Data preparation for multiple workflows and multiple study areas
The most generic case study is when several maps are available for several locations. A simple folder tree allows an easy analysis.
Input data structure
When several LCZ maps are available for several places and stored into local files, the following data structure allows an easier analysis.
–main_data_directory
|–location1
|--wf1_lcz.fgb
|--wf2_lcz.fgb
|--wfp_lcz.fgb|–location2
|--wf1_lcz.fgb
|--wf2_lcz.fgb
|--wfp_lcz.fgbThe main data directory must contain only the data subdirectories, as the loading functions will span the subdirectories searching for data.
In each subdirectory, the data are stored in files named
wfNamei_lcz.fgb where
Other spatial extensions may work, but .fgb files are
recommended and were the most tested.
General steps to compare several LCZ maps
Import. First, all the data is loaded in a list.
This list contains a sublist for each location, and each location sublist contains one sf object for each LCZ map.
Intersection. Second, the spatial units of all maps (i.e. the geometries bearing the LCZ type) are intersected. On any resulting geometry, a pair of workflow either totally agree or totally disagree. Intersection is done within lists, which allows to try to intersect only geometries of the same location. The result is concatenated in a single sf objects which contains a column for each lcz map, named after the workflow it comes from, a column with the location name, a column with the area and a column with the geometry for each intersected spatial unit.
Visualisation
Simple descriptive barplots give a general idea of the behavior of the different workflows.
Pairwise comparison allow a visual correspondence between 2 workflows.
Circular Chord Diagram allow a simultaneous visualization of correspondences between all workflows.
Numerical Comparison
Pairwise agreement between workflow and consensus value for each LCZ type quantify this comparison.
Importing maps from several locations and several workflows
In case of several workflows and several locations, the function
importMultipleLocsLCZvect allow to load all the data in a
list.
In the data structure from above section, the function would be
applied to the main_directory folder.
The small data files for this vignette are in the
inst/extdata/multipleWfs folder, which contains the
locations Arville and Redon, for which three LCZ maps are available,
produced by the following workflows : - GeoClimate based on
OpenStreetMap input data - GeoClimate based on french Geographical
Information Institute database called BDTOPO - WUDAPT, based on
satellite images and machine learning.
require(lczexplore)
twoLocsDir<-paste0(
system.file("extdata", package = "lczexplore"),"/multipleWfs")
twoLocsSfList<-importMultipleLocsLCZvect(dirPath = twoLocsDir, workflowNames = c("osm","bdt","wudapt"),
location = c("Arville", "Redon"))
#> The data will be imported from the following folders :/tmp/RtmppWIPZu/Rinst132c522a1db42/lczexplore/extdata/multipleWfs/ArvilleThe data will be imported from the following folders :/tmp/RtmppWIPZu/Rinst132c522a1db42/lczexplore/extdata/multipleWfs/Redon
#> The locations are :ArvilleThe locations are :Redon
#> The names of the LCZ types columns were mis-specified, attempt to load with default 'lcz_primary' name
#> Column lcz_primary not found,but LCZ_PRIMARY was, trying with it
#> Column lcz_primary not found,but LCZ_PRIMARY was, trying with it
#> Column lcz_primary not found,but LCZ_PRIMARY was, trying with it
#> Column lcz_primary not found,but LCZ_PRIMARY was, trying with itThe function importMultipleLocsLCZvect loads all the
input .fgb files in a list.
When comparing maps for only one location use the similar function
importMultipleLCZvect.
Intersecting geometries from the different workflows
The spatial units
In the case of only two workflows, the function
compareLCZ loads the two sf objects,
intersects their geometries, produce the array of correspondance between
the LCZ types of the two workflows.
With more than two workflows, one needs more control over the intersection steps. First, in case of multiple locations, the intersection of the several workflows must occur within each location. Second, intersecting the spatial units of multiple workflows can lead to artificially small geometries.
The createIntersect function allows to specify a minimal
area threshold. Resulting geometries whose area are lesser than this
threshold will be discarded.
twoLocsSfIntersected <- createIntersect(sfList = twoLocsSfList, columns = rep("lcz_primary", 4),
refCrs=NULL, workflowNames=c("osm", "bdt", "wudapt"), minZeroArea=0.000)
#> Warning: attribute variables are assumed to be spatially constant throughout
#> all geometries
#> Warning: attribute variables are assumed to be spatially constant throughout
#> all geometries
#> Warning: attribute variables are assumed to be spatially constant throughout
#> all geometries
#> Warning: attribute variables are assumed to be spatially constant throughout
#> all geometriesNote : the function createIntersect is able to see if
the list fed to the sfList argument has one or two levels, i.e. if there
are one or multiple locations.
## Pairwise agreements From this sfInt object, the function
