Thesis title: Traits ecology and conservation of African bats in a global change context
Extended abstract
Background
Climate and land-use changes are considered the main threats to biodiversity, therefore it is crucial to understand in detail the responses of biodiversity to these changes if we want to act in terms of conservation. Deforestation and habitat loss are progressively reducing the distance between humans and wild species, leading to a potential increase in zoonotic diseases outbreak.
In this context, statistical models can help understand the effects of global change on biodiversity and the potential outcomes on the relationship between the ecology of wild species and the probability of transmission of pathogens in the areas where they live. Species distribution models (SDMs) are commonly used to forecast the expected impacts of global changes on biodiversity but often the reliability of their results is questionable. Moreover, our knowledge of biodiversity is far from complete, presenting shortfalls about the identity, evolution, distribution, and dynamics of global biodiversity. In order to obtain ecologically reliable outcomes, it is important to account for model complexities but also to consider the proper taxon for the question at hand.
The general scope of my Ph.D. is to investigate the ecology, distribution, and conservation of African bats in a global change context. I considered bats an appropriate focus for my research because of their sensitivity to climate and land-use changes, besides the fact that they are often considered as a group of natural hosts of several pathogens.
Chapter 1 - AfroBaT: a database on African Bats Traits
Trait-based approaches are becoming extremely common in ecological modeling as well the availability of traits databases is increasing. However, data availability is often biased towards particular regions and taxa, with bats often under-represented.
In Chapter 1, I present the AfroBaT database, a compilation of trait data on 314 African bat species containing 27,004 values for 86 traits focusing on morphology, reproduction, life-history, and trophic ecology. I gathered all trait data from published literature following the ecological trait-data standard procedure and ensuring a complete harmonization with other existing databases. For missing data, I used a data imputation procedure with a machine learning approach considering both numerical and categorical traits and including species phylogeny. Trophic ecology traits showed the highest coverage in literature (73% of the species on average), while reproductive traits the lowest (27%). The data imputation procedure improved the coverage of AfroBaT especially for reproductive traits, reaching a total of 59% of the species covered. AfroBaT has a range of potential applications in macroecology and community ecology, and the availability of open-access data enables collaboration and data-sharing among researchers.
Chapter 2 - Is geographic sampling bias representative of environmental space?
Species occurrence data from public repositories are widely used in biogeography and conservation research. However, these data are prone to several sampling biases limiting their usefulness in biodiversity studies, and particularly in species distribution models (SDMs). Specifically, a geographic sampling bias can lead to overfitted SDMs while an environmental sampling bias can affect the estimate of the true environmental space occupied by a species. Several methods can be used to correct for sampling biases, but most of these are focused on the geographic component only.
In Chapter 2, I assessed if sampling bias correction based on geographical factors can represent a viable solution also for environmental bias. I focused this analysis on plants and invertebrates in Africa, being the main food resources of bats. To model sampling bias in the spatial dimension I estimated sampling rates as a function of distance from a set of potential bias factors (roads, cities, waterbodies, and coastline) in a Bayesian framework. To model sampling bias in the environmental dimension I calculated the environmental multivariate distance between the average climate in the species' occurrences and the rest of the study area by using a set of bioclimatic variables.
I quantified the spatial relationship between geographic and environmental bias calculating a Local Indicator of Multivariate Spatial Association (LISA), highlighting local clusters of spatial correlation. For both plants and invertebrates, almost 50% of Africa showed a non-significant relationship between the geographical and the environmental bias components, indicating the absence of a direct correspondence between the two types of biases. Furthermore, from 20.4% to 24.4% of the study area showed an opposite pattern between the two biases, clearly indicating that a correction in the spatial component would be detrimental in the environmental component. These findings showed that geographical factors cannot be considered as good proxies for environmental bias and, consequently, I suggest considering both geographic and environmental corrections when sampling bias is a problem.
Chapter 3 - Not only climate: The importance of biotic interactions in shaping species distributions at macro scales
Abiotic factors are usually considered key drivers of species distribution at macro scales, while biotic interactions are mostly considered important at local scales. A few studies have explored the role of biotic interactions at macro scales, but all considered a limited number of species and obligate interactions.
In Chapter 3, I examined the role of biotic interactions in large-scale SDMs by testing two main hypotheses: 1) biotic factors in SDMs can have an important role at continental scale; 2) the inclusion of biotic factors in large-scale SDMs is important also for generalist species.
I used a maximum entropy algorithm to model the distribution of 177 bat species in Africa calibrating two SDMs for each species: one considering only abiotic variables (noBIO-SDMs), and the other (BIO-SDMs) including also biotic variables (trophic resource richness). I focused the interpretation of the results on variable importance and response curves. For each species, I also compared the potential distribution measuring the percentage of change between the two models in each pixel of the study area.
All models gave AUC >0.7, with values on average higher in BIO-SDMs compared to noBIO-SDMs. Trophic resources showed an importance overall higher level than all abiotic predictors in most of the species (68%), including generalist species. Response curves were highly interpretable in all models, confirming the ecological reliability of the models. Model comparison between the two models showed a change in potential distribution for more than 80% of the species, particularly in tropical forests and shrublands.
These results highlight the importance of considering biotic interactions in SDMs at macro scales. I demonstrated that a generic biotic proxy can be important for modeling species distribution when species-specific data are not available, but a multi-scale analysis combined with a better knowledge of the species might provide a better understanding of the role of biotic interactions.
Chapter 4 - Conservation of African bats’ diversity
Biodiversity plays a crucial role in supporting human well-being, however climate and land-use changes are leading to a decline in global biodiversity. Despite the efforts to protect biodiversity, conservation programs are usually based on species richness (e.g., taxonomic diversity, TD) or different irreplaceability measures (e.g., endemism or rarity) often neglecting phylogenetic (PD) and functional (FD) diversity among species. PD has been proposed in conservation planning as a good proxy for functional traits, albeit the spatial congruence between these two facets of biodiversity has been recently questioned.
In Chapter 4, I explored multiple facets of the diversity of African bats (taxonomic, phylogenetic, functional, rarity) to evaluate their representation in the current network of protected areas in Africa. I modeled the distribution of 238 African bats at 1 km2 resolution with climatic, environmental, and trophic resource variables. I computed TD, PD, and FD through Hill numbers, while I weighted the potential distribution of each species for their global geographic ranges to obtain a range-size rarity measure (RAR). I fitted a generalized additive model (GAM) between each facet of diversity against TD, and I calculated the z-scores of the residuals to identify the areas with higher and lower values than expected from TD. I then compared the distribution of the values of each diversity index falling within the PAs with their values across the entire continent to evaluate their representation in the current network of protected areas (PAs) in Africa.
Hotspots of multifaceted diversity of African bats were located in the Congo basin, the Ethiopian plateau, and south-east Africa, with highest values of RAR in Madagascar. The residuals maps of PD and FD gave an opposite pattern when compared to TD, highlighting that not only considering species richness alone may result in inadequate conservation planning but also using facets of biodiversity as surrogates may underestimate essential evolutionary history and ecosystem functions. The current network of protected areas are not concentrated in areas of high diversity of bats, highlighting the inadequacy of the current PAs network for the African bat fauna.
These findings refute the assumption that the protection of areas with high phylogenetically distinct species would maximize the protection of ecosystem functioning. Future studies should explore integrative approaches investigating the spatial mismatch among diversity facets in order to optimize the conservation of evolutionary and functionally diverse assemblages.
Chapter 5 - Hidden diversity in bat-associated viruses in Africa: ecological correlates and global change effects
Climate change and land-use change, through habitat loss and fragmentation, are progressively reducing the distance between humans and wild species, leading to a potential increase in zoonotic diseases outbreak. Knowledge about the presence and distribution of viruses and their hosts is still lacking, with important virus reservoirs almost completely unknown, especially in highly diverse areas like tropical regions. More than 200 viruses have been isolated or detected in bats, often considered as a group of natural hosts of several pathogens albeit poorly studied compared to other taxa.
In Chapter 5, I identified the current and future distribution of known and hidden diversity of zoonotic viral families associated with African bats. I mapped the distribution of 199 African bats at 1 km2 resolution with climatic, environmental, and trophic resource variables. I projected the models across the African continent under future scenarios of greenhouse gas emissions, using five Ground Circulation Models (GCMs) and three Shared Socioeconomic Pathways (SSP) from CMIP6 scenarios. Using a trait-based approach, I predicted the potential of each species to host viruses within seven viral families with known zoonotic species (Coronaviridae, Filoviridae, Flaviviridae, Rhabdoviridae, Paramyxoviridae, Reoviridae, Hantaviridae). By using functional traits involved in bat-virus association as explanatory variables, I applied a Random forest algorithm to each viral family. To identify hotspots of current and future hidden diversity of each viral family, I mapped the spatial distribution of bat species potentially hosting each viral family.
Bat family, colony size, sampling effort, geographic range size, and reproductive traits were the most important traits in predicting viral family presence. These results suggest that the west coast, the Congo basin, the Ethiopian plateau, eastern Africa, and eastern Madagascar are currently the areas with the highest hidden diversity highlighting how much our knowledge of bat-virus association is lacking. This pattern is even stronger looking at the future scenarios maps (e.g., SSP5-8.5) providing evidence of the importance of considering both the hidden diversity and future scenarios to identify the areas where surveillance is needed. This framework can easily be adapted to include other drivers of host-virus transmission (e.g., livestock, bushmeat consumption, rural population density), supporting the identification of areas at greater risk of zoonotic emergence where a preventive approach can be implemented to avoid future potential threats to human and planetary health.
Conclusions
Although bridging the state of knowledge on biodiversity remains a fundamental and urgent challenge, particularly in the regions with the highest level of biodiversity (e.g., tropical areas), this Ph.D. project represents a starting point to fill the knowledge gap on understudied taxa such as bats, providing important caveats for modeling species distribution in a global change context.
My Ph.D. project is the first study investigating the ecology, distribution, and conservation of all African bats at a continental scale using high-resolution data, providing a better understanding of this unique order of mammals. The AfroBaT database will provide an unprecedented collection of traits completely based on published literature, with a range of potential applications in macroecology and community ecology. Identifying important biological traits for predicting the probability of presence of zoonotic viral families associated with bats can support decision-making in priority areas for surveillance, where new outbreak events may be likely to occur. Spatial mismatch among the multiple facets of bat diversity highlights the need for a more integrative conservation planning that can include the bat fauna in the African network of protected areas, currently underrepresented.