Purpose: Municipality x has recently acquired some SPOT 5 satellite data and is interested in determining its utility for monitoring the abundance (area in hectares) / location of land cover types in the region. You have been asked to deliver a report assessing what land cover types can be identified in the imagery. Your report should follow the format specified below. (Data is available for these two municipalities: Port Coquitlam, BC and Port Perry, the northeastern portion of the Greater Toronto Area. Select one of those municipalities for your report.)

OR

Purpose: Agriculture Canada has recently acquired some Landsat (ETM+) imagery (raw data description here) of an area near Brandon, MB, and is interested in determining its utility for monitoring the status of farmlands. You have been asked to deliver a report in which you identify the abundance (area in hectares) / location of different crops being grown in the area. (You need not identify the crop down to a specific type, but you will need to identify the different types that are present [e.g., Crop A, Crop B], along with identifying the other land covers present in the area.) Your report should follow the format specified below.

OR

Purpose: Montana's Department of Natural Resources and Conservaton is interested in determining the current status of the forests in an area located in northern Montana above the Libby Dam along the Kootenai River, using some Landsat 7 imagery (raw data description here). You have been asked to deliver a report in which you identify how much of the area in the Landsat image consists of recent clear-cuts (no timber standing), of older clear-cuts (logged areas that now have some vegetation growing in them), and of remaining forests. (Other land covers exist in the area, such as snow-capped mountains, river valley bottoms, clouds, etc. that you have to identify in your classified map, but in your tables that summarize the various logged areas you need not include them.) Your report should follow the format specified below. IMPORTANT NOTE: Included in the Montana ZIP file is the Landsat ETM+ panchromatic band (PROJMT_B8). Since this band has a different resolution than the other bands (15m vs 30m) you should not use it when performing your classification analyses.

Abstract: A brief summary of the purpose of the investigation and of the findings.

Introduction:

• Include a one paragraph description of the area covered by the image (Use Google Earth to provide some additional context, in addition to the images themselves).
• Include a one paragraph description of the imagery obtained and the sensor (either SPOT 5 [but the pixel size has been resampled to 20 m], Landsat 5 TM or Landsat ETM+]).
• A colour composite image to aid feature identification.
• A description of the colour composite image (bands used, features highlighted by the particular band combination).

Analysis:

Unsupervised classification:

• Perform a CLUSTER routine on the imagery using all of the available bands.
• Produce a histogram showing the number of pixels assigned to each class.
• Run the CLUSTER routine again using the "relevant" number of classes as determined from the histogram. (You may find that too many 'relevant' classes are identified [e.g., over 30], in which case you need to reduce the number of relevant classes. You can do this by changing the Cluster rule (Drop least significant clusters (e.g., the 1% rule) to something higher, or by selecting a broad versus a fine cluster. You should explore the different options to see which one produces what you consider to be the 'better' results.)
• Include an image showing the clusters of your second CLUSTER routine and update the legend to show a plausible assignment of land cover types to classes.
• Run the ISOCLUST routine using the appropriate number of classes (again, using all of the available bands).
• Include an image showing the ISOCLUST output, with the legend updated to show a plausible assignment of land cover types to classes.
• Discussion - comment on the ability of the unsupervised classification procedures to identify land cover types. Compare and contrast the output from CLUSTER and ISOCLUST and highlight areas of successful and unsuccessful classification.

Supervised classification:

• Include an image showing the training sites you delineated.
• After creating your signature files, use the SIGCOMP module to produce a figure showing the spectral separability of the training sites. Include this figure in your report along with a comment on the spectral separability of the classes.
• Run a supervised classification routine of your choosing, using all of the available bands. Justify your choice and include a description of the chosen statistical technique.
•  Include an image showing the resulting classification.
• Discussion - comment on the ability of the supervised classification procedure to identify land cover types. 

Conclusion:

• Success of classification efforts? What land cover types were clearly identified, which land cover types were clearly misclassified?
• Recommendations:
  • What classification routine would you recommend for land cover identification?
  • How could land cover identification be improved?

Notes:

• All images should include a north arrow, scale bar, title and legend (where appropriate)
• All projects must be individual projects--if we find that the content of two projects is identical or too similar to be a coincidence, you will, at most, only receive 50% of your mark.