Where Are All the Fish: Potential of Biogeographical Maps to Project Current and Future Distribution Patterns of Freshwater Species

Wolter C (2012) Where Are All the Fish: Potential of Biogeographical Maps to Project Current and Future Distribution Patterns of Freshwater Species. PLoS ONE 7(7): e40530. doi:10.1371/journal.pone.0040530 Where Are All the Fish: Potential of Biogeographical Maps to Project Current and Future Distribution Patterns of Freshwater Species Danijela Markovic 0 Jo rg Freyhof 0 Christian Wolter 0 Juan A. An el, University of Oxford, United Kingdom 0 Leibniz-Institute of Freshwater Ecology and Inland Fisheries , Berlin , Germany The dendritic structure of river networks is commonly argued against use of species atlas data for modeling freshwater species distributions, but little has been done to test the potential of grid-based data in predictive species mapping. Using four different niche-based models and three different climate change projections for the middle of the 21st century merged pairwise as well as within a consensus modeling framework, we studied the variability in current and future distribution patterns of 38 freshwater fish species across Germany. We used grid-based (11611 km) fish distribution maps and numerous climatic, topographic, hydromorphologic, and anthropogenic factors derived from environmental maps at a finer scale resolution (250 m-1 km). Apart from the explicit predictor selection, our modeling framework included uncertainty estimation for all phases of the modeling process. We found that the predictive performance of some niche-based models is excellent independent of the predictor data set used, emphasizing the importance of a well-grounded predictor selection process. Though important, climate was not a primary key factor for any of the studied fish species groups, in contrast to substrate preferences, hierarchical river structure, and topography. Generally, distribution ranges of cold-water and warmwater species are expected to change significantly in the future; however, the extent of changes is highly uncertain. Finally, we show that the mismatch between the current and future ranges of climatic variables of more than 90% is the most limiting factor regarding reliability of our future estimates. Our study highlighted the underestimated potential of grid cell information in biogeographical modeling of freshwater species and provides a comprehensive modeling framework for predictive mapping of species distributions and evaluation of the associated uncertainties. - Competing Interests: The authors have declared that no competing interests exist. Introduction Predicting future distribution patters of species following climate change projections becomes increasingly important in environmental management, species conservation and restoration planning at global, regional and local scales. Studies dealing with future species distribution patterns generally relay on niche-based species distributions models (SDMs) which relate the current conditions to the current species distributions and project these using climate change models (e.g., [1][4]). Recent studies on methodological aspects of SDMs have shown that application of consensus methods reduce model based uncertainty and increase reliability of the projections [1], [5][7]. In particular, Marmion et al. [5] have shown that consensus methods based on averaging of all methods provide robust projections and significantly increase the accuracy of species distribution forecasts. Additional important aspects of the predictive modeling process include a well substantiated predictor selection methodology and the investigation of the general predictability of future changes using the information on current conditions. Despite the overall high relevance of future projections, there is a general lack of studies comprising all these uncertainty aspects of the predictive modeling process. Species distribution patterns are affected by a combination of environmental factors acting at different spatial and temporal scales [4], [8][10]. Climate is widely acknowledged as primary factor at the continental scale whilst topography, the land use and habitats become important at regional to local scales [11]. Because dendritic river network structures constrain species dispersal ability, describing spatial distribution patterns of freshwater species is inseparable from describing hierarchy, heterogeneity and lateral connectivity of river systems [12]. The dendritic river network structure is commonly argued against grid cell related data in biogeographical modeling of freshwater species such as fish (e.g. [2], [9]). However, most studies on fish distribution patterns are void of an explicit predictor selection process and combine detailed river and fish data at the site scale with the global environmental data (20 km grid cells) resulting in serious scale mismatches and biased data samples. Freshwater biodiversity is particularly vulnerable to climate change not only because temperature is climate dependent, but also because other pressures on freshwater biodiversity such as human consumptions of ecological assets, nitrogen deposition and species invasions show increasing trends over recent decades [13] [][][16]. Since the combined effect of climate and human stressors is likely to be amplified in the future compared to their individual effects, it is expected that there will be a considerable change in species composition and diversity loss [17]. Here, we faced the challenge of modeling distribution patterns of 38 freshwater fish species across Germany at the scale grain used for the national fish records (11 km611 km grid cell). Grid cell related species distribution maps are common in presenting species occurrence information. Such data have already proven their potential to describe current and future distribution patterns of plant and terrestrial vertebrate species (e.g. [18], [19]). To our knowledge, the potential of fish occurrence maps in predictive biogeographical modeling at a comparable resolution has not been exploited. As the initial step we identified factors affecting species distributions. At the landscape to regional scale previous studies indicated the position along the upstreamdownstream gradient, mean temperature and site elevation as major predictors of species distributions [2], [9], [20]. Across river basis at continental to global scale fish species distributions were commonly described by discharge, climate, topography, net primary productivity, dam characteristics, land use properties and population density [3], [8], [16]. In a fish species traits study based on literature data Goldstein and Meador [21] showed that fish distribution patterns vary according to stream size and substrate type. Consequently, we collected information on the substrate and stream size properties of each cell as well as information on climatic, hydromorphologic, topographic and anthropogenic properties. The main objective of our study was the investigation of the potential of grid-cell data (atlas data) in b (...truncated)