Monitoring the rapid changes in mangrove vegetation of coastal urban environment using polynomial trend analysis of temporal satellite data

Abstract

Over the last few decades, anthropogenic activities have triggered the rate of change in the function of mangrove ecosystems in coastal urban areas. Satellite imagery provides valuable information for mangrove mapping and monitoring. Metrics derived from the linear regression analysis of spectral indices (SIs) derived from satellites are commonly used for change analysis. This study examines the robustness of the widely used SIs derived from Landsat satellite image to distinguish mangroves and non-mangrove features and identify the non-linear changes over mangrove forest using the polynomial trend analysis. Airborne Visible and Infra-Red Imaging Spectrometer Next Generation (AVIRIS-NG) data was resampled to simulate the spectral-response of the Landsat sensor. One-way analysis of variance (ANOVA) and mutual information (MI) were applied to the simulated data to identify the optimal SIs. Based on the statistical analysis, modular mangrove recognition index (MMRI), modified normalized difference water index (MNDWI) and normalized difference built-up index (NDBI) were identified to delineate the mangrove, inundated and built-up region. Finally, time-series profiles of the identified spectral indices were generated using Landsat-8 (L8) and Landsat-7 (L7) data to analyse the change dynamics of the mangrove vegetation from 2002 to 2019. The proposed methodology was applied to study the changes in mangroves in the coastal areas of Kochi (Ernakulam District, Kerala State, India). The temporal-non-linear changes in the mangrove area were identified based on polynomial regression that revealed abrupt decline (42% decrease) in the mangrove area due to large-scale infrastructure development projects. The proposed approach is easy to implement, which enables the frequent monitoring of extensive mangrove forests.