data-raw/dsmdata.R

In inlabru-org/inlabru: Bayesian Latent Gaussian Modelling using INLA and Extensions

# Import dsm data
#
# Converts a dsm style data set into a \link{dsdata} object.
#
# @aliases import.dsmdata
# @export
# @param covar.col Column of the original data set to extract covariate information from
# @return a \link{dsdata} object
# @examples \dontrun{ library(dsm) ; data(mexdolphins); dset = import.dsdata(mexdolphins, covar.col = 8) }
# @author Fabian E. Bachl <\email{f.e.bachl@@bath.ac.uk}>

import.dsmdata <- function(dsmdata, covar.col = NA) {
  # Extract data
  segdata <- dsmdata$segdata
  distdata <- dsmdata$distdata
  obsdata <- dsmdata$obsdata
  preddata <- dsmdata$preddata

  # 1. DEFINE BLOCKS
  # Take account of depth in defining blocks - Block.Label is identifier
  # More than one covariate can be listed
  # If no covariates, then covar.col=NA and transects are used (to be done)
  segdata <- define.blocks.f(seg = segdata, covar.col = covar.col, geometry = "euc")
  # segdata[1:2, ]

  # 2. GET THE ANGLE OF DIRECTION FOR EACH SEGMENT
  segdata <- get.direction.segment.f(data = segdata, geometry = "euc")
  # segdata[1:2, ]

  # 3. GET THE START AND END POINTS OF ALL SEGMENTS
  # NOTE: x and y get renamed to mid.x and mid.y to avoid confusion with detection coordinates
  segdata <- start_end_points_segments_f(seg = segdata, use.tran = FALSE, geometry = "euc")
  # segdata[1:2, ]

  # 4. AMALGAMATE SEGMENTS INTO NEW BLOCKS
  blocks <- get.blocks.f(seg = segdata, geometry = "euc")
  # blocks[1:2, ]

  # 5. ADD SEGMENT AND BLOCK LABELS TO DETECTIONS
  # so that detections and blocks data can be combined
  distdata <- add.labels.to.obs.f(dists = distdata, obs = obsdata, seg = segdata)
  # distdata[1:2, ]


  # Generate location for detections if location missing
  # Use do.plot=T to see the location generated for each detection
  if (!("x" %in% colnames(distdata))) {
    det.new.coords <- generate_obs_location_f(
      seg = segdata,
      dists = distdata,
      geometry = "euc",
      do.plot = FALSE
    )
    # det.new.coords[1:3, ]
    distdata <- cbind(distdata, det.new.coords)
  }

  newdata <- combine.dsmdata.f(blocks = blocks, dists = distdata)
  # Only one strata for this data so need to add this information
  newdata$strat <- 1
  # newdata[1:3, ]



  # Automated mesh construction
  loc <- rbind(
    as.matrix(segdata[, c("start.x", "start.y")]),
    as.matrix(segdata[, c("end.x", "end.y")]),
    as.matrix(preddata[, c("x", "y")]),
    as.matrix(newdata[, c("mid.x", "mid.y")]),
    as.matrix(newdata[, c("x", "y")])
  )
  ok <- rowSums(is.na(loc)) == 0
  loc <- loc[ok, , drop = FALSE]
  inner <- fm_nonconvex_hull(
    loc,
    convex = min(diff(range(loc[, 1])), diff(range(loc[, 2]))) / 20,
  )
  outer <- fm_nonconvex_hull(
    loc,
    convex = min(diff(range(loc[, 1])), diff(range(loc[, 2]))) / 2,
  )
  mesh <- fm_mesh_2d_inla(
    boundary = list(inner, outer),
    max.edge = c(
      min(diff(range(loc[, 1])), diff(range(loc[, 2]))) / 10,
      min(diff(range(loc[, 1])), diff(range(loc[, 2]))) / 2
    ),
    cutoff = min(diff(range(loc[, 1])), diff(range(loc[, 2]))) / 20
  )

  dset <- list(effort = newdata, mesh = mesh)
  class(dset) <- c("dsdata", "list")
  return(dset)
}