Effects of abandoning long-term goat grazing on species composition and species richness of pastures at La Gomera , Canary Islands

Goat grazing management is a powerful ecological force structuring plant communities which effects are highly different depending on factors such as climatic conditions, grazing intensity, type of plant community, etc. Pastures in the Canary Islands have important heritage, socioeconomic and landscape values because they are the remains of traditional livestock management and a key element of local economy. An experimental study of goat grazing exclusion has been carried out in an area of La Gomera Island to determine the main effects of grazing abandonment on species richness and species composition. The species composition of La Gomera pastures is more sensitive to grazing than to climatic variability, though effects in the species richness in relation to grazing were not detected. Because the pasture origin in this area involved artificial, removing of non useful plants in order to provide land for agriculture more than 300 years ago, and using it for pastures around 80 years ago, the sustainable use has maintained this rich plant community. Grazing abandonment can have a negative effect on the pastures’ diversity, as grazing has been a strong force in the maintenance of certain species. Public promotion of this primary activity is suggested to allow the continuation with minimal effects on the vegetation composition. This will help to conserve the high diversity of this area and the promotion of natural values as well as traditional and sustainable human activities. Additional key words: detrended correspondence analysis; plant community; rarefaction; sustainability.


Introduction
Goat grazing management has been a traditional sustainable land use in many areas of the planet (Milchunas et al., 1988;Crawley, 1997) requiring correct grazing management to maintain the species composition and the high diversity values of the plant communities of grazing lands (Baldock et al., 1994;Olff and Ritchie, 1998).In spite of this traditional and sustainable use, the effects cause remarkable and significant variation in species composition (Casado et al., 2004;Arévalo et al., 2007).
Grazing effects on species richness and species composition are contradictory in many studies (Olff and Ritchie, 1998) and there are studies concluding that herbivores can enhance, have weak effects or even have a negative impact on plant diversity and richness (Osem et al., 2002;De-Bello et al., 2007) as well as different effects on soil nutrients (McIntosh et al., 1997;Bakker et al., 2004;Peco et al., 2006;Fernández-Lugo et al., 2009).The lack of consistent responses has been attributed to factors such as the evolutionary history of grazing, productivity gradients or grazing intensity (Milchunas and Lauenroth, 1993).The high context dependence of grazing effects on ecosystems highlights the need to carry out studies that allow us to identify the intrinsic dynamics of the pastures in each region (Perevolotsky and Seligman, 1998;De-Bello, 2006).
Pastures in the Canary Islands have an anthropogenic and recent origin, but they have important heritage, socioeconomic and landscape values because they are the remains of traditional livestock management and a key element of the local economy (Mata et al., 2000;Arévalo et al., 2007).Canarian pastures are fragmented and of modest size (González et al., 1986).They are very rich plant communities, something that can be attributed to their history, management and dynamics (Arévalo et al., 2007).These ecosystems are grazed by goats, which are the most common grazer animal in the archipelago (Bermejo, 2003).These plant communities are an important traditional economic resource exploited by the local population.However, as experienced by industrialized countries, changes in the farming industry have led to these extensive livestock systems either being intensified or abandoned (García Dory and Martínez, 1988;Marrero and Capote, 2001).This process is not concluded yet (Bermejo, 2003) and the abandonment of this traditional activity becomes more evident every day.
The study of grazing abandonment effects has potential management implications (setting the maximum number of animals, suggesting promoting the traditional activity in the protected area...), environmental (plant and animal protection, exotic species...) and socioeconomic importance, and is necessary to achieve conservation of these ecosystems.The main hypothesis of this study is that grazing abandonment will significantly change species composition and species richness in these pastures, and the second hypothesis is that climatic variability will drive important changes in species composition from year to year.The results will provide useful information that will help managers to continue with this activity, which is of high cultural value at present.

Study site
Field work was carried out in two natural protected areas, Valle Gran Rey Rural Park and Natural Monument of Lomo del Carretón, located in the south western part of the island of La Gomera (28°02' N; 17°12' W), Canary Islands (Spain).Total studied area comprises 2,193 ha; from 0 m a.s.l. to 900 m a.s.l., with deep ravines and few plains.The area is characterized by a Mediterranean climate, with a dry season in summer and a humid season in winter and spring.The long-term  mean annual precipitation is 138.8 ± 23.2 mm and mean annual temperature is 14.8 ± 0.7°C (Marzol, 2000).During the study period, 2003 was a drier year (103.4 mm), 2001-2002 and 2004 received close to the mean, and 2004 received the highest rainfall (377.1 mm).The mean rainfall in the studied period was higher (243.4mm) that the long-term mean (237,209,103,377,289 mm respectively for each year).
Volcanic rock forms the majority of the area, included as recent basalts (Villalba and Santana, 2000) and the soil have been considered as brown soils and litosoils (Rodríguez and Mora, 2000).
Natural vegetation in the area varied from sweet spurge scrubland (Euphorbia balsamifera) covering the lowest parts of the slopes (0-400 m a.s.l.) and «cardonal» or E. canariensis shrubland growing above (400-600 m a.s.l.), to the upper parts which were dominated by thermophilous vegetation (600-900 m a.s.l.) mainly represented by the tree Juniperus turbinata.In the beds of ravines the palm tree Phoenix canariensis formed palm grooves (Del-Arco et al., 2006).Little of the natural vegetation remains today, mainly due to the deforestation of the thermophilous woodland to provide wood and open lands for agriculture and grazing that become an important activity around 80 years ago.
At present, main plant communities of both Natural Protected Areas are the coastal shrubland (0-600 m a.s.l.) extended in the previous distribution of natural shrubland and dominated by E. berthelotii and other shrubs; and pastures occupying abandoned terraces, which are abundant in both areas and cover the altitudinal gradient, composed of grasses (Stipa capensis, Hyparrhenia hirta, Brachypodium distachyon, Bromus spp., etc.), legumes (Medicago spp.and Trifolium spp.) and other forbs (Notoceras bicorne, Cynara cardunculus, Plantago lagopus, Ajuga iva, Convolvulus althaeoides, etc.).Fragments of palm grooves and E. canariensis scrublands persist in both areas, whereas small stands of E. balsmifera and J. turbinata forest are present in Valle Gran Rey and Lomo del Carretón respectively (Del-Arco et al., 2006).However, these natural communities were not included in the study due to their inaccessible location or in the case of palm grooves, because they have been traditionally kept for agricultural uses different from grazing.
Grazing areas of this study cover up to 1,418 ha where 31 farmers practice mainly goat grazing with animals belonging to the Canarian breeds.Mean stocking rate is 0.21 ± 0.18 animals ha -1 day -1 , using mainly a continuous grazing system.There are no cultivated pastures in the Canary Islands, and all the study areas were spontaneous pastures that appeared after removing the shrubland.This area includes 775 ha on which grazing had been abandoned for more than 10 years.Boths grazed vs. ungrazed areas are similar in terms of vegetation composition, geology, soil and land use (except for grazing management).

Design of the study
A total of 41 permanent transects, each 30 m long, were located randomly in the study area, with point samples taken every 0.3 m; i.e. 100 points in every transect (Daget and Poissonet, 1971).Twenty-five transects were distributed in grazed areas and 16 in nongrazed areas following a stratified sampling (60% of the area was under grazing management; Fig. 1).
These abandoned areas were similar on soil characteristics (Rodríguez and Mora, 2000) and the abandonment was more related with the gradually movement of rural population to urban areas.Altitude of transects is not significantly different among transects in grazed areas and in abandoned areas (Student t = -5.69,for n = 1 and p > 0.05) for our study area.Five transects were lost in 2001 (4 in grazed vs. 1 in non-grazed), two in 2002 (1 in grazed vs. 1 in non-grazed) to 2004 and one in 2005 (grazed transect) due to some location problems and vandalism.Also, as it can be seen in Figure 1, the abandonment is patchily organized.Because of that, these areas are comparable in order to environmental conditions.
All transects were georeferenced, with fixed UTM coordinates and a fixed orientation for each, in order to make possible the recording of data from the same location, every year, with high precision and to reduce sampling error.
The presented results are base on measurements taken between the years 2001 and 2005.Species composition and frequency was measured using the pointquadrat intersect method (Daget and Poissonet, 1971), recording 100 point samples in each transect.Species relative frequency was calculated as the number of points in which a species appears with regard to the sum of all points in which species are found in the transect (total of points touched by a species divided by the sum of total touches of all species; Bullock et al., 2001;Del-Pozo et al., 2006): ΣP jk where SCP ijk = specific contribution by presence of a species i in the transect j in year k; P ijk = points in which species i appears in transect j in year k; and P jk = points in which any species appear in transect j in year k.

Statistical analysis
Rarefaction represents a powerful statistical method to calculate the expected number of species as a function of the sampling effort.In this study, effort is quantified in terms of the number of transects for which the species abundance was calculated using the SCP values.Species rarefaction curve was calculated (Hurlbert, 1971;Heck et al., 1975) for transects, separating grazed and non-grazed areas at the different years.Based on the rarefaction curves, the number of species were compared for grazed and non-grazed transects, using the highest possible sampling intensity for the five years (12 transects), with a non-parametric Wilcoxon test (for a p < 0.05, n = 5).

Effects of abandoning goat grazing on species composition and richness
We analyzed species richness of transects as a dependent variable with respect two factors: grazing vs. ungrazed and the sampled year, and added altitude as a covariable (as long as it an important variable affecting species richness) using an analysis of covariance (ANCOVA, p < 0.05).Normality of the data was checked with the Komogorov-Smirnov test, and homoscedasticity of the data was examined with a Levene contrast (for p < 0.05).The post-hoc Tukey tests was used to detect signif icant differences among groups found different overall (p < 0.05).Due to the large set of statistical contrasts a multiple test procedure of Holm for independent tests (Legendre and Legendre, 1998) was also applied.
Detrended correspondence analysis (DCA) (Hill and Gauch, 1980) was used to examine how species composition differed among transects and changed over time in the different sampling years using the ordination package CANOCO (Ter-Braak and S ˘milauer, 1998).Analyses were based on species abundances calculated as SCP.Other ordination techniques are based on sample distances (multidimensional scaling) or similarities to composed groups (classification).However, the best way to analyze species composition is gradient analysis, as long as is the canonical answer of the species to environmental gradient (Gauch, 1982).In the case of the DCA, this technique improves other gradient analysis techniques, as long as correct the compression of scores of samples at the end of the axes and the quadratic relationship of firsts axis with the second (Jongman et al., 1987).
To determine if DCA axes are able to discriminate the species composition based on the treatments (grazed vs. ungrazed, the dependent variable), the coordinates of transects from the DCA for axes I and II were analyzed using logistic regression (using the statistical of Wald for a p <0.05).The DCA coordinates of axes I and II for transects were correlated with the sampling year of that transect to determine if the axes' gradients are related to the annual variability, and also these axes

Canary Islands
coordinates were correlated with the annual precipitation of the year (using the Pearson correlation coefficient, for p < 0.05 and n = 192).In order to reveal the relationship of the DCA transects' scores (axes I and II) with altitude, the Pearson correlation coefficient was calculated among these variables.Altitude is the strongest environmental gradient in the occidental Canary Islands, which are characterized by strong altitudinal gradients as occur in La Gomera (Fernández-Palacios and De-Nicolás, 1995;Arévalo et al., 2005).Basic statistical methods according to Zar (1984) were implemented using the SPSS computer package (SPSS, 1997).

Results
A total of 214 species (Appendix A) were found throughout the study, but the number of species changed from year to year (73, 119, 132, 115 and 147 from 2001 to 2005 respectively).
The generated rarefaction curves to discriminated between grazed vs. abandoned transects (Fig. 2), showed not significant differences in the number of species at a similar sampling effort (Z = -1.483,for n = 5 and p > 0.05), using the Wilcoxon test, for both the highest possible sampling effort (12 transects) and for sampling each year.
Once tested the normality and homocedasticity of the data, the ANCOVA revealed similar results as the rarefaction analysis.Grazing effect on species richness was revealed not significant (F 1,193 = =0.722,p > 0.05) as well as the interaction grazing vs. year (F 4,193 = 0.722, p > 0.05), while the year of sampling was a significant effect on species richness (F 4,193 = 0.722, p < 0.01).The Tuckey test indicated two groups of years: 2001-2004 vs. 2002, 2003 and 2005 that differed significatively (p < 0.05).
The DCA analysis revealed an important variability based on the species relative frequency, with almost 7 SD (standard deviation units) for axis I and 5 SD between transects for axis II, indicating that there are no species in common between many transects.Discrimination based on the sampled years showed not significant differences in any of the axes (Fig. 3).The Pearson correlation coefficient of the years vs. transect scores for axes I and II did not reveal significant correlation (for axis I r = 0.105, p > 0.05 and for axis II r = -0.071,p > 0.05, in both cases n = 193), meaning that the species gradient in species composition revealed by both axes cannot be attributed to the differences in year.The same analyses but using annual precipitation of the year reported similar results (for axis I r = -0.058,p > 0.05 and for axis II r = 0.079, p > 0.05, in both cases n = 193).
Discrimination based on grazed vs. non-grazed (Fig. 4) using a logistic regression revealed no relationship between the scores of transects in axis I and the treatment (B = -0.347,Wald statistic = 4.207, p > 0.05), but the relationship between the transect scores of axis II and the treatment was revealed significant (B = -1.291,Wald statistic = 25.857,p < 0.01).When we correlated the axes' scores with the altitude of transects, axis I was highly related to altitude (r = 0.777, p < 0.001)

Effects of abandoning goat grazing on species composition and richness 117
Figure 2. Mean values of species numbers obtained by rarefaction for each year and management treatment (grazed vs. ungrazed).The vertical line indicated the minimum sampling effort for which value differences in species richness were compared in grazed areas (S) vs. non-grazed (N).while this relationship was not significant for axis II (r = -0.136,p > 0.05 and n = 193 in both cases).

Discussion
Previous studies of grazing effects on plant species composition and species richness have traditionally been inconsistent and conflicting in their results, lacking a general model that predicts the response of grazing intensity or abandonment (Olff and Ritchie, 1998;Peco et al., 2006).Although fluctuations in total annual rainfall in Mediterranean areas has been considered as one of the most important parameters determining changes in pastures species composition (Figueroa and Davy, 1991;Espigares and Peco, 1995), this effect was not found important.Over the last 5 years of the study, precipitation ranged from 103 mm to 377 mm yr -1 , but it did not have a significant effect on species composition in function of the year (Fig. 3).
In contrast, it was found that goat grazing had a significant effect on species composition (Fig. 4) but not on species richness (Fig. 2), analyzing these with rarefaction and with statistical contrast.Many studies have demonstrated that grazing has an impact on species richness (Olff and Ritchie, 1998), although it is dependent on spatial-scale and highly related to climate variability (Huston, 1994;Cingolani et al., 2005) and resources availability (Milchunas et al., 1988;Osem et al., 2002).In arid environments such as the study area, it is unlikely that species richness increase in grazed areas because plant growth and diversity would be limited rather by soil resources than aboveground competition, therefore grazing did not reveal any effect on species richness (Olff and Ritchie, 1998).
Altitudinal gradients are very important in the distribution of species (Hubbard and Wilson, 1988;Ullman et al., 1995), particularly in islands with a strong altitudinal gradient like Tenerife or Gran Canaria (Arévalo et al., 2005).A strong and significant effect of altitude on species distribution was found over the 750 altitudinal gradient of transects.In spite of this gradient, it did not obscure the variation produced by grazing management in species composition.The species composition of pastures on La Gomera was more sensitive to grazing than climatic variability, but this study did not detect effects in species richness in relation with grazing.The climatic variability, as expected, was an important variable also determining species composition.The pastures origin in this area is related to artificial removing of non useful plants more than 300 years ago (González et al., 1986).Consequently, grazing abandonment can have a significant effect in the pastures' diversity, since grazing is a strong force in the maintenance of some species.Although is true that the dominant group of species are ruderal, some endemic species of La Gomera (Euphorbia bertelotii) and endemic species of the archipelago (Micromeria varia, Echium aculeatum or Kikcia scoparia) appeared in the inventories.Because of that, we suggest promotion of this traditional activity to conserve natural values as well as traditional sustainable human activities.
122 J. R. Arévalo et al. / Span J Agric Res (2011) 9(1), 113-123 Appendix A. Species list of the 214 species found in the survey sorted by their importance value (calculated as the percentage of presence in all the transects at different years).Nomenclature follows Izquierdo et al. (2004)

Figure 1 .
Figure 1.Canary Islands, showing the Valle Gran Rey Rural Park and Natural Monument of Lomo del Carretón and in grey indicating the grazed zones of the two protected areas.

Figure 3 .Figure 4 .
Figure3.Detrended correspondence analysis (DCA) transect scores for species richness by year of sampling.Polygons enclosed 90% of the transects using the minimum possible area (each year polygon for each year indicated with an arrow).Eigenvalue of axis I: 0.295 and axis II: 0.202, with a 35% of total inertia explained by both axes.

Figure 5 .
Figure5.Detrended correspondence analysis (DCA) coordinates of the species composition per transect.Species are labelled with the first four letters of the genus followed by the first four letters of the specific epithet (as appears in appendix A).
Izquierdo et al. (2004) species found in the survey sorted by their importance value (calculated as the percentage of presence in all the transects at different years).Nomenclature followsIzquierdo et al. (2004)