Dottorato in BIOLOGIA AMBIENTALE ED EVOLUZIONISTICA

Titolo della tesi: Assessing species’ exposure to climate change to support extinction risk assessments for the IUCN Red List

The IUCN Red List of Threatened Species (hereafter IUCN Red List) is the most authoritative tool to assess extinction risk of species globally. More than 160,000 species have already been assessed, with more than 46,000 identified as threatened with extinction. The IUCN Red List classifies species into nine categories of risk of extinction according to five quantitative criteria, that measure the responses of species to the drivers of extinction, such as population decline, small ranges and small population size. The importance of the IUCN Red List is widely recognized as it is used to assess progress towards international strategic plans and agreements and to evaluate the potential biodiversity impacts of development projects. Thus it is crucial to facilitate and strengthen the ability of IUCN Red List assessors to consistently identify the impact of global drivers of extinction. Climate change has rapidly become one of the major threats to biodiversity. Many organisms face shifting climatic conditions, resulting in range contractions or population decline, eventually leading to extinction. Projections of future climate change impact predict that more than 7% of species globally might face extinction. Species responses to climate change are mediated by life-history traits. Specialist species, species with slow life-histories (e.g., species living longer) and species already threatened are expected to be more vulnerable to climate change. Among them, ectothermic groups are of particular concern. In fact, their physiology and reproduction are strictly linked to environmental temperature and they might not be able to cope with changes in climatic conditions. Despite the impact of climate change has been widely recognized, its consideration into IUCN Red List assessments is still limited. Just a minority of species presented information on climate change in the last assessment and this information is biased towards endothermic groups (birds and mammals). There are indeed many factors hampering the consistent evaluation of climate change impact in IUCN Red List assessments. While an increasing body of scientific literature focuses on climate change impact on biodiversity, such studies cannot be used in Red List assessments as they do not comply with IUCN Red List guidelines, and are not even comparable among them. Species distribution models (SDMs) are often used to measure climate change impact, but methodological choices are delegated to modelers preferences, such as the choice of background or pseudoabsence points, the selection of important variables or scenarios of adaptability through dispersal. Moreover, the time frame considered in such studies is usually a fixed time frame (e.g., until 2100), while IUCN Red List criteria explicitly require a time frame based on the generation length of species. This is problematic as generation length data are available just for birds and mammals, leading to a bias in the application or Red List criteria. The overarching goal of this PhD project is to improve the ability of Red List assessors to consistently evaluate climate change impact in extinction risk assessments. There are three specific objectives: 1) providing a standardized framework to measure the impact of future climate change, complying with IUCN Red List guidelines and focusing on IUCN Red List criterion A3; 2) estimating the generation length of amphibians and reptiles; 3) suggesting how to select biologically relevant variables for SDMs applications. Each objective has been developed into a research Chapter. To achieve the first research objective, I devised a standard framework to consider the impact of future climate change on species extinction risk. This risk is evaluated in terms of change in habitat quality, to inform Red List criterion A3 on future population reduction. I considered different algorithms, general circulation models and representative concentration pathways to address uncertainties associated with models settings and future climate projections. I also considered two climate adaptability scenarios through dispersal: the no-dispersal scenario in which I assumed the species could not disperse outside the range, and a dispersal scenario, in which I calculated a dispersal buffer based on dispersal and generation length estimates, and I assumed the species being able to reach those areas. This was needed to provide a range of measures of potential climatically suitable areas lost, but also gained. I applied this framework to 1,493 species of amphibians, mammals and reptiles. Although this was a modeling exercise to show the applicability of my framework, I found 13% of analyzed species might worsen their extinction risk category only due to climate change. However, I expect more species to be threatened as climate change likely cumulates its effect with other threats. This framework can be easily adapted to further Red List requirements and it will be the starting point for the climate change module in the sRedList platform, which will ensure user-friendly access to my methodology to all assessors. As part of the second research objective, I estimated the generation length of more than 13,000 species of herptiles and filled-in this data gap for more than 73% of all reptiles and 57% of all amphibians. I modeled generation length as a function of species size, climate, life history, and phylogeny using generalized additive models and phylogenetic generalized least squares. I found that generation length was strongly related to species’ body size and temperature across all taxa, with larger species and species living in cold climates showing the highest generation length. Response of climate seasonality differed between amphibians and reptiles, but still meeting my ecological expectations. I used the fitted models to predict generation length of many species. Models performed well for some of the most specious families (e.g., Bufonidae among amphibians, Lacertidae and Colubridae among squamates, and Geoemydidae among testudines), while I found high uncertainty around the prediction of a few families, notably Chamaeleonidae. These new estimates will have a crucial role in biodiversity conservation, especially in Red List assessments. They will allow the application of many Red List criteria for herptiles, currently unused due to the lack of generation length. These estimates can also play a critical role in reducing the proportion of Data Deficient species on the IUCN Red List. Finally, they can be used in assessments focusing on climate change defining the dispersal buffer where to predict habitat suitability. Although my estimates are not meant to replace robust and validated field studies or museum records, they can help mitigate biases in conservation assessments until comprehensive data on the generation length of herptiles become available. In my third research objective, I address the issue of selecting biologically relevant variables for species distribution models. Several studies proposed a statistical variable selection based on filters for collinearity or correlation, and only limited to the standard set of 19 bioclimatic variables. However, such approaches might select variables based on spurious correlations found within available species occurrences. This is problematic when models are used to estimate probability of habitat suitability under future conditions, for example future climate, risking to provide unreliable results. Thus, when such models are used in extinction risk assessments, the resulting risk category might be inconsistent. I used three case study species (Boreus westwoodi, Crucianella maritima, and Rhinolophus hipposideros) to demonstrate that biologically relevant variables can be selected even with broad knowledge on species biology. I demonstrated that predictions from models based on expert based and naive variable selection can be dramatically different. This highlighted the need to consider species ecology when selecting relevant variables to model species distribution, as just relying on the 19 bioclimatic variables and a statistical variable selection can be misleading and not representative of the species niche. This information will be key to increase biological consistency of SDMs and consequently improve extinction risk assessments when SDMs are used to predict species extinction risk. Overall, my PhD thesis will help the work of Red List assessors. Standardizing the quantification of climate change impact in extinction risk assessments will improve consistency and reduce methodological discrepancies. My framework, which complies with IUCN Red List guidelines, can be used in formal extinction risk assessments and will help reduce bias in the quantification of climate change impact on the IUCN Red List. Estimating generation length for amphibians and reptiles addresses a key data gap, allowing a more balanced assessment of climate change impact across tetrapods, and reducing the inherent bias in application of Red List criteria. Helping the identification of biologically relevant variables for species distribution models can reduce the risk of misleading predictions about habitat suitability under changing climates. By integrating quantitative methods and ecological knowledge, this PhD research will improve extinction risk assessments that might inform more effective biodiversity conservation strategies.