Defining and evaluating predictions of joint species distribution models

Data-centric species distribution modeling: Impacts of modeler decisions in a case study of invasive European frog-bit

Premise

Species distribution models (SDMs) are widely utilized to guide conservation decisions. The complexity of available data and SDM methodologies necessitates considerations of how data are chosen and processed for modeling to enhance model accuracy and support biological interpretations and ecological applications.

Methods

We built SDMs for the invasive aquatic plant European frog-bit using aggregated and field data that span multiple scales, data sources, and data types. We tested how model results were affected by five modeler decision points: the exclusion of (1) missing and (2) correlated data and the (3) scale (large-scale aggregated data or systematic field data), (4) source (specimens or observations), and (5) type (presence-background or presence-absence) of occurrence data.

Results

Decisions about the exclusion of missing and correlated data, as well as the scale and type of occurrence data, significantly affected metrics of model performance. The source and type of occurrence data led to differences in the importance of specific explanatory variables as drivers of species distribution and predicted probability of suitable habitat.

Discussion

Our findings relative to European frog-bit illustrate how specific data selection and processing decisions can influence the outcomes and interpretation of SDMs. Data-centric protocols that incorporate data exploration into model building can help ensure models are reproducible and can be accurately interpreted in light of biological questions.

Novel community data in ecology-properties and prospects

New technologies for monitoring biodiversity such as environmental (e)DNA, passive acoustic monitoring, and optical sensors promise to generate automated spatiotemporal community observations at unprecedented scales and resolutions. Here, we introduce 'novel community data' as an umbrella term for these data. We review the emerging field around novel community data, focusing on new ecological questions that could be addressed; the analytical tools available or needed to make best use of these data; and the potential implications of these developments for policy and conservation. We conclude that novel community data offer many opportunities to advance our understanding of fundamental ecological processes, including community assembly, biotic interactions, micro- and macroevolution, and overall ecosystem functioning.

Mapping multiscale breeding bird species distributions across the United States and evaluating their conservation applications

Species distribution models are vital to management decisions that require understanding habitat use patterns, particularly for species of conservation concern. However, the production of distribution maps for individual species is often hampered by data scarcity, and existing species maps are rarely spatially validated due to limited occurrence data. Furthermore, community-level maps based on stacked species distribution models lack important community assemblage information (e.g., competitive exclusion) relevant to conservation. Thus, multispecies, guild, or community models are often used in conservation practice instead. To address these limitations, we aimed to generate fine-scale, spatially continuous, nationwide maps for species represented in the North American Breeding Bird Survey (BBS) between 1992 and 2019. We developed ensemble models for each species at three spatial resolutions-0.5, 2.5, and 5 km-across the conterminous United States. We also compared species richness patterns from stacked single-species models with those of 19 functional guilds developed using the same data to assess the similarity between predictions. We successfully modeled 192 bird species at 5-km resolution, 160 species at 2.5-km resolution, and 80 species at 0.5-km resolution. However, the species we could model represent only 28%-56% of species found in the conterminous US BBSs across resolutions owing to data limitations. We found that stacked maps and guild maps generally had high correlations across resolutions (median = 84%), but spatial agreement varied regionally by resolution and was most pronounced between the East and West at the 5-km resolution. The spatial differences between our stacked maps and guild maps illustrate the importance of spatial validation in conservation planning. Overall, our species maps are useful for single-species conservation and can support fine-scale decision-making across the United States and support community-level conservation when used in tandem with guild maps. However, there remain data scarcity issues for many species of conservation concern when using the BBS for single-species models.

Interspecific behavioural interference and range dynamics: current insights and future directions

Novel biotic interactions in shifting communities play a key role in determining the ability of species' ranges to track suitable habitat. To date, the impact of biotic interactions on range dynamics have predominantly been studied in the context of interactions between different trophic levels or, to a lesser extent, exploitative competition between species of the same trophic level. Yet, both theory and a growing number of empirical studies show that interspecific behavioural interference, such as interspecific territorial and mating interactions, can slow down range expansions, preclude coexistence, or drive local extinction, even in the absence of resource competition. We conducted a systematic review of the current empirical research into the consequences of interspecific behavioural interference on range dynamics. Our findings demonstrate there is abundant evidence that behavioural interference by one species can impact the spatial distribution of another. Furthermore, we identify several gaps where more empirical work is needed to test predictions from theory robustly. Finally, we outline several avenues for future research, providing suggestions for how interspecific behavioural interference could be incorporated into existing scientific frameworks for understanding how biotic interactions influence range expansions, such as species distribution models, to build a stronger understanding of the potential consequences of behavioural interference on the outcome of future range dynamics.

Investigating Ecological Interactions in the Tumor Microenvironment using Joint Species Distribution Models for Point Patterns

The tumor microenvironment (TME) is a complex and dynamic ecosystem that involves interactions between different cell types, such as cancer cells, immune cells, and stromal cells. These interactions can promote or inhibit tumor growth and affect response to therapy. Multitype Gibbs point process (MGPP) models are statistical models used to study the spatial distribution and interaction of different types of objects, such as the distribution of cell types in a tissue sample. Such models are potentially useful for investigating the spatial relationships between different cell types in the tumor microenvironment, but so far studies of the TME using cell-resolution imaging have been largely limited to spatial descriptive statistics. However, MGPP models have many advantages over descriptive statistics, such as uncertainty quantification, incorporation of multiple covariates and the ability to make predictions. In this paper, we describe and apply a previously developed MGPP method, the saturated pairwise interaction Gibbs point process model , to a publicly available multiplexed imaging dataset obtained from colorectal cancer patients. Importantly, we show how these methods can be used as joint species distribution models (JSDMs) to precisely frame and answer many relevant questions related to the ecology of the tumor microenvironment.

Co-occurrences enhance our understanding of aquatic fungal metacommunity assembly and reveal potential host-parasite interactions

Our knowledge of aquatic fungal communities, their assembly, distributions and ecological roles in marine ecosystems is scarce. Hence, we aimed to investigate fungal metacommunities of coastal habitats in a subarctic zone (northern Baltic Sea, Sweden). Using a novel joint species distribution model and network approach, we quantified the importance of biotic associations contributing to the assembly of mycoplankton, further, detected potential biotic interactions between fungi-algae pairs, respectively. Our long-read metabarcoding approach identified 493 fungal taxa, of which a dominant fraction (44.4%) was assigned as early-diverging fungi (i.e. Cryptomycota and Chytridiomycota). Alpha diversity of mycoplankton declined and community compositions changed along inlet-bay-offshore transects. The distributions of most fungi were rather influenced by environmental factors than by spatial drivers, and the influence of biotic associations was pronounced when environmental filtering was weak. We found great number of co-occurrences (120) among the dominant fungal groups, and the 25 associations between fungal and algal OTUs suggested potential host-parasite and/or saprotroph links, supporting a Cryptomycota-based mycoloop pathway. We emphasize that the contribution of biotic associations to mycoplankton assembly are important to consider in future studies as it helps to improve predictions of species distributions in aquatic ecosystems.

What Does the Future Hold for Páramo Plants? A Modelling Approach

Understanding the effects of climate change on tropicalpine biota remains a scientific challenge today. The Andean páramo is the largest and most diverse tropicalpine biogeographical region in the world, and also one of the most threatened as it is prone to accelerated environmental changes. My goal was to predict changes in the distribution ranges of the diverse and highly endemic páramo flora on the mid-term (50 years). First, I predicted distribution changes in páramo plant species under novel climates and considering dispersal constraints. Second, I looked for consensus areas of species losses vs. gains in the páramo, expecting to identify a gradient of increasing relative richness with elevation over time. Last, I evaluated the behavior of plant species regarding their climatic refugia since the Last Glacial Maximum (LGM) to establish if they likely remain or transcend them. Based on VegParamo vegetation data and CHELSA bioclimatic information, I performed species distribution models for a 664 species pool, that were then contrasted between the present, future (2070) and past (LGM). About 8.3% of the entire species pool (55 species) were predicted to be extirpated from the páramo by 2070, including 22 species endemics. On average, páramo plants gained 15.52% of additional distribution by 2070 (18.81% for endemics). Models predicted the most area gains for the northern páramos of Colombia and Venezuela, and the highest losses for the eastern Ecuadorian and Peruvian mountains. Moreover, area gains were more pronounced at high elevations, suggesting a future accelerated colonization process toward the northern Andean summits. Finally, only 21.41% of the species’ 2070 distribution coincided with their LGM (19.75% for endemics), and the largest climatic refugia since the LGM were found in southern Ecuador and Peru. This study is pioneer in predicting future distribution shifts for páramo plant species overall and provides solid bases to support climate change research and adaptation strategies in the tropical Andes.

Predicting species distributions and community composition using satellite remote sensing predictors

Biodiversity is rapidly changing due to changes in the climate and human related activities; thus, the accurate predictions of species composition and diversity are critical to developing conservation actions and management strategies. In this paper, using satellite remote sensing products as covariates, we constructed stacked species distribution models (S-SDMs) under a Bayesian framework to build next-generation biodiversity models. Model performance of these models was assessed using oak assemblages distributed across the continental United States obtained from the National Ecological Observatory Network (NEON). This study represents an attempt to evaluate the integrated predictions of biodiversity models-including assemblage diversity and composition-obtained by stacking next-generation SDMs. We found that applying constraints to assemblage predictions, such as using the probability ranking rule, does not improve biodiversity prediction models. Furthermore, we found that independent of the stacking procedure (bS-SDM versus pS-SDM versus cS-SDM), these kinds of next-generation biodiversity models do not accurately recover the observed species composition at the plot level or ecological-community scales (NEON plots are 400 m2). However, these models do return reasonable predictions at macroecological scales, i.e., moderately to highly correct assignments of species identities at the scale of NEON sites (mean area ~ 27 km2). Our results provide insights for advancing the accuracy of prediction of assemblage diversity and composition at different spatial scales globally. An important task for future studies is to evaluate the reliability of combining S-SDMs with direct detection of species using image spectroscopy to build a new generation of biodiversity models that accurately predict and monitor ecological assemblages through time and space.

Predicting correlated outcomes from molecular data

MOTIVATION

Multivariate (multi-target) regression has the potential to outperform univariate (single-target) regression at predicting correlated outcomes, which frequently occur in biomedical and clinical research. Here we implement multivariate lasso and ridge regression using stacked generalisation.

RESULTS

Our flexible approach leads to predictive and interpretable models in high-dimensional settings, with a single estimate for each input-output effect. In the simulation, we compare the predictive performance of several state-of-the-art methods for multivariate regression. In the application, we use clinical and genomic data to predict multiple motor and non-motor symptoms in Parkinson's disease patients. We conclude that stacked multivariate regression, with our adaptations, is a competitive method for predicting correlated outcomes.

AVAILABILITY AND IMPLEMENTATION

The R package joinet is available on GitHub (https://github.com/rauschenberger/joinet) and cran (https://cran.r-project.org/package=joinet).

SUPPLEMENTARY INFORMATION

Supplementary tables and figures are available at Bioinformatics online.