sl.grid.curvilin2unstr: Convert Curvilinear to Unstructured Grid

In FESOM/spheRlab: Spherical Geometry, Analysis, and Plotting of Geoscientific Data on Arbitrary Grids

Convert Curvilinear to Unstructured Grid

Description

Convert a curvilinear grid, where fields are provided in matrices and lat-lon coordinates are provided either in matrices (general) or vectors (only for lat-lon meshes), into an identical unstructured grid. The conversion allows to apply all the unstructured-grid algorithms also to curvilinear-grid data, without remapping.

Usage

sl.grid.curvilin2unstr(lon = NULL, lat = NULL, Nx = NULL, Ny = NULL, vars = NULL, neighnodes = TRUE, neighelems = TRUE, quad2triag = TRUE, quad2triag.mode = "zigzag", transpose = FALSE, cut.top = 0, cut.bottom = 0, cut.left = 0, cut.right = 0, close.sides = FALSE, close.top = FALSE, close.bottom = FALSE, close.topbottom.skip = "none")

Arguments

lon	an Nx X Ny matrix or an Nx vector providing the longitudes of the curvilinear grid, or NULL (default). In the latter case, Nx must be provided.
lat	an Nx X Ny matrix or an Ny vector providing the latitudes of the curvilinear grid, or NULL (default). In the latter case, Ny must be provided.
Nx	an integer specifying the number of longitudes. Default is NULL, in which case lon must be provided.
Ny	an integer specifying the number of latitudes. Default is NULL, in which case lat must be provided.
vars	a list of variables defined on the curvilinear grid. If the list elements are matrices, they are converted to vectors before they are returned.
neighnodes	a logical value specifying whether to return a matrix with the neighbour nodes for each node. Default is TRUE.
neighelems	a logical value specifying whether to return a matrix with the neighbour elements for each node. Default is TRUE.
quad2triag	a logical value specifying whether to cut rectangles into triangles. Default is TRUE (and currentls only that option is implemented).
quad2triag.mode	a character specifying how to cut rectangles into triangles (if quad2triag=TRUE). Default is 'zigzag' (and currently only that option is implemented).
transpose	a logical value specifying whether to transpose the input matrices. This will affect the order of nodes in the output mesh and which sides of the matrix are considered as 'sides', 'top', and 'bottom', which is relevant for boundary cut-offs and closure (see further arguments). Default is FALSE.
cut.top	an integer specifying how many rows at the top (first rows) are removed. Default is 0.
cut.bottom	an integer specifying how many rows at the bottom (last rows) are removed. Default is 0.
cut.left	an integer specifying how many columns at the left (first columns) are removed. Default is 0.
cut.right	an integer specifying how many columns at the right (last columns) are removed. Default is 0.
close.sides	a logical value specifying whether to close the gap at the sides by connecting the first and last columns (or rows if transpose = TRUE) of nodes across the 'zonal' gap. Default is FALSE.
close.top	a logical value specifying whether to close the 'pole hole' at the top by connecting the first row (or column if transpose = TRUE) of nodes across the grid pole. Note that 'top' and 'bottom' do not necessarity correspond with the North and South Pole!
close.bottom	a logical value specifying whether to close the 'pole hole' at the bottom by connecting the last row (or column if transpose = TRUE) of nodes across the grid pole. Note that 'top' and 'bottom' do not necessarity correspond with the North and South Pole!
close.topbottom.skip	a character, one of "none", "left", and "right". Affects the way the top and/or bottom rows are closed. If "none" (default), the first an last node of the first row are considered vis-a-vis, the second and second-last, and so forth. If "left", the first node (and the first node of the second half) of the row is skipped, that is, the second and last nodes are considered vis-a-vis, the this and second-last, and so forth. If "right", the first and second-last nodes are considered vis-a-vis, and so forth.

Details

To remove land points from the grid, use sl.grid.reduce, see the example below.

Closing the top and/or bottom boundaries works only with an even number of columns.

If the input is based on matrices (rather than lon/lat vectors), the returned element curvilin2unstr.fun is a function that can be used to convert data from the original curvilinear grid to the unstructured grid.

Value

a list with the elements lon, lat, elem, coast, openbound, neighnodes, neighelems, vars, and curvilin2unstr.fun. Some of these may be NULL, depending on the input arguments and on which options have already been implemented.

Author(s)

Helge Goessling

See Also

sl.grid.reduce

Examples

# generate an example curvilinear grid
lon.0 = seq(-60,60,5)
lat.0 = seq(-30,30,5)
grd.0 = sl.grid.curvilin2unstr(lon = lon.0, lat = lat.0)
abg = sl.lonlatrot2abg(c(5,-20,0))
rot.lonlat = sl.rot(lon = grd.0$lon, lat = grd.0$lat, alpha = abg[1], beta = abg[2], gamma = abg[3])
grd = grd.0
grd$lon = rot.lonlat$lon
grd$lat = rot.lonlat$lat

# generate corresponding example data and insert NAs that define a land mask (arbitrary,
# not matching true land coordinates)
dat = exp(-((sl.gc.dist(lon = c(180,grd$lon), lat = c(85,grd$lat), sequential = FALSE)/0.4)^2))
dat.orig = dat
dat[c(7,140,250)] = NA
for (i in 1:2) {
  for (j in which(is.na(dat))) {
    dat[grd$neighnodes[j,]] = NA
  }
}

# remove land nodes from the grid
grd.reduced = sl.grid.reduce(grd = grd, remove.points = is.na(dat), return.info = TRUE)

# compute the 0.15-contour on the reduced grid
cont = sl.contours(var = dat[grd.reduced$reduce.kept$nodes], elem = grd.reduced$elem, levels = 0.15, neighmat = grd.reduced$neighnodes, lon = grd.reduced$lon, lat = grd.reduced$lat)

# plot
pir = sl.plot.init(projection = "polar", do.init.device = FALSE, col.background = "lightgrey")
cb1 = sl.plot.field.elem(pir, dat.orig, lon = grd$lon, lat=grd$lat, elem=grd$elem,
colbar = sl.colbar(cols=c("red","yellow")))
cb2 = sl.plot.field.elem(pir, dat[grd.reduced$reduce.kept$nodes], lon = grd.reduced$lon,
lat = grd.reduced$lat, elem = grd.reduced$elem, colbar = sl.colbar(cols=c("darkblue","white")))
cont = sl.contours(var = dat[grd.reduced$reduce.kept$nodes], elem = grd.reduced$elem,
levels = 0.15, neighmat = grd.reduced$neighnodes, lon = grd.reduced$lon, lat = grd.reduced$lat)
sl.plot.contours(pir, cont, col = "white", lwd = 2)
sl.plot.end(pir, do.close.device = FALSE)

FESOM/spheRlab documentation built on April 6, 2024, 6:52 p.m.