Post provided by Teresa Goicolea and Alejandra Zarzo

Esta publicación también está disponible en español.

Species Distribution Models (SDMs) are essential tools for scientists and conservationists to predict where species are likely to be found, where they have existed in the past, and where they might appear in the future. As we face urgent issues like climate change and biodiversity loss, producing accurate predictions is more important than ever to identify key areas for conservation measures. However, SDMs often struggle with accuracy, especially due to niche truncation and environmental extrapolation issues (see later).

That’s where the new R package sabinaNSDM, comes in. Designed by our SABINA research team, this package uses a new approach to species distribution modeling called spatially-nested hierarchical species distribution models (N-SDMs). By combining broad-scale global patterns with regional finer characteristics, sabinaNSDM allows for more accurate predictions of species distributions. This makes the new package a powerful resource for conservation planning and ecological research.

The problem with traditional SDMs

Standard SDMs come with a set of limitations. Most models fall into one of two categories: regional or global.

• Regional models are focused on specific areas, like a country or region. While they may offer detailed insights about local conditions, they miss out on the broader environmental picture that shapes a species’ distribution. This leads to what’s called niche truncation, where models fail to consider the full range of conditions that a species experiences across its entire distribution (i.e., the ecological niche). These spatially-restricted models also suffer from a larger proportion of non-analog conditions, leading to problems when projecting to other areas (for example to predict the expansion of invasive species) or periods (to predict the impact of climate change on species distribution).
• Global models, on the other hand, cover a species’ entire range, but often rely on coarse, large-scale data. They usually rely only on bioclimatic data as other thematic environmental drivers are not available at large scales, and imprecise species data. As a result, they lack the fine details needed for precise, localized predictions.

The solution: Nested Species Distribution Models (N-SDMs)

Spatially-nested hierarchical SDMs (N-SDM) address these issues by combining the broad perspective of global models with the fine detail of regional models to get the best of both. Global models provide a big-picture view, capturing a species’ complete ecological niche across its entire range and accounting for factors like climate at a coarse resolution. Regional models then zoom in on finer details, such as land cover or microhabitat conditions and more precise species distribution data, that are usually available for smaller areas, such as at country level. These finer details are critical for making accurate, high-resolution predictions.

Key features of sabinaNSDM package

sabinaNSDM is designed to make this N-SDM approach accessible to researchers and conservationists. Here are some of its key features:

1. Perform N-SDMS: The package combines global and regional models.

2. Different nesting approaches: Users can choose between two methods for combining models—the covariate approach, which uses global model outputs as inputs for regional models, or the multiply approach, which averages global and regional predictions.

3. Ensembling modeling: sabinaNSDM uses ensemble modelling, a technique that combines multiple statistical models to increase the reliability and accuracy of predictions. 

4. Comprehensive workflow: the package is an end-to-end tool that integrates (a) the generation of background data; (b) preparation and spatial thinning of species occurrences (and absences if available) (c) environmental covariate selection; and (d) NSDMs calibration, evaluation and projection.

5. Proven Accuracy: In a case study involving 77 tree and shrub species in the Iberian Peninsula, sabinaNSDM outperformed traditional SDMs, offering more accurate predictions of species distributions.

6. Open-Source and User-Friendly: sabinaNSDM is freely available on GitHub, and we are working to make it available on CRAN. This package was designed to be user-friendly, making it accessible to ecologists and conservationists with various levels of programming experience.

Real-World Impact

The ability to accurately model species distributions has real-world consequences, and sabinaNSDM’s enhanced modeling capabilities can play a crucial role in shaping conservation efforts. For example, the package can predict how climate change might alter species distributions, guide restoration programs to target areas with the greatest potential to support biodiversity, or anticipate the spread of invasive species. One of our key applications has been creating a geoportal showcasing the predicted distribution of 200 woody plant species in Spain under current conditions and four future climate scenarios. This tool offers various practical applications, such as generating lists of the shrubs and trees with the highest suitability for specific locations, helping inform restoration efforts by identifying the species most likely to thrive both now and in the future. sabinaNSDM has already shown its potential in our work, and we’re excited to see how other researchers and conservationists will use it in their projects.

Get Started with sabinaNSDM

If you’re interested in trying out sabinaNSDM, you can download the package and explore its features on our GitHub repository. For a deeper dive into how it works, check out our recent paper published in Methods in Ecology and Evolution. We’ve also included supplementary materials and tutorials to help you get started with both single-species and multi-species modeling. If you’re interested in knowing more about sabinaNSDM or have any questions, feel free to reach out.

Post Edited by Lydia Morley