Long‐Distance Asthenospheric Transport of Plume‐Influenced Mantle From Afar to Anatolia

The origin of widespread volcanism far from plate boundaries and mantle plumes remains a fundamental unsolved question. An example of this puzzle is the Anatolian region, where abundant intraplate volcanism has occurred since 10 Ma, but a nearby underlying plume structure in the deep mantle is lacking. We employed a combination of seismic and geochemical data to link intraplate volcanism in Anatolia to a trail of magmatic centers leading back to East Africa and its mantle plume, consistent with northward asthenospheric transport over a ∼2,500 km distance. Joint modeling of seismic imaging and petrological data indicates that the east Anatolian mantle potential temperature is higher than the ambient mantle (∼1,420°C). Based on multiple seismic tomography models, the Anatolian upper mantle is likely connected to East Africa by an asthenospheric channel with low seismic velocities. Along the channel, isotopic signatures among volcanoes are consistent with a common mantle source, and petrological data demonstrate similar elevated mantle temperatures, consistent with little cooling in the channel during the long‐distance transport. Horizontal asthenospheric pressure gradients originating from mantle plume upwelling beneath East Africa provide a mechanism for high lateral transport rates that match the relatively constant mantle potential temperatures along the channel. Rapid long‐distance asthenospheric flow helps explain the widespread occurrence of global intraplate magmatism in regions far from deeply‐rooted mantle plumes throughout Earth history.

2 of 22 Yamamoto et al., 2007). This information is critical to understanding the role of plumes throughout Earth history, including large igneous provinces where magmas sharing a mantle source erupt over massive areas (Madrigal et al., 2016;Marzoli et al., 1999;Stern et al., 2020), and on planets like Venus that lack moving tectonic plates (Smrekar et al., 2010).
In this paper, we address these fundamental questions by exploring the source of intraplate magmatism in Anatolia and its connection to lower mantle plumes. The tectonically active Anatolian region lies at the northeast edge of the Mediterranean Sea (Figure 1a). To the north, the Anatolian plate is separated from the Eurasian plate by the North Anatolian Fault; to the south, the African plate subducts beneath Anatolia via the Cyprus trench; and to the southeast, the Anatolian plate is compressed by the Arabian plate (Bird, 2003).
Extensive intraplate volcanism exists in Anatolia, and while the erupted basaltic magmas have indicated mantle temperatures and compositions similar to hotspot volcanoes in some areas (McNab et al., 2018;Nikogosian et al., 2018;Reid et al., 2017), this region lacks evidence for a mantle plume that is locally connected to the lower mantle (French & Romanowicz, 2015;Lei et al., 2020). To explain the intraplate magmatism, regional processes within Anatolia have been proposed, including upwelling related to lithospheric extension, lithospheric delamination, and slab rollback and/or fragmentation (e.g., Delph et al., 2017;Göğüş & Pysklywec, 2008;Keskin, 2007;Lynner et al., 2022;Memiş et al., 2020). While these processes dramatically alter the lithosphere, in most cases they do not result in broad zones with significantly hotter asthenosphere and elevated mantle potential temperatures, although local temperature increases of tens of degrees may be produced (e.g., King & Ritsema, 2000;Van Wijk et al., 2008). Alternatively, other studies have proposed that hot mantle derived from upwelling beneath the East African Rift (EAR) flows horizontally over thousands of kilometers toward Anatolia (Ershov & Nikishin, 2004;Faccenna et al., 2013;Hansen et al., 2012;Wei et al., 2019). This model is consistent with regional seismic velocity and anisotropy patterns (Hansen et al., 2006;Qaysi et al., 2018;Wei et al., 2019), and also local seismic tomography results (e.g., Berk Biryol et al., 2011) that suggest that hot mantle material could flow into the Anatolian region through the slab window opened by the Arabian plate-Anatolian plate collision. This type of model provides a possible explanation for high mantle potential temperatures beneath Anatolia. Once in Anatolia, this hotter-than-average asthenosphere could interact with regional processes such as lithospheric extension, delamination, and slab roll-back and break-off.
To systematically test the plausibility of long-distance transport, several questions must be addressed: (a) What are the bounds on mantle potential temperature beneath Anatolia and are they indeed in the range of hotspot regions globally? (b) Does mantle along the transport path record a common mantle source? (c) Is the estimated mantle temperature along the transport path consistent with a single, plume-related, heat source? (d) With reasonable mantle viscosities, is the long-distant transport consistent with constraints from geodynamic models?
In the following sections, we address these questions by joint interpretation of seismic, petrological, and geochemical data, aided by geodynamical modeling, to make the case that rapid lateral transport of high temperature plume-derived asthenosphere from the EAR to Anatolia is possible. To answer question (a), in Section 2 we used novel joint modeling of seismic receiver function phases and basaltic primary magma equilibration conditions to explain the extensive upper mantle melting beneath Anatolia and to constrain upper mantle potential temperature. To answer question (b), in Section 3 we analyzed multiple seismic tomography models to verify that they are consistent with a channel of high-temperature asthenosphere, and we compared various radiogenic isotope ratios along the transport route. To answer question (c), in Section 4, we obtained mantle potential temperatures at different locations along the path with basaltic magma samples. Finally, to answer question (d), in Section 5, we used a simplified 1D pressure-driven Poiseuille flow model to demonstrate the physical plausibility of rapid long-distance asthenospheric transport.

Extensive Partial Melting Beneath Anatolia
The Anatolian upper mantle has recently been imaged with Sp receiver functions (Hua, Fischer, Wu, & Blom, 2020), an approach which uses conversions of S phases to P phases to measure the depths and polarity of velocity gradients. In particular (Hua, Fischer, Wu, & Blom, 2020) employed common-conversion point (CCP) stacking to construct a three-dimensional model of crust and mantle velocity gradients using a new approach that incorporates Sp sensitivity kernels in the spatial functions used to weight the receiver functions as they are Writing -review & editing: J. Hua, K. M. Fischer, E. Gazel, E. M. Parmentier, G. Hirth summed at each point in the stack. The stacking employed 23,787 individual receiver functions. Within the upper mantle, in addition to the commonly observed negative shear wave velocity gradient (positive receiver function amplitude) that represents the seismic lithosphere-asthenosphere boundary (LAB) (e.g., Figure S1i in Supporting Information S1), the receiver function stack reveals an unusual and widespread asthenospheric positive velocity gradient (PVG) at depths of 100-150 km (Figure 1b and Figure S1 in Supporting Information S1) that is indicated by negative receiver function amplitudes. To study the depth range over which this gradient extends, similar stacking of Sp receiver functions was performed using seismograms from a range of bandpass filters. This analysis demonstrated that the velocity gradient is gradual in depth (it extends over more than 30 km) and is better characterized with long-period data (Hua, Fischer, Wu, & Blom, 2020). Hence, in this study, we used stacks from seismograms with a 10-100 s bandpass filter to study the PVG. The shallower negative velocity gradient (positive receiver function amplitude) associated with the LAB is observed at ∼70 km depth ( Figure S1 in Supporting Information S1) (Hua, Fischer, Wu, & Blom, 2020). However, due to the thin mantle lithosphere, this phase is clearer at shorter periods (Hua, Fischer, Wu, & Blom, 2020), so we used stacks from a 2-20 s filter to study the LAB velocity gradient.
The depths of the PVG and LAB were measured from the Sp CCP stack (Hua et al., 2018, see Text S1 in Supporting Information S1 for details). The PVG is present primarily at depths of 100-150 km beneath most of the Figure 1. Constraints on seismic velocity gradients and melting onset depths from Sp receiver functions. (a) Depth of the asthenospheric positive velocity gradient (PVG) at depths of 100-160 km observed in the Sp common conversion point (CCP) stack (see Section 2.1), which we interpret as the lower boundary of a layer containing partial melt. Colors show PVG depths; dark gray regions lack a clear observation of this feature. Profile A-A' location shown by white line with green circles. The Karacadag volcanic field is labeled by the red triangle, and thick black lines show plate boundaries (Bird, 2003). Subduction boundaries are labeled with triangles and divergent boundaries with bars. (b) The amplitude of the Sp receiver function CCP stack for profile A-A' (10-100 s bandpass filter). Negative amplitudes correspond to PVGs (and positive amplitudes to negative velocity gradients). The depths of the PVG associated with the base of the partial-melt-bearing layer are shown by red circles. Error bars (black lines) are the depth ranges spanned by the Sp phase from this gradient, estimated as three times the standard deviation of the receiver function amplitude distributions (e.g., Figure S1a in Supporting Information S1). Positive amplitudes at depths of 50-90 km represent the lithosphere-asthenosphere boundary. Green circles correspond to markers in (a).
Anatolian region (Figure 1a), while the LAB phase exists mainly at depths of 60-85 km (Figure S1h in Supporting Information S1). The PVG phase is consistent with the base of a low velocity asthenospheric layer seen in two regional full waveform tomography models Fichtner et al., 2013), and the LAB phase is consistent with the top of this layer (Hua, Fischer, Wu, & Blom, 2020).
To probe the origins of the mantle PVG phase, we compared its depth to mantle geotherms inferred by scaling the regional velocity models Fichtner et al., 2013) to temperature, using a range of experimentally-derived velocity-temperature relationships (Text S2 in Supporting Information S1). Although the resulting uncertainties in mantle temperature are large, this comparison shows that the PVG depths are similar to the depths where the seismically-inferred geotherms cross to the high temperature side of a dry or slightly damp solidus with decreasing depth (Text S2 in Supporting Information S1). Thus the PVG can be interpreted as the base of an asthenospheric layer that contains partial melt.
However, a key goal is to constrain mantle potential temperature (T P ) beneath Anatolia, and to determine whether it is elevated with respect to ambient mantle, or whether the melting is related to elevated mantle water content, for example, due to the adjacent Cyprus trench, without the need for excess temperature from a plume-derived mantle. As shown in Text S2 in Supporting Information S1, the geotherms derived from scaling shear velocity to temperature are highly dependent on the chosen conversion relationship between shear wave velocity and temperature and assumptions about other mantle parameters such as water content ( Figure S2 in Supporting Information S1). We therefore do not use this type of velocity-temperature scaling to infer mantle temperatures for the remainder of this paper. Instead, we determine T P using a new, more precise, approach that combines inferred melting depths from the PVGs in the Sp receiver function stack and independent constraints on the mantle geotherm from erupted basalts. This approach does not rely on any pre-existing velocity models or empirical velocity-temperature relationships.

Primary Magma Equilibration Pressure and Temperature (P-T) Conditions
To provide independent constraints on the mantle geotherm from major element compositions of erupted basalt samples, we first selected reliable basalt samples from the Karacadag volcanic field (Figure 1a). We also estimated the temperatures and pressures of last mantle equilibration from basalts in other Anatolian volcanic fields, but the Karacadag field was the only location where a sufficient number of asthenospheric samples were found. In this study, all samples in the GEOROC database (http://georoc.mpch-mainz.gwdg.de/Georoc, downloaded in 2020) from 37°N to 38°N and 39°E to 40.5°E were classified as being from Karacadag (Data Set S1). The FeO and Fe 2 O 3 weight percentages were first converted to FeO total to represent the total iron weight percentage. Then we calculated the sum of the weight percentages for major elements, and only samples with summed weight percentage between 98% and 102% were used. For these samples, the major element weight percentages were normalized to make their sum equal to 100%. We only used samples that follow an olivine control trend based on major element (FeO, TiO 2 , CaO, Al 2 O 3 , CaO/Al 2 O 3 , Na 2 O, etc.) variations with MgO, and limiting the samples to those with MgO higher than 8 wt.%, SiO 2 higher than 45 wt.%, and CaO higher than 8 wt.% to be consistent with experimental results for peridotite sources that are the basis for modeled primary magma FeO-MgO thermometers (Herzberg & Asimow, 2008;Lee et al., 2009). The average location for the 117 selected samples is 37.58°N, 39.83°E. The Sp PVG phase at this location (Figure S1a in Supporting Information S1) is centered at 138 km depth with a phase depth extent of 28.5 km (defined as 1.5 times the distribution standard deviation, e.g., Figure  S1a in Supporting Information S1), as obtained from the receiver function phase picking.
Primary magma equilibration pressure-temperature (P-T) conditions for the basaltic magma samples were calculated based on the Lee et al. (2009) parameterization ( Figure 2a). Starting from the major element compositions of the collected melt samples, this method first adds olivine increments in equilibrium with the instantaneous melt composition back to the melt until the melt reaches equilibration with the suggested ambient mantle composition (explaining why only samples in an olivine control trend can be used); the corresponding melt composition represents the primary magma. Then, the primary magma equilibration P-T conditions are estimated from the primary magma major element compositions using an experimentally-defined empirical relationship for liquids in equilibrium with a mixture of orthopyroxene and olivine (Lee et al., 2009). Therefore, as long as the source has some of these two components, the parameterization is valid, even if the source is a pyroxenite. That said, to avoid possible olivine-free sources, we only used samples with CaO > 8 wt.% (Herzberg, 2006(Herzberg, , 2011 within our primitive population to produce the most internally consistent data. A more recently calibrated empirical relationship also exists (Plank & Forsyth, 2016), based on the same experimental data used by Lee et al. (2009); the resulting T-P estimates are similar (Plank & Forsyth, 2016). We therefore chose to continue using the well-established thermobarometer in Lee et al. (2009).
The estimated final P-T equilibration conditions also depend strongly on the assumed water weight percentage in the melt sample, oxidation state (controlling the Fe 3+ /Fe Total ) and source mantle forsterite content (Fo) (Figures 3a-3c). We assumed 1 wt.% water in the melt, a typical value for primitive intraplate basalts (Dixon & Clague, 2001;Plank & Forsyth, 2016). This value is also in the range of water contents in olivine hosted melt inclusions that were collected at a volcano ∼300 km away from Karacadag (Özdemir et al., 2011). To further validate the assumed water content, the Ce abundances of these basalt samples were used as a proxy (Dixon et al., 2002). Ce values were multiplied by 200 (Dixon et al., 2002) to obtain the estimated water content (inset in Figure 2b), which is also distributed around 1 wt.%. If the actual water in the melt is not approximately the assumed value, this parameter could introduce a ∼30°C difference in the calculated mantle equilibration temperature (e.g., Figure 3a). The effects of Fe 3+ /Fe Total and Fo on equilibration conditions and mantle poten tial temper ature are nearly identical (Figure 3d), so we fixed Fo at 90% and only solved for bounds on the value of Fe 3+ /Fe Total , using a new approach that combines information from both the LAB depth determined from the receiver functions and the primary magma equilibration depths. This method is described in Text S3 in Supporting Information S1. The resulting Fe 3+ /Fe Total value is 0.16, consistent with high temperature upper mantle conditions, as shown by samples from other intraplate locations like Hawaii and Iceland that also indicate similar Fe 3+ /Fe Total values (Moussallam et al., 2019). With these parameters, we calculated the equilibration P-T conditions of the Karacadag melt samples (Lee et al., 2009) (Data Set S1).

Joint Modeling of Receiver Function and Primary Magma Equilibrium Conditions
With both the Karacadag melting onset depth at 138 km from the Sp receiver function PVG phase and the estimated primary magma equilibration P-T conditions from the 117 reliable Karacadag basalt samples, both the mantle potential temperature (T P ) and the water content of the Karacadag volcanic field are constrained through joint modeling. We first assumed that the primary magma P-T conditions represent the actual mantle conditions, and the asthenospheric geotherm follows an adiabat. Values of T P and mantle water content were then obtained by requiring that the corresponding mantle geotherm approximates the equilibration P-T conditions, and that the geotherm intersects the hydrous solidus for the given water content (Hirschmann et al., 2009) such that the onset of melting matches the seismically-observed (PVG) melting depth (Figure 2a). The effect of changes in Fo and Fe 3+ /Fe Total on the equilibration conditions for a single sample (Sample ID 3983 in Data Set S1). Fo is fixed at 90% for triangles, and Fe 3+ /Fe Total is fixed at 0.16 for circles. (e) Equilibration depth versus sample age for the estimated and assumed parameters (1 wt.% water in melt, Fo equal to 0.9 and Fe 3+ /Fe Total equal to 0.16). Error bars show age uncertainties from published studies.
In practice, a grid search of T P and mantle water content values was employed for the modeling (Figure 4). We tested T P from 1,300°C to 1,550°C, and water from 0 to 400 wt. ppm. For each tested T P -water combination, we estimated two misfits between model predictions and conditions based on either the observed melting depth from the Sp stack or the basalt samples. One misfit is | soli − geo| (Figure 4a), where soli is the hydrous solidus temperature (Hirschmann et al., 2009) at the observed melting depth from the PVG with the assumed water content, and geo is the predicted geotherm temperature at the same depth for the values of T P and water (see Text S4 in Supporting Information S1 for how we construct geotherms). The misfit between the predicted geotherm and the basalt equilibration P-T conditions is characterized as ‖ sample − geo‖2 (Figure 4b), where in this L2-norm misfit function, sample represents the primary magma equilibration temperature at its equilibration depth, while geo represents the predicted geotherm temperature at that equilibration depth. Each of the misfit functions reaches a minimum value along a curve that reflects trade-offs in the effects of T P and water ( Figure 4). However, the intersection of the two misfit minima curves resolves the T P and water conditions that satisfy both seismic and basalt data. The best-fitting values for the Karacadag basalts are 1,420°C ± 6°C for T P and 100 ± 17 wt. ppm for mantle water content (these uncertainties are on order of the two-standard deviation uncertainties for our analysis outlined in the Text S5 in Supporting Information S1). The resulting T P value of 1,420°C ± 6.4°C for the Karacadag field is consistent with the 1,420°C T P estimate from Reid et al. (2017), obtained with a different method. It also falls within the range of elevated T P values for eastern Anatolia from McNab et al. (2018), and is consistent with the deeper melting regime and larger melt fractions inferred from rare earth elements analyses in that study. While the temperatures we computed are higher than typical ambient mantle, re-equilibration could had happened between the hot upwelling melt and a normal ambient mantle, biasing our results to lower temperatures. If this process affected our samples, then this implies the primary magma temperatures could have been even higher than is recorded by the lavas used in this study, and thus, the T P responsible for melting would have been >1,420°C.
We also tested potential bias in our interpretation of the Sp receiver function PVG phase by making different assumptions about the relationship of the phase to melting processes. In one alternative case we assumed that the mid-point depth of the PVG corresponds to 0.5% of melting instead of incipient melting, and in another case we assumed that the lower boundary of the PVG is the melting onset depth (Text S6 and Figure S3 in Supporting Information S1). Both scenarios lead to higher estimated water contents (185 and 155 wt. ppm), but similar estimated mantle T P (1,430°C and 1,425°C), suggesting that the estimated T P is robust.
To further test the estimated values of T P and mantle water content, we predicted the degree of melting for the inferred geotherm (Hirschmann, 2010;Hirschmann et al., 2009;Katz et al., 2003) (see Text S4 in Supporting Information S1). The predicted degree of melting correlates with the observed equilibration depth distribution of the basaltic samples ( Figure 2b). Since melt is more likely to be extracted at depths with higher degrees of melting, this agreement validates the T P and mantle water content estimates and the assumption that P-T conditions represent the geotherm. An alternative interpretation for the equilibration P-T values from the basalts is that they represent re-equilibration of the magmas at the base of the lithosphere (Gazel et al., 2012;Plank & Forsyth, 2016), and that the distribution of equilibration depths ( Figure 2b) is caused by lithospheric thinning. However, this interpretation requires that magma samples equilibrated at shallower depths be younger than those from deeper depths, which is not evident in the basalt data ( Figure 3e). Mantle T P values were also estimated for other regions of Anatolia. We first assumed the solidus for a mantle source with 100 wt. ppm of water (Hirschmann et al., 2009), consistent with our analyses of the Karacadag basalts with the assumption that the center of the PVG defines the melting onset depth (Figure 2). Then we determined T P across the region by requiring that its corresponding geotherm intersects the solidus at the depth of the local seismically-determined PVG, again associated with the onset of mantle melting. The resulting T P distribution for Anatolia ranges from 1,350°C to 1,450°C ( Figure S4a in Supporting Information S1).

Mantle Conditions Beneath Karacadag Relative to Ambient Mantle
Based on Herzberg and Gazel (2009), the T P of normal ambient mantle ranges from 1,300°C to 1,400°C. If we use the set of adiabat-related parameters used in that study (Iwamori et al., 1995), the estimated T P in Karacadag would be 1,435°C, higher than the value of 1,420°C that is based on parameters from Katz et al. (2003) we used here. Therefore, Karacadag is at least ∼35°C hotter than the upper bound for a normal ambient mantle. The Karacadag T P does overlap the higher end of the mid-ocean ridge T P ranges of some previous studies (Bao et al., 2022;Courtier et al., 2007;Dalton et al., 2014;Putirka, 2008). However, differences in T P measurement methods make it difficult to directly compare the Karacadag T P values to some of these studies, for example, the T P values inferred from seismic velocities in Bao et al. (2022). In addition, when determining ambient mantle T P ranges, the use of basalt samples that are close to active mantle plumes or do not lie on an olivine control trend would obscure ambient mantle conditions (Herzberg & Asimow, 2008;Herzberg et al., 2007;Madrigal et al., 2016). Hence, we choose to put more weight on the comparison of the Karacadag T P values to Herzberg and Gazel (2009), where accounting for differences in T P estimation is straightforward.

Potential Lower Mantle Roots
To test the idea that the elevated T P beneath Karacadag reflects high temperature mantle that flows into the upper mantle beneath Anatolia, we searched for the signatures of potential heat sources in multiple global and regional mantle seismic velocity models (Figure 5,Figures S5 and S6 in Supporting Information S1), including SEMUCB_WM1 (French & Romanowicz, 2015), GLAD_M25 (Lei et al., 2020), S362WMANI+M (Moulik & Ekström, 2014), EAV09 (Chang et al., 2010), CAM2016 (Ho et al., 2016), 3D2018_08Sv , Africa.ANT. Emry-etal.2018(Emry et al., 2019 and CESM_Europe (Fichtner et al., 2018). Based on these models, no low velocity mantle plume structure coming from the lower mantle is evident beneath Anatolia, and the closest low velocity anomalies that connect to the lower mantle include plume structures beneath the East African Rift system (Afar hotspot) and the Eifel hotspot, as well as the lower mantle Perm Anomaly ( Figure 5 and Figure S6 in Supporting Information S1) (French & Romanowicz, 2015;Lei et al., 2020). However, all of these regions are more than 2,000 km away from Anatolia. Although the Perm Anomaly is the closest low velocity lower mantle body, no obvious low velocity anomaly connects it to the Anatolian upper mantle (Figure 5c and Figure S6b in Supporting Information S1). In contrast, the EAR region, which lies above a broad plume of low velocity mantle extending from the core-mantle boundary to the upper mantle, is connected to Anatolia through a pronounced channel of low velocity upper mantle ( Figure 5, Figures S5 and S6 in Supporting Information S1). While a low velocity upper mantle anomaly also connects the Eifel region to Anatolia (Figure 5b and Figure  S6a in Supporting Information S1), this anomaly is weaker and shallower than the EAR-Anatolia channel, and the Eifel region is less obviously connected to a deeply-rooted mantle plume. We therefore focus on the mantle plume beneath the EAR as a potential source for the high temperature mantle beneath Anatolia, although we also evaluate the Eifel region.
We also acknowledge that other models are possible, for example, that a plume connected to the lower mantle actually exists directly beneath Anatolia but has not been imaged, or that some fortuitous combination of regional scale processes (such as extension or delamination in the upper plate lithosphere, or slab roll-back and breakoff Göğüş & Pysklywec, 2008;Keskin, 2007;Lynner et al., 2022;Memiş et al., 2020)), has contributed to the high Anatolia T P . However, our focus is to assess whether lateral flow of high temperature asthenosphere over thousands of kilometers from the EAR is plausible, rather than disproving other scenarios.
Prior studies have supported the idea that asthenosphere mantle flows from the EAR to Anatolia based on geochemical data (Ershov & Nikishin, 2004;Faccenna et al., 2013), anisotropy in seismic wave velocities (Faccenna et al., 2013;Wei et al., 2019), and geodynamical modeling in which driving forces from plume upwelling and pull from Mediterranean slabs draw upper mantle from the EAR to Anatolia (Faccenna et al., 2013). In addition, this type of flow is also evident in global mantle flow models (e.g., Conrad & Behn, 2010). In this study, we test the possibility of upper mantle transport from the EAR to Anatolia using isotopic data and T P estimates for basalts erupted along the low velocity channel.

Radiogenic Isotope Ratios
Radiogenic isotope ratios from 1,004 samples were used to assess whether basalts erupted along the proposed EAR-Anatolia transport channel share a common mantle isotopic composition ( Figure 6, Figures S7 and S8 in Supporting Information S1). The ratios were obtained from the GEOROC database from all studies including regions from the EAR to Anatolia published between 1990 and 2020. Maximum and minimum ages for the samples were obtained from their original publications (Data Set S2).
We first analyzed 143 Nd/ 144 Nd and 87 Sr/ 86 Sr isotopes. For this analysis, 143 Nd/ 144 Nd is represented by its relative deviation ε Nd which was calculated based on the present-day chondritic value of 0.512638 (Hofmann, 2007), and only mafic samples with maximum ages within 35 Ma were used. Values of ε Nd versus 87 Sr/ 86 Sr for all samples from the Arabian plate to the EAR for ages less than 35 Ma are clustered closely (Figure 6b), consistent with a common source that overlaps suggested values for the Afar plume (Rooney et al., 2012). However, samples from Anatolia on the Eurasian/Anatolian plate (Bird, 2003) show age-dependent results consistent with the region's tectonic evolution (Figure 6b and Figure S7a in Supporting Information S1). Anatolian samples with minimum ages older than 10 Ma, when the Anatolian plate was separated from the Arabian plate by a subducting slab (Cosentino et al., 2011), show high 87 Sr/ 86 Sr and negative ε Nd (Figure 6b), consistent with a large contribution from continental sediments that were potentially related to prior subduction (White, 2020). In contrast, Anatolian samples with maximum ages younger than 10 Ma show isotopic ratios that are similar to other regions along the EAR-Anatolia channel, consistent with EAR-sourced mantle entering the asthenosphere beneath Anatolia following the Arabia-Anatolia collision and slab break-off (Cosentino et al., 2011;Faccenna et al., 2013;Wei et al., 2019) (Figure 6b). A similar shift in isotope ratios was also reported in McNab et al. (2018). Some Anatolian samples with ages less than 10 Ma still show 87 Sr/ 86 Sr values higher than 0.704, but similar to what was in Supporting Information S1), which suggests that part of the Anatolian mantle could also be influenced by subduction-modified or lithospheric sources (Class et al., 1998;Gazel et al., 2012) involved in post-collision processes like lithospheric delamination (Keskin, 2007).
We also analyzed Pb isotopes for samples that are not strongly influenced by lithospheric sources to verify mantle source similarity along the EAR-Anatolia channel (Figure 6c and Figure S8a in Supporting Information S1). For the Pb isotope analyses, in addition to the SiO 2 filter, we also required 87 Sr/ 86 Sr to be between 0.7028 and 0.7040 to make sure the source compositions based on 87 Sr/ 86 Sr were similar and crustal and sea water influences were eliminated (White, 2020). We calculated the 95% confidence ellipse for Pb isotope distributions ( 206 Pb/ 204 Pb, 207 Pb/ 204 Pb, and 208 Pb/ 204 Pb, Figures 6c and Figure S8a in Supporting Information S1), assuming a normal distribution for all Pb isotope ratios in each geographic group. To avoid outliers in the Pb isotope analyses, samples outside of the 99.9% confidence ellipse were removed. In what follows, we limited our analysis of samples south of the boundaries between Anatolian/Arabian plates, Eurasian/Arabian plates and Anatolian/African plates  (Bird, 2003) related to the group division in (b) are in pink (Anatolian/Arabian, Anatolian/African, Eurasian/Arabian and Eurasian/African) and others in red. Subduction boundaries are marked with triangles and divergent boundaries with bars. The white line indicates profile D-D' in Figure 8. Gray arrows show negative horizontal pressure gradients based on the stress field from a global mantle flow model (Conrad & Behn, 2010), and are especially large south of 30°N. Blue bars show station-averaged shear-wave splitting azimuthal anisotropy fast orientations for the Arabian Plate (Qaysi et al., 2018), and bar lengths are proportional to splitting times. Transparent wedges around bars show one standard deviation of the orientation. Inset shows the 100 s Rayleigh wave phase velocity anomaly map from Yao et al. (2017), relative to the local average velocity of 3.97 km/s. Geographic location of the inset is outlined by gray lines. (b) 87 Sr/ 86 Sr versus ε Nd from samples grouped by age and tectonic plate (locations in Figure S7a in Supporting Information S1). Histograms show distribution of points in each group. (c) 206 Pb/ 204 Pb versus 208 Pb/ 204 Pb from different geographic groups; ellipses show 95% confidence (locations in Figure S7c in Supporting Information S1). Stars in (b)-(e) are mantle endmembers (Hofmann, 2007), and the Afar plume composition (Rooney et al., 2012). (pink lines in Figure 6a) (Bird, 2003) to those with maximum ages of less than 35 Ma, and limited our analysis of samples north of the boundaries to those with maximum ages less than 10 Ma. Samples along the channel between the EAR and Anatolia were divided into five geographical groups (Figure 6a and Figure S7 in Supporting Information S1): the EAR South group contains samples within 0°N-10°N and 30°E to 45°E; the EAR North group contains samples within 10°N-20°N and 38°E to 45°E; the Arabian group contains samples within 20°N-30°N and 35°E to 45°E; the Jordan/ Dead Sea group contains samples within 30°N-35°N and 30°E to 45°E; and the Anatolian group contains samples within 35°N-43°N and 30°E to 45°E. Based on the filtered samples, both the Anatolian group and the Jordan/Dead Sea group have Pb isotope ratios similar to those observed in the EAR system (Figure 6c and Figure S8a in Supporting Information S1), similar to what was found in Faccenna et al. (2013). A few Arabian samples have slightly more radiogenic 206 Pb/ 204 Pb, but the values are still within the 95% confidence range of samples from the southern EAR, while overlapping the Anatolian 95% confidence range.
In summary, although they span a wide range of values, the overlapping of Sr, Nd and Pb isotopes values from different locations along the transport channel and the data from Anatolia since 10 Ma is consistent with a common mantle source. This result, and the shift in isotope ratios in Anatolia at the 10 Ma Eurasia-Arabia collision (when mantle from the EAR region could more easily reach Anatolia), do not contradict the hypothesis that asthenospheric mantle beneath Anatolia is derived from a low velocity channel originating from the EAR upper mantle.
One potential issue is the basalts from the Red Sea, which contain isotopic signatures (Eissen et al., 1989;Volker et al., 1993) that are more similar to normal mid-ocean ridges, and differ from nearby Arabian volcanism (Bertrand et al., 2003;Moufti et al., 2012). However, while the low seismic velocity channel appears beneath the Red Sea in global tomographic models (Figure 5a and Figure S5 in Supporting Information S1), in higher resolution regional seismic models that incorporate Arabian seismic arrays, the channel is horizontally offset from the Red Sea (Chang et al., 2011;Koulakov et al., 2016;Wei et al., 2019;Yao et al., 2017). For example, based on the 100 s Rayleigh wave phase velocity tomography that is most sensitive to the shear velocity structure at ∼120 km depth (Yao et al., 2017), the channel lies east of the Red Sea (Figure 6a), consistent with the distinctive isotopic signatures of the channel versus the Red Sea, particularly in the middle latitudes of the Red Sea (Eissen et al., 1989;Volker et al., 1993). Nonetheless, because these high-resolution models do not span the entire length of the potential channel (Yao et al., 2017) or depend on body waves (Chang et al., 2011;Koulakov et al., 2016;Wei et al., 2019) with potential loss of vertical resolution, the following analyses are primarily based on larger-scale models.

Thermal Evolution During Asthenospheric Transport
If the plume rising beneath the EAR region is the main heat source for material flowing toward Anatolia, T P values should not increase along the proposed path of upper mantle transport from the EAR toward Anatolia. To estimate T P , basaltic primary magma equilibration P-T conditions for 771 volcanic rock samples with major element abundance measurements from the GEOROC database (Data Set S1) along the channel ( Figure S7f in Supporting Information S1) were calculated (Lee et al., 2009) with the same method and parameters that were used for the Karacadag samples (Figure 7). The samples were divided into the same geographic groups that were used for the isotopes applying the same age requirements. The culling operation based on major element composition that was used for Karacadag samples for the T P modeling was also applied to these samples, and we assumed the same magma water content (1 wt.%) and oxygen fugacity (Fe 3+ /Fe Total = 0.16). Since the assumed water content in the melt influences the inferred T P (e.g., Figure 3a), to justify the 1 wt.% assumption, the Ce abundances of the samples along the channel were analyzed ( Figure S9a in Supporting Information S1). These values demonstrate that 1 wt.% is a good estimate of water content. 1 wt.% also agrees with the water abundance from olivine hosted melt inclusions from samples in Ethiopia and Eritrea (Donovan et al., 2018;Field et al., 2012;Iddon & Edmonds, 2020).
To ensure that melt equilibration P-T values represent the asthenosphere, we estimated LAB depths for each geographic group by analyzing the distribution of melt equilibration depths (Figure 7), and defining the LAB as the depth where the number of samples in the 10 km beneath it most outnumbers the number of samples in the 10 km above it (similar to what we did for the Karacadag samples in Text S3 in Supporting Information S1). Because samples in each group are not from the same volcanic field and LAB depths for different sample locations may differ, to avoid samples from regions with deeper LAB and thus affected strongly by the lithosphere, we only used samples with equilibration depths more than 15 km below the estimated overall LAB depth for each group to estimate mantle T P (Figure 7).
Equilibration P-T conditions were again assumed to represent the geotherm, and for samples at depths that are not strongly affected by lithospheric conduction, the corresponding adiabatic geotherms and T P values were estimated through a grid search of mantle adiabats with T P from 1,300°C to 1,550°C with an interval of 5°C, following the approach in Section 2.3 (Figure 7, lower inset). Samples from the two EAR groups, the Arabian Peninsula and the Jordan/Dead Sea region produced distributions of T P values that overlap the T P values from Anatolia (Figure 7). Some samples from the Anatolian group are clustered at shallower depths, but as noticed before (McNab et al., 2018), these central Anatolian samples share high K/Nb and U/Nb values, indicating the potential influence of a different geodynamic process that involves water (Figures S9b and S9c in Supporting Information S1). The distribution of T P values from Anatolia has more samples at slightly higher values than the other regions (Figure 7), but given potential uncertainties in water content, these minor temperature differences are not significant. We also tested the effect of only including the 394 samples with normalized MgO higher than 9 wt.% ( Figure S9d in Supporting Information S1) instead of 8 wt.% (Figure 7), but the main conclusions about T P do not change.
The relatively constant T P values among the different geographic groups along the potential channel are consistent with upper mantle flow from the EAR to Anatolia. However, they also require that only a limited amount of conductive cooling occurs within the channel during the transport, despite the ∼2,000-3,000 km distance.
To test whether or not Eifel could be the heat source for the Anatolian mantle, we also analyzed samples from the Carpathian region within 42°N-50°N and 14°E to 30°E. If upper mantle was flowing from the Eifel region to Anatolia, along the weaker low velocity channel that connects these zones, it would pass through Carpathian-Pannonian volcanic fields (Figure 5a and Figure S7 in Supporting Information S1). Although recent Carpathian samples show isotopic signatures similar to Anatolian samples ( Figure S8 in Supporting Information S1), consistent with the same mantle source, their T P values are significantly lower than the basalts erupted on the EAR-Anatolia channel (Figure 7). This result argues against flow from the Eifel region to Anatolia, because the mantle T P trend would require an additional heat source en route. If flow from the EAR to Anatolia continues on to the Carpathians, the lower Carpathian T P would suggest that the effects of cooling become more obvious at these longer transport distances.

A Potential Mechanism for the Long-Distance Transport
As has been demonstrated in the previous sections, data from different regions along the potential transport channel from the EAR to Anatolia are consistent with a shared mantle source and similar mantle temperatures. These observations can be explained if asthenospheric material can flow laterally over 2,000 km distance without much cooling. In this model, as shown in Figure 8, hot mantle materials would rise upwards within the mantle plume beneath the EAR system, but after reaching the upper mantle, this hot material would flow toward Anatolia creating a low seismic velocity channel (Figure 6a) that lies beneath the western edge of the Arabian Peninsula, the eastern edge of the Mediterranean Sea, then finally reaching Anatolia.
The preservation of high mantle temperatures during transport along this asthenospheric channel from the EAR to Anatolia may seem surprising. Since northward plate motion along the channel since 20 Ma is only 2-5 cm/yr (Seton et al., 2012), the asthenosphere driven by plate motions would move slowly enough that it would likely have lost significant heat before reaching Anatolia. However, flow driven by lateral pressure gradients acting on the low-viscosity asthenosphere in the channel due to buoyancy forces provides an alternative mechanism.
Here, we develop a 1D model of northward pressure-driven asthenospheric flow to test whether flow rates can be high enough to explain the relatively constant T P values from the EAR to Anatolia. The presence and geometry of a low viscosity athenospheric channel are inferred from the seismic velocity profiles shown in Figures 6 and 8, and we assume that the channel has already been completely developed, as supported by various seismic tomography models (Figure 5a and Figure S5 in Supporting Information S1), and that flow is in steady-state.

Pressure Gradients Between the EAR and Anatolia
Pressure-driven flow in the direction from the EAR to Anatolia could be viewed in two ways. From the global mantle convection perspective, whole mantle buoyant flows that originate from large-scale density variations (over the whole depth of the mantle) could result in a horizontal pressure gradient within the asthenosphere. For the region of this study, buoyancy forces driving the upwelling mantle plume beneath the EAR would produce a horizontal pressure gradient that pushes hot, low-viscosity material toward Anatolia, which would be enhanced by the pull from the descending Hellenic slab (Faccenna et al., 2013). Such a northward gradient has been demonstrated in global flow models (Conrad & Behn, 2010;Natarov & Conrad, 2012), and is especially strong close to the Afar region (Figure 6a). The resulting northward flow is also consistent with azimuthal anisotropy in the mantle measured by shear-wave splitting (Qaysi et al., 2018) (Figure 6a) and body wave tomography (Wei et al., 2019). From a regional observational perspective, a horizontal asthenospheric pressure gradient could be caused by variations in the thickness of hot asthenospheric material and surface uplift to maintain overall isostatic balance. As plume material upwells beneath the EAR system, it generates uplift, expressed in both the absolute topography (Amante & Eakins, 2009) and the residual topography (Yang & Yang, 2021) which decrease from the EAR to lower values before reaching Anatolia. This upwelling and uplift would produce an excess in upper mantle pressure which pushes hot mantle material away from the surface expression of the plume (e.g., Bercovici & Lin, 1996;Olson, 1990). The thickness of the buoyant plume-derived asthenospheric layer would decrease in the direction of mantle flow, as corroborated by seismic velocity structure (French & Romanowicz, 2015) (Figure 8). The northward gradient in the thickness of the channel and its accompanying surface topography would produce a northward pressure gradient. Hence, by evaluating the thickness variation of hot asthenopheric materials from EAR to Anatolia, we could obtain the pressure gradient caused by the thinning of this low-viscosity, low-density asthenospheric layer.
In the simple model of Olson (1990), mantle flow at the top of the plume could spread to form a disk of asthenosphere. Alternatively, radial outflow at the top of the mantle plume could be concentrated in relatively narrow fingers due to the Saffman-Taylor instability (Schoonman et al., 2017), instead of distributed uniformly and radially in all directions. Fingering could occur because the upwelling mantle would have a much lower viscosity than the ambient mantle due to its higher temperature. For example, as shown in Figure 5a, while the seismic velocities suggest that hot upper mantle materials are concentrated along the channel from the EAR to Anatolia, hot mantle is also evident along the Gulf of Aden, and a potential finger extends westward as well. The development of the northward finger could also have been initially aided by the opening of the Red Sea (Schoonman et al., 2017), although the Red Sea spreading center may have rotated and is not currently overlying the channel (Chang et al., 2011). Hot material from the plume could thus flow in other than a purely northward direction with a geometry that is even time-dependent. However, our goal in the following discussion is to simply estimate the northward component of flow rate by balancing viscous shear stresses on horizontal planes with the northerly pressure gradient described below. Therefore, the exact geometry of low viscosity fingers and potential east-west horizontal inflows and outflows from/to the channel will not affect the following interpretation, and we are able 10.1029/2022GC010605 14 of 22 to ignore flow perpendicular to the channel and treat the flow as 1D to the north. To further validate the 1D flow assumption, we also checked asthenospheric pressure gradients from the global mantle flow model of Conrad and Behn (2010), which also indicates a consistent northward pressure-driven flow in region of the EAR-Anatolia channel (Figure 6a).
Asthenospheric pressure gradients and resulting flow velocities between the EAR and Anatolia were estimated from large-scale buoyant flow in a global flow model (Conrad and Behn 2010) and from observed regional asthenosphere thickness variations (Figures 9a and 9b; Details in Text S7 in Supporting Information S1). The two approaches result in similar overall pressure gradients of 6.7 and 6.9 Pa/m. Although estimated differently, the pressure gradients from large-scale buoyant flow and asthenospheric channel thickness variation could play a similar role in pushing upwelled material away from the plume. For the global model of Conrad and Behn (2010), a rigid boundary condition is imposed at the surface, which provides the required horizontal mantle pressure gradients for the channel flow. However, if a more realistic traction-free boundary condition is assumed at the surface, and both lithospheric uplift and the asthenospheric channel are included, then the variations in topography and channel thickness would contribute to the pressure gradients. Hence, the pressure gradients derived from these two estimates could act similarly to drive horizontal asthenospheric flow. However, because density variations at the depths of the channel are not included in Conrad and Figure 9. Thermal evolution during mantle transport from the East African Rift (EAR) to Anatolia. (a) Relative pressure gradient with respect to the transport end point in Anatolia at 1,150 km (left dotted line) from Conrad and Behn (2010). The original value is shown by the blue line, and the smoothed gradient averaged over length-scales of ±556 km is shown by the red line. The representative pressure drop (magenta dots) obtained from the starting point in the EAR at 3,750 km (right dotted line) and the end point. The transport length and estimated values for pressure gradient are labeled in gray. (b) Similar to (a) but for the pressure gradient estimate based on the maximum channel depth. The excursion in the blue curve at ∼2,000 km is due to the relatively weak low velocity anomaly in the reference model (Figure 8), which makes it unable to resolve a pick for maximum depth (or leading to a pick at shallower depths). (c) The estimated transport flow rate based on the stress-and-strain rate-dependent viscosity (red solid line) and its average value between 75 and 200 km depths (red dashed line) with the scale at the bottom. The pressure gradient profile (blue line) corresponds to the scale at the top. (d) The initial geotherm at the EAR before transport (solid blue line); cooled geotherms after transport to Arabia for ∼4 Myr (blue dotted line); cooled geotherms after transport to Jordan for ∼8 Myr (blue dashed line); and cooled geotherms after transport to Anatolia for ∼10.74 Myr (blue dotted-dashed line). The initial geotherm is constructed from the EAR adiabat (yellow line) estimated from deep EAR primary magma equilibration conditions (dots). The red solid line shows the dry solidus. The black dashed line shows the lithosphere-asthenosphere boundary depth defined by the north EAR samples (upper melting limit; Figure 7). The effective viscosity (shear stress divided by strain rate) profile for the initial geotherm is shown by the green solid line with its scale labeled at the top of the panel, and its averaged log-scale value between the depths of two black dotted lines shown by the green dashed line. The high viscosity at ∼150 km depth is due to the stress dependence of dislocation creep. Behn (2010), and pressure estimates based on channel thickness do not include dynamic pressures, it is difficult to judge whether pressure gradients estimated in these two ways are equivalent. In any case, we regard these two values as independent pressure gradient estimates, and because their values are similar (0.30 Pa/m difference), we use an intermediate value (6.8 Pa/m) as the reference pressure gradient in the following steps of the 1D flow calculation.
Based on the EAR-Anatolia pressure gradient constraints, we constructed a conservative depth profile for the pressure gradient (Figure 9c). In the profile, non-zero pressure gradients only exist between depths of 100 and 200 km. In the real world, the gradient could extends to deeper depths, but here we chose, conservatively, to make the lower boundary above the base of the thinnest channel as defined by the shear-wave velocities in Anatolia (Figure 8), since horizontal pressure gradients originating from asthenospheric disk thickness variation exist within the channel (Olson, 1990). The maximum pressure gradient is the EAR-Anatolia estimate of 6.8 Pa/m, and pressure is represented by a cosine shape function (Figure 9c), so the average pressure gradient within the channel is only half of the maximum. This conservative approach is supported by the findings of Natarov and Conrad (2012), who found that the pressure gradient inferred from the asthenospheric infinite strain axis is smaller than the values directly taken from the flow model (Conrad & Behn, 2010). In addition, if the temperature difference between the mantle inside and outside of the channel is less than 100°C, the estimated pressure gradient from the second approach would also be smaller.

Thermal Evolution Along the Channel
The initial EAR mantle geotherm was first estimated based on primary magma equilibration P-T conditions for magma samples from the two EAR groups with equilibration depths below 90 km (Figure 9d). Only these deep samples were employed since these depths are less influenced by the lithosphere and the less well-constrained latent heat of melting. The corresponding EAR adiabat (Figure 9d) was obtained by assuming a mantle water content of 100 wt. ppm (as found in Karacadag) and by searching for the best-fitting mantle T P with the grid search approach similar to that used in Section 2.3 that minimizes ‖ sample − geo‖2 . The resulting EAR T P value is 1,408°C.
We used a semi-analytical method to solve for the depth distribution of flow rate, assuming a stress-dependent mantle rheology that considers both dislocation and diffusion creep (Hirth & Kohlstedt, 2003) and calibrations of FTIR data in olivine (Bell et al., 2003). The details of the rheology and flow calculations are in Text S8 in Supporting Information S1. With the stress and strain rate dependence of viscosity for dislocation creep, at ∼150 km depth, a high effective viscosity is reached due to the approximately zero strain rate and deviatoric stress, with the result that most deformation is accommodated by diffusion creep at these depths ( Figure 9d). However, due to the small strain rate (Figure 9d), these depths do not strongly affect the flow rate. Flow rate is largely controlled by the lower viscosities in the high strain rate regions around the channel boundaries. Similar behavior has also been revealed in numerical modeling that incorporates stress-dependent rheology (e.g., Semple & Lenardic, 2018).
The resulting asthenospheric flow as a function of depth is shown in Figure 9c. Beneath the channel (>200 km), higher viscosity mantle is dragged by flow in the overlying channel at a relatively low rate, but between the overall LAB depth from Anatolian basalt samples (75 km, Figure 7) and the lower boundary of the channel in Anatolia (200 km, Figure 8), the average flow rate is 24.2 cm/yr. This rate, while high, is comparable to asthenospheric mantle flow rates estimated from geodynamic modeling when low viscosity and a northward channel are prescribed beneath the Arabian Plate (Faccenna et al., 2013). In addition, the flow rate is also similar to upper mantle flow velocities estimated in the South Pacific (Ballmer et al., 2013), within the range estimated near North Atlantic Ridges (Parnell-Turner et al., 2017), and lower than the estimated rates in another North Atlantic study (Hartley et al., 2011). With a 24.2 cm/yr asthenospheric flow rate, mantle materials would travel the 2,600 km distance from the EAR to Anatolia in 10.74 Myr.
Given the estimated transport time from the EAR to Anatolia, we modeled thermal evolution within the channel to evaluate whether cooling in the channel is small enough to be consistent with the minimal variation among observed mantle potential temperatures along the transport path. For the initial geotherm at the beginning of the transport in EAR, a conductive geotherm for the lithosphere was used to link the adiabatic temperature in the asthenosphere to 0°C at the surface. We also required that this geotherm crossed the dry solidus at 53 km depth in accordance with the LAB (upper melting limit) inferred from the basalt samples in the northern EAR group (Figure 7). At depths below 200 km, a 100°C temperature decrease from the adiabat defined by the EAR deep samples was imposed so that the mantle beneath the channel is at a typical ambient mantle temperature (purple line in Figure 9d). Details for how the initial geotherm was parameterized appear in Text S8 in Supporting Information S1.
To estimate cooling along the transport path, the initial geotherm was allowed to cool over time intervals corresponding to the transport times to the Arabian volcanic fields (∼4 Myr), the Jordan/Dead Sea volcanic fields (∼8 Myr) and to Anatolia (near Karacadag; 10.74 Myr) (Figure 9d). While the shallower part of the geotherm cools significantly over these time-scales, temperatures at depths greater than 75 km are largely unchanged, especially for depths over 100 km. Since the mantle T P values are based on the basalt samples with equilibration depths greater than 75 km, the cooling models explain how similar mantle T P values can be observed in the different regions along the EAR-Anatolia transport channel.
In this modeling, cooling across the lateral sides of the channel is not accounted for. However, as constrained by the high lateral resolution surface wave tomography (Figure 6a), the channel likely has a width of ∼300 km (three times its thickness in depth). This geometry, combined with the larger vertical temperature gradient outside of the channel, indicates that heat loss from the top of the channel would be greater than through its sides. In addition, as shown across the channel bottom in Figure 9d, ambient mantle cooling only penetrates 25 km into the channel during transport. Applying this result as an upper bound to cooling at the sides of the channel, the channel would still be ∼250 km wide after the transport. These arguments support the validity of 1D cooling for this problem. The rapid transport and lithospheric thickness variations might also introduce small-scale convection and shear-driven upwelling that could promote decompression melting and affect mantle temperatures (Ballmer et al., 2013;Conrad et al., 2011;Raddick et al., 2002). In this study, we consider only the consequences of flow that would be driven by lateral pressure gradients that are directly indicated by large-scale seismic or mantle flow models (e.g., Conrad & Behn, 2010;French & Romanowicz, 2015).
To further test the cooling during channel flow model, we compared the LAB depths inferred from the basalts (Figure 7, upper inset) with LAB depths predicted by the cooling. In the cooling model, as lithospheric conduction and cooling progress, the upper intersection of the geotherm and the dry solidus moves to deeper depths ( Figure 9d). The shallowest depth at which primary magmas could equilibrate thus gets deeper as well. This prediction (red line in Figure 10a) is broadly matched by the deepening of the basalt-inferred LAB (Figure 10a) along the channel from the northern EAR to Anatolia. The uncertainty of the basalt-inferred LAB was obtained by bootstrapping the basalt samples in each group that were used for determining that LAB depth (Figure 10a). The overall trends in predicted and observed depths are similar, and given uncertainties in the initial geotherm (which depends strongly on the assumed water content in the melt), local tectonic histories, and calculation of pressures of last equilibration, misfits at specific cooling times can easily be accounted for. The overall match between predicted and observed depths provides support that the cooling model is plausible for transport along the EAR-Anatolia channel.
We further tested rates of cooling in the channel for a range of viscosities (details in Text S9 in Supporting Information S1). As long as the average channel viscosity is less than 5 × 10 18 Pa‧s (viscosity is ∼2 × 10 18 Pa‧s for the case in Figure 9d), the cooling is small enough to be consistent with the observed T P distributions (Figure 10b).
Compared with published mantle viscosities estimated with different methods, the viscosities assumed in the flow rate calculations at or above 5 × 10 17 Pa‧s are reasonable. They are higher than the lower bound for asthenospheric viscosity of 5 × 10 17 Pa‧s constrained by seasonal deformation (Chanard et al., 2018), within the range of the mantle viscosity suggested from postseismic deformation beneath Indian Ocean as (0.5-10) × 10 18 Pa‧s (Hu et al., 2016), and lower than the upper bound for asthenospheric viscosity of 10 19 Pa‧s estimated from the geoid variation across an ocean basin fracture zone (Craig & McKenzie, 1986). However, we also recognize that the different time-scales of these processes need to be taken into account when making viscosity comparisons (e.g., Lau et al., 2021).

Conclusions and Implications
In this study, we demonstrated that a wide range of seismic and geochemical observations are consistent with a model in which hot asthenosphere beneath Anatolia is fed by long-distance lateral transport of upper mantle from East Africa, with the lateral flow being driven by pressure gradients created by the buoyancy of the African mantle plume. These conclusions are based on a series of analyses.
1. We determined the mantle potential temperature beneath the Karacadag volcanic field in eastern Anatolia to be ∼1,420°C, higher than typical of ambient mantle, based on joint modeling of primary magma equilibration P-T conditions from basalt samples and a seismic velocity gradient imaged by Sp receiver functions that is consistent with the onset of partial melting. 2. Tomographic models show that no obvious deep mantle plume exists beneath Anatolia, but they consistently show a channel of low seismic velocity upper mantle that connects Anatolia and the East African Rift, as has been proposed by prior studies (Ershov & Nikishin, 2004;Faccenna et al., 2013;Wei et al., 2019). 3. In a chain of magmatic fields along the proposed transport path from the East African Rift system to Anatolia, we assessed available radiogenic isotope ratios in basalt samples, including 143 Nd/ 144 Nd, 87 Sr/ 86 Sr, 208 Pb/ 204 Pb, 207 Pb/ 204 Pb, and 206 Pb/ 204 Pb, and found they are consistent with a shared mantle source. 4. To evaluate thermal evolution along the transport channel, we analyzed primary magma equilibration P-T conditions from the basalts, and found that any decreases in mantle potential temperature are minimal, indicating little cooling within the asthenospheric channel. 5. To assess whether the observed mantle potential temperatures are physically consistent with flow over thousands of kilometers in the channel, we modeled potential mantle transport time from East Africa to Anatolia for pressure-driven flow, created by the African plume, and a stress-dependent rheology. With conservative assumptions, asthenospheric mantle from East Africa could reach Anatolia in 11 Myr. By modeling conductive cooling of the geotherm for ∼11 Myr, we found that while the shallowest asthenosphere is significantly cooled, leading to an increase in the depth of the LAB that is matched by basalt sample constraints, temperature changes in the deeper asthenosphere are small, consistent with the observed elevated mantle potential temperatures all along the channel. In addition to the horizontal pressure gradient provided by the mantle plume, another factor needed to enable this type of rapid asthenospheric flow is a low viscosity, as would be expected for the elevated temperature of plume-affected mantle.
The type of rapid pressure-driven asthenospheric flow over length-scales of thousands of kilometers supported by this study (Figure 8) has important implications for heat and material transport in the Earth. The 2,000-3,000 km length-scale from East Africa to Anatolia exceeds some previously inferred cases of long-distance transport of hot plume materials from their plume roots (Ebinger & Sleep, 1998;Schoonman et al., 2017), and rivals others (Ballmer et al., 2013). Rapid asthenospheric flow from plumes and the preservation of high mantle temperatures that it enables help to explain massive Large Igneous Provinces earlier in Earth history, such as the Central Atlantic Magmatic Province (Marzoli et al., 1999) and in the Pacific (Madrigal et al., 2016;Stern et al., 2020). While the inferred rates of asthenospheric flow from the EAR to Anatolia are higher than regional plate velocities, they are actually comparable to the pressure-driven asthenospheric flow velocities implied by the age-progression of volcanic chains from the Pacific superswell toward the southern East Pacific Rise (Ballmer et al., 2013) and in the North Atlantic (Hartley et al., 2011;Parnell-Turner et al., 2017). Mantle flow driven by plume-related pressure gradients are relevant to understanding the distribution of magmatism on terrestrial planets such as Venus which have plume-like upwellings (Smrekar et al., 2010) in the absence of plate tectonics. Overall, the results of this study indicate that high temperature plume-influenced materials can reach much farther in the upper mantle and over larger volumes of the Earth than commonly assumed.

Data Availability Statement
The Sp receiver function common-conversion point stack is from Hua, Fischer, Wu, and Blom (2020). Geochemical and petrological data were downloaded from GEOROC (https://georoc.eu/georoc), and are included in the Supporting Information S1 as follows: data related to primary magma equilibration conditions are provided in Data Set S1; isotope data are provided in Data Set S2; and the original sources for data in Data Sets S1 and S2 are provided in Data Set S3. Calculated primary magma equilibration conditions are based on Lee et al. (2009), and are also provided in Data Set S1. Velocity model SEMUCB_WM1 was downloaded from the Berkeley Global Seismology Group (http://seismo.berkeley.edu/wiki_br/Broad_plumes_rooted_at_the_base_of_the_mantle_ beneath_major_hotspots). Velocity model GLAD_M25 was obtained from E. Bozdag, the velocity model in Blom et al. (2020) from N. Blom, and the velocity model in Fichtner et al. (2013) from A. Fichtner; other velocity models are from the IRIS archive (https://ds.iris.edu/ds/products/emc-earthmodels/) with names provided in this paper (e.g., in the caption of Figure S5 in Supporting Information S1). All the figures are generated with the Generic Mapping Tools (Wessel et al., 2019).