Modelling and Prediction of Land Use Changes in Jodhpur City...

S.L. Borana et al., International Journal of Research in Engineering, IT and Social Sciences, ISSN 2250-0588,

Impact Factor: 6.452, Volume 07 Issue 11, November 2017, 4-21

http://indusedu.org Page 14

This work is licensed under a Creative Commons Attribution 4.0 International License

Modelling and Prediction of Land Use

Changes in Jodhpur City using Multi-

Layer Perceptron Markov Techniques

S. L. Borana1 and S. K. Yadav

2

1,2(RSG, DL, Jodhpur-342011, Rajasthan, India)

Abstract: Land use change models techniques are effectively used in prediction of land use dynamics of the urban development. Landsat satellite data of 1990, 2000 and 2010 were used to analyse the current and predict

future land use growth of Jodhpur city. The supervised classified images were used for future prediction of land

use change and generation of projected map (year 2020 and 2030). Multi Layer Perceptron Markov

(MLP_Markov) Model and IDRISI, Land Change Modeler (LCM) were used to analyze the land use and land

cover changes between various classes during the period 1990-2010. Using multi-layer perception, the five

classes identified as built -up area, vegetation, mining area, water body and other area. The prediction of land

use change was prepared on neural network built-in module and its validation.

Keywords: Remote Sensing, Landsat Data, Land change modeler, Predicted Map.

I. INTRODUCTION Unplanned fast development of cities causes troubles to urban surroundings and natural environment. A

scientific awareness of urban expansion trends and possible urban development management should be achieved

by remote sensing temporal data. Satellites images are the resources for exposure, quantification, and land use

mapping (Abel El-kawy et al., 2011).The base data generated from remote sensing is used for spatial modeling

of urban growth and its future prediction. Modelling studies have started using cellular automata and Markov

chain (CA_Markov) together to give spatial measurement to Markov chain model which is fragile in spatial

face. The CA_Markov has an capability to predict any transition with any number of categories. Thapa and

Murayama (2011) use LCM for Nepal urban development modelling. They model years of 2020, 2030, 2040,

and 2050 by observing and analyzing satellite images of 1991, 2001, and 2010 and by using historical,

environmental, conservational scenarios. Sefyaniyan (2009) evaluates Isfahan land use changes (1987-1998) and

finds significant changes in agricultural lands. Recently, combination of Multi-layer perceptron (MLP) Neural

Network and Markov chain modeling technique with Land Change Modeler (LCM), is used in future prediction of urban growth by many researchers for various cities.

Study Area

Jodhpur also known as sun city, is centrally situated in western region of the Rajasthan state with

location at 26ºN 18' latitude and 73º E 04' and at an average altitude of 224m above mean sea leve (Fig.-1). l. In

general, the contours are falling from North to South and from North to Southeast with maximum level of 370m

and minimum of 210m. The present population is about 1.05 million and admeasures 230sq.km. Alongside,

Jodhpur has been functioning as one of the engines powering the Rajasthan state economy.


Impact Factor: 6.452, Volume 07 Issue 11, November 2017, Page 14-21



Figure1: Study Area

II. DATA USED AND METHODOLOGY Remote Sensing (RS) data scene (1990, 2000 and 2010 FCC & TCC) were used for analysis and

interpretation of land use types in the study area. The Coordinate System use for projection of satellite images

are WGS-1984 and UTM Zone-43N. with 30m spatial resolution were used in this study. The SoI maps, ground

truth data were used for land use classification and accuracy analysis. Ancillary data, such as a digital elevation

model (DEM), major road networks were also included into the analysis. ArcGIS and ERDAS Imagine software

were used to achieve land use classification mapping in a multi-temporal approach. Image classification is carried out by using the supervised classification maximum likelihood method. Five land use types i.e. built -up

area, vegetation, mining area, waterbody and other area have been identified in this study. For future prediction

of land use change, Multi Layer Perceptron Markov (MLP_Markov) Model has been used for modelling of land

dynamics. IDRISI Selva software with Land Change Modeler used in this paper for analysis of land use

changes. Methodology adopted for this research is shown in the Figure 2.





Figure2: Methodology Flow Chart

III. RESULT AND DISCUSSION Future Prediction and Analysis: MLP_Markov model has been selected for simulating land cover map of study

area for the year 2020 & 2030. Various steps involved in prediction procedure are creating Boolean and

Distance Images, generation of Land Cover Transition Image, Selecting Driving Variables, and Testing Potential of the Driving Variables, Transition Potential Modeling, Markov Chain Analysis and Analysis of the

Predicted Map.

Creating Boolean and Distance Images: Boolean images for land cover type of Built-up, Mining area,

Vegetation & other are and driver have been prepared using linear type fuzzy membership function available in

IDRISI; a value 255 indicates the highest suitability and a value 0 indicates the lowest suitability of that

particular category. The values 1 represent the areas of interest and the values 0 represent the areas of no

interest (Fig.3).

Figure3: Boolean Images of each Land Cover Type





Distance images for each of these Boolean land cover images and drivers have been generated. These

distance images are important to measure the suitability values for the pixels of land cover classes. The distance

images are produced using simple Euclidean distance function which measures the distance between each cell

from the featured image. The lowest and highest values obtained from the distance images have been used as the input for fuzzy set membership analysis (Fig.4).

Figure4: Distance Images of each Land Cover Type (2010)

Land Cover Transition Image: The basic concept of modelling with MLP neural network is to consider

the change in built-up area over the years. Transition from all land cover types to built up has been produced,

considering the transitions from all other land cover types to only built up area(Fig.5). Other changes have been

ignored. Fig.6. shows the land cover type is contributing more to net change in built-up area. It is found that

other area is contributing most converting towards built-up area followed by vegetation.

Figure5: Transition from All to Built-up Area (2000-2010)

Figure6: Contributors to Net Change by Built-up Area (Unit: per cent of Area)





Selection of Driving Variables: The issue of which variable affects the change to built-up area (2000-

2010) has been considered at this stage. Therefore, distance from all to built-up area has been chosen (Fig.7).

Another important aspect is to find out the empirical likelihood of all changes for transforming into built-up

areas based on the base image of the year 2000 (Fig.8). The highest value 0.80 is showing a high probability of converting other land cover types to built-up area. Therefore, 5 driver variables have been selected for the

model. These five driver variables are distance from all to built-up area, distance from water body, distance

from vegetation, distance from other area and empirical likelihood image.

Figure7: Distance Image of Transition from All to Builtup Area. (2000-2010)

Figure8: Empirical Likelihood Image (2000-2010)

Testing Potential of the Driving Variables The quantitative measures of the variables have been tested through Cramer’s V. It is suggested that

driving variables having Cramer’s V of about 0.15 or higher are useful. Table1 shows that the potential

explanatory values of the driving variables are useful (Cramer’s V > 0.15).

Table1: Cramer’s V of the Driving Factors

Driving Variable Class Overall V Built-up Other area Vegetation Water Body

Distance from All to Built-up 0.0843 0.000 0.1501 0.3013 0.5604

Distance from Other Area 0.1637 0.000 0.3869 0.3023 0.4903

Distance from Vegetation 0.1604 0.000 0.3987 0.3010 0.4591

Distance from Water Body 0.1250 0.000 0.1310 0.6502 0.2596

Empirical Likelihood Image 0.6601 0.000 0.8831 0.7134 0.4935





Transition Potential Modeling The MLP running statistics gives a very high accuracy rate of 99.29% (Table-2). The minimum number

of cells that transitioned from 1990 to 2000 is 4794. The RMS error curve has also been found smooth and

descent after running MLP neural network (Fig.9). It means the training result is satisfactory. Based on these running statistics four transition potential maps have been produced. The dominance of water body and fallow

land to built-up area type is clear here (Fig. 10).

Table2: Running Statistics of MLP Neural Network

Figure9: RMS Error Monitoring Curve

Figure10: Transition Potential Maps

(1- Built up area, 2- Other area, 3- vegetation, 4-Water body)

Markov Chain Analysis: Using this MLP neural network analysis, determine weights of the

transitions is included in the matrix of probabilities of Markov Chain for future prediction. The transition

probabilities are shown in Table-(3 &4).The final predicted map of 2020 and 2030 have been simulated through

Markov chain analysis based on all these pieces of information from MLP neural network (Fig.11 &12).

Table3: Transition Probabilities Grid for Markov Chain (2020)





Table4: Transition Probabilities Grid for Markov Chain (2030)

Table5: Land Cover Area of Base map and projected map

LU/ LC Area (km2) Projected Area (km2)

2010 2010 2020 2030

Built up 114.42 119.69 167.52 206.74

Other area 368.37 359.45 312.66 276.03

Vegetation 86.36 84.19 80.65 74.79

Mining area 23.38 29.32 31.74 34.83

Water body 1.47 1.35 1.43 1.61

Total 594 594 594 594

Figure11: Change in Area (%) over the Years (1990-2020)

Analysis for Prediction Map: The predicted map of 2020 and 2030 reveals that 28 % and 34.8% of the

total area will be occupied by built-up area cover type respectively. On the other hand, water body and other

area (fallow land) types are going to decrease. Similarly Gains in built-up area land cover type are prominent

while most of the areas will be persistent (Fig.12). Slight loss in water body and vegetation cover types will also

be found. But other area will decrease in good amount in near future. Long term predictions are not possible by applying the MLP_Markov model. It is giving errors in predicting land cover maps for long terms as the urban

growth rate is very high in case of Jodhpur city. It means that the study area will reach its threshold limit in

built-up area type by 2020 and 2030. Therefore, CA_Markov model is better option for long term prediction for

this research.

Figure12: MLP_Markov Projected Land Cover Map (2020 and 2030)





IV. CONCLUSION This study demonstrates different models to simulate the land cover change map of Jodhpur city. The

best-fitted model has been selected based on various Kappa statistics values and also by implementing other

model validation techniques. The ‘Multi Layer Perceptron Markov (MLP_Markov)’ Model has been qualified

as the most suitable model for this research work. Using the MLP_Markov model, the land cover map of 2020

& 2030 has been predicted. Using MLP_Markov model, Prediction map shows that 28% and 34.8% of the total

study area will be converted into built-up area cover type in 2020 and 2030 respectively.

Acknowledgment The authors are thankful to the director dl, Jodhpur and Head, Department of Mining Engineering, Jai

Narain Vyas University, Jodhpur for help and encouragement during the study.

V. REFERENCES [1] Borana S.L., Yadav S.K., Parihar S.K. and Paturkar R.T. - Integration of Remote Sensing & GIS for Urban Land Use / Cover

Change Analysis of the Jodhpur city, 33rd INCA International Congress , 19 - 21 September, 2013, Jodhpur, Rajasthan, India.

[2] Dymond, J.R., J.D. Shepherd, J. Qi. 2001. A simple physical model of vegetation reflectance for statndardising optical satell ite

imagery. Remote Sensing of the Environment, 77(2): 230-239.

[3] Bhagawat Rimal (2011). Urban Growth and Land Use /Land Cover Change of Pokhara Sub-Metropolitan City, Nepal. Journal

of Theoretical and Applied Information Technology, Vol.26, No. 2, ISSN 1992-8645

[4] Ahmed, B. and Ahmed R. (2012). Modeling Urban Land Cover Growth Dynamics Using Multi-Temporal Satellite Images: A

Case Study of Dhaka, Bangladesh, ISPRS International Journal of Geoinformation, 1, 3-31.

[5] Goodchild, M. F. (2000). Spatial analysis: methods and problems in land use management. in Spatial Information for Land Use

Management, eds. M. J. Hill and R. J. Aspinall, (Gordon and Breach Science Publishers, Singapore), 39-50.

[6] Lillesand T.M.and Keifer W(1994) “Remote Sensing Image Interpretation”, New York: John Viley.

[7] Borana S. L. (2015). Urban Settlement, Planning and Environmental Study of Jodhpur City using Remote Sensing and GIS

Technologies, JNV University, Jodhpur, PhD Thesis, pp.225 (Unpublished).

[8] GLCF – http://www.glcf.umiacs.umd.edu

[9] USGS - http://glovis.usgs.gov.

[10] Eastman, J.R. IDRISI Taiga Guide to GIS and Image Processing; Manual Version 16.02; Clark Labs: Worcester, MA, USA,

2009.

[11] Abubaker, H.M, Elhag A.M.H. and Salih,A.M.(2013). Accuracy Assessment of Land Use and Land Cover Classification (LU/LC)

Case study of Shomadi area-Renk County-Upper Nile State, South Sudan. International Journal of Scientific and Research

Publications, Volume3, Issue 5.

[12] Congalton, R.G. (1991) A Review of Assessing the Accuracy of Classifications of Remotely Sensed Data. Remote Sensing of

Environment, 37, 35-46. https://doi.org/10.1016/0034-4257 (91)90048-B.

[13] Jensen, J.R. (1996) Introductory Digital Image Processing: A Remote Sensing Pers-pective. 2nd Edition, Prentice Hall, Inc.,

Upper Saddle River, NJ.

[14] Landis, J.R. and Koch, G.G. (1977) A One-Way Components of Variance Model for Categorical Data. Biometrics, 33, 671-679.

https://doi.org/10.2307/2529465.

[15] Lillesand T. M. and Kiefer R. W. (2004), “Remote Sensing and Image Interpretation,” 5th Edition, John Wiley, New York.

[16] Griffiths, P.; Hostert, P.; Gruebner, O.; Linden, S.V.D. Mapping mega city growth with multi-sensor data. Remote Sens.

Environ. 2010, 114, 426-439.

[17] Dewan, A.M.; Yamaguchi, Y. Land use and land cover change in greater Dhaka, Bangladesh: using remote sensing to promote

sustainable urbanization. Appl. Geogr. 2009, 29, 390-401.

[18] Cabral1, P.; Zamyatin, A. Three Land Change Models for Urban Dynamics Analysis in

[19] Sintra-Cascais Area. In Proceedings of 1st EARSeL Workshop of the SIG Urban Remote Sensing, Humboldt-Universität zu

Berlin, Germany, 2–3 March 2006.

[20] Cheng, J.; Masser, I. Urban growth pattern modeling: A case study of Wuhan City, PR China. Landscape Urban Plan. 2003, 62,

199-217.

[21] Balzter, H. Markov chain models for vegetation dynamics. Ecol. Model. 2000, 126, 139-154.

[22] Oludare H. Adedeji, Opeyemi O. Tope-Ajayi1, Olukemi L. Abegunde, Assessing and Predicting Changes in the Status of

Gambari Forest Reserve, Nigeria Using Remote Sensing and GIS Techniques, Journal of Geographic Information System, 2015,

7, 301-318.

[23] Firoz Ahmad and Laxmi Goparaju, Predicting Forest Cover and Density in Part of Porhat Forest Division, Jharkhand, India

using Geospatial Technology and Markov Chain, Biosciences Biotechnology Research ASIA, September 2017. Vol. 14(3), p. 961-

976.

[24] Michael Iacono et al (2012), Markov Chain Model of Land Use Change in the Twin Cities, 1958-2005, Article in TeMA - Journal

of Land Use, Mobility and Environment · January 2012, DOI: 10.6092/1970-9870/2985.

[25] Bell, E. J., 1974, “Markov analysis of land use change: an application of stochastic processes to remotely sensed data”, Socio-

Economic Planning Sciences, 8: 311-316.

[26] Bell, E. J. and R. C. Hinojosa, 1977, “Markov analysis of landuse change: continuous time and stationary processes”, Socio -

Economic Planning Sciences, 11: 13-17.

[27] Islam, S., and Ahmed, R. (2011). “Land Use Change Prediction In Dhaka City Using Gis Aided Markov Chain Modeling”, J.

Life Earth Sci. Vol. 6(Islam), 81–89. ISSN 1990-4827 http://banglajol.info.index.php/JLES.

[28] Nadoushan, et al.:Mozhgan,Modeling Land Use/Cover Changes by the Combination of Markov Chain and Cellular Automata

Markov (CA-Markov) Models, International Journal of Earth, Environment and Health,Jan-Jun 2015,Volume,Issue 1

[29] Pontius RG, Spencer J. Uncertainty in extrapolations of predictive landchange models. Environ Plan 2005;32:211-30.

[30] Abd El-Kawy OR et al (2011) Land use and land cover change detection in the western Nile delta of Egypt using remote sensing

data. Appl Geogr 31(2):483–494.

[31] Varun Narayan Mishra, Praveen Kumar Rai, Kshitij Mohan, Prediction of Land Use Changes Based on Land ChangeModeler

(LCM) Using Remote Sensing: A Case Study of Muzaffarpur (Bihar), India, J. Geogr. Inst. Cvijic. 64(1) (111-127), UDC:

911.2:551.11(540), DOI: 10.2298/IJGI1401111M.

Modelling and Prediction of Land Use Changes in Jodhpur City...

Documents

Transcript of Modelling and Prediction of Land Use Changes in Jodhpur City...