Emerging horizons in predictive biogeography

The notion that different branches of biological sciences – including ecology, macroecology, and biogeography – should adopt a predictive focus rather than merely aiming to describe and understand the natural world has gained traction over the past decades (Peters 1991, Shrader-Frechette and McCoy 1993). This trend has been enabled both by technological advancement leading to new predictive frameworks, and by the pressing societal demands to anticipate and mitigate the effects of global change on biodiversity and the associated ecosystem services.

An early example of this trend is the work by Sánchez-Cordero et al. (2004) who contributed a chapter on predictive biogeography for conservation applications in a seminal volume on biogeography (Lomolino and Heaney 2004). While the authors did not explicitly define the term predictive biogeography, their discussion emphasized how developments in statistical ecology and mapping had allowed the description of species distributions at large spatial scales. Similarly, Thuiller et al. (2006) employed the concept of predictive biogeography in the restricted context of describing the use of stacked species distribution models (SDMs) in predicting plant richness in South Africa.

Dawson et al. (2011) subsequently highlighted SDMs as the most widely used predictive method in biogeography, but also called attention on the importance of establishing broader frameworks to anticipate changes in biodiversity, from species to ecosystems, in response to climate change. There are other biogeographic patterns that are widely used in a predictive context. Most notably, the species area relationships (SARs), which have also been important to understand and predict species extinctions (Drakare et al. 2006) driven by anthropogenic habitat fragmentation for example. However, the widespread use of SDMs, along with the fact that they remain the method of choice at large scales in ecology, has been repeatedly highlighted (Bellard et al. 2012, Araújo et al. 2019, Zurell et al. 2020, Soley-Guardia et al. 2024). Mapping biodiversity remains an essential component of large-scale spatial conservation planning (Margules et al. 2002). It is critical not only for delineating species conservation statuses, trends, and management strategies on regional to global scales, but also for interpreting the geological, historical, and anthropogenic causes and consequences for biodiversity distribution (Whittaker et al. 2005). Therefore, describing and modelling species distributions will probably remain an essential component of predictive biogeography.

However, many studies emphasize the need to move beyond individual species distributions to encompass a broader range of spatio-temporal issues at the interface between biodiversity sciences and society, such as ecosystem services, and their effects on human health and agricultural systems. This expanded scope inevitably calls for a wider definition of predictive biogeography.

In this special issue, we aim to broaden the scope and application of predictive biogeography, moving beyond the confines of SDMs to spotlight cutting edge research across different dimensions of the field. The deliberate use of the term biogeography, as opposed to ecology or macroecology, reflects our intent to include a more diverse array of approaches – statistical, evolutionary, historical, geological, and more – that contribute to understanding and forecasting distribution, abundance, and diversity across broad spatial and/or temporal scales. This scope includes not only natural systems but also productive systems (e.g. agroecosystems).

We propose a definition of predictive biogeography as a subdiscipline of biogeography that uses known ecological or evolutionary patterns and processes to predict the abundance, distribution, and diversity, whether it be at the species, intra-, or inter-specific levels, including biotic interactions and their relationship with the environment, over broad spatial and temporal scales. Over the past two decades, this research field has experienced exponential growth, driven by the increasing availability of digital data on the distribution of species and the genetic variability within them, as well as the proliferation of spatially explicit environmental data layers and increasingly fine spatial and temporal resolutions. This rapid evolution has catalysed the development of new syntheses and theories, alongside advancements in methodologies and computational capabilities. As a result, biogeography is undergoing a transformation from the primarily descriptive discipline championed by the likes of Alexander von Humboldt (1769–1859), Augustin Pyramus de Candolle (1778–1841), Alfred Russel Wallace (1823–1913), and Philip Lutley Sclater (1829–1913), amongst others, to a predictive science, capable of informing both fundamental research and practical applications in conservation, resource management, and beyond. The emergence of predictive biogeography as a discipline has been driven primarily by a pressing societal demand (Dietze et al. 2018, Enquist et al. 2024). The growing array of global challenges – including the widespread decline in biodiversity, rising food demands, and the far-reaching impacts of recent global pandemics – paired with ongoing global changes that threaten biodiversity, ecosystem services, food security, and public health, have made the ability to anticipate these changes across broad spatial and temporal scales an existential priority for humanity.

Over time, the focus of predictive biogeography has expanded. Initially, in the 1990s, its scope centred largely on modelling the past, present, and future distribution of biodiversity. Today, its applications have evolved to address challenges more directly linked to human societies, such as food production and public health (Enquist et al. 2024). This broader relevance has positioned predictive biogeography as a critical discipline underpinning advancements and applications across a wide range of fields (Araújo and Peterson 2012). These include conservation biology (Araújo et al. 2011, Fordham et al. 2013), agriculture (Meynard et al. 2017, Gerber et al. 2024, Soubeyrand et al. 2024), forestry (Zhang et al. 2022, Rosa et al. 2024), fisheries (Cheung et al. 2010, Boavida-Portugal et al. 2018), epidemiology (Aliaga-Samanez et al. 2024, Mestre et al. 2024), and paleobiology (Metcalf et al. 2014, Mestre et al. 2022), reflecting its versatility in addressing contemporary and future global issues.

The emergence of new technological advances in all areas of ecology, biology, and computer science has translated into a vast availability of high-resolution information for large geographic areas, from landscapes, to countries, continents, and even globally. Technological advances include molecular biology and sequencing, which make large-scale biodiversity monitoring, even of microscopic life, possible (Beng and Corlett 2020). These DNA recovery efforts can go so far as to sequence DNA from ancient samples, allowing the exploration of genetic diversity from old specimens stored in museum collections (Raxworthy and Smith 2021), or recovering trophic relationships through environmental samples (Pereira et al. 2023). Sequencing, along with analytical and theoretical advances, makes it possible to integrate understanding of evolutionary history, rates and processes of diversification (Morlon et al. 2010, Kergoat et al. 2018) into ecological predictions at large scales. Remote sensing to follow large-scale land use transformation (Cavender-Bares et al. 2022), integrating it with chemical properties and/or mapping of phylogenetic and functional diversity (Cavender-Bares et al. 2020), as well as large-scale or even global microclimate mapping at fine temporal resolutions (Lembrechts et al. 2020) are among the many promising developments that allow integrating fine-grain mechanisms and patterns into large-scale models. Statistical methods and computing advances have also been fundamental in this field (Record et al. 2023), as well as the computer technology that allows sharing data globally, including curated species, occurrence, trait, phylogenetic, and any other type of ecological datasets. These are just a few of the many technological advances that have allowed expanding the extent at which fine-resolution biodiversity data can be gathered. When combined, applied, and used at large scales, these new methods can greatly advance our understanding of biogeographic patterns in the past, present, and future.

Within these bounds, we can identify at least three fundamental components of any predictive biogeography framework (Fig. 1): biodiversity and environmental data, both of which must encompass large temporal and/or spatial scales to fall within the domain of biogeography; one or more scenarios that establish the context for relevant predictions; and a formal model or theory that translates our current understanding of biodiversity–environment relationships into the scenarios considered. Note that scenarios often pertain to environmental change (e.g. climate or land-use change scenarios), but they can also include evolutionary scenarios, extinction scenarios, management strategies, human behaviour, or any other processes driving large-scale predictions. Importantly, we view these components as dynamic rather than static. Advances in data and scenario development should lead to updates in models, theories, and predictions. In turn, model outputs and data requirements can guide the collection of data and the refinement of scenarios, creating a positive feedback loop (Dietze et al. 2018).

Each of these three components can involve a plethora of elements. For example, biodiversity data can include gene expression profiles, intra-specific genetic diversity, and intra- and inter-specific functional traits, among others (Fig. 1). Despite significant progress, the technologies enabling the measurement, monitoring, and characterization of biological and environmental data, as well as scenarios, continue to evolve. There remains considerable potential for innovation in relating biodiversity to environment factors, imagining new scenarios, and enhancing predictive capabilities.

When curating papers for this special issue, we aimed for broad interdisciplinary integration. However, many of the compiled studies revolve around SDMs, which remain a core predictive tool across broad spatial and temporal scales. For example, Boom and Kissling (2024) propose that tracking data can complement traditional occurrence data, improving SDM predictions. Chronister et al. (2024) demonstrate the potential of automated acoustic detectors to monitor and distinguish juvenile and adult great horned owls, opening the door for estimating demographic parameters at very large scales. By incorporating such demographic data into SDMs, researchers can explore habitat use at different life cycle stages – a critical factor to consider when setting species conservation priorities. Goicolea et al. (2024) employ hierarchical modelling to refine SDMs, combining locally calibrated models nested within regionally constrained ones. This approach mitigates the common problem of truncating the species environmental range when calibrating local distribution models (Thuiller et al. 2004). Similarly, Mowry et al. (2024) used hierarchical modelling to account for constraints related to the potential distributions of disease vector distributions – ticks, in this case – resulting in improved disease distribution estimates.

Several studies featured in this issue leverage the interplay between genetic differentiation and populations and distributions. Naughtin et al. (2024) use large-scale genetic structure data alongside SDM-based reconstructions of past potential ranges to infer, via approximate Bayesian computation (ABC) models, the most likely combinations of climate models and statistical SDMs that matches the current differentiation structure. The authors argue that this approach can help rank SDMs that are otherwise indistinguishable using standard occurrence validation methods. In another application, Mascarenhas and Carnaval (2024) employ random forest models to explore how genetic differentiation relates to life history traits, particularly dispersal and demographic characteristics. Their results highlight the importance of incorporating dispersal traits for understanding arthropod phylogeography. Hernández et al. (2024) propose a formal theoretical framework linking environmental suitability, as modelled by SDMs through deep time intervals, with genetic diversity. This theoretical integration produces interesting predictions regarding range stability over paleological periods and its relationship to current genetic structures, enabling the identification of endemic regions and poorly surveyed genetic patterns. Along similar lines, Formoso-Freire et al. (2024) relate species abundance distributions and species genetic distributions, investigating how long-term climate stability informs present-day community stability.

This special issue also includes advancements in theoretical modelling. Sharma et al. (2024) propose a framework for integrating phylogenetic constraints into niche evolution. The authors demonstrate the utility of the method with a case study using hummingbirds. Verdon et al. (2024) demonstrate the potential of combining eDNA with SDMs to estimate soil diversity for taxa that have been traditionally overlooked in biodiversity monitoring. This ambitious modelling effort incorporates numerous amplicon sequence variants (ASVs), revealing both the capabilities and limitations of current technologies and modelling approaches. As discussed by the authors, improving biodiversity dynamics models for these systems will require better estimates of species abundance from eDNA data, as well as enhanced soil-related layers and scenarios of large-scale soil change.

Another recurring theme in the issue is the incorporation of species interactions into SDMs – one of the most pressing challenges in the context of climate change. The success of species adapting their ranges to changing climates largely hinges on their interactions with other species (Araújo and Luoto 2007). Poggiato et al. (2025) tackled this issue with Bayesian modelling using current data, while González-Trujillo et al. (2024) employ the phenomenological modelling of community trophic structures proposed by Mendoza and Araújo (2019, 2022). This framework allows the authors to hindcast trophic guild distributions and richness over different latitudes, enabling exploration of the effects of past climate changes on these interactions.

Predictive biogeography beyond the traditional scope of SDMs is also represented in this issue. Park et al. (2024) present a simulation study demonstrating how plant specimens in museum collections can be used to estimate not only the median flowering dates and their relationships with mean temperatures but also the onset and termination of the flowering periods. This approach offers a valuable tool for inferring flowering phenology in species with strong museum representation, thus helping understanding phenological shifts driven by climate change.

Siders et al. (2024) capitalize on a comprehensive literature review to extract data from shark tracking devices and comparing vertical diversity distributions with and without depth-weighted information. Their results show that depth preference can add important information to understand current vertical diversity distribution among sharks, including both phylogenetic and functional components. Adding depth as a third dimension to study marine biogeographic patterns seems like a promising venue for future research, one that has only recently been made available thanks to the accumulation of biotelemetry technology as well as 3-D environmental data layers across the ocean (Fragkopoulou et al. 2023). Finally, Lertzman-Lepofsky et al. (2024) take advantage of two global biodiversity databases to explore the role of trophic interactions in explaining abundance correlations between taxa. Their analysis demonstrates that incorporating co-variations in abundance – when interactions are well-documented – enhances predictions of abundance changes over time.

In summary, these studies exemplify how technological innovations are reshaping our ability to monitor, understand, and predict various components of abundance, distribution, and diversity across broad spatial and temporal scales. From genetic population structures to taxonomic, functional, and phylogenetic diversity, the field is evolving rapidly. Emerging methods now include previously invisible or challenging-to-monitor aspects of biodiversity, facilitated by tools such as eDNA, automated detection technologies (sound and telemetry), statistical modelling, and big data integration. An exciting direction in predictive biogeography involves utilizing deep-time biodiversity patterns to inform our understanding of the present and forecast future changes. Advances in sequencing technologies have also opened new possibilities for examining genetic variation on large scales, forging compelling connections between ecology and evolution.

Despite advances, there remain important gaps to address in predictive biogeography. As is often the case with ecological publications at large (Maldonado et al. 2015, Nuñez et al. 2021), our collection includes only one study focused on tropical regions (Mascarenhas and Carnaval 2024). Moreover, the technologies represented in this issue account for only a small subset of those illustrated in Fig. 1c. While biogeography plays a crucial role in monitoring biodiversity changes on a global scale, this collection features limited exploration of monitoring strategies or the scaling of ecological theories to larger contexts. These gaps underscore the vast number of unexplored possibilities for advancing predictive biogeography. For example, could we combine text mining, automated monitoring of biodiversity, and citizen science – engaging individuals with everyday tools such as cell phones – into multi-modal models capable of real-time trend analysis? Such an approach could enable early detection of population declines and species range shifts. Could genomics and epigenetics offer deeper insights into genotype-to-phenotype relationships, improving our understanding of climate change adaptation and prioritizing populations for conservation at the range level? Furthermore, could technological innovations facilitate a ‘macroscope' for biodiversity analysis and monitoring (Gonzalez et al. 2023), bridging the gap that often leaves the Global South underrepresented in our global datasets? These questions only scratch the surface of what could be achieved as we push the boundaries of predictive biogeography.

While this special issue does not aim to represent exhaustively the global literature, the prevalence – or absence – of certain methods within it reflects real biases in the current state of predictive biogeography. For example, none of the studies in this special issue uses ancient DNA (Lagerholm et al. 2017, Raxworthy and Smith 2021) to take advantage of specimens stored in museum collections or natural environments such as middens or pollen deposits, for estimating pre-human biodiversity baselines, range shifts, or genetic diversity influenced by climate change or human intervention. Similarly, SDM developments lack a coherent theoretical framework to estimate model uncertainties. While ensemble modelling has become standard practice in SDMs (Araújo et al. 2007, 2019), there is no equivalent framework for identifying and reporting spatial or temporal uncertainties. Citizen science, or community science, also remains underrepresented, despite its growing prominence through artificial intelligence assisted applications such as Pl@ntNet (Joly et al. 2016). Links between large-scale citizen science, biogeography, and error estimation require further theoretical and applied development.

The dominance of SDMs in this issue reflects their utility but also their limitations. To advance predictive biogeography, the field must move beyond static SDMs and adopt a more mechanistic understanding of ecological processes across scales. Functional biogeography, though promising, is largely underrepresented here. A comprehensive theoretical and empirical framework linking functional ecology to predictive biogeography remains elusive (but see Violle et al. 2014, Díaz et al. 2022, Neyret et al. 2024). Similarly, ecological frameworks developed at small spatial and temporal scales must be scaled to larger extents to address global change scenarios. Such frameworks should incorporate broader ecological processes – such as trophic regulation, productivity, stability, and ecosystem functions – rather than focusing solely on individual species.

Dynamic SDM predictions that leverage real-time weather data and remote sensing are also crucial for advancing the field. Near-term ecological forecasting has been identified as a priority for making timely predictions relevant to management decisions (Dietze et al. 2018). These systems can also play a retroactive role, integrating lessons learned into ecological theories to refine mechanistic understanding and to improve forecasts (Dietze et al. 2018, Lewis et al. 2023). Achieving this requires fully replicable modelling pipelines that can incorporate near-real-time data. This highlights the importance of open science and programming literacy (Mandeville et al. 2021). Open data, models, and pipelines not only ensure reproducibility but also democratize science, allowing analysis pipelines to be easily adapted to new settings (Maldonado et al. 2015). Additionally, a broad system for archiving and synthesizing across predictions (Record et al. 2023) is needed to build forecasting systems based on past experiences.

As Enquist et al. (2024) points out, while technology has given us a large toolkit and the potential to learn about different levels of organization, not all high-resolution datasets or detailed information are equally informative for improving predictions (Meynard et al. 2023). The success of predictive biogeography will depend on reconciling three scientific cultures: one that values detail and specificity, one that emphasizes experimentation and mechanistic explanations, and one that simplifies to discern generalizable patterns. Striking the right balance between these approaches is a challenging yet worthwhile endeavour for advancing predictive science.