The climate of a retrograde rotating Earth

To enhance understanding of Earth’s climate, numerical experiments are performed contrasting a retrograde and prograde rotating Earth using the Max Planck Institute Earth system model. The experiments show that the sense of rotation has relatively little impact on the globally and zonally averaged energy budgets but leads to large shifts in continental climates, patterns of precipitation, and regions of deep water formation. Changes in the zonal asymmetries of the continental climates are expected given ideas developed more than a hundred years ago. Unexpected was, however, the switch in the character of the European–African climate with that of the Americas, with a drying of the former and a greening of the latter. Also unexpected was a shift in the storm track activity from the oceans to the land in the Northern Hemisphere. The different patterns of storms and changes in the direction of the trades influence fresh water transport, which may underpin the change of the role of the North Atlantic and the Pacific in terms of deep water formation, overturning and northward oceanic heat transport. These changes greatly influence northern hemispheric climate and atmospheric heat transport by eddies in ways that appear energetically consistent with a southward shift of the zonally and annually averaged tropical rain bands. Differences between the zonally averaged energy budget and the rain band shifts leave the door open, however, for an important role for stationary eddies in determining the position of tropical rains. Changes in ocean biogeochemistry largely follow shifts in ocean circulation, but the emergence of a “super” oxygen minimum zone in the Indian Ocean is not expected. The upwelling of phosphate-enriched and nitratedepleted water provokes a dominance of cyanobacteria over bulk phytoplankton over vast areas – a phenomenon not observed in the prograde model. What would the climate of Earth look like if it would rotate in the reversed (retrograde) direction? Which of the characteristic climate patterns in the ocean, atmosphere, or land that are observed in a present-day climate are the result of the direction of Earth’s rotation? Is, for example, the structure of the oceanic meridional overturning circulation (MOC) a consequence of the interplay of basin location and rotation direction? In experiments with the Max Planck Institute Earth system model (MPI-ESM), we investigate the effects of a retrograde rotation in all aspects of the climate system. The expected consequences of a retrograde rotation are reversals of the zonal wind and ocean circulation patterns. These changes are associated with major shifts in the temperature and precipitation patterns. For example, the temperature gradient between Europe and eastern Siberia is reversed, and the Sahara greens, while large parts of the Americas become deserts. Interestingly, the Intertropical Convergence Zone (ITCZ) shifts southward and Published by Copernicus Publications on behalf of the European Geosciences Union. 1192 U. Mikolajewicz et al.: RETRO the modeled double ITCZ in the Pacific changes to a single ITCZ, a result of zonal asymmetries in the structure of the tropical circulation. One of the most prominent non-trivial effects of a retrograde rotation is a collapse of the Atlantic MOC, while a strong overturning cell emerges in the Pacific. This clearly shows that the position of the MOC is not controlled by the sizes of the basins or by mountain chains splitting the continents in unequal runoff basins but by the location of the basins relative to the dominant wind directions. As a consequence of the changes in the ocean circulation, a “super” oxygen minimum zone develops in the Indian Ocean leading to upwelling of phosphate-enriched and nitrate-depleted water. These conditions provoke a dominance of cyanobacteria over bulk phytoplankton over vast areas, a phenomenon not observed in the prograde model.


Introduction
When being introduced to Earth's climate, after learning about how quantities such as the latitude or elevation influence the climate of a region, school-children learn about the zonal asymmetries in patterns of weather. A common exercise in this context being to compare and contrast the climate of a city on Europe's Atlantic coast with one at a similar latitude and elevation on the Atlantic coast of North America. In the mid-latitudes, the climate of the west coast is usually more maritime and milder Earth Syst. Dynam. Discuss., https://doi.org /10.5194/esd-2018-31 Manuscript under review for journal Earth Syst. Dynam. Discussion started: 6 June 2018 c Author(s) 2018. CC BY 4.0 License. et al., 1989;Philander et al., 1996) -the reason primarily related to zonal asymmetries in how continental boundaries align with prevailing wind systems. Or is its preference for the Northern Hemisphere a consequence of zonally symmetric, but hemispheric asymmetries in the energy budget (Chiang et al., 2003;Kang et al., 2008)? Even if the hemispheric asymmetries in the heat budget are important, this raises the question as to their origin. Is it from hemispheric asymmetries in the distribution of land masses, or from asymmetries in how the ocean transport heats, which might be more related to zonal asymmetries in 5 the structure of ocean basins and their effect on deep-water formation.
The global ocean conveyer belt (Broecker, 1991), with meridional overturning in the North Atlantic fueling a deep western boundary current that winds through the southern ocean, around African and into the Indian Ocean and North Pacific where waters again upwell, is a profound example of a hemispheric asymmetry that might ultimately result from zonal asymmetries in the climate system. The drivers of the conveyor belt are still inadequately understood. Though it is clear that the salt advection 10 feedback suggested by Stommel (1961) stabelizes the current mode of overturning, it also creates the potential for arresting the Atlantic overturning, which thus implies multiple equilibria of the thermohaline circulation. But why is deep water not formed in the North Pacific? Two main mechanisms have been proposed, both invoking the role of zonal asymmetries. One is that the net freshwater gain in the Pacific (and the corresponding net freshwater loss in the Atlantic) leads to saltier sub-polar water in the Atlantic and rather fresh water in the sub-polar North Pacific, thus stabelizing the fresh Pacific relative to the salty Atlantic. 15 Another idea is that the limited northward extension of the Pacific limits the cooling in the northernmost parts of the Pacific.
In contrast the Atlantic extends all the way into the Arctic allowing the surface water to be cooled down to the freezing point.
For many of the above questions, changing the direction of rotation of the Earth thus presents an easy way to obtain a completely different climate. One which includes different net freshwater forcing regimes for the ocean, while keeping the present continental geometry. But also one in which the continents interact with patterns of winds in ways that may inform 20 understanding of hemispheric asymmetries in atmospheric circulation systems, such as the position of the tropical rain-bands, or structure of the storm tracks. Two studies (Smith et al., EGU 2008;Kamphuis et al., 2011) have explored the consequences of a retrograde rotation of the Earth with ocean circulation questions in mind. But even for these more limited in scope studies the authors come to quite different findings. Smith et al. (EGU 2008) found an inverse conveyor belt circulation with strong deep water formation in the North Pacific in a simulation with FAMOUS. Using another model Kamphuis et al. (2011), 25 meanwhile, found that deep water formation in the North Atlantic weakened in their simulations, and intensified intermediate water formation became evident in the North Pacific. These simulations did not, however, show evidence of a complete reversal in the role of sub-polar North Atlantic and Pacific for deep-water formation, leading to the conclusion that a reversed freshwater flux is not sufficient to obtain the reversal, but that the continental geometry is crucial for the pattern of the overturning circulation.

30
The two aforementioned studies, as noted, focus on the effect of the direction of the Earth's rotation on dynamics with a focus on the ocean circulation. We are aware of no study that has simulated the effect of changing the sense of Earth's rotation using a full Earth system model run long-enough to reach a stationary state, and wherein Earth's oceans and biosphere (terrestrial land surface) were allowed, for the most part, to freely adapt to the changing conditions, and thereby define a new climate. Two hundred years after Humboldt first introduced the idea of showing climatic data on a map, and a century after 35

Model and experiments
All simulations were performed with the MPI-ESM (version 1.2) climate model, as developed for use in CMIP6. The model contains the atmospheric general circulation model ECHAM6.3.02, (Mauritsen et al., 2017, in preparation). This version contains some bug fixes and a different tuning, but is not structurally very different from the version used in CMIP5 . The land component JSBACH.3. 10 (Reick et al., 2013) includes a dynamic vegetation model.. Marine compo- 10 nents are the ocean general circulation model MPIOM-1.6.2p3 (Jungclaus et al., 2013) and the marine biogeochemistry model HAMOCC (Ilyina et al., 2013;Paulsen et al., 2017). The coarse resolution (CR) configuration of MPI-ESM is used. It consists of ECHAM6 run with a T31 spectral truncation and with 31 vertic hybrid (sigma-pressure coordinate) levels. In physical space, the spectral transform grid on which parameterized processes are solved corresponds to 96 by 48 horizontal grid cells on a Gaussian grid. MPIOM has a curvilinear grid, with the poles located on Greenland and Antarctica. It has 120 by 101 15 grid cells in the horizontal and 40 levels in the vertical. A dynamic-thermodynamic sea ice model is included in MPIOM. The atmosphere and ocean are coupled once a day.
The forcing applied is the same as for the CMIP5 preindustrial control run  and corresponds to 1850 climate conditions (fixed greenhouse gas conditions, insolation, aerosols, ozone). All simulations were started from an equilibrated climate state, resulting from a long spin-up with subsequent model tuning. The control run CNTRL is simply concentrations of phosphate and oxygen in such a way that, to first order, biological impacts are eliminated and thus can be used as a (quasi) conservative property, which in the prograde world is used to track the contributions of North Atlantic deep water (NADW) and Antarctic bottom water (AABW) to the ventilation of the deep Pacific and Indian Ocean (Rae and Broecker, 2018). surface module JSBACH. The anthropogenic land use was prescribed according to 1850 conditions. These shortcomings in the model/experimental set-up will be dealt with in future experiments.
The experimental setup was oriented on the piControl conditions, as this run served as spinup for CNTRL. Whereas the relative complete (with respect to the components) ESM was run long enough to reach equilibrium for the climate, some biogeochemical components revealed somewhat longer time scales as discussed above. Hence, for the analyses, the last 1000 5 years of each experiment were used. An exception is for the analysis of the atmospheric short term variability, which requires 2-hourly model output, and is based on the last 100 years of the experiments for which high-temporal resolution output was retained.
3 The atmosphere, its energy budget and the surface climate At a first glance, changing the sense of planetary rotation incurs changes that Humboldt, or even Hadley (1735), might well have 10 anticipated. The patterns of surface winds and the polarity of the temperature distribution, or the inclination of the isotherms, is reversed (Fig. 2). The change in the winds are self evident, as the trades blow from the West in RETRO and the mid-altitudes 5 Earth Syst. Dynam. Discuss., https://doi.org /10.5194/esd-2018-31 Manuscript under review for journal Earth Syst. Dynam. Discussion started: 6 June 2018 c Author(s) 2018. CC BY 4.0 License. have surface easterlies. Changes in the inclination of the isotherms is perhaps most evident in the distribution of the sea ice extent, this is shown by the feint blue lines in panels a and b of Fig. 2, as well as the sea-ice distributions themselves presented in Fig. 3. In RETRO, winter sea-ice extends southward eveloping great Britain and reaching as far as the Bay of Biscay in the Northeast Atlantic, whereas the Canadian province of Newfoundland & Labrador is free of sea-ice in winter. Similar changes are evident in both the Aleutian Basin of the North Pacific, and in the Pacific Basin of the Southern Ocean. These changes 5 are also evident in the tropical sea-surface temperatures, with the region of warmest temperatures broadening to the east in the Retrograde simulation as compared to a westward broadening in CNTRL (Fig. 2).
Over the continents the shifting isotherms mean that the Eastern continental margins warm and the western lands cool in RETRO as compared to CNTRL. In the mid-latitudes these changes are skewed (reflecting the changing inclination of Humboldt's isotherms) so that the cooling in the western lands is poleward amplified, and the warming is strengthened toward 10 the equator. Strong warming for instance is evident in the Southeast of Brazil (Rio de Janeiro), over the South Eastern states (Atlanta) of the United States of America, and over Southeastern China (Guangzhou and the Pearl river delta region). The strongest cooling is over the ocean, in association with winter sea-ice, particularly over the North and Baltic Seas, although west Africa, which is very warm in CNTRL also cools substantially as does British Columbia in present day Canada. The magnitude of these changes have the effect that -from a temperature perspective -the African-European landmasses cool, and 15 the American's warm (Fig. 2c). Over the ocean the cold-upwelling waters evident in the present day East Equatorial Pacific largely vanish, rather than shift, in RETRO. Fig. 2 shows only a hint of a cold tongue stretching eastward from East Africa in the Equatorial Indian Ocean.
To a large extent, the temperature changes also reflect changes in patterns of precipitation. Circulation changes underlying changes in the inclination of the isotherms are evident in changes in the isohyets (lines of constant precipitation; Fig. 2). The 20 simulated warming of the southeastern regions of the continents coincides with marked drying, as precipitation is displaced to areas of present-day deserts in simulations of the Retrograde Earth (Fig. 2d,e). As was the case for temperature, some extracontinental scale chages are evident as again the roles of the Americas and the Afro-European land masses are exchanged. The latter becomes substantially wetter as a whole, particularly over Africa and the Mediterranean, and the former substantially drier (Fig. 2f). Precipitation in the tropics shift from the western oceans where it is stronger north of the Equator, particularly 25 in the Pacific, to the eastern oceans and then south of the Equator, as in the Atlantic. This results in the zonally averaged precipitation becoming slightly stronger south of the Equator, a point that we rejoin further below.
RETRO also exhibits substantial changes in its monsoons and deserts, something that might have been anticipated based on earlier work by Rodwell and Hoskins (1996). Whereas in CNTRL tropical precipitation, and the velocity potential (at 200 hPa) to which it is related, is strongly focused around the Western Pacific Warm Pool and Asian-Australian Monsoon Complex,  between oceanic (warmpool) and terrestrial (monsoon) precipitation in RETRO results in a more tripolar, rather than monopolar, tropical precipitation pattern, with distinct centers of precipitation over the Middle East, extending into North Africa, over the Southeast Atlantic, and into the Central Pacific. These changes are reflected in the 200 hPa velocity (not shown) potential, which becomes seasonally more varying and less mono-polar in RETRO as compared to CNTRL.
To assess changing precipitation patterns we have also tracked individual extra-tropical storms in both RETRO and CNTRL. 5 This analysis is included in Appendix A. As expected, the storms track westward, rather than eastward in the mid-latitudes.
This analysis shows little evidence of a change in their characteristics (such as number, lifetime, growth rate or intensity), but the tracks of storms change substantially regions of the western-boundary currents to be centered over land, with particularly strong storm activity extending from the Caspian sea and Arabian Sea through the northern Mediterranean (see supplemental videos). This has large implications for the climate of North Africa and the middle east but also acts to freshen the Atlantic at shifts equatorward with a more northward tracking of the storms in RETRO as compared to CNTRL (Fig. 2d,e). This may arise from different topographic interactions, or a shift in baroclinicity evident in an equatorward shift in the sub-tropical jet. Despite large regional shifts in circulation patterns, the globally averaged energy budget of RETRO hardly differs from that of CNTRL, and those differences that do emerge tend to be smaller than our ability to match the energy budget as derived from observations (Stevens and Schwartz, 2012;Stephens et al., 2012). Fig. 4 is adapted from Stevens and Schwartz (2012) and 5 shows that RETRO is slightly cooler (by about 0.14 • C, see Table 1, with less upwelling terrestrial radiation from the surface), cloudier and wetter. The increase in cloudiness is evident in a slight increase in planetary albedo, with an additional 1 W m −2 of reflected solar radiation. Surface turbulent moisture fluxes are larger and sensible heat fluxes are smaller in RETRO. These changes are consistent with more terrestrial radiation to the surface, indicative of an atmosphere that cools and hence rains more.
The surface cooling in RETRO is manifested entirely in the northern hemisphere, as the southern hemisphere actually warms 10 by 0.28 • C. In RETRO the difference between the average temperature of the northern and southern hemisphere is −0.40 • C, as compared to 0.57 • C in CNTRL. Concomitantly, almost all of the change in the downward longwave (5.7 W m −2 ) is in the northern hemisphere, and almost all of this is over land.
Changes in the mean energy budget could be expected to effect the zonal distribution of precipitation. Building on the ideas developed by Kang et al. (2008Kang et al. ( , 2009) and Frierson et al. (2013), Bischoff and Schneider (2014) argue that an increase of   Fig. 5a). The shift is also predicted by the change in the energy budget equator (Fig. 5b, and inset), although not by the magnitude of the shift, nor does its position correlate well with the actual position of the precipitation maximum. These discrepancies perhaps being related to changes in the zonally asymmetric circulation as earlier argued by Wallace et al. (1989) and Philander et al. (1996). To the extent the changes in precipitation are consequences of the changed hemispheric energy budget, they point to a possible role for ocean circulation, and its disproportionate impact on the northern hemispheric ice sheets 5 and temperature gradients, particularly over the North Atlantic. Reversing the sense of the planetary rotation does not, however, affect the distribution of landmasses, so that differences between RETRO and CNTRL suggest that hemispheric asymmetries in the distributions of continents (i.e., the fact that Antarctica is in the southern hemisphere) is at least not sufficient to explain why the ITCZ is mostly north of the equator.
Many of the points discussed above are amplified by an inspection of the latitude-height description of the zonally aver-10 aged circulation. Fig. 6a shows the differences between RETRO and CNTRL of annually and zonally averaged temperatures. Fig. 6b presents the corresponding annually and zonally averaged zonal winds. The surface temperature anomaly pattern shows the tropical warming in RETRO and the strong northern hemispheric cooling. The subtropical jets in RETRO are shifted northwards, i.e., equatorwards in the SH and polewards in the NH, and slightly increased in magnitude. This is consistent (through the thermal wind balance) with changes in tropospheric temperatures whereby the high-latitude cooling in RETRO is 15 much more pronounced in the northern hemisphere. This pattern of tropical warming and high-latitude cooling implies greater baroclinicity at mid-latitudes, but is accompanied by relatively little change in storm activity (Appendix A), suggesting that enhanced northern hemispheric energy fluxes are mostly attributable to the changing enthalpy gradients.
As typical for climate model simulations of global warming, the tropical warming is stronger in the free troposphere than at the surface and reaches about 1.7 K around 300 hPa. Also typical for a globally warmer tropical troposphere is the increase 20 of the cold-point tropical tropopause, i.e., the coldest point in the tropical tropopause region, as one might also expect if the tropopause temperature is radiatively controlled following the ideas of Zelinka and Hartmann (2010), which in RETRO is about 0.6 K warmer than in CNTRL. A warmer cold-point tropopause and the resulting increased water vapour entry into the stratosphere is the likely cause of an about 13 % larger specific humidity in the lower to middle stratosphere in RETRO (not shown). More stratospheric water vapour is a plausible reason for the, in general, lower stratospheric temperatures. Changing the planetary rotation should, of course, to a first order cause a reversal of zonal winds. Fig. 6b shows the sum of annually and zonally averaged zonal winds of RETRO and CNTRL, which indicates 2nd-order effects, i.e., deviations from a simple change in sign. Changes in 5 the stratospheric circulation should be interpreted with caution due to the model top in the middle stratosphere at 10 hPa.
However, the strongest stratospheric cooling, which occurs in the polar northern hemisphere, is dominated by a cooling in boreal winter, which further contributes to northern hemispheric baroclinicity, and is related to a stronger polar vortex in RETRO. The average eddy heat flux entering the stratosphere in boreal winter between 40°N to 80°N is weaker in RETRO by about 15 % (not shown), thereby helping to explain the temperature changes required to balance this stronger polar vortex. The

Land surface and biosphere
Köppen, in his book Die Klimate der Erde (1923), outlined his ideas regarding how continents impart zonal asymmetries in climate using his classification system applied to an idealized continent. For the purposes of the present discussion Köppen's idealized continent has been redrafted, and is presented in Fig. 7. The expectation inherent in this idealization is that a retrograde rotating Earth would experience mirror symmetry about the North-South axis in its climate zones. 5 As might have been expected given the discussion of the previous secton, predictions based on Köppen's idealized continent, both for the distribution of climate zones in the present day climate, and for the expected mirror symmetry for a retrograde rotating Earth are well supported by CNTRL and RETRO. This is shown, using the same indication of major climate zones, in Fig. 8. CNTRL reproduces both the North-South asymmetry associated with more northern hemispheric land masses as reflected by the emergence of cold winter climates (Dw and Df) in the northern hemisphere, and the east-west asymmetry of 10 a vareity of features. For instance the shift from subtropical deserts (BW) in the continental southwest toward moist temperate climates in the continental southwest at subtropical latitudes is well evident in the contrast between West Africa and Southeast Asia. To a first approximation in a retrograde rotating Earth these features do appear with mirror symmetry (Fig. 8b).
There are however some differences that would not have been predicted from just mirroring the idealized continent. Most prominent is the shift in deserts from the Eurasia-African continental mass to the Americas, with a greater center of mass over 15 the subtropical southern hemispheric continents. This change is consistent with the inferences of the last section, whereby in RETRO many of the climate features associated with present day Europe/Africa and North/South America are exchanged.  Figure 7. Classification of climate zones for an idealized continent, the Klimarübe, following Köppen (1923). First letters in the legend are main climate types: A: equatorial, B: arid, C: warm temperate, D: snow, E: polar. Second letters are precipitation regimes: W: desert, S: steppe, f: fully humid, s: summer dry, w: winter dry, T: tundra, F: ice cap.
The complete replacement of the wide desert belt from West Africa to the Middle East by forests and humid grasslands is more quantitatively measured by changes to the leaf-area index (LAI, Figure 9). Changes in LAI also illustrate the degree to which the retreat of desert climates in Africa and Eurasia is accompanied by an extensive formation of dry climates in South and North America. Southern Brazil and Argentina become the Earth's biggest deserts and the southern States of the United States of America see a dramatic climate shift from a fully humid climate towards a complete aridification. In general, many 5 dry regions are simulated in RETRO, but extreme deserts -like the present-day Sahara -are less widespread. Changes that are, as discussed in section 3, consistent with circulation and precipitation changes over these same regions.
The global area covered by permanent deserts is reduced by about 25 % from 42 × 10 6 km 2 to 31 × 10 6 km 2 (see Table   2). Woody and herbaceous vegetation fill the new vegetated areas in about equal measure. The greening is concentrated over northern hemispheric land masses (desert areas shrink by nearly 40%) and mostly attributed to the aforementioned vanishing of 10 the wide desert belt from West Africa to the Arabian peninsula. Over southern hemispheric land masses, deserts and grasslands spread sightly, whereas the tree-covered area is reduced by about 20 %. Tropical vegetation remains largely unaffected in Africa and Asia. In South America, the Amazon rain forest shrinks substantially, though.
The greener climate in RETRO affects the global carbon storage on land, which increases by 86 PgC compared to the CNTRL simulation. After taking into account the implied changes in ocean carbon that corresponds to a decrease of around    occurs in the North Atlantic (Fig. 12). The occurrence of convection has a marked multidecadal time scale with a peak around 80 years. In the long-term mean, no clear North Atlantic deep water cell emerges in RETRO. Instead, the northeast Pacific exhibits strong formation of deep water (down to 3000 m). This leads to a strong Pacific MOC of 23 Sv, which is associated with the formation of North Pacific deep water (Fig. 11). This indicates, that the Atlantic and Pacific change their roles in CNTRL and RETRO. 25 Whereas the global poleward oceanic heat transport on the northern hemisphere is almost unchanged (1.56 PW in CNTRL vs. 1.52 PW in RETRO at 30°N , Fig. 13 a), the distribution between the basins changes strongly. In CNTRL, more than half of the heat is transported in the Atlantic. In RETRO, the Pacific transports almost 80% of the heat. The transport in the Pacific is larger, as the gyre-transport in the Pacific is stronger than in the Atlantic, which can be explained by the larger spatial extent of the Basin. Thus, also in terms of the heat transport the Atlantic and Pacific have changed their roles. Due to The changes in ocean heat transport are also reflected in SST and sea ice changes. In general (see Fig. 3), the North Atlantic is colder in RETRO, while the North Pacific is warmer. The effect of the reduced ocean heat transport in the North Atlantic 5 is amplified by the reduced wintertime mixed layer depth, which reduces the effective heat capacity of the surface ocean and, thus, leads to colder winter temperatures.
In CNTRL the net moisture fluxes are almost equal in the Atlantic and Pacific (Fig. 13b). The net moisture loss of the Pacific north of 30°S is 0.30 Sv, in the Atlantic only 0.07 Sv. In CNTRL the role of the two oceans is reversed. In RETRO, the Atlantic gains moisture from the Indian ocean by atmospheric moisture transport across the Middle East, which is strongest in spring 10 and summer (Appendix B). The transport across Central America is very small, but directed towards the Atlantic, whereas in CNTRL a strong transport is directed towards the Pacific. In both simulations the tropical Indian Ocean has the highest net loss of freshwater compared to the other oceans. Consequently, the North Pacific surface salinity in RETRO is higher than in CNTRL; in the North Atlantic surface salinity is reduced (see supplemental video).

Ocean biogeochemistry
Ocean circulation is the key driver of spatial patterns of marine biogeochemical tracers. This implies large scale patterns of these tracers follow the reversal of the ocean circulation (described in section 5) and biogeochemical water mass trackers, such as PO * 4 mirror the transient behavior of the MOC reversal though with longer equilibration time scales (Fig 1b). Planetaryscale features such as zonal and global means are largely unaffected by the direction of Earth's rotation (Tab. 3). The changes 5 in the Coriolis effect and wind patterns lead to a shift of eastern boundary upwelling systems to the western sides of ocean basins in RETRO (Fig. 14). In these upwelling systems, in general, primary production and, thus, carbon storage in particular organic matter (POM) is driven by a continuous supply of nutrients from greater depth into the sun-lit surface layers (euphotic zone). Gravitational settling and subsequent remineralization of POM, a process being referred to as biological pump, lead to pronounced vertical gradients in all biogeochemical tracers. Exported organic material is remineralized under the consumption 10 of oxygen. For this reason, hypo-or suboxid conditions are often found at mid-depths in regions with high biological production such as upwelling systems. The global volume of these oxygen minimum zones (OMZs) is nearly identical in RETRO and CNTRL (Tab. 3, see also supplemental video). Despite that, in CNTRL three major OMZs are found on the eastern side of each ocean basin while in RETRO there is one sizable OMZ in the Indian Ocean. Strong upwelling, which is also reflected in strong heat uptake, fuels biological production rates at levels not found in CNTRL. In combination with the circulation and 15 the basin geometry of the Indian Ocean this leads to very low ventilation and nutrient accumulation in the northern part of the basin (Fig. 15) and results in the development of this extended OMZ. One prominent characteristic of OMZs is that in the low oxygen environment denitrifying bacteria are able to access food energy by degradation of organic matter using nitrate (NO 3 ).
Denitrification is limited to very low oxygen conditions (in the model O 2 ≤ 0.5 mmol m −3 ) and is the only remineralization process that selectively removes bioavailable nitrogen. All other degradation processes, i.e. aerobic remineralization and sulfate 20 reduction, convert organic material into phosphate, iron, and nitrate. In RETRO global denitrification is about 50 % higher than in CNTRL and takes place predominately in the northern Indian Ocean (Tab. 3, see also supplemental video). As a consequence, upwelling of nitrate depleted and phosphate rich water (N * = NO 3 − 16 · PO 4 < 0, Fig. 16) results in a shift of the pythoplankton species composition (Fig. 15) with a dominance of cyanobacteria in RETRO. Most pythoplankton species need both nitrate and phosphate for their growth. Only cyanobacteria are able to grow on inorganic nitrogen (N 2 ), as long 25 as sufficient phosphate and iron are available (e. g. Sohm et al., 2011). In the prograde world, regions where surface water is nitrate depleted and phosphate rich occur only very localized. Thus, cyanobacteria are nearly everywhere outcompeted by other phytoplankton species (bulk phytoplankton). In contrast, in RETRO they become the dominant primary producer in the northern Indian Ocean. The change of the Earth's rotation direction and the subsequent development of an extended oxygen minimum zone in RETRO provoke plankton species compositions over large areas which have not been observed in 30 the prograde world.
It was hypothesized that a warming climate and the consequential deoxygenation of the ocean (e.g. Breitbarth et al., 2007;Hutchins and Fu, 2017) trigger such an ecosystem shift with nitrogen-fixing cynaobacteria as a potential winner. Thus, the response nicely supports the possibility of these extreme changes in the ecosystem. It also demonstrates the model's ability to 21 Earth Syst. Dynam. Discuss., https://doi.org /10.5194/esd-2018-31 Manuscript under review for journal Earth Syst. Dynam. Discussion started: 6 June 2018 c Author(s) 2018. CC BY 4.0 License. Figure 14. Net primary production (mol m −2 yr −1 ) a. RETRO b. CNTRL Table 3. Global mean net primary production rate, biogenic material export flux at 100 m, volume of OMZ, global mean denitrification and ratio of denitrification to organic matter export flux a on percentage basis Rden Net primary production (   To understand this difference, we look at the diurnal cycle of precipitation (not shown) over the impacted land masses. In agreement with the sun movement, precipitation starts falling on the eastern side of Africa in RETRO-S and moves inland.
The opposite is true in RETRO, so that precipitation starts falling with a 2 h to 4 h time lag on the eastern side of Africa (note that the output frequency is 2 h). Despite the later precipitation start, RETRO produces more precipitation. This may be related to the fact that most precipitation falls on the eastern side of Africa in the mean (see Fig. 1

Climate sensitivity
If climate sensitivity, the longterm response to increased atmospheric CO 2 , did not depend on the details of the circulation we would expect it not to depend on the direction of the planetary rotation. To investigate this we conduct experiments with abrupt quadruplings of CO 2 (Figure 18). Here the equilibrium response is estimated using linear regression on annual means of the relationship between radiation balance and global warming. The slope in such a diagram is thought of as a feedback parameter, 5 λ, which must be negative to yield a stable climate. We find, indeed that the climate sensitivity is not very different among CNTRL and RETRO, roughly 10 percent larger in the latter (4.00 vs. 4.45 K per doubling of CO 2 ). This is substantially less than the difference to the slightly higher resolution MPI-ESM1.2-LR model (also shown, 3.19 K) and much less than the spread among CMIP5 climate models (2.0-4.6 K) and scientific uncertainty as expressed by the likely range from the IPCC AR5 (1. 5-4.5 K). It should be noted that the here reported climate sensitivities are regressed over the years 21-1000, and using instead 10 the more standard regression over years 1-150 we obtain 3.46, 3.92 and 2.87 K for CNTRL, RETRO and MPI-ESM1.2-LR.
The small increase in the longterm feedback parameter for RETRO of about δλ ≈ +0.09 W m −2 K −1 stems primarily from cloud shortwave effects. Indeed, it was tempting to think that the more widespread sea ice cover in the North Atlantic in RETRO would cause a stronger surface albedo feedback, but the clear-sky shortwave component does not differ between the simulations. Likewise, the longwave feedbacks differ by merely 0.01 W m −2 K −1 . Thus, the remaining 0.08 W m −2 K −1 stems 15 from cloud shortwave feedbacks.
The difference in feedback parameter between the earlier parts of the simulation (years 1-20) and later parts (years 21-1000) is thought to be due to tropical cloud feedbacks Andrews et al., 2015;Zhou et al., 2016). One idea is that surface temperatures in the ocean upwelling regions in the tropical East Pacific warm slower than the surface waters in the Warm Pool region, causing more low-level cloudiness in the East as a result of increasing low-level stratification. The 20 striking similarity between CNTRL and RETRO is suggestive that the details of the tropical atmosphere and ocean circulations also do not matter for determining the strength of this mechanism.
Finally, we noted a striking difference between CNTRL and RETRO in the longterm equilibration for years 1001-2500. As found for an earlier version of the model by Li et al. (2013), CNTRL exhibits a third time-scale where λ becomes more negative causing the system to equilibrate faster, whereas, if anything, RETRO does the opposite. The third timescale found in CNTRL 25 and in the ECHAM5/MPIOM model used by Li et al. (2013) is yet to be found in other models partly because models are only seldom run long enough. Previously, we thought the behavior originated in the coarse resolutions used in these runs, as the MPI-ESM1.2-LR exhibits a faster equilibration, but the contrasting very slow equilibration of the coarse resolution RETRO simulation could indicate that the properties of the ocean circulation are indeed important for the equilibration process. However not all of the changes were expected, nor necessarily trivial. One of the most unexpected differences is in patterns of precipitation. In the extra tropics tracking of the storms shows greatest track densities over land, rather than over the ocean, even if the intensity and number of storms do not show detectable changes. Likewise there is a northward shift of the extratropical storms, and the main regions of baroclinicity, as cyclogenisis in the southern hemisphere shifts equatorward, and over 30 North America a major storm track develops to the North of the great lakes. The biggest change however is in the Mediterranean regions, as storms tracking from south Asia and the Indian ocean bring rains to the region. Strong centers of storms increase freshwater transport to the Atlantic. In the tropics the rainbands shift from a double ITCZ centered in and around the warm-pool Asia-Australia monsoon complex in CNTRL, to a more tri-polar structure with centers of action in the eastern tropical Atlantic, the central Pacific and a monsoon region centered in the Middle East. A southward shift in the position of the annually and zonally averaged position of the rain bands is consistent with shifts in the zero-line of the atmospheric energy transport (the intertropical convergence zone), although the shift in this line is as would be anticipated by changes in the global climate, i.e., from a warming of the tropics or increased northern hemisphere baroclinicity (Kang et al., 2008;Bischoff and Schneider, 2014).
The origin of these changes is not clear, and the magnitude and position of the rainband shifts are in less good quantitative 5 agreement with changes in the energy-flux near the equator. The latter suggests that stationary eddy transport and air-sea interaction (Wallace et al., 1989;Philander et al., 1996) is important for determining the position of the zonally and annually averaged rainbands.
Another unexpected change, albeit one inline with the changes in patterns of precipitation, is a shift in the desert climates from the Northern to the Southern Hemisphere. The simulations predict a complete replacement of the wide desert belt from 10 West Africa to the Middle East in CNTRL by more moderate, humid climates in RETRO. This retreat of desert climates in Africa and Eurasia is accompanied by an extensive formation of dry climates in South and North America, but not so dry as the African dry climates they replace. So that in RETRO Earth becomes greener with substantially more biomass and by implication a lesser airborne fraction of carbon-dioxide. Changes in biomass, and precipitation patterns are to the first order consistent with the America's adopting the climate pattern of persent day Euro-Africa, and vice versa.

15
The exchange of Euro-African, with American climate, and the shifts in fresh water transport by the atmospheric circulation is accompanied by large-changes in the ocean overturning circulations. An important and outstanding question of physical oceanography is the North Atlantic is so disproportionate in its production of deep water, with very little or no deep water formed in the North Pacific. Kamphuis et al. (2011) suggested that the more southward continental boundary of the Pacific limits deep water formation as compared to the Atlantic which extends well into the Arctic. Other studies have suggested 20 that the freshening of the Pacific relative to the Atlantic stabilizes the overturning pattern. A prominent difference between RETRO and CNTRL is the collapse of the Atlantic MOC in RETRO, with only sporadic formation of deep intermediate water in the North Atlantic. At the same time a strong meridional overturning cell emerges in the Pacific. The Pacific MOC in RETRO is similar in structure but slightly stronger than the Atlantic MOC in CNTRL. The breakdown of the Atlantic MOC and the associated decrease in meridional heat transport leads to a cooling of the North Atlantic associated with a southward 25 extension of sea ice and a significant cooling over Europe. The accompanying southward shift of the temperature maximum over the tropical Atlantic contributes to the southward shift of the Atlantic ITCZ, through enhanced baroclinicity (and hence heat transport) as argued above. In contrast, the Pacific MOC in RETRO, characterized by an enhanced northward transport of heat along the West coast of North America, results in a significant warming of the North Pacific. The switch of the deep water formation into the North Pacific shows that changes in atmospheric circulation and moisture transport are sufficient to create a 30 completely reversed conveyor belt circulation. The topographic setting (basin and continent distribution) does not exclusively allow deep water formation in the North Atlantic and Southern ocean and not in the North Pacific as has been suggested by Kamphuis et al. (2011).
In the tropics the Indian ocean takes over the role of the eastern tropical Pacific with a strong net heat uptake due to strong upwelling. As expected, the pattern of biogeochemical tracers are tightly bound to changes in the ocean circulation. Zonal and global means of biogeochemical features are very similar in CNTRL and RETRO. However, an unforeseen shift in the dominance of cyanobacteria over bulk phytoplankton is found in the northern part of the Indian Ocean. Here, the interplay between strong upwelling and basin geometry leads to the development of an extended oxygen minimum zone with increased denitrification and nutrient trapping. Upwelling of phosphate enriched and nitrate depleted water leads to favourable growth conditions for cyanobacteria. 5 Sensitivity studies suggest that differences in the circulation in the RETRO versus CNTRL simulations do not fundamentally affect Earth's climate sensitivity. Changes in the ocean in RETRO do however appear to affect the time-scale of equilibration.
Overall changes in the interaction between the atmosphere and the ocean that accompany a reversal of the sense of planetary rotation help demonstrate the important role of the ocean in climate and provide valuable out of sample tests of the robustness of climate models and the physics governing their response to perturbations. Repeating the RETRO/CNTRL pair of simulations 10 at higher resolutions and with other models could thereby provide a fascinating laboratory for testing understanding of climate models on the one hand, and by inference, of Earth's climate.

Appendix A: Storm track changes
Given the change in the planet rotation and the inversion of prevailing winds in the upper atmosphere we expect the storm tracks to be modified. However we do not know, whether the change would be just zonally symmetric or instead involve a modification of storms intensity and lifetime. In order to asses that we perform a tracking using the Mean-Sea-Level Pressure 20 (MSLP) data every 2 hours. Storms are defined as local minimum of MSLP and are tracked during their lifetime using the methodology described in Hoskins and Hodges (2002).
Since we expect the cyclone activity to be concentrated in winter we perform a DJF tracking for the Northern Hemisphere and a JJA tracking for the Southern Hemisphere. The tracking density observed for DJF in the northern hemisphere is presented in Fig. A1. As expected, cyclones are now forming over the eastern Atlantic and Pacific and are subsequently transported north- 25 westward, instead of north-eastward as in CNTRL. It is interesting to note how both the Mediterranean and north America are strongly affected by the increase in track density.
However, the zonal averages depicted in Fig. A2 show that there is little change between RETRO and CNTRL. The retrograde planet has slightly more intense cyclones (less than 2 hPa averaged difference) which last slightly more in the mid-   latitudes and have a smaller baroclinic growth rate. The difference in the baroclinic growth rate is the only one that appear to be statistically significant, especially in the Atlantic and Pacific (not shown), where it is significantly reduced in the RETRO case. This could be related to the lack of orography when the winds blow the other way, which changes baroclinicity.
In the southern hemisphere, cyclones seem to shift from the Indian and Atlantic ocean to the Pacific ocean, as presented in Fig. A3. Consistent with that, the cyclones in the Pacific Ocean in the retrograde planet are stronger, have higher growth rates, Figure A3. Track density (number of track per season for a unit area equivalent to a 5 degree spherical cup) computed for the southern hemisphere during JJA season. The data are seasonally averaged over the entire simulation period (1000 years) Shown in the leftmost plot is the difference between the retrograde and the prograde planet.

5
are are also faster (see Fig. A4). However, these local changes are compensated such that the overall zonally-averaged statistics in the southern hemisphere do not change significantly (not shown).

Appendix B: Changes in the Moisture Transport
As changes in the major wind patterns and cyclone tracks are associated with changes in the atmospheric energy transport we also analyze changes in the transport of dry-and moist-static energy. To calculate the transports, we use 2 hourly data of 10 velocities, surface pressure, temperature and specific humidity (Keith, 1995). We specifically focus on the moist static energy transport, as it appears to play an important role for the regional climates of North Africa and the Middle East as well as for the freshwater distribution in the Atlantic and Indian Ocean in RETRO, as discussed in sections 3 and 5.
The reversal of the westerlies and trade winds in RETRO leads to a significant shift in the moisture transport. Due to the westerly trade-winds in RETRO, large amounts of moist-static energy are transported from the Atlantic towards North Africa    Figure C1) for RETRO and CNTRL. Figure C1 shows the two leading EOFs which, covering the area from the East African to the South American coast (from 10 (40°E-60°W, 30°S-30°N, Fig. A-2), explain 36% and 14% of the total variance for RETRO (62% and 10% for CNTRL). The dominating variability is centered in the Indian Ocean (tropical Pacific for CNTRL). The second EOF captures variability in the western tropical Pacific.
The sectional seasonality of the sea surface temperature anomaly (SSTA, averaged between 5°S and 5°N) is diagnosed in terms of the monthly standard deviation after removing the climatological annual cycle ( Figure C3). In CNTRL, the strongest variability, which occurs in the eastern tropical Pacific in July-August, is different from the observed ENSO peaking in winter months, but the weak variability that characterizes the spring predictability barrier is clearly seen ( Figure C3). Also in CNTRL, the strongest variability occurs at the eastern tropical Pacific in July-August which differs from observed ENSO peaking in winter months. In RETRO, the center of strong variability shifts to the Indian ocean and to the middle tropical Pacific in RETRO with peaks in May-July and around October-November, respectively. The related indices characterizing only the Indian ocean 5 INDINO (10°S to 0°S for RETRO) and tropical Pacific Nino3/4 (for CNTRL) are shown in Figure C3.
The power spectral densities are presented for the Retro-Indian and CNTRL-Pacific Ocean indices: While Nino3/4 reveals a broad peak in the one to ten years period domain, the lower frequency variability in the Indian Ocean levels off as white noise.