European Journal of Forest Research

, Volume 132, Issue 1, pp 151–163

Greater abundance of Fagus sylvatica in coniferous flood protection forests due to climate change: impact of modified root densities on infiltration

  • Benjamin Lange
  • Peter F. Germann
  • Peter Lüscher
Open AccessOriginal Paper

DOI: 10.1007/s10342-012-0664-z

Cite this article as:
Lange, B., Germann, P.F. & Lüscher, P. Eur J Forest Res (2013) 132: 151. doi:10.1007/s10342-012-0664-z


Climate change is expected to modify the spatial distributions of zonal forest communities and thus, their species compositions. The aim of this paper was to study the impact of higher abundance of beech on water storage capacity in current coniferous flood protection forests due to varying root densities of the main tree species. Two forest communities in the northern pre-Alps in Switzerland with similar soil properties but varying in species composition were investigated (space-for-time substitution). It was assumed that the Vaccinio myrtillii-Abieti-Piceetum (site A) will be replaced by a Luzulo-Abieti-Fagetum (site B). We irrigated 16 hydromorphic soils (1 m2, 70 mm/h, three consecutive irrigations) at site A and 10 at site B and recorded water-content variations with time domain—and frequency domain reflectometry. Roots were extracted from soil cores taken from the positions where the water-content probes were inserted, and digitally measured. Infiltration capacity ωI was mainly limited to the upper soil at site A but was approximately constant down to 0.7 m depth at site B. Between 0.3 and 1.0 m soil depth, root densities at site B exceeded those at site A. Root density was the main predictor for ωI (R2 = 0.57) at site A as shown by a multiple linear regression analysis. Assuming that the root density in the current coniferous forest (A) will increase to that of the beech stand (B) due to the greater abundance of beech, the water storage capacity will increase by 9.2 mm in consequence of the expected forest transformation.


Climate changeFlood protection forestsWater storage capacityBeechSpruceForest transformation


Climate change is expected to increase the annual mean temperature in the northern part of Switzerland by approximately 2 °C by 2050 compared with 1990 (OcCC/ProClim-(ed.) 2007). As a consequence, the spatial distribution of zonal forest communities and thus the species compositions of these forests will be modified by temperature increase.

Approximately 36 % of the total forested area in Switzerland has an important protection function (Brändli 2010). Flood protection forests are likely to become even more necessary since floods are expected to occur more frequently due to climate change (OcCC/ProClim-(ed.) 2007). Such forests are in the catchment areas of the large Swiss rivers and are located in the montane and subalpine belts. Simulations indicate that the currently dominant Piceo-Abietion community in the northern pre-Alps in Switzerland will be replaced by Abieti-Fagion (Brzeziecki et al. 1995). Thus, the abundance of Fagus sylvatica will increase at the expense of Picea abies and, to a lesser extent, of Abies alba. This forest transformation may influence the efficiency of these flood protection forests.

Species change in forests impacts their hydrological properties directly by modifying their transpiration and interception rates. The interception rates of deciduous trees vary during the course of the year, while those of evergreen conifers remain nearly constant throughout all seasons (Christiansen et al. 2006). Furthermore, the transpiration rates of conifers and deciduous trees differ. Peck and Mayer (1996) analysed several studies about transpiration rates of trees and concluded that beech transpires on average 27 % more than Norway spruce, even though the transpiration of beech is mainly limited to the growing season. Hence, tree species significantly influence the soil water content and, consequently, how much water the soil can potentially store.

The transformation of forest composition also affects soil hydrological characteristics indirectly by modifying soil properties. The thickness of the forest floor layer declined during a Pinus sylvestris to F.sylvatica transformation in Germany, but the humus topsoil increased (Bens et al. 2007). Organic layers have high water storage capacities (Guevara-Escobar et al. 2007), but may also reduce infiltration capacity as they are water repellent (e.g. Doerr et al. 2000). Moreover, changes in tree species impact the rhizosphere since different species have different root systems and root densities, resulting in varying spatial patterns of water uptake, and thus drying out of the soils (Schwärzel et al. 2009).

Roots are important generators of preferential flow paths, which may improve the forest soil’s infiltration capacity substantially (Beven and Germann 1982). Noguchi et al. (1999) demonstrated that 70 % of the macropores in the topsoil and 55 % in the subsoil in a Japanese forest were related to roots. Lange et al. (2009, 2010) compared the root distributions in hydromorphic soils with the porosity that effectively carried preferential flow, and concluded that root density was related to infiltration capacity. Pores generated by woody roots can persist for decades, as has been shown in Denmark and Switzerland (Jorgensen et al. 2002; Hagedorn and Bundt 2002). Furthermore, saturated hydraulic conductivity ks can be improved by roots since Li and Ghodrati (1994) found that ks was five to six times higher in samples with root channels than in those without roots. Thus, roots seem to be one of the most important generators of well-connected pores in forest soils. Periodical anaerobic conditions in soils limit the root growth of certain species. Lehnardt and Brechtel (1980) found the roots of adult beech stands penetrated down to 0.6 m, whereas the maximum root depth of spruce was only 0.4 m in comparable hydromorphic soils, indicating that spruce is more liable to oxygen deficiency in the rhizosphere.

Pores generated by roots are of particular importance for infiltration in forest stands growing on poorly drainable forest soils (Lüscher and Zürcher 2003; Lange et al. 2009). A species shift in such stands may influence the soil’s infiltration capacity and, hence, the flood protective function. Even though it is widely accepted that roots are important for infiltration and that climate change will result in a forest transformation, the significance of these processes for flood protection has, as far as we know, never been explored. Thus, the main objective of this study was to investigate how an increasing abundance of beech in coniferous flood protection forest stands on hydromorphic soils in Switzerland will affect infiltration and water retention due to the resulting modified root densities and distributions. In particular, we aimed to:
  1. (i)

    compare the infiltration characteristics of two forest sites with comparable soil properties but varying abundances of beech and spruce;

  2. (ii)

    evaluate the relevance of roots and other soil properties for infiltration characteristics and

  3. (iii)

    model the modification of water storage capacity resulting from the greater abundance of beech in coniferous flood protection forests.


Materials and methods

Site and soil description

The experiments were conducted in the Gantrisch region in the Swiss pre-Alps approximately 30 km south of Berne at altitudes between 880 and 1,000 m a.s.l. Annual precipitation averages 1,700 mm and the mean annual temperature is 5.9 °C at 1,160 m a.s.l. (Zimmermann et al. 2006). The study sites are located in the Flysch zone and features mainly marled clays interlaced with stony or sandy layers. A space-for-time substitution was applied to evaluate the potential effects of greater abundance of beech in the flood protection forest, currently composed of spruce and fir. Two adult forest sites with comparable soil properties but varying in forest composition and altitude were investigated. The distance between the sites is approximately 700 m and the difference in the altitude ≈120 m. Site A defines the current forest stand and is classified as a Vaccinio myrtillii-Abieti-Piceetum (Ellenberg and Klötzli 1972), with Norway spruce (P. abies (L.) Karst.) as the most abundant species. Silver fir (A. alba Miller) and European beech (F. sylvatica L.) are secondary.

The prevailing soil types are Gleys, cambic Gleys, gleyic and stagnic Cambisols and Cambisols, according to FAO-UNESCO (1988). Beneath depths of 0.03–0.25 m, hydromorphic attributes such as iron and manganese concretions and mottles are found. The physical and chemical properties of the soil were determined from dried samples (48 h, 105 °C for bulk density and 60 °C for pH and texture). Three cylinders per horizon, with a volume of 1,000 cm3 and a height of 0.1 m, were collected to determine bulk density. pH was determined in a CaCl2 solution and the pipette method was applied for particle-size distribution separation. Considering all horizons investigated at site A, the percentage of sand was 18.6–71.6 %, silt content was 12.7–43.5 % and clay varied between 9.9 and 42.0 %. Bulk density was between 0.19 and 1.19 Mg m−3 in topsoils and 0.75 and 1.60 Mg m−3 in subsoils. The slopes of the investigated areas did not exceed 15°.

Climate change will enhance the abundance of beech in what are currently mainly coniferous forests in the montane belt (Brzeziecki et al. 1995). Thus, the second area investigated (site B) was a Luzulo-Abieti-Fagetum (Ellenberg and Klötzli 1972), which is what will presumably replace the current Vaccinio myrtillii-Abieti-Piceetum with global warming. This stand is characterised by more abundant European beech, while silver fir and especially Norway spruce are less dominant. Figure 1 shows a map of the investigated areas. The prevailing soil types are Gleys according to FAO-UNESCO (1988), and hydromorphic attributes are found beneath depths of 0.05–0.14 m. Soil texture varied between 17.5 and 73.1 % sand, 16.4 and 48.6 % silt and 8.6 and 34.6 % clay. Bulk density was 0.65–1.00 Mg m−3 in topsoils and 0.79–1.60 Mg m−3 in subsoils. The slopes of the areas investigated were less than 7°.
Fig. 1

Map of the investigated areas. Reproduced by permission of swisstopo (BA12053)

Table 1 summarises the characteristics of the study sites and Fig. 2 compares generalised soil profiles and average thicknesses of classes of soil genetic horizons. Both sites are characterised by an organic layer of 0.04–0.05 m above non-hydromorphic horizons with average thicknesses between 0.12 and 0.13 m, following by a layer with only few hydromorphic features up to approximately 0.3 m. Below, periodically water saturated horizons dominate. The soil textures at sites A and B were comparable, but varied considerably within the sites (Fig. 3). The bulk density was equal at both study sites in the top- and subsoils, while pH tended to be higher at site A (Fig. 4).
Table 1

Characteristics of the study sites


Study site



Altitude (m) a.s.l.



Slope °



Soil types

Gleys, cambic Gleys, gleyic or stagnic Cambisols, Cambisols


Forest community

Vaccinio myrtillii-Abieti-Piceetum


Main tree species

Norway spruce, silver fir

Silver fir, European beech

Number of irrigated plots


Fig. 2

Exemplary soil profiles and average depths of soil horizons at sites A and B
Fig. 3

Texture of all soil horizons at sites A and B
Fig. 4

Bulk density and pH (CaCl2) at sites A and B, shown for topsoil and subsoil separately

Sprinkling experiments

We conducted sprinkling experiments to measure variations in the volumetric water content θ(Z, t) as a function of time (t) at various depths (Z). At site A, the volumetric water content was measured with TDR equipment (time domain reflectometry). The wave guides consisted of two paired stainless steel rods, 0.15 m long, 30 mm apart and 5 mm in diameter. A 50-Ω coax cable connected the rods to an SDMX 50 coaxial multiplexer controlled by a CR10X micro-logger. A Campbell TDR100 device (Campbell Scientific, Logan, USA) generated the electrical pulses and received the signals. The transfer function of Roth et al. (1990) was applied to calculate the volumetric water content. At site B, soil moisture was recorded with FDR equipment (frequency domain reflectometry) from Decagon Device (Pullman, USA). We used 10HS soil moisture sensors (0.1 m long) and collected data with an Em50 data logger. Dielectric permittivity was converted to volumetric water content by applying the transfer function of Topp et al. (1980). Variations in water content were recorded at 60 s intervals at both sites. The wave guides were installed horizontally from a trench into each soil genetic horizon according to the soil profile descriptions. Thus, depth of water-content probes varied between the investigated plots. Each soil profile was equipped with five probes, where a soil was composed of less than five horizons, the thickest layer was equipped with two wave guides one below the other.

The sprinkler device consisted of an aluminium plate (1 m × 1 m) perforated with 100 holes in a 0.1 m × 0.1 m square pattern. The holes were attached to PVC tubes (inner diameters 2 mm) connected to a water reservoir with a constant water table. The plate was moved by a motor ±50 mm back and forth in both horizontal dimensions during the irrigation experiments so that it took approximately 1,800 s until a tube returned to exactly the same position. The intensity of irrigation was 70 mm h−1 and lasted 1 h. No visible surface run-off occurred during the experiments. A waterproof tent (3 m × 3 m) covered the set-up and the soil profiles during the experiments.

At each plot, we conducted three consecutive irrigation experiments in 23 h intervals. The first and second irrigations primarily served to saturate the finer pores in the soil where capillarity may be dominant. As shown by Lange et al. (2011) at the same location, water flow released by the third artificial rainfall event was mainly restricted to larger, fast-draining pores, where capillarity can be neglected. Thus, the principal focus of our interpretation was on the results of the third irrigation, where the pore-size spectra of pores carrying mobile water were comparable in all plots and the influence of the varying antecedent soil moisture level was negligible.

Overall, we recorded 371 time series of water-content variations from 16 plots at site A and 10 plots at site B. The plots were distributed over an area of approximately 10,000 m2 at site A and 2,000 m2 at site B and the distances of the irrigated areas to the closest tree was about 1 m (with the exceptions of three plots at site A where the distance to the stem was about 3.5 m). The depths of the probes were between 0.04 and 0.97 m. We conducted the infiltration experiments in fall 2006 and summer and fall 2007 (site A), and in summer and fall 2010 (site B).

Root measurements

Soil cores for the analyses of the root length distribution were taken subsequent to irrigation with a HUMAX soil corer (diameter 0.1 m) to depths between 0.5 and 1 m, at the same positions as those where the TDR and FDR probes were installed. The cores consisted of 0.25 m long segments in plastic tubes in which the soil was left undisturbed. The roots were sorted out and washed in a 1-mm sieve with tap water. All root fragments (woody and herbal roots) were collected and stored at 4 °C for no more than 12 weeks. Root length (cm) was measured with winRHIZO (V4.1c; Regent Instruments Inc., Quebec, Canada) for each morphological horizon separately.

Statistical analyses

To test distributions for different means, a rank-based test was applied (Mann–Whitney U test) since some data were not normally distributed. Statistical analyses were conducted with SYSTAT (V10; Cranes Software International, Bangalore, India).


Infiltration was analysed according to a Stokes-flow approach. A steady water input during the period Ts [s], beginning at t = 0 [s], to the soil surface initiates a water-content wave (wcw, see Fig. 5a). The balance between gravity g as driving force and kinematic viscosity η, which expresses momentum dissipation, is what determines flow. Capillarity is neglected. The wcw consists of thin water films moving between two neighbouring layers (solid or gaseous phase). A water film is defined with its thickness f [m], its length of contact Λ [m] in the horizontal plane and the vertical position of the wetting front zw(t) [m] at time t (Fig. 6). A water film’s specific contact length l [m−1] is the contact area at time t between mobile water and the solid soil phase per soil volume, which corresponds to the length of contact Λ per horizontal two-dimensional extent of the soil A [m2]:
$$ l = \frac{{\Uplambda \cdot z_{\text{w}} (t)}}{{A \cdot z_{\text{w}} (t)}} = \frac{\Uplambda }{A} $$
where l describes porosity which effectively carries mobile water. Single water streaks are too tiny to be measured with ordinarily applied instrumentation such as TDR equipment. Superposition of all films adds up to the entire wcw, which is measurable in situ. An entire wcw’s contact length L [m−1] per area A is:
$$ L = \sum\limits_{A} l $$
and its mobile water content per soil volume, which is the amplitude of infiltration ωI [m3 m−3], is given by:
$$ \omega_{I} = L \cdot F $$
where F is the water films’ average thickness. According to Germann and al Hagrey (2008), the wcw’s average wetting front velocity from the soil surface to depth Z [m], vW [m s−1] is given by:
$$ {{v_{\text{W}}}} = \frac{g}{3\eta }F^{2} = \frac{Z}{{t_{\text{W}} }} $$
where g [9.81 m s−2] is the acceleration due to gravity, η [10−6 m2 s−1] is the kinematic viscosity and tW [s] is the arrival time of the wetting front. Rearranging Eq. (4) leads to F:
$$ F = \left( {3\frac{Z\eta }{{t_{\text{W}} g}}} \right)^{0.5} $$
Contact length L can be determined by solving Eq. (3) for L:
$$ L = \frac{{\omega_{I} }}{F} $$
See Germann et al. (2007) for further details. Stokes flow is amenable to a certain range in the soil water content between Darcy’s saturated flow and a lower degree of saturation, where Richards’ capillary equilibrium requirement remains valid. Thus, the approach used is limited to a soil moisture range where mainly larger, fast-draining pores are involved in water flow and capillary forces can be ignored. According to Hincapie and Germann (2009), a rainfall intensity of 5 mm h−1 may present the lower limit for producing significant water-content waves. The volume flux density of our irrigation experiments was 70 mm h−1, and most likely resulted in wcws amenable to Stokes flow.
Fig. 5

a Definitions of parameters and variables of water-content waves (wcw). b Fitted increasing limb of a wcw. θin initial volumetric water content, θmax maximum volumetric water content, θF volumetric water content 75,000 s after the irrigation began, ωIθmax − θin, amplitude of infiltration, ωDθmax − θF, amplitude of drainage, ωFθF − θin, amount of non-draining water, tw arrival time of wetting front, tE end of irrigation at t = 3,600 s, tD arrival time of the drainage front
Fig. 6

Geometry of a water film flowing along a vertical plane. f film thickness, Λ length of contact between mobile water and the resting part of the soil in the horizontal plane; zW(t) vertical position of the wetting front as a function of time t

Analyses of wcws

The key points of a wcw, initial (θin) and maximal (θmax) water content were derived from the mean of ten water-content recordings between the beginning of the irrigation t0 and t0 − 540 [s] (θin) and between the arrival time of the drainage front tD and tD − 540 [s] (θmax). The final water content θF refers to the interval between 74,460 and 75,000 s (see Fig. 5a), where water content was constant during data recording, θmax was averaged over the last 540 s of the irrigation (3,060–3,600 s). A wcw is defined with its amplitudes of infiltration ωI (θmax − θin), which corresponds to the mobile water content (see Eqs. 3 and 6), and drainage ωD (θmax − θF) and the amount of non-draining water ωF (θF − θin, Fig. 5a).

The calculations of film thickness F and contact length L require an increase in volumetric water content due to irrigation since F is given by the velocity of the wetting front and the calculation of L is based on F and the amplitude of infiltration (Eqs. 5 and 6). We defined a threshold value of ωI = 0.01 m3 m−3 to separate non-reactive and reactive wcws, and only wcws with ωI ≥ 0.01 m3 m−3 were further analysed according to the Stokes-flow approach. In principle, tW can be derived from wcws with a maximum accuracy corresponding to the temporal resolution of the recordings. This may result in uncertainties if F is calculated according to Eq. (5) since the measurement intervals were 60 s. The approximately linear increasing limbs of the wcws were therefore fitted by a linear function θ(t), and the point in time when θ(t) = θin is the arrival time of the wetting front tW:
$$ t_{\text{W}} = \frac{{\theta_{\text{in}} - b}}{a} $$
where a is the slope [dθ dt−1] and b is the y intercept of the linear fitting function. Figure 5b) shows the fitted increasing limb of a wcw and the corresponding tW.

Knowing tW and depth Z, F follows from Eq. (5). Contact length L is determined according to Eq. (6), including F and the amplitude of infiltration ωI, which corresponds to the mobile water content.

Since F and L are calculated according to the maximal velocity of the wetting front (Eq. 5) between the soil surface and depth Z and on measured volumetric water content at Z (Eq. 6), L and F are also controlled by the wetting front’s advance and the water-carrying porosity above Z, which has to be considered when interpreting the water-flow geometry.


Water-content waves

Table 2 lists the number of recorded wcws and the arithmetic means of θin, ωI, ωD and ωF for sites A and B (see Fig. 5a), including the whole data set of recorded wcws independent of measurement depths. Initial water content was significantly lower at site B for all irrigations. The amplitudes of infiltration and drainage were lower at site A than at site B, but only ωI of the second irrigation differed significantly between the sites. Increasing the initial soil moisture reduced ωI. The most marked decline of ωI was found between the first and the second irrigation at site A, and between the second and the third irrigation at site B. For the first, second and third irrigation, 26.8, 36.7 and 40.5 %, respectively (site A), and 26.0, 24.5 and 32.6 %, respectively (site B) of all recorded wcws were non-responsive to the irrigation (meaning ωI < 0.01 m3 m−3). Thus, the proportion of non-responsive wcws increased with increasing antecedent soil moisture.
Table 2

Number of recorded wcws and arithmetic means of the water-content waves’ amplitudes


Recorded wcws






m3 m−3

m3 m−3

m3 m−3

m3 m−3































26.8 %

26.0 %


















36.7 %

24.5 %


















40.5 %

32.6 %


Irr irrigation, Ø arithmetic mean, A site A, B site B, θin initial water content, ωI amplitude of infiltration, ωD amplitude of drainage, ωF non-draining water

p < 0.05; ** p < 0.01; *** p < 0.0001; n.s. not significant

In italics: percentage of wcws with ωI < 0.01 m3 m−3

The distribution of ωI versus soil depth differed substantially between the two study areas (Fig. 7). At site A, ωI was clearly reduced at higher soil depths, while the averaged amplitudes of infiltration were approximately constant down to 0.7 m at site B. Thus, the infiltration capacity at site A was mainly limited to the upper soil horizons, while site B had higher ωI between 0.3 and 0.7 m depth, resulting in larger infiltration capacities in the entire soil.
Fig. 7

Averaged amplitudes of infiltration ωI at sites A and B for different soil depth ranges. n.s. not significant; *p < 0.05, ***p < 0.001

The required boundary conditions for water flow according to Eqs. (16) are: (i) an increase in the water content due to irrigation and (ii) gravity-driven flow instead of capillarity-dominated flow. As ωF indicates, the saturation of finer pores was pronounced during the first and second irrigation at site B, but only during the first irrigation at site A. Thus, the calculation of L and F was limited to wcws with ωI ≥ 0.01 m3 m−3 for the third irrigation (n = 47 at site A and n = 29 at site B). Including all wcws where film thicknesses and contact lengths were calculable, F and L did not significantly differ between the two sites (Mann–Whitney U test, p ≤ 0.05). Nevertheless, the averaged film thicknesses at site B (1.90 × 10−5 m) exceeded those at site A (1.58 × 10−5 m) by approximately 20 %. L at site A exceeded those at site B by 18 % on average (4,804 m−1 at sites A and 4,081 m−1 at site B), but the median of L was larger at site B (3,030 m−1 at site A and 3,702 m−1 at site B). These differences in water-flow geometry resulted in higher amplitudes of infiltration at site B. The cumulative frequencies of L, F and ωI are shown in Fig. 8.
Fig. 8

Cumulative frequencies of contact lengths L, film thicknesses F and amplitudes of infiltration ωI for the third irrigation at sites A (n = 47) and B (n = 29). Only water-content waves with ωI ≥ 0.01 m3 m−3 were considered

L and F at depth Z are also controlled by the wetting front’s advance above Z. Thus, only the relative progress of L and F versus soil depth was analysed. For this analysis, F and L were averaged for each depth range per site and divided by the arithmetic means of both sites’ F and L. Figure 9 summarises the progress of L and F for five depth categories (0–0.1, 0.1–0.3, 0.3–0.5, 0.5–0.7 and 0.7–1.0 m depth). F, and thus the velocity of the wetting front (Eq. 4), was approximately constant over the entire soil depth at site A, except in topsoil horizons where film thicknesses were smaller. F at site B increased versus soil depth, and the water accelerated. L was less beneath 0.1 m at site A and at Z > 0.5 m at site B. The higher average L at site A was due to few extensive contact lengths in topsoils since L at site A was less at Z > 0.1 m compared with site B. In summary, the decreasing ωI was due to the decreasing porosity carrying mobile water (L) at both sites, while film thicknesses were approximately constant (site A) or tended to increase with depth (site B).
Fig. 9

Relative film thicknesses and contact lengths for sites A and B versus soil depth. FA, FB, LA, LB arithmetic means of the film thicknesses F and contact lengths L of the specific soil depth range at sites A (FA, LA) and B (FB, LB), ØF and ØL arithmetic means of the film thicknesses and contact lengths at the specific soil depth range, including both sites A and B

Root densities

Including all soil depths, arithmetic means of root lengths per soil volume RL [cm cm−3] did not differ distinctly between the two sites (averaged RL = 0.529 cm cm−3 for site A and 0.504 cm cm−3 for site B). However, RL versus soil depth varied markedly between the investigated forest stands. In comparison with site B, site A had larger root densities in topsoils but RL decreased more pronouncedly with increasing soil depth. Below 0.1 m depth, RL was on average smaller at site A (0.238 cm cm−3) than at site B (0.332 cm cm−3). At a depth range of 0.3–0.5 and 0.7–1.0 m, RL of site B was significantly higher than that of site A. Figure 10 summarises RL versus soil depth for sites A and B, including a fitting function of RL versus soil depth.
Fig. 10

Root length densities (RL) and fitted RL versus soil depth for sites A and B. n.s. not significant. *p < 0.05, ***p < 0.001

Soil properties influencing the amplitude of infiltration

We applied a multiple linear regression for the third irrigation to ascertain the parameters determining the amplitude of infiltration ωI. The depth of the water-content probes, initial water content θin, root length density RL, bulk density BD and percentages of sand, silt and clay served as predictors. Some of the data were not normally distributed and square-root transformed to obtain approximately normal distributions (RL and depth at site A and RL at site B). Since some root data were not determinable at site A due to technical problems with the auger, the universe at site A was 67 wcws (out of 79) for the regression analyses, while all water-content recordings (n = 43) at site B were included. We applied a forward stepwise regression. Table 3 shows the resulting predictors and coefficients of determination for sites A and B. The first step of the regression led to the predictor which explains the maximum of ωI’s variance (RL at site A (R2 = 0.57) and θin at site B (R2 = 0.09)). Both relationships were significant at p < 0.05. The regression at site B could not be advanced by an additional predictor, whereas the regression coefficient at site A increased to R2 = 0.68 by integrating silt content and to R2 = 0.75 by including additional initial soil water content. More steps did not improve the regressions. Thus, root density mainly determined ωI at site A, while such soil properties as texture, bulk density and root length density were not sufficient to describe the amplitude of infiltration at site B since only θin influenced ωI significantly.
Table 3

Multiple linear regression analyses with ωI as the explanatory variable

Dependent variable

First step

Second step

Third step








Site A



RL, u


RL, u, θin


Site B



ωI Amplitude of infiltration, RL root length density, u percentage of silt, θin initial water content

Predicted amplitude of infiltration with greater abundance of beech

Since root density was the main determiner of the amplitude of infiltration at the present coniferous stand (site A), it is feasible to model the expected ωI at site A. RL versus soil depth was therefore fitted by second-order decay functions (Fig. 10), which are given by:
$$ {\text{RL}}_{\text{A}} (Z) = 0.0328 + 0.67514 \cdot {\text{e}}^{{\frac{ - Z}{0.20057}}} + 2.01966 \cdot {\text{e}}^{{\frac{ - Z}{0.07938}}} $$
$$ {\text{RL}}_{\text{B}} (Z) = - 0.06243 + 39.49142 \cdot {\text{e}}^{{\frac{ - Z}{0.02549}}} + 0.79725 \cdot {\text{e}}^{{\frac{ - Z}{0.54057}}} $$
where Z is the soil depth [m] and A and B refer to sites A and B. The coefficient of determination R2 was 0.75 for site A and 0.71 for site B. Based on the regression analyses shown in Table 3, the relationship between the present root density and the amplitude of infiltration at site A (ωI(pres)) is, considering square-root normalised data, given by:
$$ \omega_{{{\text{I}}({\text{pres}})}} (Z) = 0.08463 \cdot ({\text{RL}}_{\text{A}} (Z))^{\frac{1}{2}} - 0.00826 $$
We assume that the present root density at site A (RLA) will increase to that determined at site B (RLB) as beech becomes more abundant. Considering the predicted root density RLB at site A, the predicted amplitude of infiltration ωI(pred)(Z) is:
$$ \omega_{{{\text{I}}({\text{pred}})}} (Z) = 0.08463 \cdot ({\text{RL}}_{\text{B}} (Z)^{\frac{1}{2}} - 0.00826 $$
The averaged difference in ωI between ωI(pres) and ωI(pred), ωI(dif) from depths Z1 to Z2 is given by:
$$ \omega_{{{\text{I}}({\text{dif}})}} = \frac{{\int_{{Z_{1} }}^{{Z_{2} }} {\omega_{\text{I(pred)}} - \omega_{\text{I(pres)}} } }}{{Z_{2} - Z_{1} }} $$
and the additional water storage capacity due to the higher root density, sc(add), between Z1 and Z2 is:
$$ {\text{sc}}_{{ ( {\text{add)}}}} = \omega_{\text{I(dif)}} \cdot (Z_{2} - Z_{1} ) $$
Figure 11 shows the fitted root densities according to Fig. 10, and the present and predicted amplitudes of infiltration based on Eqs. (10) and (11) for soil depths between 0.3 and 1.0 m. Site B clearly had significantly higher root densities (with the exception of the depth range of 0.5–0.7 m, where p = 0.11). It can be expected that ωI would increase, on average, from the current 0.015–0.028 m3 m−3 at 0.3 m ≤ Z ≤ 1.0 m, resulting in additional water storage capacity of 9.2 mm.
Fig. 11

Present (RLA) and predicted root density (RLpred = RLB) according to Eqs. (8) and (9), and present and predicted amplitudes of infiltration ωI(pres) and ωI(pred) according to Eqs. (10) and (11) for site A

Figure 12 shows the predicted additional water storage capacity sc(add) according to Eq. (13) for different soil depth ranges. Between 0.3 and 0.7 m depth, sc(add) is relatively constant (1.48–1.68 mm per 0.1 m soil) and declines at Z > 0.7 m. Thus, an increase in infiltration capacity due to the greater abundance of beech is mainly restricted to soil depths between 0.3 and 0.7 m.
Fig. 12

Predicted additional water storage capacity sc(add) according to Eq. (13) for different soil depth ranges at site A

Discussion and conclusions

Infiltration properties

Averaged ωI at the beech site (B) exceeded those at the spruce site (A) by 0.005, 0.025 and 0.011 m3 m−3 for the first, second and third irrigation, respectively, due to the proportion of wcws with no increase in water content being lower at site B and larger ωI for water-content measurements with ωI ≥ 0.01 m3 m−3. This was probably because the initial soil water contents were significantly lower at site B than at site A (about 0.1 m3 m−3 for all irrigations). The total porosities of comparable hydromorphic soils measured in the Flysch zone were between 0.42 and 0.57 m3 m−3 (Richard and Lüscher 1987). Thus, averaged θin at site B was presumably at least 0.08 m3 m−3 below saturation since the mean of antecedent soil water content was at most 0.34 m3 m−3, even for the third irrigation. Nevertheless, the volumes of non-draining water within approximately 20 h, ωF, were close to zero for the third irrigation and did not differ significantly between sites A and B (Table 2). This observation suggests that mainly larger pores were involved in infiltration and the water-carrying pore-size spectra were comparable at both sites investigated for the third irrigation.

The bulk density and soil texture, and thus the total porosity, did not vary significantly between sites A and B. The fact that θin was lower for the third irrigation, although still with complete drainage, but ωI was larger, indicates that there were proportionally more larger, fast-draining pores at site B than at site A. This interpretation is supported by the contact lengths since at site B, L was larger at a depth range of 0.1–0.7 m. Consequently, potential infiltration was mainly restricted to topsoils at site A, while ωI was approximately constant down to a depth of 0.7 m at site B.

The interpretation of contact lengths and film thicknesses for the third irrigation gives additional information about water-flow geometry for wcws with ωI ≥ 0.01 m3 m−3 since the amplitude of infiltration is given by the product of L and F (Eq. 3). At site A, F, and accordingly the velocity of the wetting front (Eq. 4), was approximately constant between 0.1 and 1 m depth. Consequently, variations in ωI were due to differing L, as Hincapie and Germann (2009) also found. At site B, film thicknesses increased between 0.5 and 0.7 m and, according to Eq. (4), wetting fronts accelerated, while L decreased. This was possibly due to water films flowing together when porosity was reduced in deeper soil, resulting in thicker films and increasing vW (Eq. 4). This funnel effect requires well-connected pores. Since we did not detect such an acceleration of the wetting front at site A, we assume that the connectivity of vertical pores was less distinct at site A.

Root densities

Root densities were higher at site B, also in hydromorphic horizons. The forest sites investigated differed mainly in their abundances of beech and spruce since fir occurs in both forest communities. Accordingly, beech most likely had larger root densities, especially below depths of approximately 0.3 m. These findings are in accordance with those of Schmid and Kazda (2001), who observed that the relative number of spruce roots in mixed beech–spruce stands was lower than that of beech, and mainly concentrated in the upper soil. Furthermore, beech is less liable to temporarily anaerobic conditions in the rhizosphere than spruce (Lehnardt and Brechtel 1980). Soil acidity can substantially influence root growth and root distribution. Jentschke et al. (2001), for example, showed that fine root biomass in the mineral soil of Norway spruce forests was reduced in heavily acidified stands, but increased in organic layers. On the other hand, Finér et al. (2007) found no correlation between pH and root biomass in spruce and beech stands but demonstrated that root biomass in beech stands exceed that of spruce stands markedly. In this study, the root density was higher at the more acid site B. Consequently, the difference in root densities between the two investigated sites could have been even stronger if soil acidity would have been comparable. Thus, infiltration capacity at the current coniferous site could possibly increase more strongly than we suggested when beech will be more abundant due to climate change.

Soil properties influencing the amplitude of infiltration

At site A, the root length density was the most significant predictor of ωI, as many other studies that focused on the influence of roots on soil water flow have also found (e.g. Angers and Caron 1998; Perillo et al. 1999). It is then most likely that roots generated the porosity accessible to water. According to the Stokes-flow approach, the amplitude of infiltration ωI is given by the product of L and F (Eq. 3). Thus, for wcws with ωI ≥ 0.01 m3 m−3, where L and F were calculable, the root densities may be related to film thickness, contact length or both parameters. Since L was clearly related to RL (R2 = 0.43), but the coefficient of regression between F and RL was only 0.03, we draw the conclusion that the porosity carrying mobile water was indeed mainly generated by roots. Our investigations were conducted on a plot scale. This raises the question whether the impact of roots on soil hydrology can be assessed on larger scales in a similar way to the approach used in this study. The key issue is if the root densities measured at particular soil depths are the representative of the entire stand’s root densities. According to Brunner et al. (2004), the number of fine root segments is independent of their distance from the nearest tree in a forest community very similar to site A. Fine roots accounted for approximately 90 % of total root length density at site A, where ωI was determined by root densities. Thus, we can assume that the observed root densities based on 16 (site A) and 10 (site B) soil profiles represent the stand’s root density well enough.

The predictors applied did not enable ωI to be predicted well at site B. Larger pores, where water flow is gravity driven, can also be generated by physical processes such as thawing and freezing (Beven and Germann 1982) or shrinking and swelling (Romkens and Prasad 2006), or by biological factors such as earthworms (e.g. Lamande et al. 2003). Earthworms were not found during soil excavation at site B, possibly because of the soils low pH (Potthoff et al. 2008). Since bedrock, soil genesis, soil properties such as bulk density and texture, as well as climatic conditions, are similar at both sites, no significant differences in pore-generating processes besides roots are expected. The current root density does not necessarily represent total porosity generated by roots since decomposed roots are also able to generate pores accessible to water (Jorgensen et al. 2002). Thus, it may be possible that, even at site B, pores carrying mobile water were mainly generated by roots, but with the method of root measurement we used, decayed roots could not be detected.

Schwarzel et al. (2009) showed that the spatial distribution of soil–water variability in spruce and beech stands was related to not only soil properties but also to root distributions due to the water uptake of trees. The daily transpiration rates of beech stands can exceed 5 mm when water availability is sufficient (Marc and Robinson 2004), while spruce forests transpire at most 1.5–2 mm day−1 (Clausnitzer et al. 2011). Since the mobile water content was significantly influenced by the initial soil water content at site B and, to a lesser extent, also at site A, roots indirectly affect ωI by modifying the antecedent soil moisture due to water uptake. Thus, beech will reduce θin during the growing season more than spruce and will make additional water storage capacity available, which will decrease during winter and early spring when beech hardly transpires.

In this study, we focused on the significance of different species’ roots for gravity-driven water flow in soils, but the predicted transformation of forests can also affect the soil’s hydrological properties via other processes. These processes include the varying characteristics of organic layers (Greiffenhagen 2005) and the trees’ interception efficiency (Schume et al. 2003). The porosity generated by roots may also increase lateral water flow (Redding and Devito 2010) and therefore promote run-off generation (Uchida et al. 2005).

Change in water storage capacity

The greater abundance of beech in coniferous flood protection forests leads to increase in root densities and thus to an additional water storage capacity of 9–10 mm for soil depths between 0.3 and 1.0 m. Thus, the soil water storage capacity will be increased by approximately 15 % of a 1-h heavy precipitation with a recurrence interval of 100 years. Moreover, the initial water content will be reduced during the growing season because of the higher transpiration rates of beech, resulting in even more water storage capacity.

During forest transformation, beech and spruce will presumably coexist during a certain time period in the current coniferous stand. As shown by Schmid and Kazda (2002), fine root density in mixed beech–spruce forests growing on stagnic Cambisols was almost twice that of monospecific spruce stands. Thus, within a certain time period during forest transformation when beech initially grows in current coniferous forests, water storage capacity would possibly increase more than suggested by this study.

It is assumed, however, that climate change will lead to more frequent and intense heavy rainfall events (OcCC/ProClim-(ed.) 2007). Hence, even with greater infiltration capacity, the flood protection provided by a forest would not necessarily become more effective with the higher root densities associated with the predicted species shift, but the negative effects of more intense rainfall may be mitigated.

Hydrological models normally include vegetation by considering the transpiration rate which can be determined, for example, by root depth, leaf area index and potential evapotranspiration (e.g. Abbaspour et al. 2007). The results presented in this paper indicate that also modifications in porosity generated by roots substantially influence infiltration properties. Considering root distribution versus soil depth in water-flow models will not only enhance the assessment of transpiration rates but also the estimation of the significance of a modified pore system due to land use change or forest transformation. Thus, more detailed information about root distribution in different forest stands may increase the validity of such models considerably.

This study has begun to investigate the impact of a forest transformation on infiltration due to the different root densities of the main tree species. To assess the significance of more abundant beech in coniferous flood protection forests, future investigations should be more comprehensive. They should, for example, include further factors that affect infiltration such as transpiration, interception and organic layer, as well as the impacts of species transformation on the lateral subsurface flow, which may result in more peak flow generation.


We are grateful to Philipp Mösch and Dieter Müller for allowing us access to the sites. We thank Roger Köchli and Marco Walser for the help in the field and Silvia Dingwall for proofreading this manuscript. Two reviewers are acknowledged for their helpful and constructive comments which substantially improved the manuscript. This study was supported by the COST Action E38 (Woody Root Processes) and the research programme Forests and Climate Change (Swiss Federal Institute for Forest, Snow and Landscape Research WSL and the Federal Office for the Environment FOEN).

Open Access

This article is distributed under the terms of the Creative Commons Attribution License which permits any use, distribution, and reproduction in any medium, provided the original author(s) and the source are credited.

Copyright information

© The Author(s) 2012

Authors and Affiliations

  • Benjamin Lange
    • 1
    • 3
  • Peter F. Germann
    • 2
  • Peter Lüscher
    • 1
  1. 1.Swiss Federal Institute for ForestSnow and Landscape Research WSLBirmensdorfSwitzerland
  2. 2.Department of GeographyUniversity of BernBernSwitzerland
  3. 3.ZürichSwitzerland