Instructor: Brian Klinkenberg

Office: Room 209
Office Hours: Tues 12:30-1:30
Wed 12:00-1:00

Lab Help: Jose Aparicio

Office: Room 240D

Computer Lab: Room 239


 

 

Lab 2: Exploring Fragstats
In this lab you will explore how the landscape around London, Ontario (Google Map) has changed over time (1966 to 1976). You will do this by examining some FragStats metrics (some I've identified for you, others you'll select by yourself), as well as by creating a transition matrix that explicitly shows the transformation of land uses over time. The report you are to prepare should be written from the perspective of a consultant (who could be an urban planner, a pro-agriculturalist, an environmentalist, an outdoor enthusiast, etc.--your interpretation of the results would reflect your particular perspective on the direction / nature of the changes you observe).

You must first download the Canada Land Use Monitoring Program (CLUMP) data for London from the Geogratis web site for (i.e., LONON66u.zip and LONON76u.zip). Unzip the files, and store the data in C:\data\Lab2\. (Reminder: Lab 1 contains the full details on how to find the data.) A description of the land uses documented in CLUMP (e.g., what is included in 'outdoor recreation' lands) is available here. Don't forget: ArcMap Etiquette

Start ArcMap. Within ArcToolbox, import the interchange file (LONON##u.e00) (Conversion Tools / To Coverage / Import from E00). In the Import window, save the file in C:\data\Lab2\ as LONON## [where ## is the year of your CLUMP file] and leave the File type as Basic types. It may appear that the import process fails, but if you click on the Add file button [+] you should be able to find the coverage file in C:\data\Lab2\.

In order to exclude some of the unmapped area from our analyses, we can set the Processing Extent such that most of the unmapped areas are excluded in the vector-to-raster conversion. So, before doing the raster conversion below, set the GeoProcessing / Environments... / Processing Extent to these values:

Top
4780000
 
Left
449000
  Right
510000
  Bottom
4730000
 

In order to explore what impact spatial resolution has on the results, when converting the vector files to raster files produce two rasters (one for 1966, one for 1976) with a resolution of 100m and, for 1976 only, another raster with a resolution of 250m (i.e., you will have three raster files in the end). (ArcToolbox / Conversion Tools / To Raster / Polygon to Raster) (Don't forget to specify "USE" as the Value field in your conversion.) As Fragstats cannot read ArcMap 10.x raster / grid files you will have to export each raster to a GeoTIFF file (as described in Lab 1) before doing your Fragstats analyses.

In order to assist you in your interpretation of the Fragstat results, it is useful to explicitly state what CLUMP land uses are associated with each raster class. In order for Fragstats to know what the classes actually refer to (e.g., that class 1 represents Cropland), you need to create a class descriptors file for each year. This is a text file (with the extension fcd) that lists ID, Name, Enabled, IsBackground, where ID refers to the grid value, Name refers to the associated CLUMP land use, Enabled is a binary variable [true or false] that indicates whether that class should be included in the analyses [true] or not [false], and IsBackground is another binary variable [t or f] that tells Fragstats to either include that class in the analyses [false] or to treat that class as a background class [true]. There should not be any blank lines in the fcd file (if so, the program will not run).

Note that the 1976 raster land use map has an additional blank class not present in 1966 (i.e., there is no 'use' associated with some cells). You should provide a label for those blank cells in the fcd file--give them a name of 'unknown'.

The easiest way to create the class properties file is to open the README.TXT file that comes with the CLUMP data and to cut / paste the CLUMP land use codes into a blank text file (note that you will have to remove any commas "," associated with the land use names). Then, using the information contained in the raster attribute file (Open the Attribute Table in ArcMap; you need to replace the CLUMP codes in the README file with the ID's from the raster attribute table) you can create the class properties file. Here is an example of the steps / files you should use in creating your FCD file (note that the order of the entries doesn't matter, and that unmapped areas should be considered as the background class). Read over the Fragstats help file on creating the class descriptors file for further details (open help and search for fcd). It is important that the syntax rules for the class properties file are followed (that is, each field is followed by a comma, and there are no commas elsewhere in the file), or you will run into problems. Once you have created the fcd file, Save As the file by explicitly setting the extension to .fcd.

(Note: When producing your final maps you can import the land use codes (extracted from the readme.txt file) into ArcMap, and then join that table to your vector or raster layers in order to have the land use names displayed in the legends.)

To add the fcd file to your Fragstats analysis, you need to point to the file in the Common Tables section.

Start Fragstats, and create a New file:

  • select GDAL GeoTIFF grid as the Data input dataset type,
  • select one of your three gridded (TIFF) files as the input file,
  • select the appropriate fcd file
  • specify an approriate edge depth (e.g., 100m for the 100m resolution files, 250m for the 250m resolution file)
  • under the Analysis parameters tab, check that 'Automatically save results' is selected,
  • provide an appropriately named output file name (such as LONxx100), and
  • select Class and Landscape metrics [under Multi-level structure].
  • select the output statistics for the Classes and the Landscape, as described below.

Accept the program defaults for all of the other values. Note that you have to run Fragstats three times, once for each raster file, since the FCD file is specific to each raster file.

For the purposes of this lab you will only be examining a few selected metrics for each of the three grid files:
Class metrics (Area or Aggregation: Total Area, Percentage of Landscape, Number of Patches, Total Edge) and
Landscape metrics (Area or Aggregation: Number of Patches, Patch Density; Diversity: Shannon's Diversity Index, Shannon's Evenness Index).

Select the Class and Landscape metrics from their respective windows.

IMPORTANT NOTE: You need to select two additional Class metrics and two additional Landscape metrics and, in your written report, you must describe those four additional statistics (why you choose them and what the statistical values showed). If you read over the Fragstats help file (PDF) and, in particular, go through the documentation that provides an overview of each group of metric (e.g., Area and Edge, Aggregation), you should be able to identify some relevant metrics to employ. (If you start Fragstats, and then click on Help (Help Contents), and then go to the Contents / Fragstats Metrics / Overview, and read over the group descriptions, this should help.)

Set the Class and Landscape metrics accordingly. Run Fragstats.

Using Excel you can take the output files from Fragstats (open the *.class and *.land files as comma-delimited text files) and produce a number of plots. You must label the land uses using the appropriate CLUMP descriptors. You can then compare the results of your analyses and see how the landscape around London has changed over time, and how changes in spatial resolution affect the results.

You should ignore any results associated with those classes that have fewer than 25 cells, and the results associated with the one class that has no CLUMP code associated with it.

Transition Matrix

You will now create a transition matrix that shows how the land uses in 1966 changed over time (i.e., what uses did they become in 1976?). In order to do this we can use the 'combine' tool [ArcToolbox / Spatial Analyst Tools / Local / Combine] (use the 100m ArcGIS raster files). The result of this operation will be a raster that contains the matches between the two input rasters (that is, the attribute table shows you what a cell was classified as in1966 and what it was classified as in 1976) (an illustration of what combine does--taken from the ESRI Help file for Combine--where InRas1 and InRas2 would correspond to your two land use rasters). In order to make a transition matrix you need to export the attribute table (open the attribute table of our combined result and Select Export under Table Options) as a DBF, and then open that DBF file in Excel. Check to ensure that the dbf is being exported to C:\Data\Lab2 (read the notes below before exporting the file).

IMPORTANT NOTE: Do not store the results of the combine operation in a Geodatabase. If you do the attribute table will not contain the two 'extra' columns that contain the links to the original raster files (the join fields).

Before exporting the table, however, we should add (JOIN) the actual USE codes to the combined table. Select the combined raster, and select Joins and Relates -- Join. You will be joining a table--select the field associated with the 1966 raster as the join field (the field name should match the name of your 1966 100 m resolution raster). Select the attribute table from the 1966 raster, and select VALUE as the join field. Click on Okay. Look at the attribute table of the combined raster to ensure that the join worked properly. Repeat the join process, but this time select the 1976 raster. You should now export the attribute table.

Import the file into Excel (you'll need to set the file type to *.* in order to select the DBF file). You should tidy up the file a bit by removing the extraneous columns. To create the transition matrix you need to create a Pivot Table (Microsoft's description). Highlight the three columns you'll need to summarize [Count, showing the number of cells associated with each pair of land use codes, and the two USE columns], and then click on Insert -- Pivot Table. For the column labels select the 1976 raster values, for the row labels select the 1966 raster values, and select Count as the Σ (Sum of) field.

Once the pivot table is created, we need to convert the raw numbers into percentages, and provide proper row/column labels (e.g., changing the use codes (e.g., 8) to meaning names (e.g., Unmapped areas)). You should also delete the blank ('unknown') row / column. To make the calculation of the %'ages easier, you should copy the entire pivot table and then paste 'values' to a new worksheet. You should include this final (%) table, and a discussion of what it shows, in your report. (A question to ponder: What percentage values to calculate? That is, you could calculate the percentages based on the row totals [showing how a land use in 1966 changed over time], based on the column totals [showing where a land use in 1976 came from], or based on the sum total. We will talk about this in class.)

To be handed in

A 3-4 page report (excluding the tables and figures) on the results of your analyses. Use tables / graphs to demonstrate the changes that you observed in the landscape metrics--both as a result of changing the spatial resolution (only one paragraph on this), and as a result of the change in the land use around the area of London between 1966 and 1976 (this discussion should form the bulk of your report). In your report you must not only present the results but also explain what each metric means.

The report should be written from the perspective of a consultant hired by the city council to examine the changes in the land use around the area of London, ON (therefore, include a cover page [report title, your name, the date] and an Executive Summary [a short paragraph] as the second page. Your report should include maps showing the land uses in the two years and, using blow-ups of an area in order to highlight the differences, a map that illustrates the impact that a change in scale has on the data (you only need a paragraph discussion on the impact that the change in resolution had on the results). Ensure that you use a consistent colour scheme for all of your land use maps (i.e., a land use on all three maps should have the same colour). Two of the maps should show the entire study area, one for each time period (can be produced using the vector maps [import the legend to ensure that they correspond]), while the other map should show the effect that changing scale had on the data (therefore you must use the raster maps).(Here is an explanation of how to create a map with two zoomed-in raster maps as well as an inset map highlighting where the zoomed-in maps are located relative to the entire area.)

The results of this lab will be due in two weeks--Jan 29th, at the beginning of the lab. You should include tables showing the Fragstat statistics for the three grids (combining the three landscape metric outputs into one table), the transition matrix table, and several graphs highlighting some of the interesting statistics.

Ritters et al. (PDF) explored the relations among 55 of the landscape metrics provided by FRAGSTAT and found that there are very high correlations among many of them ( in fact, those 55 metrics statistically only represented 6 different factors). [Riitters et al., 1995. A factor analysis of landscape pattern and structure metrics. Landscape Ecology 10:23-39.] [freely available from this web site]

Two other papers of interest:

Hansen, J. A. G. 1984. Canadian small settlements and the uptake of agricultural land, 1966–1976. Social Indicators Research 15(1): 61-84. (Link)

Muller, M. R. and J. Middleton. 1994. A Markov model of land-use change dynamics in the Niagara Region, Ontario, Canada. Landscape Ecology 9(2): 151-157. PDF   [freely available from this web site]