Abstract
Marine heatwaves (MHWs) adversely impact Aotearoa New Zealand’s marine ecosystems and pose challenges for resource management. In this study, we evaluate forecast skill of monthly MHWs and sea surface temperature (SST) anomalies using a multi-model ensemble (MME) comprised of nine general circulation models and 206 members with a focus on Aotearoa New Zealand for the first time. Over the hindcast period (1993–2016), the MME outperforms individual models in forecasting SST anomalies around Aotearoa New Zealand, based on its higher anomaly correlation coefficient (ACC) and lower root mean square error (RMSE). The forecast skill of the MME varies seasonally, and is highest for forecasts initialized between June and September and lowest from October to December. Forecasts generally outperform persistence across all months and lead times, except at certain lead times between September and December. The background climate state also influences the MME skill, with higher accuracy during El Niño for forecasts initialized from December to February and during La Niña for certain lead times from March to August. Skill improves in spring under neutral (normal) conditions. We also evaluate the MME’s skill in predicting MHW events using a probabilistic framework. The MME retains skill up to two months along Aotearoa New Zealand’s western coast and upper east North Island but has negligible skill at four- and five-month lead times. Overall, these findings highlight that MHW and SST can be forecasted with reliability, especially at one and two months of lead time with important implications for marine resource management.