Effect of pine reforestation associated with soil disturbance on ant assemblages (Hymenoptera: Formicidae) in a semiarid steppe

Soil and changes in vegetation may affect ant assemblages, but the relative importance of each in different habitats is not well characterized. In particular, information on the effects of ecological restoration on arthropods is scarce. It was decided, therefore, to study how reforestation may affect an ant assemblage. Ants were sampled in area that had been reforested and adjacent grassland using pitfall traps. Soil surface and vegetation were characterized. The disturbance of the vegetation caused by reforestation resulted in a decrease in the cover of Stipa tenacissima and Cistaceae and an increase in the cover of pine. The mechanical preparation of the site also resulted in changes in the soil surface, with an increase in the cover of stones and rocks. Ant species richness and abundance were greater at the reforested site than in the grassland and more species showed a positive than a negative response to reforestation. The underlying causes of this pattern are mainly related to changes in vegetation and structure of the soil surface and are associated with the increase in the cover of pine, which most probably provided additional food resources, and the greater cover of stones and rocks that provided more shelter for the ant assemblage.


INTRODUCTION
Grasslands dominated by esparto grass Stipa tenacissima L. with a sparse cover of vegetation, are one of the commonest ecosystems in semiarid regions in the Western Mediterranean basin (Barberá et al., 2006;Maestre et al., 2007;Cortina et al., 2009).S. tenacissima is a perennial tussock grass in which the live biomass is usually surrounded by thick layers of standing dead grass (Maestre et al., 2007).The physiognomy of S. tenacissima grasslands is more similar to shrub land than to mesic grassland.The effect of grazing on S. tenacissima is low because this grass has a low palatability, but these grasslands have been intensively managed for producing fi bre (esparto) used for making baskets, shoes, mats, etc, from prehistoric times until the 1950-1960s.Although this practice has now been mostly abandoned, the abandonment has not resulted in any major changes in the habitat (Barberá et al., 2006;Maestre et al., 2007).Many authors consider esparto grass steppes to be degraded former oak or pine open forests or meteorological station of La Carrichosa (6.22 km S, 254 m a.s.l., data 1995-2012) the following annual mean values were recorded: temperature 17.2°C, relative humidity 64%, rainfall 253 mm and evapotranspiration 1166 mm (SIAM-IMIDA, 2017).
The vegetation in the study area is steppe grassland (Fig. 2), dominated by the esparto grass (Stipa tenacissima).There was a partial reforestation with Pinus halepensis in the 1960s and 1970s, although the trees developed poorly, probably due to rooting diffi culties associated with the petrocalcic horizon.In July 1994, the study area suffered a catastrophic wildfi re.Our study site is located close (1-2 km) to the eastern limit of the fi re.At the end of 1996, the area was patchily reforested again using P. halepensis.The planting of pines was preceded by mechanical preparation of the soil by subsoiling to 1-1.5 m along the plantation rows, causing a signifi cant disturbance both to the soil surface and profi le.One-year old tree seedlings were planted in rows and spaced 2.5-3 m apart.The 19-ha plot studied was an experimental plot, which was subjected to a large number of different mechanical subsoiling techniques, which differed mainly in the way organic matter was added to each line of trees (for details see Barberá et al., 2005).Pine growth was slow and mortality high.After 16 years, the trees were rarely more than 3 m high and their cover sparse (Fig. 3).Mechanical preparation of the terrain combined with the presence of a shallow petrocalcic horizon resulted in large slabs of rock and stones being brought to the surface, creating a particular microtopography.Original soil surface effects of MSP on arthropod communities (Bellocq et al., 2001;Bird et al., 2004).
As in other arid and semi-arid areas on the Iberian Peninsula, esparto grass steppes are inhabited by a great diversity of invertebrates.Two insect groups dominate: Coleoptera (mainly Tenebrionidae) and, more specifi cally, Formicidae (Doblas-Miranda et al., 2007;Piñero et al., 2011).Ants are dominant organisms in terrestrial ecosystems, have very important functions, make up a great part of the animal biomass and are ecosystem engineers (Hölldobler & Wilson, 1990;Folgarait, 1998;Read & Andersen, 2000;Frouz & Jilkova, 2008;Del Toro et al., 2012).
Vegetation and soil may infl uence ant assemblages.Shrub encroachment of grasslands results in an increase in ant diversity and/or abundance, both in mesic and semiarid grasslands (Nash et al., 2000;Bestelmeyer, 2005;Azcárate & Peco, 2012;Wiezik et al., 2013).However, the reforestation of semiarid grasslands by pines may reduce the abundance and diversity of ants, although this effect is much lower when tree cover is sparse (Corley et al., 2006).Soil characteristics may be especially relevant to understanding ant assemblages in semiarid landscapes (Bestelmeyer & Wiens, 2001).Nevertheless, there are very few studies and little current literature that deals with the direct and indirect effects of soil properties on ant assemblages.In the current study, MSP simultaneously disturbed both the soil and vegetation, which provided the opportunity to study the effects of both factors on ant assemblages.
We studied the effects of a 16-year old reforestation with P. halepensis of a semiarid S. tenacissima grassland on the local ant assemblage, looking specifi cally at the effects of changes in the soil and vegetation.The objectives of this study were: (i) to determine how reforestation following MSP affects the structure of the surface of the soil and vegetation; (ii) to determine how reforestation affects an epigeal ant assemblage and species, particularly their diversity and abundance; (iii) to determine which changes at the surface of the soil and in the vegetation are likely to cause changes in the ant assemblage.

Study Site
This study was carried out in the Sierra del Picarcho (Region of Murcia, south-eastern Spain, 394 m a.s.l., 38°20´N, 1°29´W; Fig. 1).Soils are mostly Petric Calcisols (FAO-ISRIC-ISSS, 1998), characterized by an A-Ckm1-Ckm2 profi le, with a thick petrocalcic horizon (hardpan) at a depth of 40-70 cm.The topography of the area is mostly fl at, with slopes not exceeding 6% (Barberá et al., 2006).The climate is semiarid Mediterranean.At the nearby   was smooth with a varying density of small stones on the surface, while the subsoiled area is anfractuous at the microscale with a lot of large stones and rocks at the surface, which resulted from the breaking up of the hardpan during subsoiling.

Sampling
Three 200-m transects were established in May 2012 (Fig. 4).Each transect consisted of 20 pitfall traps set 10 m apart.Ten traps were placed in the reforested area, and 10 in the grassland.Because of the patchy structure of the reforested area, the distance of a pitfall from the edge of the grassland/reforested area may be less than the distance to the middle point of the transect (Fig. 4).
The ants were sampled using pitfall traps.Polystyrene tubes (2 cm diameter, 10.5 cm long) were installed two weeks before the traps were set in order to avoid the effects of digging-in (Greenslade, 1973), with the lip of the tube fl ush with the soil surface.Tubes were fi lled with 5 ml of 50 : 50 propylene glycol and water solution with a few drops of liquid soap added to reduce surface tension (Bestelmeyer et al., 2000;Calixto et al., 2007).Pitfall traps were set for one week each season, in May, August, November 2012 and February 2013.
All ants were identifi ed and counted using a stereo microscope.Identifi ed ants were transferred to vials containing 70% ethanol, or were mounted.All specimens are deposited at the premises of Asociación de Naturalistas del Sureste (ANSE, Murcia, Spain).
To characterize the vegetation and soil surface one transect was established alongside each transect with the pitfall traps in March 2013.The purpose of these transects was to record the general characteristics of the composition of the vegetation and structure of the soil surface.Each transect was 8-m long, centred on each pitfall point and aligned with the original pitfall transect.At 17 points, at 50-cm intervals, type of soil surface (cover of soilno stones or rocks on soil surface; superfi cial stones -hereafter stones; or in-ground rocks -hereafter rocks), species of plants and their height were recorded.Only perennial species were included in the analysis.Plants were categorized into functional groups (Table 1).

Data analysis
In order to determine the effects of reforestation on the ant assemblage and infer the possible causes we carried out two analyses.The fi rst was for the whole year at the level of the assemblage.Numbers of each species caught in each pitfall on the four sampling sessions were summed and log-transformed.This matrix was submitted to non-metric multidimensional scal-ing (NMDS) followed by an Adonis test (McArdle & Anderson, 2001) to check if the between-group (grassland vs. reforestation) and within-group similarities differed signifi cantly.Adonis is a permutation test that partitions the sum of squares for distance matrices in a way analogous to MANOVA.The number of permutations was set to 999.Relation between NMDS axes and environmental variables was tested using Spearman's rank correlation.These analyses were carried out using the vegan package (Oksanen et al., 2018) in R (R Core Team, 2018).The second analysis aimed to test the effects on the abundance of individual species as well as the abundance of the assemblage (all species summed) and species richness.It was hierarchically carried out in two steps: (i) testing if reforestation using MSP is associated with changes in soil and vegetation variables as well as in ant species richness and ant total abundance for each species; (ii) testing if ant species richness, total abundance and species specifi c variation were associated with the variation in the environmental variables that were signifi cantly different in the two habitats, as tested in (i).
We hypothesized that these changes may be expressed in terms of three possible models (Fig. S1): (i) a shift in the levels of the different variables of interest in each habitat (factor); (ii) a linear trend (positive or negative) in the variable from inner part of the reforested area to that of the grassland; (iii) a unimodal trend with a maximum or a minimum value of the variable of interest close to the ecotone.In relation to environmental variables, to test for the most complex case (unimodal response) a generalized linear mixed model was fi tted to each variable with transect as a random factor and signed distance to the ecotone set to zero and distance to inner part of the grassland assigned a positive sign and that to the inner part of reforested area a negative sign and its squared value as fi xed factors.If the effect of the quadratic signed distance Table 1.Environmental variables and error function used for modelling response of the environmental and ant assemblage variables to reforestation in the framework of Generalized Linear Mixed Models (see Material and methods).Tree and perennial grass functional groups were not tested as each one is represented by only one species (Pinus halepensis and Stipa tenacissima, respectively).Note that error functions for total vegetation and functional groups were tested using a Poisson error as species in the same functional group can be overlaid on the same sampling point.was statistically signifi cant (indicating a unimodal response) the model was retained, otherwise we tested the alternative (less complicated) linear trend model with the signed distance without the quadratic term.Again, if the signed distance parameter was signifi cant the model was retained, otherwise we tested for the shift model, which includes the habitat as a qualitative factor.
For each response variable, the generalized linear mixed model was fi tted with an error function appropriate for the nature of data (binomial, Poisson, normal; details in Table 1).Models for ant species richness and total abundance were tested similarly but because ants were sampled seasonally the fi xed part of the generalized linear mixed model includes habitat and season.Ant species richness was tested at two spatial scales: pitfall and habitat × transect (10 aggregated traps in a particular habitat along a particular transect).For the species' abundance models, as autumn and winter abundances per individual species were very low, a full model, including habitat and season, is usually over parametrized.Therefore, for each species, we tested a model that included the year-round aggregated abundance and four seasonal models.Abundance data were log-transformed prior to analyses.Generalized linear mixed models were fi tted using the GLIM-MIX procedure in SAS 9.4.Details of the statistically signifi cant models are presented in Table S1.In the main text only graphical expressions of the models are shown.
Next, we determined whether the abundance and richness of ant species was correlated with environmental variables that signifi cantly varied between habitats, using Spearman's rank correlation with the Benjamini-Hochberg correction for multiple comparisons (Benjamini & Hochberg, 1995).This correction is less conservative than the Bonferroni one and is based on controlling the false discovery rate (FDR).We set FDR in this study to 0.2, that is, we accept a maximum rate of false tests of 20%.

Environmental changes induced by reforestation
Soil surface characteristics clearly vary between habitats (Figs 5-6; expected values of models are shown in Table S1).The cover of rocks was low and constant in the grassland and increased sharply towards the inner part of the reforested area.The cover of soil increased linearly from the centre of the reforested area to that of the grassland, while stone cover peaked close to the ecotone.Changes in vegetation were not associated with differences in species richness or general physiognomy, but were with total vegetation, functional groups and, most importantly, species cover (Fig. 7; predicted values of the models are presented in Table S1).Of course, pine cover was only recorded in the reforested area.The two subdominant shrubs showed opposite linear trends.In the inner part of the grassland the cover of Cistus clusii and Rosmarinus offi cinalis was similar, whereas that of C. clusii decreased towards the inner part of the reforested area while that of R. offi cinalis increased.As a result, the total shrub cover remains unchanged.Stipa tenacissima was clearly dominant   S1).S1).Cisclu -Cistus clusii; Fumthy -Fumana thymifolia; Pinhal -Pinus halepensis; Rosof -Rosmarinus offi cinalis; Stiten -Stipa tenacissima.
in grassland but far less abundant in the reforested site, to the point that in the inner part its cover was similar to that of R. offi cinalis.Total chamaephyte cover had a similar trend, but only one species of chamaephyte differed signifi cantly between habitats: Fumana thymifolia.The cover of this tiny species was very low except in the inner part of the grassland.Interestingly, total perennial plant cover was similar in the inner parts of the forest and grassland but showed a minimum in the ecotone close to the reforested area.

Effect of reforestation on ant assemblages
A total of 13 species of ants belonging to 11 genera were recorded in this study (Table 2).Ten species were found in both habitats and three only in the reforested habitat.As Temnothorax racovitzai was only caught twice it was not included in the analysis.
Species richness at the habitat × transect scale was signifi cantly higher in the reforested area (Fig. 8).The pattern did not differ between seasons.At the pitfall scale it is possible to detect a linear trend of decreasing richness from the inner part of the reforested area to that of the grassland (Fig. 9; predicted values of models are presented in Table S1).Total ant abundance follows a very similar pattern (Fig. S2; predicted values of models are presented in Table S1).
Reforestation had signifi cant effects on the year-round abundance of several species and in spring and summer, when ants are much more active (Figs 10 and S3; predicted values of models are presented in Table S1).Plagio lepis schmitzii and Gonioma blanci were the only species that linearly increased in abundance from the inner part of the reforested area to that of the grassland.Plagiolepis schmitzii shows this trend year-round and in summer, whilst G. blanci shows this trend only in spring.More species decrease in abundance from the inner part of the reforested area to that of the grassland: Aphaenogaster iberica yearround and in both spring and summer, Camponotus sylvaticus year-round and in summer.For Camponotus foreli the trend also linearly decreased from the reforested area to the grassland year-round and in spring but in summer the response was unimodal with a maximum in the ecotone.S1.
This kind of unimodal response is also characteristic of Tetramorium semilaeve year-round and in summer.

Correlation between the differences in the environmental factors in the two habitats and in the ant assemblages
There were signifi cant correlations between environmental variables and the ant assemblage variables that differed signifi cantly in the two habitats (Fig. 11, Table S2), as stated in the previous section.Ant species richness yearround and when most active in spring and summer was positively associated with the cover of P. halepensis, also species richness year-round and in summer was associated positively with the cover of stones.In autumn ant species richness was positively associated with the cover of rocks.Year-round ant species richness was negatively associated with C. clusii and a continuous cover of soil.
Total ant abundance year-round was positively associated with the cover of stones, as was total ant abundance in summer.As for ant species richness, seasonal total ant abundance in autumn was positively associated with the cover of rocks.In contrast, total ant abundance was negatively correlated year-round with a continuous cover of soil and in summer with the cover of F. thymifolia.
In respect to specifi c species, G. blanci year-round and spring abundances were positively associated with the cover of chamaephytes and with that of F. thymifolia.Aphaenogaster iberica year-round and summer abundances were positively associated with the cover of P. halepensis, like the abundance year-round and in spring of C. foreli.Also for C. foreli and the same seasons there was a negative association with the cover of C. clusii.Camponotus sylvaticus, however was positively associated with the cover of rocks in spring and negatively with a continuous cover of soil.
NMDS axis 1 clearly separated Crematogaster auberti from the rest of the species, but not the pitfall trap catches by habitat (Fig. 12).In contrast, axis 2 segregated the reforested area from grassland, although there was a considerable overlap between groups.Nevertheless, the Adonis test shows signifi cant differences between pitfall trap catches in the reforested area and grassland (pseudo-F = 3.18, p = 0.013) but the partial R 2 of the habitat factor was low (5%).NMDS axis 1 was only signifi cantly correlated with perennial plant richness (Table 3; positive correlation), while NMDS axis 2 was positively correlated with the cover of stones and Pinus halepensis and negatively with Sideritis bourgeana, chamaephytes and a continuous cover of soil.

Environmental changes induced by reforestation
Vegetation and soil were disturbed by reforestation.We did not fi nd a shift but mostly linear trends of different signs depending on the variable.Without doubt, this is related to the detailed spatial scale of the study, itself conditioned by the patchy design of the reforested area.
The disruption caused by mechanical soil preparation removed a large proportion of the dominant species of the grassland (Stipa tenacissima), which resulted in a reduction in its cover of > 60% in the inner part of the refor-   ested area compared to that in the inner part of the grassland area (Fig. 7).In our study area, the rate of emergence of seedlings of S. tenacissima is very low due to the high predation of seeds (by ants) and no germination on bare soil, with only germination occurring within tussocks of S. tenacissima (Barberá et al., 2006).Thus, the recolonization of reforested area by this species is very slow.The introduced Pinus halepensis developed poorly (Barberá et al., 2005) but in the inner part of reforested area its cover was similar to that of S. tenacissima.The abundance of the two subdominant shrubs differed in the two habitats.That of Cistus clusii and other Cistaceae was lower in the reforested habitat probably because their seed banks only become highly active after fi res (Thanos et al., 1992).In contrast, Rosmarinus offi cinalis increased in the open spaces left by the removal of S. tenacissima during reforestation, as this species has an active seed bank and good colonization potential.In summary, reforestation greatly changed the structure of the grassland, reducing especially the cover of S. tenacissima and Cistaceae, which were the main taxa in the original grassland.On a more detailed scale the changes are gradual, not abrupt.Neither plant species richness nor the general physiognomy of the vegetation (vertical structure and overlaying) was signifi cantly affected, as the poor growth of P. halepensis resulted in a shrub-like structure not very different from that of the original grassland.Soil surface was changed greatly by reforestation.The mechanical preparation of the site broke through the petrocalcic hardpan and brought large numbers of rocks to the surface of the soil.Within our study area, the surface of soil in the grassland characteristically consisted mainly of a continuous cover of soil with a few rocks on the surface and moderate abundance of stones.Cover of stones was highest in the ecotone.This can be probably explained by the disturbance caused by vehicles turning and manoeuvring between reforestation rows (pers.obs.).Consistent with this hypothesis is the lower cover of perennial plants in the ecotone.

Effect of reforestation on ant assemblages
There was higher ant species richness (increase > 70% at the scale of transect × habitat accounting for the whole year and > 50% of pitfall catches at a distance of 60 m from the ecotone in the spring-summer season) and of total abundance by > 100% in the reforested area.Nevertheless, the multivariate analysis (Fig. 12) indicates that the fi rst trend in variation (axis 1) is unrelated to reforestation and is characterized by C. auberti differing from the other species in occurring in areas with a poor plant species richness.The second trend in variation indicates that the pitfall catches differ in the two habitats and, although there is overlap this may be a result of a gradual change in the environment between two habitats.The Adonis test was, however, signifi cant.In summary, reforestation clearly changed the ant assemblage.
Previous studies on the transition from grassland to forest found a variety of effects: highest species richness at the ecotone (Downie et al.,1996;Pinheiro et al., 2010), increasing linear response in species richness from forest to grassland (Kotze & Samways, 2001;Yekwayo et al., 2016), or an increasing linear response in species richness from grassland to forest (Bieringer & Zulka, 2003), as in this study.It seems that the response of arthropods to changes in their habitat resulting from reforestation may be case-and taxon-specifi c.
Although patterns at the level of the ant assemblage were clear (specially for species richness and total abundance), on a species by species basis they are fuzzier.There are two possible reasons for this.The most abundant species (Tapinoma nigerrimum) may be a generalist not affected by changes induced by reforestation, and the others are scarce species for which random noise may obscure any effects.Nevertheless, two species were signifi cantly more abundant in the grassland, four more abundant in the reforested area, and two in the ecotone.More species responded positively than negatively to reforestation.The underlying causes of this are mainly related to the changes in the vegetation and soil.Corley et al.'s (2006) study of the effects of pine afforestation on arid grasslands dominated by Stipa spp. on the Patagonian steppe reveals that open forest plantations have ant assemblages similar to those on native steppes.Other authors report that scattered presence of trees or invasion by shrubs of open grassland is associated with an increase in species richness and abundance of ants (Bestelmayer, 2005;Azcárate & Peco, 2012).This may be due to several factors.Reyes-López et al. (2003) report that some species are more common near trees than in open areas, suggesting that milder thermal environments may favour ants.However, a high density of trees on open grassland has an adverse effect on the ant assemblage (Reyes-López et al., 2003;Corley et al., 2006).
The results of this study are consistent with an improvement in the conditions for ants in grasslands with scattered shrubs and/or trees, as in our case.Year-round ant species richness and the abundances of two species of ants (Camponotus foreli and Aphaenogaster iberica) were positively correlated with the cover of P. halepensis, and the cover of this species is also correlated with axis 2, which segregates pitfall trap results by habitat.Trees are exclusive to the reforested area, but being widely spread and small they do not constitute a forest, but shrub land, with no more vegetation cover than the grassland .But, typically, for the same cover the biomass is much greater than that of shrubs and probably the microhabitats associated with trees are quite different from those associated with S. tenacissima and dominant shrubs.Not all the effects of the changes in vegetation were positive.Goniomma blanci was more abundant in grassland than in the reforested area.Furthermore, the abundance of this species was positively associated with Fumana thymifolia (a Cistaceae species) in spite of the low cover of this plant.Goniomma kugleri, another species of this genus on the Iberian Peninsula is specialized in harvesting the seeds of Cistaceae (Bastida et al., 2009), therefore it is very probable that G. blanci is similarly specialized and this would explain its lower abundance in the reforested area where the cover of Cistaceae is considerably less than in the grassland.
Food and nesting space are the most important resources for ants (Blüthgen & Feldhaar, 2010).A higher cover of stones and rocks provide more cool shelters for ant nests (Fernández-Escudero et al., 1993;Thomas, 2002;Robinson, 2008).Indeed, thermal properties of rocks benefi t ants in two ways: by reducing high temperatures during hot weather, and by rapid and effective absorption of solar heat in cold weather, enabling better brood development (Dean & Turner, 1991;Thomas, 2002).Furthermore, rocks protect colonies against predators, and in addition the soil below these stones is moister and therefore easier to excavate (Dean & Turner, 1991).Reyes-López et al. (2003) report that in an open forest in Sierra Morena (Spain), the relationship between availability of stones for nesting and number of ant nests is positive.Dahms et al. (2010), studying semi-natural grasslands in Sweden, report a positive relationship between rock cover and ant species richness.Finally, Friedrich & Philpott (2009) suggest that intraspecifi c and interspecifi c competition may be less when there is an abundance of nest sites.
In the reforestation, the planting of pines was preceded by mechanical preparation of the site, which resulted in a major disturbance of the surface properties of the soil.The petrocalcic horizon in the soil was partially broken, fragmented and mixed with the surface horizon.This increased the frequency of stones and rocks and reduced the cover of soil.Total ant species richness and total ant abundance was positively correlated with the number of stones on the surface of the soil and the segregation by the NMDS analysis (axis 2) of pitfall trap catches by habitat was correlated with the cover of stones and soil.The abundance of Camponotus sylvaticus, which is known to shelter under stones/ rocks, is negatively correlated with the cover of soil.More interestingly, the positive association of ant species richness and abundance with rock cover occurs in autumn but not year-round or in spring-summer, indicating a positive heating effect in cooler seasons.
The reforested area in this study is a more heterogeneous habitat, especially when jointly considering soil and vegetation and not just vegetation.Small-scale habitat complexity infl uences interspecifi c competition and resource use in ant communities, which results in differences among species that contribute to resource partitioning and permit subordinate species to avoid competition resulting in a greater diversity of ants (Luque & Reyes, 2007).Tews et al. (2004) introduce the concept of "keystone structures" in terms of vegetation complexity as a distinct spatial structure, which provides resources, shelter or "goods and services" crucial for other species.It would seem that both the abundance of stones on the surface of the soil and the small size of the pines in the system studied may match the concept of a keystone structure for ant assemblages.

CONCLUSIONS
The changes in the vegetation summarised by the decrease in the cover of S. tenacissima and Cistaceae, basic elements in grassland, and an increase in the cover of pine seems to provide more resources for the ant assemblage.
The parallel disturbance of the soil seems to increase the availability of shelters.Altogether, these changes resulted in an increase in the richness and abundance of ants.Nevertheless, the positive effect of disturbing soil during reforestation is probably not the norm.Most mechanized reforestation in the Mediterranean was carried out on hillslopes and resulted in extensive erosion and large bare patches that are extremely slowly recolonized by vegetation (especially on escarpments under terraces), very different from the gentle slopes studied here.S1.

Fig. 1 .
Fig. 1.Map showing the location of the study area.

Fig. 6 .
Fig. 6.Soil profi le in the grassland (a) and reforested area (b).Note that the petrocalcic horizon (Ckm) was lifted and fragmented by mechanical reforestation.

Fig. 8 .
Fig. 8. Ant species mean richness along the transect.Bars indicate standard error as calculated from model in TableS1.Fig.9.Ant species mean richness per pitfall trap.Values predicted by the models (TableS1).

Fig. S1 .
Fig. S1.Possible models of response of a variable to the ecotone and/or changes between habitats.

Table 2 .
Total number of each species of ant captured in the pitfall traps in grassland (G) and the reforested area (R).

Table 3 .
NMDS Correlations between environmental variables and coordinates of samples (pitfall trap catches) in the NMDS space.

Table S1 .
VALDÉS A. & HERRANZ J.M. 1989: Matorrales de la Provincia de Albacete: Espartales, Romerales y Tomillares.Instituto de Estudios Albacetenses, Albacete, 72 pp.WIEZIK M., SVITOK M., WIEZIKOVÁ A. & DOVČIAK M. 2013: Shrub encroachment alters composition and diversity of ant communities in abandoned grasslands of western Carpathians.-Biodivers.Conserv.22:2305-2320.YEKWAYO I., PRYKE J.S., ROETS F. & SAMWAYS M.J. 2016: Responses of ground living arthropods to landscape contrast and context in a forest-grassland mosaic.-Biodivers.Conserv.26:631-651.Full structure and basic statistics of the parameters of the generalized linear mixed models statistically signifi cant according with the procedures described in the Methods.For each model the link function (transformation of the dependent variable) and the adequate error function is indicated.Figures in the main text are represented after inverting the link function, i.e., on the original scale of the dependent variable.Signed distance is the distance to the ecotone, with negative sign to the inner reforestation and positive sign for the inner grassland.