sdm package to accept sf, stars, and terra spatial data output formats
One of the greatest spatial analysis packages in R called raster has seen great growth in development of spatially sensitive packages such as sdm since its inception in 2010. However, maintenance of the package is likely to be halted in 2023 among other platforms like maptools, sp, rgeos, and rgdal. As a response,other spatial analysis packages have come to the surface such as sf, terra, and stars. At its present version (1.1.8), the model building process of the package does not support the incoming packages but still run on the older retiring packages. There is therefore need to shift the capabilities of sdm package, of course and many other packages depending on the older platforms, to embrace the new file structures produced by the new ‘tidy’ packages. The new platforms are praised for being able to do more, easier to use, and faster see. It is in this light that I draft this short post.
Check my GitHub for code files.
Using st_read() function from sf package, we can easily load spatial vector data into R in a tidy kind of spatial data that can be easily wrangled by other tidyverse packages. Here I load in species occurrence data that comes with the sdm pakage.
file <- system.file("external/species.shp", package = "sdm")
species <- st_read(file) # Note, instead of shapefile(file) I use st_read(file)## Reading layer `species' from data source
## `D:\R2\R-4.1.2\library\sdm\external\species.shp' using driver `ESRI Shapefile'
## Simple feature collection with 200 features and 1 field
## Geometry type: POINT
## Dimension: XY
## Bounding box: xmin: 110112 ymin: 4013700 xmax: 606053 ymax: 4275600
## Projected CRS: WGS 84 / UTM zone 30NHere, I read in raster predictor variables using read_stars() from stars package.
path <- system.file("external", package = "sdm")
lst <- list.files(path = path, pattern = 'asc$', full.names = T)
preds <- read_stars(lst) # Instead of stack() from raster, I use read_stars() from stars package.Here I use st_extract() function from sf to obtain pixel values corresponding to occurrence points. Note vifcor(), at its present version (1.1.18), does not support stars objects from stars package and SpatRaster or terra from terra package. This, accordingly, also needs improvement.
st_crs(preds) <- st_crs(species)
extract <- st_extract(preds, species)
v <- vifcor(data.frame(elevation = extract$elevation.asc,
precipitation = extract$precipitation.asc,
temperature = extract$temperature.asc,
vegetation = extract$vegetation.asc))It is however, still not possible to use the sf and stars objects above directly in sdmData() function, meaning no sdm model can be built upon them.
# d <- sdmData(formula=Occurrence~., train=species, predictors=preds)It is therefore my humble call, and I hope the larger sdm community would agree, that future versions of sdm and usdm packages should consider accommodating outputs from modern packages such as sf, stars, and terra that we use to prepare data just before building and running the model. The sdm community is also concerned with an issue while generating output gui with latest version of shiny package (I have not detailed that here).
Personally, I am not a developer but rather a user and I think identifying such looming problems could help developers work around them in time for seamless transition.
Best regards and enjoy the holidays.