produce diversity maps from a set of features

Jean-Baptiste Féret, Florian de Boissieu

2026-02-19

Once spectral features are computed from optical data, they can be used as input variables in biodivMapR.

Starting from biodivMapR v2, a simplified procedure is followed compared to biodivMapR v1. A unique function named biodivMapR_full runs the full workflow and produces a limited set of outputs.

produce diversity maps from SPCA file

SPCA was previously computed and the relevant spectral features were identified based on visual interpretation. These selected components are used to produce diversity maps with biodivMapR. The SPCA file is a unique raster file stacking all components. biodivMapR requires - the path for the input raster: it can be a unique stacked raster, or a set of individual features. The current version does not accept combination of stacked raster and individual raster layers as input.

• optionally, if biodivMapR is expected to run on a selection of features from the input raster, these can be specified with the variable selected_bands.

• here, we also take advantage of the mask previously computed.

# 1- define biodivMapR output directory 
output_dir_biodivMapR <- file.path(output_dir, 'biodivMapR/SPCA')
dir.create(output_dir_biodivMapR, showWarnings = F, recursive = T)

# 2- define parameters for biodivMapR
window_size <- 10     # window side size for computation of spectral diversity
nb_clusters <- 50     # nb of clusters (spectral species)

# 3- define path for intermediate variables to be saved
# - information related to kmeans clustering defining spectral species
Kmeans_info_save <- file.path(output_dir_biodivMapR,'Kmeans_info.RData')
# - information related to beta diversity mapping (BC dissimilarity + PCoA)
Beta_info_save <- file.path(output_dir_biodivMapR,'Beta_info.RData')

# 4- adjust parameters for multithread & computational efficiency
nbCPU <- 4            # nb of threads for parallel processing
# 5- apply biodivMapR 
selected_bands <- c(1, 4, 5, 6)
options <- set_options_biodivMapR(fun = 'biodivMapR_full')
options$nb_clusters <- nb_clusters
options$maxRows <- 1000
ab_info_SPCA <- biodivMapR_full(input_raster_path = PCA_Output$PCA_Files$PCA, 
                                input_mask_path = mask_path_PCA,
                                output_dir = output_dir_biodivMapR, 
                                selected_bands = selected_bands, 
                                window_size = window_size, 
                                Kmeans_info_save = Kmeans_info_save,
                                Beta_info_save = Beta_info_save, 
                                nbCPU = nbCPU, options = options)

ab_info_SPCA produced from the biodivMapR_full function includes two lists:

• Kmeans_info, which gathers all information related to the k-means clustering: centroid, min and max values for each feature

• Beta_info, which gathers all information related to the computation of the beta diversity

These lists are saved as Rdata files corresponding to the files defined by Kmeans_info_save and Beta_info_save. These files can then be used for an independent process with biodivMapR_full. The path for these files can then be provided for input variables Kmeans_info_read and Beta_info_read.

biodivMapR will not recompute this information, if user provides a valid path for Kmeans_info and Beta_info. This option is interesting when adjusting Kmeans_info and Beta_info for a region including multiple tiles, and processing each tile independently.

$\alpha$ and $\beta$ diversity maps produced from the analysis of selected PCs are displayed below.

Fig. 1. maps of shannon index and beta-diversity obtained from biodivMapR analysis on selected components computed from sentinel-2 acquisition.

 

Produce diversity maps from spectral indices with biodivMapR

Spectral indices computed previously will be used here to produce diversity maps with biodivMapR. Each spectral index was saved following a tiling system, and a vrt was produced. biodivMapR can use the path for the vrt as input_raster_path for biodivMapR_full. Alternatively, it can take advantage of the tiling system in order to process the image tile per tile in parallel.

# 1- define biodivMapR output directory 
output_dir_biodivMapR_si <- file.path(output_dir, 'biodivMapR/spectral_indices')
output_dir_mask <- file.path(output_dir, 'biodivMapR/spectral_indices_mask')
dir.create(output_dir_biodivMapR_si, showWarnings = F, recursive = T)
dir.create(output_dir_mask, showWarnings = F, recursive = T)

# 3- define path for intermediate variables to be saved
# - information related to kmeans clustering defining spectral species
Kmeans_info_save <- file.path(output_dir_biodivMapR_si,'Kmeans_info.RData')
# - information related to beta diversity mapping (BC dissimilarity + PCoA)
Beta_info_save <- file.path(output_dir_biodivMapR_si,'Beta_info.RData')

# 4- apply biodivMapR 
options <- set_options_biodivMapR(fun = 'biodivMapR_full_tiles')
options$nb_clusters <- nb_clusters
options$alpha_metrics <- c('richness', 'shannon', 'hill')

ab_info_SI <- biodivMapR_full_tiles(feature_dir = output_dir_si, 
                                    list_features = si_list, 
                                    output_dir = output_dir_biodivMapR_si, 
                                    mask_dir = output_dir_mask, 
                                    window_size = window_size, 
                                    plots = plots, nbCPU = nbCPU, 
                                    site_name = site_name, options = options)

Fig. 2. maps of $\alpha$-diversity (shannon’ H) and $\beta$-diversity obtained from biodivMapR analysis on spectral indices from sentinel-2 acquisition.

 

The validation can then be performed if ground information is available Validation is described in this tutorial.