Spatial Environmental Risk Assessment Through Food Webs

A workflow for spatially explicit environmental risk assessment of contaminated sites: habitat, food web, dispersal, exposure and risk. Starting from a raster of soil contamination, the package builds habitat layers for a set of species, organises them in a directed acyclic graph representing a food web, and propagates the flow of elements (e.g., food, energy, contaminants) along its links. It also provides soil interpolation by kriging, dispersal by kernel convolution or by circuit theory (through the optional 'Julia' package 'Omniscape'), land cover retrieval for France ('OCS-GE') and Europe ('EUNIS'), and Bayesian calibration of transfer models with 'Stan'.

spacemodR

Spatial environmental risk assessment through food webs

spacemodR is an R package for the ecological risk assessment of contaminated sites. It starts from a map of soil contamination and follows the contaminant through the food web, taking into account where each species lives and how it moves. The result is a set of maps: habitat, exposure and risk for each species.

Full documentation: https://qonfluens.github.io/spacemodR/

What the package does

The package covers the steps of a spatial risk assessment:

1. Soil contamination. Build a soil concentration raster from sampling points, by kriging or by kernel smoothing.
2. Habitat. Turn land cover maps (OCS-GE for France, EUNIS for Europe) into habitat rasters, with a weight for each species.
3. Food web. Describe who eats whom as a directed acyclic graph, with a weight on each link.
4. Dispersal. Spread each species around its habitat with a dispersal kernel, or with circuit theory (Omniscape, which requires Julia).
5. Exposure and transfer. Move the contaminant from the soil to each species along the links of the food web.
6. Risk. Compare the exposure maps with toxicity thresholds to map the risk.

The package also provides Bayesian models (written in Stan) to estimate transfer parameters from field data.

Installation

Install the development version from GitHub:

# install.packages("remotes")
remotes::install_github("Qonfluens/spacemodR")

# or a given release
remotes::install_github("Qonfluens/[email protected]")

System requirements. sf and terra need GDAL, PROJ and GEOS. They are already included in the CRAN binaries of these packages for Windows and macOS. spacemodR contains compiled Stan models, so installing it from source needs a C++ compiler (Rtools on Windows, Xcode command line tools on macOS) and takes a few minutes.

Pre-built packages

Each release provides:

• spacemodR_<version>.tar.gz: source package (all platforms, needs a compiler),
• spacemodR_<version>.zip: Windows binary (no compiler needed),
• spacemodR_<version>.tgz: macOS binary (no compiler needed).

Binaries are built with the current R release. With an older R version, use the source package. Download the file for your platform, then:

# install the dependencies first (binaries do not install them)
install.packages(c("dplyr", "ggplot2", "httr", "patchwork", "Rcpp", "rlang",
                   "rstan", "rstantools", "sf", "terra"))

# source package
install.packages("path/to/spacemodR_<version>.tar.gz", repos = NULL, type = "source")
# Windows binary
install.packages("path/to/spacemodR_<version>.zip", repos = NULL, type = "win.binary")
# macOS binary
install.packages("path/to/spacemodR_<version>.tgz", repos = NULL, type = "binary")

Docker

The repository contains a Dockerfile with all system dependencies, including Julia for Omniscape.

# build the image
docker build -t spacemodr .

# run the container (RStudio Server on port 8787)
docker run --rm -ti \
  -p 8787:8787 \
  -e PASSWORD=yourpassword \
  spacemodr

A short example

The example below builds two habitat layers, links them in a food web and transfers cadmium from the soil to a consumer.

library(spacemodR)

# data shipped with the package: land cover and soil cadmium (log10 scale)
data(ocsge_metaleurop)
ground_cd <- load_raster_extdata("ground_concentration_cd_compressed.tif")

# 1. habitats: all the area for the soil, herbaceous cover for the consumer
hab_soil <- habitat() |> add_habitat(ocsge_metaleurop)
hab_vole <- habitat() |>
  add_habitat(ocsge_metaleurop[ocsge_metaleurop$code_cs == "CS2.2.1", ])

stack <- raster_stack(
  list(habitat_raster(ground_cd, hab_soil), habitat_raster(ground_cd, hab_vole)),
  names = c("soil", "vole")
)

# 2. food web: the consumer takes up the contaminant from the soil
web <- trophic() |> add_link("soil", "vole")

# 3. spacemodel: the rasters and the food web together
model <- spacemodel(stack, web)
model[["soil"]][] <- ground_cd

# 4. transfer along the food web, with a Gaussian foraging kernel (sd = 200 m)
kernels <- list(soil = NA, vole = compute_kernel(radius = 200, GSD = 25))
fluxes <- flux(model) |> add_flux(from = "soil", to = "vole", value = 0.5)
exposure <- transfer(model, kernels, fluxes)

terra::plot(exposure)

Documentation

• Quick start (also in French)
• Tutorial: the whole workflow, step by step
• Guides: soil maps by kriging, habitat layers, habitat data sources, connectivity with Omniscape, food intake
• Parameter estimation: soil to target transfers, trophic transfers
• Function reference and change log

Data included

See the help page of each dataset for its source and its license.

The SPACEMOD project

spacemodR is developed within the SPACEMOD project, funded by the French agency for ecological transition (ADEME), call for projects IMPACTS 2023.

Older software for ecological risk assessment, such as TerraSys or BERISP, is no longer maintained. SPACEMOD develops a new, generic and spatially explicit tool for contaminated sites and soils, able to handle several exposure routes and mixtures of chemicals in food webs.

Partners:

• UMR Chrono-environnement (CNRS / Université de Franche-Comté), project coordinator (Pr. Renaud Scheifler)
• Qonfluens SAS
• UK Centre for Ecology & Hydrology (UKCEH)

More information on the ADEME research portal and on the Chrono-environnement website.

Getting help

Please report bugs and ask questions on the issue tracker.

For maintainers

Release a new version

1. Update Version: in DESCRIPTION and add a section to NEWS.md.
2. Commit and push to main, then tag the commit and push the tag:

   git tag -a v0.3.0 -m "Release v0.3.0"
   git push origin v0.3.0
   

3. The workflow .github/workflows/release.yaml builds the source package and the Windows and macOS binaries, and attaches them to a new GitHub release. The tag must match the version in DESCRIPTION, otherwise the workflow stops.

Large files

Large files such as fitted models are tracked with git lfs. Before adding a new large file, track it:

git lfs track "raw_data/new_fit.rda"

License

spacemodR is released under the MIT license. See the LICENSE file.