Spatial Data Models Raster uses individual cells in a matrix, or grid, format to represent real world entities Vector uses coordinates to store the shape of spatial data objects David Tenenbaum GEOG 7 UNC-CH Spring 25
Raster Data Model In the raster data model, the primary data object is the cell or pixel You are familiar with these if you have used a digital camera or viewed a computer monitor David Tenenbaum GEOG 7 UNC-CH Spring 25
Raster Data Model The raster data model represents the Earth s surface as an array of two-dimensional grid cells, with each cell having an associated value: 2 3 5 8 Cell (x,y) 4 6 8 3 9 Cell value rows 3 5 3 3 7 5 4 3 9 2 2 4 5 2 Cell size = resolution columns David Tenenbaum GEOG 7 UNC-CH Spring 25
Raster Data Model Each grid cell in a raster data layer is one unit (the minimum amount of information in the raster data model) Every cell has a value, even if it is a special value to indicate that there is no data or that data is missing at that location The values are numbers, either: absolute values OR codes representing an attribute David Tenenbaum GEOG 7 UNC-CH Spring 25
Cells - Absolute Values In this instance, the value of the cell is actually the value of the phenomenon of interest, e.g. elevation data (whether floating point or integer): David Tenenbaum GEOG 7 UNC-CH Spring 25
Pond Branch Catchment Control Topographic Index Example David Tenenbaum GEOG 7 UNC-CH Spring 25
Cells - Coded Values Here, the values stored in each cell are used as substitutes for some nominal or categorical data, e.g. land cover classes: David Tenenbaum GEOG 7 UNC-CH Spring 25
MODIS LULC In Climate Divisions Maryland CD6 North Carolina CD3 David Tenenbaum GEOG 7 UNC-CH Spring 25
Cells Coded Values The coded values can then link to one (or more) attribute tables that associate the cell values with various themes or attributes: David Tenenbaum GEOG 7 UNC-CH Spring 25
Cell Size & Resolution The size of the cells in the raster data model determines the resolution at which features can be represented The selected resolution can have an effect on how features are represented: m Resolution m Resolution 5 m Resolution David Tenenbaum GEOG 7 UNC-CH Spring 25
Rules for Assigning Cell Values Cell values are assigned to cells accorded to some set of rules, and selecting those rules differently can also effect the representation of features: David Tenenbaum GEOG 7 UNC-CH Spring 25
Raster Data Model - Objects The raster data model still represents spatial objects, but does so differently from the vector model: Geographic Primitives Points dimensional Lines dimensional Polygons 2 dimensional David Tenenbaum GEOG 7 UNC-CH Spring 25
Raster Data Model - Points + + + + point = cell What problem do we have here? How can we solve it? David Tenenbaum GEOG 7 UNC-CH Spring 25
Raster Data Model - Lines A line = a series of connected cells that portray length Is there a problem with this representation? David Tenenbaum GEOG 7 UNC-CH Spring 25
Raster Data Model - Areas Area = a group of connected cells that portray a shape What problems could we have with this representation? David Tenenbaum GEOG 7 UNC-CH Spring 25
Raster and Vector Data Model Comparison Real World Features Raster Vector A raster model tells what occurs everywhere, while a vector model tells where every thing occurs David Tenenbaum GEOG 7 UNC-CH Spring 25
Raster Data Model - Topology Y-axis rows origin (,) X-axis (2,2) columns Spatial Relations are implicit Origin: upper left corner Each cell has 8 neighbors 4 in cardinal directions 4 in diagonal directions Cells are identified by their position in the grid Location of a cell can be calculated based on its position and the cell size David Tenenbaum GEOG 7 UNC-CH Spring 25
Grids and Missing Data This TVDI image has no data values in the black portions of the grid David Tenenbaum GEOG 7 UNC-CH Spring 25
Raster Data Model - Storage There is a trade-off between spatial resolution and data storage when we use the raster data model, e.g. 6 km satellite image with m cell size 6 X 6 = 36,, cells byte of attribute value (i.e. values -255) ~36 MB of disk storage! 6 km satellite image with m cell size 6 x 6 = 36, cells 36 KB of data % the size of the other one David Tenenbaum GEOG 7 UNC-CH Spring 25
Raster Data Model Compaction Because the raster data model records a value for each and every cell in a grid, it is very storage intensive, meaning that it can use a lot of memory and disk space to represent a theme Compaction techniques are used in conjunction with raster data to reduce the amount of required storage space to a more manageable amount David Tenenbaum GEOG 7 UNC-CH Spring 25
David Tenenbaum GEOG 7 UNC-CH Spring 25,, max. cell value rows columns Problem: too much redundancy Raster Data Storage No Compaction This approach represents each cell individually in the file: 3 values
David Tenenbaum GEOG 7 UNC-CH Spring 25,,,,, 4,, 4,,2, 4,, 4,,2, 2,, 6,,2, 2,, 6,,2, 2,, 6,,2, 2,, 6,,2,, There is a tendency towards spatial autocorrelation; for nearby cells to have similar values - values often occur in runs across several cells Raster Data Storage Run Length Encoding This approach takes advantage of patterns in the data, taking advantage of the repetition of values in a row: 45 values row by row header
Raster Data Storage - Quadtrees The quadtree method recursively subdivides the cells of a raster grid into quads (quarters) until each quad can be represented by a unique cell value: 2 2 3 32 33 3 The number of subdivisions depends on the complexity of features and stores more detail in areas of greater complexity 32 33 David Tenenbaum GEOG 7 UNC-CH Spring 25
Vector to Raster Transformations Quite often, data in the vector and raster models need to be used together, and data from one model is transformed to be represented in the other model Any such transformations can cause distortions consider this line transformed from vector to raster, and then back to vector: David Tenenbaum GEOG 7 UNC-CH Spring 25
Vector Data Model - Advantages It is a good representation of the world as we see it (our visual systems automatically segments the world we see by identifying objects) The topology of a layer can be fully described and explicitly stored It is efficient in terms of data storage It only uses storage for objects of interest and does not need to store values for the spaces in between No jaggy edges (raster has these on any diagonal) Useful for network analysis and modeling flows of linear features David Tenenbaum GEOG 7 UNC-CH Spring 25
Vector Data Model - Disadvantages The data structure is more complex especially when you have fully encoded topology (i.e. using the arc-node model) It is more difficult to write computer programs to manipulate data Spatial analysis operations can be more difficult David Tenenbaum GEOG 7 UNC-CH Spring 25
Raster Data Model - Advantages The data structure is much simpler It is easy to overlay and combine layers This model better represents continuous data Raster data is easily integrated with satellite (and other remotely-sensed) data Writing programs to manipulate raster is easier It is easy to do simulation modeling due to uniform size and shape of grids (i.e. it is easy to define uniform modeling units) David Tenenbaum GEOG 7 UNC-CH Spring 25
Raster Data Model - Disadvantages Because a value must be stored for each and every cell in a grid, there is a great deal of redundancy and large storage requirements Location can be captured only as accurately as the resolution allows, which is determined by the cell size Spatial analyses that are based on topological relationships are not well supported by this model David Tenenbaum GEOG 7 UNC-CH Spring 25
Which Data Model Should You Use? This can depend upon the type of data you re using and what goals you re trying to achieve Vector model: linear features such as rivers, roads, political boundaries Raster model: temperature, elevation (continuous) Sometimes you don t have a choice, as it is determined by data availability (or other sorts of precedents) many GIS projects are works that take many years, and new works may depend on what has been done previously and what data has been collected The two data models should be seen as complementary David Tenenbaum GEOG 7 UNC-CH Spring 25