SARIMA-based forecasting of mangrove vegetation indices across India

Abstract Mangrove ecosystems are crucial to the ecological and climatic dynamics of the Asia–Pacific coastline. However, the ability to predict long-term changes is limited, hindering evidence-based resilience and adaptation strategies. The study presents a remote-sensing framework that integrates data from multiple sensors with time-series analysis to predict changes in mangrove ecosystems along India’s coastlines, aiding in conservation planning. Vegetation indices derived from MODIS were collected monthly from 2001 to 2025. The initial two decades (2001–2020) were used to develop Seasonal Autoregressive Integrated Moving Average (SARIMA) model, while the last five years (2021–2025) served as an out-of-sample validation set. The time series revealed strong intra-annual seasonality for all indices (seasonal strength F_S = 0.81–0.86), with distinct seasonal patterns for each index reflecting climatic impacts on mangrove phenology, alongside a slight but statistically significant upward trend. MSAVI exhibited a narrower dynamic range compared to NDVI and EVI, aligning with its sensitivity to soil background and atmospheric interference in coastal intertidal areas. For the 2021–2025 validation period, SARIMA models achieved MAPE ranging from 2.5% to 4.9% across the indices; NDVI notably surpassed a seasonal-naive benchmark (Diebold-Mariano p < 0.05), whereas EVI and MSAVI demonstrated similar accuracy to seasonal-naive and STL+ETS models without a statistically significant edge. Projections to 2030 suggest a continued slight upward trend for all indices, staying within historical variability limits, indicating stable to improving conditions under current human and climate influences.