Google Earth Survey

19 December 2014

Introduction

This coding assignment is part of a project looking at the landscape effects of foreclosure in Maricopa County (Phoenix), AZ. We are using Landsat Normalized Differential Vegetation Index (NDVI) satellite data to detect houses where there was a change in landscape management that coincided with foreclosure on that house. Possible scenarios (among many) include:

You will be coding a sample of 200 lots based on the amount of landscape management change that you observe using historical imagery in Google Earth. We will be using your coded data to build a model that relates the outputs of our Landsat analysis to what actually happened on the ground.

100 of the lots are actual foreclosures and 100 are a control group of randomly selected lots and dates. To avoid bias, they are mixed together and you cannot know which are which.

The historical imagery in Google Earth is quite limited. There are only a small number of dates available (which varies depending on location) and some of the imagery is of very low quality. As such, in most cases you will not be able to make a coding determination with absolute certainty. Use your best judgment and that will be fine.

Instructions

  1. Click on this Google Earth KML file link to open the map of parcels in Google Earth
  2. The label for each parcel contains a parcel ID number and a date when the foreclosure occurred
  3. Click on this Spreadsheet link to open the spreadsheet of parcels
  4. Get a range of parcels to code from Michael
  5. For each parcel in your range of parcels, observe Google Earth historical imagery within one year prior to the foreclosure date and one year after the foreclosure date. You will often need to look further back or forward if the imagery is unclear or imagery is unavailable.
  6. For each parcel, fill in these three columns in the spreadsheet based on your observations:
    • Change:
      • -2 = major decrease (1/2 or more of Potential Lawn Area)
      • -1 = minor decrease (less than 1/2 of PLA)
      • 0 = no observable change
      • +1 = minor increase (less than 1/2 of PLA)
      • +2 = major increase (1/2 or more of PLA)
    • Certainty:
      • A: high - good imagery within one year before and after
      • B: medium - imagery good only one year before OR after
      • C: low - inadequate quality both before AND after
    • Comments (optional)
  7. When you have completed all your parcels, save the spreadsheet to your local hard drive and e-mail a copy to the following addresses. E-mail a copy to yourself as well in case there is some problem with your e-mail to us.

    illinois@michaelminn.com,bcutts@illinois.edu

  8. Log your time and submit that to Bethany

Details

  1. Changes are coded positive or negative based on observed additions or removal of managed vegetation, regardless of size.
  2. If inadequate imagery is available at high enough quality or within one year before/after the date, use the nearest available scenes
  3. Mixed changes (such as removal of turfgrass and addition of desert plants in xeriscaping) are coded as positive or negative based on the whether a greater area of change was estimated to be positive or negative.
  4. Changes that appear to be caused by maturation of vegetation unrelated to observable management change (like the growth of trees in yards that exhibited no other changes) are coded as no change.
  5. Transient changes in vegetation that appear to be related solely to climate events (based on similar changes on surrounding parcels) are coded as no change.
  6. Lots containing areas for pasture or agriculture where vegetation fluctuates with weather or planting cycles should only be coded as changed if there is a clear difference between the management of those fields before and after the date.