2. The aim of this research project is:
To identify potential of Google Earth in land use change
detection and detect disadvantages and advantages of standard
land use/ land cover change detection.
The objectives of this research project are:
• To collect land use data from Google history imagery in 2001
and 2011.
• To compare and evaluate two different remote sensing
methods.
• To compare the land use for different years.
• To identify the most appropriate remote sensing method to
use for the classification of land use zones.
3. GOOGLE WORKFLOW LANDSAT
EARTH IMAGERY
2001 IMAGERY 2011 IMAGERY IMAGE
DIGITALIZATION DIGITALIZATION
PREPARATION
SUPERVISED
KML FILE CLASSIFICATION
ACCURACY
ASSESMENT
GIS ANALYSIS
WITH ArcGIS 10
LAND
LAND RESULTS OF ANALYSIS IN COVER
USE MAPS TABULAR AND GRAFHIC
MAPS
FORMAT
4. 2001 2011
1 . Digitizing polygons with Google Earth PANORAMIO PHOTO
polygon tool for 2011
2. Using Google Earth database for land use
detection (photos , 3d models, street view...)
3. Detecting changes for each polygon in 2001
and changing class if necessary
3D MODEL 2D IMAGE
4. Reshaping polygons whose land use is
changed
5. KML export to ESRI SHP file
5. RA - single Mixed
family RA - multi RA - Sport Unmaintaine Construction Agricultural - Agricultural - agricultural- Parks - green
houses family houses features Industrial d sites croplands unmaintained urban areas Rangeland Forest land
-0.75(km²) 2.25(km²) -0.27(km²) 3.27(km²) 0.03(km²) 0.04(km²) -5.64(km²) 0.06(km²) 0.14(km²) 1.49(km²) -0.67(km²) -0.09(km²)
6. 4:3:2 BAND COMBINATION
• Deciding on Land cover classes
• Image pre - processing
SUPERVISED CLASSIFICATION
• Training stage
• Supervised Classification -
ACCURACY ASSESMENT
Maximum Likelihood Classifier
ACCURACY ASSESMENT
• Accuracy Assessment – Error matrix
7. User Accuracy 2001 User accuracy 2011
Accuracy % % Class %
Agricultural 93 Urban 90
Forest 93 Agricultural 95
Urban 85
Forest 100
Water 75
Water 100
Overall 89
Overall 93
8. Supervised classification (2001-2011)
Land use/cover class Change (km2) % -Change
Urban or built up 25.3 7.21
Agricultural -11.31 -3.22
Forest -12.87 -3.67
Water -1.12 -0.32
8.00
% -Diffrence Landsat
6.00 %- Diffrence Google
4.00
2.00
0.00
-2.00
-4.00
Google Earth (2001-2011)
Land use/cover class Difference (km2) %- Change
Urban or built up 4.59 1.34
Agricultural -4.63 -1.32
Forest -0.09 -0.02
Water 0 0
9. DISCUSSION & CONCLUSION
• Results show a similar pattern in both analyses
• Urban areas are increasing and agricultural areas are decreasing
• The increase is evident, but Zagreb is still not over developed City
Google Earth Analysis Landsat Imagery Supervised
Classification
• Very useful tool for Land use
• Much faster process, but less
analysis, but with minor bugs
detailed
when converted to SHP files
• Google database (photos, 3D
• Misplaced pixels of water in
models, street view...) can be of
forested and shadow areas
great help for land use
determination
• Time consuming, not suitable for • Combination of both methods
larger areas would give even better results