Geospatial mapping and modelling of mangrove ecosystem health at Matang mangrove forest reserve

Among the multi-functional ecosystem provided by mangrove forest, the health like biotic and abiotic variables represent the most challenging components to be monitored due to their diverse characteristics and spatially distributed over mangrove area. Since the biotic and abiotic variables factors are vital for the mangrove ecosystem health (MEH), the information of remotely sensing data and spatial analysis were integrated to evaluate the health condition of mangrove. The objective of this study was (1) to examine mangrove vegetation cover by integrating NDVI and SAVI, and (2) to model the MEH using ordinary kriging (OK) for the entire MMFR. Supervised classification, NDVI and SAVI were performed to determine vegetation coverage. Since SAVI requires a suitable L-factor to be used to distinguish the vegetation areas, four different L-factors viz. 0.1, 0.25, 0.5 and 0.75 were tested with the multiple linear regressions using the stepwise regression method of backward elimination. The relationship of NDVI and SAVI in detecting mangrove vegetation covers were examined with correlation-Pearson analysis. For Objective 2, eight variables from Faridah-Hanum et al. (2019) were used as preference factors to determine the MEH. Semivariogram model and interpolation method of OK were used to generate spatial autocorrelation of MEH. Prediction accuracy was examined through ME, RMSE and RMSSE. All variables were then overlay and combined via linear weight regression (LWC) to see the overall health status. Reclassification was conducted to standardise the health value viz. 1 (worst), 2 (poor), 3 (moderate), 4 (good) and 5 (excellent). In order to verify the health status, NDVI analysis performed in Objective 1 were used to support the accuracy of MEH. Supervised classification was observed with good accuracy; Kuala Sepetang (71.8%; K=0.668), Kuala Trong (83.8%; K=0.798) and Sungai Kerang (73.5%; K=0.681). SAVI with L-factor 0.75 was found to be significant to be used for MMFR. The vegetation indices (VIs) resulting from NDVI and SAVI demonstrate the classification variations when compared to the initial supervised classification. In Objective 2, all variables had an overall prediction accuracy with 85.16% (AGB), 90.78% (crab abundance), 97.3% (soil C), 99.91% (soil N), 89.23% (number of phytoplankton species), 95.62% (number of diatom species), 99.36% (DO) and 87.33% (turbidity). The spatial prediction autocorrelation delineating an area of 307.9 ha for excellent MEH, 15935.68 ha for good MEH, 5224.34 ha for moderate MEH, 17795.63 ha for bad MEH and 715.55 ha for worst MEH. This study modelled the overall MEH through OK with selected semivariogram model and comparison to VIs consequently promoting the restoration affects, relevant management and facilities distribution, and therefore improving the MEH over the entire MMFR.

Text FPAS 2021 16 UPMIR.pdf Download (1MB)

Actions (login required)

View Item