PRODES

Data Source and Methodology

PRODES monitoring: - Uses satellite imagery from multiple sources (Landsat, CBERS, others) - Automated and manual analysis to detect clear-cut deforestation - Annual assessment of forest loss - Recently (2020+) raster data published through TerraBrasilis - Data available as both raster files and aggregated by municipality

For more information, visit INPE PRODES Project and TerraBrasilis.

1. deforestation

Clear-cut deforestation areas (complete forest loss).

• Coverage: Entire Legal Amazon region
• Time period:
  • 2007: Cumulative deforestation from 1988 to 2007
  • 2008-2023: Annual incremental deforestation (year-specific)
• Data types:
  • Raw: Raster/grid data (SpatRaster format)
  • Treated: Municipality-level area aggregates
• Variables: Deforested area (hectares) by municipality
• Use cases:
  • Track annual deforestation trends
  • Identify deforestation hotspots
  • Cumulative forest loss assessment
  • Municipality-level deforestation analysis

2. residual_deforestation

Deforestation that was not captured in previous surveys (detected through improved methodology).

• Coverage: Legal Amazon
• Time period: 2010-2023
• Data types: Raster or municipality aggregates
• Variables: Additional detected deforestation (hectares)
• Use cases:
  • More comprehensive deforestation assessment
  • Correct for historical detection gaps
  • Full accounting of forest loss

3. native_vegetation

Remaining native forest and natural vegetation areas.

• Coverage: All vegetation in Legal Amazon
• Time period: 2023 (snapshot of current state)
• Data types: Raster only
• Variables: Forest and vegetation extent
• Use cases:
  • Assess remaining forest cover
  • Identify conservation priority areas
  • Vegetation extent mapping

4. hydrography

Water bodies and hydrographic features.

• Coverage: All water features in Amazon
• Time period: 2023
• Data types: Raster only
• Variables: Water body extent
• Use cases:
  • Environmental impact assessment
  • Hydrology and water resource mapping

5. non_forest

Non-forest areas including savanna, grasslands, and other vegetation types.

• Coverage: Legal Amazon region
• Time period: 2023
• Data types: Raster only
• Variables: Non-forest vegetation extent
• Use cases:
  • Distinguish non-forest from forest areas
  • Natural vegetation assessment
  • Land-use classification

6. clouds

Cloud cover in satellite imagery (data quality indicator).

• Coverage: Legal Amazon
• Time period: 2023
• Data types: Raster only
• Variables: Cloud-covered areas where analysis is uncertain
• Use cases:
  • Assess data quality and coverage
  • Identify areas needing re-analysis
  • Quantify monitoring gaps

Raw vs. Treated Data

• Raw data (raw_data = TRUE): Returns SpatRaster objects (raster grids) from TerraBrasilis
• Treated data (raw_data = FALSE): Aggregated to municipality level showing total affected area
• Raster resolution: Typically ~30-meter pixels
• Large size: Raster data is high-resolution and may be large

Cumulative vs. Incremental Data

When analyzing time trends, be aware that 2007 includes 19 years of accumulated loss.

Data Organization

• Raster format: High-resolution grid with classified pixels (deforestation, forest, etc.)
• Municipality level: Treated data aggregates raster to municipality boundaries
• Non-spatial: Municipality aggregates don’t retain spatial location (summarized only)

1. dataset

Selects which landscape classification to download.

dataset = "deforestation"             # Clear-cut deforestation (main product)
dataset = "residual_deforestation"    # Previously undetected deforestation
dataset = "native_vegetation"         # Remaining forests and vegetation (2023 only)
dataset = "hydrography"               # Water bodies (2023 only)
dataset = "non_forest"                # Savanna and other non-forest (2023 only)
dataset = "clouds"                    # Cloud cover (2023 only)

2. raw_data

Controls data format: raster or municipality aggregates.

• TRUE: Returns raw raster data (SpatRaster objects)
• FALSE: Returns municipality-level aggregated area data

raw_data = FALSE  # logical

Important: Raster data is large; ensure sufficient storage and memory.

3. time_period

Specifies which year(s) to download.

Available by dataset:

time_period = 2020              # single year
time_period = c(2015, 2020)     # multiple years
time_period = 2015:2020         # range of years (for deforestation only)

4. language

Output language.

• "pt": Portuguese
• "eng": English

language = "eng"  # character string

Example 1: Deforestation data for a single year

# download treated deforestation data for 2023
deforestation <- load_prodes(
  dataset = "deforestation",
  raw_data = FALSE,
  time_period = 2023,
  language = "eng"
)

Example 2: Deforestation data over multiple years

# download treated deforestation data for 2008 to 2023
deforestation_series <- load_prodes(
  dataset = "deforestation",
  raw_data = FALSE,
  time_period = 2008:2023,
  language = "eng"
)

Example 3: Cumulative forest loss since 1988

# download 2007 (cumulative 1988-2007) plus all years up to 2023
all_deforestation <- load_prodes(
  dataset = "deforestation",
  raw_data = FALSE,
  time_period = c(2007, 2008:2023),
  language = "eng"
)

Example 4: Residual deforestation

# download treated residual deforestation data for 2020
residual <- load_prodes(
  dataset = "residual_deforestation",
  raw_data = FALSE,
  time_period = 2020,
  language = "eng"
)

Raw Raster Data Characteristics

• Format: SpatRaster (from terra package)
• Resolution: ~30 meter pixels
• Coverage: Complete Legal Amazon region
• Size: Very large files; may require significant storage/memory
• Classification: Pixel values indicate land cover class (deforestation, forest, etc.)
• CRS: UTM zones or standard projection; verify after loading

Municipality Aggregates

• Simplified for analysis: Loses spatial detail but much smaller files
• Aggregation: Total deforested area per municipality per year
• Ideal for: Time series, trend analysis, comparative studies

Data Quality and Accuracy

1. Detection accuracy: ~95% for clear-cut deforestation
2. Minimum mapping unit: Small forest clearances may not be detected
3. Cloud contamination: Clouds may prevent detection in some areas (see clouds dataset)
4. Seasonal effects: Deforestation easier to detect in dry season

Limitations

1. Clear-cut only: Degradation and partial logging not captured (see DEGRAD for that)
2. Legal Amazon only: Data limited to Legal Amazon definition; doesn’t cover all forest regions
3. Raster data very large: May require specialized tools and significant computing resources
4. Historical changes: INPE occasionally revises historical estimates as methodology improves
5. Recent data provisional: 2023 data may be subject to revision as final processing completes