half of the dye coverage. This implies that finger flow is only prevalent when the dye pattern is characterized by a majority of broader stained path widths, ignoring the fact that the size of the finger-like flow paths can vary over a broad range . Thus, we argue that while finger flow and macropore flow are caused by different properties and flow mechanisms, they can lead to similar dye patterns and distributions of stained path width classes. This is especially the case for macropore flow with high interaction, which creates broader stained paths that could also be assigned to finger flow. Therefore, both flow types were considered in the extended classification for this class. Furthermore, it was observed that the presence of rocks within the image analysis interrupts homogeneous blue stained areas and thus leads to smaller stained path widths. Using the original classification scheme on a soil profile with high stone content suggests a heterogeneous flow pattern, which can be classified as heterogeneous matrix flow, finger flow, or macropore flow depending on the abundance of rocks. Therefore, an additional class has been introduced, which is used when homogeneous matrix flow between rocks takes place. The classification rule for the additional flow type class is based on the proportion of blue-dye coverage of the available permeable matrix space (profile width minus the sum of stone widths per row). If at least 95 % of the permeable space is stained by blue dye, the flow type is classified as matrix flow between rocks.

Soil samples were taken during August and September of 2018 close to each dye tracer plot. For grain size analysis, two disturbed bulk soil samples per depth were taken at 10, 30, and 50 cm depth at each plot. The total of 72 samples was analyzed in the laboratory between November 2018 and January 2019 by using a combination of dry sieving (grain sizes >0.063 mm) and sedimentation analysis (grain sizes <0.063 mm) with the hydrometer method . Organic matter removal was only possible by floating off the lighter fractions prior to grain size analysis. Since three plots were selected per moraine for the Brilliant Blue experiments, six samples per depth and age class (a total of 18 samples for each moraine) were available for the grain size analysis. Grain sizes between 2 and 0.063 mm were classified as sand, between 0.063 and 0.002 mm as silt, and smaller than 0.002 mm as clay. Grain size fractions of particles <2 mm were calculated as weight percentages of total weight of particles <2 mm, thus excluding gravel and stones to avoid single larger stones shifting or dominating the distribution. The gravel and stone fraction was calculated separately as a weight percentage of the entire soil sample.

For the analysis of the structural parameters, soil samples were taken with sample rings to provide undisturbed cores which preserve the natural soil structure. At each plot, two 250 cm3 and one 100 cm3 undisturbed soil samples were taken at a depth of 10 and 30 cm. Three samples of 100 cm3 were taken at 50 cm depth. Thus, per age class, nine samples per depth were available for the determination of the structural parameters of porosity and bulk density. A detailed overview of the sampling scheme at each plot is given in Fig. .

The porosity was determined in the lab using the water saturation method. For this method, sample weights were recorded at saturation and after drying at 105 ∘C. For saturation, the samples were placed in a small basin. The water level in the basin was increased stepwise by 1 cm d-1. When the water level reached the top of the soil sample and the sample was fully saturated, the bottom of the sample was sealed, and the weight at saturation was measured. Bulk density was determined by relating the dry mass after drying at 105 ∘C to the sample volume. The loss on ignition is a measure of the organic substance in the soil and describes the proportion of the organic substance that was oxidized during annealing for 24 h at 550 ∘C. The loss on ignition was determined by drying subsamples (4–6 g) for at least 24 h at 105 ∘C and then at 550 ∘C. The ignition loss is then calculated by relating the weight loss after drying at 550 ∘C to the sample weight after drying at 105 ∘C.

The nonparametric Kruskal–Wallis test was used to test the significance of the differences in the soil texture among the four moraines of differing age classes. It can be applied when the assumption of a normal distribution cannot be made and is also valid for small sample sizes. We applied the test to each grain size fraction across the four age classes. Average values were based on 18 samples per age class. The grain size distribution at 10 cm depth of the oldest moraine was excluded from consideration, since due to the high organic content not all organic matter could be removed, and the results may therefore be erroneous.

Comparing the depth-averaged soil texture over the millennia, we find that while the soil texture at the youngest moraine mainly consists of sand, the grain sizes decrease over the millennia with silt being the largest fraction after 10 000 years (Fig. ). Clay content increased with age for the three older moraines, with the youngest moraine being the exception (having the second-highest clay content). The Kruskal–Wallis test with a 0.05 confidence level showed that differences in grain size fractions among the four age classes were statistically significant (p values <0.05; sand of p=0.0013, silt of p=0.0006, and clay of p=0.0018).

A significant reduction in grain size over time was observed, which is most pronounced between 3000 and 10 000 years of soil development. The gravel and stone fraction is roughly the same at the three younger moraines and significantly lower at the oldest moraine (Fig. b). The structural parameters of porosity and bulk density also show a clear trend with age, with porosity increasing and bulk density decreasing (Fig. ).

The porosity observed at the youngest moraine ranges between 0.22 and 0.37, with no pronounced differences among the three soil depths. The 160-year-old moraine has a higher porosity in the upper 10 cm than the 30-year-old moraine. After 3000 years the increase in porosity continues but is now also visible in 30 and 50 cm, but with much higher values at 10 cm (Fig. a). After 10 000 years the porosity at 10 cm ranges from 0.6 to up to more than 0.8. The other two depths also experienced a further increase in porosity. The decrease in bulk density is also most pronounced in the top layer of the soil (Fig. b). While after 30 years the bulk density in the upper 10 cm ranges around 1.7 g cm-3, the bulk density after 10 000 years is much smaller and ranges between 0.2 and 0.7 g cm-3. After 3000 years the trend is also visible at 30 and 50 cm.

The loss on ignition, as a measure for the organic matter content, shows an increase throughout the first 10 millennia of soil development, which is most pronounced in the upper soil layer (see Fig. c). At the two youngest moraines, the organic matter content is still very low (<2 wt %; percentage by weight). At these two age classes, the organic matter content is homogeneously distributed over the profile, with a slight tendency to higher values in the topsoil at the 160-year-old moraine. The 3000-year-old moraine shows a strong increase in the organic matter content in the surface layer. At the oldest moraine, the trend of increasing organic matter continues in all three depths. Here, the organic matter content in the topsoil makes up to two thirds of the soil material. However, the organic matter content varies distinctly with a minimum of 6 wt % and a maximum of 87 wt %. In deeper depths, the organic content also increases compared to the 3000-year-old soil but remains below 20 wt %.

Even though the differences in soil physical characteristics between 30 and 160 years are comparatively small, the rates of change during this initial phase are highest (Table ). Between 160 and 3000 years, the rates of change are significantly reduced and remain in a similar range between 3000 and 10 000 years.

Flow patterns traced with Brilliant Blue dye changed considerably with moraine age (Fig. ). The volume density profile is a measure of the amount of blue dye per depth. The profile patterns of volume and surface area density show distinct differences among age groups, while differences between the vegetation complexity levels are not as clear (Figs.  and ).

The volume density of the blue dye was classified in three selected groups of stained path width . Additionally, the volume density of rocks was also classified in three groups (Fig. ).

An analysis of the average or maximum infiltration depth based on the dye profiles was not possible because not all profiles could be excavated up to the maximum infiltration depth. In most cases, large boulders prevented further excavation or the infiltration depth was more than 1 m. The latter was mostly the case at the youngest moraines. Only at the oldest moraine could a maximum infiltration depth be determined based on the dye profiles.

Combining the volume density profiles with the surface area density profiles, it is possible to derive whether the stained area described by the volume density is made up of many small flow paths or few large ones. The shapes of the volume density profiles and the surface area density of the youngest moraine are all very similar across the vegetation complexity levels and irrigation amounts. The youngest moraine has a higher volume density of flow paths in the top half of the soil profile than all other moraines (Fig. ). There are almost no unstained areas. Beginning from approximately 30 cm depth, the volume density declines. The surface area density profiles show an opposite pattern (Fig. ). The surface area density is smaller in the upper half and increases in the lower half of the profile. The combination of both parameters indicate a homogeneous staining in the top half, where interruptions of stained areas are only caused by rocks. In the lower part of the soil profiles the flow paths are subdivided, which is indicated by the increase in surface area density (apparent at the plots at 30 cm depth for low, 25 cm depth for medium, and 50 cm depth for high vegetation complexity). This combined with a decline in volume density indicates a narrowing of the flow paths. For the plots of low and high vegetation complexity, the change in flow paths coincides with a layer of higher clay and silt content. This layer does not exist at the plot of medium vegetation complexity. In this case, the narrowing and splitting up of flow paths is caused by large rocks. No clear differences are visible between the different irrigation amounts. The proportion of classes of the stained path width is controlled by the existence of rocks. The maximum infiltration depth is either controlled by the position of the clay layer or is deeper than the profile depth and therefore cannot be determined.

Comparing the 160-year-old moraine to the youngest moraine, we find that the volume density is lower and the surface density is higher in the upper part of the profile. Also unstained areas (colored white in Fig. ) are visible, which indicates that the higher surface density is not caused by the existence of rocks that split up a homogeneously stained area as was the case for the youngest moraine. In this case there is no total dye coverage of the permeable soil, and the preferential flow paths are initialized already near or at the soil surface. The surface density profiles show a decrease in the lower half of the soil profile, which either goes along with a decrease in the blue-dye coverage and an increase in the stone coverage or an increase in the blue-dye coverage without a significant change in stone coverage. Both indicate a reduction in the amount of separate flow paths and an increase in flow path widths, even if the permeable space is reduced due to an increase in rock content. Also the fraction of stained path width (SPW) bigger than 200 mm increases, which indicates that the dye plumes widen in deeper soil depths.

Compared to the two youngest moraines, the moraine of 3000 years shows in general a higher surface area density and a lower dye coverage combined with a higher fraction of an unstained permeable soil matrix. This indicates that similar to the 160-year-old moraine, water is transported in individual flow paths, but here there are more flow paths, and they have a smaller width. This can also be seen in the less frequent appearance of stained path width higher than 200 mm (Fig. ). Similar to the 160-year-old moraine, preferential flow paths are already initialized at the top of the soil during infiltration (apparent from the white-colored areas across the profile in Fig. ).

The oldest moraine with an age of 10 000 years shows the highest surface area density and the lowest volume density combined with the lowest infiltration depths. The surface and volume density profiles show the same pattern: after a peak close to the soil surface, both density profiles show a decrease with soil depth. This means that the dyed water is only transported deeper into the soil via a few individual flow paths.

Using the information in the volume density profiles and the stained path widths to characterize flow types , we found a trend from a rather homogeneous flow pattern with matrix flow in a fast-draining coarse-textured soil at the youngest moraine to a more heterogeneous flow pattern with a mix of heterogeneous matrix flow and finger flow at both medium-age moraines (Fig. ). While the flow characteristics of both the 160- and 3000-year-old moraine are dominated by finger flow with smaller stained path widths, this is much more pronounced in the 3000-year-old moraine. By contrast, the oldest moraine stands out clearly from the other age groups. Here, only the macropore flow is predominant. For the deeper soil layers, this was confirmed during the field experiments, since the few macropore flow paths were clearly visible. For the topsoil, the result is less certain, as the blue areas were very difficult to identify during the image analysis due to the very dark color of the organic layer.

The relative frequency distribution of the flow types per moraine age class derived from the results in Fig.  shows a clear shift in the flow type distribution along the age classes (Fig. ). At the youngest moraine, all types of finger flow and matrix flow are present, and the frequency distribution does not show a distinct peak at any flow type. With increasing age, macropore flow becomes more and more important, and the peaks in the frequency distribution become more and more pronounced (Fig. ).

The observation of bulk density, porosity, and soil texture show significant differences between the age groups as well as some clear trends with age. A detailed description of this data set can be found in .

The flow type classification by was used to classify the volume density patterns into flow type categories. A comparison between the observations made during the excavation and the derived flow types showed that in this case an adaptation of the flow type classification was necessary. On the one hand, this adaptation involved the treatment of rocks to prevent misclassification, and, on the other hand, we introduced the possibility of finger-like flow paths with smaller widths. After these adaptations the derived flow types correspond well with the observations made in the field.

The observed staining patterns and derived flow types show a significant difference between the age groups, whereas no significant difference was observed with respect to vegetation complexity and irrigation amount.