Mapping species distribution ranges by means heterogeneus data

Resumo

The mapping of species ranges is one of the most relevant and widely used pieces of information in the study of biodiversity. Knowing the distribution range of species is a fundamental first step us to understand the factors that determine those distributions, as well as the patterns in the richness and abundance of species in a biogeographical context, all this being necessary information to establish conservation strategies. The distribution range is a conceptual construction that describes the area where a taxon occurs. The basic units of information for constructing these ranges are spatially and temporally referenced observations of species (i.e. records). Direct field sampling on very large spatial scales is rarely feasible, as it requires significant resources and time. Therefore, large-scale biodiversity analyses tend to be based on a variety of data reporting information on species observations or distributions, ranging from point location data obtained from databases or wildlife atlases to species distribution maps based on expert knowledge. In spite of been essential, our knowledge on the distribution of species is far from complete, even for the best studied taxa. Given the great relevance of species distribution maps, it is surprising to note that very little attention has been paid to analyse how these maps are affected by the quality of the baseline data and the diversity of methods used to construct them. This is the central axis of the thesis, which structured in four main chapters. In Chapter I we conducted a bibliographic review in order to obtain information from scientific publications that use species distribution ranges in their studies. We noted how distribution ranges have been generated and identified which are the most commonly used methods to generate distribution ranges from georeferenced data, along with the advantages and disadvantages provided by each of them. Most often researchers do not provide information on how ranges have been constructed. The lack of explicit information on the data and methods used in the construction of distribution ranges severely affect the interpretation of results. Finally, the methods commonly used to delineate the areas have been insufficiently evaluated. We urge researchers to be explicit both in what they consider the ranges of distribution of species and in the methods they use to generate them. This will allow for more robust comparisons between the ranges of distribution of species generated by different methods. In Chapter II we assessed the accuracy of five geographic algorithms commonly used to delineate species ranges with the aim of providing guidelines to minimize Type I error and maximize sensitivity of the resulting species ranges. To this aim, we generated hypothetical range areas with the same total surface but varying in shape, number of fragments, heterogeneity in fragment size and simulated sets of species records varying in numbers, spatial distribution and presence of errors and biases. The recommended algorithms have been Adaptive Local Convex Hull (a-LoCoH) and Kernel Density Estimation (KDE). KDE algorithm has the highest sensitivity and a-LoCoH algorithm has the lowest type I error rate. Both behaved similarly well when describing range fragmentation. We provide recommendations to minimize the effects of data quantity and quality, and provide guidance to choose an algorithm when defining species distribution ranges based on species observations. Chapter III of this thesis explores options for a systematic and replicable generation of range maps that take into account the different sources of variability and the exponential increase in the availability of species records. We offer a unified and repeatable methodology for building species range maps, which we compare with the existing maps of the International Union for Conservation of Nature (IUCN). The combination of IUCN distribution maps with georeferenced species data available from the Global Biodiversity Information Facility (GBIF) is a promising route to providing information on where mapped distributions are reliable and where they are uncertain. Lack of information or availability of information in certain areas makes it difficult to implement systematic approaches to the construction of distribution maps. So we also reveal priority sites for lack of information or sampling effort on a global scale. Chapter IV assesses the variability in the description of species distribution ranges based on non-systematic data gathering (e.g. using records from available databases) or on systematic and specific surveys. As a case study, we used the southern water vole (Arvicola sapidus) in peninsular Spain, using the results of a citizen science initiative specifically focussed on this species and comparing them with those of a previous atlas. The resulting distribution maps had notable differences, which were related to identification errors and heterogeneous sampling effort in the non-systematic dataset as well as to actual changes in range due to predation by invasive American mink. The likelihood of commission errors increases in areas where there are species that may be confused with the water vole and by mink predation. The probability of omission errors increases in areas with low sampling effort and the existence of rodents easily confused with the study species. We emphasize the need to be cautious in using available information sources to generate range maps, particularly in areas with little data or signs of heterogeneous spatial coverage. In conclusion, this thesis explores the different dimensions of species distribution maps and offers a necessary perspective to deal with problems posed by sciences such as ecology or conservation biology. We also try to understand the nature of the uncertainty involved in distribution maps to help interpret existing results and guide future research. The information metrics developed throughout this thesis could be incorporated into online tools that allow researchers and funding agencies to identify priority species and areas to improve information sources along with their associated distribution maps.