a helper function for making predictions

[btupper]

Hi @solanoalex @SeascapeScience @tyshawnrene @vameringer @taragorman @DanielSweeney25 @zane-chutkow @tenzytaec26 @J-Kapl @ArianaRaschidFarrokhi ,

If you are encountering issues trying to make all of your predictions, consider using the function below. Given an approach, scenario and year it will iterate through the months generating (and saving) habitat suitability predictions. When running it you may encounter one or more warnings similar to this one...

Error in read_workflow(version) : 
  file not found:~/Dropbox/code/projects/ColbyForecasting_data/2025/ben/models/g_Dec_workflow.rds

It's just a warning, and the iterator will continue on to the next month. The outputs are saved in the predictions folder with names like g_Aug_PRESENT_PRESENT.tif, and g_Aug_RCP45_2055.tif. You can read them back in using something like this pred = read_stars(data_path("predictions", "g_Aug_PRESENT_PRESENT.tif"))

#' Make a prediction (for a year) given the sampling approach, scenario and year.
#' 
#' @param approach chr, one of "greedy" or "conservative"
#' @param scenario chr one of "PRESENT", "RCP45" or "RCP85"
#' @param year chr one of "2055" or "2075", ignored if `scenario` is "PRESENT"
#' @return stars predicted object
make_predictions = function(approach = "greedy",
                            scenario = "PRESENT",
                            year = "2055",
                            path = data_path("predictions")){
  
  if(FALSE){
    approach = "greedy"
    scenario = "PRESENT"
    year = "2055"
    path = data_path("predictions")
  }
  
  # make sure the path exists
  ok = make_path(path)
  
  #make sure the user passed correctly formed args
  scene = toupper(scenario[1])
  yr = if (scene == "PRESENT") {
    "PRESENT"
    } else {
      as.character(year[1])
    }
  DB = brickman_database()
  if (scene == "PRESENT"){
    DB = DB |>
      filter(scenario == scene, interval == "mon")
  } else {
    DB = DB |>
      filter(scenario == scene, year == yr, interval == "mon")
  }
  
  COVARS = read_brickman(DB)
  
  preds = sapply(month.abb,
    function(mon_name){
      version = sprintf("%s_%s", substr(approach, 1,1), mon_name)
      cfg = try(read_configuration(version))
      if (inherits(cfg, "try-error")){
        return(NULL)
      }
      
      wflow = try(read_workflow(version))
      if (inherits(wflow, "try-error")){
        return(NULL)
      }
      
      covars = COVARS |>
        select(cfg$keep_vars) |>
        slice("month", mon_name)
      
      predfile = sprintf("%s_%s_%s.tif", version, scene, yr)
      pred = predict_stars(wflow, covars) |>
        write_prediction(file.path(path, predfile))
      return(pred)
  }, simplify = FALSE)
  
  ix = sapply(preds, is.null)
  if (all(!ix)){
    return(NULL)
  }
  preds = preds[!ix]
  nms = names(preds)
  
  p = do.call(c, append(preds, list(along = list(month = nms))))
  return(p)
}