Assessing Cyclonic Changes in Mangrove Dynamics with Multispectral to Continuous Hyperspectral Transformation

Abstract The limited spectral resolution of available multispectral satellite imagery restricts the research to assessing cyclone-induced mangrove degradation. The eight vegetation-sensitive Sentinel-2 bands cannot continuously resolve the green-to-red, red-edge, and near-infrared regions. Therefore, canopy stress, chlorophyll loss, moisture alteration, fragmentation, and subsequent recovery from multi-spectral images become difficult. This study proposes Ecologically Bounded Spectral Reconstruction using Multi-Regime Interpolation (EBSR-MR). This algorithm aims to transform eight Sentinel-2 surface-reflectance bands into a continuous 101-band hyperspectral representation spanning 400–900 nm. The methodological innovation focuses on wavelength-specific reconstruction regimes, preservation of measured Sentinel-2 bands as spectral anchors, Gaussian modeling of the nonlinear red-edge transition, smooth boundary integration between spectral regimes, and ecologically constrained visible and near-infrared interpolation. These components are needed to reconstruct a reflectance curve that follows the expected physical behavior of real mangrove vegetation. This will prevent physically implausible reflectance behavior and help in reconstructing diagnostically important wavelengths absent from the native sensor. EBSR-MR was evaluated over the cyclone-affected mangroves (pre-Amphan (2019), post-Amphan (2021), and recovery-stage (2023) of Henry Island, Indian Sundarbans.EBSR-MR bridges the spectral gap between accessible multi-spectral imagery and costly hyperspectral observations by maintaining near-perfect spectral–spatial fidelity. Also, EBSR-MR provides detailed information in the chlorophyll-sensitive red-edge region and estimates the missing 900-nm reflectance needed to identify mangrove canopy moisture, stress, cyclone damage, and recovery more accurately. EBSR-MR is sensor-accessible and does not require adversarial training or extensive labelled samples; it can be adapted for monitoring vegetation responses to floods, droughts, wildfire, and coastal erosion.