(C) PLOS One
This story was originally published by PLOS One and is unaltered.
. . . . . . . . . .



Atmosphere and ocean energy transport in extreme warming scenarios [1]

['Alyssa N. Poletti', 'Department Of Atmospheric Sciences', 'University Of Washington', 'Seattle', 'Washington', 'United States Of America', 'Dargan M. W. Frierson', 'Travis Aerenson', 'Akshaya Nikumbh', 'Geophysical Fluid Dynamics Laboratory']

Date: 2024-02

All three models produce an equatorward shift of the ITCZ in the warm period. Based on surface wind convergence in the piControl runs, the ITCZ is at 2.8–4.5° N. By 2300, the convergence shifts to approximately 1.2–2.6° N. The vast majority of the equatorward shift of the ITCZ occurs after 2100, especially in CanESM5 where the ITCZ shifts more than 3 degrees northward by 2100 and 4 degrees southward after. The changes in ITCZ location are consistent with the changes in tropical precipitation seen in Fig 6 and described above.

The bottom row of Fig 7D–7F shows the zonally averaged zonal wind anomalies. The jet stream, defined by the maximum in zonal wind, rises along with the tropopause, decreasing in pressure by 50 to 100 hPa, with a larger rise in the models that warm more. Changes in zonal wind below the tropopause are on the order of several m/s, representing large fractional changes in wind strength especially near the surface. The large poleward shift of the SH jet streams are consistent with the time series of the SH stormtrack, measured as the maximum surface zonal wind, which shows a decrease in latitude between approximately 3.0 and 7.5 degrees ( Fig 10 ).

The 30-year running mean of (A, D) the NH Polar-Ferrel Cell boundary, (B, E) the SH Polar-Ferrel Cell boundary (C, F) and the SH storm track location versus time (top row), and versus the global mean temperature anomaly (bottom row) from 1850 to 2300. The time and zonally averaged pre-Industrial values are denoted by an X. In the time series figures (top row), the historical data (1850–2015) are denoted by the dashed lines, while the SSP-Ext runs (2015–2300) are denoted by the solid lines. In the temperature anomaly figures (bottom row), the time and time-mean historical data is denoted by a circle. The piControl stormtrack latitudes for CanESM5, IPSL-CM6A-LR, and MRI-ESM2-0 are 48.7S, 48.8S, and 47.9S respectively.

Time series of the Ferrel Cell boundary in the NH and SH as well as the SH stormtrack latitude are shown in Fig 10 , both as a function of year and of global temperature change. The boundary between the Ferrel Cell and the Polar Cell also expands with warming in each model ( Fig 10 ). In every model there is greater expansion of the NH Ferrel Cell than the NH Hadley cell, which is especially pertinent in CanESM5, where the NH Hadley cell boundary shrinks equatorward, while the NH Ferrel cell boundary expands poleward. The absolute location of the SH Polar-Ferrel boundary, however, is always at higher latitude than the NH boundary; however, both the SH and NH Polar-Ferrel boundaries approach a limit in their poleward extent. Barnes & Hartmann (2012) postulate that there is a poleward limit to cyclonic wave breaking in midlatitude jets and that the SH jets have already reached this extent [ 65 ]. This is consistent with the lesser poleward motion we see in the SH Polar-Ferrel boundary compared to that in the NH.

The first value is for the NH and the second value is for the SH (except for the ITCZ which has one location in the NH). See Table 2 for methods. All values included have a p-value of less than 0.05. All of the regressions are done on yearly averages.

The first value is for the NH and the second value is for the SH (except for the ITCZ which has one location in the NH). The Hadley-Ferrel and ITCZ locations were found by interpolating for subgrid zeros in the 500hPa streamfunction. The outer Ferrel Cell boundary was found by interpolating for subgrid zeros in the zonally averaged zonal wind. The stormtrack was found by interpolating for subgrid zeros in the numerical derivative of surface zonal wind with respect to latitude. Values which do not have a p-value less than 0.05 are italicized. All of the regressions are done on yearly averages.

IPSL-CM6A-LR, by contrast, exhibits large NH tropical widening, even more than its SH tropical widening (in final extent and in rate of expansion; Tables 2 and 3 ). The larger SH Hadley Cell expansion in CanESM5 and MRI-ESM2-0 (when compared to the NH) is consistent with previous literature [ 64 ].

(A, B, C) The 30-year running mean NH and SH Hadley-Ferrel Boundary location and ITCZ location versus time, and (D, E, F) global mean temperature anomaly from 1850 to 2300. The time and zonally averaged pre-Industrial values are denoted by an X. In the time series figures (top row), the historical data (1850–2015) are denoted by the dashed lines, while the SSP-Ext runs (2015–2300) are denoted by the solid lines. In the temperature anomaly figures (bottom row), the time-mean historical data is denoted by a circle.

We measure the Hadley cell width as the boundary between the Hadley and Ferrel cells at 500 hPa, which is calculated as the latitude at which the 500 hPa meridional streamfunction crosses zero for the first time between the equator and the pole. Time series of the Hadley cell boundaries in the NH and SH are shown in Fig 9 , both as a function of year and of global temperature change. The SH tropical expansion is as large as 5.4 degrees in the CanESM5 model, and is relatively similar as a function of global warming across the three models ( Fig 9 ). The NH Hadley Cell extent, on the other hand, shows significant variability across models, and even within a single time series. In CanESM5, the NH Hadley cell extent vacillates between the years 2100 and 2200 ( Fig 9 ), before contracting to below pre-Industrial values. From Fig 8 , it is clear that the zero crossing of the streamfunction in CanESM5 is less well-determined in the hot period, in that there is a range of latitudes near 30 degrees that are close to zero streamfunction. This ambiguity in the Hadley Cell’s northern extent in CanESM5 may be a factor in the variability in the NH Hadley Cell boundary after 2100 in Fig 9 . Other studies have proposed possible mechanisms for an insignificant NH Hadley cell width change, including enhanced equatorial warming reversing seasonal Hadley Cell expansion [ 61 ], dynamic processes due to sea ice loss and ocean heat transport [ 62 ], or a large change in natural variability [ 63 ]. A more detailed analysis of the large-scale circulation in CanESM5 and other extreme warming scenarios is required to fully understand the mechanisms involved in NH Hadley Cell contraction.

The decrease in SH Hadley cell (and the SH Ferrel cell) strength occurs predominantly between the intermediate and hot period for all models, while the decrease in the NH Hadley cells occurs both before and after the intermediate period ( Fig 8 ).

The increased depth of the Hadley cells can be identified by the positive anomaly above the NH Hadley cell, and the negative anomaly above the Southern Hemisphere (SH) Hadley cell (these anomalies occur between approximately 300 and 100 hPa and between 30S and 30N; Fig 7A–7C ). The maximum Hadley circulation strength in the hot period is 55%, 57%, and 85% in the NH compared to the maximum in pre-Industrial, and 73%, 84% and 82% in the SH compared to pre-Industrial, in CanESM5, IPSL-CM6A-LR, and MRI-ESM2-0, respectively. However, the changes to the NH and SH Hadley cells occur at different times.

The streamfunction (10 10 kg/s) at 500hPa for CanESM5 ( A , red), IPSL-CM6A ( B , green) and MRI-ESM2-0 ( C , blue). The solid lines represent the hot period, the dashed lines represent the intermediate period, and the dotted lines represent the pre-Industrial period.

The color contours represent (A, B, C) the streamfunction anomaly (10 10 kg/s) and (D, E, F) the zonal wind anomaly (m/s) of CanESM5, IPSL-CM6A-LR, and MRI-ESM2-0. The black contours represent the pre-Industrial. The pre-Industrial streamfunction ranges from -9 to 9x10 10 kg/s with a contour interval of 3x10 10 kg/s. The pre-Industrial zonal wind ranges from 0 to 30 m/s with a contour interval of 6 m/s.

The zonally-averaged streamfunction in pre-Industrial (black contours) and the hot period anomaly from pre-Industrial (colors) is plotted in Fig 7A–7C . The zonally-averaged streamfunction at 500hPa for the pre-Industrial (dotted), the intermediate period (dashed), and the hot period (solid) is plotted in Fig 8 . The change in the zonally-averaged streamfunction has a complex structure, due to distinct physical processes occurring at different heights and latitudes. In each of the three models, there are streamfunction changes associated with the increase in tropopause height [ 59 ], the weakening of the Hadley cells [ 39 ], and the widening of the Hadley cells [ 60 ].

There are several areas with significant model-to-model differences in the pattern of precipitation. Equatorial Africa moistens in IPSL-CM6A-LR while it dries or exhibits more complex patterns in the other two. These precipitation changes over land explain some of the differences in the surface relative humidity changes. The Atlantic Inter-Tropical Convergence Zone (ITCZ) also exhibits a southward shift in the MRI-ESM2-0 model. While the magnitude of net precipitation changes are roughly equal before and after the intermediate period, changes near the ITCZ occur differently across the models. In CanESM5, north of the Pacific ITCZ has a positive net precipitation anomaly in the intermediate period, but a negative anomaly between the intermediate period and the hot period. Similar areas of drying north of the ITCZ in the later period can be seen in the other two models as well.

Drying in sub-Saharan Africa occurs preferentially between the intermediate and hot periods for all models ( Fig 6 ). In particular, the drying in tropical Africa between the intermediate and hot periods suggests that the more familiar CMIP wetting response in this region (e.g. between the pre-Industrial and intermediate periods in Fig 6 ) is not fundamental, corroborating the general lack of confidence in this wetting [ 57 , 58 ].

(A, B, C) The net precipitation (precipitation minus evaporation) anomaly for the hot period from pre-Industrial, (D, E, F) the intermediate period from pre-Industrial, and (G, H, I) the hot period anomaly from intermediate period anomaly. These maps are centered at 180 degrees longitudinally so as to better see the ITCZ. These maps were made using the Cartopy Robinson projection ( https://scitools.org.uk/cartopy/docs/v0.15/crs/projections.html ) which uses data from Natural Earth ( https://www.naturalearthdata.com/downloads/110m-physical-vectors/ ).

Precipitation minus evaporation (net precipitation) patterns are shown in Fig 6 . The three columns display net precipitation anomalies for the hot period from pre-Industrial, the intermediate period from pre-Industrial, and the hot period from the intermediate period (left to right). Precipitation changes are characterized by a wet-get-wetter pattern that is familiar in global warming scenarios, in which precipitation increases in ascent regions such as the deep tropics, and along the Kuroshio, Gulf Stream, and decreases in subtropical descent regions [ 26 ]. Nearly all locations in the polar regions of both hemispheres show increases in precipitation. Many subtropical areas experience drying, as do locations on the equatorward edge of the midlatitude storm track, which is in line with the multi-model-mean precipitation patterns taken during the first century of the ScenarioMIP runs [ 45 ]. Poleward shifts of circulation, which are analyzed in the next subsection, in part determine the subtropical drying.

Fig 5 displays the zonal mean of specific and relative humidity, comparing the hot period to pre-Industrial. Specific humidity in the Arctic increases by a much larger fraction than in the tropics. In latitudes greater than 60° N, average moisture increases by factors of 5.1, 4.4, and 2.9 for CanESM5, IPSL-CM6A-LR, and MRI-ESM2-0 respectively. The large increase in Arctic moisture content can be explained by the large Arctic warming observed in all three models, since the relative humidity in Figs 4 and 5 show a slight decrease.

The relative humidity changes above land masses are very different from model to model. The IPSL-CM6A-LR model shows substantial drying over nearly all continental regions by 2300. This drying occurs primarily before the intermediate period, especially in the Arctic and the Amazon. CanESM5 has even larger relative humidity changes over some land locations like the Amazon and the Congo Rainforest but has increasing humidity over Arctic and some tropical locations. The drying in the Amazon occurs primarily before the intermediate period, while the majority of drying in the Congo Rainforest occurs after 2100. The relative humidity change in the MRI-ESM2-0 model is qualitatively similar to that of CanESM5 with significantly less amplitude change. It is expected that global warming causes declines in relative humidity over land [ 12 , 54 – 56 ]. Specific humidity over land tends to increase proportionately to specific humidity over ocean; thus, the percent increase in specific humidity with climate change is controlled by ocean warming, which is smaller than land warming [ 54 , 55 ].

(A, B, C) Anomalies of surface relative humidity in the hot period relative pre-Industrial, (D, E, F) anomalies of surface specific humidity in the intermediate relative pre-Industrial, and (G, H, I) the relative humidity change from 2075–2100, calculated as the hot period relative humidity minus the intermediate period relative humidity. These maps were made using the Cartopy Robinson projection ( https://scitools.org.uk/cartopy/docs/v0.15/crs/projections.html ) which uses data from Natural Earth ( https://www.naturalearthdata.com/downloads/110m-physical-vectors/ ).

Fig 4 shows the relative humidity anomalies of the hot period and intermediate periods relative humidity anomalies from pre-Industrial for the hot period, the intermediate periods, and as well as the difference between the hot and intermediate period. The surface relative humidity change is characterized by small increases over most ocean areas in each model ( Fig 4 ). Exceptions to this are in the Arctic, where relative humidity decreases in each model.

In all three models, the spatial pattern of the percentage change in specific humidity that occurs prior to 2100 is similar in structure to that of surface temperature in regions outside of the Arctic. In CanESM5 and IPSL-CM6A-LR, the Southern Hemisphere sees more warming in the hot period than the intermediate period (blue coloration in Fig 2G–2I ), which contributes to larger increases in specific humidity in the Southern Hemisphere in the hot period. While the Arctic warms more in the intermediate period than the hot period (red coloration in Fig 2G–2I ), the specific humidity changes occur mostly in the hot period (in CanESM5) or equally in either period (IPSL-CM6A-LR); this is due to the nonlinearity of the Clausius-Clapeyron equation under the very large Arctic temperature changes.

The tropical specific humidity more than doubles in CanESM5 and IPSL-CM6A-LR and increases by a factor of 1.6 in MRI-ESM2-0. In CanESM5 and IPSL-CM6A-LR, globally the majority of specific humidity changes occur after 2100 ( Fig 3D–3F ), in part due to nonlinearity in the Clausius-Clapeyron relation. Exceptions to this include the Atlantic Ocean east of Greenland in CanESM5 and IPSL-CM6A-LR, parts of west and central Africa in CanESM5, and parts of the tropics and Northern Hemisphere (NH) land in MRI-ESM2-0, in which there are larger changes in the intermediate period.

(A, B, C) Anomalies of surface specific humidity relative to pre-Industrial, and (D, E, F) the percent change from the intermediate period to the hot period, calculated as the ratio of the surface specific humidity anomaly in the intermediate period and the anomaly in the hot period. These maps were made using the Cartopy Robinson projection ( https://scitools.org.uk/cartopy/docs/v0.15/crs/projections.html ) which uses data from Natural Earth ( https://www.naturalearthdata.com/downloads/110m-physical-vectors/ ).

Fig 3 shows the surface specific humidity anomalies for the hot period and the percentage of the total change that occurs by the intermediate period. The specific humidity is shown in units of Kelvin (which is converted by multiplying by the ratio of latent heat of vaporization of water and the specific heat at constant pressure of air). This convention has greater utility for use in interpreting moist static energy (MSE) which we examine later in the paper.

Outside of the Arctic, the greatest warming is seen over subtropical land areas, with temperature changes that are often 50% higher than the global average warming in each model. CanESM5 shows particularly large warming over Southern Africa, Australia, and equatorial South America, with temperature changes that are 70% larger than the global mean. Equatorial South America has a rather different temperature change across the three models. There is variability across these models regarding when land warming occurs. In CanESM5, the pronounced warming over Southern Africa and Australia occurs primarily between the intermediate and hot periods. In IPSL-CM6A-LR, the warming over Southern Africa, Australia, and equatorial South America occurs nearly equally before and after the intermediate period. Finally, in MRI-ESM2-0, warming over those three regions occurs primarily before the intermediate period.

The middle column of Fig 2 shows the temperature anomaly during the hot period, compared to the piControl climatology. The changes are greatest for each model in the Arctic, which is consistent with the polar amplification that has been widely studied [ 50 – 52 ]. In CanESM5 and IPSL-CM6A-LR the Arctic warming exceeds 30°C, and in MRI-ESM2-0 it gets as high as 25°C. In each model the average warming north of 60° N is 26.5°C, 23.7°C, and 16.2°C for CanESM5, IPSL-CM6A-LR, and MRI-ESM2-0 respectively. The Arctic warming of each model is between a factor of 1.5 and 1.75 greater than the global average, with MRI-ESM2-0 having the strongest Arctic amplification ratio of 1.74, then IPSL-CM6A-LR with 1.69, and CanESM5 has the lowest amplification ratio of 1.51. In each model, more than half of the Arctic warming occurs by 2100 ( Fig 2G–2I ). Namely, the impact of Arctic amplification is most prominent before 2100, likely due to substantial sea ice loss by 2100 [ 50 – 52 ]. Ono et al. (2022) showed that the rate of sea-ice loss with increasing temperature is non-monotonic in SSP5-8.5 because once the sea-ice extent is sufficiently small, the loss rate is lower than occurs in a cooler climate (even with greater amounts of warming) [ 53 ].

In the right-hand column of Fig 2 , blue coloration indicates regions where there is more warming between the intermediate period and the hot period than occurs between the pre-Industrial and intermediate periods (and vice-versa for red coloration). The plots are presented in descending order of climate sensitivity: CanESM5, IPSL-CM6A-LR, and MRI-ESM2-0. In each model there are regions where the annual mean temperature over tropical continents exceeds 40°C in the hot period, and in CanESM5, nearly the entire tropics exceeds 40°C in the annual mean. In CanESM5, tropical warming happens more between the intermediate period and the hot period, as indicated by the blue colors in Fig 2G . In IPSL-CM6A-LR, tropical heating occurs approximately equally before and after 2100 ( Fig 2H ). Whereas, in MRI-ESM2-0, tropical heating occurs primarily before the intermediate period, as shown by the red coloring in Fig 2I . In the hot period in CanESM5 and IPSL-CM6A-LR, the 0°C annual average contour only exists over Antarctic and Greenland ice sheets (the topography of which is prescribed). MRI-ESM2-0 also has a small region below freezing in the Himalayas.

(A, B, C) Near-surface temperature averaged over the hot period, (D, E, F) the hot period temperature anomalies relative to pre-Industrial, and (G, H, I) the percent change in surface temperature by the intermediate period, calculated as the ratio of the intermediate period temperature anomalies divided by the hot period temperature anomalies relative to pre-Industrial. These maps were made using the Cartopy Robinson projection. ( https://scitools.org.uk/cartopy/docs/v0.15/crs/projections.html ) which uses data from Natural Earth (https://www.naturalearthdata.com/downloads/110m-physical-vectors/).

The 2-meter surface air temperatures averaged over the hot period are shown in Fig 2 alongside their anomalies from pre-Industrial and the percentage of the total change that occurs during the simulation happens during the 21st century,

The equilibrium climate sensitivities to doubling CO 2 of CanESM5, IPSL-CM6A-LR, and MRI-ESM2-0, calculated using the method derived by Gregory et al. (2004) [ 48 ], and compiled for CMIP6 by Meehl et al. (2020) [ 49 ], are 5.6°C, 4.6°C, and 3.2°C, respectively. The transient climate responses for CanESM5, IPSL-CM6A-LR, and MRI-ESM2-0 are 2.8°C [ 42 ], 1.96°C [ 44 ], and 1.64°C [ 49 ], respectively. The ratio between the hot period anomalous temperature change and the equilibrium climate sensitivity are close to 3 for each model; the TCR has a less consistent relationship with the hot period warming, which is expected since the models have been at a stable forcing for more than a century by the end of the run ( Fig 1A ). The ratio of radiative forcing in the hot period (13.0 W/m 2 ) to the forcing of doubling CO 2 (3.7 W/m 2 ) is 3.5, which suggests that the hot period warming is slightly less than would be expected from a linear, equilibrium response to the radiative forcing. From Fig 1B , it is clear that all three models are still warming in 2300, despite the near-constant forcing over the previous century.

Global averaged 2-meter surface air temperature change is plotted for each model in Fig 1B . The globally-averaged temperature change in these three models in the intermediate and hot period, as well as the percentage of total warming that occurs by the intermediate period is shown in Table 1 and Fig 1B . Roughly half of the total warming occurs by the intermediate period, and models with more warming experience a smaller fractional change pre-2100.

3.2 Poleward energy transports

a. Latent energy, DSE, and MSE transport. The northward transport of latent energy is plotted for both pre-Industrial control and the end of SSP5-8.5ext in Fig 11A. Since the moisture content increases by such a large amount with warming, it is not surprising that the latent energy transport also amplifies quite a bit. In the SH extratropics, for instance, the maximum southward flux increases from less than 3 PW to over 4 PW in each model, with larger increases in models with more temperature change. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 11. Northward transport of energy from (A) latent, (B) moist static energy, and (C) dry static energy and (D) the Atlantic Ocean. Pre-Industrial transports are dotted and hot period transports are solid. The CanESM5 ocean heat transport data for the hot period is averaged between 2156 and 2180, due to data availability, so it is denoted with a dashed line to indicate the different averaging period. https://doi.org/10.1371/journal.pclm.0000343.g011 The CanESM5 model has a 3 PW increase in southward latent energy transport at 55° S (Fig 12A). In the NH extratropics, the maximum northward latent heat transport increases by 1–1.8 PW, depending on the model. At 45° N, poleward of its initial extratropical maximum, the northward latent energy flux in the CanESM5 model increases by 2 PW. In both the NH and SH extratropics, the latent heat transport increases more on the poleward side of the maximum than the equatorward side. The latitude of maximum poleward flux shifts poleward in both hemispheres in each model, and the fractional increase in flux reaches a factor of 3–4 in regions poleward of the flux maximum, as compared with 40–100% increase of the maximum value itself. These moisture transport changes are consistent with the large increases in precipitation at high latitudes in each model. PPT PowerPoint slide

PNG larger image

TIFF original image Download: Fig 12. Change in northward transport of energy (hot period minus pre-Industrial) from (A) latent (dashed) and DSE (solid), as well as from (B) ocean transport (dashed) and MSE (solid). https://doi.org/10.1371/journal.pclm.0000343.g012 The equatorward latent energy transport within the tropics in the pre-Industrial control is driven by the Hadley circulation. Despite large increases in humidity in the deep tropics, the equatorward transport in the NH tropics does not increase much, meaning the decrease in Hadley cell strength (Fig 7) nearly compensates. The SH tropical transport increases more, especially in the IPSL-CM6A-LR model, but the reduction in Hadley circulation strength also reduces the magnitude of the equatorward moisture flux change. The northward transport of MSE is plotted in Fig 11B, and its change in Fig 12B. The structure of the transport is similar in pre-Industrial and warmed climate, with poleward transport in both hemispheres. The poleward transport increases in all models, but with varying magnitudes. The increase in poleward flux is especially large in CanESM5 (maximum change of 1.7 PW), and small in MRI-ESM2-0 (maximum change 0.6 PW). IPSL-CM6A-LR shows the most asymmetric increase between NH and SH, with an average of twice the increase of poleward MSE flux in the SH as in the NH. The average fractional increase in poleward MSE transport in the SH extratropics is 35% in CanESM5, 25% in IPSL-CM6A, and 10% in MRI-ESM2-0. In each model, there is a part of the high northern latitudes where the poleward MSE transport decreases. This feature is most prominent in MRI-ESM2-0. Hwang et al (2011) argued that models with lesser flux increase into the Arctic were more likely to have stronger polar amplification, due to a diffusive response to high latitude warming causing a diminished energy flux into the Arctic [66]. We return to this Arctic feature later in the paper. In the NH, the CanESM5 model experiences a large equatorward shift of the maximum MSE flux, of approximately 10 degrees latitude, while the other two models keep the same latitude of maximum northward flux. The maximum in the SH in the CanESM5 model also shifts equatorward by a slightly smaller amount, while the MRI-ESM2-0 and IPSL-CM6A-LR models have a slight poleward shift. The equatorward shifts in maximum. MSE flux in the CanESM5 indicates especially large increases of MSE flux in the subtropics. We provide a more detailed explanation for the unique behavior of the CanESM5 model’s MSE flux maxima in the following subsections, based on the tropical cloud radiative effect changes. The northward transport of DSE is plotted in Fig 11C and is the difference between Fig 11A and 11B. The large decrease in poleward transport in the extratropics and increase in poleward transport in the tropics is clear in all models. These oppose the latent energy transport, and thus create a smaller change in MSE transport than in either DSE or latent energy flux individually. Opposition of the latent energy transport by DSE transport occurs at nearly every latitude in each model, and the changes are plotted together in Fig 12A to facilitate comparison. The most prominent exception is in the subtropics (especially visible in the CanESM5 model), where the poleward flux of both the DSE and latent energy increases in both hemispheres. The CanESM5 model has a particularly large decrease in southward DSE transport in the SH extratropics, where it decreases by over 50% in some places. In all models, the poleward DSE flux decreases particularly strongly in the regions between 35 and 60 degrees. The resulting DSE flux exhibits a poleward shift in the maximum extratropical flux, or the establishment of a new local maximum at high latitudes. We next consider the compensation percentage, defined as The compensation percentage C represents the effectiveness of the DSE transport in compensating for the increased latent energy transport, with 100% meaning there is no change in the MSE flux. Compensation percentages within the tropics are larger than 100%, reflecting the fact that the DSE transport increases faster than the latent energy transport. In the Northern Hemisphere between 40 and 60 degrees, compensation percentages are around 75% for each model. In the Southern Hemisphere, percentages are much lower: 65% in MRI-ESM2-0, and 50% in IPSL-CM6A-LR and CanESM5. How is the compensation percentage in the CanESM5 model in the SH so small, despite the large change in DSE flux? Examining the moisture flux change, one can see that the poleward latent energy transport increases by up to 3 PW, a number that is larger than the poleward transport of DSE in the pre-Industrial period. Thus 100% compensation is impossible within the SH midlatitudes, assuming the DSE transport remains poleward. The picture that emerges is of a SH storm track in which the latent energy transport increases substantially, working off of large humidity gradients (Figs 3 and 4) despite decreases in DSE transport, driven by some combination of weaker temperature gradients and weaker eddies. We examine these factors later in the paper.

b. Ocean heat transport. The ocean component of the anomalous northward energy transport is plotted in Fig 12B. This is calculated by integrating the anomalous surface flux after subtracting out the global mean, so includes the effect of both changes in ocean heat transport and differential storage. Since no model is in energy balance by 2300, the differential ocean heat storage is non-negligible. When heat storage is occurring more at the higher latitudes, e.g., in the Southern Ocean, there is an anomalous equatorward energy transport in this metric. In all three models, the ocean component opposes the MSE transport change. The change in ocean transport is especially close to being equal and opposite to the MSE transport change in the MRI-ESM2-0 model on a latitude-by-latitude basis, while in the CanESM5 model, the ocean transport averages around half of the poleward MSE transport change. The anomalous cross-equatorial transport is strongly negative in the MRI-ESM2-0 model, with a value of -0.5 PW, and is approximately half this value in other two models. An examination of the anomalous northward transport of energy in the Atlantic Ocean, plotted in the pre-Industrial and warmed climate in Fig 11D, shows that the cross-equatorial anomalies in Fig 12B are primarily due to reductions in strength of the AMOC. In Fig 11D, the x-axis only goes as far south as 30° S, due to the end of the ocean basin. CanESM5 only outputs data on ocean heat transport to the year 2180, so in that model we use an average of years 2156 to 2180 to represent the warmed period. In all models, the pre-Industrial ocean circulation transports a large amount of heat northward across the equator, with a similar value of 0.55–0.6 PW, while north of the equator there is substantial model spread in the poleward ocean heat flux. The MRI-ESM2-0 model has 50% larger flux in the pre-Industrial period than the other two models at 45 N. Warming causes a large decrease in the northward transport of energy in the Atlantic in all three models. AMOC damping is a result of large buoyancy fluxes that help to inhibit deep overturning circulation [67]. While buoyancy changes are primarily driven by surface fluxes [68], this buoyancy flux can also come in the form of melting sea ice, as melt water can cause abrupt changes in the salinity of surface waters [69, 70]. The MRI-ESM2-0 model has a slightly negative Atlantic cross-equatorial transport in the warmed climate, indicating a complete collapse of the AMOC. The AMOC collapse in MRI-ESM2-0 is consistent with the warming pattern in the intermediate period which shows decreased warming south of Greenland, also called the North Atlantic Warming Hole (NAWH; Fig 2I) [71, 72]. The AMOC collapse is likely also the primary cause of the southward ITCZ shift in the Atlantic identified in Fig 6F–6I as the reduction in oceanic heat transport must be met by an northward transport of MSE, which contributes to the convergence of winds occurring farther south [73, 74]. The slowdown in circulation and the accompanying ocean heat flux decrease can help explain the equatorward shift in the ITCZ shown in each model. Another consistent feature in the ocean heat flux change in Fig 12 is the decrease in strength of the ocean heat divergence within the equatorial region. The subtropical cells are wind-driven, and along with the decrease in strength of the surface winds throughout the tropics, visible both in the change in zonal winds in Fig 7 and in surface wind change plots as a function of latitude and longitude (not shown), the upwelling and overturning circulation must weaken. In turn, the MSE fluxes increase in each model to compensate for the decrease in ocean heat transport.

[END]
---
[1] Url: https://journals.plos.org/climate/article?id=10.1371/journal.pclm.0000343

Published and (C) by PLOS One
Content appears here under this condition or license: Creative Commons - Attribution BY 4.0.

via Magical.Fish Gopher News Feeds:
gopher://magical.fish/1/feeds/news/plosone/

You are viewing proxied material from magical.fish. The copyright of proxied material belongs to its original authors. Any comments or complaints in relation to proxied material should be directed to the original authors of the content concerned. Please see the disclaimer for more details.