Biology subjects

Dinnage, R.

Publications and source records attributed to Dinnage, R..

11 recordsLinked to original sources

bioRxiv2025

Petal to the metal: The slow road to automating large-scale phenology labeling for herbarium specimens

Herbarium specimens represent critical historical records of plant phenology, yet automating annotation of reproductive structures remains challenging given the diversity of floral morphologies, specimen age and quality, and image quality. Here, we present a machine learning pipeline that uses an ensemble modeling approach to detect flowers on herbarium specimens and deliver these data to the phenology research community. After testing multiple strategies for generating training data, we found in-house expert-curated annotations were essential for producing reliable results. Expert validation found relatively strong accuracy for detecting present floral structures, but still had moderately high false negative rates. Applying the ensemble to our filtered final image dataset of 22 million records resulted in 11.1 million records labeled with flowers present. However, only 2.9 million of these contained complete metadata necessary for downstream phenology research, highlighting the need for full label digitization efforts. Still, this dataset represents a large compilation of historical herbarium-derived phenology records available as a resource for the phenology community. We end by demonstrating how integrating these machine-labeled records into Phenobase, a publicly-available phenology database, expands taxonomic and temporal coverage for large-scale phenological analyses, and discuss remaining challenges and next steps.

ecology↗

bioRxiv2025

Turning a new leaf: PhenoVision provides leaf phenology data at the global scale

Plant phenology dictates many aspects of community function and ecosystem dynamics. Yet, global phenology data are still limited, especially in areas lacking monitoring programs. Here we present a new data resource, PhenoVision-Leaf, which extends a computer-vision pipeline utilizing iNaturalist digital image vouchers to produce global-scale leaf phenophase data for deciduous, woody genera. We first discuss our implementation of a new human annotation framework for leaf phenology on iNaturalist, aligning with phenophase definitions used by the larger phenology community. We then showcase the use of 165,988 crowdsourced annotated records to train a Vision Transformer model with a two-stage regime to maximize accuracy across single- and multi-image records. This approach extends Phenovision from scoring individual images to aggregating at the iNaturalist record level, better aligning with human annotation processes. Post-hoc validation showed high performance for detecting present green and colored leaves (>98% accuracy), and reasonable accuracy for breaking leaf buds (>87% accuracy). Applying PhenoVision-Leaf to over 26 million iNaturalist records yielded 5.6 million record-level phenology observations across 6,500 species and 57 families, filling geographic and taxonomic gaps. These data, now accessible through the Phenobase portal, establish a foundation for near real-time monitoring of leaf phenology, supporting global-scale synthesis analyses.

ecology↗

bioRxiv2024

NicheFlow: Towards a foundation model for Species Distribution Modelling

1. Species distribution models (SDMs) are crucial tools for understanding and predicting biodiversity patterns, yet they often struggle with limited data, biased sampling, and complex species-environment relationships. Here I present NicheFlow, a novel foundation model for SDMs that leverages generative AI to address these challenges and advance our ability to model and predict species distributions across taxa and environments. 2. NicheFlow employs a two-stage generative approach, combining species embeddings with two chained generative models, one to generate a distribution in environmental space, and a second to generate a distribution in geographic space. This architecture allows for the sharing of information across species and captures complex, non-linear relationships in environmental space. I trained NicheFlow on a comprehensive dataset of reptile distributions and evaluated its performance using both standard SDM metrics and zero-shot prediction tasks. 3. NicheFlow demonstrates good predictive performance, particularly for rare and data-deficient species. The model successfully generated plausible distributions for species not seen during training, showcasing its potential for zero-shot prediction. The learned species embeddings captured meaningful ecological information, revealing patterns in niche structure across taxa, latitude and range sizes. 4. As a proof-of-principle foundation model, NicheFlow represents a significant advance in species distribution modeling, offering a powerful tool for addressing pressing questions in ecology, evolution, and conservation biology. Its ability to model joint species distributions and generate hypothetical niches opens new avenues for exploring ecological and evolutionary questions, including ancestral niche reconstruction and community assembly processes. This approach has the potential to transform our understanding of biodiversity patterns and improve our capacity to predict and manage species distributions in the face of global change.

ecology↗

bioRxiv2024

PhenoVision: A framework for automating and delivering research-ready plant phenology data from field images

Plant phenology plays a fundamental role in shaping ecosystems, and global change-induced shifts in phenology have cascading impacts on species interactions and ecosystem structure and function. Detailed, high-quality observations of when plants undergo seasonal transitions such as leaf-out, flowering, and fruiting are critical for tracking causes and consequences of phenology shifts, but these data are often sparse and biased globally. These data gaps limit broader generalizations and forecasting improvements in the face of continuing disturbance. One solution to closing such gaps is to document phenology on field images taken by public participants. iNaturalist, in particular, provides global scale research-grade data and is expanding rapidly. Here we utilize over 53 million field images of plants and millions of human annotations from iNaturalist - data spanning all angiosperms and drawn from across the globe - to train a computer vision model (PhenoVision) to detect the presence of fruits and flowers. PhenoVision utilizes a vision transformer architecture pretrained with a masked autoencoder to improve classification success, and it achieves high accuracy for flower (98.5%) and fruit presence (95%). Key to producing research-ready phenology data is post-calibration tuning and validation focused on reducing noise inherent in field photographs, and maximizing the true positive rate. We also develop a standardized set of quality metrics and metadata so that results can be used effectively by the community. Finally, we showcase how this effort vastly increases phenology data coverage, including regions of the globe where data have been limited before. Our end products are tuned models, new data resources, and an application streamlining discovery and use of those data for the broader research and management community. We close by discussing next steps, including automating phenology annotations, adding new phenology targets, e.g., leaf phenology, and further integration with other resources to form a global central database integrating all in-situ plant phenology resources.

ecology↗

bioRxiv2023

How many variables does Wordclim have, really? Generative A.I. unravels the intrinsic dimension of bioclimatic variables.

The 19 standard bioclimatic variables available from the WorldClim dataset are some of the most used data in ecology and organismal biology. It is well known that many of the variables are correlated with each other, suggesting there are fewer than 19 independent dimensions of information in them. But how much information is there? Here I explore the 19 WorldClim bioclimatic variables from the perspective of the manifold hypothesis: that many high dimensional datasets are actually confined to a lower dimensional manifold embedded in an ambient space. Using a state-of-the-art generative probabilistic model (variational autoencoder) to model the data on a non-linear manifold reveals that only 5 uncorrelated dimensions are adequate to capture the full range of variation in the bioclimatic variables, with a clear data-driven separation between informative and redundant dimensions that eliminates arbitrary thresholds. I show that these 5 variables have meaningful structure and are sufficient to produce species distribution models (SDMs) nearly as good and in some ways better than SDMs using the original 19 bioclimatic variables. I have made the 5 synthetic variables available as a raster dataset at 2.5 minute resolution in an R package that also includes functions to convert back and forth between the 5 variables and the original 19.

ecology↗

bioRxiv2022

Mangroves deviate from other angiosperms in their genome size, leaf cell size, and cell packing density relationships

Background and AimsWhile genome size limits the minimum sizes and maximum numbers of cells that can be packed into a given leaf volume, mature cell sizes can be substantially larger than their meristematic precursors and vary in response to abiotic conditions. Mangroves are iconic examples of how abiotic conditions can influence the evolution of plant phenotypes. MethodsHere, we examined the coordination between genome size, leaf cell sizes, and cell packing densities, and leaf size in 13 mangrove species across four sites. Four of these species occurred at more than one site, allowing us to test the effect of climate on leaf anatomy. ResultsWe found that genome sizes of mangroves were very small compared to other angiosperms, and, like other angiosperms, mangrove cells were always larger than the minimum size defined by genome size. Increasing mean annual temperature of a growth site led to higher packing densities of veins (Dv) and stomata (Ds) and smaller epidermal cells but had no effect on stomatal size. Contrary to other angiosperms, mangroves exhibited (1) a negative relationship between guard cell size and genome size; (2) epidermal cells that were smaller than stomata, and (3) coordination between Dv and Ds that was not mediated by epidermal cell size. Furthermore, mangrove epidermal cell sizes and packing densities covaried with leaf size. ConclusionsWhile mangroves exhibited coordination between veins and stomata and attained a maximum theoretical stomatal conductance similar to other angiosperms, the tissue-level tradeoffs underlying these similar relationships across species and environments was markedly different, perhaps indicative of the unique structural and physiological adaptations of mangroves to their stressful environments.

ecology↗

bioRxiv2021

slimr: An R package for integrating data and tailor-made population genomic simulations over space and time

Software for realistically simulating complex population genomic processes is revolutionizing our understanding of evolutionary processes, and providing novel opportunities for integrating empirical data with simulations. However, the integration between simulation software and software designed for working with empirical data is currently not well developed. Here we present slimr, an R package designed to create a seamless link between standalone software SLiM 3.0, one of the most powerful population genomic simulation frameworks, and the R development environment, with its powerful data manipulation and analysis tools. We show how slimr facilitates smooth integration between genetic data, ecological data and simulation in a single environment. The package enables pipelines that begin with data reading, cleaning, and manipulation, proceed to constructing empirically-based parameters and initial conditions for simulations, then to running numerical simulations, and finally to retrieving simulation results in a format suitable for comparisons with empirical data - aided by advanced analysis and visualization tools provided by R. We demonstrate the use of slimr with an example from our own work on the landscape population genomics of desert mammals, highlighting the advantage of having a single integrated tool for both data analysis and simulation. slimr makes the powerful simulation ability of SliM 3.0 directly accessible to R users, allowing integrated simulation projects that incorporate empirical data without the need to switch between software environments. This should provide more opportunities for evolutionary biologists and ecologists to use realistic simulations to better understand the interplay between ecological and evolutionary processes.

evolutionary biology↗

bioRxiv2020

phyr: An R package for phylogenetic species-distribution modelling in ecological communities

O_LIModel-based approaches are increasingly popular in ecological studies. A good example of this trend is the use of joint species distribution models to ask questions about ecological communities. However, most current applications of model-based methods do not include phylogenies despite the well-known importance of phylogenetic relationships in shaping species distributions and community composition. In part, this is due to lack of accessible tools allowing ecologists to fit phylogenetic species distribution models easily. C_LIO_LITo fill this gap, the R package phyr (pronounced fire) implements a suite of metrics, comparative methods and mixed models that use phylogenies to understand and predict community composition and other ecological and evolutionary phenomena. The phyr workhorse functions are implemented in C++ making all calculations and model estimations fast. C_LIO_LIphyr can fit a variety of models such as phylogenetic joint-species distribution models, spatiotemporal-phylogenetic autocorrelation models, and phylogenetic trait-based bipartite network models. phyr also estimates phylogenetically independent trait correlations with measurement error to test for adaptive syndromes and performs fast calculations of common alpha and beta phylogenetic diversity metrics. All phyr methods are united under Brownian motion or Ornstein-Uhlenbeck models of evolution and phylogenetic terms are modelled as phylogenetic covariance matrices. C_LIO_LIThe functions and model formula syntax we propose in phyr serves as a simple and unified framework that ignites the use of phylogenies to address a variety of ecological questions. C_LI

ecology↗

bioRxiv2018

Spatiophylogenetic modelling of extinction risk reveals evolutionary distinctiveness and brief flowering period as risk factors in a diverse hotspot plant genus

Comparative models used to predict species threat status often combine variables measured at the species level with spatial variables, causing multiple statistical challenges, including phylogenetic and spatial non-independence. We present a novel bayesian approach for modelling threat status that simultaneously deals with both forms of non-independence and estimates their relative contribution, and we apply the approach to modelling threat status in the Australian plant genus Hakea. We find that after phylogenetic and spatial effects are accounted for, species with greater evolutionary distinctiveness and a shorter annual flowering period are more likely to be threatened. The model allows us to combine information on evolutionary history, species biology, and spatial data, to calculate latent extinction risk (potential for non-threatened species to become threatened), and estimate the most important drivers of risk for individual species. This could be of value for proactive conservation decision-making that targets species of concern before they become threatened.

ecology↗

bioRxiv2018

Habitat loss is information loss: Species distribution models are compromised in anthropogenic landscapes

Species distribution models (SDMs) are valuable tools to estimate species distributions, but are vulnerable to biases in the probability of a species being observed. One such bias is habitat loss, which has affected a substantial and increasing proportion of the Earth. In regions of severe habitat loss, data on a species occurrence may represent a small, non-random subset of sites it once occupied. This could cause distorted reconstructions of species distributions, and misleading inferences of evolutionary and ecological processes. We present a statistical approach for quantifying the influence on SDMs of habitat loss, and generating distribution predictions that are robust to these biases. We explored some of the effects of accounting for habitat loss on inferences from common downstream biogeographic and ecological analysis methods.\n\nWe used herbarium record data to model the distribution of 325 plant species in the genera Banksia and Hakea across Australia, using point process models. We accounted for biases in the models by including a proxy variable representing habitat loss, and compared the fit of models without this variable to those with it. We explored the influence of habitat loss by mapping biodiversity patterns predicted with and without accounting for it.\n\nGenerally, accounting for habitat loss in SDMs led to increases in the mean area of modelled species distributions of ~10% for Banksia and ~12% for Hakea across Australia (in some cases, up to several 100,000 km2 increases in predicted range), with somewhat greater average increases (11% and 15%) for species in the southwest Australian biodiversity hotspot. Accounting for habitat loss leads to an increase in predicted species richness (Alpha and Gamma diversity), but a decrease in compositional turnover (Beta diversity), across most of Australia. Accounting for habitat loss in SDMs had minimal influence on a downstream macroevolutionary analysis (Age-Range Correlation) that utilizes species distributions, seemingly because exposure to habitat loss did not show a phylogenetic pattern in this taxonomic group.\n\nThe influence of habitat loss on species distributions estimated with SDM is likely to be context-dependent and difficult to generalize, but will tend to cause underestimates of range sizes. This may have consequences for mapping spatial patterns of diversity and for some downstream analyses of biogeographic, evolutionary, or ecological processes, based on species distributions, as well as conservation measures that rely on accurate species mapping.

ecology↗

bioRxiv2018

Larger legume plants host a greater diversity of symbiotic nitrogen-fixing bacteria

A major goal in microbial ecology is to understand the factors that structure bacterial communities across space and time. For microbes that have symbiotic relationships with plants, an important factor that may influence their communities is host size or age, yet this has received little attention.\n\nUsing tree diameter size as a proxy for age, we quantified the diversity of rhizobia that associate with an endemic legume, Acacia acuminata, of variable size across a climate gradient in southwest Australia. We examined the 16S rRNA diversity (V1-V3 hypervariable region) of rhizobia at the taxonomic level and at higher sequence level diversity within taxonomic groups.\n\nWe identified 3 major taxonomic clades that associated with Acacia acuminata: Bradyrhizobiaceae, Rhizobiaceae, and Burkholderiaceae. Within these groups, we found extensive genetic variability, especially within Bradyrhizobiaceae. Using binomial multivariate statistical models that controlled for other factors that affect plant size and rhizobia community structure (climate and local soil characteristics), we determined that soil sampled at the base of larger Acacia trees was much more likely to contain a greater number of taxonomic clades and cryptic genetic variants within the Rhizobiaceae clade.\n\nDespite strong influences of climate and highly heterogeneous soil conditions on rhizobial diversity, our results show that host tree size is a prominent factor in structuring nitrogen-fixing symbionts diversity across a large landscape. The identification of a positive relationship between plant host size and microbial diversity raise interesting questions about the role of plant host size in driving ecological processes that govern microbial community assembly. Specifically, our results suggest that hosts may modify the habitat of their surrounding soil to enhance growth (niche construction hypothesis) or that symbiotic microbes have large differences in dispersal capability. Our results also suggest that host plants may be analogous to islands, where larger legume hosts may accumulate diversity over time, through migration opportunities or in situ diversification. From a practical perspective, including plant size as an additional variable may assist sampling and analyses designs of future soil microbial studies.

ecology↗