ff_polygonize is used to create focused, discrete polygons from the prediction raster that comes out of ff_predict or ff_run. It creates vector data that can be written as a Shapefile, GeoJSON or other spatial data format.

Exercise

# Load required libraries
library(ForestForesight)

# Let's assume we have downloaded predictions for Suriname into our forestforesight_data folder
prediction_file <- "forestforesight_data/predictions/SUR/SUR_2024-01-01_predictions.tif"

# Basic usage with automatic thresholding for "high" risk areas
high_risk_areas <- ff_polygonize(
    input_raster = prediction_file,
    threshold = "high",     # Automatically determine threshold for high-risk areas
    verbose = TRUE         # Show the chosen threshold value
)

# Save the polygons to a shapefile
ff_polygonize(
    input_raster = prediction_file,
    output_file = "high_risk_areas.shp",
    threshold = "high",
    minimum_pixel_count = 10,  # Slightly larger minimum area
    smoothness = 3            # Increased smoothing
)

# Create a series of risk levels to compare
risk_levels <- c("medium", "high", "very high")
risk_polygons <- lapply(risk_levels, function(level) {
    ff_polygonize(
        input_raster = prediction_file,
        threshold = level,
        verbose = TRUE
    )
})
names(risk_polygons) <- risk_levels

Key points about ff_polygonize:

Automatic Thresholding:
- "medium" - Uses a balanced threshold suitable for general risk areas
- "high" - More selective, focusing on higher probability areas
- "very high" - Most selective, only highest probability areas
Important Parameters:
- minimum_pixel_count: Controls minimum size of risk areas (default=5)
- window_size: Controls initial smoothing (default=7)
- smoothness: Controls polygon smoothing (default=2)
Output Attributes:
- risk: Average risk value within the polygon
- size: Area in hectares
- riskfactor: risk × size
- threshold: The threshold value used
- date: Generation date
Tips:
- Start with threshold = "high" for a balanced result
- Increase minimum_pixel_count to remove small polygons
- Adjust smoothness higher for smoother boundaries
- Use verbose = TRUE to see the automatically chosen threshold

Forest Foresight

ff_polygonize: creating actionable data from predictions

Analytics

Exercise

Related content