MakeStacks.R

library(INLA)


# load functions for building INLA stacks
source("Functions/MakeSpatialRegion.R")
source("Functions/MakeIntegrationStack.R")
source("Functions/GetNearestCovariate.R")
source("Functions/MakeProjectionGrid.R")
source("Functions/MakeBinomStack.R")
source("Functions/MakePointsStack.R")

# load warbler data created in ExtractData.R
load("Data/BTWarblerData.RData")

# Create the mesh to approximate the area and the spatial field
Meshpars <- list(max.edge = c(0.05, 0.4), offset = c(0.1, 0.4), cutoff = 0.1)

Mesh <- MakeSpatialRegion(
  data = NULL,
  bdry = PA,
  meshpars = Meshpars,
  proj = proj
)

# Make stack for background mesh
stk.ip <- MakeIntegrationStack(
  mesh = Mesh$mesh,
  data = covariates_eBird,
  area = Mesh$w,
  tag = "ip",
  InclCoords = TRUE
)

# make data for projections

# This sets the resolution of the predictions. For the final version we use 
# Nxy.scale <- 0.01, but have changed it here so the code will run more quickly.
if(!exists("Nxy.scale")) Nxy.scale <- 0.1

Boundary <- Mesh$mesh$loc[Mesh$mesh$segm$int$idx[, 2], ]
Nxy.size <- c(diff(range(Boundary[, 1])), diff(range(Boundary[, 2])))
Nxy <- round(Nxy.size / Nxy.scale)

# Make stack for projections
stk.pred <- MakeProjectionGrid(
  nxy = Nxy,
  mesh = Mesh$mesh,
  data = covariates,
  tag = "pred",
  boundary = Boundary
)

# Create data stacks for each of the different data types
stk.BBS <- MakeBinomStack(
  observs = BBS_sp,
  data = covariates,
  mesh = Mesh$mesh,
  presname = "NPres",
  trialname = "Ntrials",
  tag = "BBS",
  InclCoords = TRUE
)

stk.eBird <- MakePointsStack(
  presences = eBird_sp,
  data = covariates_eBird,
  mesh = Mesh$mesh,
  tag = "eBird",
  InclCoords = TRUE
)

stk.BBA <- MakeBinomStack(
  observs = BBA_sp,
  data = covariates,
  mesh = Mesh$mesh,
  presname = "present",
  tag = "BBA",
  InclCoords = TRUE
)

save(
  stk.ip, stk.pred, stk.eBird, stk.BBS, stk.BBA, Mesh,
  file = "Data/Stacks.RData"
)