Geographically Weighted Regression

Geographically Weighted Regression

gwr(.ref, ...)

# S3 method for character
gwr(
  .ref,
  .tar,
  varid = NULL,
  coordinates = smpds::CRU_coords,
  res = 0.5,
  xy_buffer = 1.5,
  z_buffer = NA,
  cpus = 1,
  bandwidth = 1.06
)

# S3 method for numeric
gwr(
  .ref,
  .tar,
  coordinates = smpds::CRU_coords,
  res = 0.5,
  xy_buffer = 1.5,
  z_buffer = NA,
  cpus = 1,
  bandwidth = 1.06
)

Source

This function was adapted from a code developed by Yunke Peng (yunke.peng@usys.ethz.ch) - ETH Zürich: https://github.com/yunkepeng/gwr

Arguments

.ref

Reference data from which the data will be interpolated (see the details section).

...

arguments to be passed to other functions

.tar

Table with geographical target data, including: latitude, longitude and elevation.

varid

String with the identifier of the main variable inside the NetCDF file pointed by .ref (if applicable).

coordinates

Reference data set with columns for latitude, longitude and elevation. Default: CRU_coords.

res

Numeric value for the mask resolution. Default: 0.5 degrees.

xy_buffer

Numeric value to be used as the boundary for the search area in the x and y axes.

• latitude < .tar$latitude + xy_buffer

• latitude > .tar$latitude - xy_buffer

• longitude < .tar$longitude + xy_buffer

• longitude > .tar$longitude - xy_buffer

z_buffer

Numeric value to be used as the boundary for the search area in the z axis:

• elevation <= .tar$elevation * z_buffer

• elevation >= .tar$elevation / z_buffer

cpus

Number of CPUs to be used in parallel, default = 1.

bandwidth

bandwidth used in the weighting function, possibly calculated by gwr.sel

Value

Table with interpolated values from the .ref data for each record/row in .tar.

Details

The input reference data can be in any of the following formats:

• Matrix: this should be a 3-dimensional object with spatial components (latitude and longitude) and a temporal component for representing each time step to be used for the extraction of the data.

• String: this should point to a valid path on disk where the reference NetCDF file is stored. Note that the parameter called varid should be used to indicate the identifier of the main variable inside the NetCDF file (e.g., "tmp", "pre", "cld", etc.).

References

Peng, Y., Bloomfield, K.J. and Prentice, I.C., 2020. A theory of plant function helps to explain leaf‐trait and productivity responses to elevation. New Phytologist, 226(5), pp.1274-1284. doi:10.1111/nph.16447

Examples

if (FALSE) {
`%>%` <- magrittr::`%>%`
data <- tibble::tibble(entity_name = "University of Reading",
                       latitude = 51.44140,
                       longitude = -0.9418,
                       elevation = c(61, 161, 261, 361))
smpds::gwr(.ref = "/path/to/reference-tmp.nc",
           .tar = data,
           varid = "tmp")

ncin <- ncdf4::nc_open("/path/to/reference-tmp.nc")
reference_data <- ncdf4::ncvar_get(ncin, varid)
ncdf4::nc_close(ncin)
reference_data %>%
  smpds::gwr(.tar = data)
}