Ecological niche modelling of species of the rose gall wasp Diplolepis (Hymenoptera: Cynipidae) on the Iberian Peninsula

Diplolepis (Hymenoptera: Cynipidae) are gall wasps that induce conspicuous galls on Rosa spp. (Rosaceae). These species are distributed globally and in Europe some are especially common and are founder organisms of biological communities composed of different insects. However, the ecological niches of these species have not been studied in detail. We modelled the potential distributions of these species using the locations of the galls of the four most abundant species of Diplolepis on the Iberian Peninsula (Diplolepis mayri, Diplolepis rosae, Diplolepis eglanteriae and Diplolepis nervosa, the galls of latter two are indistinguishable) using four different algorithms and identifi ed the resulting consensus for the species. We compared the potential distributions of these species, considering their spatial complementarity and the distributions of their host plants. We found that D. mayri and D. eglanteriae/nervosa have complementary distributions on the Iberian Peninsula. The former species is found in the Mediterranean region, while D. eglanteriae and D. nervosa are distributed mainly in the Eurosiberian region. Diplolepis rosae has the widest distribution on the Iberian Peninsula. Our models constitute the fi rst effort to identify suitable areas for species of Diplolepis species on the Iberian Peninsula and could be useful for understanding the evolutionary ecology of these species throughout their distribution in the western Palearctic.

Currently, there are a few cynipid ecological models, all of them related to Fagaceae hosts (Rodríguez et al., 2015;Gil-Tapetado et al., 2018). This paper presents, for the fi rst time, an accurately niche suitability modelling of Diplolepis communities on Rosaceae hosts.
The main aim of this study is to determine potentially favourable areas for gall-inducing cynipids associated with shrubs of the family Rosaceae on the Iberian Peninsula. This objective is achieved by comparing different ecological models incorporating predictive climatic and environmental variables, Diplolepis records and a suitability model of host plants, which is used to limit the distribution. These favourability models are used to analyse the association of species of Diplolepis with biotic and abiotic variables and draw conclusions about their distribution on the Iberian Peninsula.

Selection of presence data
To create the niche models, we compiled a dataset of the presence of fi ve species of Diplolepis recorded in southern Europe: D. eglanteriae/nervosa, D. spinosissimae, D. rosae and D. mayri (Fig. 1). This was compiled using the published and georeferenced records of each species. All records for synonymous species names were also considered acceptable and included in the data matrix (Nieves-Aldrey, 2001a). This data was compiled from different sources up to 2017 (Table 1, Fig. 2A-D, Table S1). As species of Diplolepis are dependent on Rosa species, we used the modelled distribution of the host plant to defi ne the limits of the cynipid distributions. Records of Rosa, consisting of 17,943 presences (Fig. 2E), were compiled from GBIF (Global Biodiversity Information Facility) datasets (GBIF Data Portal, 2016) and used in niche models. Both cynipid and host plant records were cleaned by eliminating geographically redundant (we included only one presence record per km 2 ) and low accuracy georeferenced data.

Se lection of variables
Bioclimatic and environmental variables can be used to predict the presence of each species of gall wasp. In the present study, different variables were used to model cynipid-gall and host Rosa, because it is important to distinguish those variables that have a direct effect on cynipid biology and those that infl uence the wasp through its host plant (Rodriguez et al., 2015;Gil-Tapetado et al., 2018).
The WorldClim version 1.4 (Hijmans et al., 2005) variables at a resolution of 30 arc seconds were used in the niche models of Diplolepis. As they emerge from their galls in spring (March, April, and May), only the spring variables Bio04 (temperature seasonality), Bio08 (mean temperature in wettest quarter), Bio13 (precipitation in wettest month), Bio15 (precipitation seasonality) and Bio16 (precipitation in wettest quarter) were used in these analyses, because they directly affect adult Diplolepis. Galls isolate the wasps from external environmental conditions, although some variables, such as snow cover, glycerol concentration or mild winter temperatures, affect the survival of those that overwinter in galls (Somme, 1964;Shorthouse, 1980;Williams et al., 2003). The free-living stage (i.e., adult cynipid), however, is affected directly by bioclimatic conditions (Shorthouse & Rohfrisch, 1992). nervosa and D. spinosissimae induce mainly small spherical and unilocular galls on leaves, and D. rosae, D. mayri and D. fructuum cause larger, multilocular galls on stems, leaves or hips.
D. eglanteriae and D. nervosa form a complex of cryptic galls that are morphologically indistinguishable, potentially leading to misidentifi cation. Both species are univoltine and bisexual. On the other hand, the galls of D. spinosissimae are produced on the leaves, the fruits and sometimes the stems of different species of the genus Rosa, mainly shrubs of Rosa pimpinellifolia L.
The species D. rosae and D. mayri induce conspicuous and striking galls and are thus the most collected and recorded within the genus. The galls of both species develop on buds or twigs, but sometimes on leafl ets or fruits. The galls of D. mayri (Fig. 1G) have a sparse coating of stiff spines instead of the fi lamentous appendix of the so-called 'rose bedeguar' galls of D. rosae (Fig. 1D-F). The life cycle of both species is univoltine, but their modes of reproduction vary with their geographical range, particularly on the Iberian Peninsula. While males of D. rosae are very scarce or virtually absent in most areas on the Iberian Peninsula, the sex ratio of this species is closer to 1 : 1 in the rest of Europe. In fact, some studies in other areas have linked latitude with the relative presence/abundance of males of D. rosae (Askew, 1960;Stille, 1984Stille, , 1985. Males of D. mayri (Fig. 1A) are not abundant in non-Iberian Europe, but other authors have discovered differences in the reproductive biology of this species on the Iberian Peninsula, where it does not undergo thelytokous parthenogenesis as it does in other countries. Instead, the Iberian D. mayri has a bisexual generation and a sex ratio close to 1 : 1 (Pujade-Villar, 1983;Nieves-Aldrey, 1989;Nieves-Aldrey, 2001a).
From the biogeographical and macroecological point of view, determining the potential distribution of organisms is an important tool for the ecological and biological conservation of animals (Guisan & Zimmermann, 2000;Guisan et al., 2006;Peterson, 2006). Using mathematical algorithms to model the potential distribution of species has allowed the determination of those areas where there is a greater probability of fi nding them. Niche models, suitability models or predictive habitat distribution models are empirical or mathematical approximations of the ecological niche of a species constructed from their presence and/ or absence records and variables that limit and defi ne this niche (Araújo & Guisan, 2006;Austin, 2007;. It is important that these niche models are based on environmental conditions in the regions studied and that the potential distribution models provide a distribution of habitat suitability for each species (Franklin, 2009;Zimmermann et al., 2010). The variable selection procedure is quite important since these variables are going to limit and defi ne the habitat of a species. In addition, selecting appropriate predictor variables will decide if the model is essential for predicting species distributions (Guisan & Zimmermann, 2000;Jiménez-Valverde et al., 2011). Climatic variables We used cluster dendrogram analysis of correlation distances to identify possible correlation biases between the variables. First, we chose all the variables that exceeded a value of 0.3 (or less than 70% correlation). Second, we selected all the uncorrelated variables (those variables that did not form a cluster); fi nally, we chose the variable with the greatest biological signifi cance for Diplolepis from the clusters of correlated variables. In cases in which a variable did not have a clear biological meaning, the most derived variable was chosen (i.e., the variable that refers to a specifi c period of the year). Bio16 was excluded from the modelling of Diplolepis due to its high correlation with the rest of the variables (Fig. 3A). In addition, we applied a forward elimination process, variance infl ation factor (VIF), to the variables. VIF measures how much the variance in a regression coeffi cient increases when predictors are correlated. We computed the VIF of the variables using R version v 3.3.1 ( R Development Core Team, 2008) and RStudio 0.99.903 (RStudio Team, 2016) in the package HH 3.1-32 (Heiberger & Holland, 2015). The threshold value of 5 was used to determine which variables were strongly corre-lated with other factors, thereby indicating that it was appropriate to eliminate them from the analysis (Kutner et al., 2004;O'Brien, 2007;Lin et al., 2011). All the variables in the dendrogram satisfi ed the threshold value condition (VIF > 5) and no variable was excluded from the analysis.
The selection of environmental variables for the construction of the Rosa model was diffi cult because these plants are widely distributed and it is a group with many species (Zhang & Gandelin, 2003;Yan et al., 2005;Koopman et al., 2008). The ecological and edaphic conditions selected are general, which makes it difficult to determine those conditions that are most likely to infl uence its presence (Castroviejo et al., 1998). We used the modelling methods previously mentioned and selected the most general climatic variables from WorldClim (Bio01 (annual mean temperature), Bio02 (mean diurnal range), Bio03 (isothermality), Bio07 (temperature annual range) and Bio12 (annual precipitation)) and other non-climatic variables (available water capacity (ESDB, 2004), percentage of clay (ESDB, 2004), lithology (IGME, 2020), proximity to water masses (rivers, lakes, reservoirs and other continental water bodies) (modifi ed from EEA, 2009) and land uses from CORINE Land Cover 2006 (EEA, 2012) resized to 1 km 2 ). Bio2 and available water capacity were not included in the fi nal set of variables (Fig. 3B).
For both the Rosa and Diplolepis models, we performed background point generation tests following and improving upon the methodology of Gil-Tapetado et al. (2018), using the maximum and minimum values of the predictor variables as parameters of habitability for each species. The sum of the resulting areas between the maximum and minimum values of each predictive variable was classifi ed as habitable area in which to generate random background points; the excluded areas were classifi ed as nonhabitable zones in which to generate the pseudo-absences.
Four individual models (GLM, GAM, MaxEnt and RF) were developed for the host plant taxa (genus Rosa), D. rosa and D. mayri, and the group formed by D. eglanteriae and D. nervosa, the galls of which are not externally morphologically distinguishable. We could not develop models for D. spinosissimae (Fig. 1H) because of the few records for this species from the Iberian Peninsula ( Fig. 2D, Table 1). With respect to host plant taxa, models were made including different species of the genus Rosa.
We evaluated these models using area under the curve (AUC) statistics. Finally, we selected models that had a maximum suitability value of at least 0.80 to obtain an improved estimation of the potential geographic distribution of a species. The models selected for each species were joined to create an average consensus map.
ArcGIS 10.1 software (ESRI, 2011) was used to transform highly suitable areas (i.e., areas with a suitability greater than or equal to 0.70) indicated by the host plant consensus models into buffer zones of high suitability -i.e., zones with a radius of one kilometre around the presence data for Rosa, which constitute possible action areas for the species. Presence data for the genus Rosa were also included, and corresponding buffers with a one-kilometre radius were generated around the presence points. We overlapped this mask layer with each consensus model of Diplolepis species to obtain the fi nal models. These results refl ect the high and low probabilities of the presence of each cynipid within their host plant areas on the Iberian Peninsula.
The complementary maps of the Diplolepis species were developed in ArcGIS 10.1 (ESRI, 2011) based on the most favourable areas (areas with values greater than 0.70). Finally, these maps were combined by twos. In addition, the possible associations of environmental variables Bio04 (temperature seasonality), Bio15 (precipitation seasonality) and Bio16 (precipitation in wettest quarter) from WorldClim version 1.4 (Hijmans et al., 2005) with the potential distribution areas of the Diplolepis species were checked using boxplots in the ggplot2 package in RStudio.

RESULTS
The results from the RF, GLM and GAM algorithms were selected for the species D. eglanteriae/nervosa and D. rosae because they had a maximum suitability of more than 0.80. In the case of D. mayri, only the results of the RF and GAM met this requirement. In all these cases, Max-Ent did not exceed a maximum suitability of 0.80 and we Table 1. Record of presence of different species of Diplolepis on the Iberian Peninsula and t he sources of the information (Source) from which they were compiled. N indicates the number of records per source.

Species
Source N Presence data
The models indicate that southern areas of the Iberian Peninsula are more suitable for D. mayri (Fig. 4C) and northern areas more suitable for D. eglanteriae/nervosa (Fig. 4B), while D. rosae is a generalist, with highly suitable areas throughout the Iberian Peninsula (Fig. 4A).
Below, are the detailed results for each of the species of Diplolepis.
Diplolepis rosae (Figs 1C-F, 4A) is the most common species of this genus on the Iberian Peninsula. According to the presence points and areas of maximum suitability, this species occurs in areas with a mean annual temperature of approximately 11°C and annual precipitation of 700-800 mm and at an average altitude of approximately 900-1000 m a.s.l. In the areas in which D. rosae occurs, in the warmest quarter, the mean temperature was 19°C and there was approximately 160 mm of precipitation, while in the coldest quarter, the mean temperature was 4°C and there was approximately 190 mm of precipitation (Table  3). The presence areas for D. rosae are diverse, with suitability percentages of > 0.70 in all areas. The areas with the most suitable ecological conditions are concentrated in the region of the Catalonian Pyrenees (north eastern Iberian Peninsula) and Montes de León (north western Iberian Peninsula) (Fig. 4A).
The most favourable areas for D. eglanteriae and D. nervosa (Fig. 1I, Fig. 4B) are in mountain ranges. The results indicate that areas in the Cantabrian Mountains and Pyrenees (in the north of Iberian Peninsula), both the Central and Iberian Mountain Ranges (central Iberian Peninsula), and part of the Baetic System (in the south of Iberian Peninsula) are the most suitable areas (with suitability percent- Fig. 3. Cluster dendrogram analysis based on the correlation coeffi cient distance for the following datasets. A -cynipid-related variables: Bio04 (temperature seasonality), Bio08 (mean temperature of wettest quarter), Bio13 (precipitation of wettest month), Bio15 (precipitation seasonality) and Bio16 (precipitation of wettest quarter); B -host plant-related variables: Bio01 (annual mean temperature), Bio02 (mean diurnal range), Bio03 (isothermality), Bio07 (temperature annual range), Bio12 (annual precipitation), available water capacity, percentage of clays, lithology, proximity to water masses and CORINE Land Cover (derived from land uses in the CORINE Land Cover 2006). ages > 0.70) for this species complex on the peninsula. The most suitable areas are characterized by an average mean annual temperature of 9.97°C and an annual precipitation of 908 mm (Table 4). The models indicate that the most suitable areas for these species are coastal areas in Galicia, Asturias, Cantabria, Basque Country and Catalonia and the mainland in Navarre and Aragon (northern Spain; Fig. 4B). Diplolepis mayri (Figs 1A, B and G, 4C), like D. rosae, is common on the Iberian Peninsula. The mountainous areas in the central and south eastern parts of the peninsula are highly suitable for D. mayri (Table 5), with the Catalonian Pyrenees (north eastern Iberian Peninsula) the most suitable. These areas are the parts of the Iberian Peninsula with a Mediterranean climate.
Regarding the bioclimatic variables in the highly suitable areas for each species of Diplolepis (Tables 3, 4 and 5), D. mayri is the species which can occur at higher altitudes (956 m a.s.l. on average) and D. eglanteriae/nervosa in areas with the highest rainfall (787.1 mm on average). Moreover, combining the potential distribution maps revealed suitable areas for all of the species of Diplolepis. There is a clear separation in the highly suitable areas for D. eglanteriae/nervosa and D. mayri (Fig. 5A), with those of the former mainly in the northern parts and the latter in of the central and southern parts of the Iberian Peninsula. The central mountain ranges are of low suitability for D. eglanteriae/nervosa and D. rosae, whereas the mountainous areas in the northern part of the Iberian Peninsula are the most favourable for them (Fig. 5B). The species D. rosae and D. mayri share the same potential distribution, with a small area in northern Catalonia region (north eastern Iberian Peninsula) with highest suitability for both species (Fig. 5C). The distributions of species of Diplolepis seem to be associated with particular climatic variables (Fig. 6). The boxplots provide a visual approximation of the predictions of the models: areas highly suitable for D. mayri are in areas with higher temperatures and lower rainfall than those highly suitable for the other species of Diplolepis. It is also noted that D. eglanteriae/nervosa has the highest tolerance range for rainfall and D. rosae is the most generalist of the species (Fig. 6).

Suitability models for the different species of Diplolepis
The presence data indicate a greater occurrence of all the species of Diplolepis studied in the northern than in the southern part of the peninsula (Fig. 2). This may be because cynipids have been sampled less in southern Spain than in central and northern Spain (Nieves-Aldrey, 2001a). However, more recent samples collected from 2017 to 2019 by Nieves-Aldrey (unpubl. data) confi rm that D. rosae is absent or less abundant in the southern part of the Iberian Peninsula. In addition, it is noteworthy that the records obtained from the citizen science platform Biodiversidad Virtual (BVdb, 2016), which were collected by many observers and make up ~ 50% of the total records for species of Diplolepis (Table 1), few of these records are for the southern Iberian Peninsula. Moreover, many galls are striking structures that are specifi c to the inducer wasp, which makes them easily identifi able and photographic evidence of their presence at a particular place. Such georeferenced data can be used along with previously published records in biogeographical studies of the Iberian Peninsula (Goula et al., 2013;Jiménez-Valverde et al., 2019). Based on this information and that D. rosae is the most common species on the Iberian Peninsula, it is possible that the data truly refl ect a latitudinal gradient in the abundance of these species on this Peninsula.
On the other hand, an important aspect of these models is the environmental conditions selected and their effects on the distribution of the species analysed. The insects studied are obligatorily dependent on their Rosa hosts, on which they induce galls (Nieves-Aldrey, 2001a). Thus, the limiting variable is the distribution of their host plant. The suitability models developed for Rosa were constructed using all the data for wild rose plants, without considering host specifi city or preference of each species of Diplolepis. Although there is currently no evidence of species of Diplolepis being host specifi c (Stille, 1984;Kohnen et al., 2011), except in the case of D. spinosissimae for R. pimpinellifolia, it is possible that the distributions of specifi c species of Rosa may be a factor affecting the distributions of these cynipids. There are 27 species of the genus Rosa on the Iberian Peninsula, which are highly polymorphic and hybrids are also present, making their identifi cation by non-specialists diffi cult (Castroviejo, 1998;Cueto & Giménez, 2009;Calvo & Ross-Nadié, 2016;Tomljenović & Pejić, 2018). For this reason, and in addition to the fact that there are no reliable data on the distributions of species of Rosa, all the plant data used were at the genus rather than the species level.
Other variables not considered were the effects of habitat fragmentation due to natural and anthropogenic causes and the microhabitat or microtopographic conditions. Increasing land use intensity and habitat fragmentation are important threats to global biodiversity, especially in agricultural landscapes. Some studies report that landscape diversity has little or no effect on the species richness of Diplolepis (Looney & Eigenbrode, 2010) and that the homogenization of the landscape provides a perfect habitat for gall inducers and their community members (László et al., 2018). On the other hand, microtopographic conditions can mask temperature differences along altitudinal or latitudinal gradients over small scales. The interaction between microtopography, plant cover and solar radiation result in microhabitat conditions that are not represented by climatic variables (Scherrer & Körner, 2010Scherrer et al., 2011).
Diplolepis rosae has a wider potential distribution than the rest of the species. For this reason, this species is found in most rose bushes on the Iberian Peninsula (Fig. 4A). Most of the D. rosae data used in the present article do not specify the sex of the galler or the composition and abundance of its parasitoid communities. However, in the summer of 2017, Nieves-Aldrey (unpubl. data) found a single male specimen of D. rosae in the Aracena Mountains in south western Iberian Peninsula (Nieves-Aldrey, 2001a). This observation matches those of other authors, which indicate the possibility of a latitudinal gradient (from south to north) in the sex ratio of this species. For example, Askew (1960) and Hoffmeyer (1925) suggest that there is a higher abundance of male D. rosae in the northern parts of the peninsula. However, some authors (Rizzo & Massa, 2006;  László & Tóthmèrèsz, 2011;Todorov et al., 2012) report the lack of a latitudinal gradient and that the presence of D. rosae males depends on more complicated factors, and also that the endosymbiotic bacteria Wolbachia affect the sex ratio of this species (Schilthuizen & Stouthamer, 1998;Rizzo & Massa, 2006). The record of a male of D. rosae on the Iberian Peninsula indicates the presence of males in southern Europe, although at a low abundance. However, to determine the actual sex ratio of D. rosae, more intensive sampling is needed in southern Europe.
The D. eglanteriae/nervosa model (Fig. 4B) reveals that the presence data and the maximum suitability for these two species occur in mountain forest areas and they avoid river-valleys (Table 4). In addition, D. eglanteriae and D. nervosa are also abundant in areas near the northern coast of the Iberian Peninsula. It is possible that in mountainous areas, species of Rosa have local characteristics, e. g. separation of vegetation patches, shrub size, or distance to the soil, that are lacking in riverbanks, and that these conditions are favourable for both cynipid species. An example is provided by Mifs ud (2016), who reports that in mountainous areas, where wild roses grow in the shade of conifer trees, they do not fl ower and galls are not recorded. On the other hand, mountainous areas, particularly mountain forests, are favourable in many other ways; for example, the spatial confi guration of trees favours optimal radiation, temperature and precipitation conditions. The conditions found in mountainous areas, together with being isolated from other areas, may have resulted in there being more or different species of Rosa, which could be specifi c or more suitable hosts for D. eglanteriae and D. nervosa, than at lower altitudes. However, the host preference of these spe-cies of Diplolepis has not been determined due to the taxonomic diffi culties of identifying each species of wild rose, among other reasons.
Diplolepis mayri occurs in the western Palearctic in areas of north Africa and Asia Minor (Nieves-Aldrey, 2001a). The distribution model for this species (Fig. 4C) revealed that the areas with the highest suitability were mainly in mountainous regions in the Mediterranean area of the Iberian Peninsula. Unlike D. rosae and D. eglanteriae/nervosa, there is an area of high suitability for D. mayri in the southern third of the Iberian Peninsula. This species is distributed in areas with relatively more precipitation seasonality and temperature seasonality, and relatively less rain (Fig. 6, Table 5). This might indicate that D. mayri is adapted to the Mediterranean climatic conditions in the southern Iberian Peninsula, unlike the rest of the species of Diplolepis studied.

Habitat complementarity
A comparison of the potential distributions of D. eglanteriae/nervosa and D. mayri revealed that they occur in complementary areas, e.g., areas on Iberian Peninsula with highly suitable environmental conditions for D. eglanteriae and D. nervosa are of low suitability for D. mayri and vice versa (Fig. 5A). This phenomenon occurs between species in other genera, such as Copris hispanus (Linnaeus, 1758) and Copris lunaris (Linnaeus, 1758) (Coleoptera: Scarabaeidae) (Chefaoui et al., 2004), different species of the red-striped oil beetle of the genus Berberomeloe Bologna, 1989 (Coleoptera: Meloidae) (Sánchez-Vialas et al., 2020) and among species that compete for resources (De Smedt et al., 2016). It is possible that the species of Diplolepis studied are allopatric. The areas of high suitability for D. eglanteriae and D. nervosa occur in the north of the Iberian Peninsula (Euro Siberian areas) and those for D. mayri in central and southern parts of the Iberian Peninsula (Mediterranean areas). These zones are separated by the northern plateau of the Iberian Peninsula where host plants are either absent or scarce. Studies on the structure and genetic diversity of these species are likely to reveal signifi cant genetic differentiation between them, possibly due to this physical barrier. The molecular characteristics of the cytochrome b and 12S genes and different character states ('number of larval chambers per gall', 'organ bearing the gall' and 'surface area of the gall') of species of the Diplolepis reveals that the species included in our study form two species groups: the D. eglanteriae group (including D. eglanteriae and D. nervosa) and the D. rosae group (including D. rosae, D. fructuum, D. mayri and D. spinosissimae) (Plantard et al., 1997). Recently, Zhang et al. (2019) published a phylogeny of Diplolepis and Periclisttus based on COI (cytochrome c oxidase subunit I) that divides the genus Diplolepis into two major monophyletic clades, the 'fl anged femur' clade and 'leaf galler' clade, the latter of which includes three subclades: the Nearctic leaf galler subclade, the Palearctic multichambered subclade and a mixed leaf gall subclade. The Palearctic multichambered subclade includes the D. rosae group (D. fructuum, D. mayri and D. rosae), D. spinosissimae and two undescribed species. Finally, the rest of the species included in our study are classifi ed in the mixed leaf gall subclade, which distinguishes between the D. eglanteriae Palearctic group (D. japonica and D. eglanteriae) and the D. eglanteriae Nearctic group and D. nervosa. The different lineages of the subclades that contain D. mayri and D. rosae (Palaearctic multichambered subclade) and D. eglanteriae and D. nervosa (mixed leaf gall subclade) could partially explain the observed habitat complementarity between D. mayri and D. eglanteriae/ nervosa.
There is an overlap in the most suitable areas for Diplolepis eglanteriae/nervosa and D. rosae in the northern mountain ranges (Fig. 5B); therefore, there could potentialy be interspecifi c competition between these species in this area. However, D. rosae and D. eglanteriae/nervosa exploit their host plants in different ways. Diplolepis eglanteriae/nervosa induce galls on the petioles or undersides of leaves and D. rosae on the buds of the stems, leaflets or fruits (Nieves-Aldrey, 2001a). It is possible that if two different species have overlapping distributions, they might use different parts of the same resource or develop strategies that allow them to exploit the same resource without competing. For example, the distributions of other Hymenoptera [Apis mellifera (Linnaeus, 1758) and Bombus terrestris (Linnaeus, 1758)] also largely overlap with respect to the plants they pollinate; however, their pollination activities have different schedules (Perez, 2013). For the two species with most records (D. rosae and D. mayri), the most suitable areas occur in the north eastern and central parts of the Iberian Peninsula (Fig. 5C) where they overlap the suitable habitat for D. eglanteriae/nervosa, as mentioned above.

An overview of the distribution of Diplolepis in Europe
Focusing on the western Palearctic, the glaciation events in the Quaternary/Pleistocene are considered to be a key determinant of the distribution of plants and therefore that of cynipids (Taberlet et al., 1998;Hewitt 1999;Rokas et al., 2003;Stone et al., 2001;Güçlü et al., 2008). Like the Iberian Peninsula, the Anatolian Peninsula was also a glacial refugium during the European glaciations and have a rich fauna, including cynipids of the genus Diplolepis. Therefore, a comparative study of these two peninsulas, with either similar or parallel scenarios, can be used to defi ne the evolutionary history of this group in the western Palearctic. In addition, the Anatolian Peninsula is the transition point between the continents of Europe, Asia and Africa, and has a great variety of natural habitats, ranging from Mediterranean, Aegean and Black Sea beaches to towering coastal and interior mountains, including deeply incised valleys, expansive steppes and fertile alluvial plains along with arid, rocky slopes. Ronquist & Liljeblad (2001) report that many of the ancestral cynipid relationships occur in the eastern Mediterranean and Turano-Eremial region, indicating that this area was possibly the centre of speciation for cynipid gallers (Mete & Demirsoy, 2012).
The absence of D. fructuum from the Iberian Peninsula (the only other species of western Palearctic Diplolepis missing in this territory) and the presence of all the European species of this genus on the Anatolian Peninsula seem to indicate an east-west gradient in the richness of species of cynipids. In the past, D. fructuum was considered a geographic race of D. mayri, but molecular techniques confi rm they are different species (Plantard et al., 1997;Güçlü et al., 2008;Zhang et al., 2019). It is possible that Diplolepis spread from eastern Europe to the west; in this case, the Iberian Peninsula is the western limit to the distribution of this genus, with greater species richness in the Nearctic and eastern Palearctic regions.
In the case of the suitable areas for D. eglanteriae and D. nervosa there are similarities between the Anatolian and Iberian peninsulas. The most suitable areas on the Iberian Peninsula for Diplolepis eglanteriae/nervosa are around high-altitude large mountain ranges. This is also the case on the Anatolian Peninsula, where D. eglanteriae and D. nervosa occur in provinces enclosed by mountain ranges at altitudes greater than 2,000 m (Güçlü et al., 2008;Katılmıs & Kıyak, 2010;Mete & Demirsoy, 2012). This supports the hypothesis that both the Iberian and Anatolian Peninsulas were Pleistocene glacial refugia of these species. In addition, D. eglanteriae and D. nervosa spread from western Asia into Europe along major mountain ranges, such as the Carpathians in Romania, Serbia and the Ukraine (Melika, 2006;Marković, 2015;Prázsmári et al., 2017), Malta (Mifsud, 2016) and the Scandinavian Mountains at their northern latitudinal limit (Bergqvist, 2010). Based on these facts and our distribution models, which indicate that D. eglanteriae and D. nervosa occur in the Eurosiberian region, it is likely that glaciation events may have limited the distributions of these species, which were able to adapt to cold and high-altitude conditions (Fig. 6, Table 4). On the other hand, D. rosae seems to have been able to spread and adapt to different temperate conditions (Table 3), although the scarcity of this species in the southern parts of the Iberian Peninsula seems to indicate it has a Euro Siberian distribution. Diplolepis mayri seems to have an affi nity for Mediterranean conditions as the highly suitable areas for this species are on the southern plateau of the Iberian Peninsula and complementary to those of D. rosae and Diplolepis eglanteriae/nervosa.