added accuracy assesment scripts

scripts/create_loss_layer.R

@@ -0,0 +1,149 @@
+library(terra)
+library(dplyr)
+library(tidyr)
+
+
+######################### data setup ######################### 
+lt_data_dir <- file.path("rasters/lt_nbr_ftv_1985_2021_combo6")
+lt_name <- "lt_nbr_1985_2021_combo6" 
+ftv_file_name <- "lt_nbr_ftv_1985_2021_combo6" 
+vertex_file_name <- "lt_nbr_is_vertex_1985_2021_combo6" 
+dir.create(paste0(lt_data_dir, "/ftv_tiles"))
+dir.create(paste0(lt_data_dir, "/vertex_tiles"))
+dir.create(paste0(lt_data_dir, "/masked_ftv"))
+dir.create(paste0(lt_data_dir, "/losses"))
+
+######################### merge tiles function ######################### 
+#ftv stacks from LT are often very large, and as such are split into several tiles. This function will stitch those tiles back together
+#if you have a single ftv file, you can skip this step
+#requires gdal installation
+merge_tiles <- function(source_dir, dest_file, pattern = ".*tif") {
+  vrt_file <- gsub("(tiff|tif)", "vrt", dest_file)
+
+  files <- list.files(source_dir, full.names = TRUE, pattern = pattern)
+  index_file <- file.path(source_dir, "files.txt")
+  writeLines(files, index_file)
+  system(
+    glue::glue("gdalbuildvrt {vrt_file} -input_file_list {index_file}")
+  )
+  system(
+    glue::glue(
+      paste(
+        "gdal_translate -co COMPRESS=DEFLATE -co PREDICTOR=3 -co TILED=YES",
+        "-co BIGTIFF=YES {vrt_file} {dest_file}"
+      )
+    )
+  )
+  unlink(index_file)
+  unlink(vrt_file)
+  return(dest_file)
+}
+
+merge_tiles(lt_data_dir, paste0(lt_data_dir, "/", ftv_file_name, "_all.tif"), pattern = "lt_nbr_ftv.*tif" )
+
+
+######################### retile ftv and vertex layers ######################### 
+# load merged ftv and vertex layers
+ftv <- terra::rast(paste0(lt_data_dir, "/", ftv_file_name, "_all.tif"))
+names(ftv) <- 1985:2022
+is_vertex <- terra::rast(paste0(lt_data_dir, "/", vertex_file_name, ".tif"))
+names(is_vertex) <- 1985:2022
+
+
+#analysis requires much smaller tiles, can adjust size of tiles based on processing ability of your computer
+tile_grid <- terra::rast(
+  extent = terra::ext(ftv),
+  ncols = 30, 
+  nrows = 30
+)
+
+terra::makeTiles(
+  ftv,
+  tile_grid,
+  path.expand(paste0(lt_data_dir, '/ftv_tiles/ftv_', ".tif")),
+  overwrite = TRUE
+)
+
+terra::makeTiles(
+  is_vertex,
+  tile_grid,
+  path.expand(paste0(lt_data_dir, '/vertex_tiles/is_vertex_', ".tif")),
+  overwrite = TRUE
+)
+
+######################### mask smoothed ftv traj with vertices  ######################### 
+#removing pixels that are not breakpoints in spectral trajectory
+vertex_files <- list.files(paste0(lt_data_dir,"/vertex_tiles/"), pattern = "is_vertex_.*.tif", full.names = TRUE)
+ftv_files <- list.files(paste0(lt_data_dir,"/ftv_tiles/"), pattern = "ftv_.*.tif", full.names = TRUE)
+
+lapply(1:length(ftv_files), \(i) {
+  message(paste0("Working on iteration ", i))
+  vertex_file <- vertex_files[i]
+  ftv_file <- ftv_files[i]
+  index <- stringr::str_split(gsub(pattern = "\\.tif", "", basename(vertex_file)), "_")[[1]] |> tail(1)
+  masked_ftv <- file.path(paste0(lt_data_dir,"/masked_ftv/masked_ftv_", index, ".tif"))
+
+  if (!file.exists(masked_ftv)){
+    terra::mask(terra::rast(ftv_file), terra::rast(vertex_file), maskvalues = 0, filename = masked_ftv)
+  } 
+})
+
+######################### create all loss layer  ######################### 
+#using masked ftv to identify decreasing segments
+#value of previous vertex greater than value of current vertex
+
+detect_loss <- function(model, data, ...){
+  # Collect these for later
+  years <- names(data)
+
+  # Build a place holder data.frame
+  results_container <- data.frame(
+    matrix(nrow = nrow(data), ncol = length(years))
+  )
+  names(results_container) <- years
+
+  # Compute magnitudes for loss vertices, with > 1 segment
+  data <- data |> 
+    mutate(id = 1:nrow(data)) |> # keep track of pixel order
+    pivot_longer(-id, names_to = "year") |>
+    mutate(year = as.numeric(year) + 1) |> # LT defaults adds a year to the vertex date to assign year of disturbance
+    group_by(id) |>
+    filter(!is.na(value)) |>
+    filter(n() > 2) |>
+    mutate(change = value - lead(value)) |>
+    filter(change > 0) |> 
+    select(-value) |>
+    ungroup() |>
+    pivot_wider(names_from = "year", values_from = "change")
+
+  # Pad the results, filling in the years that don't have disturbances
+  fill_years <- setdiff(years, names(data))
+  fill_cols <- lapply(fill_years, \(fy) {
+    return(rep(NA, nrow(data)))
+  })
+  names(fill_cols) <- fill_years
+  data <- bind_cols(data, fill_cols) |>
+    dplyr::relocate(all_of(years))
+
+  # Insert results in the correct location
+  results_container[data$id, ] <- data |> select(-id)
+  return(results_container)
+}
+
+#use loss funtction to create loss tiles
+masked_ftvs <- list.files(paste0(lt_data_dir, "/masked_ftv/"), full.names = TRUE)
+lapply(masked_ftvs, \(masked_file) {
+  index <- stringr::str_split(gsub(pattern = "\\.tif", "", basename(masked_file)), "_")[[1]] |> tail(1)
+  losses <- file.path(paste0(lt_data_dir,"/losses/losses", index , ".tif")) 
+
+  if (!file.exists(losses)){
+    terra::predict(terra::rast(masked_file), list(), detect_loss, filename = losses)
+  } 
+})
+
+#merge loss tiles into new loss.tif 
+merge_tiles(paste0(lt_data_dir, "/losses"), paste0(lt_data_dir,"/losses/", lt_name,"_all_losses.tif"), pattern = "*.tif" )
+all_losses <- rast("rasters/lt_nbr_rt1_1985-2022/losses/rt1_1985_2022_all_losses.tif")
+names(all_losses) <- 1985:2022
+all_losses <- subset(all_losses, 6:38) #remove pre 1990 years, optional step
+writeRaster(all_losses, paste0(lt_data_dir,"/losses/", lt_name,"_all_losses.tif"), overwrite = TRUE)

scripts/harvest_brick.R

@@ -0,0 +1,59 @@
+library(terra)
+######################## set up######################## 
+years <- 1990:2021 #study years
+harvest_directory <- file.path("../GIS_Files/harvest_rasters")
+dir.create(paste0(harvest_directory, "/harv_layers"))
+harv_records <- vect("../GIS_Files/albers_combined_harv_records.gpkg")
+#need study area bounds
+harv_boundaries <- vect("../GIS_Files/combined_tract_boundaries.gpkg")
+
+crs <- labrador.client::get_cafri_crs()
+
+
+######################## rasterize harvests  ######################## 
+harv_records$value <- 1 #dummy value for creating harvest raster
+
+#use your LT output as reference raster here, want harvest layer to have same extent
+ref_raster <- rast("rasters/combo12_all_losses2.tif")[[1]]
+values(ref_raster) <- NA
+
+lapply(1:length(years), \(i) {
+  message(paste0("Working on iteration ", i))
+
+  harv_subset <- terra::subset(harv_records,
+                               (years[i] >= harv_records$start_date & years[i] <= harv_records$end_date)
+                               | as.integer(harv_records$start_date) + 1 == years[i]
+                               | as.integer(harv_records$end_date) + 1== years[i]
+                               | as.integer(harv_records$start_date) - 1 == years [i]
+                               | as.integer(harv_records$end_date) - 1 == years[i] )
+
+  harv_file <- file.path(paste0(harvest_directory, "/harv_layers/harv_layer_", years[i], ".tif"))
+  harv_exists <- file.exists(harv_file)
+
+  if (!harv_exists & nrow(harv_subset) != 0){
+    terra::rasterize(harv_subset, ref_raster, field = "value", filename = harv_file)
+  } else if (!harv_exists){
+    writeRaster(ref_raster, filename = harv_file)
+  }
+})
+
+######################## combine havrvest rasters into single harvest brick layer ######################## 
+rasters <- list.files(paste0(harvest_directory, "/harv_layers/"), full.names = TRUE)
+harv_raster <- rast(rasters)
+names(harv_raster) <- years
+harv_raster[is.na(harv_raster)] <- 0
+writeRaster(harv_raster, paste0(harvest_directory,"/harvest_brick.tif"), overwrite = TRUE)
+harv_raster <- rast(paste0(harvest_directory,"/harvest_brick.tif"))
+
+######################## mask harvest brick with study area boundaries ######################## 
+#terra throws a fit if the crs info doesn't match exactly, just making the labels match here, its the same crs
+crs(harvest_rast) <- crs
+
+#dummy value for making raster
+harv_boundaries$rast_value <- 1
+harv_boundaries <- rasterize(harv_boundaries, harv_raster, field = "rast_value")
+#areas outside harvests set to NA
+harv_raster <- mask(harv_raster, harv_boundaries) 
+
+writeRaster(harv_raster, paste0(harvest_directory,"/harvest_brick.tif"), overwrite = TRUE)
+rast