Guide - Geospatial File Format Selection

Overview

Selecting the right geospatial file format depends on your use case: web display, analysis, archival, or data exchange. This guide helps you choose the optimal format for your needs.

Quick Decision Matrix

Vector Data

Use Case	Recommended Format	Alternatives
Web mapping	GeoJSON, FlatGeobuf	TopoJSON
Database storage	PostGIS, GeoPackage	-
Data exchange	GeoPackage, GeoJSON	Shapefile
Analytical processing	GeoParquet, GeoPackage	-
Archival	GeoPackage	GeoJSON
ArcGIS workflows	File Geodatabase	Shapefile

Raster Data

Use Case	Recommended Format	Alternatives
Cloud storage	Cloud Optimized GeoTIFF (COG)	-
Local analysis	GeoTIFF	-
Web display	COG, Tiles (PNG/JPEG)	-
Multi-dimensional	NetCDF, Zarr	HDF5
Data exchange	GeoTIFF	-

Tile Archives

Use Case	Recommended Format
Vector tiles	PMTiles, MBTiles
Raster tiles	PMTiles, MBTiles
Self-hosted	PMTiles (no server needed)

---

Vector Formats

GeoJSON

Best for: Web applications, API responses, data exchange

Aspect	Details
Format	JSON with geometry
Size	Larger (text-based)
Streaming	Not natively
CRS	WGS84 (EPSG:4326) only per spec

{
  "type": "Feature",
  "geometry": {
    "type": "Point",
    "coordinates": [-122.4, 37.8]
  },
  "properties": {
    "name": "San Francisco"
  }
}

GeoPackage

Best for: Desktop GIS, data exchange, archival

Aspect	Details
Format	SQLite database
Size	Compact
Features	Vector + raster, multiple CRS
Standard	OGC official

GeoParquet

Best for: Cloud analytics, big data, columnar queries

Aspect	Details
Format	Apache Parquet with geometry
Size	Highly compressed
Streaming	Yes (row groups)
Best With	DuckDB, Spark, Arrow

FlatGeobuf

Best for: Web streaming, fast random access

Aspect	Details
Format	Binary, seekable
Size	Compact
Streaming	Yes (HTTP range requests)
Best With	MapLibre, Leaflet

Shapefile (Legacy)

Best for: Legacy compatibility only

Aspect	Details
Limitations	2GB max, 10-char field names, no NULL, single geometry type
Recommendation	Migrate to GeoPackage

---

Raster Formats

Cloud Optimized GeoTIFF (COG)

Best for: Cloud storage, web access, remote sensing

Aspect	Details
Format	GeoTIFF with internal tiles and overviews
Access	HTTP range requests
Best With	STAC catalogs, cloud storage

NetCDF / Zarr

Best for: Multi-dimensional data (climate, oceanography)

Aspect	Details
NetCDF	Mature, widespread support
Zarr	Cloud-native, chunked arrays
Dimensions	Time, depth, bands, etc.

---

Tile Formats

PMTiles

Best for: Serverless tile hosting

Aspect	Details
Format	Single-file tile archive
Hosting	Static file hosting (S3, CloudFlare)
Types	Vector (MVT) and raster

MBTiles

Best for: Tile archives with server

Aspect	Details
Format	SQLite container
Server	Requires tile server

Mapbox Vector Tiles (MVT)

Best for: Web map rendering

Aspect	Details
Format	Protocol Buffers
Standard	OGC Community Standard
Best With	MapLibre, Mapbox

---

Format Conversion

Using GDAL/OGR

# shapefile to geopackage
ogr2ogr -f GPKG output.gpkg input.shp
 
# geojson to geoparquet
ogr2ogr -f Parquet output.parquet input.geojson
 
# geotiff to cog
gdal_translate -of COG input.tif output_cog.tif

Using R

library(sf)
 
# read shapefile, write geopackage
sf::st_read("input.shp") |>
  sf::st_write("output.gpkg")
 
# read, write geoparquet
sfarrow::st_read_parquet("input.parquet") |>
  sfarrow::st_write_parquet("output.parquet")

Using Python

import geopandas as gpd
 
# read shapefile, write geoparquet
gdf = gpd.read_file("input.shp")
gdf.to_parquet("output.parquet")

---

Format Comparison

Format	Type	Size	Streaming	Cloud-Native	Standard
GeoJSON	Vector	Large	No	No	IETF RFC
GeoPackage	Vector/Raster	Medium	No	No	OGC
GeoParquet	Vector	Small	Yes	Yes	Community
FlatGeobuf	Vector	Small	Yes	Yes	Community
Shapefile	Vector	Medium	No	No	De facto
COG	Raster	Medium	Yes	Yes	OGC
PMTiles	Tiles	Small	Yes	Yes	Community

---

Notes

• Default choice for vector: GeoPackage (most compatible)
• Cloud analytics: GeoParquet with DuckDB
• Web streaming: FlatGeobuf or PMTiles
• Raster in cloud: Always use COG
• Avoid Shapefile for new projects

---

Appendix

Created: 2024-12-22 | Modified: 2024-12-22

See Also

• MOC - Geospatial
• GeoJSON
• GeoPackage
• GeoParquet
• Cloud Optimized GeoTIFF (COG)
• PMTiles

Backlinks

---

(c) No Clocks, LLC | 2024