Hydrological consequences of declining land use and elevated CO2 in alpine grassland

Authors


Correspondence author. E-mail: nicole.inauen@bluewin.ch

Summary

  1. Large areas of alpine pastures and meadows currently face declining land use or abandonment, which leads to tall-grass transition ecosystems with higher leaf area index (LAI), potentially increased evapotranspiration (ET) and thus, reduced water yield. Elevated atmospheric CO2, on the other hand, is known to reduce stomata opening and hence, leaf-level transpiration, which may translate into higher soil moisture and enhanced total runoff. Here, we quantify these opposing effects of global change on the water balance of alpine grassland in a field experiment in the Swiss Alps (2440 m a.s.l.).
  2. Rates of ET and deep seepage (percolation water) of four alpine grassland types (dominated by Agrostis, Nardus, Carex or forbs) were measured using intact monoliths in 51 weighing lysimeters. A part of the monoliths was clipped to simulate sheep grazing during three seasons (2008–2010). Another set was exposed to elevated CO2 (580 ppm) using free-air CO2 enrichment (FACE) during the 2009 growing season.
  3. Simulated grazing reduced bright day ET by on average −12% across all years, with the most pronounced effects in the high-stature swards. Correspondingly, the higher biomass and LAI in unclipped grassland lowered the seasonal sum of deep seepage by −13% in a drier summer (2009) and by −5% in a rather wet summer (2010) compared to clipped swards.
  4. CO2 enrichment reduced ET in all grassland types by −3 to −7%, increased δ18O in foliage and enhanced soil moisture, but not deep seepage. Hence, future CO2 slightly counteracts the land use effects at canopy level, however, not in terms of water yield.
  5. Synthesis. Our results indicate that both grazing and elevated CO2 are mitigating the effects of dry spells on alpine vegetation. The net effect of the continuous decline in the land use and of elevated CO2 is negative for catchment water yield and thus, for potential hydroelectric power production. Although these economic ‘costs’ are rather moderate per hectare of alpine grassland, sums are substantial when scaled to the vast areas potentially affected in the Alps. These calculated ‘costs’ attribute economic value to the eco-hydrological benefits of land care at these high elevations.

Introduction

Does the management of alpine grassland matter for the ecosystem water balance and catchment water yield? Large areas of mountainous high-elevation terrain are used agriculturally world-wide (mainly livestock grazing), some are maintained in a sustainable way, and others are either overgrazed or face abandonment. While overgrazing and associated soil erosion have received substantial attention (e.g. Morgan 2005), the hydrological consequences of declining land use and abandonment have not. A dense, tall leaf canopy is likely to transpire more, depletes soil water stores faster and yields less deep seepage (percolation water). The water yield of a catchment should thus become reduced and likewise the hydroelectric potential, when land use on alpine pastures and meadows is reduced or even abandoned. Hence, in addition to aspects of the preservation of the available agricultural area, of soil and biodiversity conservation (Tasser, Mader & Tappeiner 2003; Fischer et al. 2008), there are eco-hydrological consequences of alpine land abandonment that await quantification, given the large extent of the current transitions in land use.

Across the European Alps, often referred to as the water tower of Europe (Weingartner, Viviroli & Schädler 2007), 33% of all farms were abandoned between 1980 and 2000 (Streifeneder et al. 2007). As a result, large areas of poorly accessible alpine grassland are currently no longer used, and traditional, labour intensive practices of land management are being abandoned (e.g. decline in shepherding, substitution of meadows by pastures) with a concurrent intensification in land use on more easily accessible and more productive parcels (MacDonald et al. 2000; Tasser & Tappeiner 2002; Gellrich et al. 2007). In Switzerland, mountain pastures and meadows currently cover almost 28% of montane and alpine terrain (> 1000 m a.s.l.), representing 35% (537 800 ha) of the total Swiss agricultural land (FSO 2005), of which 60% are at risk of being abandoned in certain regions of the Alps in future (Gotsch, Flury & Rieder 2004). These abandoned grasslands either develop into forest and/or shrubland (74%; FSO 2005; Gellrich et al. 2007) or – at high elevation – end up as dwarf shrub heaths or tall swards with high leaf area index (LAI), necromass and plant litter production (Tappeiner & Cernusca 1989; Wohlfahrt et al. 2003; Merz et al. 2009). Because tall vegetation intercepts and also transpires more water, these land-cover changes are likely to affect evapotranspiration (ET) and the water balance (Körner, Wieser & Cernusca 1989; Asner et al. 2004; Tasser, Tappeiner & Cernusca 2005; Garcia-Ruiz & Lana-Renault 2011). Here, we quantify these catchment-wide eco-hydrological consequences of land use change, which may also help policy, landscape planners as well as stakeholders in a comprehensive economic assessment of sustainable management vs. land abandonment in alpine terrain.

Atmospheric CO2 has been shown to affect the water consumption of grassland as well and is likely to reach more than twice pre-industrial concentration by the end of this century (Meehl et al. 2007). Elevated concentrations of CO2 are known to reduce stomata opening and hence, leaf-level transpiration with water savings inevitably leading to temporally increased soil water content (SWC) in grassland (Morgan et al. 2004) and possibly translating into greater total runoff. The leaf-level effects are commonly diminished at ecosystem scale for reasons associated with aerodynamic coupling, rainfall patterns, soil water storage capacities or plant growth responses to elevated CO2 (Leuzinger & Körner 2010), and thus, are hard to predict.

In this study, we aim to quantify these partly opposing effects of reduced land use or abandonment and elevated atmospheric CO2 concentration on ET, soil moisture and growing season runoff in alpine grassland of the central Swiss Alps. We employed in situ lysimeters with intact monoliths of alpine grassland subjected to grazing simulation (clipping) and to free-air CO2 enrichment (FACE). This approach is novel by combining a high number of monoliths of different grassland types with FACE and applying mass balance lysimeters to assess the treatment effects on different time-scales and components of the water balance. Integrated over time, effects on seasonal sums of deep seepage (percolation water from soil surface to the subsoil and groundwater) and ET are quantified. We expect abandonment of grazing to reduce water yield in alpine grassland, and a future double pre-industrial CO2 atmosphere to increase it.

Materials and methods

Study site

In the Ursern Valley (central Swiss Alps), where the present study site is located (Furka Pass, 46°34′N 8°25′E, 2440 m a.s.l.), alpine grassland and dwarf shrub heaths cover 68% of the land surface area of the alpine zone, between 2100 m and 2700 m a.s.l. (T. van den Bergh, unpubl. data). This zone represents 58% of the upper catchment of the river Reuss down to the village of Andermatt (22 730 ha), a typical high-alpine catchment. Grassland dominated by Carex curvula, Nardus stricta (lower alpine zone) and variants of forb-rich turf cover vast areas of alpine terrain in the Alps including our study area. The grassland surrounding our study site is currently grazed at different intensities by small stationary herds of sheep and bigger migrating flocks during summer and autumn (on average every second year). Grassland occurs on relatively deeply weathered profiles of partly podzolized alpine brown earth on siliceous bedrock.

Precipitation generally increases with elevation in the Alps and annual sums average at c. 1900 mm at the study site (Atlas of Switzerland 3, Institute of Cartography, ETHZ). The growing season lasts between 2.5 and 3.5 months at this elevation with snowmelt in June and plant senescence in September. Meteorological conditions during the growing season were monitored by a weather station (Wireless Vantage Pro2 PlusTM; Davis Instruments, Hayward, CA, USA) set up at the study site at 1.5 m above-ground. Precipitation was additionally measured at 40 cm above-ground level by a rain gauge (RAINEW 111; RainWise Inc., Bar Harbor, ME, USA) and at 5 cm by two totalisators (funnel and bottle). Year-round climate data records are available from the meteorological station ‘Gütsch’ (2287 m a.s.l.; 17 km northeast) operated by the Swiss Federal Office of Meteorology and Climatology, MeteoSwiss. The 2009 growing season was warmer and drier than the other experimental seasons (2008 and 2010) with particularly low precipitation in August (see Table S1 in Supporting Information).

Methods of water balance measurement

We employed weighing lysimeters, set up in summer 2008, to solve the hydrological water balance equation in monoliths of alpine grassland vegetation (Fig. 1). In its simplest form, precipitation (P) equals the sum of ET (incl. interception), total runoff (R, the sum of surface runoff and deep seepage) and differences in SWC (∆S).

display math
Figure 1.

A diagram and photographs of weighing lysimeters at Furka Pass, Switzerland (2440 m a.s.l.). The bottom of the inner bucket of the weighing lysimeters is perforated to allow collecting percolation water in the reservoir of the outer bucket. The filter pad at the bottom of the inner bucket prevents soil particles from being washed out with seepage water. Monolith volume averages at 10 L. The top right picture shows a clipped grassland monolith with the weather station in the background, and the bottom right picture shows six unclipped grassland monoliths in a FACE ring (free-air CO2 enrichment).

The sum of evaporation from soil surface and transpiration by foliage (ET) was measured on a daily basis by weighing the lysimeters on bright days. Interception contributes only a small fraction to the total water balance of alpine grassland (c. 7% of growing season P; Körner, Wieser & Cernusca 1989). The seepage volume, collected in reservoirs below the lysimeters, is a measure of total percolation water (R), that is, deep seepage plus surface runoff, because the c. 2 cm collar at the upper edge of the lysimeters prevents lateral water flows into and out of the lysimeters. As soils are naturally saturated with water after snowmelt in spring and because of repeated saturating rain events during the growing season, the ∆S term (variation in volumetric SWC) of the ecosystem water balance usually becomes negligible on a seasonal basis. Thus, knowing the amount of seepage water, precipitation and ∆S over one season, seasonal ET can be calculated (including days with weather conditions unsuitable for assessing ET by weighing).

The lysimeters consisted of two plastic buckets of the same diameter (27.5 cm, surface area: 0.059 m2): a shallower (21 cm) inside a deeper one (30 cm; Fig. 1). The inner bucket was filled with a largely undisturbed alpine grassland monolith and had a perforated bottom, which permitted to collect seepage water in the outer bucket. The soil monoliths were all taken from the surrounding area (< 1 km) of the study site and similar elevation (2440–2480 m a.s.l.). They were shaped with a knife to precisely fit into the slightly conical buckets avoiding any air gaps between the monolith and the bucket wall. Skeleton-free soil volume averaged at 10 L with soil depths of c. 19 cm. Although rooting in alpine grassland is often deeper, c. 75% of all roots are confined to the top 10 cm, and c. 65%, to the uppermost 2.5 cm (Körner 2003). Given the rainfall regime in this area, the water stored in that profile depth prevented any moisture limitation, except for longer rainless periods, when monoliths received known amounts of water (as was considered necessary once during an exceptionally dry period in Aug 2009). Soil moisture sensors (EC-5, EC-10, EM50; Decagon Devices Inc., Pullman, WA, USA) were inserted into the root zone of each lysimeter at 5–10 cm below soil surface, and SWC was recorded hourly (plus seven probes outside the lysimeters served as controls). The lysimeters were sunk into the ground with the upper rim of the buckets slightly (1–2 cm) above the surrounding soil surface.

Daily ET was directly measured by weighing the grassland monoliths in the morning and evening hours of days without rain using a balance placed in an on-site shelter (BBK422-35LA; Mettler-Toledo Inc., Greifensee, Switzerland; precision ± 0.1 g). Monoliths were weighed at least once after sunrise at 6 a.m. and once after the sun disappeared from the study site between 6 and 7 p.m. depending on day length. On a few days, monolith weight was assessed every 2 h for diurnal courses of ET. Bright day ET was recorded for 19, 31 and 15 days in 2008, 2009 and 2010, between the end of June and mid September (except for 2008, when the experiment started in mid-July). Seepage volumes were determined after major rain events: around 10 times per growing season (mid-June–mid-September) and summed up to seasonal sums. Two heavy rain events caused overnight flooding of the lysimeters in 2010. In these two cases, the amount of seepage of each monolith was thus estimated using SWC before and after the rain events and the amount of precipitation. However, this calculation probably underestimated the effect of clipping on the 2010 seepage sum.

Land use experiment

For the land use experiment, we chose four sward types representing the most abundant vegetation in the study area. The different types of swards were dominated either by the grasses Agrostis schraderiana (sub-dominant species: Ligusticum mutellina; nomenclature following Lauber & Wagner 2007) or Nardus stricta (Ligusticum mutellina, Leontodon helveticus), by the sedge Carex curvula (Ligusticum mutellina, Leontodon helveticus) or by different forb species (Geum montanum, Trifolium alpinum, Ranunculus montanus, Potentilla aurea) to cover a wide range of vegetation structures (regarding vegetation height, density or the proportion of horizontal vs. vertical structures) and of plant biomass or LAI (Table 1). Agrostis schraderiana-dominated swards, although not as abundant as the other grassland types, were chosen to mirror the structure of grassland of the lower alpine belt under decreasing land use intensity, leading to tall-grass canopies. In the following, we address the grassland types by the abbreviations As for A. schraderiana, Ns for N. stricta, Cc for C. curvula and forb. We set up six monoliths of each of the four grassland types (24 in total), of which half (= 3) were clipped (to a vegetation height of 3–4 cm) at the peak of standing biomass (24 July 2008, 25 July 2009, 2 August 2010) as is the typical grazing by sheep (the common land use practice in the studied alpine grassland). Vegetation was kept short during the remaining weeks of the season. Our treatment is an approximation of sheep grazing, which is, by our experience in this area, more uniform than cattle grazing at lower elevation. Although the species N. stricta is avoided by most grazers as long as other fodder is abundant, we included vegetation with N. stricta, which is more abundant in the lower alpine zone, for the sake of completeness. The LAI of such swards is also reduced by grazing through the selective removal of all other species than N. stricta, thus, not what we could simulate. Clipped phytomass was separated into necromass (attached, dead material) and biomass of different plant functional types (grasses, sedges, forbs, dwarf shrubs), dried at 80 °C and weighed. Leaf area was measured using the LI-3000C Portable Area Meter in combination with the LI-3050C Transparent Belt Conveyer Accessory (LI-COR Biosciences Inc., Lincoln, NE, USA). Maximum gross canopy height (excluding inflorescences) was recorded at peak season in each experimental year (before clipping).

Table 1. Characterization of the grassland types used in the land use experiment: plant dry matter partitioning, cryptogam mass, leaf area index (LAI; both assessed in autumn 2010, including phytomass removed by clipping) and maximum gross canopy height prior to clipping (means ± SD;= 6 lysimeters)
  Agrostis schraderiana Nardus stricta Carex curvula ForbP-values
  1. P-values show results of anova, bold values are statistically significant (P < 0.05); means with different letters differ significantly based on Tukey's test (< 0.05).

Plant mass partitioning (g m−2)
Biomass229 ± 44a208 ± 91a102 ± 33b246 ± 47a < 0.001
Necromass8 ± 3a40 ± 16b24 ± 9c14 ± 9d < 0.001
Litter136 ± 41237 ± 120138 ± 49137 ± 50n.s.
Cryptogams19 ± 18a86 ± 89ab87 ± 34b90 ± 94ab 0.040
LAI (m2 m−2)2.6 ± 0.5a1.2 ± 0.5b0.8 ± 0.3b2.5 ± 0.5a < 0.001
Max. canopy height (cm)11.8 ± 3.5a8.0 ± 1.8b5.8 ± 1.7b7.0 ± 2.6b 0.003

CO2 enrichment experiment

To explore ET responses to elevated CO2, we set up a second series of identically constructed weighing lysimeters in summer 2008 (12 N. stricta, 12 C. curvula and 12 forb-dominated) and exposed half of them to Free-Air CO2 Enrichment (MiniFACE). Elevated CO2 was provided by three 1.3-m diameter FACE rings, delivering pure CO2 (Inauen, Körner & Hiltbrunner 2012), from 1 July to 9 September 2009. The target CO2 concentration was 580 ppm, and CO2 concentrations were recorded and controlled every second, for each FACE ring separately. FACE was switched off during 10 single days to obtain daily ET signals without CO2 enrichment for standardization, because the CO2 effect on ET was expected to be small with considerable variance among monoliths. The monoliths were randomly assigned to CO2 treatments, and two lysimeters of each grassland type were placed in each of the three FACE rings, adding up to a total of six lysimeters per ring (Fig. 1). For all monoliths in the CO2 experiment, we randomized positions within treatment groups (ambient or elevated CO2) every 2 weeks, which allowed us testing each lysimeter as a spatially uncorrelated replicate (= 6). All monoliths in the CO2 experiment remained unclipped. The unclipped control monoliths of the land use experiment (three of each grassland type Ns, Cc and forb) also served as controls for the CO2 experiment; hence, the total number of monoliths across both experiments was six As, 15 Ns, 15 Cc and 15 forb (51 lysimeters in total).

For the analysis of the ratio of heavy (18O) to light (16O) stable oxygen isotope concentration in leaf tissue under ambient and elevated CO2 (expressed as δ18O, relative to the international standard VSMOW), we collected leaf samples of the most common species in all grassland types used in monoliths on 20 August 2009 (six graminoids and six forbs; for species names see Table 4). In order not to disturb lysimeters and given the same soil moisture readings inside and outside monoliths, leaf samples were taken immediately next to the monoliths from the native grassland growing inside and outside FACE rings. For each species, leaf samples (current season green foliage only) of several individuals were pooled (one pooled sample per FACE ring and three pooled samples taken from control plots outside FACE rings, = 3). The samples were dried for 48 h at 80 °C, ground to fine powder, put in tin capsules (0.5–0.7 mg) and after pyrolysis analysed on a continuous flow isotope ratio mass spectrometer (EA-1108; Carlo Erba Thermoquest, Milan, Italy; CONFLO II and DELTA PLUS XP; Thermo Finnigan, Bremen, Germany; precision of δ18O analyses ± 0.3‰).

After three seasons of operation (September 2010), above-ground plant material of all monoliths was harvested down to ground level in steps of 4 cm and treated as described above (categories: phytomass, cryptogams and litter, i.e., detached, dead material). The soil monoliths were subsequently oven-dried at 100 °C. Actual volumetric SWC (mm) of each monolith was then calculated from dry weight for any weighing date during the experimental period. The highest SWC after 1 day of drainage of saturated soils was assumed to represent the water content at ‘field capacity’ (θfc). For monoliths, this θfc is an approximation and does not necessarily reflect θfc of soil outside lysimeters due to the interruption of the capillary continuum by the lysimeters, but served to standardize soil moisture data measured by soil moisture sensors. To explore the relative differences in SWC among grassland types and treatments, all SWC data were set to the same initial SWC (across all sensors) on the first weighing occasion of each growing season.

Statistical analyses

The statistical analyses of the results of the land use experiment and the CO2 experiment were performed separately using the open-source software r, version 2.13.2 (R Development Core Team 2011) with the package ‘nlme’ (Pinheiro et al. 2011). In the land use experiment, mean daily ET rates (averaged across all three seasons), sums of seasonal runoff and mean SWCs (across the period after clipping) were analysed using analysis of variance (anova) with the factors ‘grassland type’ and ‘clipping’. In all models, residuals were tested visually for normality and homogeneity of variances. In case of non-normality and inhomogeneity, we applied power or log transformations to the response variables, or we fit a constant variance function using ‘gls’ (generalized least squares fit by ‘REML’). All models were subsequently reduced to the minimum adequate model by removing non-significant factors or interactions (not shown in tables of results of statistical tests) and merging grassland types that did not differ significantly in their response. Model selection was performed by comparing nested models, using log-likelihood ratio tests. For a posteriori comparison, we applied Tukey's HSD to test multiple group means. CO2 effects on mean daily ET rates, mean SWC (both across the 2009 season) and the seasonal sums of deep seepage were analysed, employing models with the factors ‘grassland type’, ‘CO2 treatment’ and, for ET only, the covariate ETCO2off, that is, the mean ET rates of each monolith across all days when FACE was switched off. In all analyses, P-values < 0.05 were considered statistically significant and P-values < 0.1 marginally significant.

Results

Effects of land use change on the water balance

Mean total biomass and LAI in monoliths employed in the land use experiment were highest in A. schraderiana (As) and the forb-dominated communities (forb), slightly lower in N. stricta (Ns) and lowest in C. curvula swards (Cc) by the end of the experiment (autumn 2010; Table 1). In the graminoid-dominated swards, graminoid cover averaged at 62 ± 11% (largely vertical structures), and in the forb-dominated swards, forb cover was 69 ± 7% (largely horizontal structures; means ± SD). The fraction of necromass in monoliths ranged from 4% (As) to 24% (Cc) of phytomass. The amount of plant litter plus cryptogams (lichens and mosses) surpassed the amount of phytomass in grassland types dominated by N. stricta and C. curvula. We found a statistically significant, positive linear correlation between the amount of biomass and mid-season ET (July) of these monoliths prior to the clipping treatment (Fig. 2). However, the grassland types behaved clearly differently (F3,46 = 15.5, < 0.001) with A. schraderiana showing higher ET at similar amounts of biomass than the other, lower stature grassland types. Better atmospheric coupling of the taller stands is a likely explanation. Bright day courses of ET closely followed the course of vapour pressure deficit, with a peak between 11 a.m. and 1 p.m. (data not shown). ET was not reduced under dry weather conditions, even when monoliths reached the lowest SWCs that were measured during the three test seasons, namely 62 ± 8% of the water content at ‘field capacity’ (θfc: 103 ± 9 mm or 54 ± 5% vol across all monoliths; means ± SD). Hence, soil moisture never constrained ET at this elevation.

Figure 2.

Bright day rates of evapotranspiration (ET) of four alpine grassland types before clipping (July 2010) in relation to above-ground biomass (harvest in September 2010; dashed line: regression across all grassland types). Regressions by grassland type As, Agrostis schraderiana R2 = 0.24, n.s.; Ns Nardus stricta R2 = 0.32, = 0.016; Cc, Carex curvula R2 = 0.28, = 0.024; Fo forb-dominated R2 = 0.33, = 0.014).

Depending on year, the clipping treatment reduced mean bright day ET by −21 to −32% in the high-stature A. schraderiana, by −8 to −14% in N. stricta and least (+2 to −8%) in the lowest stature swards (sedge- or forb-dominated; Fig. 3, Table 2a). On average, this reduction by clipping amounted for −12% of daily ET across all years and all grassland types. The difference in the clipping effect between grassland types corresponded to the amount of biomass removed by clipping (As: 47%, Ns: 21%, Cc: 16%, forb: 18%). In general, clipping effects were highest during the first few weeks after clipping and diminished towards the end of the growing season, although the vegetation was kept short during the whole growing season after first clipping. Similarly, daily ET rates decreased by late season due to plant senescence and lower temperatures and lower daily irradiance sums (Table S1).

Table 2. The effect of grassland type (GT), simulated grazing (clipping) and their interaction on (a) mean daily evapotranspiration (ET; across all years, 2008–2010), (b) volumetric soil water content, SWC (averaged across all dates in 2009) and (c) seasonal sums of deep seepage
 d.f. F P-values
  1. Bold P-values are statistically significant (P < 0.05).

(a) Mean daily ET
GT1, 204.4 0.049
Clipping1, 2016.9 < 0.001
GT × Clipping1, 2015.0 < 0.001
(b) Mean SWC
GT2, 183.6 0.047
Clipping1, 1816.7 < 0.001
GT × Clipping2, 183.10.070
(c) Deep seepage sum
2008
GT2, 185.3 0.015
Clipping1, 1839.2 < 0.001
GT × Clipping2, 1817.0 < 0.001
2009
GT2, 183.7 0.046
Clipping1, 1827.1 < 0.001
GT × Clipping2, 185.9 0.011
2010
GT1, 208.7 0.008
Clipping1, 2014.7 0.001
GT × Clipping1, 203.50.077
Figure 3.

Mean daily evapotranspiration (ET) rates (means ± SE,= 3 weighing lysimeters) of unclipped (white bars) and clipped (grey bars) alpine grassland swards across the years 2008–2010 (period after first clipping). Bars with different letters are significantly different.

The reduced water vapour loss due to clipping increased SWC in monoliths already a few days after the clipping treatment was applied (shown for 2009, the season when the monoliths were weighed most frequently; Fig. 4, top panels, Table 2b). These water savings were pronounced in A. schraderiana and N. stricta, but rather small in the small-stature C. curvula and forb-dominated lysimeters. The maximum differences in SWC due to clipping were reached during the warm and dry period in autumn 2009 and averaged at +24 mm (As), +13 mm (Ns), +4 mm (Cc and forb). During drying cycles, soil moisture of clipped monoliths decreased more slowly and remained at higher levels than unclipped ones (Fig. 4, bottom panel). Heavy rain events equalized SWC between clipped and unclipped lysimeters.

Figure 4.

Top panels: volumetric soil water content (SWC) of unclipped and clipped lysimeters (means ± SE,= 3; Carex curvula and forb pooled) on 22 weighing days after clipping on day 206 in 2009. Bottom panel: SWC measured by soil moisture sensors in unclipped (dashed line) and clipped (solid line) monoliths across all grassland types (means ± SE,= 4, SE: grey shaded area, standard error across grassland types) and precipitation (black bars; 14 mm of water added on day 230, grey bar). θfc: 103 ± 9 mm or 54 ± 5% vol across all monoliths (means ± SD).

The amount of deep seepage water collected in lysimeters (equals total ecosystem water loss by runoff) provides a temporally integrated signal of the effects of vegetation cover or clipping on the seasonal water balance. Accordingly, clipping reduced ET and as a result led to higher seepage sums during the time after clipping, especially in the higher stature As and Ns (Fig. 5, Table 2c). We found significantly reduced total seepage, but increased absolute and relative clipping effects in the warm and dry 2009 season compared to the relatively wet seasons 2008 and 2010 (meteorological conditions see Table S1). However, the effect of clipping on the 2010 seepage sum is likely to be underestimated (due to overnight flooding of the lysimeters). Across the full growing season (including the time before clipping; not available for 2008 due to installation), the clipping treatment significantly increased deep seepage by +5 to +38% in 2009, depending on grassland type, (F1,18 = 8.3, = 0.010) and by only +4 to +8% in the rather wet 2010 season (F1,18 = 9.0, = 0.008). In other words, declining land use or complete land abandonment (tall grass) reduced seepage by −13% (−35 mm) over the whole comparatively dry growing season in 2009 and by −5% (−22 mm) in 2010 (a relatively wet season).

Figure 5.

The effect of clipping on the seasonal sum of deep seepage (means ± SE,= 3; white bars: unclipped, grey bars: clipped) of four grassland types, measured after clipping (42 days in 2008; 55 days in 2009; 43 days in 2010, precipitation sums are given in brackets). Bars with different letters are significantly different within the same season. Note the relatively bigger signals in the warmer and drier 2009 season.

In line with the higher plant water use in unclipped compared to clipped swards, mean bright day ET increased significantly with increasing canopy height (in unclipped monoliths), which is expected to occur with extensification or after abandonment. Daily ET was +31% higher in the highest stature A. schraderiana than in the lowest stature C. curvula swards across all three growing seasons (2008: after mid-July; 2009/2010: full growing seasons). These increased ET rates in A. schraderiana depleted soil water stores faster and caused the seasonal seepage sums to be significantly smaller than in C. curvula-dominated grassland: by −33 mm (−15%) in 2008 (after mid-July), −82 mm (−30%) in 2009 and −42 mm (−10%) in 2010 (full growing seasons). Hence, seasonal water yield was substantially reduced with increasing vegetation height.

From the seepage sums, precipitation and changes in SWC, we calculated evaporative water losses (ET) for unclipped monoliths during the full growing seasons from snowmelt to the last seepage measurement at the end of the 2009 and 2010 growing seasons: ET sums averaged at 288 mm (60% of total ecosystem water losses) in A. schraderiana (the highest stature swards) and 196 mm (42%) in C. curvula (the lowest stature swards) in the warm and dry 2009 season, and at 220 mm (37%) in A. schraderiana and 173 mm (29%) in C. curvula in the cool and wet 2010 season. Mean daily ET rates across all grassland types (unclipped monoliths) and the full growing season were thus 2.4 mm in 2009 (100 days) and 2.0 mm in 2010 (96 days).

Effects of elevated CO2 on the water balance

FACE was operating on 59 of 71 days between 1 July and 9 September 2009. On 10 days in total, FACE was turned off to obtain ET signals without CO2 enrichment, and on another 2 days, FACE was interrupted because of snow cover. Across all enrichment days, CO2 concentration was very close to the target level in two rings (584 ± 10 ppm, 592 ± 14 ppm) and slightly higher in one ring 634 ± 20 ppm. Of all 1-s values 88%, 75% and 52% were within ± 20% of the target CO2 concentration of 580 ppm in the three rings. Potential influences of the differences between FACE ring CO2 concentrations have been avoided by regular randomization of the lysimeters between FACE rings.

CO2 enrichment reduced daily ET rates consistently, in all three grassland types studied in the CO2 experiment and during the whole growing season by on average –3.6% in N. stricta, −5.7% in C. curvula and −8.7% in forb-dominated monoliths (F1,34 = 4.2, = 0.048; Fig. 6). However, there was considerable residual variance, which was ascribed to single monolith properties that affect ET (e.g. biomass, vegetation structure). When the mean daily ET rates (across growing season) were standardized using the covariate ETCO2off, the ET reduction by elevated CO2 was −6.6% in N. stricta, −3.2% in C. curvula and −6.0% in forbs and significantly different between grassland types (Table 3a). In forb-dominated monoliths subjected to elevated CO2, the ET rates on days when FACE was switched off became steadily reduced with the progression of the growing season relative to the controls grown at ambient conditions (not in other grassland types). This could have been caused by slight seasonal reductions in LAI under elevated CO2 (in forb only; n.s.; data not shown). However, the CO2 effect on ET standardized by ETCO2off was significant irrespective of any possible biomass changes. Hence, across all grassland types, elevated CO2 clearly reduced ET by on average −5.3%. This reduction in ET was also supported by a significantly higher stable oxygen isotope ratio (δ18O) in organic leaf material in seven of 12 tested species grown under elevated (580 ppm) compared to plants grown at ambient CO2 (Tables 3b and 4). δ18O in leaves as well as the CO2 effect on δ18O were highly species-specific.

Table 3. The effect of elevated CO2 on (a) average daily evapotranspiration (ET, across all dates in 2009), (b) δ18O in foliage, (c) volumetric soil water content (VWC, averaged across all dates in 2009) and (d) seasonal sums of deep seepage in the three different grassland types (GT)
 d.f. F P-values
  1. Bold P-values are statistically significant (P < 0.05).

(a) Mean daily ET
ETCO2off (covariate)1, 299758.2 < 0.001
GT2, 2964.9 < 0.001
CO21, 29150.4 < 0.001
GT × CO22, 2911.1 < 0.001
(b) Leaf material δ18O
Species11, 4914.3 < 0.001
CO21, 411.2 0.029
Species × CO211, 492.7 0.009
(c) Mean VWC
GT1, 3316.8 < 0.001
CO21, 3323.4 < 0.001
GT × CO2  n.s.
(d) Deep seepage sum
GT  n.s.
CO2  n.s.
GT × CO2  n.s.
Table 4. δ18O in leaf tissue under ambient and elevated CO2 (nine grassland species grown inside and outside FACE rings; means ± SE,= 3 rings)
Speciesδ18Oambient (‰)δ18Oelevated (‰)P-values
  1. Bold P-values are statistically significant (P < 0.05).

Graminoid species
 Agrostis schraderiana 23.70 ± 0.7024.17 ± 0.680.662
 Anthoxanthum odoratum 24.16 ± 0.2225.29 ± 0.380.061
 Carex curvula 22.42 ± 0.5923.51 ± 0.310.137
 Helictotrichon versicolor 24.14 ± 0.3025.30 ± 0.06 0.019
 Nardus stricta 22.77 ± 0.1323.99 ± 0.400.053
 Poa alpina 25.44 ± 0.2925.39 ± 0.540.931
Forb species
 Geum montanum 23.29 ± 0.2823.32 ± 0.460.946
 Homogyne alpina 22.01 ± 0.3724.48 ± 0.50 0.016
 Leontodon helveticus 22.73 ± 0.0923.08 ± 0.03 0.024
 Ligusticum mutellina 22.84 ± 0.0822.48 ± 0.350.292
 Potentilla aurea 23.49 ± 0.1025.44 ± 0.44 0.013
 Trifolium alpinum 22.08 ± 0.2323.46 ± 0.37 0.033
Figure 6.

The effect of elevated CO2 on daily evapotranspiration (ET) rates (white triangles: ambient CO2, black triangles: elevated CO2) and deep seepage (grey bars: ambient CO2, black bars: elevated CO2, right y-axes) of three grassland types in 2009 (means ± SE,= 6). The inset bar charts show mean daily ET rates (averaged across the growing season) under ambient (white) and elevated CO2 (black bars).

Already a few days after the start of CO2 enrichment, the reduced ET rates under elevated CO2 resulted in higher SWC than in control monoliths in all grassland types, a difference that was perpetuated across the remaining part of the 2009 growing season (Table 3c, Fig. S1). The maximum increase in SWC under CO2 enrichment was 5.6 mm (+7.0%) in N. stricta, 4.7 mm (+6.5%) in C. curvula and 8.0 mm (+11.0%) in the forb-dominated monoliths during the rather dry period in autumn 2009. The water savings due to higher atmospheric CO2 translated into slightly higher seasonal seepage sums in the forb-dominated grassland only (Fig. 6, Table 3d; forb-dominated: +18 mm; F1,10 = 3.6, = 0.086).

Discussion

This 3-year field study provided clear evidence for reduced growing season water yield when grazing in alpine grassland is reduced or fully abandoned. Unexpectedly, simulated atmospheric CO2 enrichment did not affect the water yield significantly, even though we measured small, but significant reductions in ET of the different grassland canopies.

Water balance of alpine grassland

The ET rates presented here (2.0–2.4 mm per day and 173–288 mm per season) correspond to ET rates at similar elevation (2000–2550 m a.s.l.) collected in several lysimeter studies in the Austrian Alps, where seasonal mean ET of alpine grassland was 2.1 mm day−1 and 165–255 mm per growing season (summarized by Wieser, Hammerle & Wohlfahrt 2008). In line with our observations, there were little, if any, indications of water stress on plant transpiration in the Austrian Alps, despite reductions in soil water availability in combination with high evaporative demand, especially during the summer drought of 2003 (Brilli et al. 2011).

Effects of land use change on ET and the water balance

Bright day ET rates were reduced by −12% across all years and grasslands types when grassland was kept short during the second half of the growing season (by simulated grazing), matching signals obtained in a small pilot study with Carex curvula in the Austrian Alps (−12% ET; Körner, Wieser & Cernusca 1989). The lower ET led to increased seasonal deep seepage in short grassland. The additional water yield, resulting from naturally small stature and/or clipping, was higher in the rather warm and dry 2009 season compared to the cooler and wetter growing seasons in 2008 and 2010. In other words, sustainable land use (i.e. non-destructive land use that maintains livestock carrying capacity) in this area produces higher water savings during dry summers, because soil moisture differences accumulate over longer periods.

The average, clipping-induced reduction in ET and the corresponding increase in water yield across all 24 monoliths used in the land use experiment are attributing equal weight to the tall (A. schraderiana) and the three short (N. stricta, C. curvula and forb-dominated) grassland types. In reality, the abundance of tall grass is somewhat lower in the immediate surroundings of the study site, but much higher at only 100–200 m lower elevation in the lower alpine belt. Furthermore, the sedge-dominated monoliths employed were less lush (40% lower biomass) than average Carex curvula-dominated swards examined elsewhere (Klug-Pümpel 1982; Körner et al. 1997), likely to cause an underestimation of the clipping effect in this highly abundant alpine grassland type. Because the clipping treatment was only applied by mid-season, in line with the recommended timing of sustainable sheep grazing in the region, it influenced the second half of the growing season only, as is apparent from the reduction of the treatment effect across the whole snow-free season compared to the period after clipping only. Overall, the −5% (−22 mm) to −13% (−35 mm) reduction in the seasonal sum of deep seepage by increased biomass, presented in this study, reflects a reasonable estimate of the potential, landscape-wide effects of land use reduction in alpine grassland. At lower elevation, in the lower alpine or upper montane belts, where grassland canopies are generally much higher than at the study site, effects of grazing are expected to be in the range observed here for A. schraderiana, that is, arrive at reductions in ET of −25% and seasonal seepage gains of +8% (+30 mm) to +38% (+74 mm) for the two complete seasons (wet vs. dry). Should future rates of atmospheric nitrogen deposition (currently 3.3–5 kg N ha−1 a−1 wet deposition in the studied region; Hiltbrunner, Schwikowski & Körner 2005) exceed the ‘critical load’ for such alpine systems (5–10 kg N ha−1; Bobbink et al. 2010), this may change vegetation in favour of more vigorous species, likely to enhance the land abandonment effect on ET, as shown here.

Effects of elevated CO2 on ET and the water balance

At ecosystem scale, the widely reported reduction in stomatal conductance and thus, transpiration in grassland exposed to elevated CO2 are diminished by soil evaporation (which is not directly affected), interception losses and a strong aerodynamic component of canopy resistance to transpiration (Field, Jackson & Mooney 1995; Leuzinger & Körner 2010). For example, in Swiss calcareous grassland, CO2 enrichment halved stomatal conductance in the dominant species Bromus erectus (Lauber & Körner 1997), which, due to other, less responsive species and the above processes, translated into a marginal reduction in ecosystem ET of c. 6% only (Stocker, Leadley & Körner 1997). In the present study, we found very similar ET savings under elevated CO2, which were supported by the increased δ18O in leaf tissue of plants exposed to 580 ppm CO2, providing an integrated, qualitative signal across the growing season. The increased δ18O most likely resulted from reduced stomatal conductance and thus, reduced leaf-level transpiration under CO2 enrichment (Barbour 2007; Farquhar, Cernusak & Barnes 2007), as the isotope composition of soil water and the relative air humidity was assumed to be equal between treatments (same location and same meteorological conditions and isotopic signal in precipitation). In an isotope study in the Eastern Alps, an increase in leaf δ18O by 1.3–1.8‰ in the alpine species Festuca rubra and Potentilla aurea was linked to a significant reduction in stomatal conductance by around 30% (Scheidegger et al. 2000).

Similar to the data presented here, the small reduction in ET also resulted in higher SWC under elevated CO2 throughout the growing season in the calcareous grassland (Niklaus, Spinnler & Körner 1998). In another CO2 experiment in alpine Carex curvula-dominated grassland in the same area in the Swiss Alps, the CO2 effect on stand ET was below the detection limit of gas flux measurements (Diemer 1994; no weighing lysimeters). Soil water savings by CO2 enrichment were commonly more pronounced in drier and more productive lowland grassland (Niklaus, Spinnler & Körner 1998; Nelson et al. 2004). We thus assume that, at ecosystem level, soil moisture savings by elevated CO2 are too weak to translate into significantly higher sums of deep seepage or runoff in non-water-limited, temperate systems, in which CO2 effects on runoff mainly depend on day-to-day rainfall patterns (Leuzinger & Körner 2010). In addition, initial water savings by elevated CO2 (after saturating rainfall) are consumed by plants during prolonged rainless periods, if water availability limits ET and, thus, do not result in increased runoff (Morgan et al. 2004). Such effects of scaling (to higher levels of complexity, larger spatial scales or longer time spans) as well as possible atmospheric feedback should be considered in more detail in hydrological modelling of CO2 effects (Leuzinger et al. 2011).

There are well-known drainage implications with lysimeters (Flury, Yates & Jury 1999). The interruption of the capillary continuum at the bottom of any lysimeter may cause the rate of deep seepage to slow and thus, cause a periodic increase in soil moisture compared to control soils in situ. Our fine resolution series of soil moisture readings inside vs. outside lysimeters did not show such a slowed drainage effect. Hence, it seems, the generally high soil moisture and regular rainfall did not permit such signals to materialize and the same relative depletion of soil moisture inside vs. outside lysimeters during rainless periods indicated very similar water consumption by vegetation inside and outside lysimeters at this high-alpine location.

Ecological implications

This article quantifies the way vegetation interacts with the ecosystem water balance, and it offers a basis for estimating the hydrological consequences of land transformation and atmospheric change at catchment scale. Our data evidence that the higher biomass and leaf area associated with extensification or abandonment of alpine pastures and meadows, which increase daily ecosystem vapour loss (ET), exhibit much stronger effects on the water balance than the slightly counteracting future (double pre-industrial) CO2 concentrations. The higher evaporative water consumption reduces growing season deep seepage by −220 to −350 m3 per hectare of unmanaged alpine grassland (wet vs. dry summer). A decrease in catchment water yield following reduced land use implies more intense soil moisture depletion during dry periods and a reduction in fresh water provision of alpine catchments to rivers and lowland ecosystems. These changes also have repercussions for the potential of hydroelectric power production, all important ecosystem services provided by alpine regions (Viviroli et al. 2011; Grêt-Regamey, Brunner & Kienast 2012). While the effect of increased evaporative water consumption on soil water availability is unlikely to cause plant water shortage in such humid alpine grasslands, our data illustrate the magnitude of such effects that would become much more important in drier regions or under future climate scenarios, which predict more frequent and prolonged dry spells in the Alps in combination with elevated summer temperatures and increased ET at high elevation (Calanca et al. 2006; Horton et al. 2006).

Assuming an average 1500 m vertical potential energy of water originating from the alpine belt of a mountain catchment (between 2400 m and sea level, of which potentially more than half is usable within the region, down to c. 800–1000 m a.s.l.), an overall efficiency of the hydroelectric installation of 8.5 kN m−3 and a current average price of hydroelectric power yield (0.09 CHF per kWh; ElCom 2012; 1 CHF currently equals c. 0.83 Euros or 1.10 USD), the seasonal reduction in water yield of 220–350 m3 (wet vs. dry summers) per ha of abandoned high-alpine grassland would correspond to a loss of revenues from power production of 70–112 CHF per ha and year. For the alpine zone of the currently studied test region in the headwaters of the river Reuss (9000 ha of vegetated area between 2100 and 2700 m a.s.l., mostly alpine grassland), these ‘costs’ for the loss of water for potential hydroelectric power production by tall vs. regularly grazed alpine grassland would amount to 630 000–1 000 000 CHF per year (assuming that 100% of the ‘additional’ water by grazed terrain is used for power production). These estimated ‘costs’ of declining land use are probably an upper limit for low-stature alpine vegetation, because it is unlikely that all grassland of an alpine catchment is grazed uniformly. Furthermore, effects may get diminished with the accumulation of dead plant material in grassland that is dominated by certain species (e.g. Festuca rubra) and has been abandoned for several years (Tappeiner & Cernusca 1989; Rosset et al. 2001; Merz et al. 2009). On the other hand, the reduction in water yield by extensification or abandonment of grassland is likely to increase as vegetation gets taller at lower elevation and as shrubs encroach. Alterations in soil properties due to changes in land use, for example, reduced or increased soil compaction or erosion induced by trampling of grazers (strongly depending on shepherding), will exert effects on the water balance that are not explored here. Our data show that the magnitude of land cover or land use effects on ET and runoff depend on the climatic conditions (e.g. dry vs. wet summers) as was suggested by hydrological (modelling) studies of alpine catchments (de Jong, Mundelius & Migala 2005; Verbunt, Groot Zwaaftink & Gurtz 2005). Climatic warming is likely to reduce discharge of alpine catchments during prolonged summer seasons, due to earlier snowmelt, reduced summer precipitation, increased summer temperatures and declining contributions of glacier melt water to total discharge (Beniston, Stoffel & Hill 2011; Huss 2011; Viviroli et al. 2011). Under such conditions, the hydrological consequences of declining land use are likely to become significantly more important as is reported here for the rather dry 2009 season.

The data presented here substantiate the need to account for hydrological consequences in any judgement of land use change in alpine regions. Furthermore, our results permit to attribute economic value to pastoralism that goes beyond agronomic and ecological aspects (such as the loss of usable agricultural land, decrease in landscape diversity and species diversity). This broader view at the benefits of sustainable pastoralism in alpine catchments has obvious political implications including those for the energy sector and freshwater production and thus, offers novel assets for the promotion of sustainable land use in remote areas (MacDonald et al. 2000; Flury, Gotsch & Rieder 2005; Marini, Klimek & Battisti 2011). We advocate that land use and change in vegetation cover should receive more attention in hydrological models of alpine catchments.

Acknowledgements

This study contributes to the informal tri-national project Bio-CATCH on sustainable land use and catchment water yield. We are especially grateful to our project partners U. Tappeiner and N. Obojes (EURAC, Bolzano and University of Innsbruck), S. Lavorel, J.C. Clément and J. Lochet (UJF Grenoble) for fruitful cooperation. This project received funding by BLW (Swiss Federal Office for Agriculture, project Bio-CATCH, U. Gantner), the University of Basel, the SNF (Swiss National Science Foundation, project VALURsern, CR30I3-124809), the FAG (Freiwillige Akademische Gesellschaft Basel), and profited from collaboration with the BAFU project N-alpin (Swiss Federal Office for the Environment, 09.0084.PJ/I225-1307). We thank T. van den Bergh, T. Bühlmann, G. Gisler, L. Reißig, D. Scherrer, M. Studer and T. Zehnder for their help with fieldwork, M. Saurer and R. Siegwolf for advice on and analyses of 18O. We acknowledge the Korporation Ursern for allowing us to use their land for years and the Alpine Research and Education Station Furka (ALPFOR) for providing facilities and accommodation.

Ancillary