Plant guilds drive biomass response to global warming and water availability in subalpine grassland



    1. Laboratory of Plant Ecology and Forest Botany, Technological Forestry Centre of Catalonia, 25280 Solsona, Spain; and
    2. Agricultural Engineering School, University of Lleida, 25198 Lleida, Spain
    Search for more papers by this author

Maria-Teresa Sebastià, Centre Tecnològic Forestal de Catalunya, Pujada del Seminari s/n, 25280 Solsona, Spain (fax 34 973481392; e-mail


  • 1The consequences of global warming and changes in resource availability were investigated in subalpine grasslands in the Pyrenees. These communities are considered to be especially vulnerable to climate change because of their position at the south-western edge of the semi-natural grassland biome in Europe.
  • 2Changes in patterns of above- and below-ground biomass were assessed for different plant guilds in two experiments, in which turves were transplanted from upland to lowland locations. The first experiment aimed to evaluate general responses to warming and drought, and the second to disentangle the effects of possible underlying mechanisms through resource manipulation by means of a nitrogen × phosphorus fertilization experiment.
  • 3The increased above-ground biomass in grassland turves transplanted to lowlands suggested that biomass production was more temperature-limited than water-limited. The enhancement effect found in the upland turves following phosphorus addition supported the hypothesis of a strong limitation arising from reduced nutrient availability, confirming the central role played by phosphorus in these grasslands and its potential importance in the response to global change.
  • 4Nitrogen addition did not stimulate total biomass but affected guild composition. Grasses dominated the uplands and at high resource levels, while forbs dominated the lowlands and when water and nutrients decreased. The counterintuitive effect of increased biomass with decreased water in the lowlands was related to shifts in dominance from grasses to forbs, probably enabled by decreased nutrient availability under drought conditions.
  • 5Synthesis and applications. Environmental factors interacted in complex ways, producing changes in biomass distribution and guild proportions in subalpine grassland. In addition, the results suggested that the capability of high-altitude grasslands to provide quality forage in summer time could be threatened in the northern Mediterranean region under climate change conditions because of: (i) a decrease in their reliability as a result of complex biomass interactions with temperature, water and nutrient dynamics; (ii) expected feedback mechanisms; and (iii) compositional shifts.


Climate analyses show that summers in Europe during the last decade have been the warmest in 500 years (Luterbacher et al. 2004). In addition to warming, many zones are expected to experience decreased rainfall under conditions of climate change (Watson, Zinyowera & Moss 1998). High mountains in the northern rim of the Mediterranean region have cold-temperate climates, with moderate summer temperatures and high precipitation. However, some years have shown distinct Mediterranean influences, with summer drought periods. Extreme climatic events are predicted to increase with climate change (Houghton et al. 2001) and the summer drought periods are expected to occur more frequently in the northern Mediterranean high mountains (Llebot 2005).

In cold ecosystems, low temperatures limit many biological processes, and warming has been found to produce an enhancement in primary productivity (Rustad et al. 2000). In Mediterranean ecosystems, water is considered to be a key factor limiting plant growth (Roy, Garnier & Jackson 1987), being a critical resource for plant survival (Lloret, Peñuelas & Estiarte 2004) and competition (Vilà & Sardans 1999). The projected diminishing water input, combined with increased temperature, will produce further reductions in water availability to plants in the Mediterranean region. Increased frequency of extreme climatic events involving recurrent drought is expected to reduce ecosystem resilience to stress in the Mediterranean (Lloret, Siscart & Dalmases 2005).

At an individual level, responses to changes in environmental conditions can be achieved through physiological adjustments of existing plants. Among these adjustments, shifts in biomass allocation between above- and below-ground organs are considered to be important mechanisms used by plants to maintain their productivity in the face of environmental change and limiting factors (Davidson 1969; Van der Werf & Lambers 1996). Plants maintain a balanced functional economy by partitioning photosynthate and absorbed minerals into roots and shoots. The functional equilibrium model assumes that the relative allocation to shoots and roots varies to compensate for changes in root and shoot activity (Van der Werf & Lambers 1996). When there is nutrient or water shortage, plants invest in root growth and more biomass is allocated to the root system (Beck 1996; Van der Werf & Lambers 1996). Therefore, the root mass:shoot mass ratio balances the photosynthetic rate:absorption rate, setting limits to plant responses (Davidson 1969).

Disturbance has been defined in terms of changes in resource availability (Bazzaz 1996). If many individuals in an ecosystem respond similarly to disturbance, this will produce an impact at the ecosystem level. However, not all species are expected to perform alike and this can lead to shifts in species composition and plant functional types. Groups of plants found to respond similarly to changes in resources have been classified into guilds (sensu Root 1967; for a recent discussion applied to plant groups see Blondel 2003). In grasslands, the most important guilds have long been considered the grasses and forbs (Bazzaz & Parrish 1982). Changes in species and guild composition under changing environmental conditions are expected to modify the overall biotic functional traits present in the ecosystem, therefore affecting ecosystem processes, which, in turn, feeding back, can modify species and functional type composition (Chapin et al. 1997).

Subalpine grasslands in the Pyrenees are similar to other grasslands from cold-temperate regions in many structural and functional aspects (Canals & Sebastià 2000; Sebastià 2004). They play a key role in the stockbreeding cycle because they are used as summer pastures, and constitute an inexpensive, sustainable source of food for livestock at a time when productivity is limiting in more Mediterranean, low-altitude grasslands. However, they are at the south-western edge of the semi-natural grassland biome's distribution in Europe (Rychnovska 1993) and therefore particularly vulnerable to changing climatic conditions (Sala et al. 2000). In order to assess the vulnerability of subalpine grasslands in the Pyrenees to warming and changes in water availability, two transplanting experiments with turves were performed (for a discussion of this technique see Shaver et al. 2000; for some examples see MacGillivray, Grime & The Integrated Screening Programme Team 1995; Ineson et al. 1998). In the first experiment, general responses of the ecosystem in terms of biomass production and allocation were explored. In the second experiment, several resources were manipulated to test possible explanatory mechanisms for the observed patterns. In particular, the following questions were addressed: (i) Does yield decrease with decreased water availability, as expected in Mediterranean ecosystems (Roy, Garnier & Jackson 1987), or increase with increased temperature, as found in cold ecosystems (Rustad et al. 2000)? (ii) Does the ecosystem shift patterns of below-ground allocation as a mechanism to cope with the altered climatic conditions (Davidson 1969; Van der Werf & Lambers 1996)? (iii) Do responses to altered resource availability with warming involve changes in plant guild composition (Dormann & Woodin 2002)? Expectations were that: (i) the decreased biomass associated with reduced water availability will overcome the enhancement effect of increased temperature; (ii) water shortage will stimulate biomass allocation to roots; and (iii) functional changes in the ecosystem will be related to changes in guild composition.


experiment 1: warming and drought

Most climate change scenarios project warmer and drier conditions in north-eastern Spain (Llebot 2005). Experiment 1 was designed to assess general responses in plant biomass to warming and drought through transplanting from upland (moist and cold) to lowland (dry and warm) areas. Response variables included above- and below-ground biomass, biomass allocation to plant parts and biomass distribution into guilds. The experiment was performed on vegetation at Pla de Rus, Cadí-Moixerò Natural Park, in the eastern Pyrenees (1·993°E 42·276°N). The same design was replicated in two different sites, M1 and M2, located at around 2000 m a.s.l. and approximately 1 km apart. Both sites held a vegetation of subalpine perennial mesic Festuca nigrescens grasslands on limestone (for more information on the structure and ecology of these grasslands see Sebastià 2004). Festuca nigrescens Lam., Carex caryophyllea Latourr., Anthoxanthum odoratum L., Potentilla neumanniana Reichenb., Galium verum L., Thymus pulegioides L. and Koeleria macrantha (Ledeb.) Schultes were among the most abundant species in the grassland. At each site a plot of around 400 m2 protected from livestock by electric fence was established. Inside the plot, 60 turves were extracted. The turves were cut with shovels close to 40 × 40 × 20 cm depth and sawn to make them fit exactly plastic trays of 40 × 40 × 20 cm depth. The trays were pierced with 25 holes, each 0·6 cm diameter. Turf depth was sufficient to minimize damage to roots, as most below-ground material occurs within the first 20 cm of topsoil (Sebastià 2004). In total, 120 turves were excavated.

Within each site, half of the excavated turves were selected at random and placed back in the grassland in their trays (upland location). The other half was transported to the experimental fields of the Agricultural Engineering School of the University of Lleida at around 400 m (lowland location) and placed there at random, blocked by site of origin. All the trays were buried at soil level, to prevent warming of the tray walls.

Climatic conditions in the two locations in this experiment were determined from the Atlas Climàtic Digital de Catalunya (, accessed 2006; Ninyerola, Pons & Roure 2002). In the upland (Pla de Rus, 2014 m a.s.l.), mean annual rainfall is 1214 mm and mean annual temperature 5 °C. In the lowland (Lleida, 400 m a.s.l.), mean annual rainfall is 390 mm and mean annual temperature 14 °C. During the experiment, 10 mm of water were added weekly in the lowland location to decrease the difference in water availability between both locations, in order to reach around half the precipitation expected in the upland for the duration of the experiment.

Half of the turves were used for non-destructive measurements, while 56 were selected at random to perform two harvests, half of these at 10 weeks (harvest 1) and half at 17 weeks (harvest 2) after the beginning of the experiment (mid-May). At both harvests, above-ground biomass in a 20 × 20-cm quadrat was harvested in the centre of 28 randomly selected turves (seven per site and locality). A cube of soil of 10 × 10 × 20 cm depth was extracted from the middle of each turf after harvesting above-ground biomass. Roots and other below-ground material in the 0–10- and, for some samples, 0–20-cm soil layer were separated and carefully washed. Comparison of the 0–10- and the 10–20-cm layers did not reveal different patterns in response to the treatment factors. Therefore, the more complete data set on below-ground biomass in the 0–10-cm soil layer was selected for detailed analysis. Above-ground plant biomass by species and guilds (grasses, sedges and forbs) and above-ground dead matter were separated. Above- and below-ground biomass were oven-dried to constant mass separately, and the dry weight was determined. The root weight ratio (RWR) was calculated as below-ground biomass/total biomass. The forb weight ratio (FWR) was calculated as biomass of forbs/total biomass.

The most dominant species in experiment 1 (both in the upland and the lowland) were selected and their forage value determined by means of the specific index (Is) as proposed by Daget & Poissonet (1972) in their pastoral value method. The specific index ranged from 0, in species that had null forage interest, to 5, in species that made excellent forage. The specific indexes for the species in the study were determined from existing data sets [Daget & Poissonet 1991; Cavallero et al. 1992; CEMAGREF (, accessed 2005) and CTFC (, accessed 2006) databases; R.M. Canals, M. Taull & M.T. Sebastià, unpublished data].

experiment 2: warming, water availability and nutrient manipulation

Experiment 2 was designed to disentangle the relative effect of resource availability (nutrients and water) and temperature on changes in biomass patterns after transplanting from the upland (moist and cold) to the lowland (dry and warm) through resource manipulation. The experiment was undertaken in Alinyà, located in the eastern Pyrenees (1·414°E 42·188°N). This area has a more continental climate than Pla de Rus (experiment 1), with higher temperatures and lower rainfall (Sebastià 2004). In this experiment, 128 turves were extracted from subalpine grasslands at two different sites, at about 2000 m a.s.l. All turves were put in plastic trays as in experiment 1. Half of the turves were placed back in the grassland in Alinyà (upland location) and half were transported to experimental fields at Can Mascaró (lowland location). Climatic conditions in the two locations were obtained from the Atlas Climàtic Digital de Catalunya (Ninyerola, Pons & Roure 2002). In the upland (Alinyà, 2004 m a.s.l.), mean annual rainfall is 1132 mm and mean annual temperature 5 °C. In the lowland (Solsona, 664 m a.s.l.), mean annual rainfall is 670 mm and mean annual temperature 12 °C.

The methodology for turf extraction and experimental design was similar to experiment 1 but, in this second experiment, in addition to warming through transplanting, resources (nutrients and water) were manipulated to disentangle the effects of several possible factors driving biomass patterns. The following four nutrient treatments were applied to 64 turves (eight per treatment and location, upland and lowland, selected at random): (i) no nutrient addition, (ii) only nitrogen added (60 kg ha−1 ammonium nitrate added bi-weekly), (iii) only phosphorus added (25 kg ha−1 calcium bi-phosphate added bi-weekly) and (iv) both nitrogen and phosphorus added (in the same amounts as single nutrient treatments).

To investigate the effects of water, historical rainfall patterns were used as a basis of water adjustments aimed to create overlapping gradients of moisture in the two locations, upland and lowland. Watering was 44 and 179 L m−2 in the upland and 44, 89 and 134 L m−2 in the lowland. Adding this amount to the estimated rainfall during the experiment, turves received about 505 or 642 mm of water in the upland and 471, 516 or 561 mm in the lowland. The watering treatments were randomly assigned to turves in both locations. Because the application of these adjustments did not result in identical water regimes in upland and lowland, water input was used as a quantitative variable in the analysis.

The 64 turves were harvested on 3 October, 18 weeks after the beginning of the experiment, and above-ground biomass processed and treated as described for experiment 1. Plant samples were separated into three guilds: grasses, sedges and forbs. One 20-cm diameter core of soil was taken from each turf after harvesting. Roots and other below-ground material in the upper 10-cm soil layer were separated, following the methodology described for experiment 1.

statistical analysis

In experiment 1, two-way anova was used to explore the relationship between total above-ground biomass and the treatment factors (site, location). The same analysis was applied to live above-ground biomass (yielding similar results to total above-ground biomass) and to below-ground biomass, total biomass, RWR and FWR. Biomass was log-transformed when necessary to meet the assumptions of the analysis. The allometric relationship between log-transformed below-ground biomass and log-transformed above-ground biomass was analysed, with site and location included in the analyses. The relationships of RWR, above-ground biomass and below-ground biomass with proportion of forbs and FWR, including site and location, were modelled as previously described. Transformations were applied when necessary. Non-parametric Mann–Whitney tests were used to compare differences between locations in biomass of guilds and of some abundant species.

In experiment 2, linear modelling was used to model above-ground biomass, total and by guilds, as a function of water input (rainfall during the experiment plus added water), with site, location, nitrogen and phosphorus fertilization introduced as factors (with values 0 = no added and 1 = added). In all models, first- and second-order interactions were initially introduced, except for site, but were removed when not significant. In addition to regression techniques applied to RWR, allocation was analysed following the allometric approach, by modelling log-transformed below-ground biomass as a function of log-transformed above-ground biomass, plus the treatment factors. This was done to separate allocation responses as a result of changes in resources from size-dependent allometric effects (Müller, Schmid & Weiner 2000; Weiner 2004).

Compositional data analysis (CDA) was applied to predict changes in the proportion of guild biomass with the treatment factors. CDA takes into account the fact that proportions of all components must add up to 1 and therefore predicts these proportions with higher accuracy than other methods (Muldowney, Connolly & Keane 2001). This method has been applied successfully in ecology (for some applications on vegetation data see Sebastià 2004; Connolly & Wayne 2005; Ramseier, Connolly & Bazzaz 2005). Following Muldowney, Connolly & Keane (2001), the log-transformed abundances of grasses and forbs, divided by sedges, were modelled in parallel. In the models all the treatment factors, plus log-transformed above-ground live biomass and the interactions, were introduced. The predicted values of grasses and forbs, respectively, were calculated from the resulting final model, obtaining the proportion of sedges by difference. The predicted guild proportions were calculated at values of above-ground live biomass predicted by the full model for this variable at each treatment level.


biomass responses and changes in below-ground allocation patterns (experiment 1)

Despite huge increases in above-ground biomass of the turves transplanted to the lowland in experiment 1 (P < 0·001; Table 1), below-ground biomass did not change under the experimental conditions (Table 1). Therefore, biomass allocation to below-ground parts (RWR) was lower in the lowland than in the upland (P < 0·001; Table 1).

Table 1.  Experiment 1: mean values ± 1 SE of several biomass variables by location, upland and lowland, for harvest 1 (31 July) and harvest 2 (19 September). For variables analysed using parametric tests, letters compare mean values among treatments for a given variable by the Tukey mean separation test: treatments not sharing any letter are significantly different from each other at the P < 0·05 level
  Harvest 1Harvest 2
Live above-ground biomass (g m−2)Upland152 ± 11·7a190·59 ± 12·6a
Lowland605 ± 40·5b718·70 ± 77·7b
Above-ground total biomass (g m−2)Upland156 ± 11·9a   198 ± 13·0ab
Lowland611 ± 40·7bc   731 ± 78·0c
Above-ground forb biomass (g m−2)Upland 90 ± 11·1    88 ± 9·7
Lowland262 ± 28·4   358 ± 51·5
Above-ground sedge biomass (g m−2)Upland 18 ± 3·3    34 ± 6·4
Lowland144 ± 26·2   169 ± 17·8
Above-ground grass biomass (g m−2)Upland 44 ± 4·5    69 ± 8·0
Lowland199 ± 10·4   192 ± 57·6
Below-ground biomass (g m−2)Upland  1852 ± 76
Lowland  1786 ± 106
RWRUpland   0·90 ± 0·006a
Lowland   0·70 ± 0·023b
Total biomass (g m−2)Upland  2048 ± 83a
Lowland  2549 ± 135b

Adding up the values of all turves by location, above-ground biomass increased in the lowland compared with the upland in experiment 1 for all three guilds, forbs, grasses and sedges. However, increases differed among the groups. Overall, grasses increased three times, sedges four and forbs five times in the lowland compared with the upland. Mean above-ground biomass per sample was significantly higher in the lowland than in the upland for forbs and sedges (P < 0·001, Mann–Whitney tests), but not for grasses (P = 0·874), which showed high heterogeneity in response among samples (Table 1).

Below-ground biomass increased with the proportion of forbs in the lowland. FWR explained up to 80% of below-ground biomass variation in the lowland but only 1% in the upland in this experiment. Above-ground biomass did not change significantly with forb proportion in experiment 1, but several forb species showed significant increases in above-ground biomass in the upland by the final harvest and became part of the dominant species (Table 2). The most abundant grasses, in contrast, decreased in relative biomass (and, in the case of Festuca nigrescens, in absolute biomass as well; Table 2). In the upland, the most abundant grasses were those with a higher forage specific index (Table 2). In the lowland, these were substituted in dominance by sedges and forbs with mediocre or null forage value (Table 2).

Table 2.  Experiment 1: summary of changes in the eight most abundant species in the transplant experiment considering absolute (left) and relative (right) biomass values. The specific index, Is, is an indicator of the species’ forage value. Only species significantly changing in the lowland compared with the upland (control) in harvest 1 (31 July), or in harvest 2 (19 September) are shown. +, increased biomass compared with the control; –, decreased biomass compared with the control. Mann–Whitney test significance level < 0·05. Bonferroni correction < 0·006 (0·05/8 comparisons = 0·006); symbols between parentheses indicate significant individual differences, but Bonferroni correction > 0·006. The ranks of the species showing significant changes in relative biomass between the upland and the lowland are shown
HarvestAbsolute biomassSign of changeRelative biomassSign of changeIs*Ranking in the uplandRanking in the lowland
  • *

    Is, specific index. Quality index that ranges from 0, in species that have null forage value, to 5, when they make excellent forages (pastoral method; Daget & Poissonet 1972).

1Carex caryophyllea+Carex caryophyllea(+)1 21
Anthoxantum odoratum+Festuca nigrescens(–)3 14
Festuca nigrescens +     
Euphorbia cyparissias+     
Potentilla neumanniana(+)     
Achillea millefolium(+)     
2Carex caryophyllea+Achillea millefolium (+)2131
Potentilla neumanniana +Potentilla neumanniana(+)0 84
Euphorbia cyparissias L.+Carex caryophyllea(+)1 22
Achillea millefolium L.(+)Thymus pulegioides0 68
Koeleria macrantha(+)Anthoxanthum odouratum2 36
Festuca nigrescensFestuca nigrescens3 18

disentangling biomass response mechanisms through resource manipulation (experiment 2)

Above-ground biomass increased in the lowland compared with the upland in experiment 2, as in experiment 1. However, modelling revealed that this effect decreased as water input increased (Table 3; Fig. 1). Phosphorus addition increased above-ground biomass, and this effect was higher in the upland than in the lowland (Table 3). In fact, in the upland, predicted above-ground biomass was no different from that in the lowland when phosphorus was added (Table 3). On the contrary, biomass enhancement by nitrogen addition was weak (non-significant effect; Table 3).

Table 3.  Experiment 2: results from regression analysis on above-ground biomass, total and by guild, and below-ground and RWR. The variables above-ground and forb biomass were log-transformed to meet the assumptions of the analysis. RWR was transformed using the arcsin(square root) method. P-values are presented for variables and interactions included in the final models
 Above-ground total biomass (g m−2)Below-ground biomass (g m−2)RWRGrass biomass (g m−2)Forb biomass (g m−2)Sedge biomass (g m−2)
  1. N, nitrogen; P, phosphorus.

Constant  0·001  0·042< 0·001  0·134  0·991  0·186
Site   0·137    0·008< 0·001  0·012
Location< 0·001  0·018  0·027  0·559< 0·001< 0·001
N  0·186    0·008  0·004 
P  0·010   0·016 < 0·001  0·232 
Water< 0·001  0·094  0·042  0·026  0·177 
Location × P  0·001   0·023  0·016  0·001 
Location × water  0·001   0·063   
P × water  0·042   0·038   
inline image (%) 52   7 19 49 77 25
P-value0·0010·049< 0·006<0·001< 0·001< 0·001
Figure 1.

Experiment 2: mean + 1 SE total above-ground biomass estimated from regression for each location, at different water levels, set between the maximum and minimum water input values (rainfall plus watering for the duration of the experiment) in the experiment. Predictions estimated at mean values for other factors in the model. Significance of differences between locations for a given water input are indicated by the P-value. NS, not significant.

In this experiment, below-ground biomass was significantly higher in the lowland than in the upland, and independent of FWR (Table 3). The RWR decreased significantly when phosphorus was added in the upland (P < 0·01) but not in the lowland (P = 0·993; see Table 3 for interaction). However, RWR increased at low water levels with phosphorus shortage (Table 3). Allometric analysis revealed that log-transformed below-ground biomass was independent of log-transformed above-ground biomass. The factors included in the final model were location and water input (Table 3).

Guilds were very responsive to changes in resources in experiment 2, and these responses were complex and most often dependent on location (Table 3). Grasses increased with increased water availability and with nitrogen and phosphorus fertilization (Table 3). However, the effect of phosphorus on grass biomass was weaker in the lowland than in the upland (significant location × phosphorus interaction; Table 3). Forbs thrived in the lowland (Fig. 2), particularly when resource levels were low (mainly water but nitrogen and phosphorus as well). In the lowland, forb biomass was suppressed with the addition of phosphorus, particularly at the highest water availability, and this effect was opposite to what was found in the upland (significant location × water input × phosphorus interaction; Table 3).

Figure 2.

Experiment 2: mean guild biomass proportion estimated from CDA in the upland and the lowland at different water levels, set between the maximum and minimum water input values in the experiment. Standard errors predicted from regression on logit transformations. All estimates were calculated at the predicted value of above-ground live biomass for the given treatment. All predictions estimated at mean values for other factors in the model. Fertilization treatments: P0 and N0 indicate that nitrogen or phosphorus was not applied, N1 and P1 that it was.

The above responses led to changes in the proportion of guilds with warming and altered resource availability, as revealed by CDA (Table 4). In the upland, most of the biomass was made up of grasses, particularly at high levels of nutrients (Fig. 2). In the lowland, the grass proportion decreased and sedges and forbs increased, with forbs becoming the dominant guild in the lowland at low nutrient levels (Fig. 2). Water was not significant as a main effect nor in interaction with the other treatment factors, and therefore was not included in the final CDA model (Table 4).

Table 4.  Experiment 2: results from CDA on guild proportion. The dependent variable was the log-transformed biomass grass or forb over sedges. Step-wise backwards regression was used to find the minimum common model for both components, grasses and forbs. The third component, sedges, was calculated by difference. P-values are presented for variables and interactions included in the final model
 CDA grassesCDA forbs
Site0·707< 0·001
Location × N0·1600·009
Location × P0·0160·212
Location × biomass0·0990·239
N × P0·0320·056
N × biomass0·0780·069
P × biomass0·0030·010
Location × N × P< 0·0010·003
Location × N × biomass0·1290·006
Location × P × biomass0·0030·092
N × P × biomass0·0090·041
Location × N × P × biomass< 0·0010·004
inline image (%)4861
P-value < 0·001< 0·001


functional responses

Above-ground biomass increased in the lowland compared with the upland in both experiments, in spite of decreased water availability (for experiment 1 see Table 1; for experiment 2 see Fig. 1). These results suggest that productivity in subalpine grasslands in the Pyrenees might be more temperature-limited than water-limited, contrary to our first hypothesis.

Temperature accelerates metabolic rates, favouring plant growth. It also stimulates microbial activity, enhancing biologically driven processes such as organic matter mineralization (Shaver et al. 2000). The mineralization of organic matter provides nutrients to plants if nutrient immobilization by microbes is low (Shaw & Harte 2001). Many experiments show a stimulation in yield with warming in cold-temperate regions (Rustad et al. 2000), which in part has been attributed to the increased nutrient availability through mineralization (Epstein et al. 2000; Shaver et al. 2000). Further nutrient fertilization has been found to occur by decreased leachate when transplanting turves to lower altitudes (Ineson et al. 1998).

The decrease in RWR in the lowland turves compared with the upland turves in both experiments (Tables 1 and 3) and the enhancement of above-ground biomass with phosphorus addition in experiment 2 (Table 3) are consistent with the hypothesis of higher nutrient availability in the lowland through increased organic matter mineralization with warming. The selective positive effect of phosphorus addition on above-ground biomass, restricted to the upland location (Table 3), suggests a critical threshold for which phosphorus is a limiting factor in subalpine grasslands in the Pyrenees under natural conditions (Sebastiàet al. 2004) and demonstrates the important role of phosphorus in these ecosystems (Table 3). Phosphorus has already been shown to be critical for primary productivity in limestone grasslands in the Alps (Köhler et al. 2001) and a wide survey of subalpine grasslands on limestone in the Pyrenees demonstrates the importance of phosphorus in the distribution of plant communities across the landscape (Sebastià 2004). Nitrogen addition in this experiment failed to produce any total above-ground biomass response (Table 3). However, at the guild level, there were significant changes in response to the nitrogen addition, the guilds showing diametrically opposite effects (Table 3).

Consistent with the balanced growth hypothesis (Davidson 1969; Van der Werf & Lambers 1996), biomass allocated to below-ground parts decreased with increased below-ground resources, when phosphorus was added and, in the upland, when water availability increased. However, decreased below-ground allocation and increased above-ground biomass with water shortage in the lowland indicate other mechanisms at work (see below). The independence of below-ground biomass from above-ground biomass in the allometric analysis suggests that this was not a size-dependent effect (Weiner 2004).

The biomass enhancement with warming in these short-term experiments could be transitory (Fig. 1); it is most likely to disappear once feedback mechanisms, including the consumption of the mobilized nutrients, increased water depletion by a higher biomass and changes in vegetation composition increase in importance (Sebastiàet al. 2004). However, these initial responses to disturbance could be critical in grassland communities and may determine the whole future trajectory of the ecosystem (Sebastià, Canals & Gamarra 1998).

the role of guilds

The results from both experiments suggest strong guild responsiveness to the experimental factors influencing the functional responses of the ecosystem to climate changes and resource alteration, as expected from our third hypothesis. Increases in dominant perennial forb species and decreases in dominant grasses have been reported with warming and summer drought in the UK (Sternberg et al. 1999). In our experiments, grasses made up most of the biomass in the upland, while forbs and sedges predominated in the lowland (for experiment 1 see Table 2; for experiment 2 see Fig. 2).

Changes in biotic interactions have been shown to affect responses to climate change in alpine species (Klanderud 2005). In experiment 2, high resource availability facilitated the maintenance of a significant grass proportion in the lowland (Fig. 2) while, at low resource levels, only forbs were resistant enough to thrive and yield high biomass (Fig. 2). The results suggest (i) grasses were strongly competitive in the upland; (ii) grasses were highly sensitive to warming in the lowland, independently of other experimental conditions; (iii) adverse conditions for grasses in the lowland reduced their competitiveness and facilitated the development of other guilds, i.e. sedges and forbs; and (iv) resource addition ameliorated the negative effects of warming on grasses. In addition, two important properties of forbs emerged: (i) they were strongly responsive to resources in complex ways (many significant interactions among experimental factors; Table 4) and (ii) the revealed patterns were consistent within this group (a high proportion of the variability explained in all models applied to either forb biomass or proportion; e.g. Table 3). The counterintuitive result of increased above-ground biomass with decreased water availability (also reported by Krueger-Mangold et al. 2004) could be related to increased forb biomass (Figs 1 and 2).

In experiment 1, below-ground biomass increased with the proportion of forbs in the lowland location but above-ground biomass did not change with this, nor did RWR (Table 1), suggesting an increase in forbs with high below-ground biomass in the lowland not mirrored by an increased overall below-ground allocation. However, in experiment 2, where conditions between upland and lowland were less extreme, below-ground biomass was not dependent on FWR. Differences in vegetation composition between both experiments could be behind this inconsistency. In experiment 1, forb proportion in the upland was similar (FWR = 0·47) to experiment 2 (FWR = 0·46) but sedges were much more abundant. In the lowland, forb proportion was much lower in experiment 1 (FWR = 0·53) than in experiment 2 (FWR = 0·74), because in the former a sedge (Carex caryophyllea) was the most dominant species after warming (Table 2). This suggests different responses depending on differences in climate and the local species pool (de Bello, Lepš & Sebastià 2005). Differences in climate and the local species pool among locations could also explain differences in the sensitivity of grasslands and other open ecosystems to changes in climate factors (Dunnett et al. 1998; Dormann & Woodin 2002; Stampfli & Zeiter 2004).

Although there was no effect of nitrogen fertilization on overall biomass, above- or below-ground (Table 3), this nutrient was important for guild yield and distribution (Tables 3 and 4). The nitrogen fertilization enhanced grass and sedge biomass but depressed forbs (Fig. 2). The positive effect of nitrogen on grasses was reduced when phosphorus was limiting (Fig. 2). The opposite effects of nitrogen on guilds could be responsible for the lack of overall above-ground biomass responsiveness to this nutrient (Fig. 2).

It is important to take into account that, although absolute biomass responses in all guilds and most species in this study point towards increased biomass in the lowland, differences in responses in relative guild biomass (Table 2 and Fig. 2) could determine strong compositional changes at the community level in the short term. This could induce changes in community structure and functioning in the mid- and long term.

implications for management

It has been suggested that the current understanding of how species and ecosystems respond to climatic changes will be a key tool for future management adapted to changed conditions (Hulme 2005). High-altitude grasslands are used as summer pastures in many mountain regions around the globe. In the northern rim of the Mediterranean, they constitute a much needed resource for extensive stockbreeding during a period in which grasslands at lower altitudes are depleted, parched or both (de Bello, Lepš & Sebastià 2005).

This traditional management is threatened by climate (Arenas, Serra & Sebastià 2004) and land-use changes (Taull, Casals & Sebastià 2006). Our results indicate that the reliability of these precious resources, available when other food supplies are not accessible in lower altitude pastures, will decrease as a result of complex biomass interactions with temperature, water and nutrient dynamics (Table 3 and Fig. 1). In addition, we found that increased biomass does not equal higher forage quality, and some of the best grass forage species in experiment 1, such as Festuca nigrescens, were negatively affected by the treatment conditions, while other less-palatable forbs, such as Potentilla neumanniana and Euphorbia cyparissias, became abundant (Table 2). Analyses of the species’ (Table 2) and plots’ (Sebastiàet al. 2005) forage values suggest that high-altitude grasslands in the northern Mediterranean region could decrease their capability of providing good forage in summer time under climate change conditions. An increased frequency of extreme events such as droughts is expected to decrease further the reliability of these food sources (Hulme 2005).

Interactions between climate and management could be relevant in this context. Grazing is an important environmental filter that selects specific groups of species (Louault et al. 2005). In both experiments described here, turves were protected from grazing because of the intrinsic complexity of adding this treatment to the experimental set-up. However, de Bello, Lepš & Sebastià (2005) found that functional types selected by grazing changed depending on climate conditions. Therefore the observed patterns could be modified by grazing.

Opposite responses in traits relevant for grassland utilization by livestock (e.g. increased biomass but decreased forage quality) suggest that in the long term the future of these grasslands as food sources in stockbreeding activities must be carefully reconsidered.


I would like to thank S. Duaigües, V. Gràcia, B. Mola, M.C. Franceschini, P. Ferrer and P. Gayoso for extensive field work and sample handling. Particular thanks to J. Weiner for guidance and encouragement, and J. Connolly for stimulating comments and statistical advice. Two anonymous referees provided very useful comments. The Fundació Territori i Paisatge, the Natural Park Cadí-Moixerò and the School of Agricultural Engineering of the University of Lleida facilitated the development of the experiments. This research was partially supported by the Catalan Government and the Thematic Catalan Network of Climate Change (CARBOCAT project), the Fundació Territori i Paisatge, the Paeria of Lleida and the Technological Forestry Centre of Catalonia. The manuscript was initially developed while the author was a visiting professor at the University College Dublin, with a fellowship from the Government of Catalonia.