Raster Data
Geospatial data represented as a grid of pixels (cells), where each pixel stores a value. Satellite imagery, elevation models, and temperature maps are raster data. Common formats include GeoTIFF and COG. Best for continuous data and remote sensing products.
Overview
Raster data represents geographic information as a regular grid of cells (pixels), where each cell stores one or more numeric values. Satellite imagery, elevation models, temperature maps, and land cover classifications are all raster data. Each pixel covers a fixed ground area (its spatial resolution), and the grid is georeferenced to a coordinate system so that every pixel maps to a real-world location. Raster is the natural format for continuous spatial phenomena and for any data acquired by imaging sensors.
How It Works
A raster is defined by its origin (the geographic coordinate of one corner), cell size (ground dimensions of each pixel), number of rows and columns, coordinate reference system, and the numeric values stored in each cell. Multi-band rasters store multiple values per pixel — a Sentinel-2 image has 13 bands, each recording reflectance at a different wavelength. Operations on rasters include map algebra (arithmetic between layers), focal statistics (neighborhood operations like smoothing), zonal statistics (summarizing raster values within polygon zones), and reclassification.
Common formats include GeoTIFF (the universal standard), Cloud Optimized GeoTIFF (COG, optimized for HTTP access), NetCDF (multi-dimensional arrays for climate data), and HDF5 (hierarchical data for complex products). GDAL is the standard open-source library for reading, writing, and transforming raster data across all formats.
Key Facts
- Spatial resolution = ground area per pixel (e.g., 10 m × 10 m for Sentinel-2 visible bands).
- Standard formats: GeoTIFF, COG, NetCDF, HDF5. GDAL reads/writes all of them.
- Multi-band rasters store multiple values per pixel — spectral bands, time steps, or variables.
- Map algebra performs pixel-by-pixel arithmetic between raster layers (e.g., NDVI = (B8 - B4) / (B8 + B4)).
- Raster data scales quadratically — doubling resolution quadruples file size and processing time.
Applications
Satellite Imagery
All optical and radar satellite sensors produce raster data — grids of reflectance or backscatter values across spectral bands.
Elevation and Terrain
DEMs, DSMs, and slope/aspect maps are raster surfaces representing topographic properties at each grid cell.
Continuous Environmental Variables
Temperature, precipitation, air quality, soil moisture, and other environmental measurements are naturally represented as raster surfaces.
Classification Maps
Land cover maps, habitat suitability models, and risk surfaces assign categorical or continuous values to every pixel across a landscape.
Limitations & Considerations
Raster resolution is fixed — it cannot represent features smaller than one pixel, and sub-pixel detail is lost. Increasing resolution quadratically increases storage and processing requirements. Rasters are inefficient for sparse discrete features (a single point-of-interest would require an entire grid). Mixed pixels at boundaries between land cover types are a persistent challenge. Large multi-temporal raster stacks (time series) require specialized formats (NetCDF, Zarr) that GeoTIFF cannot natively represent.
History & Background
Raster data in GIS emerged alongside satellite remote sensing in the 1970s — Landsat 1 produced the first widely available raster imagery. The GeoTIFF format (1995) standardized georeferenced raster files. GDAL (2000) unified raster I/O across formats. Cloud Optimized GeoTIFF (2016) adapted the format for cloud-native access. Today, raster data dominates Earth observation, with petabytes of satellite imagery produced annually and processed through cloud platforms like Google Earth Engine and Microsoft Planetary Computer.
Analyze Raster Data data with LYRASENSE
Use our agentic notebook environment to work with satellite data and compute indices like Raster Data — no setup required.