This content will become publicly available on March 1, 2027

Title: CISO: Species distribution modelling Conditioned on Incomplete Species Observations
Abstract Species distribution models (SDMs) are widely used to predict species' geographic distributions, serving as critical tools for ecological research and conservation planning. Typically, SDMs relate species occurrences to environmental variables representing abiotic factors, such as temperature, precipitation, and soil properties.However, species distributions are also strongly influenced by biotic interactions with other species, which are often overlooked in traditional models. While some methods, such as joint species distribution models (JSDMs), partially address this limitation by incorporating biotic interactions, they often assume symmetrical pairwise relationships between species and require consistent co‐occurrence data. In practice, species observations are often sparse, and the availability of information about the presence or absence of other species varies significantly across locations.To address these challenges, we propose CISO, a deep learning‐based method for species distribution modelling Conditioned on Incomplete Species Observations. CISO enables predictions to be conditioned on a flexible number of species observations alongside environmental variables, accommodating the variability and incompleteness of available biotic data. We demonstrate our approach using three datasets representing different species groups: sPlotOpen for plants, SatBird for birds, and a new dataset, SatButterfly, for butterflies.Our results show that including partial biotic information improves predictive performance on spatially separate test sets. When conditioned on a subset of species within the same dataset, CISO outperforms alternative methods in predicting the distribution of the remaining species for plants and birds. Furthermore, we show that combining and conditioning on observations from multiple datasets can improve the prediction of species occurrences in scenarios with sufficient co‐occurrences between datasets to train CISO effectively. Our results show that CISO is a promising ecological tool, capable of incorporating incomplete biotic information and identifying potential interactions between species from disparate taxa.  more » « less
Award ID(s):
2330423
PAR ID:
10714323
Author(s) / Creator(s):
 ;  ;  ;  ;  ;  ;  
Publisher / Repository:
British Ecological Society
Date Published:
Journal Name:
Methods in Ecology and Evolution
Volume:
17
Issue:
3
ISSN:
2041-210X
Page Range / eLocation ID:
947 to 962
Subject(s) / Keyword(s):
Sponsoring Org:
National Science Foundation
More Like this
  1. Environmental conditions are dynamic, and plants respond to those dynamics on multiple time scales. Disequilibrium occurs when a response occurs more slowly than the driving environmental changes. We review evidence regarding disequilibrium in plant distributions, including their responses to paleoclimate changes, recent climate change and new species introductions. There is strong evidence that plant species distributions are often in some disequilibrium with their environmental conditions.This disequilibrium poses a challenge when projecting future species distributions using species distribution models (SDMs). Classically, SDMs assume that the set of species occurrences is an unbiased sample of the suitable environmental conditions. However, a species in disequilibrium with the environment may have higher‐than‐expected occurrence probabilities (e.g. due to extinction debts) or lower‐than‐expected occurrence probabilities (e.g. due to dispersal limitation) in different areas. If unaccounted for, this will lead to biased estimates of the environmental suitability.We review methods for avoiding such biases in SDMs, ranging from simple thinning of the occurrence dataset to complex dynamic and process‐based models. Such models require large data inputs, natural history knowledge and technical expertise, so implementing them can be challenging. Despite this, we advocate for their increased use, since process‐based models provide the best potential to account for biases in model training data and to then represent the dynamics of species occupancy as ranges shift.Synthesis. Occurrence records for a species are often in disequilibrium with climate. SDMs trained on such data will produce biased estimates of a species' niche unless this disequilibrium is addressed in the modelling. A range of tools, spanning a wide gradient of complexity and realism, can resolve this bias. 
    more » « less
  2. Abstract Correlative species distribution models (SDMs) are essential tools in conservation biology, global change assessment and reserve prioritization, linking species occurrences with environmental conditions. These models often rely on coarse‐scale spatial and temporal predictors, overlooking fine‐scale environmental conditions experienced by organisms. Additionally, SDMs commonly use short‐term occurrence data to make long‐term predictions, which can reduce reliability.We hypothesized that long‐term and finer temporal resolution data would provide more accurate predictions by capturing population variability under microclimatic conditions. Using data from 37 bird species in the H. J. Andrews Experimental Forest (Oregon, USA), we built SDMs with a 10‐year (2010–2019) dataset of breeding season observations at 184 sites. We evaluated four modelling frameworks that differed in temporal extents (short‐term [1 year] vs. long‐term [10 years]) and resolution (fine vs. coarse) of environmental data. Predictors included hourly microclimate temperatures beneath the forest canopy and LiDAR‐derived vegetation variables. We evaluated interannual transferability and compared model performance based on temporal extent, resolution and species traits.Temporally dynamic (long‐term) models with higher resolution microclimate data outperformed static and short‐term models (AUC and TSS difference ~ 0.06, difference in unreliability index of ~0.04) and were more accurate and spatially consistent, particularly for migratory species. Increased temporal resolution improved performance for small‐bodied species, whereas long‐lived, larger species performed similarly in short‐ and long‐term models.Synthesis and applications. To our knowledge, this is the first empirical study to demonstrate the benefits of long‐term dynamic SDMs with spatially matched predictor variables. If predicting the future of biodiversity under land‐use and climate change is the goal, practitioners should consider additional investment in multi‐year biodiversity monitoring rather than single‐year ‘snapshots’ of species distributions. 
    more » « less
  3. Summary Anthropogenetic climate change has caused range shifts among many species. Species distribution models (SDMs) are used to predict how species ranges may change in the future. However, most SDMs rarely consider how climate‐sensitive traits, such as phenology, which affect individuals' demography and fitness, may influence species' ranges.Using > 120 000 herbarium specimens representing 360 plant species distributed across the eastern United States, we developed a novel ‘phenology‐informed’ SDM that integrates phenological responses to changing climates. We compared the ranges of each species forecast by the phenology‐informed SDM with those from conventional SDMs. We further validated the modeling approach using hindcasting.When examining the range changes of all species, our phenology‐informed SDMs forecast less species loss and turnover under climate change than conventional SDMs. These results suggest that dynamic phenological responses of species may help them adjust their ecological niches and persist in their habitats as the climate changes.Plant phenology can modulate species' responses to climate change, mitigating its negative effects on species persistence. Further application of our framework will contribute to a generalized understanding of how traits affect species distributions along environmental gradients and facilitate the use of trait‐based SDMs across spatial and taxonomic scales. 
    more » « less
  4. Abstract Anthropogenic pressures on biodiversity necessitate efficient and highly scalable methods to predict global species distributions. Current species distribution models (SDMs) face limitations with large-scale datasets, complex interspecies interactions, and data quality. Here, we introduce EcoVAE, a framework of autoencoder-based generative models trained separately on nearly 124 million georeferenced occurrences from taxa including plants, butterflies and mammals, to predict their global distributions at both genus and species levels. EcoVAE achieves high precision and speed, captures underlying distribution patterns through unsupervised learning, and reveals interspecies interactions viain silicoperturbation analyses. Additionally, it evaluates global sampling efforts and interpolates distributions without relying on environmental variables, offering new applications for biodiversity exploration and monitoring. 
    more » « less
  5. Abstract Intraspecific trait variation (ITV) is an increasingly important aspect of biodiversity and can provide a more complete perspective on how abiotic and biotic processes affect individuals, species' niches and ultimately community‐level structure than traditional uses of trait means. Body size serves as a proxy for a suite of traits that govern species' niches. Distributions of co‐occurring species body sizes can inform niche overlap, relate to species richness and uncover mechanistic drivers of diversity.We leveraged individual‐level body size (length) in freshwater fishes and environmental data from the National Ecological Observatory Network (NEON) for 17 lakes and streams in the contiguous United States to explore how abiotic and biotic factors influence fish species richness and trait distributions of body size. We calculated key abiotic (climate, productivity, land use) and biotic (phylogenetic diversity, trait diversity, community‐level overlap of trait probability densities) variables for each site to test hypotheses about drivers of ITV in body size and fish diversity.Abiotic variables were consistently important in explaining variation in fish body size and species richness across sites. In particular, productivity (as chlorophyll) was a key variable in explaining variation in body size trait richness, evenness and divergence, as well as species richness.This study yields new insights into continental‐scale patterns of freshwater fishes, possible only by leveraging the paired high frequency, in situ abiotic data and individual‐level traits collected by NEON. 
    more » « less