Research Article 
Corresponding author: Andres Baselga ( andres.baselga@usc.es ) Academic editor: Jorge SantiagoBlay
© 2016 Andrea Freijeiro, Andres Baselga.
This is an open access article distributed under the terms of the Creative Commons Attribution License (CC BY 4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Citation:
Freijeiro A, Baselga A (2016) Spatial and environmental correlates of species richness and turnover patterns in European cryptocephaline and chrysomeline beetles. In: Jolivet P, SantiagoBlay J, Schmitt M (Eds) Research on Chrysomelidae 6. ZooKeys 597: 81–99. https://doi.org/10.3897/zookeys.597.6792

Despite some general concordant patterns (i.e. the latitudinal richness gradient), species richness and composition of different European beetle taxa varies in different ways according to their dispersal and ecological traits. Here, the patterns of variation in species richness, composition and spatial turnover are analysed in European cryptocephaline and chrysomeline leaf beetles, assessing their environmental and spatial correlates. The underlying rationale to use environmental and spatial variables of diversity patterns is to assess the relative support for niche and dispersaldriven hypotheses. Our results show that despite a broad congruence in the factors correlated with cryptocephaline and chrysomeline richness, environmental variables (particularly temperature) were more relevant in cryptocephalines, whereas spatial variables were more relevant in chrysomelines (that showed a significant longitudinal gradient besides the latitudinal one), in line with the higher proportion of flightless species within chrysomelines. The variation in species composition was also related to environmental and spatial factors, but this pattern was better predicted by spatial variables in both groups, suggesting that species composition is more linked to dispersal and historical contingencies than species richness, which would be more controlled by environmental limitations. Among historical factors, Pleistocene glaciations appear as the most plausible explanation for the steeper decay in assemblage similarity with spatial distance, both in cryptocephalines and chrysomelines.
Beta diversity, biogeography, Chrysomelidae , Chrysomelinae , Cryptocephalinae , species richness
The assessment of largescale biogeographic patterns has proven a fruitful research discipline for understanding the ecological, evolutionary and historical processes that have determined the current distribution of biological diversity (
Recent examples of this approach have proven particularly successful. A few invertebrate taxa (e.g., diurnal Lepidoptera) have been particularly well sampled, so diversity patterns are relatively well known (
The studies cited above have shown that different beetle taxa have different diversity patterns across continental Europe. Regarding species richness, the steepness of the negative latitudinal richness gradient has been shown to be related to the dispersal capacity of taxa (
The study area included continental Europe. Thirtysix inventories of Cryptocephalinae and Chrysomelinae were obtained from
Three sets of variables were obtained for each country: (i) area; (ii) spatial position: mean, minimum and maximum longitude (Long, Long_{min}, Long_{max}), mean, minimum and maximum latitude (Lat, Lat_{min}, Lat_{max}) and longitudinal and latitudinal range (Long_{ran} and Lat_{ran}); and (iii) environmental factors: mean altitude (Alt); altitudinal range (Alt_{ran}); annual mean temperature (T_{ann}); spatial range of T_{ann} (T_{ran}); maximum temperature of the warmest month (T_{max}); minimum temperature of the coldest month (T_{min}); annual precipitation (P_{ann}); spatial range of P_{ann} (P_{ran}); precipitation of driest quarter (P_{dri}); and spatial range of P_{dri} (P_{drn}). Topographic and climatic variables were obtained from WorldClim 1.4 layers (
The relationship between diversity attributes (species richness, species composition and spatial turnover) and the aforementioned variables was independently assessed for cryptocephalines and chrysomelines:
1. Variation in species richness. Multiple relationships between species richness and the explanatory variables were analysed using regression modelling (
2. Variation in species composition. The variation in species composition among countries was analysed with the Simpson’s index of dissimilarity (β_{sim}) (
3. Variation in spatial turnover. Given the observed patterns of spatial variation in species composition (i.e. spatial turnover), we assessed whether the rates of turnover with spatial distance were significantly different in northern and southern Europe. To do this, we transformed the pairwise dissimilarity matrix based on β_{sim} on similarities (i.e. 1β_{sim}), and split the data into two groups: northern European countries, with mean latitude higher than 48 degrees (n=19), and southern European countries (n=17). Thereafter, we assessed the decay of assemblage similarity with spatial distance in Northern and Southern European datasets, using nonlinear regression on similarity matrices to fit exponential decay curves expressed as y=a*e^{}^{bx}, where y is similarity at distance x, a initial similarity and b the rate of distance decay. Spatial distance was computed in km as the Euclidian distance between the UTM centroids of countries. Finally, to assess for significant differences in distance decay slopes between Northern and Southern regions, the frequency distributions of the parameters were estimated by bootstrapping. A frequency distribution of 1,000 slopes was retrieved by bootstrapping, using the boot package (
The assessment of variation in species richness revealed a clear latitudinal gradient, with a significant reduction of diversity to the North in both groups (Figure
Patterns of variation in species richness (a, b), hierarchical clustering based in β_{sim} (c, d) and mapping of 4 major clusters (e, f) for cryptocephalines (left column: a, c, e) and chrysomelines (right column: b, d, f). Colours correspond to the 4 major clusters. Countries’ abbreviations follow those of
Relationships between cryptocephaline species richness and variables, and models for each group of variables. The sign of the relationships and percentage of explained variance (%) are shown, with their respective F parameters, degrees of freedom (d.f.) and pvalues. A, S and E are the area, spatial and environmental models, respectively. f^{2} is the second order polynomial of the variable considered.
Variable  Function  Variance (%)  F  d.f.  p 

Area  logarithmic (+)  12.6  4.9  1, 34  0.034 
Long  ns  6.0  2.2  1, 34  0.150 
Long_{min}  ns  10.0  3.8  1, 34  0.060 
Long_{max}  ns  4.0  1.4  1, 34  0.244 
Long_{ran}  quadratic (+,)  21.9  4.6  2, 33  0.017 
Lat  linear ()  22.2  9.7  1, 34  0.004 
Lat_{min}  linear ()  29.7  14.4  1, 34  0.001 
Lat_{max}  linear ()  15.4  6.2  1, 34  0.018 
Lat_{ran}  quadratic (+,)  16.7  3.3  2, 33  0.050 
Alt  quadratic (+,)  19.6  4.0  2, 33  0.027 
Alt_{ran}  linear (+)  28.3  13.4  1, 34  0.001 
T_{ann}  linear (+)  16.9  6.9  1, 34  0.013 
T_{ran}  linear (+)  30.1  14.6  1, 34  0.001 
T_{max}  linear (+)  16.5  6.7  1, 34  0.014 
T_{min}  ns  10.8  4.1  1, 34  0.050 
P_{ann}  ns  0.1  0.0  1, 34  0.894 
P_{ran}  quadratic (+,)  31.8  7.7  2, 33  0.002 
P_{dri}  ns  0.7  0.3  1, 34  0.619 
P_{drn}  linear (+)  34.7  18.1  1, 34  <0.001 
Model for A  log(Area)  12.6  4.9  1, 34  0.034 
Model for E  T_{max} + P_{ran} + P_{drn} + f^{2} Alt  76.1  19.2  5, 30  <0.001 
Model for S  f^{2} Long_{ran} + Lat_{min}  51.3  11.3  3, 32  <0.001 
Model S + A  f^{2} Long_{ran} + Lat_{min} + log(Area)  52.3  8.5  4, 31  <0.001 
Model E + S  T_{max} + P_{ran} + P_{drn} + f^{2} Alt + f^{2} Long_{ran} + Lat_{min}  82.0  15.3  8, 27  <0.001 
Model E + A  T_{max} + P_{ran} + P_{drn} + f^{2} Alt + log(Area)  81.5  21.3  6, 29  <0.001 
Model E + S +A  T_{max} + P_{ran} + P_{drn} + f^{2} Alt + f^{2} Long_{ran} + Lat_{min} + log(Area)  83.2  14.3  9, 26  <0.001 
Relationships between chrysomeline species richness and variables, and models for each group of variables. The sign of the relationships and percentage of explained variance (%) are shown, with their respective F parameters, degrees of freedom (d.f.) and pvalues. A, S and E are the area, spatial and environmental models, respectively. f^{2} is the second order polynomial of the variable considered.
Variable  Function  Variance (%)  F  d.f.  p 

Area  logarithmic (+)  17.1  7.0  1, 34  0.012 
Long  ns  9.9  3.7  1, 34  0.062 
Long_{min}  linear ()  17.0  6.9  1, 34  0.013 
Long_{max}  ns  5.2  1.9  1, 34  0.179 
Long_{ran}  quadratic (+,)  24.6  5.4  2, 33  0.009 
Lat  linear ()  14.3  5.7  1, 34  0.023 
Lat_{min}  linear ()  19.4  8.2  1, 34  0.007 
Lat_{max}  ns  8.8  3.3  1, 34  0.079 
Lat_{ran}  quadratic (+,)  23.3  5.0  2, 33  0.013 
Alt  linear (+)  18.0  7.4  1, 34  0.010 
Alt_{ran}  quadratic (+,)  41.1  11.5  2, 33  <0.001 
T_{ann}  ns  7.3  2.7  1, 34  0.110 
T_{ran}  linear (+)  33.9  17.4  1, 34  <0.001 
T_{max}  ns  5.0  1.8  1, 34  0.191 
T_{min}  ns  6.6  2.4  1, 34  0.129 
P_{ann}  ns  2.2  0.8  1, 34  0.385 
P_{ran}  quadratic (+,)  43.0  12.4  2, 33  <0.001 
P_{dri}  ns  2.8  1.0  1, 34  0.331 
P_{drn}  linear (+)  44.5  27.3  1, 34  <0.001 
Model for A  log(Area)  17.1  7.0  1, 34  0.012 
Model for E  P_{dri} + f^{2} P_{ran}  60.0  16.0  3, 32  <0.001 
Model for S  Long_{min} + f^{2} Long_{ran} + Lat + Lat_{ran}  71.2  14.8  5, 30  <0.001 
Model S + A  Long_{min} + f^{2} Long_{ran} + Lat + Lat_{ran} + log(Area)  75.1  14.6  6, 29  <0.001 
Model E + S  P_{dri} + f^{2} P_{ran} + Long_{min} + f^{2} Long_{ran} + Lat + Lat_{ran}  88.0  24.8  8, 27  <0.001 
Model E + A  P_{dri} + f^{2} P_{ran} + log(Area)  65.6  14.8  4, 31  <0.001 
Model E + S +A  P_{dri} + f^{2} P_{ran} + Long_{min} + f^{2} Long_{ran} + Lat + Lat_{ran} + log(Area)  89.1  23.5  9, 26  <0.001 
In Chrysomelinae, in addition to the negative relationship with latitude, species richness was also significantly and negatively related to minimum longitude and significantly and positively related to country area, mean altitude, annual mean temperature and spatial range of precipitation of driest quarter. In the case of longitudinal, latitudinal and altitudinal range and spatial range of annual precipitation the relationships were curvilinear, with maximum richness at intermediate values. After excluding collinear variables, the spatial model for Chrysomelinae consisted of mean latitude, latitudinal range, minimum longitude and the quadratic function of longitudinal range, explaining 71.2% of the variance in species richness. The environmental model comprised the spatial range of precipitation of the driest quarter, and the quadratic function of spatial range of annual precipitation, explaining 59.3% of the variance in richness. Variance partitioning showed that the largest portion of variation in richness (34.5%) was jointly explained by the environmental and spatial models, with a small contribution of country area (Figure
The assessment of variation in species composition (beta diversity) revealed the existence of similar patterns in both subfamilies, with the presence of singular faunas in the Iberian Peninsula, the Greek Peninsula, and Southern Russia, while the remaining European regions formed a relatively uniform cluster of countries with similar composition (Figure
In Cryptocephalinae, the environmental model for variation in species composition included mean altitude, mean annual temperature, minimum temperature of the coldest month, and precipitation of the driest quarter as significant variables (pseudoF_{4,31} = 6.73, p < 0.001) and explained 46.5% of the variation (Table
Relationships between variation in cryptocephaline species composition (β_{sim}) and variables, and models for each group of variables. Percentages of variation explained are shown, with their respective PseudoF parameters, degrees of freedom (d.f.) and pvalues. S and E are the spatial and environmental models, respectively.
Variable  Variation (%)  PseudoF  d.f.  p 

log(Area)  3.5  0.77  1, 34  0.288 
Long  12.7  4.96  1, 34  0.002 
Long^{2}  6.3  2.30  1, 34  0.058 
Long^{3}  5.2  1.87  1, 34  0.094 
Lat  25.1  11.37  1, 34  <0.001 
Lat^{2}  23.8  1.74  1, 34  0.085 
Lat^{3}  22.4  9.82  1, 34  <0.001 
Long*Lat  11.4  4.37  1, 34  0.004 
Long^{2}*Lat  6.1  2.22  1, 34  0.067 
Long*Lat^{2}  11.4  4.35  1, 34  0.003 
Alt  8.6  3.19  1, 34  0.023 
Alt_{ran}  3.6  1.28  1, 34  0.252 
T_{ann}  23.2  10.27  1, 34  <0.001 
T_{ran}  3.2  1.12  1, 34  0.339 
T_{max}  22.1  9.67  1, 34  <0.001 
T_{min}  18.6  7.79  1, 34  <0.001 
P_{ann}  6.1  2.22  1, 34  0.038 
P_{ran}  5.2  1.85  1, 34  0.089 
P_{dri}  6.0  2.17  1, 34  0.047 
P_{drn}  4.0  1.42  1, 34  0.196 
Model for E  46.5  6.73  4, 31  <0.001 
Model for S  54.0  9.08  4, 31  <0.001 
Model E+S  66.3  6.64  8, 27  <0.001 
In Chrysomelinae, the environmental model for variation in species composition included mean altitude, mean annual temperature, maximum temperature of the warmest month and minimum temperature of the coldest month (pseudoF_{4,31} = 4.13, p < 0.001), explaining 34.7% of total variability (Table
Relationships between variation in chrysomeline species composition (β_{sim}) and variables, and models for each group of variables. Percentages of variation explained are shown, with their respective PseudoF parameters, degrees of freedom (d.f.) and pvalues. S and E are the spatial and environmental models, respectively.
Variable  Variation (%)  PseudoF  d.f.  p 

log(Area)  5.2  1.88  1, 34  0.112 
Long  13.6  5.36  1, 34  <0.001 
Long^{2}  8.2  3.03  1, 34  0.009 
Long^{3}  7.4  2.71  1, 34  0.042 
Lat  17.6  7.25  1, 34  <0.001 
Lat^{2}  16.8  6.87  1, 34  <0.001 
Lat^{3}  16.0  6.48  1, 34  <0.001 
Long*Lat  12.1  4.66  1, 34  <0.001 
Long^{2}*Lat  7.7  2.82  1, 34  0.027 
Long*Lat^{2}  10.9  4.18  1, 34  0.002 
Alt  4.5  1.60  1, 34  0.119 
Alt_{ran}  3.5  1.24  1, 34  0.268 
T_{ann}  16.7  6.81  1, 34  <0.001 
T_{ran}  2.4  0.84  1, 34  0.564 
T_{max}  15.1  2.82  1, 34  <0.001 
T_{min}  15.2  6.08  1, 34  <0.001 
P_{ann}  4.9  1.76  1, 34  0.088 
P_{ran}  2.5  0.87  1, 34  0.525 
P_{dri}  4.9  1.76  1, 34  0.069 
P_{drn}  1.8  0.62  1, 34  0.794 
Model for E  34.7  4.13  4, 31  <0.001 
Model for S  47.5  7.01  4, 31  <0.001 
Model E+S  57.1  4.49  8, 27  <0.001 
Assemblage similarity was significantly related to spatial distance in both subfamilies and European regions (Northern and Southern regions, Figure
Distance decay of similarity with spatial distance in Northern Europe (solid dots, solid line) and southern Europe (hollow dots, dashed line) for cryptocephalines (a, blue), chrysomelines (b, red) and longhorn beetles (c, green). The density plots in (d) show the distribution of 1000 bootstrap replicates of the distance decay slopes (solid lines: northern Europe, dashed lines: southern Europe, colours corresponding to a, b, and c).
The most prominent pattern in species richness of both cryptocephalines and chrysomelines is the existence of a clear latitudinal gradient. This is an almost universal macroecological pattern (
Other significant variables of species richness in both cryptocephalines and chrysomelines were country area, temperature, altitude and the range of precipitations. The major difference between both groups was the significant negative relationship between richness and minimum longitude in chrysomelines but not in cryptocephalines. Therefore, species richness models were qualitatively similar in both groups, although a relevant quantitative difference was also observed. While in cryptocephalines the environmental variables were relatively more important than the spatial ones (unique contributions of 31% vs. 2%, respectively), in chrysomelines the situation was reversed, with a largest contribution of spatial rather than environmental factors (unique contributions of 24% vs. 14%, respectively). These figures may be interpreted as cryptocephaline richness being more determined by environmental conditions (mostly temperature), while chrysomeline richness being more subject of undetermined factors causing a spatial structure independent of climatic factors, likely historical factors linked to diversification and limited dispersal. This interpretation is in accordance with the fact that flightless species are common among chrysomelines but not among cryptocephalines. In a wider context, the differences between species richness models for cryptocephalines and chrysomelines suggest that the preponderance of nichebased processes related to energy and water availability dynamics (
The patterns of variation in species composition were, in contrast, very similar in both groups: several European southern regions, namely Iberian and Greek peninsulas and southern Russia, presented singular species composition differing from the rest of the continent, which harboured a relatively homogeneous fauna. Therefore, the environmental correlates of variation in species composition were very similar in both groups, including altitude and several temperaturerelated variables. The only difference was the significant relationship between species composition and precipitation of the driest quarter in cryptocephalines but not in chrysomelines, suggesting that cryptocephaline species are filtered by water availability but chrysomeline species are not. Likewise, the spatial determinants are similar in both groups, including both latitudinal and longitudinal components. The partition of variation in species composition between environmental and spatial variables revealed that a strong spatial structure (independent of environmental factors) exist in both groups, as also found in other beetle taxa (
The effects of Pleistocene glaciations seem also to be behind the differences in the patterns of distance decay of similarity between northern and southern regions. As it could be predicted from the visual inspection of maps illustrating the distribution of major clusters, the spatial turnover of cryptocephaline and chrysomeline assemblages with spatial distance is much steeper in southern Europe compared to northern Europe. As previously suggested elsewhere (
This work was supported by the Spanish Ministry of Economy and Competitiveness, and the European Regional Development Fund (2007–2013) through grant CGL201343350P.