SPATIO-TEMPORAL URBAN CHANGE DETECTION, ANALYSIS AND PREDICTION OF KATHMANDU VALLEY

SPATIO-TEMPORAL URBAN CHANGE
EXTRACTION AND MODELING OF
KATHMANDU VALLEY
SUPERVISORS :
Asst. Prof. Nawaraj Shrestha
Er. Uma Shanker Panday
8/3/2014 Department of Civil and Geomatics Engineering 1
PROJECT MEMBERS:
Dhruba Poudel
Janak Parajuli
Kamal Shahi

CONTENTS
1. INTRODUCTION
2. OBJECTIVES
3. SCOPE OF PROJECT
4. METHODOLOGY
5. OUTCOMES
6. LIMITATIONS AND RECOMMENDATIONS
7. CONCLUSION

1. INTRODUCTION
 Spatial extension of the cities in temporal dimension
 Continuous process all over the world BUT showing more effects on developing
countries
 Universal socio-economic phenomenon occurring world wide, Nepal not an
exception
 urban system is considered as the complex system having characteristics of:
• Non-determinism and tractability
• Limited functional decomposability
• Distributed nature of information and representation
• Emergence and self-organization
URBANIZATION

BACKGROUND
 Half of the world's population would live in urban areas by the end of 2008
(UNFPA 2007)
 By 2050, 64.1% and 85.9% of the developing and developed world respectively
will be urbanized (UNFPA 2007)
 Hence urbanization is skyrocketing

Figure 1.Nepal as fast growing urban area (Source: - UN-HABITAT Global Observatory)

Fig 2. (A) and (B) Urban growth around Bouddhanath Area
(A) Is 1967 satellite image from CORONA
(B) Is 2001 IKINOS satellite image
Source: HABITAT INTERNATIONAL(www.elsevier.com/locate/habitatint)

PROBLEM STATEMENT
Kathmandu among fastest growing city in the world.
Limited information on city growth and urbanization patterns.
Limited quantitative information on urban growth rate and direction
Need of informed decision making tool based on which future
strategic plan and action can be made to counterpart fast urban growth.

2. OBJECTIVES
 To detect, analyze and visualize the extent of spatial-temporal urban
growth based on multi-temporal Landsat Satellite imagery.
 To quantify the spatial-temporal pattern of urban growth and
landscape fragmentation using spatial metrics.
 To simulate or forecast the urban growth of the study area using
SLEUTH model.

3. SCOPE OF PROJECT
This research is attempted in order to:
 Extract the urban area of the Kathmandu valley over different time scales,
 Quantify that urban extent,
 Analyze the changeover difference time periods and
 Predict the future scenario of the urbanization considering the factors affecting the urban
growth
Using following applications:
 Remote sensing
 Geographic Information system (GIS)
 FRAGSTATS to calculate Spatial metrics
 SLEUTH model using Cellular Automata (CA) as UGPM

4. METHODOLOGY
Kathmandu is the capital city of Nepal and also one of the fastest growing cities of Asia.
This valley is bounded approximately within 27° 32' 00" N to 27° 49'16" N and longitude
85°13'28" E to 85°31'53" E (UTM coordinate system) covering the area of approximately 58 sq.
km.
The population of valley is more than 2.5 million and has population density of 129,250 per sq.
km
a. Project Area
Figure 3. Project Site(Thapa & Muriyama, 2010)

S.N. Sensor Date of
Acquisition
Resolution Source WRS Sun Elevation
(degrees)
Sun Azimuth
(degrees)
1 Landsat 5 1989-10-31 30*30 USGS website 141/04100 41 144
2 Landsat 7 1999-11-04 30*30 USGS website 141/041 42.98952434 152.67113676
a. Landsat TM
b. Data Used

S.N. Data Layers Year Projection System Website
1 Contour - WGS 1984 geoportal.icimod.org, accessed on 2014-06-15
2 Landuse 1978 & 1995 WGS 1984 geoportal.icimod.org, accessed on 2014-06-15
3 River - WGS 1984 geoportal.icimod.org accessed on 2014-06-15
4 Road 2010 WGS 1984 geoportal.icimod.org, accessed on 2014-06-15
5 Spot height - WGS 1984 geoportal.icimod.org, accessed on 2014-06-15
6 Kathmandu
Boundary
- WGS 1984 geoportal.icimod.org, accessed on 2014-06-15
b. Geographic Data layers

S.N. Software Use in the Project
1 ENVI  Used for image pre-processing, index-based image processing, supervised classification, accuracy
assessment and confusion matrix calculation, image differencing
2 ESRI’s ArcGIS  To prepare data for spatial metrics, store classified data, visualize them and prepare map
 Accuracy assessment using GCPs
 Used to prepare raster data for SLEUTH
 Process model output
3 FRAGSTATS  To quantify the landscape pattern
4 Map Source  Create and view waypoints along routes and tracks
 To deal with gpx format file
 Accuracy assessment of classified binary map
5 SLEUTH model  To predict future urban growth
6 PC-Pine  Edit scenario files to execute SLEUTH model
7 Cygwin  Used as Linux emulator to run SLEUTH model
8 Others  Expert GPS, Google Earth, GPS Visualizer used for various purposes.
 Photoshop and Paint used to create gray scale 8 bit image in GIF format
13
d. Software and instruments Used
8/3/2014 Department of Civil and Geomatics Engineering

e. Overall Work Flow
Figure 4. Work Flow
Image preprocessing
Landsat Image
Accuracy Assessment
Signature Extraction
Image Classification
Classified Map
No
Yes
ReferenceData
Multi-temporal
growth maps
Quantify landscape
Pattern
Analyze and forecast
Urban growth
Spatial metrics SLEUTH Modeling
Multi-temporal
Classified
Map
Final outcomes
1989
2014
2009
1999

METHODOLOGICAL:WORK
FLOW
1. RS IMAGE CLASSIFICATION
AND ANALYSIS
2. QUANTIFY URBAN GROWTH
PATTERN USING SPATIAL
METRICS
3.CHANGE DETECTION
4. PREDICTING URBAN
GROWTH PATTERN USING
SLEUTH MODELING
1.1 Landsat TM Image acquisition
1.2 Image Preprocessing
 Image calibration
 Atmospheric Correction
 Topographic Correction
1.3 Index images generation
 Normalized Difference Built-up Index:
NDBI=(MIR-NIR)/(MIR+NIR)
 Soil Adjusted Vegetation Index:
SAVI=(NIR-Red)(1+L)/(NIR+Red+L)
L is constant 1>L>0
 Modified Normalized Difference Water Index:
MNDWI=(Green-MIR)/(Green+MIR)
 Index based Built-up Index(IBI)
IBI=[NDBI-(SAVI+MNDWI)/2]/[NDBI+(SAVI-
MNDWI)/2]
Click here to see sample index images
contd…
1.4 Signature Extraction via Region of Interest
 Built-up ROIs
 Non-Built up ROIs
1.5 Supervised Image Classification
 using maximum Likelihood Algorithm
 Classified into two classes i.e. Built and Non-Built
1.6 Accuracy Assessment
 Confusion Matrix
i. Using Ground Truth ROIs in ENVI
ii. Using GPS sample points in GIS
 Visual Interpretation
i. Google earth Overlay
ii. Openstreet map Overlay
iii. Combined Overlay with GPS sample
points
1.7 Multi-Temporal Image analysis

PATTERN
 Spatial metrics is used to quantify the dynamic
patterns of landscape so will be used to quantify the
urban growth
 Fragstats software was used
 Three categories of metrics were calculated
 Patch metrics
 Class metrics
 Landscape metrics
 Nine types of parameters were calculated
i. Class Area(CA) vi. Edge density(ED)
ii. Number of patches(NP) vii. Cotagion(CONTAG)
iii. Patch density(PD) viii. Shannon’s Diversity
Index(SHDI)
iv. Largest Patch Index(LPI) ix. Shannon’s Eveness
Index(SEVI)
v. Area Weighted Mean Patch
Fractal dimension (AWMPFD)
METHODOLOGICAL:WORK
FLOW
AND ANALYSIS
METRICS
3.CHANGE DETECTION
4. PREDICTING URBAN
SLEUTH MODELING
1999
2009
1989
2014

3.CHANGE DETECTION
2.1 Image differencing of multi-temporal
classified image
2.2 Post classification comparison in GIS
METHODOLOGICAL:WORK
FLOW
AND ANALYSIS
METRICS
3.CHANGE DETECTION
4. PREDICTING URBAN
SLEUTH MODELING

4. PREDICTING URBAN GROWTH PATTERN
USING SLEUTH MODELING
 SLEUTH Stands for Slope, land use, exclusion,
urban extent, transportation and hill shade and
consist of urban modeling module and land cover
change transition model
Click here to see model inputs
 Uses five controlling coefficients of growth to
simulate the change
i. Dispersion : simulates spontaneous growth
ii. Breed: simulates new spreading center
iii. Spread : simulates edge growth
iv. Road Gravity : simulates road influenced growth
v. Slope : determines the effect of slope on the
probability of pixel being urbanized
 Model validation
METHODOLOGICAL:WORK
FLOW
AND ANALYSIS
METRICS
3.CHANGE DETECTION
4. PREDICTING URBAN
SLEUTH MODELING

5. OUTCOMES

a. Remote Sensing Image Classification

Analyzing Multi-Temporal Image with respect to present road NetworkURBAN MAP 1989

1.Confusion Matrix
Calculated via two methods:
 Providing Region of Interests(ROI) of classified image classes in ENVI
 Using Arc GIS’s combine and pivot table tools using input Ground control Points(GCP)
of classified image area and classified image of that date.
Results from confusion matrix:
Year Kappa Coefficient Overall Accuracy
(ROI methodI) (GCP method) ROI method GCP method
1989 0.89 0.87 90.02% 89.28%
1999 0.85 0.84 87.11% 85.61%
2009 0.88 0.86 89.87% 87.48%
2014 0.91 0.89 93.21% 89.77%
b. Accuracy Assessments

2. Visual Interpretation
i. Google earth Overlay
ii. Openstreet Map Overlay

Year CA NP PD LPI ED LSI
Non-
Built Built
Non-
built Built
Non-
Built Built
Non-
Built Built
Non-
Built Built
Non-
Built Built
1989 57411.36 873.99 52 1606 0.0892 2.7554 98.4721 0.3181 11.5943 8.8128 7.0482 43.2374
1999 56159.64 2125.71 140 3417 0.2402 5.8625 96.2464 0.8488 23.3956 20.6244 14.3842 65.0487
2009 52905.42 5379.93 1118 3735 1.9181 6.4081 88.8658 6.5222 37.582 34.8108 23.7992 69.1534
2014 49025.61 9259.74 2694 6735 4.6221 11.5552 81.3187 11.4145 66.6682 63.9392 43.8477 96.7477
1. CLASS METRICS
c. Quantification of Classified Image
 Increase in urban class area(CA) from 1989-2014 with increase in number of patches(NP)
 Increased number of patches indicating landscape fragmentation
 Fragmentation is high relative to the urban growth resulting increase in patch density(PD)
 Largest patch index, edge density are also in continuous trend of increasing for built-up class
CAN ANAALYZED FURTHER WITH THE HELP OF FOLLOWING GRAPHS:

2 . L A N D S C A P E M E T R I C S
Year TA NP PD LPI ED LSI FRAC_AM
CONTA
G PR PRD SHDI SHEI
1989 58285.35 1658 2.8446 98.4721 11.6046 7.0019 1.1913 90.778 2 0.0034 0.0779 0.1123
1999 58285.35 3557 6.1027 96.2464 23.411 14.1255 1.2586 81.1899 2 0.0034 0.1566 0.2259
2009 58285.35 4853 8.3263 88.8658 37.5974 22.6851 1.2921 65.2776 2 0.0034 0.3078 0.4441
2014 58285.35 9429 16.1773 81.3187 66.7048 40.2475 1.3455 48.1171 2 0.0034 0.4378 0.6316
Besides the metrics discussed above, FRAC_AM, CONTAG, SHDI, SHEI descries the complexity
of the patches
Which all are increasing for built up class, increasing the complexity of the landscape patches

3 . PAT C H M E T R I C S
Sample of patch metrics

d. Change Detection

0
100
200
300
400
500
600
700
800
1989-1999 1999-2009 2009-2014
125.172
325.422
775.962
Change Area(Ha/year)
1989-1999
1999-2009
2009-2014
1989-1999 1999-2009 2009-2014
growth rate 2.14 5.58 13.33
0
2
4
6
8
10
12
14
Growthrate(%)
growth rate
 Growth rate is increasing in very high rate
 Growth trend suggests that it will further increase for some decades
 Present growth rate is sufficient to double the urban area of valley in less than 15 years
 Migration, population growth, transportation development and many other new projects
on valley tends to increase more urban growth rate

e. SLEUTH Modeling Click here for animation
1. Comparative probability map

SPATIO-TEMPORAL URBAN CHANGE DETECTION, ANALYSIS AND PREDICTION OF KATHMANDU VALLEY

 Figure 1 shows the dominance of growth coefficients over different time period and
fluctuation in the coefficients
 Fluctuation is due to self modification functionality of model
 Figure 2 suggests the rapid growth up to 2022 and decrease in growth rate
2. Comparative analysis of coefficients of model and probable urban area

8/3/2014 Department of Civil and Geomatics Engineering
35
3. Coefficient based probability map

Types of Growth Patterns in the valley
1. Spontaneous Growth2. New Spreading Centre3. Edge growth4. Road Influenced Growth

Types of Growth observed
Infill Development

Edge expansion

Outlaying Development

a. Limitations
 Image classification is binary classification to built-up and non-built up only (not land use
mapping)
 Quantification is based on the binary classified map so spatial metrics are calculated on the
basis of only those landscape class
 Change detection is overall class based but not patch oriented
 Prediction of model is totally based on the factors supported by SLEUTH model
 Political condition, socio-economic and demographic factors lacks even they are the major
factors of urban growth)
6.LIMITATIONS AND RECOMMENDATION

 Use of high resolution image enhances better extraction of built-ups
 Land use classifications of landscape may be more informative than binary classification
 Patch based analysis could have detect the process urban growth trend precisely
 OSM over leesalee metrics could make made model more robust
 SLEUTH-3r would have counter the some of the limitations of SLEUTH model
b. Recommendation

7. CONCLUSIONS
 Index based Supervised classification of Landsat TM images can be used for
built-up extraction
 Urban Growth rate of Kathmandu is skyrocketing (from 2.14%-13.315 during
1989-2014)
 Spatial metrics can be used for quantification of landscape to analyze the trend
of urban growth rate and pattern
 Probability map of SLEUTH model is suitable for Regional level of planning
and policy formulation.

Only the matter is “HOW it comes???”

THANK YOU
For any detail: kamal.shahi502@gmail.com

SPATIO-TEMPORAL URBAN CHANGE DETECTION, ANALYSIS AND PREDICTION OF KATHMANDU VALLEY

More Related Content

What's hot (20)

Similar to SPATIO-TEMPORAL URBAN CHANGE DETECTION, ANALYSIS AND PREDICTION OF KATHMANDU VALLEY (20)

Recently uploaded (20)

SPATIO-TEMPORAL URBAN CHANGE DETECTION, ANALYSIS AND PREDICTION OF KATHMANDU VALLEY

Editor's Notes