Developing and evaluating an inversion model for retrieving coastal water quality and benthic coral reef properties from water color

Coral reefs prefer to reside in warm, clean, clear waters with high oxygen content. Any deterioration of their environment would affect the life of coral reefs. Therefore, coral reefs serve as an important indicator of the environmental condition. Kenting National Park enjoys the most abundant coral reefs around Taiwan. However, recent extreme weather events, such as Typhoon Morakot in 2009, destroyed 50% of coral reefs in this area. The technique of water color remote sensing is promising in assessing the status of coral reefs at both high spatial and high temporal resolutions. However, to retrieve the water quality and the properties of benthic coral reefs directly from the water color signal requires a robust algorithm that has been validated against comprehensive in situ data and model simulations. In this research, we improve upon a genetic algorithm/semi-analytical model by taking into account the properties of benthic coral reefs, classifying the bottom into six different types; coral reefs, sand, sea grass, and green, red, and brown algae. A spectral library of bottom reflectance is established from in situ data measured in Kenting National Park and data simulated by the HydroLight radiative transfer model. Our new model, Genetic Algorithm and Shallow water Semi-Analytical model (GA-SSA), is able to iterate for an optimized solution of water quality and the properties of the benthic coral reef from the input of bottom reflectance spectrum data. These solutions are then compared to the conditions of water quality and benthic coral reef properties, under which the bottom reflectance spectra are measured in situ or simulated by the Hydrolight algorithm. Our results demonstrate that our new model is able to achieve accuracy as high as 80%. In addition, we also used a hyperspectral imager to collect a coral ecosystem spectral database in the Kenting area.