Warm conveyor belts in present-day and future climate simulations. Part II: Role of potential vorticity production for cyclone intensification

. Warm conveyor belts (WCBs) are strongly ascending, cloud and precipitation forming airstreams in extratropical cyclones. The intense cloud-diabatic processes produce low-level cyclonic potential vorticity (PV) along the ascending airstreams, which often contribute to the intensification of the associated cyclone. This study investigates how climate change affects the cyclones’ WCB strength and the importance of WCB-related diabatic PV production for cyclone intensification, based on present-day 5 (1990-1999) and future (2091-2100) climate simulations of the Community Earth System Model Large Ensemble (CESM-LE). In each period, a large number of cyclones and their associated WCB trajectories have been identified in both hemispheres during the winter season. WCB trajectories are identified as strongly ascending air parcels that rise at least 600 hPa in 48 hours. Compared to ERA-Interim reanalyses, the present-day climate simulations are able to capture the cyclone structure and the associated WCBs reasonably well, which gives confidence in future projections with CESM-LE. However, the amplitude of the 10 diabatically produced low-level PV anomaly in the cyclone center is underestimated in the climate simulations, most likely because of reduced vertical resolution compared to ERA-Interim. The comparison of the simulations for the two climates reveals an increase in the WCB strength and the cyclone intensification rate in the Southern Hemisphere (SH) in the future climate. The WCB strength also increases in the Northern Hemisphere (NH), but to a smaller degree, and the cyclone intensification rate is not projected to change considerably. Hence, in the two hemispheres cyclone intensification responds differently to an 15 increase in WCB strength. Cyclone deepening correlates positively with the intensity of the associated WCB, with a Spearman correlation coefficient of 0.68 (0.66) in the NH in the present-day (future) simulations, and a coefficient of 0.51 (0.55) in the SH. The number of explosive cyclones with strong WCBs, referred to as C1 cyclones, is projected to increase in both hemispheres, while the number of explosive cyclones with weak WCBs (C3 cyclones) is projected to decrease. A composite analysis reveals that in the future climate C1 cyclones will be associated with even stronger WCBs, more WCB-related diabatic 20 PV production, the formation of a more intense PV tower, and an increase in precipitation. They will become warmer, moister, and slightly more intense. The findings indicate that (i) latent heating associated with WCBs (as identified with our method) will increase, (ii) WCB-related PV production will be even more important for explosive cyclone intensification than in the present-day climate, and (iii) the interplay between dry and moist dynamics is crucial to understand how climate change affects cyclone intensification.

The idealised simulations as well as the climate model studies demonstrate that the role of enhanced diabatic heating for cyclone development in a warming climate is not yet fully understood. As the strongest diabatic heating in extratropical cyclones occurs in WCBs, the question arises how WCB-related diabatic PV production changes with global warming. In an accompanying study by Joos et al. (2022), a climatology of WCB trajectories has been calculated for the first time in climate model 100 simulations with the Community Earth System Model Large Ensemble (CESM-LE; Kay et al., 2015), and their geographical distribution, seasonal frequency and characteristics have been investigated in the present-day and end-of-century climate. It was shown that the present-day simulations are able to realistically capture the geographical distribution and frequency of occurrence of the WCBs in many regions. In the future simulations, overall the WCB frequency maxima are located in similar regions as in the present climate, but there are also some geographical shifts, and the total number of WCB trajectories 105 increases. In regional WCB hotspots like the North and South Atlantic, North Pacific and Mediterranean, there is an increase in the WCB inflow moisture and in the precipitation and the diabatic heating rate along the ascending trajectories, and the maximum in the diabatic heating rate shifts upward.
The projected overall increase in the number of WCB trajectories and in the diabatic heating rate along the trajectories in a warmer climate potentially influences the development of the associated cyclones. In this study, we use the same WCB 110 climatology as in Joos et al. (2022) and combine it with a cyclone climatology to evaluate how WCBs and their associated diabatic PV production affect cyclone intensification in present-day and future CESM-LE simulations during NH and SH winter. In addition to a climatological analysis of all cyclones, we investigate potential future changes in the characteristics of a subset of cyclones with particularly strong WCBs and explosive deepening, referred to as C1 cyclones (see Binder et al., 2 Data and Methods

Climate simulations and reanalysis data
The study is based on output from an initial condition ensemble of the Community Earth System Model (CESM), version 1 (Hurrell et al., 2013). The ensemble data was created by rerunning the CESM Large Ensemble (CESM-LE; Kay et al., 2015) 130 with restart files from the original simulations (see Röthlisberger et al., 2020, for a detailed description). The re-simulations were produced to obtain high-resolution three-dimensional model level output, such as wind fields, which are required for the computation of WCB trajectories. The data is available every 6 h at a horizontal resolution of 1.25 • longitude by ∼ 0.9 • latitude and on 30 vertical levels. Two time periods are simulated, 1990-1999 for the present-day climate (hereafter referred to as CESM-HIST), which is based on historical forcing (Lamarque et al., 2010), and 2091-2100 for the future climate (hereafter 135 referred to as CESM-RCP85), based on the RCP8.5 emission scenario Meinshausen et al., 2011). Both time periods are simulated with five ensemble members, which only differ by small perturbations in the initial atmospheric temperature field. In total, this yields 50 simulated years for each time period. The analysis is confined to the winter season in both hemispheres, i.e., December-February for the NH and June-August for the SH.
To validate the ability of CESM-LE to represent extratropical cyclones and their associated WCBs, the present-day simula- 140 tions are compared to the corresponding fields in ERA-Interim reanalyses of the European Centre for Medium-Range Weather Forecasts (Dee et al., 2011) for the period 1979-2014. The ERA-Interim data set is the same as in Binder et al. (2016). The variables are available every 6 h on 60 levels in the vertical, and they have been interpolated from a spectral resolution of T255 to a 1 • by 1 • horizontal grid. As for the CESM simulations, we restrict the analysis to the winter season in both hemispheres.

145
Extratropical cyclones are identified based on the algorithm of Wernli and Schwierz (2006) that has been refined in Sprenger et al. (2017). Cyclones are defined as two-dimensional features delimited by the outermost closed isobar that encompasses a local sea level pressure (SLP) minimum or several minima. In addition, a tracking algorithm determines for each SLP minimum the most probable continuation among the minima identified 6 h later and thereby follows the cyclones from genesis to maturity to lysis. The analysis is restricted to cyclones with a lifetime of at least 48 h, and to exclude tracks with unrealistic lifecycles, 150 the pressure difference between the outermost closed SLP contour and the SLP minimum at the beginning of the track needs to be less than 5 hPa (see Binder et al., 2016, for details). Furthermore, we only consider cyclones located poleward of 25 • N and S during at least a 24 h interval, to exclude tropical cyclones.
The identification of WCB trajectories is the same in ERA-Interim and the CESM simulations and is described in detail in Madonna et al. (2014) and Joos et al. (2022), respectively. Using the Lagrangian analysis tool (LAGRANTO; Wernli and 155 Davies, 1997; Sprenger and Wernli, 2015), 48 h forward trajectories are started every 6 h from a horizontally equidistant grid with ∆x = 80 km resolution and from vertically equidistant (∆p = 20 hPa) levels in the lower troposphere (1050-790 hPa).
To be selected as WCB air parcels, the trajectories must undergo an ascent of at least 600 hPa within 48 h. In addition, their horizontal position must coincide with the surface field of an extratropical cyclone at least once during the 48 h. ries are not able to cross the year end. This implies that the last four days of each year are not included in the WCB climatology.
However, we expect that this only has a negligible impact on the results.

Link between cyclone deepening and WCB strength
To study the link between cyclone deepening and WCB strength (as defined below), the WCB trajectories are assigned to cyclones. In ERA-Interim, each WCB trajectory is assigned to the first cyclone with which it overlaps during the 48 h ascent, 165 as in Binder et al. (2016). In CESM-LE, WCB trajectories located close to each other are first clustered with a similar clustering algorithm as the one described in Catto et al. (2015b). All trajectories in the cluster are attributed to the same cyclone, namely the cyclone with which the highest number of WCB trajectories overlaps when they are located between 850 and 500 hPa.
Both in ERA-Interim and in CESM-LE this yields for every cyclone during the entire lifecycle the number of associated WCB trajectories and their properties. The slightly different procedures to link WCB trajectories with cyclones is expected to only 170 negligibly influence the results. Indeed, the number of WCB trajectories per cyclone is very similar in ERA-Interim and the present-day simulations of CESM-LE, as will be discussed in Section 3.
Cyclone deepening is measured in Bergeron units (Sanders and Gyakum, 1980). Over each 24 h interval during a cyclone lifecycle, the latitude-adjusted SLP deepening within 24 h is calculated as follows: where ∆SLP is the change in the minimum SLP and φ represents the cyclone's average latitude during the 24 h time period.
For each cyclone, the deepening rate ∆SLP B,max is then determined as the largest value in ∆SLP B over all 24 h intervals during the cyclone lifecycle. Sanders and Gyakum (1980) defined explosively intensifying cyclones (so-called "bombs") as cyclones with a deepening rate of more than 1 Bergeron, which corresponds to a SLP drop of more than 24 hPa in 24 h at 60 • latitude.

180
The cyclone's WCB strength is quantified by the number of WCB trajectories assigned to that cyclone and located in the lower troposphere (pressure > 500 hPa) at any time during the 24 h of strongest deepening. The trajectories do not necessarily need to be located within the cyclone area during these 24 h, as they or -for CESM-LE -part of their cluster can overlap with the cyclone area at any time during their two-day ascent. Since the WCB trajectories were started from an equidistant grid (see Section 2.2), each trajectory is associated with the same air mass, i.e., ∆m ≈ 1 g (∆x) 2 ∆p ≈ 1.3×10 12 kg, with g = 9.81m s −2 .

185
Thus, multiplication of the number of WCB trajectories with ∆m yields the WCB air mass.  Table 1). Figure 1a shows, for NH winter, percentile curves of the intensification rates of all cyclones (∆SLP B,max ), separately for CESM-HIST and ERA-Interim. Hereby, the curve indicates the percentage of cyclones (with respect to the total number as indicated in Table 1) with a specific intensification rate or lower. For instance, 90% of the cyclones have a deepening rate ≤ 1.35 Bergeron in CESM-HIST and ≤ 1.5 Bergeron in ERA-Interim, while the remaining 10%, i.e., the 10% most strongly  To assess whether CESM-HIST is able to adequately represent the structure and properties of extratropical cyclones and the associated WCBs, we investigate the average fields of a subset of cyclones with a compositing method. We restrict our analysis 225 to cyclones with particularly strong deepening and intense WCBs, referred to as C1 cyclones. The category borders of the C1 cyclones are displayed in Fig. 2. They are identical to those defined in Binder et al. (2016), i.e., explosively deepening cyclones (> 1 Bergeron) with a WCB intensity of at least 2.78 × 10 15 kg (corresponding to 2130 WCB trajectories). In ERA-Interim, this yields 500 C1 cyclones in the NH, which corresponds to 9.9% of the total cyclone number, and 330 C1 cyclones in the  the NH and 509 in the SH), but the fraction from the total cyclone number is with 9.4% and 7.5%, respectively, very similar to the one in ERA-Interim. The composites are created in the middle of the 24 h interval of strongest cyclone intensification and centred at the location of the SLP minimum. 1 Note that here we are mainly interested in the differences between CESM-HIST and ERA-Interim, a detailed description of the structure and evolution of the C1 composite cyclone in ERA-Interim for NH winter can be found in Binder et al. (2016). Here we only show C1 composites for NH winter, but the findings are very similar 235 for SH winter.  Low-level (upper-level) WCB frequencies indicate the fraction of C1 cyclones that has at least one WCB trajectory located at p > 500 hPa (p < 500 hPa) at a specific position. The spatial distribution of the frequencies agrees well between reanalysis and climate model. In both cases, high low-level WCB frequencies occur in the warm sector and above the warm front of the composite cyclone, with maximum values of more than 80% close to the cyclone centre. The highest upper-level WCB frequencies amount to about 60% and are located slightly to the north of the peak low-level values. Additionally shown is the 245 frequency of low-level WCB air parcels located between 900 and 700 hPa that reach high PV values above 1 pvu (HLPV WCB frequencies, green contours), which highlights the regions with most pronounced WCB-related diabatic PV production. The HLPV trajectories are co-located with the cyclone centre and their maximum frequencies amount to approximately 65% in ERA-Interim and 55% in CESM-HIST.

Climate model evaluation
The potential temperature structure and the horizontal temperature gradient at 850 hPa are very similar in ERA-Interim 250 and CESM-HIST (Fig. 3c,d). At upper levels, in both data sets a jet streak is located southwest of the cyclone centre. Also the precipitation pattern agrees well between ERA-Interim and CESM-HIST, with non-zero values in the entire warm sector (Fig. 3c,d). Maximum values near the cyclone centre amount to 17 mm h −1 in ERA-Interim and 16 mm h −1 in CESM-HIST.
In both cases, they coincide with the region of strongest WCB-related PV production (green contours in Fig. 3a tribute to the further intensification of the composite cyclone, until 12 h later, at the end of the strongest intensification phase, the positive low-and upper-level PV anomalies align vertically and form a troposphere-spanning PV tower (Hoskins, 1990;Rossa et al., 2000;Badger and Hoskins, 2001), which is associated with an intense cyclonic wind field (for ERA-Interim, see The number of cyclones per winter is projected to decrease in the future climate by 3% in NH winter (from 132 to 128 cyclones) and 7% in SH winter (from 136 to 126 cyclones; Table 1). This is consistent with the findings from many previous studies (e.g., Bengtsson et al., 2009;Grieger et al., 2014;Priestley and Catto, 2022). To evaluate potential changes in cyclone deepening 280 rates and WCB strengths in the future climate, we look at the differences in the percentile curves between the future and the present-day climate simulations (CESM-RCP85 minus CESM-HIST; Fig. 4). Positive differences for specific percentiles indicate an increase and negative differences a decrease in the deepening rates (Fig. 4a,b) and WCB strengths (Fig. 4c,d) in the future climate. To detect statistically significant differences, a confidence interval has been constructed by performing a bootstrap resampling. Hereby, our null hypothesis was that the cyclone intensification rates and the WCB strengths per 285 cyclone, respectively, do not differ between the present-day and future climate simulations, i.e., that they belong to a common distribution. From this common distribution, 100'000 resamples have been created by randomly attributing half of the values to CESM-HIST and half of the values to CESM-RCP85 and then calculating the difference in the percentile curves. The 2.5th and the 97.5 percentiles of the ranked differences of the resampling distribution correspond to the lower and upper limits of the 95% confidence interval. If the actual difference lies outside this interval, it is statistically significant with a probability of 290 95%.
In NH winter, the cyclone intensification rate is projected to increase slightly in the future climate for the weakest as well as the most rapidly intensifying cyclones and to decrease for the 50th-90th percentile, but the changes are smaller than 0.05 Bergeron and not statistically significant for almost all percentiles (Fig. 4a). In SH winter, there is a statistically significant increase in the intensification rates by about 0.03-0.05 Bergeron for medium-strong cyclones (70th-90th percentile) and by 295 about 0.05-0.1 Bergeron for strongly intensifying cyclones above the 90th percentile (Fig. 4b). Intensification rates typically range from about −0.5 to 3 Bergeron (Fig. 2), such that an increase by 0.05-0.1 Bergeron for the most strongly intensifying  In both hemispheres, in cyclones that are associated with WCBs, the WCBs become stronger in the future climate (Fig. 4c, Table 1). In the SH, the changes are statistically significant for almost all percentiles (Fig. 4d). The largest increases amount to about 2 × 10 15 kg, which corresponds to 1500 WCB trajectories per cyclone or 11%. In the NH, the changes are much smaller and only statistically significant for the 80th-95th percentile (Fig. 4c).
There, the increases amount to about 0.4 − 0.6 × 10 15 kg, i.e., 300-460 WCB trajectories per cyclone or 2-3% 305 The significant increase in the cyclone intensification rate in the SH but not in the NH could in part be due to the much stronger increase in the WCB strength and accordingly the WCB-related latent heating in the SH. In addition, the differences in the intensification rates between the two hemispheres are consistent with the opposite changes in low-level baroclinicity expected with global warming (e.g., Harvey et al., 2014;Catto et al., 2019, see also introduction): In the SH, low-level baroclinicity is expected to increase, which favourably interacts with the increased WCB strength and leads to stronger cyclones, 310 whereas in the NH it is expected to decrease, which counteracts the effects of the increased WCB strength such that the cyclone intensity does not change considerably or even decreases.
The statistical correlation between the cyclone deepening rate and the WCB strength is shown in the two-dimensional histograms in Fig. 2c-f. In both hemispheres, cyclone deepening correlates positively with the intensity of the associated WCB in the present-day and future climate simulations, indicating that WCBs continue to play an important role for cyclone 315 intensification in a warming climate. In the NH, the Spearman correlation coefficient has a value of 0.68 in CESM-HIST ( Fig. 2c) and 0.66 in CESM-RCP85 (Fig. 2e). In the SH, it increases from 0.51 in CESM-HIST (Fig. 2d) to 0.55 in CESM-RCP85 (Fig. 2f).
Despite the overall decrease in the number of cyclones per winter in both hemispheres (Table 1), the number of C1 cyclones, i.e., explosively deepening cyclones with strong WCBs, is projected to increase (Fig. 5). In the NH, it increases by 11% from 320 12.3 to 13.7 C1 cyclones per winter (Fig. 5a) and in the SH by 23% from 10.2 to 12.5 cyclones (Fig. 5b). In both hemispheres, there is also a ∼25% increase in the percentage of so-called C2 cyclones, which are defined as in Binder et al. (2016) to have a similar WCB strength as C1 (at least 2.78 × 10 15 kg) but a weak deepening rate of less than 0.8 Bergeron (see Fig. 2 for the category boundaries). At the same time, the percentage of C3 cyclones, i.e., explosively intensifying cyclones with weak WCBs of less than 0.33 × 10 15 kg, decreases by 11% in the NH and 21% in the SH. The total number of explosive cyclones 325 irrespective of the associated WCB strength (so-called bombs) decreases by 6% in the NH and increases by 2% in the SH. The changes in the phase-space diagram in Fig. 2, i.e., the increase in the number of C1 and C2 cyclones and the decrease in C3, are in line with the overall increase in the cyclone-related WCB strength observed in Fig. 4c,d and indicate that cyclones become more diabatic in a future climate.
To sum up, in the SH CESM-LE projects an increase in the WCB strength, the cyclone intensification rate and the correlation 330 between the two in the future climate winter. An increase is also projected in the total number of explosive cyclones and those with strong WCBs (C1), and in the number of weak cyclones with strong WCBs (C2), whereas the number of "dry bombs" with weak WCBs (C3) decreases. In the NH, the WCB strength is also projected to increase, but to a smaller extent than in the SH, and there are no significant changes in the cyclone intensification rates. While the number of C1 and C2 cyclones increases, there is a decrease in the total number of explosive cyclones and in the number of C3 cyclones. In the following, 335 we investigate whether the C1 cyclones themselves will change with regard to spatial distribution, structure and temporal evolution, in addition to their increase in number.

Future changes in the characteristics of C1 cyclones
At the beginning of their strongest intensification, both in the present-day and future climate simulations most C1 cyclones are located in the western and central parts of the NH and SH oceans, where WCBs are particularly frequent (not shown).

340
However, in the future simulations, the intensification starts about 1.5 • farther poleward in the NH and 2 • farther poleward in the SH (Table 2). In both hemispheres, the poleward shift is relatively small in the Atlantic and larger in the Pacific basin. The shift is in agreement with previous studies (e.g., Yin, 2005;Bengtsson et al., 2009;Priestley and Catto, 2022).
The time evolution of minimum SLP and the WCB air mass in the lower troposphere shows for both simulations the explosive deepening of the C1 cyclones and, concomitantly, a rapid increase in the WCB strength, which peaks during the strongest 345 deepening phase (Fig. 6a,b). Consistent with Fig. 4c,d, in both hemispheres but especially in the SH the peak WCB intensity is higher in the future than in the present-day simulations. At the same time, in the future simulations the deepening rate of the C1 cyclones increases slightly and the minimum SLP at the time of the cyclone's strongest intensity is 2 hPa deeper in the  NH and 4 hPa deeper in the SH. Thus, in addition to their increase in number (Fig. 5), C1 cyclones also become slightly more intense in a warming climate. 8c,e,g). However, in the NH the PV gradient along the trough is stronger and the jet more intense. Also the equator-to-pole potential temperature gradient at 850 hPa is stronger than in the SH (cf. Figs. 7e and 8e). On the other hand, C1 cyclones in the SH are associated with higher low-level humidity, a larger WCB frequency, higher diabatically produced low-level PV and deeper SLP (cf. Figs. 7a,c,g and 8a,c,g).
The composite cyclones' surroundings become significantly warmer and moister in the future climate. In the cyclone centre, 360 potential temperature, equivalent potential temperature and specific humidity at 850 hPa increase by about 3 K, 6 K and 1.5-2 g kg −1 in both hemispheres. (Figs. 7a,b,e,f and 8a,b,e,f). The relative humidity at 850 hPa, on the other hand, does not change considerably near the cyclone centre, as it is already close to saturation in CESM-HIST (Figs. 7a,b and 8a,b). The equator-topole potential temperature gradient at 850 hPa in the environment of the C1 cyclones decreases slightly in the NH (Fig. 7e,f) and increases slightly in the SH (Fig. 8e,f), consistent with the overall changes in low-level baroclinicity projected for the two 365 hemispheres.  climate, i.e., it becomes more positive in the NH (Fig. 9a,b) and more negative in the SH (Fig. 9c,d). The upper-level disturbance to the west of the cyclone centre also corresponds to a cyclonic PV anomaly, and it intensifies as well in the future climate.
However, the relative increase of the peak values is much larger for the cyclonic low-level PV anomaly than for the cyclonic upper-level PV anomaly (25.5% vs 11.1% in the NH and 14% vs. 7.2% in der SH, Tables 3 and 4). This suggests that the enhanced deepening and stronger intensity of the C1 cyclones in the future simulations observed in Fig. 6 is mainly associated 390 with the enhanced WCB-related diabatic PV production, and in the SH additionally with the increased low-level baroclinicity.
Downstream of the cyclone centre, the elevated tropopause goes along with anticyclonic upper-level PV anomalies, which also amplify in the future climate ( Fig. 9 and Tables 3, 4). This amplification is consistent with the intensification of the WCB and the WCB-related latent heating, which results in a stronger transport of anticyclonic PV into the tropopause region (see introduction).

395
We also investigated the composite structure of the C1 cyclones 12 h later, at the end of the 24 h interval of explosive deepening (not shown). Also at this time step, the WCB-related diabatic PV production intensifies in the future climate (Fig. 6) and contributes to the formation of a stronger PV tower (not shown) and 1.7 hPa deeper minimum SLP in the NH and 2.5 hPa deeper minimum SLP in the SH, respectively (Fig. 6). The maximum values of the cyclonic low-level PV anomalies increase by almost 50% in the NH and 19% in the SH in the future climate, whereas the peak values of cyclonic upper-level PV only showing the PV anomaly (pvu; shading), the 2 pvu contour (thick black contour) and potential temperature (thin black contours every 5 K).
The fields are shown for (a, b) NH winter and (c, d) SH winter in (a, c) CESM-HIST and (b, d) CESM-RCP85.  increase by about 12% and 5%, respectively (Tables 3, 4). The anticyclonic upper-level PV anomalies downstream of the cyclone centre also increase (Tables 3, 4) and push the tropopause slightly further upward and toward the poles (not shown).
In summary, in both hemispheres it is projected that C1 cyclones will have even stronger WCBs, stronger WCB-related diabatic PV production, an extension of the diabatically produced PV toward higher levels and an increased precipitation rate in the future climate. They will become warmer, moister and slightly more intense.

Summary and conclusions
In this study, we quantified the role of WCBs and their diabatically produced PV anomalies for cyclone intensification in present-day (1990)(1991)(1992)(1993)(1994)(1995)(1996)(1997)(1998)(1999) Meinshausen et al., 2011). The aims were to (i) assess whether the climate model is able to adequately represent the cyclone properties and the associated WCBs by comparing the present-day simulations with ERA-Interim reanalyses, and (ii) evaluate how climate change affects the importance of WCB-induced diabatic PV production for cyclone intensification in the winter season in both hemispheres. Such a detailed investigation of the role of WCBs for cyclone intensification in climate models is not straightforward and was only possible because CESM-LE has been re-run to obtain 6-hourly three-dimensional 415 output of the wind fields. To address the aims, a large number of cyclones and their associated WCB trajectories have been identified in the climate simulations and in ERA-Interim during NH and SH winter. Cyclone deepening has been measured by the maximum 24-h SLP change during the cyclone lifecycle, adjusted by latitude (Sanders and Gyakum, 1980), and the WCB intensity by the WCB air mass located in the lower troposphere (p > 500 hPa) during the 24 h interval of strongest intensification of the associated cyclone. Based on the questions posed in the introduction, the key findings of the study can be 420 summarised as follows: 1. Compared to ERA-Interim, the climate model is able to represent the properties and three-dimensional structure of extratropical cyclones, as well as the associated WCBs, remarkably well. Particularly in SH winter, there are very small differences between the present-day simulations and the reanalyses in terms of the cyclone deepening rates, the WCB strength and the statistical relationship between the two. In NH winter, the present-day simulations capture the deepening 425 rates of the weak and medium-strong cyclones, but they underestimate them for the most explosive cyclones by 0.1-0.3 Bergeron. The WCB strength and the link between WCB strength and cyclone deepening rate is well captured also in the NH. In the subgroup of explosive cyclones with intense WCBs (C1), in both hemispheres the model is able to reproduce the composite fields of the cyclones during the deepening phase in terms of upper-level PV and jet structure, low-level potential temperature, SLP, the precipitation pattern and the position of the WCB trajectories, but it 430 underestimates the diabatically produced low-level PV anomaly by about 0.5 pvu.
2. In the SH, comparison of the simulations reveals an increase in the WCB strength and -for the medium-strong and strongly intensifying cyclones -the cyclone intensification rate in the future climate. In the NH, the WCB strength is also projected to increase, but to a smaller extent than in the SH, and overall there are no significant changes in the cyclone deepening rates. The enhanced cyclone deepening rate of the medium-strong and strongly deepening cyclones 435 in the SH but not in the NH could partly be associated with the stronger increase in the WCB strength and accordingly the WCB-related diabatic heating in the SH. Furthermore, it is consistent with the opposite changes in low-level baroclinicity projected in the two hemispheres with global warming (e.g., Harvey et al., 2014;Catto et al., 2019), i.e., an increase in the SH, which favourably interacts with the moist dynamics to create stronger storms, and a decrease in the NH, which counteracts the direct effects of the moist dynamics such that the storm strength does not change significantly.
full-daily/levtype=sfc/, last access: July 2022). The CESM source code that was used for the CESM-LE simulations is available from https://www.cesm.ucar.edu/models/cesm1.0/ (last access: July 2022). The model output of the CESM-LE reruns and the WCB and cyclone data used in this study are available from the authors upon request.
Author contributions. HB designed and performed this study, using WCB trajectories previously identified by MS and HJ. HB wrote the manuscript, with feedback about the results and text from all co-authors.