Skip to content

Latest commit

 

History

History
687 lines (607 loc) · 30.3 KB

NotesOn_LearningGeospatialAnalysisWithPython_2ed.md

File metadata and controls

687 lines (607 loc) · 30.3 KB

Notes on Learning Geospatial Analysis with Python, 2nd Ed.

By Joel Lawhead, Packt Publishing, Dec. 2015; ISBN 9781783552429

Chapter 1: Learning Geospatial Analysis with Python

GIS concepts

  • Thematic maps - portray a specific theme
    • Minimal geographic features to avoid distracting from the theme
    • Mostly include political boundaries but avoid navigational marks
    • Use landmarks to orient the reader
    • Common uses:
      • Visualizaing health issues
      • Election results
      • Environmental phenomena
    • Tell a story, serve a purpose
    • Still only models / narratives about reality
  • Spatial databases
    • Organized collection of information
    • May be a DBMS, or not
    • Most DBMS's typically contain scalar and blob data
    • Spatial / geodatabases extend RDBMSs to store / query data in two or three dimensional space
    • Some account for time series data as well
  • Spatial indexing
    • Organizing geospatial data for faster retrieval
  • Metadata
    • Provides traceability for source and history of datasets
    • Summarizes technical details of the dataset
    • Several possible standards:
      • ISO 19115-1, which gives hundreds of fields to describe a geospatial dataset
      • ISO 19115-2 gives extensions for imagery and gridded data
    • Example fields:
      • spatial representation
      • temporal extent
      • lineage
    • Open Geospatial Consortium (OGC) created the Catalog Service for the Web (CSW) to manage metadata
    • pycsw library implements the CSW standard
  • Map projections
    • Basic problem is always 3D projected onto 2D
    • Choice is always a compromise about what to preserve and what to give up
    • Projections are typically a set of 40+ parameters
    • Format is either XML or Well-Known-Text (WKT), that defines the transform
    • Intl. Association of Oil and Gas Producers (IOGP) maintains a registry of most common projections
    • That was formerly European Petroleum Survey Group (EPSG)
    • Entries are still called EPSG codes
    • 5000 plus entries in the registry
    • Previously projections were a huge hassle in terms of data storage and maintenance
    • Many data formats for geospatial don't actually include the projection info, so it has to be in metadata, which is risky from a management perspective
    • Thankfully high speed transfer and cheap storage addresses a lot of that
    • Most data formats now have metadata formats defining projection, or it lives in the file header
    • There are global basemaps, that give you common projections like Google Mercator/Web Mercator
    • Web Mercator is EPSG:3857 (or deprecated EPSG:900913)
    • Modern software can often reproject on the fly
    • Closely related issue: geodetic datums
    • A datum is a model of the earth's surface that matches the location of features on Earth to a coordinate system. Most common is probably WGS 84, used by GPS devices
  • Rendering
    • Geodata vector data exists as 2- or 3-tuples of (x,y[,z])
    • Geodata is stored in a coordinate system representing a grid overlaid on the earth (3 dimensional)
    • Screen coordinates / pixel coordinates are a 2d grid
    • XY world coordinates map to pixel coordinates pretty simply by a scaling algorithm
    • The z coordinate has to go through a transform to be mapped onto the 2d plane
    • For remote sensing / raster data, challenge is file size and compression
    • Lossless compression is possible, though rendering is still computationally expensive

Remote Sensing Concepts

  • Sources of raster geodata can include ground based radar, laser range finders, specialized devices and detectors for gas, radiation, EM, etc.
  • For purposes of this text, looking at remote sensing platforms that capture large amounts of Earth data, including Earth imaging systems, some elevation data, and some weather systems
  • Images as data
    • Raster data is captured as square tiles
    • If it's multi-spectral, it will contain multiple bands in arrays
    • An array item in one of these 2d arrays represents both space and some other value, reflectance or whatever
  • Remote sensing and color
    • Visible light data in RGB can be captured and displayed per pixel
    • Non-visible EM can be shown in false color

Common vector GIS concepts

  • Data structures
    • Vector data has at minimum x/y values per point, sometimes a z value
    • Coordinates are combined to form points, lines, and polygons
    • Typically represents elevation better than raster data
    • It's more costly to collect than raster data
    • Two important concepts for vector data structures:
      • Bounding box / min bounding box - smallest rectangle bounding all points in a set
      • Convex hull - smallest convex polygon bounding all points in a set
    • the bounding box always contains the convex hull
  • Buffer
    • Buffer operations can apply to points, lines, and polygons
    • Buffers create a polygon around the original object at a fixed distance
    • Buffers are used for proximity analysis
  • Dissolve
    • Creating one polygon out of two or more adjacent ones
  • Generalize
    • Reducing the number of points being used to define an object, for efficiency
  • Intersect
    • Discover whether one part of a feature intersects with one or more additional features
  • Merge
    • Combines two or more non-overlapping shapes into a multishape object
    • Multishape objects maintain separate geometries but are treated as a single feature with a single attribute set
  • Point in Polygon
    • Check to see whether a point falls inside or outside a given polygon
    • Most commonly done with ray casting
      • Check to see if the point is on the polygon boundary
      • Draw a line from the point in a single direction
      • Count the number of times the line crosses the polygon boundary until it reaches the bounding box of the polygon
      • If the number is odd, the point is inside, if even, the point is outside
  • Union
    • Combine two or more overlapping polygons in a single shape
  • Join
    • SQL join, combines tabular data
    • Spatial join defined by a spatial db extension will combine features similar to an SQL join, but does so based on feature proximity. For instance, you could derive county name from features inside a county.
  • Geospatial rules about polygons
    • These are rules of thumb about how geospatial polygons differ from mathematical polygons:
      • Polygons must have at least four points, first and last must be the same
      • A polygon boundary should not overlap itself
      • Polygons in a data layer should not overlap
      • A polygon in a layer inside another polygon is a hole in the underlying polygon
    • Different geospatial software resolves those rules differently
    • Best practice is to make sure your polygons obey those rules
    • By definition a polygon is a closed shape. Some software will error if you haven't explicitly duplicated the first point as the last point in the polygon's dataset, some will close it automatically without complaining.
    • How the polygon is defined is dictated by the data format

Common raster data concepts

  • Band math
    • Multidimensional array math, typically matrix ops from linear algebra
  • Change detection
    • Highlight changes between images taken at different times
  • Histogram
    • Statistical distribution of values in the dataset
    • Used for analysis and operations like contrast increases
  • Feature extraction
    • Manual or automatic digitization of features in an image to form vector data
  • Supervised classification
  • Unsupervised classification

Creating the simplest possible Python GIS

Chapter 2: Geospatial Data

Overview of common data formats

  • Typical data sources may be:
    • Spreadsheets, CSV, TSV
    • Geotagged photos
    • Lightweight binary points, lines, polygons
    • Multi-GB satellite or aerial rasters
    • Elevation data like grids, point clouds, integer based images
    • XML
    • JSON
    • Databases
    • Web services
  • Before 2004 it was hard to get geodata
  • Then Google Maps and Google Earth came out, Microsoft launced TerraServer, and Open Geospatial Consortium developed Web Map Service to 1.3.0.
  • Esri also released v9 of ArcGIS server
  • These all drove things towards common basemaps and Googles web map tiling model
  • Gmaps was served dynamically with AJAX, and scaled really solidly
  • Also offered people the ability to mashup data with an API
  • There are distributed geospatial layers like OpenLayers, that gives an API richer than Gmaps
  • Complimentary to that is OpenStreetMap, which has global, street-level vector data
  • Incentives changed, siloed data started opening up some
  • Analysts benefited:
    • Data distribution started being standardized to Mercator
    • Google standardized datum on WGS 84, as does GPS

Data structures

  • Common traits of geodata across formats
    • Geolocation
    • Subject information

Spatial indexing

  • Indexing algorithms
    • Quadtree index
      • series of different algorithms based on a common theme
      • Each node in the index has four children
      • Child nodes are typically square or rectangular
      • When a node contains a specified number of features, it splits if additional features are added
      • Dividing space into nested squares speeds up spatial search
      • Software only has to handle five points at a time and use gt/lt comparisons to check for point inclusion in a node
      • Most commonly found as file based indices
    • R-tree index
      • More sophisticated than quadtrees
      • Handle 3D data
      • Optimized to store the index in a way compatible with how a database uses disk / memory
      • Nearby objects are aggregated into hierarchical nodes that are balanced at each level
      • Bounding boxes of objects may overlap across nodes (unlike quadtree)
      • Due to complexity / db integration, these are typically found in databases, not files
    • Grids
      • Spatial indexes may use integer grids
      • Coordinates are typically floating point with a fixed degree of precision
      • Float computations are expensive compared to integers
      • Indexed search is about initial eliminating possibilities that don't require precision (clearly outside the bounds of hte query)
      • Most spatial indexing algorithms therefore map floats to a fixed size integer grid for initial search, then switch to full resolution for the narrowed dataset
      • Grid sizes can be wildly variant

Overviews

  • Most overview data is raster
  • They're resampled, lower resolution versions that give thumbnail or preload views at different scales
  • Also known as pyramids, 'pyramiding' an image is creating one
  • Usually preprocessed and stored with the full resolution data, embedded or in a separate file
  • Vector data also has a concept of overviews, typically created on the fly because vectors are scalable
  • Sometimes vector data is rasterized by converting to a thumbnail image, stored with or embedded in the image header

Metadata

  • Most data formats contain the footprint or bounding box of the data on the earth
  • Detailed metadata is typically stored outside the data file
  • Formats include
    • US Federal Geographic Data Committee (FGDC) Content Standard for Digitial Geospatial Metadata (CSDGM)
    • ISO
    • Newer EU format, Infrastructure for Spatial Information in the European Community (INSPIRE)

File Structure

  • Human readable formats are easy to poke around in

  • Binary data you can get at with the struct module or a third party library

  • Example of parsing the bounding box out of a shapefile:

    import struct
bb = {}
with open("hancock.shp", "rb") as f:
    f.seek(36)
    bb['min_lon'] = struct.unpack("<d", f.read(8))
    bb['min_lat'] = struct.unpack("<d", f.read(8))
    bb['max_lon'] = struct.unpack("<d", f.read(8))
    bb['max_lat'] = struct.unpack("<d", f.read(8))

Vector data

  • OGC has 16+ formats for vector data
  • Stores only geometric primitives, as associated points
  • Geospatial vector data stores Earth-based points, not screen based
  • Typically linked to info about the object being represented
  • Geospatial vector data typically contains no styling information
  • Geospatial vectors typically include very primitive geometries for points, lines, and polygons, no curves
  • You can store vector data in human readable formats like CSV, text, GeoJSON, XML
  • In the early 90's data was all binary, then switched to XML. File sizes are greater, but they're more portable / genericized.
  • Open source library OGR has 86+ supported vector formats
  • Commercial counterpart, Safe Software's Feature Manipulation Engine (FME) has 188+

Shapefiles

  • Most ubiquitous format, Esri shapefile
  • Released the spec as an open format in 1998
  • Basically all GIS software reads it
  • OGR works on it, as do the modules Shapely and Fiona (based on OGR)
  • The format consists of multiple files (3 at minimum, up to 15)
  • Required file types within the standard:
    • .shp
      • Purpose: shapefile, contains the geometry
      • Some software will accept only .shp without the .shx or .dbf
    • .shx
      • Purpose: shape index file; fixed sized record index referencing the geometry for fast access
      • Meaningless without the .shp
    • .dbf
      • Purpose: database file, holds geometry attributes
      • Can be accessed separately sometimes, as the format predates .shp. Openable by spreadsheets.
  • Optional file types within the standard:
    • .sbn
      • Purpose: spatial bin file, the shapefile spatial index
      • Has bounding boxes of features mapped to a 256x256 integer grid
    • .sbx
      • Purpose: Fixed sized record index for the .sbn file
      • Ordered record index of a spatial index
    • .prj
      • Purpose: Map projection info stored in well known text format
      • May also accompany raster data, needed for on the fly reprojection
    • .fbn
      • Purpose: Spatial index of read only features
      • Very rarely used
    • .fbx
      • Purpose: Fixed-size record index of the .fbn spatial index
      • Very rarely used
    • .ixs
      • Purpose: Geocoding index
      • Common in geocoding applications including driving directions
    • .mxs
      • Purpose: Another type of geocoding index
      • Less common than .ixs
    • .ain
      • Purpose: Attribute index
      • Mostly legacy format, rarely used now
    • .aih
      • Purpose: Attribute index
      • Accompanies .ain files
    • .qix
      • Purpose: Quadtree index
      • Spatial index created by open source community because Esri .sbn and .sbx files were undocumented
    • .atx
      • Purpose: Attribute index
      • More recent, Esri specific attribute index for fast queries
    • .shp.xml
      • Purpse: Metadata
      • Geospatial metadata container, can be any of multiple XML standards
    • .cpg
      • Purpose: Code page file for .dbf
      • Used for internationalization of the .dbf files
  • If you want to rename a shapefile, you must rename all associated files as well
  • Records in these files are not numbered
  • Records include geometry, the .shx index record, and the .dbf record
  • Those are stored in a fixed order
  • Records are numbered dynamically on opening them but the numbers are not saved
  • Deleting a record will bump its number next time it is opened
  • Don't base anything on record number unless you create a new, secondary identifier in the .dbf file and assign each record a number.
  • If you edit shapefiles, do it with software that manages the file(s) as a shared dataset.

CAD files

  • CAD formats are mostly from Autodesk via AutoCAD
  • Two typically seen formats are
    • Drawing Exchange Format (DXF)
    • Drawing (DWG, autocad native)
  • Features of these file types:
    • Curves
    • Surfaces
    • 3d solids
    • Text rendered as objects
    • Text styling
    • Viewport configuration
  • If you encounter CAD data, best to ask if you can get shapefiles

Tag-based and markup-based formats

  • Typically XML

  • May also be well-known text (WKT) for .prj

  • XML formats include

    • Keyhole Markup Language (KML)
    • OpenStreetMap (OSM)
    • Garmin GPX for GPS data
    • Open Geospatial Consortium's Geographic Markup Language (GML)
    • That's the basis for the OGC Web Feature Service (WFS)
    • GML has been largely superseded by KML and GeoJSON

  • XML files have more than just geometry, may have attributes and rendering instructions

  • KML is now a fully supported OGC standard

  • XML is attractive to geospatial analysts because:

    • It's human readable
    • Can be edited with a text editor
    • Well supported by programming languages
    • By definition easily extensible

  • Issues with XML:

    • Inefficient for large data storage
    • Can be cumbersome to edit
    • Errors in datasets are common, most parsers deal with them poorly

  • SVG (scalable vector graphics) is a widely supported XML format

  • SVG is not a geographic format

  • WKT is an older OGC standard

  • Example WKT for WGS84 coordinate system:

    GEOGCS["WGS 84",
    DATUM["WGS_1984",
        SPHEROID["WGS 84",6378137,298.257223563,
            AUTHORITY["EPSG","7030"]],
    PRIMEM["Greenwich",0,
        AUTHORITY["EPSG","8901"]],
    UNIT["degree",0.01745329251994238,
        AUTHORITY["EPSG","9122"]],
    AUTHORITY["EPSG","4326"]]

  • Standards codes can be quite long, EPSG created a numerical coding system to reference projections by shorthand

GeoJSON

  • Standard way to define geometry, attributes, bounding boxes, and projection information
  • More compact than XML, though less than binary formats
  • Key component of REST web APIs

Raster data

  • Raster datasets are two dimensional arrays
  • Can be stored as ASCII or BLOB in databases
  • Geospatial raster resolution is not dots per inch, it's ground distance per cell
  • May contain multiple bands, meaning that different wavelengths of light can be collected over the same area
  • Typical range is 3-7 bands, but can be several hundred
  • Can be viewed individually or swapped in and out as the RGB bands of an image
  • Can be recombined into a derivative single-band image, and recolored using a set number of classes representing like values
  • Raster data often shows up in scientific computing, which uses complex formats including Network Common Data Form (NetCDF), GRIB, and HDF5, which store entire data models
  • Raster data can come in a bunch of formats
  • Geospatial Data Abstraction Library (GDAL) includes 130+ raster formats

TIFF files

  • Tagged Image File Format, most common geospatial raster format
  • Flexible tagging system lets it store any type of data in a single file
  • Can have overview images, multiple bands, integer elevation data, basic metadta, internal compression, and a variety of other data typically stored in additional supporting files by other formats
  • Anyone can unofficially extend the format by adding tagged data to the file structure
  • May mean a "valid" TIFF does not work in some applications, if it has been extended beyond what that application recognizes
  • GeoTIFF defines how geospatial data is stored
  • May show up as .tiff, .tif, or gtif

JPEG, GIF, BMP, PNG

  • Common image formats, can be used for geospatial data
  • Typically rely on accompanying support files for georeferencing
  • JPEG is fairly common for geospatial data
  • JPEGs have a built in metadata tagging system, EXIF

Compressed formats

  • Geodata rasters tend to be stored using compression because they're real big
  • Latest open standard is JPEG 2000
  • That includes wavelet compression and georeferenced data
  • Multi-resolution Seamless Image Database (MrSID, .sid) and Enhanced Compression Wavelet (ECW, .ecw) are proprietary wavelet compression formats that come up in geospatial contexts
  • TIFF supports compression including Lempel-Ziv-Welch (LZW)
  • Compressed data is fine for part of a base map, but should not be used for remote sensing processing

ASCII grids

  • Common for elevation data
  • File format created by Esri, but now an unofficial standard, widely supported
  • Simple text file with x,y values as rows and columns
  • Spatial info for the raster is in a simple header
  • Not terribly efficient, but don't require any special libraries either
  • Often distributed as zip files
  • Headers contain:
    • number of columns
    • number of rows
    • x-axis cell center coordinate | x-axis lower-left corner coordinate
    • y-axis cell center coordinate | y-axis lower-left corner coordinate
    • cell size in mapping units
    • the no-data value (typically 9999)

World files

  • Simple text files that provide geospatial referencing info to any image externally for file formats with no native spatial info
  • Recognized by geospatial software due to naming convention
  • Most common way to name a world file is to use the raster file name and then alter the extension to remove the middle letter and add w to the end
  • Examples:
    • World.jpg == World.jpw
    • World.tif == World.tfw
    • World.bmp == World.bpw
    • World.png == World.pgw
    • World.gif == World.gfw
  • File structure is simple, it's a six line text file:
    • Line 1: Cell size along the x axis in ground units
    • Line 2: Rotation on the y axis
    • Line 3: Rotation on the x axis
    • Line 4: Cell size along the y axis in ground units
    • Line 5: Center x-coordinate of the upper left cell
    • Line 6: Center y-coordinate of the upper left cell
  • Rotation is crucial because remote sensing images are often rotated due to teh data collection platform
  • Great tool when working with raster data in python

Point cloud data

  • Any data collected as the (x,y,z) location of a surface point based on some sort of focused energy return
  • You can get it from lasers, radar, acoustic soundings, or other waveform generators
  • Spacing between points is arbitrary and depends on the type and position of the collecting sensor
  • Book mostly concerned with LIDAR data and radar data
  • Radar point cloud data mostly comes from space missions, LIDAR is terrestrial or airborne
  • LIDAR uses laser range finding to model the world at high precision
  • LIDAR is a combination of the words light and radar, maybe Light Detection and Ranging
  • LIDAR sensors are high-speed, continuous, and have a wide field of view (often 360 from the sensor), so the data doesn't tend to have a regular footprint like other raster datasets
  • LIDAR datasets are point clouds because the data is a stream of 3 tuples, where the z value is the distance from the laser to the ranged object, and the x,y are the projected location of that object calculated from the location of the sensor.
  • Most common format for LIDAR is LIDAR Exchange Format (LAS)
  • Can be represented in many ways including a text file with one tuple per line
  • Can be used to create 2D elevation rasters, can be colorized

Web Services

  • Most common protocols are Web Map Service (WMS) that returns a rendered map image and Web Feature Service (WFS) that typically returns GML
  • Many WFS services can also return KML, JSON, zipped shapefiles, and other formats

Chapter 3: The Geospatial Technology Landscape

  • Most software, open source and commercial, derives from a few key packages
  • High level core capabilities for geospatial libraries:
    • Data access
    • Computational geometry (incl. data reprojection)
    • Visualization
    • Metadata tools
    • Image processing for remote sensing (very fragmented category)
  • Most image processing software is based on the core data access libraries with custom image processing logic layered on top
  • Common examples of this type of software:
    • Open Source Software Image Map (OSSIM)
    • Geographic Resources Analysis Support System (GRASS)
    • Orfeo ToolBox (OTB)
    • ERDAS IMAGINE
    • ENVI
  • Core libraries in use by most other packages:
    • GDAL
    • OGR
    • GEOS
    • PROJ.4

Data access

  • Data access libraries underpin everything else in GIS work
  • Accuracy and precision are incredibly important
  • Libraries must manage memory efficiently

GDAL

  • Geospatial Data Abstraction Library
  • Gives a single, abstract data model for raster data types
  • Consolidates data access libraries for different data formats
  • Gives a common read/write API
  • Abstraction of the GDAL dataset:
    • GDAL Rasterband
      • Raster (0 to n bands)
        • Width in pixels
        • Height in lines
      • Data Type
        • Byte
        • UInt16 / Int16 / UInt32 / Int32
        • Float32 / Float 64
        • CInt16 / CInt32
        • CFloat32 / CFloat64
      • Block size
        • Data chunk read size
    • Projection
      • Coordinate system
      • Georeferencing
        • Affine geo-transform
        • Ground control points
    • Metadata
      • Arbitrary tagging system
        • Some default tags
        • XML storage ability
    • Overviews
      • Freestanding bands
        • Reduced resolution
      • 0-n overview bands

OGR

  • OGR Simple Features Library is the vector companion to GDAL
  • At least partial support for 70+ vector formats
  • Originally stood for OpenGIS Simple Features Reference Implementation
  • Not actually a reference implementation
  • Licensed X11/MIT
  • Capabilities:
    • Uniform vector data and modeling abstraction
    • Vector data re-projection
    • Vector data format conversion
    • Attribute data filtering
    • Basic geometry filtering including clipping and point-in-polygon testing
  • Architectural sections / objects in OGR
    • Geometry
    • Feature definition
    • Feature
    • Spatial reference
    • Layer
    • Data source
    • Drivers
  • The Geometry object represents the data model for points, lines, linestrings, polygons, geometrycollections, multipolygons, multipoints, and multilinestrings
  • Feature object ties Geometry and Feature Definition info together
  • Spatial Reference contains an OGC Spatial Reference definition. What?
  • Layer represents features grouped as layers in a data source
  • Data Source is the file or database object accessed by OGR
  • Driver contains translators for the multiple underlying data formats
  • Works smoothly with one quirk: Layer concept is used even for data formats that only contain a single layer. Mild inconvenience.

Computational Geometry

  • Algorithms for ops on vector data
  • Most geospatial libraries are separate from graphics libraries because of geospatial coordinate systems
  • Screen coordinates are almost entirely positive, geo coordinates can be positive or negative across a meridian
  • Some features of OGR move beyond data access into computational geometry
  • Geospatial algorithms are pretty hard to implement from scratch.

PROJ.4 projection library

  • Created by Jerry Evenden at USGS in the 90s
  • Now a project of the Open Source Geospatial Foundation
  • Purpose is to transform data betwene thousands of coordinate systems
  • The math to do so is super complex, and nothing approaches PROJ.4
  • Uses a simple syntax capable of describing any projection
  • Used in virtually every major GIS package that does reprojection
  • Has its own command line tools
  • Also available through GDAL and OGR
  • Access it directly to reproject individual points

CGAL

  • Computational Geometry Algorithms Library
  • Originally from late 90s
  • Not specifically for geospatial analysis, but commonly used
  • Useful for operations like resizing a polygon

JTS

  • Java Topology Suite
  • Implements the Open Geospatial Consortium (OGC) Simple Features Specification for SQL
  • Has been ported to other languages
  • Has a great test program, JTS Test Builder, that gives you a GUI for testing individual functions without writing wrapper scripts
  • Lets you interactively test algorithms to verify data or just understand a process
  • Has not seen development in a long time

GEOS

  • Geometry Engine - Open Source
  • C++ port of the JTS library
  • Much larger impact on geospatial work than JTS
  • Lots of infrastructure exists for things like python bindings
  • Most common usage is via APIs
  • Capabilities:
    • OGC Simple Features
    • Geospatial predicate functions
    • Intersects
    • Touches
    • Disjoint
    • Crosses
    • Within
    • Contains
    • Overlaps
    • Equals
    • Covers
    • Geospatial Operations
    • Union
    • Distance
    • Intersection
    • Symmetric Difference
    • Convex Hull
    • Envelope
    • Buffer
    • Simplify
    • Polygon assembly
    • Polygon validation
    • Area
    • Length
    • Spatial Indexing
    • OGC well-known text and well-known binary IO
    • C and C++ API
    • Thread safety
  • Can be compiled with GDAL to give OGR all its capability

PostGIS

  • Most common spatial database
  • Module on top of PostgreSQL
  • Much of the power comes from GEOS library
  • Implements OGC Simple Features Specification for SQL
  • Allows you to execute both attribute and spatial queries against a dataset
  • Spatial operations are available via SQL functions
  • Feature set
    • Geospatial geometry types including points, linestrings, polygons, multipoints, multilinestrings, multipolygons, and geometry collections
    • Spatial functions for testing geometric relationships
    • Spatial functions for deriving new geometries
    • Spatial measurements including perimeter, length, area
    • Spatial indexing using R-Trees
    • A basic geospatial raster data type
    • Topology data types
    • US Geocoder based on TIGER census data
    • New JSONB datatype that allows indexing/querying JSON and GeoJSON

Other spatially-enabled databases

  • PostGIS is the standard, but there are others to be aware of
  • Oracle Spatial and Graph
    • geospatial data schema
    • spatial indexing system based on R-Tree
    • SQL API for geometric operations
    • Spatial data tuning API for optimizing datasets
    • Topology data model
    • Network data model
    • GeoRaster data type
    • 3D data types including Triangulated Irregular Networks and LIDAR point clouds
    • Geocoder
    • Routing engine for driving direction queries
    • OGC compliance
  • ArcSDE, Esri's Spatial Data Engine
    • Rolled into ArcGIS Server
    • Mostly DB independent, supports multiple DB backends
  • Microsoft SQLServer
  • MySQL

SpatiaLite

  • Extension to SQLite
  • Spatial data types and indexing are in SQLite
  • This adds OGC Simple Features Specification compliance and map projections

Routing

  • Niche area of computational geometry
  • Main contenders for dealing with it are Esri Network Analyst and pgRouting engine for PostGIS

Desktop tools (including visualization)