Model prediction

Koen Hufkens

2026-03-31

Read in the reference data from an independent run of the original code. The kernel bit is deterministic if given a set of parameters.

reference <- readr::read_csv(here("data-raw/validation/objective_function_detail.csv"))
reference$likelihood[reference$likelihood == -9999] <- NA

Set the coefficients (using config_best_Mmem_fitting.txt) as used in the above reference run. Config values are provided as a nested list, with the last values the resource selection coefficients.

params <- list(
  r_l = 27.5332236990522,
  w_l = 0,
  r_d = 0.018257876686841,
  w_d = 0.9999,
  r_dist = 0.0412536435305482,
  w_dist = 0.9999,
  step_length_dist = 0.00216275705935606,
  step_length_shape = 1.14267311221975,
  threshold_approx_kernel = 7000,
  threshold_memory_kernel = 1000,

  # resource selection coefficients should be
  # a named list for driver data layer validation
  # and correct data processing
  coef = c(
    "slope" = 0.272835968106296,
    "slope_sq" = -0.093687792157105,
    "tcd_325grain"= 0.177991482087775,
    "tcd_325grain_sq" = -0.140639949444926,
    "landcover_5322" = 0.591063382485486,
    "landcover_agri" = -0.811974081226742
  )
)

Using GDAL and the terra package load all raster layers (*.asc) files.

# load raster stack
r <- terra::rast(list.files(here("data-raw/drivers/"),"*.asc", full.names = TRUE))
plot(r)

r <- subset(r, names(params$coef))
plot(r)

r <- as.array(r)

Pass everything to the hr_predict() function. I set optimization to TRUE this will run the kernel method and should result in the same output as the original code (given the above parameters).

output <- hr_predict(
  data = r,
  par = params,
  obs = here("data-raw/Aspromonte_roedeer_traj.txt"),
  resolution = 25,
  optimization = TRUE,
  verbose = TRUE
)

Plot a 1:1 figure to confirm this.

plot(output$likelihood, reference$likelihood)
abline(0,1)

Now, run a true prediction (simulation) run with the fit model parameters by setting the optimization parameter to FALSE (the default).

output <- hr_predict(
  data = r,
  par = params,
  obs = here("data-raw/tracks/Aspromonte_roedeer_traj.txt"),
  resolution = 25,
  steps = 10,
  runs = 10,
  optimization = FALSE,
  verbose = TRUE
)

This returns a hr_predict class for output, which can return a summary plot through a method for quick review.

plot(output)