Seasonal amplification of subweekly temperature variability over extratropical Southern Hemisphere land masses

Seasonal amplification of subweekly temperature variability over extratropical Southern Hemisphere land masses | npj Climate and Atmospheric Science

Abstract

Temperature variability has substantial socioeconomic impacts through its association with the frequency and severity of heat extremes. Under anthropogenic influence, climate models project seasonally-dependent amplifications of near-surface temperature variability over some sectors of the Southern Hemisphere, and robust positive trends have already been observed in recent decades. Here we show that the amplification of subweekly temperature variability simulated by the multi-model ensemble mean of the sixth phase of the Coupled Model Intercomparison Project (CMIP6) over South Africa, Australia, and South America is often substantially smaller than in reanalyses in recent decades, reaching a similar amplification only at the end of the 21st century due to a weaker amplification of subweekly variance generation efficiency. Analysis of a large model ensemble indicates that this discrepancy may be due to internal climatic variability suggesting that the recent rapid amplification seen in reanalyses may slow down or even temporarily reverse in the near future.

Similar content being viewed by others

Introduction

Subweekly variability in the extratropics is primarily associated with migratory cyclones and anticyclones that are organized in major storm tracks 1, 2, 3 and cause severe socioeconomic impacts via their influence on temperature fluctuations and precipitation. In the Southern Hemisphere (SH), climate models project an overall wintertime amplification 4, 5, 6, 7, 8 of the storm track activity and a summertime poleward shift 4, 7, 9, 10, 11. Concerning the associated changes in synoptic-scale temperature variability, which is primarily caused by temperature advection 12, 13, CMIP5 models project an amplification over SH mid-latitude landmasses: over South Africa and Australia in winter and South America in summer 11, 14. This amplification has been robustly captured over the recent decades in reanalysis datasets and surface observations with the largest amplification of subweekly temperature variability occurring in September-October-November (SON) for South Africa and Australia and December-January-February (DJF) for South America 15. Since temperature variability is important in setting the severity of heat extremes 16, 17, improving future projections is vital for the planning of adaptation measures.

Projected trends in subweekly variability, whether due to an overall SH-wide amplification or latitudinal shift of the major storm tracks are often linked with the equator-to-pole gradient of temperature 7, 18 and associated changes in the westerly jets 19, 20, 21. The projected strengthening and poleward shift result from comparatively weaker upper-level warming of the SH high-latitude troposphere 7 affecting both the meridional temperature gradient and static stability to impact the development of eddies 20, 22. Tropopause height rise 23, and changes in eddy phase speed and meridional propagation 24 can also contribute to the trends. The impact of the large-scale atmospheric thermal structure is also identified in idealized model experiments 20, 25. Concerning temperature variability, the detailed spatial distribution of trends, which is inhomogeneous, depends on the distribution of regional trends in the meridional temperature gradient. The locally amplifying gradient allows meridionally traveling air parcels to cause greater temperature anomalies 11, 14, despite it is sometimes counteracted by a reduction of their meridional displacement 11. From an Eulerian perspective, the regional amplification of seasonal-mean temperature gradients acts to increase the temperature variance generation efficiency of subweekly eddies by horizontal temperature advection 15.

It was recently observed that the amplification of the wintertime storm track activity as captured in the vertically-integrated eddy kinetic energy (EKE) is happening at a much faster pace in observations-based reanalyses compared to models 6 of the sixth phase of the Coupled Model Intercomparison Project (CMIP6) 26 on average. It is happening to such an extent that the amplification projected by the multi-model mean to be realized by 2100 has already been seen in reanalyses and the greatest amplification projected among all models is less than half of what is seen in reanalyses. This important discrepancy raises concerns about the ability of climate models to reproduce the processes driving trends in atmospheric variability and it remains unknown how these discrepancies affect projections of other weather observables such as temperature variability which was not investigated along with EKE 6. Understanding how these discrepancies might affect populated areas in the SH is vital for optimal planning of adaptation strategies by policymakers. We thus present in this article a detailed and updated assessment of trends in near-surface subweekly temperature variability affecting midlatitude SH landmasses as projected by CMIP6 models and their performance in comparison to atmospheric reanalyses. The dynamical mechanisms underlying these differences as well as inter-model uncertainties are also investigated.

Results

Future projections

We first assess long-term (1980–2100) trends in 850-hPa subweekly temperature variability (TVAR; see methods) over the SH in 26 models that participate in CMIP6 (Fig. 1). The 850-hPa level is commonly used to analyze processes driving temperature variability and trends 11, 14, 27. Important TVAR amplifications are projected to affect midlatitude SH landmasses. The amplification is especially large over South Africa in June-July-August (JJA) and SON, Australia in SON, and South America in JJA, SON, and DJF, representing TVAR amplifications around its climatological maxima. These amplifications are notably larger in the most pessimistic scenario (SSP5-8.5) compared to a scenario representing the stabilization of global warming (SSP2-4.5). Even in the latter, the aforementioned trends are statistically significant as supported by a remarkable agreement among the diverse models. The differences between the two scenarios are found primarily in the amplitude of trends rather than in their spatial distribution. These trends generally agree with those projected by previous model generations for similar temperature variance metrics 11, 14.

Linear TVAR trends (shadings; K2 year-1) are assessed for 1980–2100 for each season (rows) in the (left; a, c, e, g) SSP2-4.5 and (right; b, d, f, h) SSP5-8.5 scenarios. For reference, the climatology of TVAR is shown with black contours at intervals of 1.5 K2. Sectors where there is no agreement on the sign of the trends among 2/3 of reanalyses or models are stippled with green stars. Sectors where trends are not significant are stippled in magenta. The false discovery rate method 50 is used to assess significance with a false discovery control value (αFDR) of 0.1. Sectors used for area-averaging are shown as dashed magenta rectangles in the bottom-right panel. Sectors below topography are masked in gray.

Fig. 2 summarizes the evolution of TVAR over South America, South Africa, and Australia (see Fig. 1 for their boundaries). Overall, the long-term evolution of TVAR projected under the SSP2-4.5 and SSP5-8.5 scenarios tend to be similar, except for South Africa in JJA and SON and for Australia in SON, where SSP5-8.5 begins to show a markedly larger amplification after ~2049, although its multi-model mean remains within the inter-model spread of SSP2-4.5.

Time series of area-mean TVAR (K2) are shown for different seasons (rows) and landmass sectors in the SH (columns) for the mean of reanalyses and CMIP6 models. The three sectors investigated are illustrated with dashed rectangles in the bottom-right panel of Fig. 1. The inter-model or inter-reanalysis spread is illustrated as ± 1 σ shaded intervals. The beginning of the satellite era in 1979 is illustrated with dotted lines.

Contrasting recent trends in simulations and reanalyses

We then compare the spatial distribution of TVAR trends between reanalyses and CMIP6 models for 1980–2022 (Fig. 3). TVAR trends in reanalyses are spatially inhomogeneous in the SH 15. In the SH extratropics, both positive and negative trends are observed, although positive trends prevail overall, especially in the mid-to-high latitudes. Over SH landmasses, robust positive trends are found over South Africa and Australia in SON as well as over South America in DJF and March-April-May (MAM), corresponding to local amplifications of climatological maxima of TVAR. In contrast, the CMIP6 multi-model mean depicts weaker and more spatially homogeneous changes that are structurally similar, but weaker than long-term trends (Fig. 1). This is likely due to the cancellation of internal variability that affects the regional scale temperatures 28 in the multi-model mean. Still, they are nonetheless not completely homogeneous throughout the SH, suggesting that it can be vital in climate models to adequately represent stationary waves giving rise to these regional asymmetries and their evolution under anthropogenic forcing for accurate attribution and projection of subweekly temperature variability. In JJA, the models simulate a tendency for TVAR trends to amplify equatorward of the climatological maximum and decrease poleward over the Atlantic and Indian Ocean sectors. This is an indication of an overall equatorward shift of the TVAR maxima that affects South Africa. Over Australia and South Africa in SON, by contrast, the simulated amplification is collocated with the climatological local maxima in TVAR along the southern coasts of both sectors. In DJF, TVAR is simulated to amplify poleward and weaken equatorward of the SH-wide climatological maxima, similar to the projected change in EKE in CMIP5 models 9, indicative of its poleward shift. Although overall the spatial distribution is not similar to that of reanalyses, an amplification is also simulated over South America in all seasons.

Linear TVAR trends (shadings; K2 year-1) are assessed over 1980–2022 for each season (rows) in the (left; a, c, e, g) reanalysis ensemble mean, and (right; b, d, f, h) the CMIP6 model mean. The CMIP6 mean is based on the combined historical and SSP5-8.5 scenarios. For reference, the climatology of TVAR is shown with black contours at intervals of 1.5 K2. Sectors where 2/3 of reanalyses or models agree on the sign of the trends are hatched in green. Significant trends are hatched in magenta. The false discovery rate method 50 is used to assess significance with a false discovery control value (αFDR) of 0.1. Sectors used for area-averaging are shown as dashed green rectangles in the bottom-right panel. Sectors below topography are masked in gray.

Over the satellite era (1979–2022), we observe relatively similar TVAR evolution in South America between the reanalyses and models, especially in JJA and SON (Fig. 2). Over South Africa, the reanalyses show generally larger TVAR than the model average for all seasons except DJF. Reanalyzed and simulated evolutions of TVAR in Australia are comparatively more similar, especially in JJA and MAM. In the AMIP simulations where observed sea surface temperature (SST) and sea ice concentration are imposed (1979–2014), as opposed to freely evolving in the historical, SSP2-4.5, and SSP5-8.5 simulations, some of the model-reanalysis discrepancies in TVAR are reduced. This is especially obvious for South Africa in JJA and MAM. However, the mitigation is insufficient, and over some sectors and seasons, such as in DJF over South America, it results in a slightly larger discrepancy from reanalysis compared to coupled simulations.

Prior to the introduction of satellite observations into reanalysis data assimilation in 1979, some of the largest discrepancies between reanalyses and models are observed. During that period, TVAR tends to be larger in the models in almost all seasons and locations except South Africa and Australia in JJA and MAM. The ability of reanalyses to adequately represent TVAR is most likely hindered during this pre-satellite era due to the reduced availability of upper-air observations as observational constraints. We note however that surface-input reanalyses that do not assimilate upper-air observations (20CRv2c 29, 20CRv3 30, ERA-20C 31) depict an evolution of TVAR that is reasonably consistent with full-input reanalyses both before and after 1979 15. This result suggests that surface observations could have mostly constrained 850-hPa temperature over SH landmasses and therefore that the evolution of TVAR before 1979 may be trustable to some extent. The long-term evolution of TVAR was also shown to be broadly consistent with gridded surface temperature observations from the Berkeley Earth dataset 32. Nonetheless, we hereafter focus on the satellite era for further comparisons to benefit from enhanced observational constraints.

Uncertainties in trends cannot be inferred from Fig. 2 alone. To this end, a more detailed analysis of the TVAR trends follows in Fig. 4. Trends in reanalyses over the 1980–2022 period are clearly larger than simulation averages in Australia in JJA, South Africa in SON, and South America in DJF. We note that if considering the 1970–2022 period, reanalyzed trends are also significantly larger than simulation averages in South Africa in JJA (not shown; cf. Fig. 2). The reanalyzed trends are within the large spread in trends simulated among the individual CMIP6 models, which could result from different model physics and/or internal variabilities. For other sectors and seasons some positive or negative differences are also found but still within the spread among the CMIP6 models. Whereas not all models agree concerning TVAR trends over the 1980–2022 period due to the low signal-to-noise ratio, they almost unanimously predict positive trends over 1980–2100 to affect South Africa in all seasons except DJF, and Australia in JJA and SON. We note that the CMIP6 models agree less concerning future trends projected to affect South America.

TVAR trends (K2 decade-1) over the three landmass sectors (columns) and four seasons (rows) investigated are assessed for two different periods delimited by vertical dashed lines and indicated at the bottom of each panel. Mean trends in reanalyses or CMIP6 model scenarios that are available for each period are reported with colored bars with the color convention shown at the bottom. The % change per decade relative to the reference period 1980–2014 is indicated beside each bar. The inter-quartile range and median among reanalyses or CMIP6 models are shown with boxes and the full range is shown with whiskers. A large ensemble of 30 members is shown for the SSP5-8.5 scenario of MPI-ESM1-2-LR. Transparency indicates trends that are not significant at the 5% level as assessed with a t-test.

In almost all seasons and locations, TVAR trends in reanalyses are included within the large spread among members of the MPI-ESM1-2-LR large ensemble 33, indicating that internal climatic variability can explain the difference between reanalyses and simulated forced changes, in contrast to the model deficiency previously speculated in the case of EKE 6. An exception is found for South Africa in SON when some of the reanalyses evaluated present larger trends compared to the most extreme ensemble member, suggesting that some model bias is partly contributing to the differences between models and reanalyses.

Comparing trends in reanalyses to trends in the MPI-ESM1-2-LR large ensemble from which the ensemble mean is removed (Supplementary Fig. 1), we observe that the earlier rise above the 95th percentile of ensemble trends in South America in DJF and Australia in JJA, suggesting that these amplifications are emerging out of the range that can occur by internal variability alone. The trend is also above the 95th percentile in South Africa in SON, but because of the potential model discrepancy as mentioned above, the emergence is inconclusive.

Supplementary Fig. 2 shows the same analysis repeated for the 1980–2014 period to allow comparison with AMIP experiments. Differences between reanalyses and the historical simulations are generally similar to those found over 1980–2022. A notable exception is found over Australia in SON when the model difference with respect to reanalyses becomes negative over 1980–2022, as opposed to 1980–2014, due to the prominent amplification of TVAR in the reanalyses over the past decade (Fig.

Role of baroclinic processes

An explanation for the discrepancies in TVAR trends between the reanalyses and CMIP6 multi-model mean is sought in the dominant generation term of temperature variability (Fhoriz) which describes how TVAR is generated by subweekly horizontal advective fluxes of heat that are against the seasonal-mean temperature gradient (see Methods). The efficiency of the meridional component of TVAR generation \(({F}_{y}^{{eff}}\); Fig. 5) is substantially larger than its zonal counterpart \(({F}_{x}^{{eff}}\); Supplementary Fig. 3) and as a result is responsible for most of TVAR generation. \({F}_{y}^{{eff}}\) is typically larger in the vicinity of midlatitude land masses such as the eastern coast of South America and just south of South Africa and Australia, which correspond to sectors of stronger temperature gradients (Supplementary Fig. 4) associated with the land-sea temperature contrasts. These are also sectors where the correlation between \(-v\mbox{'}\) and \(T\mbox{'}\) tends to exhibit local maxima, denoting optimal eddy structure for the production of meridional heat fluxes (Supplementary Fig. 5). The magnitude and spatial distribution of these climatological properties are similar between reanalyses and CMIP6 models.

Same as Fig. 3 but for the efficiency of the meridional component of Fhoriz, \({F}_{y}^{{eff}}\) (shaded; K m-1 year-1). For reference, the climatology of \({F}_{y}^{{eff}}\) is shown with black contours at intervals of 1.5 × 10-1 K m-1.

Figure 5 however illustrates a sharp contrast in \({F}_{y}^{{eff}}\) trends between reanalyses and simulations. Whereas both show prominent amplifications of \({F}_{y}^{{eff}}\) over sectors of positive TVAR trends (comparing with Fig. 3), the CMIP6 models exhibit more moderate trends on average. Like TVAR trends, trends in \({F}_{y}^{{eff}}\) are typically within the wide range of trends among the models, except for South Africa in SON and MAM when the positive trend is almost three times larger than the simulation average and for South America in MAM when the trends are opposite (Supplementary Fig. 6). \({F}_{x}^{{eff}}\), although generally weaker, plays a nonnegligible role over the eastern coasts of South Africa and South America (Supplementary Fig. 3). We find that the differences in \({F}_{y}^{{eff}}\) between the reanalyses and CMIP6 models as well as their spatial distributions are strongly associated with trends in the regional temperature gradients (Supplementary Fig. 4), confirming the role of the background temperature gradient in setting TVAR trends 11, 14. But there is an additional contribution from changes in the correlation between \(v\mbox{'}\) and \(T\mbox{'}\) that describe the eddies’ effectiveness at producing heat fluxes (Supplementary Fig. 5). Trends in the latter are substantially more prominent in reanalyses compared to CMIP6 models. Simulated long-term trends (1980–2100) in these efficiencies exhibit overall similar geographical distribution to their historical counterpart (not shown).

As further evidence of the role of \({F}_{y}^{{eff}}\), we find that the inter-model differences in TVAR trends are strongly linked to the magnitude of \({F}_{y}^{{eff}}\) in the vicinity of the land mass sectors examined (Fig. 6). Inter-model correlations can reach up to ~0.6–0.8, especially near South Africa (SON) and Australia (SON) and South America (DJF). This relationship is not expected to be limited to the land mass sectors of TVAR trends as TVAR is redistributed horizontally by the climatological mean-flow advection which is typically eastward. Another factor explaining nonlocal correlations is that inter-model diversity in TVAR trends are organized in broader patterns. For example, TVAR trend differences over South Africa (SON) are linked with TVAR trend differences that span a large portion of SH mid-latitudes and subtropics indicating that local diversity is associated with broader inter-model differences in the large-scale circulation. Similar correlation patterns are found in temperature gradients, further highlighting the importance of trends in the broader-scale thermal structure of the atmosphere (Supplementary Fig. 7) for uncertainties in temperature variability.

Correlation among models of 1980–2100 trends in \({F}_{y}^{{eff}}\) at each grid point with trends in TVAR for the same period (shadings) averaged over three representative sectors illustrated in individual panels in magenta. Significant correlations (p < 0.05) are shown with white hatching. The season, which is also indicated in each panel, differs among the sectors. For reference, the correlation is also assessed for TVAR trends at each grid point (black contours; 0.2 intervals; solid and dashed lines for positive and negative correlations, respectively; the 0 lines are omitted). Sectors below topography are masked in gray.

Discussion

Given the impact of changing lower-tropospheric temperature variability on the occurrence of weather extremes 16, it seems crucial to further improve its representation in climate models. The large differences between reanalyzed and simulated mean TVAR trends over some sectors, like the differences in EKE 6, may be interpreted as a deficit in the representation of synoptic variability in climate models. The reanalyzed TVAR trends assessed for recent decades are nevertheless within the inter-model spread that arises from differences in model physics and internal variability, as also suggested by the MPI-ESM1-2-LR large ensemble 33 where the ensemble spread results from internal variability alone. It is thus possible that the discrepancies between the reanalyses and CMIP6 ensemble mean originate from internal variability that is present in reanalyses but largely averaged out of the model ensemble mean. These findings, therefore, do not undermine the general capability of CMIP6 models to project future TVAR changes. The range of possible future outcomes may be narrowed in future work by considering the external forcing response of stationary waves, which lead to spatial inhomogeneities in future changes in the background temperature gradient, and the strength of synoptic eddies’ feedback to these mean flow changes. This approach is supported by the significant role of temperature anomaly generation efficiency in inter-model uncertainty.

If the discrepancy between reanalyses and CMIP6 models is indeed due to internal variability that has accelerated the rise in TVAR in the recent reanalysis record, we may expect a slowdown, or even a temporary reversal of the trends in the future, akin to the global warming “hiatus”, which temporarily paused the warming in the early 2000s 34, 35. This is supported by the large MPI-ESM1-2-LR ensemble whose members with the largest amplification over 1980–2020 show signs of a slowing or even a reversing of trends in the subsequent decades depending on the location and season (Supplementary Fig. 8).

Since ocean-atmosphere coupling was found to be a key factor in the future poleward shift of the summertime storm track 9, we briefly assessed whether the representation of oceanic variability could be a source of model bias. While the AMIP simulations (1980–2014) can reduce biases in the basic state and trends in some instances, their influence on trends, as well as their discrepancies from reanalyses, is inconsistent across seasons and sectors. These inconsistencies may stem from internal variability that leads to a weak signal-to-noise ratio when assessing trends over such a short period.

Notwithstanding the discrepancies between the reanalyses and CMIP6 models, we find that all CMIP6 models project important positive TVAR trends (1980–2100) affecting South Africa and Australia in JJA and SON (Figs. 1 and 4). The largest projected increases in South Africa occur in JJA and represent a TVAR amplification of 2.50 ± 0.53% per decade (relative to 1980–2014) in the SSP2-4.5 scenario which may represent the most likely outcome 36, or 4.59 ± 0.97% in what may be the worst case scenario (SSP5-8.5). For Australia, the projected trends are markedly larger in SON with 1.38 ± 0.21% and 2.58 ± 0.40% increases per decade in SSP2-4.5 and SSP5-8.5, respectively. If proven accurate, these sectors should prepare themselves for a substantial increase in subweekly temperature variability along with profound socioeconomic impacts.

Finally, contrasting TVAR trends to EKE trends (Supplementary Figs. 9 and 10), we observe that they do not always agree. For instance, the SON TVAR amplification over Australia in the reanalyses is accompanied by a decrease in EKE in its vicinity both at 250 and 850 hPa. There is also a general lack of agreement between the patterns of TVAR and EKE trends in CMIP6 models. This decoupling between near-surface temperature variance and EKE indicates that conventional storm-track metrics based on winds alone are not appropriate to assess surface impacts, at least concerning the potential impact on heat extremes. Instead, a direct assessment of lower-tropospheric variability is advised based on this study.

Methods

Data

Atmospheric data from reanalyses are used as an approximation of the past state of atmospheric circulation. The Climate Forecast System Reanalysis version 1 (CFSR 37) and version 2 (CFSv2 38) are combined into a single dataset at their transition between December 2010 and January 2011. Other datasets include the ECMWF Reanalysis v5 (ERA5 39), the Japanese 55-year Reanalysis (JRA-55 40), and the Modern-Era Retrospective analysis for Research and Applications, Version 2 (MERRA-2 41). Details about their models and data assimilation techniques have been compiled as part of the S-RIP project 42. Climate model projections are obtained from 26 models participating in the CMIP6 project 26: ACCESS-CM2, ACCESS-ESM1-5, CAMS-CSM1-0, CESM2-WACCM, CMCC-CM2-SR5, CNRM-CM6-1, CNRM-CM6-1-HR, CNRM-ESM2-1, CanESM5, EC-Earth3, EC-Earth3-CC, EC-Earth3-Veg, EC-Earth3-Veg-LR, FGOALS-g3, GFDL-CM4, HadGEM3-GC31-LL, INM-CM4-8, INM-CM5-0, IPSL-CM6A-LR, MIROC-ES2L, MIROC6, MPI-ESM1-2-HR, MPI-ESM1-2-LR, MRI-ESM2-0, NESM3, NorESM2-LM. Model integrations forced by the historical, Shared Socioeconomic Pathway scenario 2-4.5 (SSP2-4.5), and Shared Socioeconomic Pathway scenario 5-8.5 (SSP5-8.5) are investigated. To assess long-term trends, the historical (before 2014) and SSP (after 2015) scenarios are combined. AMIP-type experiments in which SST is prescribed are also considered. The first ensemble member of each model (i.e., ‘r1i1p1’) is used in this study except for MPI-ESM1-2-LR 33 for which we use 30 ensemble members to assess uncertainties associated with internal climate variability. MPI-ESM1-2-LR is chosen for the large ensemble analysis because it provides the largest number of ensemble members for which daily data is available for both historical and future climate scenarios. We note that MPI-ESM1-2-LR is not necessarily the model that shows the best agreement with respect to reanalysis data. Like many other models, differences are improved or worsened compared to the CMIP6 model mean depending on location and season. Similar results are obtained with the CanESM5 43 large ensemble for the period 1980–2014 (not shown) but the analysis cannot be extended to 1980–2022 due to the limited number of ensemble members for which daily data is available for future climate scenarios. Variables analyzed include temperature (\(T\)), pressure velocity (\({\rm{\omega }}\)), and the zonal and meridional wind components (\(u\) and \(v\), respectively). Daily averages are used.

Subweekly temperature variability and its leading sources and sinks

By decomposing temperature and winds into the different timescales they are composed of (e.g., \(T=\overline{T}+T^{\prime} +T^{\prime\prime}\), where the overline, ′, and ″ denote the seasonal mean, subweekly, and subseasonal time scales, respectively), substituting these components into the atmospheric thermodynamic equation, and then multiplying by \(T^{\prime}\), we obtain an equation expressing the generation and dissipation of subweekly temperature variability (\(\overline{{{T}^{{\prime} }}^{2}}\) or TVAR):

In Eq. 1, only the leading generation and dissipation terms are retained. They are the generation by horizontal wind (1st RHS term, or Fhoriz) and vertical wind (2nd RHS term, or Fvert). \(\chi\) includes interactions between the subweekly and subseasonal time scales, advection by the seasonal-mean winds which redistribute TVAR spatially but do not act as overall sources/sinks, diabatic processes, and in the case of reanalysis data, inconsistencies introduced by data assimilation (or analysis increment). Subweekly variability is here extracted through a 10-day high-pass filter (denoted with primes). By design, the filter allows oscillations with periods up to ~7 days to fully pass through while oscillations with a 10-day period are damped by ½.

Fhoriz and Fvert are considered separately because of the different processes they represent from the perspective of atmospheric energetics 44, 45. While both Fhoriz and Fvert are linked to the conversion of available potential energy (APE) from the seasonal-mean horizontal temperature gradient to subweekly eddies, Fvert receives an additional contribution from the conversion of eddy APE (~TVAR) to EKE. In fact, Fvert is primarily influenced by the latter. In agreement with the typical flows of energy 46, TVAR is usually generated by Fhoriz and dissipated by Fvert 15. In this work, we consider only the impact of the source term Fhoriz.

Like TVAR, Fhoriz is a function of the product of eddy terms. Because of this, the amplification of subweekly variability can both amplify TVAR and Fhoriz, thus obscuring the underlying source of the amplification. In this context, it is more informative to assess changes in the efficiency of energy conversion relative to eddy amplitude. To assess this efficiency, we here scale Fhoriz by the typical magnitude of wind and temperature variations to obtain

for the meridional component of Fhoriz efficiency and

for its zonal component. We note that with such scaling, the eddy terms now describe the correlation between wind and temperature fluctuations rather than their covariance. Their correlation is determined by the structure of subweekly eddies that will influence, via their horizontal motions, how wind and temperature fluctuations are synchronized in time. It is a measure for baroclinic eddy growth that generates subweekly variability in the midlatitudes to have westward-tilting structures with height, which is equivalent to negatively correlated \(v^{\prime}\) and \(T^{\prime}\) in the SH thus producing poleward heat fluxes against the equatorward background temperature gradient. We note that the efficiency defined here differs from the efficiency in the energetics framework 47, 48, 49 where the source and sink terms are normalized by the total energy (EKE + EAPE).

In addition to TVAR, we also briefly investigate trends in EKE defined as

in Supplementary Figs. 9 and 10.

These diagnostics are assessed independently in each reanalysis, model, and ensemble member before assessing the ensemble means. When reported as area averages, these diagnostics are averaged over the grid points located above the land masses included in each sector illustrated in Figs. 1, 3, and 5.

Data availability

JRA-55, CFSR, and CFSv2 data were obtained from the research data archive ( https://rda.ucar.edu/). MERRA-2 was obtained from the NASA Goddard Earth Sciences Data and Information Services Center ( https://disc.gsfc.nasa.gov/datasets?project=MERRA-2) and ERA5 was obtained from the climate data store ( https://doi.org/10.24381/cds.bd0915c6). CMIP6 model data was obtained from the Earth System Grid Federation (e.g., https://esgf-node.llnl.gov/projects/cmip6/).

Code availability

Code can be provided upon request.

1.
Nakamura, H., Sampe, T., Tanimoto, Y. & Shimpo, A. Observed associations among storm tracks, jet streams and midlatitude oceanic fronts. Geophys. Monogr. 147, 329–345 (2013).
2.
Hoskins, B. J. & Hodges, K. I. A new perspective on Southern Hemisphere storm tracks. J. Clim. 18, 4108–4129 (2005).
3.
Nakamura, H. & Shimpo, A. Seasonal variations in the Southern Hemisphere storm tracks and jet streams as revealed in a reanalysis dataset. J. Clim. 17, 1828–1844 (2004).
4.
O’Gorman, P. A. Understanding the varied response of the extratropical storm tracks to climate change. Proc. Natl Acad. Sci. USA 107, 19176–19180 (2010).
5.
Shaw, T. A. et al. Storm track processes and the opposing influences of climate change. Nat. Geosci. 9, 656–664 (2016).
6.
Chemke, R., Ming, Y. & Yuval, J. The intensification of winter mid-latitude storm tracks in the Southern Hemisphere. Nat. Clim. Chang. 12, 553–557 (2022).
7.
Harvey, B. J., Shaffrey, L. C. & Woollings, T. J. Equator-to-pole temperature differences and the extra-tropical storm track responses of the CMIP5 climate models. Clim. Dyn. 43, 1171–1182 (2014).
8.
Lehmann, J., Coumou, D., Frieler, K., Eliseev, A. V. & Levermann, A. Future changes in extratropical storm tracks and baroclinicity under climate change. Environ. Res. Lett. 9, 084002 (2014).
9.
Chemke, R. The future poleward shift of Southern Hemisphere summer mid-latitude storm tracks stems from ocean coupling. Nat. Commun. 13, 1730 (2022).
10.
Chang, E. K. M., Guo, Y. & Xia, X. CMIP5 multimodel ensemble projection of storm track change under global warming. J. Geophys. Res. Atmos. 117, 1–19 (2012).
11.
Tamarin-Brodsky, T., Hodges, K., Hoskins, B. J. & Shepherd, T. G. A dynamical perspective on atmospheric temperature variability and its response to climate change. J. Clim. 32, 1707–1724 (2019).
12.
Röthlisberger, M. & Papritz, L. A global quantification of the physical processes leading to near‐surface cold extremes. Geophys. Res. Lett. 50, 1–10 (2023).
13.
Holmes, C. R., Woollings, T., Hawkins, E. & de Vries, H. Robust future changes in temperature variability under greenhouse gas forcing and the relationship with thermal advection. J. Clim. 29, 2221–2236 (2016).
14.
Schneider, T., Bischoff, T. & Płotka, H. Physics of changes in synoptic midlatitude temperature variability. J. Clim. 28, 2312–2331 (2015).
15.
Martineau, P., Behera, S. K., Nonaka, M., Nakamura, H. & Kosaka, Y. Seasonally dependent increases in subweekly temperature variability over Southern Hemisphere landmasses detected in multiple reanalyses. Weather Clim. Dyn. 5, 1–15 (2024).
16.
Simolo, C. & Corti, S. Quantifying the role of variability in future intensification of heat extremes. Nat. Commun. 13, 7930 (2022).
17.
Schär, C. et al. The role of increasing temperature variability in European summer heatwaves. Nature 427, 332–336 (2004).
18.
Yin, J. H. A consistent poleward shift of the storm tracks in simulations of 21st century climate. Geophys. Res. Lett. 32, 1–4 (2005).
19.
Vallis, G. K., Zurita‐Gotor, P., Cairns, C. & Kidston, J. Response of the large‐scale structure of the atmosphere to global warming. Q. J. R. Meteorol. Soc. 141, 1479–1501 (2015).
20.
Butler, A. H., Thompson, D. W. J. & Heikes, R. The steady-state atmospheric circulation response to climate change–like thermal forcings in a simple general circulation model. J. Clim. 23, 3474–3496 (2010).
21.
Son, S.-W. & Lee, S. The response of westerly jets to thermal driving in a primitive equation model. J. Atmos. Sci. 62, 3741–3757 (2005).
22.
Wu, Y., Ting, M., Seager, R., Huang, H.-P. & Cane, M. A. Changes in storm tracks and energy transports in a warmer climate simulated by the GFDL CM2.1 model. Clim. Dyn. 37, 53–72 (2011).
23.
Lorenz, D. J. & DeWeaver, E. T. Tropopause height and zonal wind response to global warming in the IPCC scenario integrations. J. Geophys. Res. Atmos. 112, 1–11 (2007).
24.
Chen, G. & Held, I. M. Phase speed spectra and the recent poleward shift of Southern Hemisphere surface westerlies. Geophys. Res. Lett. 34, L21805 (2007).
25.
Lim, E.-P. & Simmonds, I. Effect of tropospheric temperature change on the zonal mean circulation and SH winter extratropical cyclones. Clim. Dyn. 33, 19–32 (2009).
26.
Eyring, V. et al. Overview of the Coupled Model Intercomparison Project Phase 6 (CMIP6) experimental design and organisation. Geosci. Model Dev. Discuss. 8, 10539–10583 (2016).
27.
Garfinkel, C. I. & Harnik, N. The Non-Gaussianity and spatial asymmetry of temperature extremes relative to the storm track: the role of horizontal advection. J. Clim. 30, 445–464 (2017).
28.
Arias, P. A. et al. Technical Summary. in Climate Change 2021 – The Physical Science Basis 35–144 (Cambridge University Press, 2023).
https://doi.org/10.1017/9781009157896.002.
29.
Compo, G. P. et al. The Twentieth Century Reanalysis Project. Q. J. R. Meteorol. Soc. 137, 1–28 (2011).
30.
Slivinski, L. C. et al. Towards a more reliable historical reanalysis: Improvements for version 3 of the Twentieth Century Reanalysis system. Q. J. R. Meteorol. Soc. 145, 2876–2908 (2019).
31.
Poli, P. et al. ERA-20C: an atmospheric reanalysis of the twentieth century. J. Clim. 29, 4083–4097 (2016).
32.
Rohde, R. A. & Hausfather, Z. The Berkeley earth land/ocean temperature record. Earth Syst. Sci. Data 12, 3469–3479 (2020).
33.
Olonscheck, D. et al. The New Max Planck Institute Grand Ensemble With CMIP6 Forcing and High‐Frequency Model Output. J. Adv. Model. Earth Syst. 15, 1–21 (2023).
34.
Kosaka, Y. & Xie, S.-P. The tropical Pacific as a key pacemaker of the variable rates of global warming. Nat. Geosci. 9, 669–673 (2016).
35.
Fyfe, J. C. et al. Making sense of the early-2000s warming slowdown. Nat. Clim. Chang. 6, 224–228 (2016).
36.
Hausfather, Z. & Peters, G. P. Emissions – the ‘business as usual’ story is misleading. Nature 577, 618–620 (2020).
37.
Saha, S. et al. The NCEP climate forecast system reanalysis. Bull. Am. Meteorol. Soc. 91, 1015–1058 (2010).
38.
Saha, S. et al. The NCEP Climate Forecast System Version 2. J. Clim. 27, 2185–2208 (2014).
39.
Hersbach, H. et al. The ERA5 global reanalysis. Q. J. R. Meteorol. Soc. 146, 1999–2049 (2020).
40.
Kobayashi, S. et al. The JRA-55 reanalysis: general specifications and basic characteristics. J. Meteorol. Soc. Jpn. Ser. II 93, 5–48 (2015).
41.
Gelaro, R. et al. The Modern-Era Retrospective Analysis for Research and Applications, Version 2 (MERRA-2). J. Clim. 30, 5419–5454 (2017).
42.
Fujiwara, M. et al. Introduction to the SPARC Reanalysis Intercomparison Project (S-RIP) and overview of the reanalysis systems. Atmos. Chem. Phys. 17, 1417–1452 (2017).
43.
Swart, N. C. et al. The Canadian Earth System Model version 5 (CanESM5.0.3). Geosci. Model Dev. 12, 4823–4873 (2019).
44.
Lorenz, E. N. Available potential energy and the maintenance of the general circulation. Tellus 7, 157–167 (1955).
45.
Oort, A. H. On estimates of the atmospheric energy cycle. Mon. Weather Rev. 92, 483–493 (1964).
46.
Sheng, J. & Derome, J. An observational study of the energy transfer between the seasonal mean flow and transient eddies. Tellus A 43, 128–144 (1991).
47.
Kosaka, Y. & Nakamura, H. Structure and dynamics of the summertime Pacific–Japan teleconnection pattern. Q. J. R. Meteorol. Soc. 132, 2009–2030 (2006).
48.
Tanaka, S., Nishii, K. & Nakamura, H. Vertical structure and energetics of the western pacific teleconnection pattern. J. Clim. 29, 6597–6616 (2016).
49.
Martineau, P., Nakamura, H., Kosaka, Y. & Yamamoto, A. Importance of a vertically tilting structure for energizing the North Atlantic Oscillation. Sci. Rep. 10, 12671 (2020).
50.
Wilks, D. S. The stippling shows statistically significant grid points”: how research results are routinely overstated and overinterpreted, and what to do about it. Bull. Am. Meteorol. Soc. 97, 2263–2273 (2016).

Acknowledgements

This study is in part supported by the Japan Society for the Promotion of Science (JSPS) through Grants-in Aid for Scientific Research (JP19H05702, JP19H05703, JP20H01970, JP22H01292, JP23K22573, and JP24H02223), by the Japanese Ministry of Education, Culture, Sports, Science and Technology (MEXT) through the Arctic Challenge for Sustainability II (ArCS-II; JPMXD1420318865) and the Advanced Studies of Climate Change Projection (SENTAN; JPMXD0722680395), by the Japan Science and Technology Agency through COI-NEXT (JPMJPF2013), by the Japanese Ministry of Environment through Environment Research and Technology Development Fund JPMEERF20242001.

Ethics declarations

Competing interests

The authors declare no competing interests

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License, which permits any non-commercial use, sharing, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if you modified the licensed material. You do not have permission under this licence to share adapted material derived from this article or parts of it. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc-nd/4.0/.

About this article

Cite this article

Martineau, P., Nakamura, H., Kosaka, Y. et al. Seasonal amplification of subweekly temperature variability over extratropical Southern Hemisphere land masses. npj Clim Atmos Sci 7, 258 (2024). https://doi.org/10.1038/s41612-024-00804-0

-
Received:
-
Accepted:
-
Published:
-
DOI: https://doi.org/10.1038/s41612-024-00804-0