A GIS-based modelling approach to identify natural drivers of coral reef abundance in the Northern Myeik Archipelago, Myanmar

Abstract

To predict the coral abundance of the Northern of Myeik, I compared various progressions of model techniques, including stepwise regression (using OLS), GWR, and Random forest analysis methods, to investigate relationships between coral abundance survey data and environmental variables such as depth, slope, aspect, rugosity, chlorophyll, sea surface temperature, and turbidity. Depth and SST have the most significant effect on predicted coral species abundance. Increased reef abundance was associated with a reduction in sea surface temperature stability and shallower optimum depths. Even then, GWR outperformed the other studied approaches in places with a substantial degree of input-output disagreement. The GWR model production was used to produce a final predicted coral abundance modelling map. The accuracy of the GWR model was determined by using Random forest predict modelling to map and comparing the higher R2 and predicted and observation graphs to the slope and interest value of each model. This sampling tool for a reef prediction model can be used in preference of potential species abundance modelling (e.g., seagrass, mangrove) in future Myanmar coastal management projects, resulting in more accurate predictions and more educated species management decisions. It can assist the Department of Fisheries in making fisheries management decisions and help to keep fish stocks stable in the long run by fostering a greater understanding of key environmental variables.