Skillful Land and Marine Heatwave Forecasting Through Hybrid Statistical Dynamical Modelling

Skillful forecasting of global heatwaves is crucial for mitigating their escalating impacts on human societies and ecosystems across various sectors. While global climate models generate invaluable dynamical temperature forecasts, the crucial role of model output statistics (MOS) in enhancing global heatwave forecasting has not been thoroughly investigated. In this study, we unravel the potential of hybrid statistical dynamical modelling in generating heatwave forecasts on a global scale. Specifically, a pioneering MOS toolkit is developed to iteratively take into consideration key attributes—bias, spread, trend, and association—within raw forecasts through a series of methodical one-factor-at-a-time experiments. A case study is devised for forecasts of 2-meter air temperature over land and sea surface temperature generated by the National Center for Environmental Prediction’s Climate Forecast System version 2. Our analysis exposes the detrimental impacts of biases and unreliable ensemble spreads within raw temperature forecasts, leading to an abundance of false positives and negatives, ultimately diminishing the skill of heatwave predictions, often plunging below − 100%. At the lead time of 0 months, integrating incremental considerations of bias, spread, trend, and association results in substantial skill enhancements across global land and marine grid cells. Notably, land heatwave forecast skill sees a remarkable ascent from a staggering − 171.63%±290.42% to a promising 5.61%±15.74%, while marine heatwave forecast skill improves from − 75.74%±206.68–23.96%±23.47%. Despite the anticipated degradation of skill with lead time, our results underscore MOS’s efficacy in leveraging raw forecast data to maintain positive forecasting outcomes.