Wetland conversion risk assessment of East Kolkata Wetland: A Ramsar site using random forest and support vector machine model

摘要

East Kolkata Wetland (EKW) is a Ramsar site located adjacent to the Kolkata megacity. EKW is one of the resourceful wetland ecosystems of the world which offers a bundle of direct and indirect ecosystem services to the Kolkata megacity region. The rapid expansion of built-up in the surrounding urban agglomeration of the EKW is putting immense pressure on the EKW and the rate of wetland loss has been highest in recent decades. To ensure that this distinct ecosystem is conserved, an efficient means of identifying wetland conversion risk is needed. This study aims to assess the risk of EKW conversion using two advanced data-driven Machine Learning (ML) models, viz, Random Forest (RF), and Support Vector Machine (SVM). The novelty of the paper is in the fact that ML models have been widely applied to groundwater potential, flood susceptibility, and landslide susceptibility, their applicability to wetland conversion risk assessment has not yet been explored. The advantage of RF and SVM is that both of the ML models can overcome the limitations of pre-assumption based conventional methods of wetland risk assessment. A total of eight factors are selected which can be categorized into ecological, bio-physical, demographic, and physical infrastructure groups. Both results indicate that around 60% area under medium to very high-risk zones. A comparison is also made between these two methods to identify the most precise prediction method for this study area. The results of the models are quantitatively validated applying the Receiver Operating Characteristics (ROC) method, where both of this method identifies SVM as a more precise predictive model for this study with 91.12% accuracy. The spatial pattern of encroachment and shrinkage of EKW triggered by urban expansion is successfully captured by RF and ME. Policy analysts and land-use planners can use the outcome derived from RF and SVM models and associated maps to identify the risk zones, assess the effectiveness of wetland conservation programs, design effective policies to stop further degradation of the wetlands, and adopt long-term sustainable planning for this precious ecosystem. (C) 2020 Elsevier Ltd. All rights reserved.