Tree Canopy Detection

Annual Landsat Raster Reference

Dataset overview

Each annual directory (2000-2025) contains 16 single-band GeoTIFFs: 14 spectral/derived predictors and two NDVI texture measures. Landsat Collection 2 Level-2 surface-reflectance scenes are selected for July 1-August 31, with scene-level reported cloud cover below 70%. Invalid, cloud, cloud-shadow, and snow pixels are masked. Predictors are calculated per scene and combined as a per-pixel median of valid observations. Texture measures are calculated from the resulting annual NDVI composite.

All outputs use a nominal 30 m grid, EPSG:32617 (UTM Zone 17N), clipped to the Oak Ridges Moraine Groundwater Program study area. Missing pixels are represented by NaNs. Reflectance bands are scaled using $\text{DN} \times 0.0000275 − 0.2$, as specified by the USGS. Different predictors’ annual median values at the same pixel need not come from the same observation date.

Raster descriptions

Spectral surface reflectance (6)

• Blue (Blue) - Visible blue reflectance; helps distinguish soil, water, built surfaces and vegetation, and contributes to EVI and tasseled-cap indices.
• Green (Green) - Visible green reflectance; captures part of the vegetation reflectance peak and contributes to tasseled-cap indices.
• Red (Red) - Visible red reflectance; strongly absorbed by chlorophyll in healthy vegetation and used in NDVI, EVI and SAVI.
• Near-infrared (NIR) - NIR reflectance; sensitive to leaf and canopy structure, used in all five vegetation indices.
• Shortwave infrared 1 (SWIR1) - SWIR reflectance sensitive to vegetation/soil water content; used in NDMI.
• Shortwave infrared 2 (SWIR2) - Longer-wavelength SWIR reflectance sensitive to moisture and disturbance; used in NBR.

Vegetation indices (5)

• NDVI (ndvi) - (NIR − Red) / (NIR + Red); indicates photosynthetically active vegetation, but does not by itself distinguish trees from crops or grass.
• NDMI (ndmi) - (NIR − SWIR1) / (NIR + SWIR1); emphasizes vegetation moisture differences.
• NBR (nbr) - (NIR − SWIR2) / (NIR + SWIR2); sensitive to vegetation moisture, disturbance and burn-related changes.
• EVI (evi) - 2.5 × (NIR − Red) / (NIR + 6×Red − 7.5×Blue + 1); vegetation-vigour index designed to reduce some soil/atmospheric effects and saturation.
• SAVI (savi) - 1.5 × (NIR − Red) / (NIR + Red + 0.5); vegetation index with soil-background adjustment.

Tasseled-cap components (3)

Each is a weighted linear combination of the six reflectance bands, calculated per scene using sensor-specific coefficients (Landsat 5/7 versus Landsat 8/9).

• Brightness (tc_brightness) - Overall spectral brightness; often associated with exposed soil and other bright surfaces.
• Greenness (tc_greenness) - Spectral component associated with green vegetation abundance and vigour.
• Wetness (tc_wetness) - Spectral component associated with vegetation and soil moisture conditions.

NDVI neighbourhood texture (2)

Calculated on the annual median NDVI raster using an 11 × 11 pixel moving window (approximately 330 × 330 m at 30 m resolution), excluding NaNs from local calculations.

• NDVI local mean (ndvi_mean_11x11) - Average NDVI in the neighbourhood; describes surrounding vegetation greenness.
• NDVI local standard deviation (ndvi_std_11x11) - Spatial variability of NDVI; describes local heterogeneity, not individual tree-crown texture.

Summary and comparison

Output filename pattern ({year})	Type	Primary information	Typical interpretation for canopy mapping
ormgp_blue_summer_median_{year}.tif	Reflectance	Blue visible light	Spectral separation of vegetation and nonvegetation
ormgp_green_summer_median_{year}.tif	Reflectance	Green visible light	Vegetation colour and spectral contrast
ormgp_red_summer_median_{year}.tif	Reflectance	Red visible light	Chlorophyll absorption
ormgp_nir_summer_median_{year}.tif	Reflectance	Near-infrared light	Leaf/canopy structure and vegetation vigour
ormgp_swir1_summer_median_{year}.tif	Reflectance	Shortwave infrared 1	Vegetation and soil moisture
ormgp_swir2_summer_median_{year}.tif	Reflectance	Shortwave infrared 2	Moisture and disturbance sensitivity
ormgp_ndvi_summer_median_{year}.tif	Index	Vegetation greenness	Identifies vegetation, not necessarily trees
ormgp_ndmi_summer_median_{year}.tif	Index	Moisture-related contrast	Helps separate vegetation types
ormgp_nbr_summer_median_{year}.tif	Index	Moisture/disturbance contrast	Additional vegetation-condition signal
ormgp_evi_summer_median_{year}.tif	Index	Enhanced vegetation vigour	Useful where NDVI saturates
ormgp_savi_summer_median_{year}.tif	Index	Soil-adjusted greenness	Useful in sparse/mixed vegetation
ormgp_tc_brightness_summer_median_{year}.tif	Tasseled cap	Overall brightness	Soil/built-surface versus vegetation contrast
ormgp_tc_greenness_summer_median_{year}.tif	Tasseled cap	Vegetation greenness	Complementary vegetation signal
ormgp_tc_wetness_summer_median_{year}.tif	Tasseled cap	Moisture-related spectral signal	Canopy/soil moisture differences
ormgp_ndvi_mean_11x11_{year}.tif	Texture	Neighbourhood mean NDVI	Local vegetation context
ormgp_ndvi_std_11x11_{year}.tif	Texture	Neighbourhood NDVI variation	Edges and land-cover heterogeneity

Units and cautions: Reflectance bands are unitless surface reflectance; indices and tasseled-cap components are unitless derived quantities; NDVI mean and standard deviation inherit NDVI’s unitless scale. The script constrains NDVI, NDMI, NBR, EVI and SAVI to [-1, 1]. The layers are predictors, not a tree-canopy classification. Annual observation counts, sensor differences and mixed 30 m pixels affect comparability across years.