methods overview

The map series are created based on Landsat TM5 imageries by using the object-based backdating classification method. Updating/backdating methods have been increasingly used for long-term classification and change detection analysis over large areas. This approach greatly improves efficiency by identifying and classifying only the changed areas based on a reference map. In this project, the images from 1985 to 2005 are normalized using the 2010 image as the reference. The 2010 map is first created using the supervised object-based method (eConigtion software). Large training samples of different land use/land cover types are selected based on the 1-m 2010 map to ensure high accuracy of the 2010 30-m map. Then, images from 1985 to 2005 are segmented and the change vector analysis is preformed to detect the changed areas. Multiple thresholds are used to identify changes between different classes. Next, a set of rules are created to classify the changed area into individual land use/land cover classes. Land use maps and other GIS layers from the Maricopa Association of Governments are used as ancillary data in this step. Finally, post-classification analysis is preformed to correct the misclassified areas and improve the map consistency. Stratified random sampling is used in the accuracy assessment.

raster value categories

Class ID. Class Name: Description

1. Water: This class includes lakes, canal and other small water bodies.
2. Asphalt/Road: This class mainly identifies transportation lines and other paved surface covered by asphalt, such as parking lots and commercial centers. This class doesn't include most of the road segments within a residential block.
3. Concrete/Buildings: This class mainly identifies commercial and industrial buildings. It also includes concrete covered areas.
4. Urban mixture: This class identifies mixture of constructed materials, fragmented vegetation and paved surface.
5. Residential: This class includes various dwellings such as neighborhoods, apartments, and rural lots.
6. Residential (white rooftops): This class identifies cluster of residential buildings that have bright (usually white) rooftops, such as aluminum roofed house trailers and foam-based flat roofing.
7. Active crop: This class identifies croplands that covered by green vegetation.
8. Inactive crop: This class identifies croplands that don't have green vegetation.
9. Cultivated vegetation: This class includes urban green space such as golf courses and parks.
10. Natural vegetation: This class includes vegetation grown in desert or riparian area, such as shrubs.
11. Soil/Desert: This class includes bare soil, desert, and any undeveloped or open land.

accuracy assessment: Kappa

Class ID. Class Name: Kappa %

1. Water: 0.94
2. Asphalt/Road: 0.82
3. Concrete/Buildings: 0.89
4. Urban mixture: 0.89
5. Residential: 0.88
6. Residential (white rooftops): 0.91
7. Active crop: 0.97
8. Inactive crop: 0.92
9. Cultivated vegetation: 0.94
10. Natural vegetation: 0.97
11. Soil/Desert: 0.88

accuracy assessment: producer's accuracy

Class ID. Class Name: Producer's accuracy %

1. Water: 100
2. Asphalt/Road: 87.5
3. Concrete/Buildings: 85.7
4. Urban mixture: 85.7
5. Residential: 81.1
6. Residential (white rooftops): 100
7. Active crop: 100
8. Inactive crop: 87.7
9. Cultivated vegetation: 95.1
10. Natural vegetation: 97.5
11. Soil/Desert: 94

accuracy assessment: users's accuracy

Class ID. Class Name: User's accuracy %

1. Water: 95
2. Asphalt/Road: 83.3
3. Concrete/Buildings: 90
4. Urban mixture: 90
5. Residential: 89.5
6. Residential (white rooftops): 92.5
7. Active crop: 98
8. Inactive crop: 93.4
9. Cultivated vegetation: 95.1
10. Natural vegetation: 97.5
11. Soil/Desert: 90.9