Seasonal Prediction of Arabian Sea Marine Heatwaves

Marine heatwaves are known to have a detrimental impact on marine ecosystems, yet predicting when and where they will occur remains a challenge. Here, using a large ensemble of initialized predictions from an Earth System Model, we demonstrate skill in predictions of summer marine heatwaves over large marine ecosystems in the Arabian Sea seven months ahead. Retrospective forecasts of summer (June to August) marine heatwaves initialized in the preceding winter (November) outperform predictions based on observed frequencies. These predictions benefit from initialization during winters of medium to strong El Nino conditions, which have an impact on marine heatwave characteristics in the Arabian Sea. Our probabilistic predictions target spatial characteristics of marine heatwaves that are specifically useful for fisheries management, as we demonstrate using an example of Indian oil sardine (Sardinella longiceps). Marine heatwaves (MHWs) are prolonged extreme events associated with exceptionally high ocean water temperatures. Such events impose heat stress on marine life, and thus predicting such events is beneficial for management applications. In this work we show that the occurrence of MHWs in summer in the Arabian Sea can be skilfully predicted seven month in advance. Our prediction system benefits from the information of sea surface temperature anomalies in the eastern Pacific Ocean in the preceding winter, among other aspects. Our predictions suggest potential for using climate information in fisheries management in this region. Summer marine heatwaves in the Arabian Sea are predictable seven months in advanceThe prediction skill in summer is mainly associated with a preceding El Nino event in winterProbabilistic predictions of Arabian Sea area under heatwave can be tailored to benefit fisheries

Automated summary

Using an Earth System Model, we are able to predict summer marine heatwaves in the Arabian Sea with skill seven months in advance. These predictions outperform those based on observed frequencies and are influenced by medium to strong El Nino conditions during winter initialization.