Please use this identifier to cite or link to this item:
|Title:||Integrated Modelling of Agro-Industrial Landscape Dynamics in India|
|Keywords:||Department of Applied Mechanics and Hydraulics;LCLU change;Land Use policy;System Dynamics;Change Modelling;Dyna-CLUE|
|Publisher:||National Institute of Technology Karnataka, Surathkal|
|Abstract:||The biophysical compositions of the earth surface are land covers and the different human uses of land covers are land uses. Hence, both are not same. Increasing human uses are altering Earth’s natural environment and land cover in rapid pace. Despite the structured development in mathematical modelling and characterization of LCLU using remote sensing techniques; systematic understanding of LCLU change process remain underexplored. In India, due to agricultural and industrial policy reforms, land conversions are taking place at a high rate. There is an apt need for scientific studies on LCLU system. The objective of this study is to investigate location specific LCLU change driving factors and find the consequent effects of these drivers LCLU changes in pre-industrial and industrialized landscape. Applicability of integrated LCLU change model for Indian condition is evaluated by integrating Dyna-CLUE model with System Dynamics (SD) model. An attempt has also been made to investigate the utility of temporal remote sensing data (acquired through space borne sensors at regular interval) to produce LCLU maps. The current study has been carried out in a typical agriculture dominated landscape as this region is undergoing rapid industrialisation and urbanisation since last decade. LCLU dynamics are assessed using multi date remote sensing data. Google Earth (GE) is used for collecting training and validation samples as well as historical infrastructure information. Data are also gathered from field, secondary and tertiary sources. LCLU map is prepared from multi date IRS remote sensing data using Maximum Likelihood supervised Classification algorithm to a satisfactory level of accuracy. A set of eight commonly accessed drivers are identified from relevant literature. Further, nine more drivers are listed with the help of local knowledge. Gathered data spatially mapped as proximate drivers and used as independent variable in Binary Logistic regression. Binary Logistic Regression (BLR) analysis is carried out to investigate the LCLU change drivers. For modelling pre-industrial landscape, year 1997 LCLU map is used as base. For this a linear interpolation model is used to estimate the aggregate LCLU change. Then Dyna-CLUE model is used to spatially model the changes. R 2 , RMSE(ha) and RSR are employed to evaluate the model’s performance. Evaluation using estimated LCLU data has demonstrated very good result. However, when error evaluation is done using actual LCLU data, accuracy has dropped significantly. Validation is also carried out for spatial domain using actual LCLU images as validator. With the advancing time steps, accuracy has reduced. Hence it is anticipated that introduction of more complex model to estimate the aggregate LCLU change may improve the accuracy. Modelling of industrialized landscape, is facilitated using System Dynamic model. Each of the LCLU classes are modelled as sub-systems. Besides, Population is also modelled as a subsystem, which pertains one-way influence on LCLU classes. LCLU maps of the year 1997, 2003 and 2005 are used for the calibration of model. Coefficient of determination (R 2 ) for calibration is 0.94. Validation using 2007 and 2010 LCLU maps is 0.96. Estimated LCLU quantities from SD are used in Dyna-CLUE model. This time “Other Land Uses” class is merged with “Waste land” class as it is <1% of total area. In Logistic regression year 1997, 2005, 2007 and 2010 LCLU maps are used as dependent variables. LCLU map of the year 2010 fared better Area under ROC curves (AUC) values in comparison to other years. Hence, it is used in final modelling. LCLU demands between the Year 1997 and 2009 (Baseline Scenario) are used for model calibration. Kappa statistics are employed to evaluate the agreement between GE samples and consecutive simulation result of the same year. The overall accuracy is found to be of average values. Model results are significantly improved during scenario simulation. This study has demonstrated the capability of virtual earth and temporal remote sensing data to LCLU time series database creation. LCLU change drivers available in the study area, are also examined. Apart from the commonly used drivers, location specific drivers are also tested successfully. After the analysis, it is found biophysical drivers are more dominant followed by population density. Researcher’s defined drivers such as Ground water, Road density are providing further insight into the change process. For better insight, drivers are separately analysed for industrialized landscape. Model’s simulation capability is sensitive toward the scale and resolution of drivers. Hence, insignificant drivers are removed during calibration of the final model. It is also observed that, class wise accuracy has relationship with the LCLU dynamics. This research also highlights model’s response to a sudden LCLU change. Such abrupt changes can only be modelled using a base map with hotspots of changes are clearly visible.|
|Appears in Collections:||1. Ph.D Theses|
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.