To monitor the coral reefs, I use statistics to detect the anomalies in the corals from satellite imagery. I include more time frames into the model as we collect more data sets from Planet satellite products, and apply advanced time-series analysis, including autoregressive integrated moving average (ARIMA) and Prophet into the model pipelines. I also apply scikit-learn machine learning package and integrate unsupervised learning into the time-series analysis. Through this pioneering project, I expect to discover the fundamental understanding of changes in the coral reefs, which will lead to high impact journal publications, and may lead to changes in the policy in Hawaii.
• For Moorea, working on the logistics of overlaying the hotspot with satellite imagery.
• As an experimental method, bleaching hotspot locations with three levels of bleaching severity could be identified and aggregated through ArcMap spatial analysis. Will work to integrate into the monitoring system using Google Maps API in two to three weeks.
• Working on a three-week moving average compared to a weekly hotspot indicator in order to test which method is more appropriate for the monitoring system. The purpose is to improve the signal/noise ratio by applying a three-week moving average to deal with the overestimated anomaly situation, which can possibly happen when using weekly data. This is being examined in ArcMap and the results will be reported in about one week.