Hawth's Analysis Tools for ArcGIS

 You are here: Home > Hawths Tools > Tools Descriptions & Help

 Summary

LANDSCAPE CHARACTERIZATION (FAST ROVING WINDOW)
Input: a raster layer
Output: new raster layer(s)
Features:

• this tool is designed specifically to perform very large scale roving windows analyses on raster layers (thematic or continuous data)
• the user can specify multiple scales at once
• for categorical/thematic rasters, one output raster is created per category, per scale; the output value represents a count of values of that category within the roving window
• for numerical/continuous rasters, the mean, sum, and count can be calculated
• a count layer is also created that keeps track of the number of values that contributed to the summary statistics at each cell location
• for large roving windows, this tool is much faster than the built-in Spatial Analyst Neighbourhood Analysis functions

Limitations:

• this tool requires a considerable amount of RAM
• this tool, while fast, can still take many many hours (or days) to run depending on the size of the raster and the number of scales that are selected
• the output can require a great deal of disk space (sometimes many gigabytes)
• this program should be run on a computer that does not automatically reboot at night or after security updates
• this program shoud be run on a computer that can be completed dedicated to this one program
• always assumes the roving window is a circle (not a rectangle)

[Click for larger view]

 Help

Getting started. A raster layer must be loaded into ArcMap in order to use this tool. The raster layers in ArcMap are listed in the Numerical or Categorical boxes as follows: all floating point rasters are only listed in the Numerical box, all integer rasters are listed in both the Categorical and Numerical boxes. Select the rasters you wish to process by checking the box next to the raster name. Do not select the same raster in both the numerical and categorical boxes (not only is this not appropriate, it will also cause the program to fail).

Input raster data type. There are about a dozen raster data types that can be used to represent raster data. Of these, there are only four that can be represented as variables in Visual Basic. If you run this tool and receive an error message (an "automation error"), then you must convert the type of the input raster dataset. Use the Raster Pixel Type Conversion Tool to do this (refer to that tool for more information).

Scales / Radii. This program processes multiple scales (i.e. multiple roving window sizes) simultaneously. Type in a comma delimited list of roving window radii into the text box (e.g. 500, 1000, 2000, 3000). Note that although you can specify as many scales as you wish, that you must ensure you have enough disk space to support this number of output rasters (remember that an entire set of output rasters is created for each input grid, at each scale - this can be many gigabytes of data).

Process edge cells. Roving window analyses are complicated by the fact that for cells near the edge of the raster (i.e. at a distance less than the radius of the roving window) the sample size of cells within that window is less than the sample size for a window in the middle of the raster (i.e. sample size diminishes the closer to the edge you get). Depending on your application/analysis, it may be most appropriate not to process these cells in the roving window analysis. This is a statistical bias issue and you must make your own informed decision. On the other hand, for large window sizes, this may eliminate a large portion of your study area. You therefore have the option to process edge cells or not. Note that processing the edge cells will add considerably to processing time.

Numerical grid outputs. This tool is fast because of the coding mechanisms used (what I call the "slider approach"). Unfortunately, this approach prohibits the calculation of some metrics that are commonly calculated for numerical rasters, namely: minimum, maximum, range and standard deviation. What can be calculated is the sum, mean, and count. If you need the other metrics, you sould use the Neighbourhood Analysis tools in Spatial Analyst.

Output layers. Because of the number of output layers created by this program, the only output information you are asked to provide is the name of an empty folder. As the naming conventions in this program are automatic, it is very important that the folder is empty (to prevent any naming conflicts).