• Energy Research

AbstractQuestionsSpecies–area relationships (SARs) are fundamental for understanding biodiversity patterns and are generally well described by a power law with a constant exponent z. However, z‐values sometimes vary across spatial scales. We asked whether there is a general scale dependence of z‐values at fine spatial grains and which potential drivers influence it.LocationPalaearctic biogeographic realm.MethodsWe used 6,696 nested‐plot series of vascular plants, bryophytes and lichens from the GrassPlot database with two or more grain sizes, ranging from 0.0001 m² to 1,024 m² and covering diverse open habitats. The plots were recorded with two widespread sampling approaches (rooted presence = species “rooting” inside the plot; shoot presence = species with aerial parts inside). Using Generalized Additive Models, we tested for scale dependence of z‐values by evaluating if the z‐values differ with gran size and tested for differences between the sampling approaches. The response shapes of z‐values to grain were classified by fitting Generalized Linear Models with logit link to each series. We tested whether the grain size where the maximum z‐value occurred is driven by taxonomic group, biogeographic or ecological variables.ResultsFor rooted presence, we found a strong monotonous increase of z‐values with grain sizes for all grain sizes below 1 m². For shoot presence, the scale dependence was much weaker, with hump‐shaped curves prevailing. Among the environmental variables studied, latitude, vegetation type, naturalness and land use had strong effects, with z‐values of secondary peaking at smaller grain sizes.ConclusionsThe overall weak scale dependence of z‐values underlines that the power function generally is appropriate to describe SARs within the studied grain sizes in continuous open vegetation, if recorded with the shoot presence method. When clear peaks of z‐values occur, this can be seen as an expression of granularity of species composition, partly driven by abiotic environment.

Species distribution models (SDMs) are useful tools for future potential distribution patterns of species in the face of climate change. Turkey is expected to be affected considerably from climatic change i.e., up to 6°C increase in temperature and 50% decrease in precipitation by 2070. Therefore, there is an urgent need for conservation and management practices for future patterns of species. It is aimed current and future (using CMIP5 projected to 2070) potential distribution areas of Campanula lyrata Lam., which is formerly an endemic species. To do this, presence-only data was used, which is obtained from the Global Biodiversity Information Facility (GBIF). Bioclimatic data from was downloaded from WorldClim dataset with 10 km2 resolution. Species distribution modelling was performed using R program. Two regression techniques and two machine learning techniques namely Generalized Linear Models (GLMs), Generalized Additive Models (GAMs), Support Vector Machine (SVM) and Random Forest (RF), were used, respectively. The bootstrapping method as partitioning resampling was also used for all analysis. Considerably high model performances as well as AUC values for all possible models were found. Significant range shifts between current and future climatic conditions were found. The most relevant relative importance variables were precipitation seasonality and precipitation of the wettest month. This study reveals the importance of the future distributional areas of species.Species distribution models (SDMs) are useful tools for future potential distribution patterns of species in the face of climate change. Turkey is expected to be affected considerably from climatic change i.e., up to 6°C increase in temperature and 50% decrease in precipitation by 2070. Therefore, there is an urgent need for conservation and management practices for future patterns of species. It is aimed current and future (using CMIP5 projected to 2070) potential distribution areas of Campanula lyrata Lam., which is formerly an endemic species. To do this, presence-only data was used, which is obtained from the Global Biodiversity Information Facility (GBIF). Bioclimatic data from was downloaded from WorldClim dataset with 10 km2 resolution. Species distribution modelling was performed using R program. Two regression techniques and two machine learning techniques namely Generalized Linear Models (GLMs), Generalized Additive Models (GAMs), Support Vector Machine (SVM) and Random Forest (RF), were used, respectively. The bootstrapping method as partitioning resampling was also used for all analysis. Considerably high model performances as well as AUC values for all possible models were found. Significant range shifts between current and future climatic conditions were found. The most relevant relative importance variables were precipitation seasonality and precipitation of the wettest month. This study reveals the importance of the future distributional areas of species.