Multi-spatiotemporal analysis of changes in mangrove forests in Palawan, Philippines: predicting future trends using a support vector machine algorithm and the Markov chain model

Abstract

Multi-temporal remote sensing imagery can be used to explore how mangrove assemblages are changing over time and facilitate critical interventions for ecological sustainability and effective management. This study aims to explore the spatial dynamics of mangrove extents in Palawan, Philippines, specifically in Puerto Princesa City, Taytay and Aborlan, and facilitate future predictions for Palawan using the Markov Chain model. The multi-date Landsat imageries during the period 1988–2020 were used for this research. The support vector machine algorithm was sufficiently effective for mangrove feature extraction to generate satisfactory accuracy results (>70% kappa coefficient values; 91% average overall accuracies). In Palawan, a 5.2% (2693 ha) decrease was recorded during 1988–1998 and an 8.6% increase in 2013–2020 to 4371 ha. In Puerto Princesa City, a 95.9% (2758 ha) increase was observed during 1988–1998 and 2.0% (136 ha) decrease during 2013–2020. The mangroves in Taytay and Aborlan both gained an additional 2138 ha (55.3%) and 228 ha (16.8%) during 1988–1998 but also decreased from 2013 to 2020 by 3.4% (247 ha) and 0.2% (3 ha), respectively. However, projected results suggest that the mangrove areas in Palawan will likely increase in 2030 (to 64,946 ha) and 2050 (to 66,972 ha). This study demonstrated the capability of the Markov chain model in the context of ecological sustainability involving policy intervention. However, as this research did not capture the environmental factors that may have influenced the changes in mangrove patterns, it is suggested adding cellular automata in future Markovian mangrove modelling.