Predicting coral mortality in South East Asia using open-source data

The best model in predicting coral mortality utilized artisanal fishing data, the model had an R2 of 0.8673 and an accuracy of 48.51%. This model identified the locations of the highest mortality to be in Timor-Leste at Blue Ventures management unit sites TL1 and TL5, however adding economic value in the concern index then Papua New Guinea site PNG7 is more at risk. I hypothesize that a lack in variability and standardized approach in the coral mortality/bleaching explanatory data decreases the accuracy and real-world applicability of the model.

The majority of explanatory variables were obtained from the online database Reef Base, some of the observations showed extreme outliers which were all indicated by citizen scientists. Going forward I would recommend that further education and standardization is included in observations to ensure the open source data has a higher accuracy to make better models.