Merge branch 'feat/global-landcover' into dev

gabrieldansereau · Feb 14, 2020 · 9d4ee02 · 9d4ee02
2 parents 4f2eae5 + 6679b1a
commit 9d4ee02
Show file tree

Hide file tree

Showing 70 changed files with 128 additions and 99 deletions.
diff --git a/.gitignore b/.gitignore
@@ -1 +1,2 @@
 tags
+assets/landcover/landcover_copernicus_global_100m*
diff --git a/assets/landcover/lc_bare-5m.tif b/assets/landcover/lc_bare-5m.tif
diff --git a/assets/landcover/lc_bare_10m.tif b/assets/landcover/lc_bare_10m.tif
diff --git a/assets/landcover/lc_bare_5m.tif b/assets/landcover/lc_bare_5m.tif
diff --git a/assets/landcover/lc_crops-5m.tif b/assets/landcover/lc_crops-5m.tif
diff --git a/assets/landcover/lc_crops_10m.tif b/assets/landcover/lc_crops_10m.tif
diff --git a/assets/landcover/lc_crops_5m.tif b/assets/landcover/lc_crops_5m.tif
diff --git a/assets/landcover/lc_grass-5m.tif b/assets/landcover/lc_grass-5m.tif
diff --git a/assets/landcover/lc_grass_10m.tif b/assets/landcover/lc_grass_10m.tif
diff --git a/assets/landcover/lc_grass_5m.tif b/assets/landcover/lc_grass_5m.tif
diff --git a/assets/landcover/lc_moss-5m.tif b/assets/landcover/lc_moss-5m.tif
diff --git a/assets/landcover/lc_moss_10m.tif b/assets/landcover/lc_moss_10m.tif
diff --git a/assets/landcover/lc_moss_5m.tif b/assets/landcover/lc_moss_5m.tif
diff --git a/assets/landcover/lc_shrub-5m.tif b/assets/landcover/lc_shrub-5m.tif
diff --git a/assets/landcover/lc_shrub_10m.tif b/assets/landcover/lc_shrub_10m.tif
diff --git a/assets/landcover/lc_shrub_5m.tif b/assets/landcover/lc_shrub_5m.tif
diff --git a/assets/landcover/lc_snow-5m.tif b/assets/landcover/lc_snow-5m.tif
diff --git a/assets/landcover/lc_snow_10m.tif b/assets/landcover/lc_snow_10m.tif
diff --git a/assets/landcover/lc_snow_5m.tif b/assets/landcover/lc_snow_5m.tif
diff --git a/assets/landcover/lc_tree-5m.tif b/assets/landcover/lc_tree-5m.tif
diff --git a/assets/landcover/lc_tree_10m.tif b/assets/landcover/lc_tree_10m.tif
diff --git a/assets/landcover/lc_tree_5m.tif b/assets/landcover/lc_tree_5m.tif
diff --git a/assets/landcover/lc_urban-5m.tif b/assets/landcover/lc_urban-5m.tif
diff --git a/assets/landcover/lc_urban_10m.tif b/assets/landcover/lc_urban_10m.tif
diff --git a/assets/landcover/lc_urban_5m.tif b/assets/landcover/lc_urban_5m.tif
diff --git a/assets/landcover/lc_water-permanent-5m.tif b/assets/landcover/lc_water-permanent-5m.tif
diff --git a/assets/landcover/lc_water-permanent_10m.tif b/assets/landcover/lc_water-permanent_10m.tif
diff --git a/assets/landcover/lc_water-permanent_5m.tif b/assets/landcover/lc_water-permanent_5m.tif
diff --git a/assets/landcover/lc_water-seasonal-5m.tif b/assets/landcover/lc_water-seasonal-5m.tif
diff --git a/assets/landcover/lc_water-seasonal_10m.tif b/assets/landcover/lc_water-seasonal_10m.tif
diff --git a/assets/landcover/lc_water-seasonal_5m.tif b/assets/landcover/lc_water-seasonal_5m.tif
diff --git a/fig/06_relationship_lcbd-richness-transf.png b/fig/06_relationship_lcbd-richness-transf.png
diff --git a/fig/06_relationship_lcbd-richness.png b/fig/06_relationship_lcbd-richness.png
diff --git a/fig/quantiles/03_raw_richness_quantiles.pdf b/fig/quantiles/03_raw_richness_quantiles.pdf
diff --git a/fig/quantiles/03_sdm_richness_quantiles.pdf b/fig/quantiles/03_sdm_richness_quantiles.pdf
diff --git a/fig/quantiles/05_raw_lcbd-transf_quantiles.pdf b/fig/quantiles/05_raw_lcbd-transf_quantiles.pdf
diff --git a/fig/quantiles/05_raw_lcbd_quantiles.pdf b/fig/quantiles/05_raw_lcbd_quantiles.pdf
diff --git a/fig/quantiles/05_sdm_lcbd-transf_quantiles.pdf b/fig/quantiles/05_sdm_lcbd-transf_quantiles.pdf
diff --git a/fig/quantiles/05_sdm_lcbd_quantiles.pdf b/fig/quantiles/05_sdm_lcbd_quantiles.pdf
diff --git a/fig/quantiles/06_relationship_lcbd-richness-dbtransf-nbsp_quantiles.png b/fig/quantiles/06_relationship_lcbd-richness-dbtransf-nbsp_quantiles.png
diff --git a/fig/quantiles/06_relationship_lcbd-richness-transf_quantiles.png b/fig/quantiles/06_relationship_lcbd-richness-transf_quantiles.png
diff --git a/fig/quantiles/07_raw_endemism_quantiles.pdf b/fig/quantiles/07_raw_endemism_quantiles.pdf
diff --git a/fig/quantiles/07_sdm_endemism_quantiles.pdf b/fig/quantiles/07_sdm_endemism_quantiles.pdf
diff --git a/fig/raw/01_raw_sp-Setophaga_petechia.pdf b/fig/raw/01_raw_sp-Setophaga_petechia.pdf
diff --git a/fig/raw/01_raw_sp-Setophaga_townsendi.pdf b/fig/raw/01_raw_sp-Setophaga_townsendi.pdf
diff --git a/fig/raw/02_raw_Y-rowcolsorted.png b/fig/raw/02_raw_Y-rowcolsorted.png
diff --git a/fig/raw/03_raw_richness.pdf b/fig/raw/03_raw_richness.pdf
diff --git a/fig/raw/04_raw_diversity-pielou.pdf b/fig/raw/04_raw_diversity-pielou.pdf
diff --git a/fig/raw/04_raw_diversity-pielou2.pdf b/fig/raw/04_raw_diversity-pielou2.pdf
diff --git a/fig/raw/05_raw_lcbd-transf.pdf b/fig/raw/05_raw_lcbd-transf.pdf
diff --git a/fig/raw/05_raw_lcbd.pdf b/fig/raw/05_raw_lcbd.pdf
diff --git a/fig/raw/07_raw_endemism.pdf b/fig/raw/07_raw_endemism.pdf
diff --git a/fig/sdm/01_sdm_sp-Setophaga_petechia.pdf b/fig/sdm/01_sdm_sp-Setophaga_petechia.pdf
diff --git a/fig/sdm/01_sdm_sp-Setophaga_townsendi.pdf b/fig/sdm/01_sdm_sp-Setophaga_townsendi.pdf
diff --git a/fig/sdm/02_sdm_Y-rowcolsorted.png b/fig/sdm/02_sdm_Y-rowcolsorted.png
diff --git a/fig/sdm/03_sdm_richness.pdf b/fig/sdm/03_sdm_richness.pdf
diff --git a/fig/sdm/04_sdm_diversity-pielou.pdf b/fig/sdm/04_sdm_diversity-pielou.pdf
diff --git a/fig/sdm/04_sdm_diversity-pielou2.pdf b/fig/sdm/04_sdm_diversity-pielou2.pdf
diff --git a/fig/sdm/05_sdm_lcbd-transf.pdf b/fig/sdm/05_sdm_lcbd-transf.pdf
diff --git a/fig/sdm/05_sdm_lcbd.pdf b/fig/sdm/05_sdm_lcbd.pdf
diff --git a/fig/sdm/07_sdm_endemism.pdf b/fig/sdm/07_sdm_endemism.pdf
diff --git a/src/00b_data_landcover-copernicus.jl b/src/00b_data_landcover-copernicus.jl
@@ -5,52 +5,20 @@ using Distributed
 ## Conditional arguments
 # save_figures = true # should figures be overwritten (optional)
 
-## Bash commands to download & prepare data, to run in terminal in ../landcover/
-#=
-cd ~/github/landcover/
-# Download Copernicus land cover data, many downloads to cover whole extent
-wget https://s3-eu-west-1.amazonaws.com/vito-downloads/ZIPfiles/W160N80_ProbaV_LC100_epoch2015_global_v2.0.1_products_EPSG-4326.zip -O landcover_W160N80.zip
-wget https://s3-eu-west-1.amazonaws.com/vito-downloads/ZIPfiles/W160N60_ProbaV_LC100_epoch2015_global_v2.0.1_products_EPSG-4326.zip -O landcover_W160N60.zip
-wget https://s3-eu-west-1.amazonaws.com/vito-downloads/ZIPfiles/W160N40_ProbaV_LC100_epoch2015_global_v2.0.1_products_EPSG-4326.zip -O landcover_W160N40.zip
-wget https://s3-eu-west-1.amazonaws.com/vito-downloads/ZIPfiles/W140N80_ProbaV_LC100_epoch2015_global_v2.0.1_products_EPSG-4326.zip -O landcover_W140N80.zip
-wget https://s3-eu-west-1.amazonaws.com/vito-downloads/ZIPfiles/W140N60_ProbaV_LC100_epoch2015_global_v2.0.1_products_EPSG-4326.zip -O landcover_W140N60.zip
-wget https://s3-eu-west-1.amazonaws.com/vito-downloads/ZIPfiles/W140N40_ProbaV_LC100_epoch2015_global_v2.0.1_products_EPSG-4326.zip -O landcover_W140N40.zip
-wget https://s3-eu-west-1.amazonaws.com/vito-downloads/ZIPfiles/W120N80_ProbaV_LC100_epoch2015_global_v2.0.1_products_EPSG-4326.zip -O landcover_W120N80.zip
-wget https://s3-eu-west-1.amazonaws.com/vito-downloads/ZIPfiles/W120N60_ProbaV_LC100_epoch2015_global_v2.0.1_products_EPSG-4326.zip -O landcover_W120N60.zip
-wget https://s3-eu-west-1.amazonaws.com/vito-downloads/ZIPfiles/W120N40_ProbaV_LC100_epoch2015_global_v2.0.1_products_EPSG-4326.zip -O landcover_W120N40.zip
-wget https://s3-eu-west-1.amazonaws.com/vito-downloads/ZIPfiles/W100N80_ProbaV_LC100_epoch2015_global_v2.0.1_products_EPSG-4326.zip -O landcover_W100N80.zip
-wget https://s3-eu-west-1.amazonaws.com/vito-downloads/ZIPfiles/W100N60_ProbaV_LC100_epoch2015_global_v2.0.1_products_EPSG-4326.zip -O landcover_W100N60.zip
-wget https://s3-eu-west-1.amazonaws.com/vito-downloads/ZIPfiles/W100N40_ProbaV_LC100_epoch2015_global_v2.0.1_products_EPSG-4326.zip -O landcover_W100N40.zip
-wget https://s3-eu-west-1.amazonaws.com/vito-downloads/ZIPfiles/W080N80_ProbaV_LC100_epoch2015_global_v2.0.1_products_EPSG-4326.zip -O landcover_W080N80.zip
-wget https://s3-eu-west-1.amazonaws.com/vito-downloads/ZIPfiles/W080N60_ProbaV_LC100_epoch2015_global_v2.0.1_products_EPSG-4326.zip -O landcover_W080N60.zip
-wget https://s3-eu-west-1.amazonaws.com/vito-downloads/ZIPfiles/W080N40_ProbaV_LC100_epoch2015_global_v2.0.1_products_EPSG-4326.zip -O landcover_W080N40.zip
-wget https://s3-eu-west-1.amazonaws.com/vito-downloads/ZIPfiles/W060N80_ProbaV_LC100_epoch2015_global_v2.0.1_products_EPSG-4326.zip -O landcover_W060N80.zip
-wget https://s3-eu-west-1.amazonaws.com/vito-downloads/ZIPfiles/W060N60_ProbaV_LC100_epoch2015_global_v2.0.1_products_EPSG-4326.zip -O landcover_W060N60.zip
-
-# Unzip files in separate directories
-for i in *.zip; do unzip "$i" -d "${i%%.zip}"; done
-
-# Delete zip files
-rm *.zip
-
-# Create repositories if needed
-mkdir coverfraction
-(cd coverfraction; mkdir bare crops grass moss shrub snow tree urban water-permanent water-seasonal)
-
-# Batch commands for each land cover variable
-for i in $(ls coverfraction/)
-do
-    # Copy landcover data in 1 folder per variable
-    for j in landcover*; do cp "$j"/*"$i"-coverfraction-layer* coverfraction/"$i"/; done
-    # Set resolution to 10 arc-minutes & merge all layers in one
-    gdalwarp -tr 0.166667 0.166667 -r average coverfraction/"$i"/W*.tif ../BioClim/assets/landcover/lc_"$i"_10m.tif
-    # Set resolution to 5 arc-minutes & merge all layers in one
-    gdalwarp -tr 0.0833333 0.0833333 -r average coverfraction/"$i"/W*.tif ../BioClim/assets/landcover/lc_"$i"-5m.tif
-done
-=#
-
-##  Test landcover variables
+## Run bash scripts to download & coarsen landcover data from Zenodo (if files missing)
+lc_files = readdir("assets/landcover/")
+# Check if landcover files are missing
+if any(startswith.(lc_files, r"^lc_"))
+    # Check if full resolution files are missing
+    if any(startswith.(lc_files, r"^landcover_copernicus_global_100m"))
+        # Download full resolution files
+        run(`bash src/bin/landcover_download.sh`)
+    end
+    # Coarsen resolution
+    run(`bash src/bin/landcover_coarsen.sh`)
+end
 
+## Test landcover variables
 # Define coordinates range
 lon_range = (-145.0, -50.0)
 lat_range = (20.0, 75.0)

diff --git a/src/01_distributions.jl b/src/01_distributions.jl
@@ -40,19 +40,12 @@ lat_range_obs = extrema(df.latitude)
 
 ## Get environmental data (with different training resolutions)
 # WorldClim data
-@time wc_vars = map(x -> worldclim(x, resolution = "10")[lon_range, lat_range], [1,12]);
+wc_vars = map(x -> worldclim(x, resolution = "10")[lon_range, lat_range], [1,12]);
 # Landcover data
-@time lc_vars = pmap(x -> landcover(x, resolution = "10")[lon_range, lat_range], 1:10);
+lc_vars = map(x -> landcover(x, resolution = "10")[lon_range, lat_range], 1:10);
 # Training data with finer resolution
-if outcome == "raw"
-    # Set resolution to 10 # CAN'T BE FINER FOR RAW ANALYSES FOR NOW
-    @time wc_vars_train = map(x -> worldclim(x, resolution = "10")[lon_range_obs, lat_range_obs], [1,12]);
-    @time lc_vars_train = map(x -> landcover(x, resolution = "10")[lon_range, lat_range], 1:10)
-elseif outcome == "sdm"
-    # Set resolution to 5
-    @time wc_vars_train = map(x -> worldclim(x, resolution = "5")[lon_range_obs, lat_range_obs], [1,12]);
-    @time lc_vars_train = map(x -> landcover(x, resolution = "5")[lon_range, lat_range], 1:10)
-end
+wc_vars_train = map(x -> worldclim(x, resolution = "5")[lon_range_obs, lat_range_obs], [1,12]);
+lc_vars_train = map(x -> landcover(x, resolution = "5")[lon_range_obs, lat_range_obs], 1:10);
 
 # Combine environmental data
 env_vars = vcat(wc_vars, lc_vars)
@@ -70,7 +63,7 @@ if (@isdefined create_distributions) && create_distributions == true
         # @time distributions = @showprogress pmap(x -> presence_absence(x, env_vars_train[1], full_range = true, binary = false), warblers)
     elseif outcome == "sdm"
         # Get sdm distributions (with different training resolutions)
-        @time distributions = @showprogress pmap(x -> bioclim(x, env_vars, train_vars = env_vars_train), warblers);
+        @time distributions = @showprogress pmap(x -> bioclim(x, env_vars, training_layers = env_vars_train), warblers);
     end
 end
 

diff --git a/src/06_relationship_lcbd-richness.jl b/src/06_relationship_lcbd-richness.jl
@@ -69,15 +69,15 @@ relationtr_plot = scatter(vec(rel_richness[1]), vec(raw.LCBD[2].grid),
          grid=:none)
 scatter!(relationtr_plot, vec(rel_richness[2]), vec(sdm.LCBD[1].grid),
          markersize = 3, c = :orange, msw = 0, label = "SDM predictions")
-relationdbtr_plot = scatter(vec(rel_richness[1]), vec(raw.LCBD[3].grid),
+relationdbtr_plot = scatter(vec(rel_richness[1]), vec(raw.LCBD[2].grid),
          markersize = 2,
          c = :skyblue,
          msw = 0,
          label = "Raw occurrence data",
          legend = :topright,
          xlabel = "Species richness (\\alpha\\/\\gamma)", ylabel = "LCBD (relative to maximum)",
          grid=:none)
-scatter!(relationdbtr_plot, vec(rel_richness[2]), vec(sdm.LCBD[3].grid),
+scatter!(relationdbtr_plot, vec(rel_richness[2]), vec(sdm.LCBD[2].grid),
          markersize = 3, c = :orange, msw = 0, label = "SDM predictions")
 
 ## Save result

diff --git a/src/bin/landcover_coarsen.sh b/src/bin/landcover_coarsen.sh
@@ -0,0 +1,32 @@
+#!/bin/bash
+
+## Bash commands to download & prepare landcover data from Zenodo
+# Just run following in bash terminal:
+# bash src/bin/landcover_coarsen.sh
+
+# Make sure file is executable. If not (e.g. Permission denied error), run:
+# chmod +x src/bin/landcover_coarsen.sh
+
+####
+
+## Go to landcover directory
+cd assets/landcover/
+
+## Coarsen resolution
+# BEWARE, can be very long and will take 10 cores, took 30 min and ~ 16GB of RAM in my case
+landcover_variables=(bare crops grass moss shrub snow tree urban water-permanent water-seasonal)
+for i in "${landcover_variables[@]}"
+do
+    # Set resolution to 10 arc-minutes
+    gdalwarp -tr 0.166667 0.166667 -r average --config GDAL_CACHEMAX 500 -wm 500 -multi landcover_copernicus_global_100m_v2.0.2_"$i".tif lc_"$i"_10m.tif &
+done
+wait
+echo "10 arc-minutes - All done"
+
+for i in "${landcover_variables[@]}"
+do
+    # Set resolution to 5 arc-minutes
+    gdalwarp -tr 0.0833333 0.0833333 -r average --config GDAL_CACHEMAX 500 -wm 500 -multi landcover_copernicus_global_100m_v2.0.2_"$i".tif lc_"$i"_5m.tif &
+done
+wait
+echo "5 arc-minutes - All done"
diff --git a/src/bin/landcover_download.sh b/src/bin/landcover_download.sh
@@ -0,0 +1,25 @@
+#!/bin/bash
+
+## Bash commands to download & prepare landcover data from Zenodo
+# Just run following in bash terminal:
+# bash src/bin/landcover_download.sh
+
+# Make sure file is executable. If not (e.g. Permission denied error), run:
+# chmod +x src/bin/landcover_download.sh
+
+####
+
+## Go to landcover directory
+cd assets/landcover/
+
+## Download Copernicus global land cover data from Zenodo
+# BEWARE, can be very long, 25 GB of data in total
+# Launching downloads in parallel, 1 core/variable = 10 cores at most, not much RAM needed
+landcover_variables=(bare crops grass moss shrub snow tree urban water-permanent water-seasonal)
+for i in "${landcover_variables[@]}"
+do
+    wget https://zenodo.org/record/3243509/files/ProbaV_LC100_epoch2015_global_v2.0.2_"$i"-coverfraction-layer_EPSG-4326.tif -O landcover_copernicus_global_100m_v2.0.2_"$i".tif &
+done
+wait
+echo "Downloads - All done"
+rm wget-log*
diff --git a/src/lib/bioclim.jl b/src/lib/bioclim.jl
@@ -2,51 +2,48 @@
 
 # import SimpleSDMLayers: bioclim
 
-# 1st part of BIOCLIM model for single environmental variables
-function bioclim_singlevar(occ::Union{GBIFRecords,DataFrame}, pred_vars::SimpleSDMLayer; train_vars::SimpleSDMLayer=pred_vars)
+# Small internal function for BIOCLIM. Ensures values are equal for both tails and between 0-1
+bcscore(x) = isnan(x) ? NaN : (x > 0.5 ? 2*(1-x) : 2x)
+
+# 1st part of BIOCLIM model for single environmental layer
+function bioclim_layer(occ::Union{G,D}, layer::T; training_layer::T=layer) where {T <: SimpleSDMLayer, D <: DataFrame, G <: GBIFRecords}
     # occ: occurences of a single species as a DataFrame with latitude and longitude columns
-    # pred_vars: environmental variables used for prediction
-    # train_vars: optional, training environmental variables, can use a different resolution
+    # layer: layer of environmental variables used for prediction
+    # training_layer: optional, training environmental layer, can use a different resolution
 
     # Get observed environmental values (training values)
-    observed_values = train_vars[occ]
-    # Create ECDF function to extract quantile value
-    qfinder = ecdf(observed_values)
-    # Create empty array for local quantile values
-    lq = zeros(Float64, size(pred_vars))
-    # Loop for all sites (prediction values)
-    for i in eachindex(pred_vars.grid)
-        if isnan(pred_vars.grid[i])
-            # Set value to NaN if env value is already NaN
-            local_quantile = NaN
-        else
-            # Get quantile rank if not NaN
-            local_quantile = qfinder(pred_vars.grid[i])
-            # Replace values greater than 0.5 to set both tails equal
-            if local_quantile > 0.5
-                local_quantile = 1.0-local_quantile
-            end
-            # Scale back between 0 and 1
-            local_quantile = 2.0 * local_quantile
-        end
-        # Collect quantile values
-        lq[i] = local_quantile
+    obs = training_layer[occ]
+    filter!(!isnan, obs)
+    # Create empty layer for predictions
+    pred = similar(layer)
+    # Make predictions
+    if length(unique(obs)) <= 1 # special case if 1 unique value only
+        # If 1 unique value only, set 1.0 as prediction for all sites with this value
+        pred.grid = replace(x -> x == obs[1] ? 1.0 : 0.0, layer.grid)
+    else
+        # Create ECDF function to extract quantile values
+        qfinder = ecdf(obs)
+        # Get prediction score
+        pred.grid = bcscore.(qfinder.(layer.grid))
+    end
+    # Restore NaNs from original layer
+    for idx in findall(isnan, layer.grid)
+      pred.grid[idx] = NaN
     end
-    # Convert to SimpleSDMLayer, with same coordinates as prediction layer
-    prediction = SimpleSDMResponse(lq, pred_vars.left, pred_vars.right, pred_vars.bottom, pred_vars.top)
+    return pred
 end
 
-# Complete BIOCLIM model on all environmental variables, including 1st part function
-function bioclim(occ, pred_vars; train_vars=pred_vars, binary=true, with_threshold=false, threshold::Float64=0.05)
+# Complete BIOCLIM model on all environmental layers, including 1st part function
+function bioclim(occ, layers; training_layers=layers, binary=true, with_threshold=false, threshold::Float64=0.05)
     # occ: occurences of a single species as a DataFrame with latitude and longitude columns
-    # pred_vars: environmental variables used for prediction
-    # train_vars: optional, training environmental variables, can use a different resolution
+    # layers: layers of environmental variables used for prediction
+    # training_layers: optional, training environmental layers, can use a different resolution
     # binary: optional, convert result to binary presence-absence values
     # with_threshold: optional, apply threshold to remove lower predictions
     # threshold: optional, set threshold to use (default = 0.05)
 
     # Apply 1st part of BIOCLIM on each environmental variable
-    predictions = [bioclim_singlevar(occ, pred_vars[i], train_vars = train_vars[i]) for i in 1:length(pred_vars)];
+    predictions = [bioclim_layer(occ, layers[i], training_layer = training_layers[i]) for i in eachindex(layers)];
     # Reduce to single layer with minimum values
     prediction = reduce(minimum, predictions);
     # Apply threshold (if requested, default is false)

diff --git a/src/lib/landcover.jl b/src/lib/landcover.jl
@@ -23,10 +23,19 @@ function landcover(layers::Vector{Int64}; resolution::AbstractString="10", path:
     landcover_mat = [permutedims(v[:,end:-1:1]) for v in values_int]
     # Replace 255 (default no data values) by NaN
     [replace!(l, 255 => NaN) for l in landcover_mat]
-    # Add additionnal line of NaNs down South (coordinates prime did not exactly match)
-    landcover_mat = [vcat(fill(NaN, (1, size(l, 2))), l) for l in landcover_mat]
-    # Convert to SDMLayers
-    landcover_layers = SimpleSDMPredictor.(landcover_mat, -160.0, -40.0, 20.0, 80.0)
+
+    # Fill missing latitudes with NaNs (latitude extent is only (-60,80) for landcover data)
+    nlat, nlon = size(landcover_mat[1])
+    slim, nlim = abs(-60), 80
+    res = Int64(nlat/(nlim+slim))
+    south_nans = fill(NaN, ((90-slim)*res, nlon))
+    north_nans = fill(NaN, ((90-nlim)*res, nlon))
+    landcover_grids = [vcat(south_nans, l, north_nans) for l in landcover_mat]
+
+    # Convert to SimpleSDMLayers
+    landcover_layers = SimpleSDMPredictor.(landcover_grids, -180.0, 180.0, -90.0, 90.0)
+
+    return landcover_layers
 end
 
 """

diff --git a/src/lib/presence-absence.jl b/src/lib/presence-absence.jl
@@ -7,6 +7,9 @@ function presence_absence(species::DataFrame, copy_layer::SimpleSDMLayer; full_r
     # full_range: return species full range, even outside range of interest
     # binary: convert to binary presence-absence values per site
 
+    # Extract coordinates
+    lon_range = (copy_layer.left, copy_layer.right)
+    lat_range = (copy_layer.bottom, copy_layer.top)
     # Filter observations to range of interest (default)
     if full_range == false
         filter!(x -> lon_range[1] < x[:longitude] < lon_range[2], species)

diff --git a/src/master.jl b/src/master.jl
@@ -1,6 +1,4 @@
 #### Master file - Raw observations
-## NOT TESTED
-
 import Pkg; Pkg.activate(".")
 using Distributed
 @time @everywhere include("src/required.jl")
@@ -16,7 +14,7 @@ using Distributed
 
 ## 0b. Landcover data preparation (if needed)
 # save_figures = true # should figures be overwritten (optional)
-@time include("00a_data_ebd-preparation.jl")
+@time include("00b_data_landcover-copernicus.jl")
 
 #### Analyses
 
@@ -44,3 +42,6 @@ outcome = "raw" # desired outcome (required)
 
 ## 6. Run LCBD-richness relationship analysis
 @time include("06_relationship_lcbd-richness.jl")
+
+## 7. Run endemism analysis
+@time incluce("07_endemism.jl")
Original file line number	Diff line number	Diff line change
		@@ -1 +1,2 @@
		tags
		assets/landcover/landcover_copernicus_global_100m*