gsMPGstitch: Extract a minimum planar graph (MPG) model in strips from a...
In grainscape: Grains of connectivity and minimum planar graph modelling of landscape connectivity (Windows only)

Description Usage Arguments Details Value Note Author(s) References See Also Examples

This function is used to extract a minimum planar graph (MPG) in strips, and can be used as an alternative to gsMPG. It can save time where there are a large number of patches by dividing the landscape into a number of strips and stitching them back together. It can also extract MPGs for each strip in parallel (with the parallel package). Using this function is necessary when there are a very large number of patches (usually in excess of 2000), as SELES and consequently gsMPG will fail on such landscapes.

Stripping and extracting separate MPGs may sometimes result in small artefactual differences in graph topology and Voronoi tessellations. Researchers may wish to quantify the impact of stripping by using a small subregion of their landscape, and comparing the results from gsMPG(x) and gsMPGstrip(x, numStrips=2, ...).

Note that gsMPGstitch cannot be used to generate MPGs for use in lattice-based grains of connectivity modelling. gsMPG must be used for this purpose.

1
2
3

gsMPGstitch(cost, patchid, numStrips, percentOverlap,
disttype="Cost", cpu=1, outputFolder=NULL, filterPatch=NULL,
spreadFactor=0, selesPath = system.file("SELES", package = "grainscape"))

`cost`	A raster of class `RasterLayer` giving a landscape resistance surface, where the values of each raster cell are proportional to the resistance to movement, dispersal, or gene flow for an organism in the landscape feature they represent. Missing values `NA` are acceptable (but see below). Negative values are not. To extract an MPG with Euclidean links (i.e. and not least-cost path links) set `cost[] <- 1`.
`patchid`	PLEASE WRITE. Note that it is different from gsMPG which requires a binary patch raster as input.
`numStrips`	PLEASE WRITE
`percentOverlap`	PLEASE WRITE
`disttype`	PLEASE WRITE
`cpu`	For parallel computation of each MPG strip. If `cpu=1` serial computation is used. If `cpu > 1` a cluster of up to this many processors will be initiated using the `parallel` package. Using this feature is recommended if processors are available. Typically one less processor than the available number of virtual cores works well (e.g. 7 on an i7 chip).
`outputFolder`	Optional. If not supplied this function creates files for use by `SELES` in a temporary folder placed in the R working directory that is deleted following successful execution. Another location may be specified instead. If supplied the location is not deleted after the analysis completes. This can be useful for debugging purposes.
`filterPatch`	Optional. Remove patches from the analysis that are smaller than a given number of cells.
`spreadFactor`	Optional. Fine-grained control over the accuracy of Voronoi polygons. To reduce accuracy and increase speed, set this as `spreadFactor=10` or `spreadFactor=100`.
`selesPath`	Optional. The location of the `SELES` installation. By default this is the folder in the package installation.

Use this function to create a minimum planar graph (MPG) that can be further analyzed using igraph routines.

It is also the first step in grains of connectivity (GOC) modelling.

A gsMPG object, consisting of a list of objects.

The main elements:
$mpg is the minimum planar graph as class igraph
$patchId is the input patch raster with patch cells assigned to their id (RasterLayer)
$voronoi is the Voronoi tessellation of the patches and resistance surface (RasterLayer)
$lcpPerimWeight gives the paths of the links between patches and their accumulated costs (RasterLayer)
$lcpLinkId gives the paths of the links between patches and their id (RasterLayer)
$lcpPerimType gives the paths of the links between patches and their type (RasterLayer; see notes)
$mpgPlot provides a quick way of visualizing the mpg (RasterLayer)

The $mpg has useful vertex and edge attributes. Vertex attributes give attributes of patches including patch area, the area of patch edges, the core area of each patch, and the coordinates of the patch centroid. All areal measurements are given as raster cell counts. Edge attributes give attributes of the graph links including link weights giving accumulated resistance/least-cost path distance, Euclidean distance, and the start and end coordinates of each link.

SELES has been compiled for Windows. Therefore use of gsMPG is limited to Windows-based platforms.

MORE DETAILS IN HERE

Four types of links are identified in the MPG (1=Nearest neighbour; 2=Minimum spanning tree; 3=Gabriel; 4=Delaunay;)

Areal measurements are given as raster cell counts. If the raster projection is one where cell sizes are approximately constant in area (e.g. UTM), or the raster covers a relatively small geographic extent (e.g. < 1000 km in dimension) areal measurements will often be adequate. Reprojection of rasters should be considered to minimize these effects in other cases (see projectRaster).

Bronwyn Rayfield (bronwyn.rayfield@mail.mcgill.ca), Paul Galpern (pgalpern@gmail.com), Andrew Fall

Fall, A., M.-J. Fortin, M. Manseau, D. O'Brien. (2007) Spatial graphs: Principles and applications for habitat connectivity. Ecosystems. 10:448:461

Galpern, P., M. Manseau, P.J. Wilson. (2012) Grains of connectivity: analysis at multiple spatial scales in landscape genetics. Molecular Ecology 21:3996-4009.

gsGOC, gsThreshold

## Not run: 

## Create a resistance surface
frag <- raster(system.file("extdata/fragmented.asc", package="grainscape"))
fragCost <- reclassify(frag, rcl=cbind(c(1,2,3,4), c(1,10,8,3)))

## Unlike with gsMPG() it is necessary to create a patchid raster
## before-hand.  This can be done with raster::clump()
fragPatchId <- raster::clump(fragCost==1, directions=8, gaps=FALSE)
fragPatchId[is.na(fragPatchId[])] <- 0

## Create an MPG graph by stitching together 3 strips
## Do this in parallel with two processor cores
fragMPGstitch <- gsMPGstitch(cost=fragCost, patchid=fragPatchId,
    numStrips=3, percentOverlap=50, outputFolder=NULL, cpu=2)


## End(Not run)