The transient impact of the African monsoon on Plio-Pleistocene Mediterranean sediments

. Over the Plio-Pleistocene interval a strong linkage exists between Northern African climate changes with the supply of dust over the surrounding oceans and continental runoff towards the Mediterranean Sea. Both these signatures in the sedimentary record are determined by orbital cycles inﬂuencing on the one hand glacial variability and on the other hand Northern African monsoon intensity. In this paper, we use the intermediate complexity model CLIMBER-2 to simulate African climate during the Plio-Pleistocene between 3.2 and 2.3 million years ago (Myr ago) and compare our simulations with existing and 5 new climate reconstructions. The CLIMBER-2 model is externally forced with atmospheric CO 2 concentrations, ice-sheet to-pography and orbital variations, all of which strongly inﬂuence climate during the Pliocene and Pleistocene. Our simulations indicate that the records of Northern Africa climate oscillate in phase with climatic precession. For the Earth’s obliquity cycle, the time lag between the 41,000-year component in insolation forcing and the climatic response increased after inception of Northern Hemisphere (NH) glaciation around 2.8 Myr ago. To test the outcome of our simulations, we have put emphasis 10 on the comparison between the simulated runoff of grid boxes encompassing the Sahara desert and the Sahel region and the sedimentary records of marine sediment cores ODP Site 659 (Atlantic Ocean) and ODP Site 967 (Mediterranean). In this study we will show for the ﬁrst time an extended Ti/Al record of Site 967 down to 3.2 Myr ago. This record strongly correlates with runoff in the Sahara and Sahel regions, whereas correlation with the dust record of Site 659 is moderate and slightly improves after NH ice-sheet inception. We investigated the transient variability of the individual and combined contributions of the Sahel 15 and Sahara regions and found signiﬁcant transient behaviour, overlapping with the Plio-Pleistocene transition and inception of NH ice sheets. Prior to 2.8 Myr ago, a larger contribution from the Sahara region is required to explain variability of Mediterranean dust input. After this transition, we found that a more equal contribution of the two regions is required, representing an increased inﬂuence of Sahel runoff and wet periods.

. Climatic forcing records used as input for CLIMBER-2 over the time studies from 3.2 to 2.3 Myr ago. a) Eccentricity in black, climatic precession in red and b) obliquity in blue (Laskar et al., 2004). c) The benthic LR04 δ 18 O record (Lisiecki and Raymo, 2005) used for reference. d) Global ice volume reconstruction (in 10 6 km 3 on land) for the Antarctic ice sheet (AIS) in orange and the total NH in blue, from de Boer et al. (2014). e) Atmospheric CO2 concentrations (in parts per million; ppm) generated using the methodology described in van de Wal et al. (2011).

Climatic dust records
We will compare the output of CLIMBER-2 with three different sites presenting dust or terrestrial variations driven by climate variability over the African continent ( Figure 1). The core drilled at ODP Site 967 is an exceptional record of paleoclimate with detailed cyclic variability (Lourens et al., 2001). It is located in the eastern Mediterranean Sea, south of Cyprus. The 110 5 https://doi.org/10.5194/cp-2020-97 Preprint. Discussion started: 4 August 2020 c Author(s) 2020. CC BY 4.0 License. interval consists of six sapropels, also analysed through colour reflectance (Lourens et al., 2001). The sapropels are cyclostratigraphically correlated to the sections of Hilgen (1991), Lourens et al. (1996), and Kroon et al. (1998), which all originate from different sites within the Mediterranean. The Ti/Al ratio proxy is used to reconstruct variations in climate, whereas variability in Ti/Al represents variation in relative contribution of aeolian (i.e. Ti-rich dust particles) and fluvial (i.e. Al-rich suspended clay components) terrigeneous input in the sediment core (Wehausen and Brumsack, 2000;Lourens et al., 2001;Konijnendijk 115 et al., 2014;Grant et al., 2017). It is suggested that Sahara dust is a prior contributor to the aeolian flux, while a large proportion of the fluvial input originates from the river Nile. Hence, low Ti/Al values indicate humid conditions, for example during periods of sapropel formation in the eastern Mediterranean, and high Ti/Al values correspond to more arid conditions at the northern part of the African continent (Wehausen and Brumsack, 2000;Lourens et al., 2001). An astronomically-tuned Ti/Al time series was established for the 2.9-2.4 Myr ago time interval by correlating minimum (maximum) values in the Ti/Al 120 record to their representative maximum (minimum) values in the Laskar et al. (2004) summer insolation curve at 65 • N latitude (Lourens et al., 2001). Here we extend the tuned Ti/Al time series down to 3.2 Myr ago, using up to now unpublished Ti/Al data that has been generated using similar procedures as described in Lourens et al. (2001).
ODP Site 659 is located on top of the Cape Verde Plateau, northwest of Africa ( Figure 1). Its dust record encompasses the past 5 Myr, and is mostly influenced by the African easterly jet stream, which transports dust from the Sahara-Sahel region 125 towards the Atlantic Ocean (Tiedemann et al., 1994). We have re-tuned the dust record of ODP Sites 659 (Wang et al., 2010) to the LR04 benthic stable isotope stack (Lisiecki and Raymo, 2005).

Orbital pacing of CLIMBER-2 modelled runoff
Runoff over the Northern African continent as modelled in CLIMBER-2 results from the difference between precipitation (P) and evaporation (E), i.e. Runoff = P -E over land. As can be seen in Figure 1, grid box 12 is mainly covering the Sahara desert, 130 whereas grid box 11 includes the Sahel region, which is more dominated by grassland. This difference is clearly represented by the variation in runoff over the Plio-Pleistocene transition ( Figure 3a). It is evident that the minimum runoff of grid box 11 is approximately the equivalent of the maximum runoff of grid box 12. Grid box 11 has large variations in amplitude, with runoff maxima occurring during precession minima (Figure 3a). For the Sahel region the runoff is strengthening following the African summer monsoon, driven by an increase in NH insolation. In contrast, the runoff values of grid box 12 (Sahara desert) 135 do not increase by the strengthened monsoon, but show peaks of low runoff during precession maxima. Although precipitation is enhanced during the summer monsoon when the air from the Atlantic Ocean reaches land, higher temperatures provide more room for water to evaporate. In the case of grid box 12, this additional precipitation is therefore compensated by an increase in evaporation.
The clear dominance of precession variability during this time interval is well depicted by the global power spectra of both 140 regions ( Figure 4a). Over time, the evolutive power spectra confirm the dominance of precession, although it is lower during eccentricity minima (Figure 3b,c). Clearly high latitude influence is more present for the Sahara (grid box 12), showing an increased obliquity power after inception of NH glaciation (Figure 3b). Table 1. Calculated time lags in kyrs between orbital frequencies and runoff over grid box 11 and 12 for the full period 3.2 to 2.3 Myr ago.
Frequencies are extracted from the data with a Gaussian filter for precession at 23 kyr (0.0435 ± 0.003) and for obliquity at 41 kyr (0.0245 ± 0.003). Data analysed with AnalySeries (Paillard et al., 1996), using a Blackman-Tukey spectral analysis with a Parzen window with 90% lags. CLIMBER-2 experiments are O: orbital, OG: orbital and CO2, OI: orbital and ice sheets, OIG: orbital, CO2 and ice sheets. box 12 (Sahara) and a much weaker correlation with box 11 (Sahel). Note that for the correlation we have used the inverse relationship, correlating high runoff with low dust output, and corrected for the time lags of 1 kyr for the chronologies of ODP Site 659.

165
The influence of the North African monsoon and its relation with orbital forcing is also clear through the presence of Mediterranean sapropels (e.g. Hilgen, 1991;Lourens et al., 2001). Sapropels are organic-rich sedimentary layers, resulting from anoxic events caused by an increase in runoff that is associated with a strong summer monsoon. Over the Plio-Pleistocene transition we compare the Ti/Al record with CLIMBER-2 runoff form grid box 11 and 12 ( Figure 5). Clearly, both the wet-and dry periods correspond well with the simulated runoff, although the runoff maxima from grid box 11 (Sahel) largely link to low 170 Ti/Al peaks, and runoff minima of grid box 12 (Sahara) overlay with maxima of the Ti/Al record. The sapropel layers that have been found in ODP Site 967 (Emeis et al., 1996;Kroon et al., 1998;Lourens et al., 2001) can be correlated with the high runoff peaks of grid box 11. This shows that the Sahel region does not only indicate the increased African monsoon phases, but also correlates to the corresponding sapropel layers in the Mediterranean Sea. The Sahara region correlates with the high Ti/Al values, indicating dry phases. These dry phases of the Sahara region can therefore be correlated to the marls, i.e. periods 175 of high dust flux, in between the sapropel layers of ODP Site 967 Lourens et al. (2001).
Although the general pattern correlates very well (Table 2), there are also clear differences between the CLIMBER-2 simulations and the Ti/Al record. For example, the relative amplitude variations of the humid phases are not always the same.
Sapropel 64, is clearly enhanced by obliquity, whereas the runoff at this time has a relatively low maximum. On the other Figure 5. A comparison between CLIMBER-2 runoff from grid box 11 (blue) and 12 (orange) and the Ti/Al record from Site 967 (green) from 3.2 to 2.3 Myr ago. Sapropel layers are labeled (S61 to S80) as adopted from Kroon et al. (1998) and Lourens et al. (2001) corresponding to Ti/Al minima and runoff maxima. Note that the left y-axis for the Ti/Al is reversed, with wet phases on top.
hand runoff is significantly high during the formation of sapropel 65, which is one of the less distinct sapropel layers. This is 180 remarkable, as a large runoff in Northern Africa would suggest a lower Ti/Al ratio due to more riverine input in the Mediterranean Sea. The most apparent discrepancy between the Ti/Al record and the CLIMBER-2 simulated runoff is shown during the 400-kyr related eccentricity minimum around 2.8 and 2.4 Myr ago. During these long-term eccentricity nodes, the Ti/Al record suggests on average drier climate conditions in northern Africa than indicated by the model simulations At this time interval, both the runoff amplitudes are significantly lower and the variations for grid box 12 are even out of phase with the 185 sapropel layers S68 and S69 ( Figure 5).

Combined runoff and overall correlation
It appears that the influence of wetter periods (grid box 11) overshadows the effect of dry periods (grid box 12), given that the amplitude variations of grid box 12 (Sahara) are a lot less than runoff from grid box 11 (Sahel). Both regions include the catchment area of the river Nile (e.g. Bosmans et al., 2015b), and hence a combination of runoff of the two sites would be more with the data. The combination of 35% runoff 11 plus 65% runoff 12 has a correlation of 0.792 with the Ti/Al record, and 20% runoff 11 plus 80% runoff 12 has a correlation of 0.602 with the dust record of Site 659. For both sites this correlation is higher than with the individual grid boxes, as shown in Table 2   greater parts of Northern African climate and the Mediterranean. Although the CLIMBER-2 model is of low resolution, it also shows to have a strong coherence with sedimentary records from the Mediterranean especially for precessional frequencies.