fast: Create a GRaster or GVector

In adamlilith/fasterRaster: Faster Raster and Spatial Vector Processing Using 'GRASS GIS'

Create a GRaster or GVector

Description

fast() creates a GRaster or GVector from 1) a file; 2) from a SpatRaster, SpatVector, or sf vector; or 3) from a numeric vector, matrix, data.frame, or data.table. Behind the scenes, this function will also create a connection to GRASS if none has yet been made yet.

GRASS supports loading from disk a variety of raster formats (see the GRASS manual page for r.in.gdal (see grassHelp("r.in.gdal")) and vector formats v.in.ogr (see grassHelp("v.in.ogr")'), though not all of them will work with this function.

Note that GVectors may fail to be created if they contain issues that do not coincide with the topological data model used by GRASS. The most common of these is overlapping polygons. See Details on how to fix these kinds of issues.

Note also that GRASS (and thus, fasterRaster) is not very fast when loading vectors. So, if the vector is large and you only want a portion of it, consider using the extent argument to load the spatial subset you need.

Usage

## S4 method for signature 'character'
fast(
  x,
  rastOrVect = NULL,
  levels = TRUE,
  extent = NULL,
  correct = TRUE,
  snap = NULL,
  area = NULL,
  steps = 10,
  dropTable = FALSE,
  resolve = NA,
  verbose = TRUE,
  ...
)

## S4 method for signature 'SpatRaster'
fast(x, ...)

## S4 method for signature 'SpatVector'
fast(
  x,
  extent = NULL,
  correct = TRUE,
  snap = NULL,
  area = NULL,
  steps = 10,
  dropTable = FALSE,
  resolve = NA,
  verbose = TRUE
)

## S4 method for signature 'sf'
fast(
  x,
  extent = NULL,
  correct = TRUE,
  snap = NULL,
  area = NULL,
  steps = 10,
  resolve = NA,
  dropTable = FALSE,
  verbose = TRUE
)

## S4 method for signature 'missing'
fast(x, rastOrVect, crs = "")

## S4 method for signature 'numeric'
fast(x, crs = "", keepgeom = FALSE)

## S4 method for signature 'data.frame'
fast(x, geom = 1:2, crs = "", keepgeom = FALSE)

## S4 method for signature 'data.table'
fast(x, geom = 1:2, crs = "", keepgeom = FALSE)

## S4 method for signature 'matrix'
fast(x, geom = 1:2, crs = "", keepgeom = FALSE)

Arguments

x	Any one of: A SpatRaster raster. Rasters can have one or more layers. A SpatVector or sf spatial vector. See especially arguments correct, area, snap, steps, and verbose. A character string or a vector of strings with the path(s) and filename(s) of one or more rasters or one vector to be loaded directly into GRASS. The function will attempt to ascertain the type of object from the file extension (raster or vector), but it can help to indicate which it is using the rastOrVect argument if it is unclear. For rasters, see especially argument levels. For vectors, see especially arguments correct, resolve, area, snap, steps, and verbose. A vector with an even number of numeric values representing longitude/latitude pairs. See arguments geom, keepgeom, and crs. A data.frame, data.table, or matrix: Create a points GVector. Two of the columns must represent longitude and latitude. See arguments geom, keepgeom, and crs. Missing: Creates a generic GRaster or GVector. You must specify rastOrVect; for example, fast(rastOrVect = "raster"). Also see argument crs.
rastOrVect	Either NULL (default), or "raster" or "vector": If x is a filename, then the function will try to ascertain whether it represents a raster or a vector, but sometimes this will fail. In that case, it can help to specify if the file holds a raster or vector. Partial matching is used.
levels	(GRasters only): Any of: Logical: If TRUE (default) and at least one layer of a raster is of type integer, search for a "levels" file, load it, and attach levels. A levels file will have the same name as the raster file, but end with any of "rdata", "rdat", "rda", "rds", "csv", or "tab" (case will generally not matter). If such a file is not found, no levels will be assigned. The levels file must contain either a data.frame, data.table, or list of data.frames or data.tables, or NULL. A data.frame, data.table, or list of data.frames or data.tables with categories for categorical rasters. The first column of a table corresponds to raster values and must be of type integer. A subsequent column corresponds to category labels. By default, the second column is assumed to represent labels, but this can be changed with activeCat<-. Level tables can also be NULL (e.g., data.fame(NULL)). You can also assign levels after loading a raster using levels<-. NULL: Do not attach a levels table. #'
extent	(GVectors only): Either a NULL (default), or a GVector, a SpatVector, a SpatExtent object, an sf vector, a bbox object, or a numeric vector of 4 values providing a bounding box. If provided, only vector features within this bounding box are imported. If extent is a numeric vector, the values must be in the order west, east, south, north. If NULL, the entire vector is imported.
correct	Logical (GVectors only): Correct topological issues. See Details for more details! By default, this is TRUE.
snap	GVectors only: Numeric or NULL (default). The value of snap indicates how close vertices need to be for them to be shifted to to the same location. Units of snap are map units (usually meters), or degrees for unprojected CRSs. For lines and polygons vectors, a value of NULL will invoke an iterative procedure to find an optimal, smallest value of snap. To turn snapping off, set snap = 0. See Details for more details!
area	Polygon GVectors only: Either a positive numeric value or NULL (default). Remove polygons with an area smaller than this value. Units of area are in square meters (regardless of the CRS). If NULL, then an iterative procedure is used to identify a value of area that results in a topologically correct polygon vector. For point and lines vectors, this argument is ignored. To turn area removal off, set area = 0. See Details for more details!
steps	GVectors only: A positive integer > 1 (default is 10). When using automatic vector correction (i.e., either snap = NULL and/or area = NULL), this is the number of values of snap and/or area to try to generate a correct topology, including no snapping or polygon removal (i.e, snap = 0 and area = 0).
dropTable	GVectors only: Logical. If TRUE, then drop the data table associated with a vector. By default, this is FALSE. See Details for more details!
resolve	GVectors only: Character. If a GVector would be topologically invalid after a first attempt at creating it, then this method will be used to resolve the issue and create a valid GVector. Partial matching is used. "disaggregate": Coerce each area of overlap between polygons into its own geometry. The output will not have a data table associated with it. "aggregate": Coerce all geometries into a "multipart" geometry so it acts like a single geometry. The output will not have a data table associated with it. NA (default): Do neither of the above and if either snap or area is NULL, keep trying to create the GVector. Upon success, the GVector will retain any data table associated with it unless dropTable is FALSE.
verbose	GVectors only: Logical. Displays progress when using automatic topology correction.
...	Other arguments:: table (GVectors–useful mainly to developers, not most users): A data.frame or data.table with one row per geometry in a GVector. Serves as an attribute table. xVect (GVectors–useful mainly to developers, not most users): The SpatVector that corresponds to the file named by x.
crs	String: Coordinate reference system (CRS) WKT2 string. This argument is used for creating a GVector from a numeric vector or a data.frame or similar, or from fast(rastOrVect = "vector") or fast(rastOrVect = "raster"). By default, the function will use the value of crs() (no arguments), which is the CRS of the current GRASS "project/location" (see vignette("projects_mapsets", package = "fasterRaster")).
keepgeom	Logical: If x is a set of numeric coordinates, or a data.frame or similar, then they can be coerced into a points GVector. If keepgeom is TRUE, then the coordinates will be included in the data table of the GVector. The default is FALSE.
geom	Character or integer vector: If x is a data.frame, data.table, or matrix, this specifies which columns of x represent longitude and latitude. Columns can be given by name (a character vector) or index (a numeric or integer vector). The default is to use the first two columns of x.

Details

GRASS uses a "topological" model for vectors. Topological issues generally arise only with polygon vectors, not point or line vectors. Sometimes, polygons created in other software are topologically incorrect–the borders of adjacent polygons may cross one another, or there may be small gaps between them. These errors can be corrected by slightly shifting vertices and/or removing small polygons that result from intersections of larger ones that border one another. A topological system also recognizes that boundaries to adjacent polygons are shared by the areas, so should not be ascribed attributes that belong to both areas (e.g., the shared border between two countries "belongs" to both countries).

By default, fast() will try to correct topological errors in vectors. There are three levels of correction, and they are not necessarily mutually exclusive:

1. Automatic correction: By default, fast() will apply automatic topology correction. You can turn this off using the correct = FALSE argument, though in most cases this is not recommended.

2. Manual snapping and/or area removal: In addition to correction from step 1, you can cause vertices of polygons close to one another to be "snapped" to same place and/or polygons that are smaller than some threshold to be removed. Problems with mis-aligned vertices arise when adjacent polygons are meant to share borders, but slight differences in the locations of the vertices cause them to mis-align. This mis-alignment can also produce small "slivers" of polygons that are the areas where they overlap. You can snap vertices within a given distance of one another using the snap argument followed by a numeric value, like snap = 0.000001. Units of snap are in map units (usually meters) for projected coordinate reference systems and degrees for unprojected systems (e.g., WGS84, NAD83, NAD27). You can also remove polygons that are smaller than a particular area using the area argument followed by a numeric value (e.g., area = 1). The units of area are in meters-squared, regardless of the coordinate reference system. Note that using snap and area entails some risk, as it is possible for nearby vertices to actually be distinct and for small areas to be legitimate.

3. Automatic snapping and/or area removal: In addition to the correction from step 1, you can use automatic snap and/or area correction on polygons vectors by setting snap and/or area to NULL (i.e., their default values). If just snap is NULL, only automatic snapping will be performed, and if just area is NULL, then only automatic area removal will be performed. Regardless, you will also need to set an integer value for steps, which is the number of steps to take between the smallest value of snap and/or area and the maximum value attempted. The function will then proceed by first attempting snap = 0 and/or area = 0 (i.e., no snapping or area removal). If this does not produce a topologically correct vector, GRASS will (internally) suggest a range for snap. The fast() function then creates steps values from the lowest to the highest values of this range evenly-spaced along the log values of this range, then proceed to repeat the importing process until either the vector is imported correctly or the maximum value of snap is reached and results in a failed topology. Smaller values of step will result in more fine-grained attempts so are less likely to yield overcorrection, but can also take more time. The value of area in automatic correction is set to snap^2. NB: Automated snapping and area removal are only performed on polygons vectors, even if snap or area is NULL. To snap lines or points, you must set snap equal to a numeric value. The area correction is ignored for points and lines.

Issues can also arise due to:

• Data table-vector mismatching: If your vector has a data table ("attribute table") associated with it, errors can occur if there are more/fewer geometries (or multi-geometries) per row in the table. If you do not really need the data table to do your analysis, you can remove it (and thus obviate this error) using dropTable = TRUE.

• Dissolving or aggregating "invalid" geometries: Using the resolve argument, you can create a topologically valid vector by either coercing all overlapping portions of polygons into their own geometries (resolve = "disaggregate"), or by coercing them into a single, combined geometry (resolve = "aggregate"). Aggregation/disaggregation will be implemented after loading the vector into GRASS using the settings given by snap and area. Aggregation/disaggregation will cause any associated data table to be dropped (it forces dropTable to be TRUE). The default action is to do neither aggregation nor disaggregation (resolve = NA).

If none of these fixes work, you can try:

• Correction outside of fasterRaster: Before you convert the vector into fasterRaster's GVector format, you can also try using the terra::makeValid() or sf::st_make_valid() tools to fix issues, then use fast().

• Post-conversion to a GVector: If you do get a vector loaded into GVector format, you can also use a set of fasterRaster vector-manipulation tools or fillHoles() to fix issues.

Value

A GRaster or GVector.

See Also

rgrass::read_RAST() and rgrass::read_VECT(), vector cleaning, fillHoles(), plus GRASS modules v.in.ogr (see grassHelp("v.in.ogr")) and r.import (see grassHelp("r.import"))

Examples

if (grassStarted()) {

# Setup
library(sf)
library(terra)

# Example data
madElev <- fastData("madElev") # integer raster
madCover <- fastData("madCover") # categorical raster
madCoast4 <- fastData("madCoast4") # polygons vector
madRivers <- fastData("madRivers") # lines vector
madDypsis <- fastData("madDypsis") # points vector

### Create GRasters from SpatRasters

# Create an integer raster:
elev <- fast(madElev)
elev

# Create a categorical raster:
cover <- fast(madCover)
madCover
levels(madCover) # category levels

# Create a GRaster from a file on disk:
rastFile <- system.file("extdata", "madForest2000.tif", package = "fasterRaster")
forest2000 <- fast(rastFile)
forest2000

# Create a 1's raster that spans the world:
ones <- fast(rastOrVect = "raster", crs = "epsg:4326")
ones

### Create GVectors

# Create a GVector from an sf vector:
coast4 <- fast(madCoast4)
coast4

# Create a GVector from a SpatVector:
madRivers <- vect(madRivers)
class(madRivers)
rivers <- fast(madRivers)
rivers

# Create a GVector from a vector on disk:
vectFile <- system.file("extdata/shapes", "madCoast.shp",
   package = "fasterRaster")
coast0 <- fast(vectFile)
coast0

# Import only Dypsis occurrences in a restricted area:
ant <- coast4[coast4$NAME_4 == "Antanambe"]
dypsisRestrict <- fast(madDypsis, extent = ant)
dypsis <- fast(madDypsis)

plot(coast4)
plot(ant, col = "gray80", add = TRUE)
plot(dypsis, add = TRUE)
plot(dypsisRestrict, col = "red", add = TRUE)

# Create a generic GVector that spans the world:
wallToWall <- fast(rastOrVect = "vector", crs = "epsg:4326") # WGS84
wallToWall

# Create a GVector from a numeric vector
pts <- c(-90.2, 38.6, -122.3, 37.9)
pts <- fast(pts, crs = "epsg:4326") # WGS84

# Create a GVector from a matrix (can also use data.frame or data.table):
mat <- matrix(c(-90.2, 38.6, -122.3, 37.9), ncol = 2, byrow = TRUE)
mat <- fast(mat, crs = "epsg:4326", keepgeom = TRUE) # WGS84

}

adamlilith/fasterRaster documentation built on Dec. 21, 2024, 2:04 a.m.