Prediction of landuse/landcover using CA-ANN approach and its association with river-bank erosion on a stretch of Bhagirathi River of Lower Ganga Plain

The present study documents the landuse and landcover change on a stretch of the Bhagirathi river in the last three decades in response to river bank erosion. Landuse and landcover maps were prepared with the help of supervised image classification techniques and validated through accuracy assessment by kappa statistics. Landuse transformation matrix was prepared for different periods to understand the nature of landuse change in the study area. Simulation and prediction of LULC for the year 2035 was done to explore future riverine landscape dynamics based on Cellular Automata-Artificial Neural Network (CA-ANN) model through MOLUSCE (Modules of Landuse Change Evaluation) plugin of QGIS software. The kappa statistics coefficients are 0.85, 0.83 and 0.93 for the years 1990, 2005 and 2019 respectively. The result of the investigation reveals that during the last three decades (1990-2019) area under agricultural land and settlement increased by an annual rate of 10.20 and 1.03 km(2) respectively, while area under fallow land, vegetation cover, water body and wetland shows a decreasing trend. Likewise, the simulation result of CA reveals that area under agriculture and settlement will continue to increase in future at the cost of other LULC classes like vegetation and fallow land. The Kappa overall, kappa histogram and kappa local values were 0.78, 0.92 and 0.85 respectively with 84.71% of correctness. This validates the prediction results that the area under agricultural land and settlement will increase at an annual rate of 2.65 km(2) and 0.82 km(2) respectively due to increase in accretional lands and population in the study area. The transformation of the area under vegetation and fallow land to agricultural practices and the unplanned human settlement in the riparian area indicate an alarming future scenario as far as loss of human life and property is concerned because the bank erosion is occurring in both the banks. Therefore, the study regarding the LULC change and its future scenario in the erosion affected floodplain will give a holistic understanding the land and landuse dynamics, for proper planning and management of the situation for the sustainability of the floodplains and its occupants.

Automated summary

The study investigates the landuse and landcover change along the Bhagirathi river over the past three decades due to river bank erosion. Classification techniques and kappa statistics were used to prepare landuse and landcover maps and assess their accuracy. A simulation model was developed to predict future changes in the riverine landscape. The results suggest an increase in agricultural land and settlement at the expense of vegetation and fallow land, indicating potential risks to human life and property.