Coral Reef modeling using Support Vector Regression and applying spectral indices

Coral reef communities face unprecedented pressures at local, regional and global scales as a consequence of climate change and anthropogenic disturbance. Remote sensing, from satellites or aircraft, is possibly the only means to measure the effects of such stresses at appropriately large spatial scales. Coral reefs are indicators of environmental, climate and sea surface change which shows the reefs damages. For this purpose, an algorithm that includes relation between those variables was used. Australia’s Heron reef island is considered to be studied. In order to modeling the coral reef health condition by support vector regression (SVR), water and coral important spectral indexes and features was found using landsat-8. From this method, the SVR is achieved the high performance for modeling the statistical problem .In this study we used genetic algorithm (GA) and sequential feature selection (SFS) for selecting suitable features in order to estimate model. The results for the testing data in this area of best model is [RMSE= 0.53 and R² =0.983] that show the high SVR performance.