ABSTRACT
Mangrove ecosystem is providing a substantial amount of ecosystem services (ESs) that has a positive impact on human well-being. Indian Sundarbans is one of the biodiversity hotspots in the world and declared as “World Heritage Site” by UNESCO. The present chapter presents the application of satellite remote sensing data to assess the economic value of the Indian Sundarbans. Ten machine learning (ML) supervised classification models were employed for land use land cover classification and subsequent interpretation. Multiple accuracy assessment tests, including user’s accuracy, producer’s accuracy (PA), kappa statistics, and Jaccard similarity test, were performed for validating the accuracy of the models. Economic valuation of natural capitals was computed using benefits transfer approach. Among the models, maximum likelihood classification (MLC) algorithm has the lowest accuracy observed throughout the study period, except for the year 1973. While the random forest (RF) and support vector machine (SVM) models performed most accurately. Also, among the models, the high similarity is found for SVM, RF, Bayes and artificial neural network (ANN) models. However, a comparably lower similarity estimates had found for MLC model. This suggests the superiority and functional capability of SVM and RF models in capturing land dynamics. For all reference years, the highest ES values (ESVs, in million US$) was found for waste treatment service, followed by erosion control, habitat, food production, disturbance regulation, genetic, soil formation, water supply, recreation, climate regulation, raw material production, water regulation, cultural, nutrient cycling, biological control and pollination services, respectively. The present research has demonstrated that machine models could be a feasible solution for accounting importance of natural capitals across the regions. Secondary sourced freely available remote sensing data are also found highly cost-effective for moderate to large-scale land use decision-making. The valuation approaches and methods adopted in this study could be a reference for future ES studies in other regions and could be replicated easily for similar research interest across the ecosystems.
