process_preds.Rmd

---
title: "Code to process predictors for urban niche model"
author: "Mikko Jimenez, Josh Carrell, and Caitlin Mothes"
date: "'r Sys.Date()'"
output: html_notebook
---

## Process predictors

As part of a broader pipeline for running a niche model in urban settings, this script is designed to provide a flexible process for reading in predictor variables, setting them to a consistent coordinate reference system, and cropping them to an extent of interest. 

Note: it is currently unclear if predictors will be expected to be downloaded prior to running this script or if we will be pulling predictors directly from an API. I've written code for each situation below. 

## set up environment
```{r}
source("setup.R") 

# read in spatial layer that includes full extent of interest, in this case a boundary .shp
aoi <- read_sf("data/nature_in_the_city/gis/ft_collins_GMA_boundary.shp")

# list of paths to predictors (shapefiles, .tifs, etc.)
path_dist <- c("data/input_raw/Hydrology/Hydrology.shp",
               "data/input_raw/Buildings/Buildings.shp",
               "data/nature_in_the_city/gis/ft_collins_GMA_boundary.shp")
path_nodist <- c("data/input_raw/Impervious_Surface/impervious_surfaces_fort_collins_2015.shp",
                 "data/input_raw/Natural_Habitat/bc8f7f07-7560-4191-aa2e-8f274a1c29e7202046-1-133oxp5.v8wfi.shp",
                 "data/input_raw/Buildings/Buildings.shp",
                 "data/input_raw/Green_Space/nic_green_space_2015.shp")

# set desired resolution
resolution <- 100

# set output directory
output_path <- "data/input_processed"
```

## write function that reads in predictors and matches CRS and extent to the aoi
```{r}
# write function that reads in predictors and matches CRS and extent to the aoi
process_preds <- function(predictor_path, aoi, resolution, distance=FALSE, save, output_path) {
  
  # Initialize variables to hold the processed objects
  rast_shapefile <- NULL
  raster_file <- NULL
  output_object <- NULL

  if (grepl("\\.shp$", predictor_path, ignore.case = TRUE)) {
    # Read in shapefile
    shapefile <- st_read(predictor_path)

    # Set CRS to match the aoi
    shapefile <- st_transform(shapefile, crs = crs(aoi))

    # Set extent to match the aoi (crop if necessary)
    shapefile <- st_crop(shapefile, st_bbox(aoi))
    
    ## Rasterize the shapefile
    # Create an empty raster aoi with the specified resolution
    raster_aoi <- rast(ext(aoi), resolution = resolution)
    # Rasterize shapefile 
    rast_shapefile <- terra::rasterize(shapefile, raster_aoi, fun = mean)
    # Set CRS to match the shapefile
    crs(rast_shapefile) <- st_crs(shapefile)$proj4string
    
    if(distance == TRUE) {
      rast_shapefile <- terra::distance(rast_shapefile)
    }

    # Extract filename without extension for naming 
    file_name <- tools::file_path_sans_ext(basename(predictor_path))

    # Store the processed object with the filename
    output_object <- rast_shapefile
    
  } else if (grepl("\\.tif$", predictor_path, ignore.case = TRUE)) {
    # Read in raster and aoi files
    raster_file <- rast(predictor_path)

    # Set aoi temporarily to raster CRS - so we don't need to project full, large rasters
    shapefile_temp <- st_transform(aoi, crs = crs(raster_file))
    
    # crop to the temp shapefile
    raster_file <- crop(raster_file, shapefile_temp)
    
    # Convert the aoi to a raster with the same extent and resolution
    aoi_raster <- rast(ext(aoi), resolution = resolution, crs = crs(aoi))
  
    # Project cropped raster into aoi CRS
    raster_file <- project(raster_file, aoi_raster)
    
    # Align the raster to the aoi's resolution and extent
    # raster_file <- resample(raster_file, aoi_raster, method = "bilinear")
    
    if(distance == TRUE) {
      raster_file <- terra::distance(raster_file)
    }
    
    # Extract filename without extension for naming
    file_name <- tools::file_path_sans_ext(basename(predictor_path))
    
    # Store the processed object with the filename
    output_object <- raster_file
    
  } else {
    stop("Unsupported file type.")
  }
  
  # Save the processed objects to the specified directory if 'save' is TRUE
  if (save) {
    if (!dir.exists(output_path)) {
      dir.create(output_path, recursive = TRUE)
    }
    
    if (!is.null(rast_shapefile)) {
      # Write raster file
      writeRaster(rast_shapefile, filename = file.path(output_path, paste0(file_name, ".tif")), overwrite = TRUE)
    }
    
    if (!is.null(raster_file)) {
      # Write raster file
      writeRaster(raster_file, filename = file.path(output_path, paste0(file_name, ".tif")), overwrite = TRUE)
    
    }
  }
  
    # Assign the output_object to the global environment with the filename as the variable name
    #assign(file_name, output_object, envir = .GlobalEnv)
    
    # Return the processed object with the filename
    return(output_object)
} 
```

# process predictors in parallel 
```{r}
# process distance predictors
pred_dist <- purrr::map(path_dist, process_preds, aoi, resolution, distance=TRUE, save=FALSE, output_path) %>% set_names(tools::file_path_sans_ext(basename(path_dist)))

# process non distance predictors
pred_nodist <- purrr::map(path_nodist, process_preds, aoi, resolution, distance=FALSE, save=FALSE, output_path) %>% set_names(tools::file_path_sans_ext(basename(path_nodist)))
```
# stack all the rasters and save
```{r}
# stack rasters
combined_rast <- terra::rast(c(pred_dist, pred_nodist))

# Save the predictor raster stack
writeRaster(combined_rast, filename = file.path(output_path, "pred_stack.tif"), overwrite = TRUE)
```