Skip to contents

Environmental thinning

Source: vignettes/environmental-thinning.Rmd
environmental-thinning.Rmd

Load the package

Step 3: Objective Grid Resolution using Nearest Neighbors

The most critical parameter in environmental gridding is the grid_resolution. Instead of guessing this value, we can derive it objectively from the data by analyzing the density of points in environmental space.

The find_env_resolution() function uses a geometric “elbow” method based on nearest-neighbor distances. This identifies the exact distance where dense artificial clustering transitions into natural data spacing, eliminating the need for arbitrary quantiles.

data(origin_dat_prepared, package = "bean")
# Find the objective resolution using the elbow method
resolution_results <- find_env_resolution(
  data = origin_dat_prepared,
  env_vars = c("bio_1", "bio_4", "bio_12", "bio_15")
)
#> Calculating nearest neighbor environmental distances and detecting elbows...

plot(resolution_results)

# Let's use this objective resolution in the next step
grid_res <- resolution_results$suggested_resolution
grid_res
#>       bio_1       bio_4      bio_12      bio_15 
#> 0.002405167 0.098876953 1.000000000 0.034934998

Step 4: Apply Thinning

Now that we have an objective, data-driven grid_resolution, we can apply the thinning. We offer two methods: stochastic and deterministic.

Method A: Stochastic Thinning with thin_env_nd

This method randomly samples exactly one point from each occupied grid cell, enforcing strict presence-only logic within the environmental hypercube.

# Apply the stochastic thinning (using a seed for reproducibility)
thinned_stochastic <- thin_env_nd(
  data = origin_dat_prepared,
  env_vars = c("bio_1","bio_4", "bio_12", "bio_15"),
  grid_resolution = grid_res
)

# Print the summary of the thinning results
thinned_stochastic
#> --- Bean Stochastic Thinning Results ---
#> 
#> Thinned 1024 original points to 78 points.
#> This represents a retention of 7.6% of the data.
#> 
#> --------------------------------------
head(thinned_stochastic$thinned_data)
#>           species        y         x    bio_1 bio_12   bio_15    bio_4
#> 945 Rusa unicolor 14.45169 101.29557 24.11074   1166 77.36796 173.1696
#> 807 Rusa unicolor 16.45394 101.72157 24.19963   1010 80.99183 225.2455
#> 968 Rusa unicolor 16.25104 101.57840 24.35942   1031 80.01744 209.8779
#> 919 Rusa unicolor 16.29003 101.49113 24.40832   1048 80.30025 207.4802
#> 812 Rusa unicolor 15.67431  99.28283 24.51126   1315 78.66610 186.7147
#> 484 Rusa unicolor 12.79893  99.45413 24.52396   1115 73.28954 110.6435

Method B: Deterministic Thinning with thin_env_center

This method provides a simpler, non-random alternative. It returns a single new point at the exact center of every occupied grid cell, regardless of how many original points fell within it.

# Apply the deterministic thinning
thinned_deterministic <- thin_env_center(
  data = origin_dat_prepared,
  env_vars = c("bio_1","bio_4", "bio_12", "bio_15"),
  grid_resolution = c(0.5,0.5,0.5,0.5)
)

# Print the summary of the thinning results
thinned_deterministic
#> --- Bean Deterministic Thinning Results ---
#> 
#> Thinned 1024 original points to 78 unique grid cell centers.
#> This represents a retention of 7.6% of the data.
#> 
#> --------------------------------------
head(thinned_deterministic$thinned_points)
#>   bio_1  bio_4  bio_12 bio_15
#> 1 23.75 175.25 1414.25  78.75
#> 2 24.75 180.25 1289.25  78.25
#> 3 24.25 173.25 1166.25  77.25
#> 4 25.25 186.25 1309.25  78.75
#> 5 23.25 182.75 1134.25  76.25
#> 6 27.75 123.25 1269.25  74.25

Step 5: Visualize the Thinning Results

The plot_bean() function provides a powerful way to visualize the effect of thinning by overlaying the thinned points on the original data within the environmental grid.

data(occ_data_raw, package = "bean")
data(origin_dat_prepared, package = "bean")
data(thinned_stochastic, package = "bean")
data(thinned_deterministic, package = "bean")

# Visualize the stochastic thinning results
plot_bean(
  original_data = origin_dat_prepared,
  thinned_object = thinned_stochastic,
  env_vars = c("bio_1","bio_4", "bio_12", "bio_15")
)


# Visualize the deterministic thinning results
plot_bean(
  original_data = origin_dat_prepared,
  thinned_object = thinned_deterministic,
  env_vars = c("bio_1","bio_4", "bio_12", "bio_15")
)

# Visualize the spatial distribution of the occurrence and thinned points

# Load the environmental raster layers
thai_env_file <- system.file("extdata", "thai_env.tif", package = "bean")
env <- terra::rast(c(thai_env_file))

# Combine the data and add a label for the partition
plot_data <- dplyr::bind_rows(
  occ_data_raw %>% dplyr::mutate(Data_Type = "Original (Raw)"),
  thinned_stochastic$thinned_data %>% dplyr::mutate(Data_Type = "Thinned (Stochastic)")
)

# Lock the order so "Original" is always on the left partition
plot_data$Data_Type <- factor(plot_data$Data_Type, levels = c("Original (Raw)", "Thinned (Stochastic)"))

# Plot with a partition (facet_wrap)
ggplot(plot_data, aes(x = x, y = y)) +
  geom_raster(data = as.data.frame(env[[1]], xy = TRUE), aes(x = x, y = y), fill = "gray80") +
  geom_point(alpha = 0.5, color = "darkred") +
  coord_fixed() +
  facet_wrap(~Data_Type) + # <--- This creates the partition line!
  labs(title = "Spatial Point Distribution",
       x = "Longitude", y = "Latitude") +
  theme_classic() +
  theme(strip.background = element_rect(fill = "grey90"),
        strip.text = element_text(face = "bold"))