During DJF, the northern mid- to high-latitude regions (North America, Northern Asia, and Northern Europe) are projected to experience a decrease in extreme DTDT changes relative to the historical climate, driven by consistent changes in atmospheric circulation and physical processes, as illustrated in Figs. 4–6 and S5–S11. This section focuses exclusively on the results for North America. Composites of anomalies from the seasonal climatology are shown for both days involved in an extreme DTDT change for geopotential height (GP) at 500 hPa, wind at 850 hPa, and sea level pressure (SLP) in Fig. 4, together with changes in the day-to-day circulation differences in Fig. S5.

In the simulated historical period, DTDT warming events are associated with pronounced shifts in GP, SLP, wind anomalies, and air mass origin between days t-1 and t (Figs. 4a–b and 5a–b). On day t-1, the grid box lies within a high–pressure area downstream of an eastward–propagating ridge at 500 hPa, in a transition zone between southwesterly wind anomalies to the west and northeasterlies to the east (Fig. 4a), associated with the advection of relatively cool air masses (Fig. 5a, m). By day t, the ridge moves over the grid box, featuring southwesterly wind anomalies in between an upstream surface low and the high that shifted downstream (Fig. 4b) and the advection of much warmer continental air masses (Fig. 5b, m). This shift to southwesterly advection (see also Fig. S5a) contributes an average warming of +6.2 °C to the DTDT change (Figs. 5m and 6a). In terms of vertical transport, weaker subsidence at t compared to t-1 reduces the adiabatic warming by -2.6 °C (Figs. 5i, 6a), partially offsetting the temperature increase. Conversely, enhanced diabatic heating during the final 24 h of the event day contributes +4.8 °C to the DTDT change (Fig. 5q), further amplifying the warming. The strength of these process contributions varies across events, as illustrated by the box–and–whisker plots (Fig. 6a). In summary, the simulated DTDT warming events during the historical period are primarily driven by a shift toward southwesterly warm–air advection, with diabatic heating reinforcing the effect and adiabatic changes slightly dampening it (Fig. 6a). These model results generally align with the patterns and process contributions identified from ERA5 reanalysis data (see Figs. 4a–c and 5k in Part 1). A quantitative comparison indicates that CESM–LE exhibits slightly smaller advection, along with greater reduced–adiabatic and enhanced–diabatic contributions than ERA5.

In general, the future synoptic–scale pattern for DTDT warming events resembles the historical pattern, though SLP and GP anomalies weaken and shift southeastward (Fig. 4e–f). This leads to an earlier shift to southwesterly flow at t-1, shifting air mass origins southwestward (Figs. 4i, 5e). By day t, GP anomalies move further downstream, reducing westerly and enhancing southerly meridional flow (Figs. 4j, 5f). Thus, both days exhibit a stronger southerly component, though the southwesterly strengthening is more pronounced at t-1, resulting in a northeasterly anomaly between days (Fig. S5c). These relatively subtle changes in circulation patterns are associated with an average net reduction of -1.1 °C in the advective contribution to DTDT warming compared to the historical period (Fig. 6c). This advective reduction is because the projected warming at the origin of the traced air masses (at -3 d) is 9.4 °C in day t-1 but only 8.3 °C in day t (Fig. 5n), which is both considerably above global mean warming (Fig. S6a–c). Nevertheless, from Arctic Amplification alone, one would expect an even larger reduction of the advective contribution (Fig. S7a and c), indicating that changes in circulation mitigate the effect of AA. Furthermore, projected changes in vertical motion indicate slightly increased subsidence on day t-1 (by 5–6 hPa in 3 d, Fig. 5j) compared to day t, which modestly enhances adiabatic warming, reducing the warming between t-1 and t by an additional -0.5 °C. Diabatic heating decreases relative to the historical period on both days, with a slightly larger reduction on the event day (mainly in the last 24h before arrival; Fig. 5r), contributing an additional -0.5 °C to the overall decrease in warming. Overall, the weakening of all three contributions together results in a mean reduction in the DTDT warming of -2.1 °C, with the reduced advective contribution accounting for approximately half of this signal. However, changes in diabatic and adiabatic warming also play a significant role (Fig. 6c).

During historical DTDT cooling events, a southeastward transition of a trough–ridge pattern and associated SLP anomalies between days t-1 and t dominate the temperature drop (Figs. 4c–d, 5c–d, and S5b). Specifically, the shift from a southwesterly advection of a warm air mass downstream of a low–pressure anomaly on day t-1 to northwesterly transport of a colder air mass between an upstream surface high and the downstream low on day t drives the cooling. This pattern results in a large advective contribution of -9.8 °C to the temperature decrease (Figs. 5o and 6b). The advective cooling is partially offset by moderate adiabatic warming (+2.9 °C), resulting from enhanced subsidence and descent of air masses on the event day (Fig. 5k and o). Additionally, diabatic cooling on day t contributes -2.4 °C (Fig. 5s), further intensifying the overall temperature decline. Collectively, the strong cold–air advection, reinforced by diabatic cooling, is the primary driver of DTDT cooling events across North America and other mid- to high-latitude regions (Figs. 6b and S5–S11). A comparison of North America cooling events shows that, while diabatic cooling is similar between CESM–LE and ERA5, cold–air advection is slightly weaker in CESM-LE, and adiabatic warming is marginally more pronounced compared to ERA5 (see Figs. 4d–f and 5l in Part 1).

The synoptic circulation pattern characterising future DTDT cooling events is very similar to that of historical events (Figs. 4c–d, g–h, k–l, S5d). On days t-1, a slight northward increase of the low-pressure anomaly is associated with strong northerly winds and enhanced air mass origins over Northern Canada, accompanied by downstream weakening of the westerly flow across the selected grid box (Figs. 4k and 5g). By the event day, the intensified low-pressure anomaly shifts southward, with enhanced northwesterly flow and airmass origins extending west-northward (Figs. 4l, 5h, S5d). Nonetheless, the average temperature of the air masses 3 d prior to their arrival rises by 10.3 °C on the day of the event, compared to a warming by only 7.5 °C for the trajectories initialized at t-1, which is due to the generally larger warming at higher latitudes, where the air masses on the day of the event originate, associated with Arctic amplification (Figs. 5h, p, S6d–f, and S7b, d). As a result, the contribution of advective cooling is substantially reduced in future events relative to the historical period (+2.8 °C, Fig. 6d). Vertical motion changes also show slightly enhanced subsidence at t (approximately 4–5 hPa in 3 d; Fig. 5l), which contributes to a modest increase in adiabatic warming (+0.4 °C) and further limits surface cooling. Diabatic cooling is similarly weaker in the future climate, with a slightly greater reduction on day t-1 – particularly during the final 24 h (Fig. 5t) – leading to an additional change of +0.4 °C. Overall, the dominant factor limiting future DTDT cooling (+3.6 °C) is the reduced strength of cold air advection associated with Arctic amplification (Fig. S7d), with smaller but reinforcing contributions from enhanced adiabatic warming and reduced diabatic cooling (Fig. 6d).

The CESM–LE future projections suggest that extreme DTDT changes will also weaken during JJA in some extratropical regions, including eastern and western North America, Northeastern Asia, Greenland, and southern South America. As an example, a grid box in western North America is examined in detail in the following section (Figs. 7–9). Further examples are shown in Figs. S13–S16.

In the historical climate simulations, DTDT warming events are typically associated with a positive SLP anomaly near the selected grid box at t-1, but with a cyclonic 500 hPa anomaly with a maximum to the east (Fig. 7a). Winds near the grid box are weak, and air masses originate in the vicinity, mostly over the continent (Fig. 8a). By day t, the cyclonic anomaly at 500 hPa is replaced by a mature anticyclonic anomaly that began developing along the west coast on day t-1 (Fig. 7b). This transition is associated with the accumulation of warmer air masses over the continent beneath an amplified upper–level ridge (Figs. S13a and 8b). Nevertheless, wind changes near the grid box between days t-1 and t are small, and air masses have nearly identical source regions (Fig. 8a–b) and source temperatures (Fig. 8m). Accordingly, the contribution of advection to the warming is small and even slightly negative (-0.4 °C); however, there is significant variability between individual events (Fig. 9a). The warming is thus driven by adiabatic and diabatic processes during the 3 d of air mass transport, with increased adiabatic warming due to stronger subsidence within the anticyclonic circulation anomaly at t compared to t-1, contributing +2.4 °C (Figs. 8i, m, 9a). Near the surface, the descent weakens, allowing substantial diabatic warming likely linked to surface sensible heat fluxes, which further elevate surface temperatures (Fig. 8q) and contribute approximately +3.4 °C to the temperature change. Overall, warm advection plays a negligible role for DTDT warming events during the historical period, which are instead driven by increasing descent and diabatic heating (Fig. 9a). However, note that the roles of advection and adiabatic processes in CESM–LE are reversed and are larger in magnitude than in ERA5 (Fig. S5c in Part 1). Nevertheless, according to ERA5, diabatic warming is the primary contributor to DTDT warming events, with a magnitude nearly equal to that in CESM–LE.

In the future climate scenario, the synoptic pattern resembles that of the historical climate, albeit with subtle changes. DTDT warming events are linked to smaller positive SLP anomalies and a weakened anticyclonic circulation over the West Coast at t-1 (Fig. 7e, i). This goes along with a further reduction in the transport of oceanic air masses and increases the presence of northerly and localised air masses near the target grid point at t-1 (Fig. 8e). By day t, the anticyclonic anomaly centered over the grid point is also weaker (Fig. 7f, j) and air mass origins shift northward (Fig. 8f). Although there is a slight shift towards warm air advection (Fig. 8m, n), contributing approximately +0.7 °C to the projected change in DTDT magnitude, it does not compensate for the overall reduction (Fig. 9c). The primary driver of the projected decline in DTDT warming is a significant decrease in adiabatic warming (-1.2 °C) due to a smaller mean descent of air masses (by -11 hPa) particularly on the day of the event (Fig. 8i, j), which is linked to a weakening of the anticyclonic anomaly at 500 hPa described above (see again Figs. 7j, S13c). Surface diabatic heating plays a secondary role. Since changes in heating are slightly greater on day t-1 than on day t, it contributes an additional -0.3 °C to the overall reduction (Fig. 8r). Thus, the decline in adiabatic warming, combined with a modest reduction in surface diabatic heating, is the key factor decreasing the magnitude of DTDT warming events under future climate conditions (Fig. 9c). However, when considering only the last day before arrival, the changes in diabatic heating are the main factor weakening the DTDT warming (by -0.9 °C), while adiabatic warming changes are weaker and of opposite sign (+0.3 °C). This suggests that the process decomposition of DTDT warming events over western North America depends on the temporal limits of the analysis.

During historical DTDT cooling events, an anticyclonic circulation anomaly at 500 hPa over and to the east of the selected grid box at day t-1 shifts further eastward and is replaced by an eastward–moving cyclonic anomaly from the Pacific Ocean on the day of the event (Fig. 7c–d). This is accompanied by an eastward shift of a negative SLP anomaly and stronger westerly winds on the southern flank of the cyclonic anomaly reaching the grid box on day t, such that the density distribution of air masses extends further into the western Pacific region compared to t-1 (Fig. 8c–d). As a result, a slight shift towards cold air advection leads to an average temperature drop of -1.6 °C (Figs. 8o, 9b). This advective cooling is partly offset by a modest increase in adiabatic warming (+1 °C) due to stronger air mass descent on the day of the event (Fig. 8k, o). However, the main factor driving the DTDT decrease is reduced diabatic heating near the surface on day t relative to t-1 (Fig. 8s), mainly on the last day before the trajectories arrive (Fig. 9b), with a mean contribution of -5.5 °C. These processes are similar in ERA5, but CESM–LE overestimates adiabatic and diabatic contributions and underestimates advection in its representation of DTDT cooling events (see Fig. S5d in Part 1).

In projected DTDT cooling events, the configuration of circulation anomalies remains similar to present–day, but their magnitude significantly weakens (Fig. 7g–h, k–l). At t-1, air masses mainly originate from the north with a slight oceanic extension, whereas by day t they shift towards the continent, accompanied by a significant reduction in oceanic air masses due to weakened westerly winds, which are linked to a weakening of the southern flank of the cyclonic anomaly (Fig. 8g–h, k–l, and S13d). Together, this leads to a future reduction in the temperature difference between days t-1 and t at the air mass origin (Fig. 8o, p), thereby reducing the contribution of cold air advection to the DTDT cooling by approximately +0.4 °C (Fig. 9d). The cooling is further suppressed by increased adiabatic warming (+0.7 °C), linked to stronger mean ascent (by 6 hPa) at day t (Fig. 8l, o). Near the surface, diabatic cooling is projected to increase slightly by -0.3 °C in the future (Fig. 8s, t). Therefore, future reductions in DTDT cooling events are mainly driven by reduced advection and further increases in adiabatic warming, with changes in diabatic cooling playing only a minor role (Fig. 9d).

During DJF, many tropical regions (the Amazon, southern Africa, and maritime continents) are projected to experience an increase in extreme DTDT changes compared to the historical climate. To investigate the mechanisms behind extreme DTDT changes, we focus on a specific location in tropical South America (Figs. 10–12; see Figs. S17–18 for another location in tropical southern Africa).

In the CESM historical simulation, air masses at -3 d cluster around the selected target grid box, indicating that local conditions mainly drive DTDT changes (Fig. 10a–d), similar to the ERA5 analysis (see Fig. 9a–d in Part 1). During warming events, the temperature and pressure evolution along the backward trajectories is very similar for trajectories initialised at t-1 and t between 3 and 1 d before reaching the target location (Fig. 10i and m), while diabatic heating within the last 24 h plays a significant role for the DTDT increase (Fig. 10q). Therefore, we analyse the physical processes on a 1 d time scale (Fig. 11a), in contrast to the 3 d–analysis presented above for the extratropics. Composite precipitation and cloud cover are further analysed in Fig. 12 to explore these local diabatic effects. On day t-1, significant precipitation (5–8 mmd-1) and high cloud cover (80 %–90 %) reduce solar radiation and thus diabatic heating, leading to lower temperatures. Conversely, on day t, decreased precipitation (≤4 mmd-1) and less cloud cover (70 %–80 %) enhance diabatic heating, contributing to higher temperatures (Figs. 10m, 12b). This shift from wet and cloudy to dry and less cloudy conditions highlights the role of albedo changes and solar heating in driving a +2 °C increase over 1 d, along with enhanced diabatic heating (+2.2 °C; Fig. 11a). The magnitudes of DTDT warming events in ERA5 and CESM–LE are nearly identical, reflecting a similar scale of diabatic heating but an underestimation of the adiabatic contribution in CESM-LE (see also Fig. 9k in Part 1 and Fig. 11a).

The localized patterns persist during future DTDT warming events, with high air particle densities at -3 d continuing to cluster in the grid box and towards its northeast (reduced in the South) on both days (Fig. 10e–f). Thus, changes in remote advection and adiabatic warming play only a minor role in future DTDT changes (Figs. 10j, n and 11c). The projected warming is primarily driven by diabatic heating (nearly +1 °C) near the surface over the last 24 h, emphasising the enhanced role of radiative heating (Figs. 10r and 11c). This finding is further supported by a significant increase in precipitation and cloud cover on day t-1, which, however, subsequently decreases on day t, contributing to the intensified DTDT warming in the future climate (Fig. 12i–j). Also, for tropical Southern Africa (Fig. S17c) and Southeast Asia (not shown), the intensification of DTDT warming events is mainly driven by diabatic heating.

Also, for the historical DTDT cooling events, the distribution of air masses on both days is concentrated around the grid box, with remote advection and adiabatic heating playing little role for the temperature decrease (Figs. 10c–d, k, o, s, 11b). These events are primarily driven by local diabatic effects near the surface in the 24 h preceding the arrival of the air masses at the target location, with magnitudes comparable to those in the ERA5 reanalysis, but biases in the adiabatic contribution (see Fig. 9l in Part 1). Similar to DTDT warming events, on day t-1, precipitation (8–10 mmd-1) and cloud cover (70 %–80 %) are smaller, leading to increased diabatic heating and higher temperatures, compared to the larger precipitation (>10 mmd-1) and cloud cover (80 %–90 %) on day t, which result in decreased diabatic heating and lower temperatures (Figs. 11b and 12c–d). This transition from dry, less cloudy to wet, cloudy conditions highlights the significant role of changes in albedo and solar heating in driving surface diabatic cooling.

In future DTDT cooling events, local effects are projected to remain dominant, with increased air particle densities around the grid box and its north (reduced in the southeast) compared to the historical climate (Fig. 10g–h). While the contributions of changes in advection to the future intensification of DTDT cooling are small, a reversal in the role of adiabatic warming is the primary factor for this intensification (Fig. 11d). In contrast to the historical climate, in which a stronger descent and adiabatic warming on day t compared to t-1 slightly offsets the diabatically–driven cooling (Figs. 10k, 11b), in the future, the ascent and adiabatic warming are projected to weaken on day t, leading to increased cooling (Fig. 10l, p). In addition, an intensification of diabatic cooling (Figs. 10t, 11d) associated with a marked decline in precipitation and cloud cover on day t-1, followed by an increase on day t (Fig. 12k–1), further enhances DTDT cooling. A consistent pattern is also evident in Southeast Asia (not shown). A synergy between adiabatic and diabatic processes, albeit with a larger contribution from diabatic changes, leads to future intensification of DTDT cooling events over tropical Southern Africa (Fig. S17d).

During JJA, some sub–tropical and mid–latitude regions, such as the Sahel, parts of Europe, Southern Asia, Central America, and the Amazon, are projected to experience an increase in extreme DTDT changes relative to the historical climate. To study the underlying atmospheric circulation and physical processes, we select a grid box over central Europe (Figs. 13–15). Further examples for locations in the subtropics (Figs. S19–S21) and midlatitudes (Figs. S22–S24) are shown in the Supplement.

During JJA warming in central Europe in the historical CESM–LE simulations, a mature trough–ridge anomaly pattern shifts southeastward from day t-1 to t (Figs. S19a, 13a–b), bringing central Europe under the influence of southeasterly winds and warm continental air masses on the event day (Fig. 14a–b), which aligns with the ERA5 synoptic pattern (see Figs. 6a–c and S6a–b in Part 1). This transition enhances warm air advection (Fig. 14m), which plays a crucial role in driving the temperature increase, resulting in an average temperature rise of +3.2 °C (Fig. 15a). In contrast, adiabatic warming makes a minor negative contribution, averaging -0.3 °C, due to slightly more substantial subsidence on day t-1 than on day t (Fig. 14i, m). However, the effect of adiabatic warming varies considerably across individual events, shown by the bars and whiskers (Fig. 15a). Enhanced diabatic heating contributes an average of +2.5 °C to the DTDT increase, with a more pronounced rise in potential temperature on day t than on day t-1 (Fig. 14q). Overall, warm air advection and diabatic heating are the primary contributors to DTDT warming events, with adiabatic warming playing a minor role in the historical climate (Fig. 15a). A quantitative comparison reveals that CESM–LE indicates contributions from advection and diabatic heating similar to those of ERA5, but the adiabatic warming is slightly underestimated (see Fig. 7k, Part 1).

In general, the synoptic–scale flow pattern associated with future DTDT warming events resembles the historical pattern (Fig. 13e, f). Notable differences are nonetheless evident upstream over the North Atlantic, where the low-pressure anomaly weakens in future projections (Fig. S19c). Concurrently, the adjacent continental ridge experiences a structural reorganisation: its southwestern branch weakens, while the northeastern branch slightly intensifies and shifts eastward at day t (Fig. 13i–j). Collectively, these circulation changes reduce the inflow of southwesterly maritime air masses and shift toward continental sources, contributing to enhanced DTDT warming (Figs. 14e–f, S19c). These subtle changes in circulation patterns result in a modest increase in warm–air advection, contributing approximately +0.5 °C to the DTDT change (Fig. 15c). Meanwhile, vertical descent intensifies on day t-1 (by ∼5 hPa) but diminishes notably on day t (by ∼10 hPa), leading to a negative adiabatic warming contribution that offsets the DTDT increase by -1.1 °C (Fig. 14j, n). Conversely, diabatic heating increases on both days compared to the historical period, with a greater increase on day t (Fig. 14r). This intensified diabatic heating adds +1.9 °C to the DTDT change. Overall, the combined effect of slightly increased warm air advection and amplified diabatic heating – despite being partly offset by enhanced adiabatic warming – appears to be the main factor driving the projected intensification of DTDT warming events (Fig. 15c). A similarly important role of diabatic heating intensification is also evident in future DTDT warming events over the selected subtropical regions: Southern Asia and the Sahel (Figs. S20–S21), as well as in Central America and northern Amazon (not shown). In contrast, changes in advection are more significant for events in mid-latitudes: southern South Africa, Northern Asia, Southern Australia (Figs. S22–S24), and northern Europe (not shown).

During JJA cooling events in the historical climate, there is a transition from warm continental air masses to colder maritime air masses, coinciding with the development and eastward shift of a North Atlantic ridge (Figs. 13c–d, 14c–d and S19b). This cold air advection causes an average decline in surface temperature of -5.5 °C (Figs. 14o, 15b). The cooling is partly offset by a modest increase in adiabatic warming (∼0.7 °C), driven by stronger descent of air masses on day t (Fig. 14k, o). Additionally, reduced diabatic heating contributes a further temperature drop of -0.5 °C (Fig. 14s). Overall, strong cold air advection and slightly reduced diabatic heating are the main drivers of DTDT cooling events over central Europe under historical conditions (Fig. 15b), which also agrees with the ERA5 reanalysis (see Fig. 7l in Part 1). In the CESM–LE, the contributions from advection and adiabatic warming are somewhat smaller, but the diabatic contribution is similar to that in ERA5.

For projected DTDT cooling events, on both days, the synoptic atmospheric circulation over the continent remains largely similar to that in the historical climate (Fig. 13c–d, g–h). Nevertheless, notable changes are expected along the North Atlantic ridge. The high-pressure anomaly weakens at both the surface and 500 hPa (Fig. S19d), and an eastward shift of the ridge becomes evident at day t(Fig. 13k–l). Regarding air-mass origins, a reduction in southwesterly maritime air masses and a concurrent increase in continental northeasterly air masses are evident, both more pronounced on day t-1 than on day t (Figs. 13k–l, and 14g–h). Together, these changes slightly reduce the contribution of horizontal temperature advection to the cooling signal (+0.3 °C). Projected changes in adiabatic warming are negligible (Figs. 14l, 15d), such that changes in diabatic heating are the primary contributor to the future DTDT cooling (-2.2 °C) intensification, which results from a much stronger intensification of the diabatic heating of air masses on day t-1 compared to day t (Figs. 14t, 15d). Changes in diabatic heating are also the main driver of the future intensification of DTDT cooling over Southern Asia, the Sahel, Northern Asia (Figs. S20–21, S23), Central America, the northern Amazon, and Northern Europe (not shown). Either advection or all processes contribute more equally to events over southern South Africa, Northern Asia, and Southern Australia (Figs. S22–S24).