Quantifying the quality of coral bleaching predictions

Abstract

Techniques that utilize sea surface temperature (SST) observations to predict coral reef bleaching are in common use and form the foundation for predicted global coral reef ecosystem demise within this century. Yet, quality assessments of these methods are typically qualitative or anecdotal. Quality is the correspondence of forecasts with observations and has standard quantitative measures. Here a forecast verification method, commonly used in meteorology, is presented and used to measure the quality of the degree heating weeks (DHW) technique as an exploration of insights that can be gleaned from this methodology. DHW values were calculated from NOAA Optimum Interpolation SST version 2 data and compared to a database of bleaching observations from 1990–2007. Quality is expressed with an objective measure, the Peirce Skill Score (PSS). The quality at varying DHW thresholds above which bleaching was projected to occur is calculated. By selecting the thresholds that maximize quality, the predictive technique is objectively optimized. This results in optimal threshold maps, showing reefs more prone and more resistant to bleaching. Optimization increases the quality of DHW as a predictor of bleaching from PSS = 0.55 to PSS = 0.83, in global average, but the optimal PSS and corresponding DHW values vary significantly from location to location. The coral reef research and management community are urged to adopt the simple, but rigorous tools of forecast verification routinely used in other disciplines so that bleaching forecasts can be quantitatively compared and their quality improved.