fillNA {AMAPVox}

Fill missing values (NA) with averaged neighboring data

Description

Fill missing values of a given variable in a VoxelSpace object with averaged neighboring values.

Neighboring values are selected among voxels within a user-defined radius in meter and whose sampling rate (number of pulses that went through the voxel) is above a user-defined threshold. Distance between voxels is the euclidian distance between voxel centers. Fill-value may be capped by user-defined minimal and maximal values.

Default radius (if not defined by user) is set to largest dimension of voxel size max(getVoxelSize(vxsp)). It guarantees that default neighborhood is isotropic.

In some cases, for instance poorly sampled area, neighboring values may all be missing or discarded. A fallback value can be provided to "force fill" suche voxels. An other option is to run again the function with larger radius or lower sampling threshold.

Usage

fillNA(
  vxsp,
  variable.name,
  variable.min = -Inf,
  variable.max = Inf,
  variable.fallback,
  radius,
  pulse.min = 10
)

Arguments

vxsp

a VoxelSpace object.

variable.name

a character, the name of a variable in the VoxelSpace

variable.min

a numeric, minimal value for the fill values

variable.max

a numeric, maximal value for the fill values

variable.fallback

a numeric, optional fallback value in case no fill value can be estimated from neighboring voxels.

radius

a numeric, the radius in meter that defines the neighborhood of a voxel. The function looks for the voxels whose center is inside a sphere of radius radius centered at current voxel center. Default is set to max(getVoxelSize(vxsp))

pulse.min

a numeric, minimal sampling intensity (i.e. number of pulses that went through a voxel) to include neighboring voxel in the estimation of the averaged fill value.

Examples

# read voxel space
vxsp <- readVoxelSpace(system.file("extdata", "tls_sample.vox", package = "AMAPVox"))
# Randomly add some NA in PAD variable
vx <- vxsp@data
ind <- sample(vx[PadBVTotal > 0, which = TRUE], 3)
# print initial values
vx[ind, .(i, j, k, PadBVTotal)]
vx[ind, PadBVTotal := NA]
# fill NA in PAD variable
fillNA(vxsp, "PadBVTotal", variable.max = 5)
# print filled values
vx[ind, .(i, j, k, PadBVTotal)]