Spectral classification of AVIRIS NG hyperspectral data for discriminating coastal foredunes based on vegetation species: a case study from Cuddalore district of Tamil Nadu, South India

Coastal dune distinction is an essential for monitoring, conservation and sustainable management of fragile coastal ecosystem. Most of the coastal scientists classified the coastal sand dunes by conventional methods as frontal dunes (foredune) and back dunes (stabilized dune). Due to the significance in the developing stage of the dunes and the coastal protection, foredunes are very significant in this context. Studies demarcating the boundary of coastal dune field based on its spectral signatures of vegetation species have not been attempted by any authors earlier. In this view, the goal of this study was to determine whether it is possible to demarcate the foredune from the dune field using vegetation species with the aid of Airborne Visible InfraRed Imaging Spectrometer—Next Generation (AVIRIS-NG) hyperspectral data. To achieve this, the study scrutinized the major vegetation species and the end members were identified using Pixel Purity Index algorithm (PPI), N-Dimensional visualizer, Insitu spectra and Linear Spectral Unmixing algorithm. Identified training pixels falling in the dune field were grouped into six training classes to perform classification techniques (Spectral Angle Mapper (SAM), Spectral Informative Divergence (SID) and Support Vector Machine (SVM)). This study also discussed about various methods of selecting training pixels for demarcation of foredunes. Accuracy assessment reveals that SVM shows an overall accuracy of 91.97% compared to SID of 68.25% and SAM of 58.52% with kappa coefficient of 0.89, 0.53 and 0.49 respectively. The study found the indicator species such as Ipomoea Pes-caprae, Spinifex Littoreus and Launaea Sarmentosa blankets the foredunes in this study area, which could be used to demarcate the boundary of the foredunes. The methodology employed in this study can be applied to distinguish the foredune from the dune field in any coastal dune ecosystem, provided that the researcher has access to a database of the flora present in the area. The proposed methodology and this research might bring new insight into the classification and demarcation of foredune using the vegetation species present over the dune field. The research allows more efficient and cost-effective identification of foredunes from vast coastal environments, providing valuable information for the conservation and sustainable management of ecologically sensitive ecosystems.