Obliquity forcing of low-latitude climate

The influence of obliquity, the tilt of the Earth’s rotational axis, on incoming solar radiation at low latitudes is small, yet many tropical and subtropical paleoclimate records reveal a clear obliquity signal. Several mechanisms have been proposed to explain this signal, such as the remote influence of high-latitude glacials, the remote effect of insolation changes at midto high latitudes independent of glacial cyclicity, shifts in the latitudinal extent of the tropics, and changes 5 in latitudinal insolation gradients. Using a sophisticated coupled ocean-atmosphere global climate model, EC-Earth, without dynamical ice sheets, we performed two idealized experiments of obliquity extremes. Our results show that obliquity-induced changes in tropical climate can occur without high-latitude ice sheet fluctuations. Furthermore, the tropical circulation changes are consistent with obliquity-induced changes in the cross-equatorial insolation gradient, suggesting that this gradient 10 may be used to explain obliquity signals in low-latitude paleoclimate records instead of the classic 65◦N summer insolation curve.


Introduction
The influence of obliquity (axial tilt) on low-latitude insolation is very small; the difference over tropical latitudes between high and low obliquity is less than 10 W/m 2 in the summer or winter Clemens et al., 1996;Mantsis et al., 2014;Bosmans et al., 2015a). Here we focus on the Summer Inter Tropical Insolation Gradient (SITIG), I 23 • N -I 23 • S at June 21 st (light blue line in Figures 2 and   60 3), which shows a better fit to the sapropel record than insolation at 23 • N or monsoon index M. A stronger inter-hemispheric insolation contrast drives a stronger (winter) Hadley circulation (Mantsis et al., 2014) and thus a stronger cross-equatorial moisture transport into the summer hemisphere.
Inter-hemispheric insolation gradients can therefore explain the obliquity signal in the sapropels (through monsoonal runoff) as well as the Indian ocean proxy records without high-latitude mecha-65 nisms. The insolation curves and insolation gradients are discussed in more detail in Section 3.1. We note that obliquity may also affect climate through other insolation gradients (e.g. Reichart, 1996;Leuschner and Sirocko, 2003;Raymo and Nisancioglu, 2003;Antico et al., 2010;Mantsis, 2011), which will be considered in the Discussion (Section 4).
Modelling results previously suggested that the influence of obliquity on the tropics resulted from 70 high latitude forcing, i.e. consistent with the application of the 65 • N insolation curve. According to Tuenter et al. (2003), an increase in obliquity results in higher temperature and humidity at high latitudes. The resulting strengthening of southward moisture transport as well as a stronger Asian low pressure system act to strengthen the monsoons. This remote control (north of 30 • N) accounts for 80-90% of the total obliquity signal in the North-African monsoon, without any changes in land 75 ice (Tuenter et al., 2003). The model they used is EC-Bilt (Opsteegh et al., 1998), a quasi-geostrophic climate model of intermediate complexity. Due to simplifications in model physics as well as low resolution one can question its suitability for modeling tropical climate. Bosmans et al. (2015a) showed, using a state-of-the-art high resolution fully coupled ocean-atmosphere model, EC-Earth, that mechanisms behind the orbital forcing of the North-African monsoon are very different than 80 previously established from the EC-Bilt model. From EC-Earth it emerged that while the effect of precession was larger, a clear obliquity signal appeared in the monsoon as well without high-latitude influences.
In this study we use the EC-Earth model to investigate the influence of obliquity signal on the entire tropics, without land ice changes. Specifically, we investigate whether our results are in line with 85 the SITIG mechanism. The model and experimental set-up are described in Section 2. Section 3.1 describes insolation curves that are used in paleoclimate literature and describes the SITIG mechanism in greater detail. Section 3.2 describe EC-Earth model results such changes in atmospheric circulation that can help explain paleoclimate records, which usually reflect changes in precipitation (for example dust records or the sapropels) and / or wind (such as Arabian Sea records). The results 90 are followed a discussion in Section 4 and a conclusion in Section 5. Figures of model results show the difference between high and low obliquity.

Model and experimental set-up
Here we use the new, state-of-the-art high resolution fully coupled ocean-atmosphere model, EC-Earth, to investigate influence of obliquity signal on the tropics. EC-Earth is used for the Fifth As-95 sessment Report of the Intergovernmental Panel on Climate Change (Hazeleger et al., 2011) and was also used to perform the pre-industrial and Mid-Holocene experiments of the Paleoclimate Modelling Intercomparison Project (Bosmans et al., 2012). Following Tuenter et al. (2003, we performed two idealized obliquity experiments, one with a low obliquity (22.04 • , Tmin) and one with a high obliquity (24.45 • , Tmax). Eccentricity is set to zero, so the Earths orbit is perfectly round and 100 there is no influence of precession. All other boundary conditions are fixed at pre-industrial levels, therefore there are no changes in land ice or in vegetation. The experiments were run for 100 years each, which is sufficient for atmospheric and surface variables, including sea surface temperature, to equilibrate to the insolation forcing. For more details see Bosmans et al. (2015a), where the same obliquity experiments are used to investigate the North African monsoon. insolation has a relatively strong obliquity signal (red line in Figures 2, 3). The latter therefore matches better with the sapropel record (sapropels from core RC9-181, as used by e.g. Rossignol-Strick (1983, 1985 are shown at the bottom of Figure 2). Although insolation peaks (precession minima) at 23 • N match the occurrence of sapropels, 23 • N summer insolation cannot explain the 115 thick-thin alternations in the sapropel record related to obliquity (e.g. Lourens et al., 1996). 65 • N shows a better match, for instance with thinner sapropels S4 and S7 matching lower insolation peaks.  Figure 2), results in an inso-lation curve very similar to that of 65 • N at June 21 st . In SITIG, the obliquity signal originates from the fact that, in contrast to precession, obliquity induces insolation increases on one hemisphere and, at the same time, insolation decreases on the other hemisphere ( Figure 1). Therefore SITIG has a much stronger obliquity signal, relative to precession, than insolation at a single low latitude.

130
At times of high obliquity (Tmax) the insolation gradient over the tropics (SITIG) is stronger than during low obliquity. This also holds for SITIG in austral summer (I 23 • S -I 23 • N at December 21 st ).
This strengthens the interhemispheric pressure gradient between the two limbs of the winter hemisphere Hadley cell, which drives monsoon winds into the summer hemisphere. A strong SITIG may therefore result in stronger cross-equatorial winds and moisture transport into the summer hemi-135 sphere, associated with an intensified winter Hadley cell (Reichart, 1996). A schematic overview of the effect of obliquity on the (winter) Hadley cell given in Figure 11 of Mantsis et al. (2014).
Like insolation at 65 • N at June 21 st , SITIG matches the sapropel record, including the obliquityinduced thick-thin alternations ( Figure 2). An insolation gradient similar to SITIG was introduced by Leuschner and Sirocko (2003), the Indian Summer Monsoon Index (I 30 • N -I 30 • S at August 1 st ).

140
Despite a small lag (due to ISMI being defined on August 1 st and SITIG on June 21 st ) they are rather similar and both have a relatively strong obliquity signal ( Figure 3). 65 • N insolation is often used to interpret paleoclimate record because of the matching patterns, but requires an explanation of how high-latitude insolation affects low-latitude climate. Here, we investigate whether changes in tropical climate and circulation patterns match the SITIG mechanism based on the EC-Earth 145 experiments (next Section, 3.2).
Note that we focus on (peak) summer insolation and climate, under the assumption that the paleoclimate proxies such as the sapropels are influenced by seasonal phenomenon such as monsoons.
Changes in obliquity result in annual mean insolation changes (right side of Figure 1) per latitude as well as changes in the annual mean equator-to-pole insolation gradient. This is briefly discussed in 150 Section 4.3.

EC-Earth obliquity experiments
EC-Earth shows statistically significant differences in net precipitation over the tropics between high (Tmax) and low obliquity (Tmin, Figure 4). Precipitation differences between Tmax and Tmin reach up to and over 100%, for instance over the Sahara and South America (not shown). There is an 155 overall intensification of the North African and Asian monsoons during boreal summer (June-July-August), with a redistribution of precipitation from ocean to land and stronger landward monsoon winds (Figure 4a, Bosmans et al. (2015a)). Differences in wind speed can also be as large as 100%, for instance in the areas of summer monsoon winds into North Africa and India (not shown). Over the equatorial and southern Pacific wind speed changes are small while winds around the North Pacific as 160 well as the North Atlantic Highs are generally stronger. During austral summer (December-January-February) net precipitation and wind changes are smaller than during boreal summer, likely related to the smaller land mass and therefore weaker monsoons on the Southern Hemisphere (SH). The largest net precipitation increases occur over part of the South American and South African summer monsoon regions as well as the Atlantic and Indian Ocean, while net precipitation over the NH 165 tropics is reduced (Figure 4b).
Our experiments indicate, in agreement with the SITIG mechanism, strengthened surface winds towards the summer hemisphere during Tmax (Figures 4a and 4b). The zonal mean cross-equatorial surface winds are northward and they are indeed stronger during boreal summer, extending slightly further into the NH (Figure 5a). With these stronger surface winds the moisture transport into the 170 NH is strengthened as well during Tmax (Figure 5b). Moisture transport into the North-African and Asian monsoon areas is generally higher during boreal summer, with enhanced northward crossequatorial transport mostly over the Indian Ocean (see Figure 7a). Changes over the Pacific are small, which could be related to the absence of land masses which have a stronger response to insolation changes.

Discussion
We have shown that changes in low-latitude climate can arise as a direct result of obliquity-induced insolation changes, using the sophisticated model EC-Earth without land ice changes. Here we discuss that these changes support the SITIG theory (Reichart, 1996;Leuschner and Sirocko, 2003), what the implications are for the interpretation of obliquity signals in low-latitude paleoclimate records and how obliquity-induced gradients may influence global climate.

Model support for the SITIG theory 205
The simulated changes in winter Hadley cell strength during boreal and austral summer are in accordance with the SITIG theory (Reichart, 1996;Leuschner and Sirocko, 2003). symmetric about the equator (as is assumed in the original SITIG hypothesis, Reichart (1996)), nor are the changes in wind and moisture transport zonally invariant, likely due to differences in the land-sea distribution. Nonetheless, a stronger SITIG during high obliquity (Tmax) results in stronger zonal mean winds and moisture transport into the summer hemisphere and a stronger Hadley cell in our EC-Earth results. The Hadley cell as well as the meridional winds and moisture transport also 215 extend farther into the summer hemisphere. This is in agreement with the poleward shift of the latitude of the tropics during Tmax (Rossignol-Strick, 1985;Larrasoaña et al., 2003). In these studies, meridional shifts in the Hadley cell and the tropics are associated to changes in the equator-to-pole insolation gradient, which in summer has a strong obliquity component. These studies also suggest that the insolation gradient over the austral winter hemisphere causes temperature and trade wind 220 changes that can influence the intensity and poleward penetration of the boreal summer monsoons.
However, the winter (intrahemispheric) insolation gradient does not vary with obliquity, but with precession (Davis and Brewer, 2009), so changes in winter (intrahemispheric) hemisphere insolation gradients cannot be used to explain (low-latitude) obliquity signals. Also, we suggest that while the Hadley cell, and thus precipitation patterns, might indeed shift on obliquity time scales due to 225 changes the (summer) equator-to-pole gradient, such a shift does not explain the changes in precipitation amounts, the strength of the Hadley circulation and the strength of cross-equatorial winds and moisture transport that we identify in our obliquity experiments. Further sensitivity studies could shed more light on the relative roles of inter-and intra-hemispheric gradients (see Section 4.3).

Implications for the interpretation of paleoclimate records 230
Obliquity signals in low-latitude paleoclimate records are often interpreted using the 65 • N 21 st June insolation curve based on the matching precession-obliquity interference in the records and the 65 • N insolation curve (Figure 2). The model study of Tuenter et al. (2003), indicating that ∼80-90% of the obliquity signal in the North-African monsoon is due to high latitude influences, supported the use of the 65 • N insolation curve. Some studies, on the other hand, used a P-1 2 T curve to interpret 235 paleoclimate records (e.g. Lourens et al., 1996). This combination of the precession and obliquity parameters is very similar to the 65 • N 21 st June insolation curve, but by using P-1 2 T no direct assumptions on climate mechanisms are made. Our results, based on a much more sophisticated (and realistic) model with fixed land ice, clearly suggest a low-latitude mechanism for obliquity patterns at low latitudes through a direct response to changes in the cross-equatorial insolation gradient. Fur-240 thermore, there is a strong resemblance between (boreal) SITIG and the 65 • N 21 st June insolation curve ( Figure 2, Reichart (1996); Leuschner and Sirocko (2003)) as well as the P-1 2 T curve (Lourens et al., 2001). Hence, the widely applied 65 • N 21 st June insolation curve (Tiedemann et al., 1994;Hilgen et al., 1995;Lourens et al., 1996Lourens et al., , 2001Hilgen et al., 2000;Sierro et al., 2000) needs to be reconsidered in favour of SITIG. SITIG instead of 65 • N 21 st June insolation relies on a physical basis 245 as described by our model results rather than pattern matching, and explains the obliquity influence on tropical climate independently of glacial-interglacial variability. It thus provides an explanation for the obliquity influence in the sapropel record of the past 14 million years during both the recent glacial cycles as well as earlier warmer times.
We note that the original SITIG theory was based on the Mediterranean sapropels (Reichart, 250 1996), which were originally linked to North African monsoon strength (e.g. Rossignol-Strick, 1985;Ruddiman, 2007), but have recently also been attributed to changes in Mediterranean winter precipitation through changes in Atlantic storm track activity (e.g. Tzedakis, 2007;Brayshaw et al., 2011;Kutzbach et al., 2013). In our obliquity experiments, changes in Mediterranean winter precipitation are unrelated to the Atlantic storm tracks, but are of equal magnitude to summer monsoonal runoff 255 into the Mediterranean. In terms of percentages, however, changes in monsoonal runoff are larger (Bosmans et al., 2015b). Another study stating the importance of cross-equatorial insolation gradients, Leuschner and Sirocko (2003), is based the insolation difference between 30 • N and 30 • S, which drives Indian summer monsoon strength through pressure differences (pink line in Figure   2). Leuschner and Sirocko (2003) state that their record of continental dust transport (indicative of 260 monsoon strength) responds immediately to changes in the cross-equatorial insolation gradient. This matches with our results, showing changes in Indian summer monsoon strength as well as crossequatorial winds and moisture transport that are particularly strong over the Indian Ocean. Also, other paleoclimate records in the Arabian Sea have been used to rule out global ice volume as a primary forcing mechanism for the occurrence of obliquity-related variability in the Indian monsoon 265 (Clemens and Prell, 2003;Clemens et al., 1991). Therefore we conclude that, despite the need to determine the relative role of (winter) precipitation over the Mediterranean in the formation of the sapropels, there is sufficient evidence from tropical and sub-tropical paleoclimate records suggesting a low-latitude, direct response to obliquity forcing.

Obliquity-induced gradients and their influence on global climate
270 Leuschner and Sirocko (2003) and Reichart (1996) suggested that through monsoon-induced changes in atmospheric moisture content, a strong greenhouse gas, the Summer Inter Tropical Insolation Gradient (SITIG) may drive glacial-interglacial variability. Indeed we find a significant obliquityinduced change in cross-equatorial moisture transport (Figures 5b, 5d). However, whether changes in atmospheric moisture content resulting from changes in low-latitude atmospheric circulation on 275 orbital time scales can indeed result in global climate change will need to be investigated with longer model experiments including dynamic ice sheets (not included in EC-Earth). Furthermore, the role of precession-induced changes in moisture content would need to be assessed as well.
Another mechanism by which obliquity can affect high-latitude climate and glacial cycles through latitudinal insolation gradients has been proposed by Raymo and Nisancioglu (2003); Vettoretti and  (2011). These studies suggest that the poleward transport of heat, moisture and latent energy is increased during minimum obliquity due to the intensified intrahemispheric (equator-to-pole) insolation and temperature gradient. The increased moisture transport towards the poles combined with low polar temperatures during low obliquity is favourable for ice growth. In our EC-Earth experiments we also find stronger poleward moisture transport outside the 285 tropics during minimum obliquity (Tmin) during both boreal and austral summer as well as in the annual mean (not shown). The response to changes in the equator-to-pole insolation and temperature gradient can be further intensified by albedo changes. In our experiments land ice is fixed but sea ice at the poles responds to polar insolation changes (not shown). Furthermore, changes in poleward energy transport by ocean currents can play a role in obliquity's effect on global climate (Khodri et al.,

Conclusions
The low-latitude SITIG mechanism proposed here is fundamentally different from high-latitude mechanisms previously proposed to explain the obliquity patterns at low latitudes. Our results, based on the sophisticated model EC-Earth, suggest that these patterns may arise from a direct response to changes in the cross-equatorial insolation gradient, i.e. without any influence of ice sheets or other high-latitude mechanisms. Despite such mechanisms, related to ice sheets and / or equator-to-pole gradients, requiring further research, our results suggest that the widely applied 65 • N 21 st June insolation curve needs to be reconsidered in favour of SITIG. the editor and two anonymous reviewers for their constructive remarks, which helped improve this paper. Dr. Tzedakis, P. C.: Seven ambiguities in the Mediterranean palaeoenvironmental narrative, Quaternary Science S12 S11 S10 S9 S8 S5 S7 S6 S4 S3 S2 S1 RC9-181 Analyseries (Paillard et al., 1996). All are for June 21st, except M caloric (green) which is for the summer half year. The lowest part of the figure shows sapropels in core RC9-181 (Cita et al., 1977;Vergnaud-Grazzini et al., 1977). We note that the obliquity signal also appears in older parts of the insolation and sapropel records (e.g. Lourens et al., 1996;Hilgen et al., 1995Hilgen et al., , 2003Zeeden et al., 2014).  The full wind field is given in Figure 6. Wind and moisture transport into the summer hemisphere is stronger during Tmax for JJA (a,b) and DJF (c,d). Solid parts of the blue line indicate where the difference is statistically significant at 95% (based on a two-sided Student t-test). Note that the vertical scales are different.  Results are only shown where the differences are statistically significant at 95% (based on a twosided Student t-test).