Seasonal trends of hot spells and predictability of driving mechanisms in northern hemisphere midlatitudes

One of the most detrimental climate change impacts in the midlatitudes is the occurrence of extreme temperature events, particularly hot spells. These periods of anomalously warm events are getting more frequent, intense, and durable under climate change. Therefore, their accurate prediction and projection are crucial for appropriate adaptation and preparedness in vulnerable communities. This thesis explores both aspects by interrogating state-of-the-art weather forecast and climate models. The occurrence of hot spells is associated with Quasi-Stationary Rossby Waves (QSWs) in the midlatitude atmosphere. The first research chapter shows that weather forecast models exhibit limited Subseasonal-to-Seasonal (S2S) skill in winter QSWs, except in the North Pacific region. Further investigation showed that this region exhibits variable skill over time, which can be associated with different sources of predictability in the climate system. Particularly, certain phases of the Madden-Julian Oscillation, Sudden Stratospheric Warming, Quasi-Biennial Oscillation, and El Niño Southern Oscillation were found to enhance predictability in the North Pacific. This is valuable for future work aimed at improving the forecast skill of temperature extremes, as these sources of predictability and the improved QSW skill can provide pathways for improvement. The second research chapter assesses trends in hot spells across the boreal seasons and their representation in climate models. Previous studies overlooked hot spells outside the summer season, though they have detrimental impacts on ecosystems and land processes. Here, I show the existence of comparable trends in all seasons and that Coupled Model Intercomparison Project phase 6 (CMIP6) models underestimate these trends, particularly in spring. Future projections mimic the historical seasonality of CMIP6, potentially continuing to misrepresent trend seasonality. The final research chapter shows that bias-correction and downscaling techniques don't substantially reduce these trend biases when examined across the Canadian region. Using the significant relationship between temperature trends and changes in hot spells, the potential of emergent constraints to reduce these biases is assessed. However, the technique only reduces biases in some regions, with noisy corrections observed across the midlatitudes. These findings highlight that the most reliable way forward is to advance model representation, accounting for the seasonality of hot spell processes.