LM Artificial landscapes

In this unit, you will learn how to generate environmental data using Neutral Landscape Models (NLMs), which we will later use in the next unit to create virtual species.

In this course, we aim to evaluate the performance of species distribution models when certain modeling conditions are altered. A key requirement for species distribution modeling is the availability of species occurrence data. For modeling, real-world occurrence data often comes from platforms like the Global Biodiversity Information Facility (GBIF), while environmental variables are commonly sourced from databases such as WorldClim (Fick & Hijmans 2017), which provides bioclimatic data.

To compare different species distribution models, there is a benchmark dataset available which consists of over 200 anonymized species from six regions around the world. However, the test data provided in this dataset are presence-absence data. Therefore, this dataset, as well as GBIF and WorldClim data are helpful but limited, as they offer no direct way to compare a predicted distributions against actual, real-world distributions. To overcome this limitation and better evaluate model quality, we will use artificially generated data in this course.

By using simulated data, we can not only calculate performance metrics based on presence-absence information but also directly compare the modeled species distribution with its “true” known distribution. This is why we will work with virtual species in this unit.

Virtual species and artificial landscapes

Virtual species are generated by calculating a probability of occurrence based on environmental variables. A commonly used environmental dataset for this is WorldClim (Fick & Hijmans 2017), which provides global climate data (see e.g. Leroy et al., 2016). However, in real-world scenarios, these environmental variables often provide only an incomplete picture of species distributions (Bald et al., 2024). Species distributions can be influenced not only by natural conditions, such as climate, but also by human activities, making data on e.g. land use equally important.

To address these complexities, we will use artificially generated environmental variables in this course. Specifically, we will employ neutral landscape models (NLMs), which are algorithms designed to simulate landscapes (Sciaini et al., 2018). Originally developed as null models for studying landscape-scale ecological processes, NLMs now have numerous applications, including the simulation of landscapes for virtual species (Sciaini et al., 2018; Grimmett et al., 2021).

Neutral Landscape Models (NLMs)

In this course, we will use the NLMR package in R, which offers a wide array of models for generating artificial landscapes. These include, but are not limited to, the following types of NLMs (Sciaini et al., 2018):

  • Random NLMs: Generate completely random landscapes.

  • Distance Gradient NLMs: Simulate landscapes with a gradient effect based on distance.

  • Linear Gradient NLMs: Create landscapes with linear environmental gradients.

  • Gaussian Random Field Models: Model landscapes with complex spatial autocorrelation.

These NLMs can also be combined or manipulated to create more complex and realistic environments for testing species distribution models (see R package landscapetools).

“The speciation of NLMs is really only limited by imagination” (With & King 1997)

In the image below you can see examples of the NLMs that are implemented into the NLMR package which we will use in this course.

Collection of all neutral landscape models *Image: Collection of all neutral landscape models (NLM) implemented in the NLMR package. Image from Sciaini et al., 2018.

Further reading:

Grimmett, L., Whitsed, R., & Horta, A. (2021). Creating virtual species to test species distribution models: The importance of landscape structure, dispersal and population processes. Ecography, 44(5), 753–765. https://doi.org/10.1111/ecog.05555

Leroy, B., Meynard, C. N., Bellard, C., & Courchamp, F. (2016). Virtualspecies, an R package to generate virtual species distributions. Ecography, 39(6), 599–607. https://doi.org/10.1111/ecog.01388

Sciaini, M., Fritsch, M., Scherer, C., & Simpkins, C. E. (2018). NLMR and landscapetools: An integrated environment for simulating and modifying neutral landscape models in R. Methods in Ecology and Evolution, 9(11), 2240–2248. https://doi.org/10.1111/2041-210X.13076

Updated: