Boost Spatial Data Science Workflows with GRASS GIS and R

Veronica Andreo and the GRASS GIS community

NCSU GeoForAll Lab at the Center for Geospatial Analytics
NC State University

useR! Conference, Salzburg, July 8-11, 2024

veroandreo.github.io/grass-gis-talks

After a talk with Jakub Nowosad at FOSS4G Europe last week :)

About the presenter

Visiting scholar at NC State's Center for Geospatial Analytics
Lic. & PhD in Biology, MSc in Applications of the Spatial Information
Researcher & lecturer at Gulich Institute in Argentina
GRASS GIS: Development Team, PSC chair
OSGeo: Charter member

What is GRASS GIS?

Geek open-source command line GIS

What is GRASS GIS?

Open-source desktop GIS

What is GRASS GIS?

Processing backend in QGIS

Image source: baharmon.github.io/grass-in-qgis

What is GRASS GIS?

Data analytics tool in R or quarto notebooks

What is GRASS GIS?

Geovisualization and data analytics tool in Python notebooks

What is GRASS GIS?

Geoprocessing engine running in an HPC environment

What is GRASS GIS?

Geospatial platform for developing custom models

What is GRASS GIS?

Cloud geoprocessing backend

Image source: neteler.gitlab.io/actinia-introduction/

GRASS is all of that!

Robustness, Innovation with Stability, Versatility, Scientific foundations

Why using GRASS?

Mature and robust geoprocessing engine

500+ tools in core, 400+ addons

Time series support

Remote Sensing toolset

Hydrology & Geomorphology

Simulations

Unique tools for diverse applications!!

Tools with parallel support

r.slope.aspect

r.neighbors

r.mfilter

r.series

r.patch

r.sun

v.surf.rst

r.sim.water

r.sim.sediment

r.resamp.interp

r.resamp.filter

r.univar

And more at grass.osgeo.org/grass-stable/manuals/keywords.html#parallel

API tooling for parallelizing Python scripts

Data parallelization: GridModule
Task parallelization: multiprocessing, ParallelModuleQueue


            from grass.pygrass.modules.grid import GridModule
        
            grd = GridModule("v.to.rast", input="roads", output="roads",
                             use="val", processes=4)
            grd.run()


            r.mapcalc.tiled "log_dist = if (dist == 0, 0, log(dist))" nprocs=4

HPC ready

Recipes for High Performance Computing setups
Better integration with conda environments: conda package on the way!
Reduced potential for race conditions
Easily create parallel workflows

by Vaclav Petras, Markus Neteler, Nicklas Larsson, …

Streamlined access to data

Automated download and import of diverse datasets i.eodag, i.sentinel, i.modis, i.landsat, r.in.usgs, r.in.nasadem, m.crawl.thredds, t.rast.import.netcdf, t.stac (on the way!), …

by Fondazione Edmund Mach, OpenGeoLabs, mundialis, GSoC, NC State University, CONICET, Norwegian Institute for Nature Research, …

Optimized for large data

392 billion cells!

Fast External Data Links

r.external links (opens) external raster data (GeoTiffs, …) much faster.
(2-5× faster, or almost no time for some workflows)

by Markus Metz

Great for workflows when only portion of the data is processed in GRASS GIS.

R packages using GRASS in the background

hydrographr: wrappers for GDAL and GRASS GIS functions to efficiently work with Hydrography90m and spatial biodiversity data.
rdwplus: implementation of the IDW-PLUS (inverse distance weighted percent land use for streams) algorithm with GRASS GIS tools.
fasterRaster: handle large-in-memory/large-on-disk spatial rasters and vectors in R through GRASS GIS.

Only a (dedicated) subset of GRASS tools available through these packages

R packages interfacing with GRASS

qgisprocess: provides an R interface to the geoprocessing algorithms of QGIS, included GRASS provider (only a selection).
link2GI: functions to connect to common open source GI software packages, via wrapper functions and direct API-use with system calls.
rgrass: interpreted interface between GRASS GIS and R, provides facilities for using all GRASS tools from the R command line.

rgrass

initGRASS(): starts a GRASS session from R
execGRASS(): executes GRASS commands from R
read_VECT() and read_RAST(): read vector and raster maps from a GRASS project into R
write_VECT() and write_RAST(): write vector and raster objects from R into a GRASS project

Access the full stack of tools and extensions in GRASS GIS from R!!

How to use GRASS through the rgrass package?

Choose your own adventure!

Use GRASS GIS tools within your R spatial workflows

Create a temporary GRASS project with the raster or vector map you want to process with GRASS
Write the R map object into the GRASS project
Run the desired GRASS tool(s) from R
Export the results back to R or your disk

Use GRASS GIS tools within your R spatial workflows

Assuming we have a terra SpatRast object, r, within our R session:


    library(rgrass)

    grassbin <- system("grass --config path", intern = TRUE)

    initGRASS(gisBase = grassbin, 
              home = tempdir(), 
              SG = r, 
              override = TRUE)

    write_RAST(r, "terra_elev")

    execGRASS("r.slope.aspect", 
              elevation = "terra_elev", 
              slope = "slope", 
              aspect = "aspect")

    grass_maps <- read_RAST(c("aspect", "slope"))

Alternatively...


    execGRASS("r.import", input=r, output="terra_elev")
    execGRASS("r.external", input="file.tif", output="file") # much faster

Use R tools within GRASS GIS workflows

From R, start GRASS within your project
Read vector or raster data from GRASS into R
Do your analysis, modeling, visualizations with R tools
Optionally write vector and raster outputs back into GRASS

Use R tools within GRASS GIS workflows


    grassdata <- path.expand("~/grassdata/")
    project <- "nc_basic_spm_grass7"
    mapset <- "PERMANENT"

    initGRASS(gisBase = grassbin,,
              gisDbase = grassdata, 
              location = project, 
              mapset = mapset, 
              override = TRUE, 
              remove_GISRC= TRUE)

    execGRASS("g.list", parameters = list(type="raster"))

    elev <- read_RAST("elevation")

    # ... plot, run analysis and models, get outputs

    write_RAST(elev, "new_elevation")

Support for gdal-grass driver added to read directly from GRASS projects without exporting, thanks to Floris Vanderhaeghe

Use R tools within GRASS GIS workflows: an example

veroandreo.github.io/grass_ncsu_2023

GRASS resources

Documentation:
- Learn center
- See full list of core tools
- See tools in Addons repository
Tutorials (my three favorites):

Support

Community support
Commercial support
- mundialis GmbH & Co KG
- OpenGeoLabs (& GISMentors)
- Center for Geospatial Analytics, NC State University
- …

Take home messages

Let's (re)use the best tools available for the tasks at hand
Let's focus on enhancing bridges!

Ask a question, get a special sticker

veroandreo@gmail.com
veroandreo.github.io/grass-gis-talks
Funded by NSF award 2303651