Geospatial data models

Helena Mitasova

GIS/MEA582 Geospatial Modeling and Analysis NCSU

Learning objectives

• Define raster and vector data models
• Distinguish between continuous and discrete phenomena
• Understand and use data models transformations
• Recognize geospatial data formats
• Identify and use on-line geospatial data repositories and services

Geospatial data models

Mapped data, results of modeling or analysis are represented in GIS using

• raster (regular grid) data model
• vector (feature) data model
• specialized representations: meshes

Geospatial phenomena

• Continuous fields
  • elevation surfaces
  • temperature, precipitation
  • concentration of chemicals in soil or water bodies

• Discrete features: lines, points or areas with attributes
  • roads, buildings, cell towers
  • land use types, administrative units

• Some phenomena can be treated as both types
  • agricultural fields (crop type vs. crop height), soil properties
  • population densities

Continuous fields

• each point in space is assigned a distinct value, change in values between neighboring points is small
• mathematical representation: bi-variate or multi-variate continuous functions $w=f(x,y), w=f(x,y,z), w=f(x,y,z,t)$
• often represented by a raster data model
• vector model is also used: isolines, meshes, or points.

Discrete objects / features

• points, lines, or areas (polygons) with attributes
• represented by vector data model as geometry(shape) with attribute table
• raster representation is also used

Raster data model: 2D

• header: spatial extent and resolution, followed by matrix of values (INT, FP, DP),
• continuous field : value assigned to a grid point
• discrete object : category value assigned to pixel (area)

Raster data model: continuous fields

Raster data model: discrete features

Land use classes at 30m resolution, qualitative data

Raster data model: discrete features

Roads: Speed limits for roads and walking speed for off-road areas, 30m resolution, quantitative data

Raster data model: 3D hybrid

Cross-sections through 3D model of soil horizons

Raster data model: 3D grid

• header + 3D matrix of values, voxel model
• spatial extent N,S,E,W,Top, Bottom
• vertical resolution is usually much finer than horizontal
• often used for 3D continuous representation $w=f(x,y,z)$

Soil properties: Percent organic carbon, soil pH reaction

Vector data model

Abstract representation of complex features
school – point, road – centerline, park - polygon

• Geometry:
  • Points [x,y,(z)] represent point, line, or polygon(area) features
  • Set of points create elements of feature geometry: line: nodes, vertices; polygon: centroid, boundary
  • Geospatial Topology defines how the elements of feature geometry are interrelated and organized (don't confuse with topography!)
  • Topology ensures integrity of features and efficiency by defining shared coincident geometry (shared boundaries, nodes).

• Attributes are stored in data management systems

Vector data model geometry

Vector data model geometry

Elements: vertices(red), nodes(blue), centroids(green)

Polygons: vertices+nodes=boundaries, centroids

Vector data model examples

Geometry and attributes for point, line and polygon data:

Vector data: 3D models

Full 3D meshes: representation of structures

Modifications of data representation

• Changing raster resolution, e.g., when model inputs are rasters at different resolutions

• Changing vector geometry type, e.g., when model input requires different geometry than given data (points instead of lines)

Raster data - changing resolution

Resolution: size of the grid cell (pixel) in map units (m)

• continuous fields: interpolation
  • the higher resolution raster values are interpolated using the values of the neighboring lower resolution cells
  • methods: bi-linear, bi-cubic, spline.

• discrete raster data: nearest neighbor resampling
  • assigns the higher resolution cell the same value as the nearest lower resolution cell
  • resulting raster has only the values present in the input raster

Increasing resolution: continuous

Elevation at 30m resolution resampled to 10m resolution

Nearest neighbor creates "flats" in the resampled DEM, interpolation preserves smooth surface.
See equations for bi-linear interpolation

Increasing resolution: discrete

Geology at 30m resolution resampled to 10m resolution

Raster values are classes of observed geology

Increasing resolution: compare

Effect of resampling / reinterpolation on the results
More complex downscaling techniques using additional variables and machine learning may be needed if the difference in resolution is large

Decreasing resolution

Nearest neighbor resampling of 10m DEM to 30m and 20m DEMs

Decreasing resolution

Nearest neighbor resampling of 10m soil typemap to 30m and 20m maps

Modifying vector data

• Converting vector data type
  • lines to points, areas to lines or points
  • points to lines: network building or interpolation may be needed
  • usually preserves the shape

• Generalization
  • simplifying geometry while preserving important information
  • both data geometry and type can be modified
  • line to simplified line, polygon to simplified polygon or point
  • selecting subset of features
  • important when combining local, state and national scale data

Changing vector data type

Data geometry is not modified, but subset is extracted and stored in a different data structure

Topology building is required for conversions point to line, line to polygon

Conversion between data models

• vector to raster
  • continuous: spatial interpolation (covered by a separate topic)
  • discrete: nearest neighbor

• raster to vector
  • continuous: point sampling, isolines
  • discrete: nearest neighbor, grid center or boundary

Continuous: vector to raster data

Discrete: vector to raster data

• lines, areas: nearest neighbor
• areas: attribute value applies to the entire polygon

Raster to vector data

Continuous data: isolines, sampling points

Raster to vector data

• points – centers of grid cells
• lines, polygon border lines: connected grid cell centers
• thinning and smoothing is often performed for lines

Raster to vector data

• areas – boundary, centroid, requires building topology
• connects points on grid cell boundary

Common geospatial data formats

Format: specific implementation of data model,
open standard or proprietary

Raster

• GIS (ascii and binary): GeoTIFF, ArcGRID, GRASS, SURFER
• Imagery: MrSID, GeoTIFF, BIN, JPEG2000, IMG
• Graphics: GIF, JPG, PNG, Bitmap
• HDF, NetCDF

Vector

• GeoPackage exchange format, KML, Shape, ArcSDE, GML, MapInfo, TIGER
• PostGIS, OracleSpatial

Geospatial data format conversion

Format description is usually stored with data: automated format recognition and conversion

General library for geospatial raster and vector format conversions:
Geospatial Data Abstraction Library (GDAL/OGR)
gdal.org

given format - single abstract model - new format

GDAL includes command line utilities for data processing

Coupled with PROJ library it also provides coordinate system transformations

Data repositories

• Web mapping Service
• Web Processing Service

See Webpage with links to relevant services

Summary

• raster and vector data models
• modifying raster and vector data representation
• converting between raster and vector data models
• geospatial data formats
• data repositories, wms services