raster analysis and terrain analysis chapter 10 & 11 raster analysis
Post on 17-Jan-2016
Embed Size (px)
Raster Analysis and Terrain AnalysisChapter 10 & 11Raster Analysis
Raster Data Model
Raster cells store data (nominal, ordinal, interval/ratio)
Excellent for terrain and hydrological modelingComplex constructs built from raster data-Connected cells can be formed in to networks -Related cells can be grouped into neighborhoods or regions
Examples of Raster Analysis
Predict fate of pollutants in the atmosphereThe spread of diseaseAnimal migrationsCrop yieldsEPA - hazard analysis of urban superfund sitesMarket analysisWatershed analysisTerrain analysis
Map algebraConcept introduced and developed by Dana Tomlin and Joseph Berry (1970s)
Cell by Cell combination of raster data layers
-Each number represents a value at a raster cell location-Simple operations can be applied to each number-Raster layers may be combined through operations such as addition, subtraction and multiplication
Scope: Local operations
Many Local Functions(page 412 of text)
Logical OperationsANDNon-zero values are true, zero values are falseN = null values
Logical OperationsORNon-zero values are true, zero values are falseN = null values
More Local Functions logical comparisons
An Example of a Logical Operation
Raster Clip Operation
Be Careful of Ambiguity
Be Careful of Ambiguity
Raster Overlay in Idrisi
First * Second0*0=00*1=01*1=1
First + Second0+0=00+1=11+1=2
Scope: Neighborhood operations
Moving Windows(Windows can be any size; often odd to provide a center)
Kernels vs. Moving Window
Neighborhood Operations: Separate edge kernals can be used
Example:Identifying spatial differences in a raster layer
Moving windows and kernals can be used with a mean kernal to reduce the difference between a cell and surrounding cells. (done by average across a group of cells)
Raster data may also contain noise; values that are large or small relative to their spatial context.(Noise often requiring correction or smooth(ing))
Know as high-pass filters
The identified spikes or pits can then be corrected or removed by editing
Scope: Global operation
ArcMaps Raster CalculatorFrom ArcGIS Help Files
Raster Calculator tool dialog box example
The minimum cost of reaching cells in a layer from one or more sources cells
travel costsTime to school; hospital; Chance of noxious foreign weed spreading out from an introduction point
Units can be money, time, etc.
Distance measure is combined with a fixed cost per unit distance to calculate travel cost
If multiple source cells, the lowest cost is typically placed in the output cell
Friction Surface (version of a Cost Surface)
The cell values of a friction surface represent the cost per unit travel distance for crossing each cell varies from cell to cellUsed to represent areas with variable travel cost.
Notes:Barriers can be added. Multiple paths are often not allowedCost and Friction Surfaces are always related to a source cell(s); from somethingThe center of a cell is always used the distance calculations
Digital Elevation Models (DEM)/Terrain Analysis
Terrain determines the natural availability and location of surface water, and hence soil moisture and drainage.
Water quality through control of sediment entrainment and transport, slope steepness and direction defines flood zones, watershed boundaries and hydrologic networks.
Terrain also strongly influences location and nature of transportation networks or the cost and methods of house and road construction.
Digital Elevation ModelsTerrain AnalysisSlope and Aspect
Used for: hydrology, conservation, siteplanning, other infrastructure development.
Watershed boundaries, flowpaths and direction, erosion modeling, and viewsheddetermination all use slope and/or aspect data as input.
Slope is defined as the change is elevation (a rise) with a change in horizontal position (a run).
Slope is often reported in degrees (0 is flat, 90 is vertical)
Slope (continued)Measured in the steepest direction of elevation change
Often does not fall parallel to the raster rows or columns
Which cells to use?
Several different methods:
Four nearest cells3rd Order Finite Difference
Elevation is ZUsing a 3 by 3 (or 5 by 5) moving windowEach cell is assigned a subscript and the elevation value at that location is referred to by a subscripted Z value
The most common formula:
Slope (continued)for ZoZ/x = (49 40)/20 = 0.45
Z/y = (45 48)/20 = -0.15
Slope (continued)Slope calculation base on cells adjacent to the center cellThe distance is from cell center to cell centerfor ZoZ/x = (49 40)/20 = 0.45
Z/y = (45 48)/20 = -0.15Generalized formula forZ/x and Z/y
Z/x = (Z5 Z4)2*Z/y = (Z2 Z7)2*
Using the four nearest cells
* = times cell width
Slope (continued)Z/x = (49 40)/20 = 0.45Z/y = (45 48)/20 = -0.15Kernal for Z/xKernal for Z/yMultiply (kernal, cell by cell)Add (results)Divide by #cells x cell widthUse slope formula
Multiply (kernal, cell by cell)Add (results)Divide by #cells x cell widthUse slope formula
Slope in ArcGIS 10From ArcGIS 10 Help
The orientation (in compass angles) of a slope
Aspect = tan-1[ -(Z/y)/(Z/x)]
As with slope, estimated aspect varies with the methods used to determine Z/x and Z/y
Aspect calculations also use the four nearest cell or the 3rd-order finite difference methods
Aspect in ArcGIS 10From ArcGIS 10 Help
ViewshedThe viewshed for a point is the collectionof areas visible from that point.
Views from any non-flat location are blocked byterrain.
Elevations will hide a point if they are higher than the viewing point, or higher than the line of site between the viewing point and target point
Shaded Relief Surfaces
From ArcGIS 10 Desktop HelpFlow Direction
High pass filters
Return: Small values when smoothly changing values.Large positive values when centered on a spikeLarge negative values when centered on a pit
High Pass FilterRaster data may also contain noise; values that are large or small relative to their spatial context.A mean kernal is used to reduce the difference between a cell and surrounding cells. (done by average across a group of cells)The identified spikes or pits can then be corrected or removed by editing
An area that contributes flow to a point on the landscapeWater falling anywhere in the upstream area of a watershed will pass through that point.
Many be small or large
Identified from a flow direction surface
A set of cells through which surface water flows
Based on the flow direction surface
Routing and AllocationRoutingFinding the shortest path between any nodes in a networkOptimal pathEach link in the net can also be assigned an impedance valueUsing an accumulated distance and the impedance factorMost efficient route can be found, rather than just the shortestNodes can also be coded with stops and barriersPreventing movement and forcing traffic along another pathAlthough routing can be done in raster, it is much easier if performed in a vector system
Routing and AllocationAllocation Process used to define the areal extent of services areasService areas are defined around a siteRegion is formed that includes a defined areaLocation/allocation model (optimizes network efficiency)Technique for the evaluation of multiple facility locations Determining the configuration of facilities (location)Assigning demand for the facilities (allocation)
****Incompatible cell sized and boundaries can cause problems with multi-layer raster operations. What cell value would you choose from layer 2 to combine with layer 1n and a portion of cell B is undefined.
Ambiguities are best resolved by resampling prior to layer combination*As with vector analysis, the scope of raster analysis can be classified as local , neighborhood or global
*****Raster reclassification assigns output values that depend on a specific st of input values.**Many differences between vector and raster overlay due to differences in the data models. Raster overlay is often restricted to nominal data. Overlay with continuous data creates to many unique cell combinations to be of much use,The output layer includes only source values indicated by the output layer.*Vector neighborhoods are not well-defined and depends upon the size and shape of the feature and neighboring feature, Raster neighborhood is better defined.
****This is an edge detection operation The first kernel amplifies differences in the x direction th