Preferential flow (PF) is an important process in soils, facilitating the rapid transport of substantial amounts of water through the soil . During preferential flow, water flows through preferred pathways rather than the surrounding soil matrix, bypassing much of the soil . Preferential flow does not occur during every rainfall or infiltration event, but when it is activated, it can account for a significant proportion of annual flow though the soil . Once formed, these pathways can persist across multiple events, with soil saturation playing a key role in determining which flow features become active .

Understanding the occurrence of preferential flow is important for predicting runoff generation and contaminant transport. Preferential flow pathways can transport water much faster than the soil matrix , with reported lateral velocities ranging from 0.3 ms-1 up to 0.8 ms-1 . These high velocities can lead to rapid responses in streamflow. Additionally, the complex and interconnected nature of preferential flowpaths allows water to move through the soil with minimal interaction with the soil matrix . This reduces a soil's capacity to filter and retain contaminants, potentially leading to the rapid transport of pollutants into the groundwater or rivers .

Three primary modes of preferential flow can occur, also simultaneously, in soils: fingered flow, funneled flow and macropore flow .

Fingered flow occurs when the infiltration wetting front reaches a more permeable soil layer and instabilities cause downwards flow in “fingers” . It is driven by gravity with the highest water content at the finger tip, widening through diffusion into the soil matrix . Flow velocity within a finger decreases over time, thus more fingers form to transport an equal amount of water .

Funneled flow occurs when infiltrating water reaches an impermeable or less conductive soil-layer, which then causes the flow to be channeled laterally . This funneling effect can guide water into vertical preferential flowpaths like macropores .

Macropore flow is the predominant preferential flow process in many soils, transporting up to 95 % of water through less than 0.3 % of the total soil volume . This type of flow includes various forms such as biopore flow, inter-aggregate flow, and preferential flow within soil aggregates . Macropores are created through multiple mechanisms, including biological activity from soil fauna (e.g. earthworms, gophers, beetles) , plant roots , soil cracks caused by shrinkage and soil pipes formed by internal erosion . These macropores can persist for years, and in some cases, they may remain active for decades . Within a soil profile, smaller diameter flowpaths can restrict preferential flow throughout the whole soil . However, even small continuous macropores can transport significant volumes of water if they maintain connectivity . Macropore flow is initiated at or close to topsoil saturation . During water flow in macropores, they are typically not filled entirely with water but a water film forms along the walls in the pore, leaving the central part of the pore empty . The complex network of macropores, particularly soilpipes in hillslopes, can extend for hundreds of meters and can contribute substantially to overall discharge in streams . This extensive and efficient transport network underscores the importance of macropore flow in both hydrological and contaminant transport processes within soils.

In hillslopes the interplay of funneled and macropore flow is also called subsuface stormflow (SSF) . Macropores can rapidly transport infiltrating water to depths where it laterally discharges through soilpipes and funneled flow . Observed SSF velocities in preferential flowpaths are higher than in the surrounding soil matrix .

Detecting preferential flow is challenging due to the inherent complexity and heterogeneity of the relevant soil structures. It occurs through a network of earthworm channels, inter-aggregate spaces, and root meso- and micropores . However, identifying these voids alone is insufficient, as many do not actively transmit water during preferential flow events . Also, field surveys are only a snapshot of an environmental system, but flowpath arrangements can change completely within one year . Additionally, even when preferential flow features are successfully identified at the plot scale, these findings can not be easily extrapolated to the whole catchment. This limitation arises because hillslopes exhibit varying hydrological responses influenced by micro-topography, soil properties, and curvature, leading to significant spatial variability in preferential flow pathways. . As a result, understanding and predicting preferential flow at broader scales remains a complex challenge that requires integrating spatial and temporal variability in soil structure and hydrological processes.

Many different methods exist to identify and partly quantify preferential flow in soils, but each has its own limitations:

Dye tracers and sprinkling experiments are commonly used to visualize flowpaths in soils but have the disadvantage of destroying the site during excavation for surveying . Additionally, these techniques may miss flowpaths that transport pre-event water, as these pathways will not be marked by the dye of the event water .

This study aims to detect lateral and vertical preferential flowpaths in soils using stable water isotopes. Stable water isotopes are controlled by several factors . The most relevant factor, fractionation, results from the phase transitions of water from the liquid to the gaseous phase and vice versa. Water vapor, which contains the heavier isotopes 2H and 18O, has a higher vapor pressure than water vapor of the lighter isotopes 1H and 16O, leading to a slower diffusion rate of heavy water molecules. This results in a preference for light water molecules in the during evaporation and for heavier isotopes during condensation . The fractionation effect is temperature-dependent, with colder conditions enhancing the enrichment of heavy isotopes in the liquid phase and lighter isotopes in the vapor phase .

Stable isotope signatures in precipitation vary spatially due to factors such as the continental effect, where precipitation is progressively depleted of heavy molecules with increasing distance from the ocean. By favouring heavy water molecules in condensation, the air masses gradually loose heavy isotopes . Conversely, high-intensity precipitation events can lead to rain-out fractionation, which favours lighter isotopes . Altitude also plays a role, as cooler temperature at higher elevations enhance condensation, leading to a depletion of heavier molecules . In the Alps, this effect causes a change of -0.2‰ per 100 m elevation for δ18O .

Seasonal variations in temperature also influence isotope ratios, especially in temperate regions. During summer, precipitation tends to have more negative isotope signatures compared to winter . In particular the seasonal variation and rain-out effects provide a dynamic input of isotopic signatures into soils, making them an ideal tracer to determine water origin and flow processes within the soil . Different flow mechanisms within the soil modify these signatures, creating distinct patterns in isotope depth profiles. By analyzing the isotopic signature of precipitation and soil cores, it is possible to infer the flow mechanisms operating within the soil .

During a rain event, precipitation first infiltrates into the topsoil, displacing older stored water and altering the isotopic signature with that of the recent precipitation . The infiltrating water causes vertical transport (translatory flow) of the event water propagating through the soil matrix . This new event water pushes the older pre-event water deeper into the soil . This process preserves the seasonal isotopic signature of precipitation within the soil, for example for observations during summer with higher isotopic ratios from recent summer events near the surface and lower ratios from older winter precipitation at greater depths .

However, other processes can modify the isotopic seasonality within the soil. Evaporation causes an enrichment of the heavier isotopes in the topsoil . Advective and dispersive transport further dampens the seasonal isotope signal with increasing depth due to stronger mixing with preexisting event water . This mixing can also occur due to rising and falling groundwater levels . In hillslopes, uplslope soils preserve the precipitation signal, but further downslope, lateral flow in the soil leads to mixing and a damped isotopic signal. At the footslope, the seasonality signal may be entirely lost .

Preferential flow also distinctly alters the seasonality signal . With preferential flow, young event water can bypass older, stationary water in the soil matrix, and reach deeper layers, thereby modifiying the isotopic profiles locally . Initially, water in preferential flow paths reflects the isotopic signature of the recent event, Over time, however, the signatures shift towards those of pre-event water due to lateral infiltration and exchange processes between the preferential pathways and the surrounding matrix . This exchange causes mixing, resulting in a composite isotopic signature .

To assess the occurrence of preferential flow at the catchment scale and its relation to landuse and topography, we focus on identifying isotopic alterations in the soil matrix that indicate the occurrence of preferential flow. Previous studies have derived these alterations by comparing isotope profiles either to modeling results of soils without preferential flow , or by analyzing soil cores sampled at different times . In contrast, we compare simultaneous profiles against “reference profiles” derived from a subset of the data to identify deviations indicative of preferential flow.

The developed method allowed us to identify numerous isotopic deviations from the reference profile, which we interpreted as preferential flowpaths in the soil profiles. Scaling the δ18O values to the δ2H range and creating individual reference profiles for each sampling campaign was effective for this purpose. However, the method is subject to certain assumptions and trade-offs which we will now discuss.

Our soil core drill sampling approach offers several practical advantages for catchment-scale assessments, compared to traditional methods for identifying preferential flow, such as dye tracer sprinkling experiments, trenches, and soil moisture sensors. While conventional techniques are effective at the plot scale, they often require intensive excavation and incur high material and maintenance costs, especially for permanent installations. In contrast, soil coring is rapid, cost-effective and logistically straightforward, enabling broad spatial coverage across many hillslopes in a short period. This makes it especially useful for identifying the presence and distribution of preferential flow features across large catchments. However, the method has key limitations. It provides only a narrow, point-scale view (based on a 6 cm drill diameter) and intersecting a PF path is largely a matter of chance. Furthermore, it only detects macropore features that have interacted with the surrounding matrix, so bypass flow paths that do not mix with matrix water remain undetected . Additionally, unlike trench installations or soil moisture sensor networks, the soil coring method does not allow preferential flow fluxes to be quantified. In contrast, though spatially limited, dye tracing experiments can visualize continuous flow paths over a larger plot-scale footprint and provide more detailed information on flow path structure, being less hit-or-miss.

One disadvantage is that a sufficient number of samples are required , with more samples improving analysis quality. BF.summer had the fewest profiles used in a sample group, with 26; Alps.fall had the most, with 96 (Table ). Taking more samples may allow for robuster and clearer reference profiles. Improved reference profiles reduce uncertainty, enabling more accurate detection of preferential flow outliers. Increasing the number of samples would take more time, but it is an efficient method for achieving a catchment-wide estimation of preferential flow compared to other approaches. While the number of profiles is important for constructing robust reference profiles, it is equally critical that they are representative of the catchment. In order to characterize soil matrix seasonality reliably, it is essential to capture the full range of isotopic variability. To this end, we took samples from across the full range of land use types, slope gradients and aspect orientations present in each catchment.

The presented method assumes that preferential flow signatures are similar to those of topsoil. This is very likely as many tracer sprinkling experiments e.g. have shown that preferential flow consists of new infiltrated water that also seeped into the topsoil matrix. However, this is not always the case. Preferential flow composition may differ from the topsoil signatures. For example, macropores can store older water that is remobilized and transported preferentially during infiltration events . Also topsoil hydrophobicity due to dryness or soil freezing funnels water into flowpaths without seeping into the topsoil matrix. In both cases, the signatures of preferential flow do not necessarily match those of the topsoil, and thus are not classified as preferential flow. Because sampling was not conducted during soil freezing condition and all soil samples were sufficiently moist for isotope analysis, the hydrophobicity effect can be dismissed.

The method allows us to observe the points where the preferential flow interacted with the soil matrix. A complex, interconnected preferential flowpath system may transport water through soils without interacting with the matrix . The signature of preferential flow would then be undetectable in the soil profile, as the water rushed through and out of the profile. When there is no interaction of preferential flow with the matrix, the method can not identify these flowpaths. As a result, if the preferential flow does not interact with the matrix, the method will not detect it, and the overall occurrence of preferential flow may be higher than our estimates. However, studies that reported the missing interaction found this only in a few cases, while identifying pore-matrix interactions was typical in most cases .

Aside from preferential flow, sampling and analysis errors can also affect the soil water isotope profile. Because the sampling cores are only one meter long, they must be removed after each meter of drilling. As a result, the borehole remained empty at depths of about 1 and 2 m for a short period during sampling. During this time, water-containing soil aggregates from the overlying soil layers may have fallen to the bottom of the borehole, altering the isotopic signature at those depths. To address this, we removed the top 5 cm of each soil core and excluded all identified flowpaths from our estimates where the sample matched the drilling cutoff. To reduce the impact of errors during sample analysis, we scaled the δ18O values to the δ2H range and only considered a sample as correct if both values exceeded the reference profile. This proved to be highly effective, as 103 profiles diverged in one isotope from the reference profile while the other was within the uncertainty band.

In order to effectively differentiate between preferential flowing water and slowly moving matrix water, the catchment must have measurable isotopic seasonality. For example, the alpine catchment exhibits much larger isotopic seasonality in precipitation (range 10 ‰ in δ18O and 100 ‰ in δ2H) compared to the other catchments (e.g. EG range 6 ‰ in δ18O and 60 ‰ in δ2H). This greater seasonal contrast in isotopic composition improves the ability to distinguish between different isotopic seasons. As a result, the pronounced isotopic variations in the alpine catchment provide clearer and more distinct signals, making it easier to identify and distinguish between preferential flowing water and slowly flowing water in the soil matrix. This level of differentiation is less noticeable in other catchments with lower isotopic seasonalities, where overlapping isotopic seasons can obscure the distinction between the two flow pathways or the two water ages. In catchments with low isotopic seasonality in precipitation, or where high groundwater levels or riparian zone mixing obscure seasonal signals in the soil profile, this method is not applicable.

Furthermore, the alpine catchment's soil properties may be better suited to preserve seasonal isotopic signals. In contrast, the other catchments showed mostly uniform profiles, indicating extensive mixing processes like lateral flow or advective and dispersive transport that obscure the seasonal signals . According to , upslope locations exhibit isotopic seasonality in their soil profiles, while further downslope, this signal is weakened due to mixing from lateral subsurface flow. Furthermore, the influence of well mixed groundwater with typically higher water ages, which increases at the base of the slopes in the riparian zone , erases the seasonality signal from the soils. Groundwater influence was also visible in the soil cores of the lower mountain ranges through redoximorphic patterns that were not present in the alpine catchment. This extensive mixing complicates the identification of the “real” seasonality profile in those regions, making it impossible to identify preferential flow clearly.

Although our sampling only provides a snapshot of the catchment, some of the PF profiles identified may represent features that are active year-round. In locations where only one PF profile was detected, it is possible that the surrounding soils lack similar characteristics. In future studies, multiple sampling at different times in the year could help to tackle this issue. However, hillslopes such as in the TA-fall (Fig. ) consistently exhibit PF signatures across all profiles. This widespread occurrence strongly suggests that lateral preferential flow is a persistent feature throughout this section of the hillslope and is likely to transport water across much of the hydrological year.

These findings are difficult to validate due to the lack of comparable quantitative data on preferential flow pathways in the study catchments. However, subsurface flow trenches on selected hillslopes provided direct evidence of preferential flow features within the soils . The observed decline in the occurrence of preferential flow with depth (Fig. ) is consistent with the depth-dependent patterns reported in other studies . This consistency across studies supports the plausibility of the results presented here.

The mixing models used make several assumptions and are subject to uncertainty. The reference profile represents the average value of the stored pre-event matrix water in the soil, rather than the real, local value, which is difficult to determine. The topsoil signature, on the other hand, is a true measured value, albeit subject to measurement uncertainties. These measurements were performed either manually or with VapAuSa, with the following uncertainties: VapAuSa δ2H=±4.5‰, δ18O=±0.58‰, and manual δ2H=±5.7‰, δ18O=±0.37‰. As a result, the mixing proportions are only estimates and can vary up to 20 % depending on the end member ranges. However, the end members of the reference profile and the topsoil are spaced farther apart than the measurement uncertainties, thus eliminating the possibility of no mixing.

The mixing model results revealed significant mixing at the depths where preferential flow was identified. This is consistent with previous research, which suggests that preferential flow is a mix of older, stored water and new event water. After a rain event, preferential flow interacts with the soil matrix , resulting in a mixed isotopic signature . Although we did not sample during rain events, frequent rainfall in the preceding days or hours supports the observed mixing.

As previously noted, it is challenging to draw conclusions from the relatively small number of profiles with identified preferential flow (only 16 % of the total), yet they still offer the potential for meaningful insights. We found no relationships between land use and the overall occurrence of preferential flow; however, we did observe a preference for shallower preferential flow in forested areas within the alpine catchment. Other studies have also identified land use-specific depth variations, but their findings were reversed: shallower preferential flow in grasslands and deeper pathways in forests . Other researcher indicated that land use significantly impacts the distribution and type of preferential flow paths, leading to varied discharge responses . It was concluded that shallower preferential flow paths lead to quicker discharge generation, whereas deeper pathways create a sponge-like behavior, quickly retaining water and then releasing it slowly .

Understanding the spatial and temporal variability of the isotopic composition of precipitation could aid in the analysis. The alpine catchment with the highest elevation gradient showed the largest uncertainty in the reference profiles (±25‰) whereas Sauerland with a low gradient only had an uncertainty of ±5‰. In catchments with a high elevation gradient, distributed isotopic precipitation data would allow for the quantification of the elevation effect. As this effect locally alters the temporal variability of the isotopic composition, scaling all profiles to one elevation could reduce the uncertainty of the reference profiles, allowing for higher precision to detect preferential flow. Furthermore, isotope precipitation data would allow for the construction of reference profiles using soil hydrological models. They could be used to derive an idealized reference pattern, assuming matrix flow is the only process involved . This model-based approach could potentially provide a clearer comparison against observed profiles.

The methodology used in this study was effective in detecting isotopic deviations that can be attributed to preferred flow processes. Notably, the alpine catchment exhibited the highest number of identifiable profiles. This is most likely due to the region's greater seasonality of stable isotopes in precipitation, which makes it easier to detect isotopic divergences. However, the analysis found no significant inter- or intra-catchment correlations with most topography variables or land use. The only significant influence on the occurrence of preferential flow was the aspect of hillslopes, with north-facing slopes having more preferential flow identified than other slopes. The soil coring approach used in this study provided a practical and cost-effective method of detecting preferential flow at catchment scale. Its minimally invasive nature and operational efficiency enabled extensive spatial coverage and the identification of vertical and lateral flow paths. However, the method has important limitations. It may underestimate preferential flow where bypass occurs without matrix interaction, or where isotopic contrasts are weak. The method's effectiveness also depends on sufficient sampling density, and it may be less effective in areas influenced by groundwater mixing or low topographic variability, where isotopic signals are less distinct. The study focused on soils in hillslopes, neglecting riparian zones, where lateral flow can induce mixing, affecting isotopic signatures. It would be valuable to extend this investigation to riparian or into flat areas, where vertical flow predominates, to understand how it might alter isotopic signatures differently. Additionally, incorporating other environmental tracers such as environmental DNA (eDNA) or dissolved organic carbon (DOC) in future studies could enhance the robustness and resolution of the analysis. This multidisciplinary approach could provide deeper insights into the dynamics of preferential flow and their underlying mechanisms.