4.3.4 Change Detection Method
The change detection method that used in this study is post-classification change
detection. LULC information of each year was extracted to detect changes. To fully
understand the change occurred in recent three decades, multi-temporal (in 6-year
interval) and time-serial change analyses were performed. The logic of change
detection analysis and some major outputs are illustrated in Fig. 4.5.
In order to minimize the vegetation phenological effect and periodic cultivation
cycle of agriculture in suburban and rural area, green cropland, fallow, and grassland were combined into one category, which is called “vegetated area”. Forest land
was not combined into the new class because forest is important natural resource
that needs to be considered individually. Other classes also remained the same.
Then LULC information was extracted for change detection analyses. Analyses
were conducted both qualitatively and quantitatively. Multi-temporal analysis has
detected changes based on classification maps in 6-year interval, which were 1984,
1990, 1996, 2002, 2008, and 2013 classification maps. Time-serial analysis is used
in all images to illustrate the trajectories of each LULC type change dynamics over
these three decades. Various graphics and tables have been created to help interpret
and analyze the results shown in Fig. 4.5. For example, change maps demonstrating
LULC
classification
maps (8 classes)
Classes combination
LULC
classification maps
(6 classes)
Multi-temporal change
detection analysis (84, 90, 96,
02, 08, 13)
Bar chart of
multi-temporal
changes
Statistics of the
changes
Bar chart of timeserial LULC
trajectories
Time-serial change
dynamics analysis (all
images)
Input data
Process
Output data
Workflow direction
Fig. 4.5 Workflow chart of change detection analysis
4 Long-Term Change Dynamics Using Landsat Archive for the Region of Waterloo. . .
77
The change detection method that used in this study is post-classification change
detection. LULC information of each year was extracted to detect changes. To fully
understand the change occurred in recent three decades, multi-temporal (in 6-year
interval) and time-serial change analyses were performed. The logic of change
detection analysis and some major outputs are illustrated in Fig. 4.5.
In order to minimize the vegetation phenological effect and periodic cultivation
cycle of agriculture in suburban and rural area, green cropland, fallow, and grassland were combined into one category, which is called “vegetated area”. Forest land
was not combined into the new class because forest is important natural resource
that needs to be considered individually. Other classes also remained the same.
Then LULC information was extracted for change detection analyses. Analyses
were conducted both qualitatively and quantitatively. Multi-temporal analysis has
detected changes based on classification maps in 6-year interval, which were 1984,
1990, 1996, 2002, 2008, and 2013 classification maps. Time-serial analysis is used
in all images to illustrate the trajectories of each LULC type change dynamics over
these three decades. Various graphics and tables have been created to help interpret
and analyze the results shown in Fig. 4.5. For example, change maps demonstrating
LULC
classification
maps (8 classes)
Classes combination
LULC
classification maps
(6 classes)
Multi-temporal change
detection analysis (84, 90, 96,
02, 08, 13)
Bar chart of
multi-temporal
changes
Statistics of the
changes
Bar chart of timeserial LULC
trajectories
Time-serial change
dynamics analysis (all
images)
Input data
Process
Output data
Workflow direction
Fig. 4.5 Workflow chart of change detection analysis
4 Long-Term Change Dynamics Using Landsat Archive for the Region of Waterloo. . .
77