One of the base units of analysis for biogeography and conservation science is the species range map. Once we know where a species is, we can ask questions like "Why is it there?", "How did it get there?", or "What can we do to make this place better for it?" Especially these days, I am very interested in mapping marine fish distributions, which, it turns out, is not as simple as mapping terrestrial species.

The problem is that ecological niche modeling and species distribution modeling methods and theory were largely developed by people working in terrestrial systems. Specifically, correlative modeling approaches were developed based on environmental conditions extracted at two-dimensional coordinates across a horizontal landscape. Sure, birds and insects fly (and some of them at very high altitudes), but their interactions  with the upper troposphere are generally brief and unlikely to directly effect the distributions of these species.

In comparison, ocean fishes may live their whole lives at the ocean's surface or the seafloor, but they may also swim freely somewhere between the two. The largest migration in the world occurs DAILY, when a million tons of mesopelagic fishes travel between the cold, dark safety of the deep sea and warm, food-rich surface waters. If you modeled the ecological niche of one of these pelagic species based on surface conditions where it was caught at night, you might drastically underestimate its temperature tolerances. If distribution maps based on these inaccurate models are used in downstream analyses, they may bias results and cause all kinds of other problems.

This is where voluModel comes in. voluModel is an R package I developed in collaboration with my advisor and co-author, Carsten Rahbek, during my Marie Curie Fellowship postdoc. In our new paper, out in Methods in Ecology and Evolution this week, we introduce a workflow to extract environmental conditions based on the three-dimensional coordinates where fishes are observed, as well as their background environments. These data can then be used in any typical ecological niche modeling algorithm that accepts data in a  points-with-data format; the model can then be projected back into three-dimensional space using a simple loop. While three-dimensional modeling has been done before, this is the first set of tools to efficiently move through a 3-D workflow instead of using case-specific custom code. Accompanying the package are vignettes that provide 1) an overview of how voluModel works, 2) raster processing tools, 3) 3D environmental data sampling, 4) visualization tools, and 5) a basic overview of how to generate a generalized linear model with 3D data.

Development of voluModel is ongoing--there have been three major updates since we first submitted the paper. Largely, this is due to the rapidly-developing landscape of faster, more efficient geoprocessing R tools using terra and other successors to the raster and rgeos packages, which are being phased out. However, we have also implemented several suggestions from manuscript reviewers and early package adopters. If you have suggestions for improvements or missing features, or if you have found bugs, you may report them here, or even better, send me a (detailed) pull request!

My aspiration is that voluModel helps niche modelers (including me) to efficiently generate more accurate estimates of pelagic species distributions for downstream biogeographic analyses and conservation assessments. This is especially useful for data-poor species that may not be subject to the same exhaustive study as target fisheries species, but which are nonetheless important pieces in the puzzle both for biogeographers and conservation scientists.

Read our new paper in Methods in Ecology and Evolution HERE.
Access the voluModel website HERE.
View voluModel on CRAN HERE.

I’m thrilled to have been part of a paper out in Ecography today: “wallace 2: a shiny app for modeling species niches and distributions redesigned to facilitate expansion via module contributions”. It updates the original wallace R package, which is a really useful tool for teaching niche modeling in the R ecosystem without requiring students to be exceptionally proficient at coding first. All a person needs do is execute two lines of code, and a graphical user interface pops up that walks them through the steps of a niche modeling analysis, while documenting the decisions made along the way in R code file that can re-run to repeat the analyses.

There is a lot of potential for this approach to be used to teach students the workflows and methods involved in fairly complex niche modeling analyses. I have incorporated it into teaching materials for Masters' and PhD level courses, as well as non-academic workshops, and the new features really expand the scope of applications one can cover. wallace 2.0 also makes it easier to add custom modules to the wallace workflow (essentially analysis options, like specific statistics or data sources), and adds several such modules (including my occCite occurrence citation package).

Here's the package website, which includes links to tutorials in multiple languages, including English, Spanish, and Japanese: wallaceecomod.github.io

Here’s the citation: 
Kass, J.M., Pinilla-Buitrago, G.E., Paz, A., Johnson, B.A., Grisales-Betancur, V., Meenan, S.I., Attali, D., Broennimann, O., Galante, P.J., Maitner, B.S., Owens, H.L., Varela, S., Aiello-Lammens, M.E., Merow, C., Blair, M.E. and Anderson, R.P. (2023), wallace 2: a shiny app for modeling species niches and distributions redesigned to facilitate expansion via module contributions. Ecography e06547. https://doi.org/10.1111/ecog.06547

By Hannah L. Owens and Jamie M. Kass, on behalf of all co-authors*
 
There are billions of species occurrence records served by aggregator databases. The Global Biodiversity Information Facility (GBIF) serves over 1.8 billion occurrence records for species from across the tree of life (GBIF Secretariat 2021), and the Botanical Information and Ecology Network (BIEN) serves over 200 million plant observations (Botanical Information and Ecology Network 2021). The primary datasets these aggregators serve are the result of millions of hours of work by museums and community science initiatives (among others) and are constantly updated as taxonomy changes and data are accrued. Citing the primary datasets that supply data to GBIF and BIEN, together with accession dates, facilitates reproducibility and scientific transparency. These citations also support primary data providers by acknowledging their role as an essential link in the research chain.
 
However, when researchers download occurrence datasets from multiple primary providers via aggregator databases (such as those used in broad-scale biogeographic and macroecological studies), managing and effectively communicating the metadata can be incredibly time-consuming. This is where our new R package, occCite, comes in. occCite is designed to facilitate searches of dataset aggregation services (currently, GBIF and BIEN) that store and manage metadata on primary data providers, database accession dates, DOIs, and taxonomic sources in a unified framework within the R environment. Search results are organized as single objects that can be passed to functions to generate visual and statistical summaries and generate formatted citations.
 
occCite’s Two Main Steps

1. Search. occQuery() provides several ways of generating and optimizing  occurrence searches while storing detailed metadata. occQuery() returns an occCiteData object with information on the type of query made, date of the query, taxonomy used, species names used, database aggregators searched, and a named list of search results corresponding to each species.
2. Cite. The occCiteData object generated by occQuery() can be passed to occCitation() to automatically generate citations with accession dates and DOIs. occCitation() returns a named list with entries corresponding to the taxonomic names used to build a query. The print() method can be used to turn these results into a formatted and alphabetized set of references, either as a single block of text for all species, or as blocks of text for each species individually.

 
Additional Features
Taxonomic Rectification. By default, occQuery() checks species’ names against the GBIF backbone taxonomy. The user may instead elect to use studyTaxonList() to prepare a data object with the species’ names to be searched that has been checked against a taxonomy of their choice from the Global Names Index (http://gni.globalnames.org/).
 
Text Summaries. When the print() method is used on an occCiteData object, tables summarizing taxonomic cleaning results, search results with counts of occurrences for each species from each dataset aggregator, and the GBIF DOIs associated with each species’ search are returned.
 
Summary Plots. occCite provides three types of plots for results from occQuery() when the plot() method is used on an occCiteData object: a histogram showing occurrences by year, a waffle plot showing the proportion of results supplied by GBIF versus BIEN, and a waffle plot showing the proportion of occurrences supplied by each primary data provider. These plots can be generated either for all search results or by species.

Maps. Interactive leaflet maps can be generated from occCiteData objects via the occCiteMap() function, for all search results or by species. Users can specify occurrence point marker colors and symbologies. Hovering over a point in the interactive map provides information on the species name, coordinates, date, dataset, and dataset aggregator that supplied it.

 
The Future of occCite
occCite has been integrated as a module in the development version of Wallace, a modular, R-based graphical user interface for modeling species’ ecological niches and geographic distributions (Kass et al. 2018). When Wallace users opt to include data source citations in occurrence data searches, occCite will be invoked to run the search and generate citations.
 
In the future, we plan to expand the number of database aggregators that occCite queries, and add various fit-for-purpose filtering actions (e.g., duplicate removal, temporal downsampling, geographic and environmental outlier removal). We also plan to add comparative summary plots for raw vs. filtered data or comparing different occCiteData objects. We hope you’ll keep up-to-date via our GitHub website (hannahlowens.github.io/occCite/) for these and other exciting developments!
 
 
Further Reading
Manuscript: https://onlinelibrary.wiley.com/doi/10.1111/ecog.05618
 
CRAN release: https://CRAN.R-project.org/package=occCite
 
YouTube Tutorial: https://www.youtube.com/watch?v=7qSCULN_VjY&t=17s
 

References
Botanical Information and Ecology Network. 2021. BIEN, the Botanical Information and Ecology Network. bien.nceas.ucsb.edu, accessed 6 August 2021.
 
GBIF Secretariat. 2021. GBIF: Global Biodiversity Information Facility. gbif.org, accessed 6 August 2021.
 
Kass, JM, Vilela, B, Aiello‐Lammens, ME, Muscarella, R, Merow, C. and Anderson, RP. 2018. Wallace: A flexible platform for reproducible modeling of species niches and distributions built for community expansion. Methods in Ecology and Evolution, 9: 1151-1156. DOI: 10.1111/2041-210X.12945

*Originally written for Ecography blog

It's official! Starting today, I am a Marie Skłodowska-Curie Fellow, with a project titled "MARDIGRAS: Elucidating MARine DIversity GRAdients with Empirical and Theoretical ModelS". I'm thrilled for the opportunity to take my experiences of the last few years working on biodiversity patterns in butterflies and birds, and puzzle through how to infer biodiversity patterns and their underlying mechanisms in marine fishes.

The aim of the project is to take a deep dive (sorry not sorry for the pun) into understanding worldwide biodiversity patterns for three groups of marine fishes (my beloved Gadiformes, aka codfishes; Scombriformes, aka mackerels and tunas; and Beloniformes, aka flyingfishes), develop a mechanistic model of how biodiversity patterns arise in marine systems, and then contrast diversity patterns among the three groups of fishes and with the mechanistic model.

While there is extensive macroecological literature on diversity gradients in terrestrial systems, especially regarding latitude, marine systems seem to be following a different set of rules. First, geographic patterns of diversity appear to be neither clear nor ubiquitous. Recently, Chaudhary et al. (2016) found a bimodal diversity curve with respect to latitude (that is, more species were found at middle latitudes than at the equator), whereas Rabosky et al (2018) found a unimodal curve with diversity concentrated at the equator (which is the expected pattern for terrestrial groups). Admittedly, these studies had different organismal foci and employed different methods, but this disparity is striking.

Second, the mechanisms underlying diversity patterns in marine systems appear to be quite different. Generally, in terrestrial systems it is thought that speciation is highest in the tropics, as this is where the most energy is concentrated (among other explanations). However, the aforementioned study by Rabosky and colleagues found that speciation (in marine fishes) was highest at high latitudes! Some of this may be attributable to the unique properties of ocean ecosystems compared to terrestrial ones. As such, one of my project goals is to adapt a mechanistic model that was developed for terrestrial tropical biodiversity (Rangel et al. 2018) and adapt it for the marine context, using diversity patterns in codfishes, mackerels, and flyingfishes to evaluate how realistic the model is.

Stay tuned as I make progress in this exciting area, either here or by following #mardigrasProj on Twitter!