Introduction

Mixing plant species in agroecosystems may result in “niche complementarity”, that is increased resource exploitation, as well as spatial and temporal partitioning of resource use by coexisting species (Hector et al. 1999). For example, Kahmen et al. (2006) showed that plant species from different functional groups may be complementary in their specific N-uptake strategies, suggesting a potentially beneficial effect of functional group diversity on ecosystem functioning. Ecological and economic benefits of multispecies systems relative to homogeneous pure cropping systems depend, however, on the net effects of facilitative and competitive interactions on resources use between coexisting species (Malézieux et al. 2009). To develop ecologically and economically profitable agroecosystems, it is therefore necessary to consider resource limitations, species competition for nutrients and other resources, and nutrition partitioning as affected by mobility and other nutrient properties (see reviews by Schroth 1998, and García-Barrios and Ong 2004). Here, we focused on multilayered Mediterranean dehesas as a suitable model for the study of differences in nutrient partitioning among plants belonging to different functional groups.

Dehesas are silvopastoral systems that are used by cattle, sheep, goats, pigs and horses, and are limited in their agricultural potential, owing to dry climatic conditions and low soil fertility (Gómez-Gutierrez and Pérez-Fernández 1996). Dehesas are among the most important and widespread agroforestry land-use systems in Europe (Eichhorn et al. 2006). These agroecosystems consist of scattered oaks, chiefly holm oak (Quercus ilex L. ssp. ballota), within a mosaic of grasslands, croplands and shrublands, where livestock is extensively raised. Livestock is undoubtedly the most important output of the dehesas, and its diet depends largely on pasture production. As a foundation tree species (sensu Ellison et al. 2005) in the dehesas, Q. ilex can drive vegetation dynamics and influence the distribution and abundance of many dependent community members. Shrub species develop in dehesas as a natural process resulting from the reduction in grazing pressure. These shrubs offer “safe-sites” that may promote tree regeneration (Pulido et al. 2010).

This study aimed to discern the ability of different vegetation layers to compete for different soil nutrients. We compared dry matter production and nutritional status of pasture, shrubs and trees under a no fertilisation regime and under NPK fertilisation in two farmer-designed silvopastoral sites. These study sites differed in the species that formed the shrub layer. We selected two prominent Mediterranean shrub species that vary in their specific ecological strategies and niches: a N2-fixing and sparse deep-rooting shrub (Retama sphaerocarpa (L.) Boiss, hereafter Retama), versus a dense shallow-rooting shrub (Cistus ladanifer L, hereafter Cistus). We used vector analysis to explore the competitive ability of each vegetation layer and to speculate on long-term maintenance of pasture and tree nutrition in shrub encroached-dehesas, where low soil N availability is limiting plant growth. We hypothesised that pasture species would be more responsive to fertilisation than shrubs (Retama and Cistus) and Q. ilex because of the conservative nutrient-use strategies of evergreen Mediterranean trees and shrubs (Aerts 1995; Valladares et al. 2000). Based on our previous studies (Rivest et al. 2011; Rolo and Moreno 2011), we also hypothesised that the magnitude of pasture and Q. ilex response to fertilisation would be higher in the site encroached by the shrub “competitor” Cistus (i.e., contributes to increased limiting soil nutrients) than that encroached by the shrub “facilitator” Retama (i.e., contributes to increased N, usually the most limiting of soil nutrients).

Materials and methods

Study sites and experimental design

The study was conducted at Retortillo (39°57′ N, 6°6′ W) and San Esteban (39°56′ N, 6°11′ W) in the north of the Extremadura region of central western Spain. The climate is Mediterranean, with dry and hot summers, and cool and rainy winters. The region is characterised by an annual average temperature of 15.7°C, and 568 mm of annual precipitation (Aldehuela del Jerte station, 40°00′ N, 6°13′ W, 2000–2010 records). Monthly average air temperature and precipitation during the study (from January 2009 to July 2010) are shown in Fig. 1. Soils are loam or sandy loam Distric Cambisols (IUSS Working Group WRB 2006). Both study sites included a Q. ilex overstorey, a shrub layer, and a pasture understorey dominated by annual native Mediterranean herbaceous species (Fernández-Moya et al. 2010) with a weak presence of legumes. The shrub layer was composed of Retama at Retortillo, hereafter called the Retama site, and Cistus at San Esteban, hereafter called the Cistus site. Most annual species germinate in autumn, develop in early spring, and desiccate in late spring, when soil water becomes limiting. In Iberian dehesas, the mature sclerophyll Q. ilex typically grows in pure and low-density stands (15–45 trees ha−1) and develops deep (i.e., mostly beneath roots of herbaceous species) and laterally extended root systems (Moreno et al. 2005). Retama is a leafless N2-fixing shrub with photosynthetic stems (i.e., cladodes) that develops a deep-root system and grows in scattered patches (Haase et al. 1996). Cistus is an evergreen semi-deciduous shrub that forms dense populations, and develops shallow, dense-root systems (Silva and Rego 2004). Relevant vegetation and soil characteristics for each site are summarised in Table 1.

Fig. 1
figure 1

Monthly average precipitation (filled bars) and air temperature (line) at the Aldehuela del Jerte station (40°00′ N, 6°13′ W) in central-western Spain. O, J, A, Jl, October, January, April, July

Full size image
Table 1 Main characteristics of the Retortillo and San Esteban experimental sites in central-western Spain
Full size table

In each site, the experimental design compared two fertilisation treatments (unfertilised vs. fertilised) that were introduced in January 2009 and which were maintained over the next two years. The two treatments were applied in a completely randomised design. Each treatment was applied in eight plots, corresponding to a total of 16 plots in each site. Each circular plot measured ca. 315 m2 (radius of 10 m), with a single Q. ilex at its centre, and a mosaic of pasture and shrub patches around. Plots were separated by buffer bands of 20–30 m width. In each year, fertilised plots received hand-applied inorganic fertiliser in January (25 kg N ha−1, 21.8 kg P ha−1, 62.3 kg K ha−1), and March (55 kg N ha−1). These fertiliser rates were chosen according to Moreno et al. (2007). All plots had not been cropped, fertilised, or tree-pruned during the 10-y period prior to this study.

Foliar sampling and analysis

Pasture was manually harvested (cut at ground level) in early June, corresponding to the peak in green biomass of 2010, to determine the effect of fertilisation on nutrient concentration and pasture dry mass. In each plot, a sample was collected in a 0.5 m × 0.5 m quadrat that was protected from livestock grazing by an exclusion cage (1 m × 1 m at the base × 0.5 m in height). Each protected quadrat was located in the subcanopy area of a randomly selected shrub, which was itself positioned in the tree canopy periphery (ca. 5 m east of the tree base). Pasture samples were oven-dried (at 65°C for 48 h) to determine dry mass.

Current-year shrub and tree shoot samples were collected at the end of July 2010, after shoot elongation and leaf expansion had been completed, to determine the effect of fertilisation on leaf (Q. ilex and Cistus) and cladode (Retama; one cladode = one shoot) dry mass and nutrient concentrations. Ten shrub shoots were collected from each of four randomly selected shrub plants per plot (one per cardinal direction, totalling 40 shoots per plot). Cistus leaves that had been removed from twigs and Retama cladodes were bulked into one composite sample per plot. Forty Q. ilex shoots (10 per cardinal direction) were collected from the mid-crown of each sampled tree (i.e., one per plot), after which leaves were removed from twigs and bulked into one composite sample per plot. Q. ilex and Cistus foliar samples and Retama cladode samples were oven-dried (at 65°C for 48 h) to determine leaf dry mass per shoot and unit cladode dry mass.

Dried pasture, shrub and tree material was ground to pass a 250 μm mesh, and 0.5 g subsamples were wet-digested with H2SO4/H2O2. Nitrogen, P and K concentrations in the digest were measured by the Kjeldahl method, vanadomolybdophosphoric acid yellow method, and atomic absorption spectrometry, respectively.

Data analyses

The two sites (Retama and Cistus) were analysed independently. The effects of fertilisation treatments on dry mass, nutrient concentration and nutrient content of pasture, Q. ilex and Cistus leaves and Retama cladodes were tested by one-way ANOVA. When necessary, data were log-transformed to meet assumptions of normality and homoscedasticity. A P value ≤ 0.05 was considered statistically significant. All tests were performed using Statistica 7.0 (StatSoft, Inc., Tulsa, OK, USA).

Vector analysis was performed to diagnose the nutritional responses to fertilisation, and to identify possible nutrient conditions such as deficiency, sufficiency, dilution and antagonism. Detailed description regarding the development and use of the technique can be found in Haase and Rose (1995) and Isaac and Kimaro (2010). Briefly, the vector diagrams reflect the function that nutrient content in a plant (bottom horizontal axis) is the product of its nutrient concentration (vertical axis) multiplied by its dry mass (top horizontal axis). Changes in these variables relate to two key processes driving plant growth: nutrient uptake and dry matter production that depict differing treatment responses. Responses are expressed relative to the control (here, the unfertilised treatment) that is normalised to 100 to simplify comparisons between treatments and nutrients. Individual responses are integrated into a vector defined by its orientation and magnitude. Vector length represents response magnitude and vector direction identifies diagnostic interpretations and possible nutrient disorders (see Haase and Rose 1995).

Results

Fertilisation significantly increased pasture dry mass by 460% and 1,090% in the Retama and Cistus sites, respectively (Table 2). Fertilisation also induced a significant increase in pasture concentrations of N at both sites, P at Retama site and K at Cistus site, and increased pasture N, P, and K contents at both sites. There were no significant effects of fertilisation treatments on dry mass and nutrient status of Retama cladodes and Cistus leaves (Table 3). At the Retama site, fertilisation significantly decreased K concentration and content of Q. ilex (Table 4). Leaf dry mass and P content of Q. ilex also tended to be lower with fertiliser application (P = 0.06). At the Cistus site, leaf dry mass, N concentration and N content of Q. ilex were higher in the fertilised than in unfertilised treatment. Potassium concentration decreased in the fertilisation treatment and P concentration tended to increase (P = 0.051).

Table 2 Dry mass, nutrient concentration, and nutrient content of pasture species as influenced by N, P, and K application in central-western Spain
Full size table
Table 3 Dry mass, nutrient concentration, and nutrient content of Retama sphaerocarpa cladodes and Cistus ladanifer leaves as influenced by N, P, and K application in central-western Spain
Full size table
Table 4 Leaf dry mass, nutrient concentration, and nutrient content of Quercus ilex leaves as influenced by N, P, and K application in central-western Spain
Full size table

Vector analysis (Fig. 2; Table 2) indicated that pasture species had a N and P deficiency in the unfertilised treatment (shift A, positive shift in dry mass, nutrient concentration and content) at the Retama site (P ≤ 0.04) and N and K deficiency at the Cistus site (P ≤ 0.01). At the Retama and Cistus sites, plant uptake of N, P, and K in fertilised plots was 868–1,840%, 778–1,450%, 732–1,617% higher than in unfertilised plots, respectively (Fig. 2). This ranking and the comparison of vector length indicated that N uptake was most greatly stimulated by fertilizer application at both sites, and was probably the major pasture limiting nutrient. At the Retama site, K concentration and content of Q. ilex decreased (shift B, P ≤ 0.04), which reflects a likely antagonism with another nutrient, most probably N. At the Cistus site, K concentration of Q. ilex also decreased but this occurred with gains in dry mass without significant increases in nutrient uptake, depicting dilution effects (shift C, P ≤ 0.003). Nitrogen dynamics in Q. ilex at the Cistus site followed a shift A vector (P ≤ 0.003), indicating an enrichment response to a nutrient deficiency. Retama and Cistus had N, P and K sufficiency (i.e., no marked differences in dry mass, nutrient concentration and content, P > 0.3). Overall, our results revealed that the magnitude of relative changes in nutrient uptake and dry matter production of pasture species in response to fertilisation was markedly higher than those of Q. ilex, Retama and Cistus.

Fig. 2
figure 2

Vector nutrient diagnosis of Quercus ilex, Retama sphaerocarpa, Cistus ladanifer and pasture species response to fertilisation at the Retortillo and San Esteban agroforestry sites in central-western Spain. All values are normalised with respect to unfertilised treatment values (set to 100), represented by the open circles. Only statistically significant vectors are depicted in each nomogram. Shift A vectors indicate nutrient deficiency; shift B vectors indicate nutrient antagonism; shift C vectors indicate nutrient dilution; all other points reflect nutrient sufficiency

Full size image

Discussion

Dehesas typically have limited growth potential because of low soil fertility (Moreno and Obrador 2007). This assumption was revealed in our study by the fact that leaf nutrient concentrations of Q. ilex in the unfertilised treatment were relatively low. For instance, foliar N concentrations (10.4–10.7 mg g−1) were near the lower values of the range reported by Stephan et al. (1997) for Q. ilex all over Europe: 9.9–21.6 mg N g−1. The oaks studied here also showed low foliar P and K concentrations following ranges reported by these same authors. Our results further showed that fertilisation dramatically increased dry matter production and nutrient uptake (especially N, Fig. 2) of pasture. However, shrub species were not affected by fertilisation, and Q. ilex responded weakly to this treatment, with slight increases in leaf N concentrations and decreases in leaf P and K concentrations. This finding is likely the result of a higher nutrient-use plasticity potential of pasture that may enhance its nutrient competitive ability (Casper and Jackson 1997). This conjecture is supported by results from Aerts et al. (1991), who found that the deciduous grass Molinia caerulea had higher competitive ability at high nutrient supply than the evergreen shrubs Erica tetralix and Calluna vulgaris. In contrast, the low responsiveness of Q. ilex, Retama and Cistus to fertilisation in our silvopastoral systems is probably the consequence of their conserving nutrient use strategy (i.e., growth characteristics of evergreens lead to a low responsiveness to environmental change), which seems to be adaptative for evergreen woody plants in Mediterranean-type ecosystems (Valladeres et al. 2000). The latter nutrient use strategy typically leads to a low potential growth rate and a low competitive ability in nutrient-enriched soils (Aerts 1995; Chapin 1980).

Our previous studies offered presumptive evidence that the shallow-rooted Cistus may compete strongly with pasture and Q. ilex for belowground resources, whereas the sparse deep-rooting Retama may act as a plant “facilitator” (Rivest et al. 2011; Rolo and Moreno 2011). Likewise in the present study, the sign and magnitude of fertilisation effects on dry matter production of pasture and Q. ilex are site-specific and affected by the dominant shrub species. For example, vector analysis revealed that pasture N enrichment response to N limitation at the Cistus site was several orders of magnitude higher than that at the Retama site, where soil mineral N was higher (Table 1). The more nutrient-limited an individual is, the more its production increases in response to a large addition of the limiting nutrient (Chapin et al. 1986). Moreover, our results also highlighted that N was a major determinant of Q. ilex growth limitation in the unfertilised treatment at the Cistus site, but not at the Retama site. We believe this difference reflects a stronger soil N depletion potential of Cistus compared to the N2-fixing Retama. For example, the two-year average accumulation of inorganic-N in Retama microsites was found to be 5.3 times higher than that in open areas lacking Retama (Delgado-Baquerizo et al. 2010). Foliar analyses further suggest that, of the nutrients made readily available through the addition of fertilisers (i.e., N, P and K), only N at the Cistus site had been taken up by Q. ilex. This may result from a higher mobility potential of N compared to P and K (Lehmann and Schroth 2003), combined with weak root length density of Q. ilex in the uppermost soil layer (i.e., low ability of Q. ilex to take up the less mobile P and K) (Moreno et al. 2005). Q. ilex had lower dry matter production (Retama site) and foliar K and P concentrations (both sites) in the fertilisation treatment, thereby suggesting that management practices oriented to improve soil chemical fertility may have positive effects on pasture understorey productivity and quality, but with potentially detrimental consequences for long-term productivity on tree overstorey.

A worldwide fundamental issue that currently faces extensive silvopastoral systems, including Iberian dehesas, is the progressive loss of tree cover resulting from the lack of natural tree regeneration (Gibbons et al. 2008). Shrub encroachment has been proposed as an effective means to resolve this issue and ensure the long-term persistence of dehesas (Pulido et al. 2010). Nevertheless, management of shrub-encroached silvopastoral systems remains crucial to soil nutrient dynamics and overall system productivity (i.e., sum of yields for both pasture and tree components). Our findings address this concern and suggest that a dense cover of Cistus may affect more severely and negatively the growth and nutrient uptake (especially N) of Q. ilex than scattered patches of Retama. Partial clearing of Cistus in dehesas (i.e., retaining only some patches) could be therefore a profitable means of minimising soil nutrient competition, but this approach needs further research. At the Retama site (richer in soil mineral-N), an important pasture limitation in N (and other nutrients) indicated a limited capacity of the N2-fixing Retama to supply N to pasture understorey (Rodríguez-Echeverría and Pérez-Fernández 2003). Future research is needed to compare the effects of several fertilisers with varying solubility and composition on growth and nutrition of pasture and trees in dehesas that differ in soil fertility and shrub species understory composition. This would help our understanding of the nutrient foraging ability of different co-occurring plants in multilayered silvopastoral systems, and planning of fertilisation regimes that may optimise pasture and tree productivity.