Land Use / Land Cover Change in the Phoenix Metropolitan Area
1984 - 2011Lori Krider & Melinda Kernik
1984 2011
Introduction
• Why Phoenix?o One of 10 fastest growing cities from 1990 -
2000 (Perry & Mackun, 2001)o Arid regions with high population are water
stressedo Water use is reflected by how the land is
used and managedo How is the landscape changing and how
does this effect water use?
Objective
• Use remote sensing software to assess land use / land cover change in Phoenix from 1984 – 2011o Expect to see dramatic changes due to rapid
population growth Increase in urban and suburban areas (sprawl) Increase in cultivated areas on edges of
metropolitan area Decrease in natural vegetation
Objective
• Study Areao Phoenix-Mesa
Metropolitan Area South-central
Arizona 16,200 km2
Phoenix, Mesa, Tempe, Chandler, Gilbert, Scottsdale, Glendale, Sun City, Peoria, and Avondale
Google Maps
Preparation
• Tools: ERDAS IMAGINE 2011, USGS GLOVIS, ArcGIS 10, Google MapsTM and Google EarthTM
• Materials: Landsat TM images from 1984 and 2011 (two from each year, 30 m res., 7 bands, June), 2006 NLCD
• Pre-classification processingo Stack bands, mosaic and crop images for each yearo View NLCDo Unsupervised classification (5, 6 & 7 classes)
Analysis• Supervised classification
o Anderson Hierarchical Classification (levels 1 and 2) Altered, unaltered, developed and water
Altered Human-assisted: healthy and stressed crops, golf
courses Uncultivated: fields not reflecting in IR Unaltered Natural: upland and scrub/shrub (not in IR) Hydrophillic vegetation: depressional vegetation often
associated with water (in IR) Water: lakes, rivers and large golf course water hazards Developed suburban (dwellings)
& urban/roads (commercial/industrial)
Analysis
• Training Areaso 15 - 45o Why?
Errors in first run with less training areas Combination of smaller category classes (i.e. healthy
crop + stressed crop) Reduce confusion and capture variety
• Change Detectiono Thematic: 1984 -> 2011o Difference
to identify areas of significant change and overall patterns
10, 20, and 30% thresholds
Post-classification• Accuracy Assessment
o stratified randomo same mosaics as reference
added Google MapsTM for 2011o switched "trainers"o 140 reference points (20 per class)
http://www.cartoonstock.com/directory/b/bad_appraisal.asp
1984 2011
Purple: Change to SuburbanLight Blue: Change to Urban
Thematic Change
Detection
1984
2011
Purple = changed to SuburbanBlue = changed to Urban
Green = more than 20% increase in NIRBlue = more than 20% decrease in NIR
Thematic Change Detection
1984 2011Limitations!
Accuracy Assessment
For future classifications:
• Clip to the smallest possible boundaries– More ontological classes = more classification
confusion
• Complications using 30m resolution images for reference data and the same image.
• Use this technique, to generate water infrastructure policy for Phoenix …probably not
References
1. Perry, M. J. & P. J. Mackun. Population Change and Distribution 1990 - 2000: Census 2000 Brief. April 2011. United States Census Bureau. 12 Nov. 2011. <http://www.census.gov/prod/2011pubs/c2kbr01-2.pdf>.