Produce diversity maps from raster data with biodivMapR

Jean-Baptiste Féret, Florian de Boissieu

2026-09-08

biodivMapR was initially developed to produce $\alpha$ and $\beta$ diversity maps Féret & Asner (2014) from airborne imaging spectroscopy data acquired over tropical forests. biodivMapR v1 used spectral transformations as initial pre-processing step. This included reflectance normalization with a continuum removal, followed by a spectral transformation (PCA, SPCA, MNF…).

Input raster data fom any type of sensor can be used, as long as the input rasters share the same footprint, spatial resolution and projection. Starting from biodivMapR v2.0, the workflow aims at being ‘sensor agnostic’, meaning that users can provide any type of raster data as input, in order to produce diversity metrics from a collection of rasters sharing the same projection, spatial resolution and extent. The workflow described in v1 is still available, but functions have been renamed.

The ecological relevance of the input data used with biodivMapR is the sole responsibility of users. While latest versions of this package provide functions to perform large scale mapping with computational efficiency and capacity to take advantage of cloud resources, it remains a method dedicated to research. Existing publications provide guidelines for the production of diversity maps. However, the ecological relevance of ‘spectral diversity’ metrics depends on the type of input data, the parameterization of the method, and the type of ecosystem under study. In situ observations remain fundamental to identify the proper parameterization,
test hypotheses and produce validation for the diversity maps.

The tutorial provided with biodivMapR describes a possible processing workflow to take advantage of Sentinel-2 imagery. Sentinel-2 imagery is downloaded from a STAC catalog using the package preprocS2, which is a wrapper of the rstac and CDSE packages. The default STAC collection is sentinel-2-l2a from the Planetary computer catalog.

biodivMapR v2 now allows user to run the preprocessing step of their choice. This includes spectral indices, vegetation biophysical variables, spectral transformation, or any type of preprocessing dedicated to a specific sensor / sensor combination.

We provide a few examples here. Users can experiment with the method of their choice.

• Definition of the processing parameters This includes the following information:

  • input / output files paths
  • output spatial resolution (window size)
  • pre-processing and processing options

• Computation of the diversity maps This includes several diversity metrics:

  • $\alpha$ and $\beta$ diversity maps as defined in Feret & Asner (2014) and Feret & de Boissieu (2019)
  • Functional diversity maps, including Richness, Evenness, and Divergence as defined in Villeger et al. (2008), as well as Dispersion and Rao entropy. These functional diversity metrics are computed with the package fundiversity
  • Since biodivMapR v2 and v3, additional $\alpha$ and $\beta$ diversity metrics have been implemented. $\alpha$ diversity metrics include Hill index (with a Hill number to be defined by the parameter Hill_order). In addition to Bray-Curtis dissimilarity (bray), $\beta$ diversity metrics include Bray-Curtis turnover (brayturn), Jaccard dissimilarity (jaccard), Jaccard turnover (jaccardturn), Simpson dissimilarity (simpson_diss) and Sorensen dissimilarity (sorensen). Please refer to the R package bioregion for the definition of these $\beta$ diversity metrics

biodivMapR v1

Please refer to the branch dev_v1 of the package for the previous versions and tutorials. Vignettes and tutorials corresponding to this version are still available here.

We recommend using the most-up-to-date versions of biodivMapR, as the previous ones will not be maintained.