Architecture M.Arch

Master in Architecture (GIS)

1. Details of Module and its Structure

Architecture M.Arch

Urban Landscape Design

Subject Name M.Arch – Architecture

Paper Name GIS

Module Name/Title Raster Analysis Part - 2 Objectives:

Objectives

 Process of GEO REFERENCING

 Method of converting from ‘R2V

 Digitization methods

 Updating data

Architecture M.Arch

Urban Landscape Design

3. 2. Development Team

Role Name

National Coordinator

Subject Coordinator Dr. Monsingh D. Devadas Paper Coordinator Dr. Pratheep Moses Content Writer/Author (CW) Dr. S. Vijaysagar Content Reviewer (CR) Dr. Pratheep Moses Language Editor (LE)

Architecture M.Arch

Urban Landscape Design

e-Text:

Raster Analysis:

Raster data (also known as grid data) represents the fourth type of feature:

surfaces. Raster data is cell-based and this data category also includes aerial and satellite imagery. There are two types of raster data: continuous and discrete. An example of discrete raster data is population density. Continuous data examples are temperature and elevation measurements. There are also three types of raster datasets: thematic data, spectral data, and pictures (imagery).

Digital Elevation Model (DEM) showing elevation.

Each cell contains one value representing the dominate value of that cell. Raster datasets are intrinsic to most spatial analysis. Data analysis such as extracting slope and aspect from Digital Elevation Models occurs with raster datasets. Spatial hydrology modeling such as extracting watersheds and flow lines also uses a raster-based system.

Spectral data presents aerial or satellite imagery which is then often used to derive

Architecture M.Arch

Urban Landscape Design

vegetation geologic information by classifying the spectral signatures of each type of feature.

Raster data showing vegetation classification. The vegetation data was derived from NDVI classification of a satellite image.

What results from the effect of converting spatial data location information into a cell based raster format is called stairstepping. The name derives from the image of exactly that, the square cells along the borders of different value types look like a staircase viewed from the side.

Unlike vector data, raster data is formed by each cell receiving the value of the feature that dominates the cell. The stairstepping look comes from the transition of the cells from one value to another. In the image above the dark green cell represents chamise vegetation. This means that the dominate feature in that cell area was chamise vegetation. Other features such as developed land, water or other vegetation types may be present on the ground in that area. As the feature in the cell becomes more dominantly urban, the cell is attributed the value for developed land, hence the pink shading.

Raster data mode:

– Grid: matrix

• Origin (top left corner)

• Number of lines

• Number of columns

• Pixel size: resolution

– To suit the resolution to the studied phenomenon

• A specific value for cells with unknown value (optional)

• Attribute table:

– The pixel value: integer or real number

Architecture M.Arch

Urban Landscape Design

– If necessary, other columns (any type)

– Transformation: image coordinates to map coordinates

Architecture M.Arch

Urban Landscape Design

Basic raster structure

The raster data model is quite easy to understand. Raster or cellular GIS datasets are constructed to represent spatial phenomena or geographic features within a framework of uniform cells. All cells in a raster dataset are the same size. Cells are organized into columns and rows that cover the extent of the area mapped. Each raster cell represents an area of the world and contains some value that represents information about that area.

Because a single raster layer typically represents a single theme, such as land use, soils, roads, streams, or elevation, multiple raster datasets are needed to fully depict an area.

Raster datasets are composed of equally sized cells organized into rows and columns.

Each cell is a square. Cell size is determined by the length of one cell side; the area of a cell is the cell size squared.

Raster data is generally divided into two types, thematic rasters (figure a below) and image rasters (figure b gelow). ESRI (2004) describes a thematic raster dataset as being like a map because it describes the features and characteristics of an area and their relative positions in space. However, there are no points, lines, or polygons, only cells. The cell values in thematic raster data represent some measured quantity or classification of a particular phenomenon, such as pollution concentration, population, or land cover.

Thematic raster datasets are usually stored in a single band 1, and cell values can be integers or real numbers.

Architecture M.Arch

Urban Landscape Design

Raster datasets are thematic (a) or image (b). Thematic raster data is stored in a single band and requires less processing than an image raster during analysis.

Rasters can be described by many other important characteristics. Rasters may be discrete or continuous. Discrete rasters represent features that have distinct boundaries.

Examples of discrete datasets include land use, habitat, roads, urban boundaries, and water wells. Most discrete rasters use integer numbers for feature identification, and generally the integer value represents a unique feature type. For example, in a land-cover map the value 5 may represent forest and the value 7 may represent water. Continuous rasters, on the other hand, do not have well-defined boundaries or have no boundaries at all. Examples of continuous data include noise levels, elevation, soil acidity, and seismic activity. Most continuous rasters use floating-point real numbers to represent feature characteristics. For example, cell values for an elevation raster (often called a digital elevation model, or DEM) might include 0, 10.5, 27.2, 2026, or anything in between or higher, depending on the surface topography. Raster values and other attributes are stored in the value attribute table (VAT). A thematic raster contains at least two items in its VAT, a value and a count, and may contain other user-defined attribute data. The value, as described above, represents some characteristic being mapped. Each value item constitutes a distinct zone. Say, for example, you have a raster that represents soil type.

One type is assigned a value of 4. All cells with a value of 4 (in other words, all areas with that particular soil type) belong to the same zone. The size of each zone is determined by the count item, which is the number of cells that share the same value. A set of contiguous cells with the same value (therefore from the same zone) is called a region. Zones can be composed of as many regions as necessary to represent a feature, and the number of cells that make up a region has no practical limit (ESRI 2004). The number of regions for a single zone depends on the connectedness of the cells in the zone. If the zone represents numerous disconnected areas, it has many regions. The connections used to define regions may be diagonal and perpendicular or just perpendicular. When diagonal cells are

Architecture M.Arch

Urban Landscape Design

excluded, only perpendicular connections are used, and the regions created are termed orthogonal (refer figure below).

Another unique characteristic of raster data is the method used for dealing with missing or unknown information. When a cell is vacant it is assigned the NoData value, indicating either no information or insufficient information exists about the location the cell represents. The NoData value, sometimes also referred to as the null value, is treated differently from any other value by all operators and functions in a raster GIS (ESRI 2004).

Although NoData represents a cell for which no cell value is recorded, don’t assume that every cell containing a NoData value does not contain useful information; in fact, the absence of information in many cases may be just the information you are looking for.

What is important about NoData for ESRI rasters is the fact that unless specifically requested, NoData always remains NoData. That means any application combining or using one or more ESRI rasters will retain NoData values in the output raster. For example, when multiplying two rasters using the Times tool, any NoData values from either of the input rasters are retained as NoData in the output raster. This is because NoData implies there is no information for the input cell, so generating a real value for the output cell is always incorrect.

Cells with the same value make up zones. Contiguous cells of the same zone make up regions. Regions may be created using the Region Group tool, using either diagonal or

orthogonal connections.(Soil data courtesy of the USDA.)

Architecture M.Arch

Urban Landscape Design

NoData can be a very powerful value. For example, it is effective to use NoData to identify areas where the GIS analyst does not wish to compute real values. Suppose that you are generating a suitability surface for residential development potential. There is absolutely no way you will be able to build residential housing in existing conservation lands. Since the conservation lands are protected by the government (state, federal, or local), you might assign NoData values to the areas identified as conservation land before proceeding with suitability analysis. The assignment of NoData to areas the analyst wishes to remove from consideration is called masking, and the raster containing the masked areas is called a mask raster.

Why use raster GIS for models?

Why should you use raster GIS for analysis when your data is stored in vector format? That’s a typical, and good, question. Whether to use raster or vector analysis techniques really depends on the analyst’s preference. Complex analysis operations are often more easily accomplished using raster GIS, which may convince an analyst to convert her vector data to raster format. Of course, this choice requires computer processing time and disk space, the ability to manage the additional raster layers, and an understanding of raster analysis tools and concepts. All of this requires additional energy and knowledge if you are primarily comfortable with vector GIS concepts, data, and tools.

However, to answer complex questions with vector GIS requires the user to continually manage, combine, aggregate, and disaggregate multiple layers to form complex polygons or networks. Usually the combination of geospatial data using vector overlay tools is more time-consuming than in raster GIS. Raster models, on the other hand, have the ability to simultaneously evaluate cells from multiple raster datasets representing the same location, rapidly accomplishing complex spatial analysis and overlay. With this ability, raster analysis facilitates the incorporation of complex mathematical expressions in a model. For example, a search of multiple rasters for cells with elevation values above 500 feet, next to steep slopes with open water below, and within 4 miles of a major roadway, would assist an architect in locating a scenic overlook for a client’s new building, and can be written as a complex raster expression that will solve the problem quickly.

Architecture M.Arch

Urban Landscape Design

Combining vector layers can be a complex task, but with raster analysis, it’s easy to overlay cells from many rasters at the same time.

Raster datasets have clear advantages for overlay analysis. Database size can be greatly reduced when datasets are stored as cells, rather than as polygons, because of raster compression techniques. Additionally, we maintain that modeling with Map Algebra, the language for Spatial Analyst that allows you to write expressions for processing multiple raster layers, is easier, faster, and more flexible than using vector overlay techniques. And, the ability to develop suitability models with raster data is more efficient for complex

analyses where

many layers are required. While much of the source data for the LUCIS case study was originally in the vector format, the data was converted to raster during preprocessing or early in analysis models.

Basic raster analysis tools:

The following sections discuss the benefits of raster analysis and highlight some commonly used raster tools within the Spatial Analyst Tools toolbox, all of which are employed in the LUCIS models. As with vector GIS, raster-based analysis is

accomplished by the programmatic use of basic functions. These functions typically fall into one of three conceptual categories: local, focal (or neighborhood), and zonal.

Architecture M.Arch

Urban Landscape Design

Local functions act on raster data one cell at a time until all cells in the raster have been processed. They may process many rasters simultaneously. Local functions are most commonly used to calculate mathematical equations (simple and complex) between rasters. Local functions may be conditional; they may be used to extract values from a raster, to reclassify values in a raster, or to combine information from two or more rasters.

The output from a local tool is a new raster with a single value or multiple values assigned to each cell. The output cell values are determined solely by the value of the correlating cells at the same location in the input and the process performed (adjacent input cells are not considered when calculating the output cell value). Basic addition, subtraction, multiplication, or division of a raster using a constant value or another raster is accomplished using simple local functions. A more complex local function is reclassifying data by assigning new values to raster cells. An example would be the reclassification of habitat data into a more simplified wetland/upland classification. Reclassifying habitat cells in a raster into wetland and upland areas can be accomplished by setting the value of the output raster to 1 for all wetland cells in the input raster, and setting the output value to 3 for all upland cells in the input raster. Complex queries can also be performed using local functions; for example, a local tool could be used to query many rasters to determine cell locations that are suitable for particular uses.

Local functions act on thematic or image rasters one cell at a time. Each output cell value depends on the value of the corresponding input cell value and the process performed.

Architecture M.Arch

Urban Landscape Design

The Plus, Minus, and Less Than tools perform simple mathematic operations on two or more input rasters.

Math tools perform mathematical analysis of raster data. The Math tools reside in their own toolset in the Spatial Analyst toolbox. All Spatial Analyst tools in the Math toolset are considered simple local functions. The above figure shows three examples of local functions performed with Math tools: (1) addition using the Plus tool, (2) subtraction using the Minus tool, and (3) a simple expression, based on a constant, using the Less Than tool (found in the Logical subset of the Math toolset). The Plus tool adds two rasters together or adds a constant value to a raster using single cell-by-cell processing. The Minus tool performs the same cell-by-cell operation for subtraction. The Less Than tool provides a Boolean analysis that returns a value of 1 (true) in the output raster where the first raster is less than a constant or a second raster, and returns a value of 0 (false) where it is not.

Conditional tools perform analysis based on satisfying a condition or set of conditions. The following figure illustrates the Con tool, one of three tools in the Spatial Analyst Tools >

Conditional toolset in ArcToolbox. The Con tool processes raster data by selecting cells

Architecture M.Arch

Urban Landscape Design

based on “if-thenelse” processing. The Con tool may process multiple rasters by checking the conditional statement and picking cells from the appropriate raster for the output raster.

In the figure, if any cell in GRID1 equals 1, then the output cell in the same geographic location is given a value of 1. Otherwise, the output cell is given the value found in the coincident cell in GRID2. The Con tool may be nested in a model or nested for complex conditional statements using the Single Output Map Algebra tool.

Conditional raster processing using the ‘Con’ tool. The result of the conditional statement is shown in the output raster. If the cell value in GRID1 is equal to 1, then transfer that cell value into the output raster’s coincident cell; otherwise, transfer the cell value from GRID2.

Extraction tools, found in the Spatial Analyst Tools > Extraction toolset, extract values from a raster based on attributes of the raster or characteristics of a second control raster, a polygon, or bounding coordinates. The following figures show two extraction tools, the Extract by Attributes tool and the Extract by Mask tool. The Extract by Attributes tool selects cells for the output raster if they meet the selection criteria for an attribute in the input raster. It requires that the raster data be in integer data format, not floating-point data. The Extract by Mask tool selects cells for the output raster that have valid values in the input control raster.

Architecture M.Arch

Urban Landscape Design

The Extract by Attributes operation transfers only the cell values that fulfill the query to the output raster. All other cells are assigned a NoData value in the output.

The Extract by Mask operation transfers to the output raster only the cell values in GRID2 that are not NoData cells in GRID1.

Local tools, like other local functions, are cell-specific. Local tools perform cell statistics (majority, minority, mean, median, sum, variety, maximum, and minimum), combine raster values, perform frequency analysis, determine raster position (highest and lowest), and determine raster rank and popularity. Figure shows one of the tools found in the Spatial Analyst Tools > Local toolset: the Cell Statistics tool, used to collect statistics for coincident cells within multiple rasters. The Cell Statistics tool assigns an output value for each cell statistic created from all input cell values in a set of rasters (ESRI 2004). For example, the tool can calculate the mean and standard deviation for all coincident cells in a set of rasters. There are two tools in particular that should be highlighted, as you will see them used often in the LUCIS models. These are the Reclassify tool and the Single Output Map Algebra tool, both of which reside in unique toolsets, and both of which are considered local functions.

The Reclassify tool (found in the Spatial Analyst > Reclass toolset) is used to reassign raster values in order to create new categories. There are many reasons to reclassify data.

For example, soils data may be reclassified into nine levels of suitability for construction.

It’s much easier to examine data that has been grouped into a digestible number of categories rather than be faced with a potentially enormous number of unique values. The Reclassify tool will be covered in greater detail in chapter 4, Logical arguments for suitability, as a method for creating raster suitability layers.

The Single Output Map Algebra tool (Spatial Analyst > Map Algebra toolset) is used to write single line equations with Map Algebra expressions. Map Algebra is the analysis language for ArcGIS Spatial Analyst that enables users to build simple or complex expressions and process them as a single command (ESRI 2004). The Single Output Map

Architecture M.Arch

Urban Landscape Design

Algebra tool allows modelers to add complexity to models without requiring the use of multiple tools in a more lengthy sequential analysis. For example, the following equations may be processed as one operation by the Single Output Map Algebra tool:

CON(GRID1 GE 30, GRID3, CON(GRID3 LE 100, GRID3, GRID2)) CON(GRID1 EQ 9, GRID1, (GRID3 * 0.5 + GRID4 * 0.35 + GRID 5 * 0.35))

In the first equation on the previous page, the CON tool is used to transfer the cells from GRID3 to a new output dataset where the cells in GRID1 are greater than or equal to 30, otherwise a nested CON tool is used to transfer the cells in GRID3 that are less than or equal to 100, otherwise all other cells in the output are assigned the value of the cells in GRID2. The resulting output raster dataset has cell values from only two input rasters,

Architecture M.Arch

Urban Landscape Design

GRID3 and GRID2, but part of the cells selected for the output are controlled by values in GRID1.

In the second equation, the CON tool is used to transfer cell values from GRID1 that are equal to 9, otherwise the output cell values are the weighted values for GRID3, GRID4, and GRID5. This statement produces an output raster with data values from four individual raster datasets that are a weighted combination of three of the rasters with a value (9) for the fourth raster. The Single Output Map Algebra tool is very as a method for weighting multiple raster suitability layers in the LUCIS project, and for complex conditional selection of raster data between multiple raster datasets.

Architecture M.Arch

Urban Landscape Design

Focal functions process data using neighborhoods. The output cell value is affected not only by the correlating input cell value, but also by the values of surrounding cells, which compose the neighborhood. It’s this behavior that differentiates focal functions from local functions, which only consider the correlating input cell when calculating the output. A neighborhood boundary is measured by cells, map units, or geometric shapes. For example, a neighborhood defined as 4 cells by 3 cells is a rectangular neighborhood 4 cells high and 3 cells wide. The output from a focal tool is a raster where the individual cells have summary information from the input raster gathered from the moving neighborhood. For example, the variety of habitat around a specific cell in an input habitat raster can be calculated by counting the types of habitat within its neighborhood; then, the neighborhood moves and types of habitat around the next cell are calculated, and so on until an output raster is produced where each cell value reflects the total number of different habitats inside that cell’s neighborhood. Understanding focal functions is easy once it becomes clear how the neighborhood moves across the input raster layer. As shown in the figure below, the neighborhood starts in the upper left corner of the input raster layer and summarizes information for the cells located within the neighborhood.

Once the focal function has finished summarizing data for that location, the neighborhood shifts one cell to the right and the function again summarizes information for cells within the new neighborhood. The process continues until all cells in the input raster layer have been visited by the moving neighborhood.

Moving 3-by-3 neighborhood method. The neighborhood moves from left to right one cell at a time. Focal functions summarize data from the input raster within the extent of the

moving neighborhood to provide values for the output raster.

Neighborhoods come in many shapes: rectangular, circular, wedge, and annulus (doughnut shaped). A neighborhood is used to calculate the majority, minority, sum, mean, median, standard deviation, range, and variety. Neighborhoods can be standardized to act as filters for high- or low-pass analysis. The low-pass filter analysis is for averaging, while

Architecture M.Arch

Urban Landscape Design

the high-pass filter analysis is for edge detection. The wedge filter is used to provide direction influence, for example, wind in a fire model. The tools in the Neighborhood toolset of the Spatial Analyst toolbox are focal functions. Additionally, some of the tools in the Surface toolset use neighborhoods, and therefore are also focal functions.

our potential neighborhood shapes: the circle, wedge, annulus, and rectangle.

Neighborhoods are used to collect information from an input raster within the moving boundary. The information collected using the neighborhood is placed into an output raster

one cell at a time.

Zonal functions

Every cell in a raster belongs to a zone. Some rasters have only a few zones, while others have a lot. Recall that all cells with the same value belong to the same zone. A zone can consist of cells that are connected, disconnected, or both. Recall that areas in zones where the cells are connected are referred to as regions. Assemblages of large entities, such as soil types in a county, or wetlands in the state, are phenomena that are most likely to be represented by zones constructed of disconnected cells. Zonal functions summarize data from a raster by using specified zones from a control raster or polygon feature class.

For example, by using zones from a habitat raster layer, a zonal tool could calculate the minimum, maximum, mean, median, and standard deviation of elevations for each habitat and store the results in a tableor new raster.

All Zonal Statistics tools use zones from a raster or vector polygon dataset to calculate statistics for those same zones determined by data values in a second raster. The output from the Zonal Statistics as Table tool (Spatial Analyst Tools > Zonal toolset) is a table, not an output raster. The table contains summary statistics such as minimum, maximum, mean, and standard deviation from the input raster for the zones found in the control layer.

For example, the table in following figure shows the statistics for the distance from major roadways collected for agricultural cropland use within Alachua County, Florida.

Architecture M.Arch

Urban Landscape Design

The Zonal Statistics as Table tool produces a tabular output. In this case, it displays statistics regarding distance from major roadways. The control layer is agricultural

croplands.

Architecture M.Arch

Urban Landscape Design

Displaying the Ithacamos DEM (for example sake)

We will be using the Ithacamos to demonstrate many of the raster analyses. This file contains elevation information for the Ithaca region in 10m x 10m cells. (How do we find this out?) Elevation values are in meters and the coordinate system is UTM18N NAD27.

This DEM is added to the map view in the same way other data are added. When you open the Ithacamos you will probably see something like the first figure below.

However, the look of the DEM will depend on the symbolization that is being used. In this case, I’m using a continuous symbolization (Stretched) where the elevations are shown in a gray scale from high elevations in white to low elevations in black (below). You may also use a classification approach where you group elevations into different colors.

Architecture M.Arch

Urban Landscape Design

The slope function calculates the maximum rate of change from every cell to its neighbors.

The function is calculated over a 3x3 set of cells and can yield slope in angular degrees (0- 90) or in percent, which is a measure of vertical rise over horizontal run. To create a slope map from the Ithacamos DEM:

Go to ArcToolbox|Spatial Analyst Tools|Surface|Slope. You will be presented with the window below.

Architecture M.Arch

Urban Landscape Design

 Identify the input surface, which is the Ithacamos.

 Select degree for the output measurement.

 Keep the Z-Factor at the default. The Z Factor is used as a conversion when the elevation units are different from the X-Y coordinate units.

 IMPORTANT – Click on the little folder icon next to Output raster and identify a location to save the new slope raster file. Make the raster grid name short and a single word (the name must not exceed 13 characters and cannot start with a number).

 Your new slope grid will be displayed in the map view.

Architecture M.Arch

Urban Landscape Design

Aspect identifies the slope direction in compass degrees (0=north, 180=south, etc.). As was the case with slope, the calculation is based on a 3x3 grid neighborhood. To create an aspect grid from the Ithacamos DEM: Go to ArcToolbox|Spatial Analyst Tools|Surface|Aspect.

Aspect is measured clockwise in degrees from 0 (due north) to 360, (again due north, coming full circle). The value of each cell in an aspect dataset indicates the direction the cell's slope faces. Flat areas having no downslope direction are given a value of -1.

 Identify the input surface, which is the Ithacamos.

 Identify a location and name for the output raster.

 After clicking OK, a map showing aspect will be added to the map view. The default symbology for this map uses colors for aspect direction. Keep in mind that the raster grid cells contain the actual aspect direction measure.

Hillshade

Hillshade allows us to determine the illumination of a surface (the DEM in the case) given a direction and angle of a light source (i.e. the sun). The resultant grid contains values ranging from 0-255 with 0 representing complete darkness. To calculate,

 Go to ArcToolbox|Spatial Analyst Tools|Surface|Hillshade.

Architecture M.Arch

Urban Landscape Design

The azimuth is the angular direction of the sun, measured from north in clockwise degrees from 0 to 360. An azimuth of 90 is east. The default is 315 (NW).

The altitude is the slope or angle of the illumination source above the horizon. The units are in degrees, from 0 (on the horizon) to 90 (overhead). The default is 45 degrees.

 Select the input surface (Ithacamos).

 Set the direction (azimuth) and angle (altitude) of the light source.

 Identify an output location and filename for the hillshade raster.

 When you click OK you will see a map similar to that below.

Architecture M.Arch

Urban Landscape Design

Note how the hillshade grid gives you a 3-D feel for the landscape. In fact, we can enhance this effect by using the transparency tool with the DEM and hillshade grids. To do this:

 Move the DEM to the top of the view window and make sure both it and hillshade grids are turned on.

 Open the “Effects” toolbar from the GUI|Toolbars.

 In the effects toolbar, set the layer to the DEM and click the transparency button.

Slide the transparency bar to 50%. You will now be able to see through the DEM (but not completely) to the underlying hillshade. (More effects, such as flicker, dim and contrast, are also available.) (You can also do this by right-clicking on

ithacamos DEM, Properties. Under the Display tab, change the transparency to 50%.)

Architecture M.Arch

Urban Landscape Design

A viewshed allows you to determine which areas on a landscape can be seen from a feature, such as a point location. This calculation is based entirely on the elevation and does not include trees, buildings, etc. As such, it is limited.

Open a new map view and add the Ithacamos and the CornellTree shapefile. Find the fattest tree in the database. (Hint: It’s a pignut hickory.) Export out the single point to its own shapefile.

Then go to ArcToolbox|Spatial Analyst Tools|Surface|Viewshed

Architecture M.Arch

Urban Landscape Design

 Select the DEM input surface and the observer point shapefile. The Earth Curvature is useful if working over very long distance (not the case in this example).

 Leave the Z factor as the default.

 Set the output raster name and click OK. You will be given a window like Fig. 12.

The green areas indicate locations that are visible from the viewpoint. The actual grid values are 0=not visible and 1=visible.