GDAL — Geospatial Data Abstraction Library

Introduction

GDAL (Geospatial Data Abstraction Library) is the Swiss Army knife of geospatial data processing. It provides read and write access to over 200 raster and vector formats, and serves as the foundation for nearly every open-source GIS tool including QGIS, PostGIS, and MapServer.

What GDAL Does

• Translates between 200+ raster and vector geospatial formats
• Reprojects spatial data between coordinate reference systems
• Provides command-line utilities for batch geospatial processing
• Offers C/C++ and Python (via SWIG bindings) APIs
• Powers geoprocessing operations like warping, mosaicking, and rasterization

Architecture Overview

GDAL uses a driver-based architecture where each file format is implemented as a loadable driver plugin. The core library exposes a unified dataset/band model for rasters and a layer/feature/geometry model for vectors. Format drivers register themselves at runtime, making the library extensible without recompilation.

Self-Hosting & Configuration

• Available via system package managers (apt, brew, conda)
• Build from source with CMake for custom driver selection
• Configure via environment variables (GDAL_DATA, GDAL_DRIVER_PATH)
• Python bindings installable via pip (GDAL or rasterio wrapper)
• Docker images available for containerized batch processing

Key Features

• Format-agnostic data access through a single API
• Built-in coordinate transformation via PROJ library
• Virtual raster (VRT) format for zero-copy mosaics
• Cloud-optimized GeoTIFF and S3/Azure/GCS remote access
• Multithreaded I/O and processing for large datasets

Comparison with Similar Tools

• Rasterio — Pythonic wrapper around GDAL with cleaner API; still uses GDAL under the hood
• Fiona — Python vector I/O built on GDAL OGR; higher-level but same engine
• QGIS — Desktop GIS that uses GDAL for all format I/O
• MapServer — Web mapping server that depends on GDAL for data access

FAQ

Q: How do I install GDAL Python bindings? A: Use conda (recommended): conda install -c conda-forge gdal. Pip install requires matching system GDAL version.

Q: Can GDAL read files from cloud storage directly? A: Yes. Use /vsicurl/, /vsis3/, /vsigs/, or /vsiaz/ virtual filesystem prefixes to read remote files without downloading.

Q: What is the difference between GDAL and OGR? A: GDAL handles raster data, OGR handles vector data. Both are part of the same library since version 2.0.

Q: Is GDAL thread-safe? A: Read operations are thread-safe. Write operations require external synchronization or separate dataset handles per thread.

Sources

• https://github.com/OSGeo/gdal
• https://gdal.org/en/stable/