Petrogenesis of Siletzia: The world’s youngest oceanic plateau

Siletzia is an accreted Palaeocene-Eocene Large Igneous Province, preserved in the northwest United States and southern Vancouver Island. Although previous workers have suggested that components of Siletzia were formed in tectonic settings including back arc basins, island arcs and ocean islands, more recent work has presented evidence for parts of Siletzia to have formed in response to partial melting of a mantle plume. In this paper, we integrate geochemical and geochronological data to investigate the petrogenetic evolution of the province. The major element geochemistry of the Siletzia lava flows is used to determine the compositions of the primary magmas of the province, as well as the conditions of mantle melting. These primary magmas are compositionally similar to modern Ocean Island and Mid-Ocean Ridge lavas. Geochemical modelling of these magmas indicates they predominantly evolved through fractional crystallisation of olivine, pyroxenes, plagioclase, spinel and apatite in shallow magma chambers, and experienced limited interaction with crustal components. Further modelling indicates that Siletzia magmatism was derived from anomalously hot mantle, consistent with an origin in a mantle plume. This plume has been suggested to have been the same as that responsible for magmatism within the Yellowstone Plateau. Trace element compositions of the most primitive Siletzia lavas are similar to suites associated with the Yellowstone Mantle Plume, suggesting that the two provinces were derived from compositionally similar sources. Radiogenic isotope systematics for Siletzia consistently overlap with some of the oldest suites of the Yellowstone Magmatic Province. Therefore, we suggest Siletzia and the Yellowstone Mantle Plume are part of the same, evolving mantle plume system. Our new geochronological data show the province was emplaced during the time when Eocene sea surface temperatures were their highest. The size of Siletzia makes the province a potential contributing factor to the biospheric perturbation observed in the early Eocene.


Introduction
Large Igneous Provinces (LIPs) represent large volume ( > 0.1 Mkm 3 ; frequently above > 1 Mkm 3 ), mainly mafic (-ultramafic) magmatic events that are not associated with 'normal' tectonic processes such as mid-ocean ridge spreading or subduction ( Bryan & Ernst, 2008 ;Bryan & Ferrari, 2013 ;Coffin & Eldholm, 1994 ), and are typically characterised by pulses of magmatism that last a maximum of a few tens of m.y. ( Bryan & Ferrari, 2013 ;Ernst & Bleeker, 2010 ). In oceanic settings, LIP magmatism often manifests itself as oceanic plateaus that comprise volcanic successions (which may be emplaced both subaerially and subaqueously), a plumbing system of sheets, sills and layered intrusions, and a crustal magmatic underplate ( Coffin & Eldholm, 1994 ;Kerr, 2014 ;Ernst et al., 2019 ). Accreted oceanic plateaus are commonly found at geological time, the source region is required to be fundamentally different to the ambient mantle, i.e. it must be either, hotter, more fertile, more volatile rich, or some combination of these factors (e.g., Campbell & Griffiths, 1990 ;Kerr, 2014 ). It has been determined that decompression partial melting of a dominantly peridotite mantle source with a higher than ambient T P (potential temperature, that is the temperature mantle material would have were it adiabatically decompressed to the surface of the earth ( McKenzie & Bickle, 1988 )) can produce the magmatic activity that characterise oceanic plateaus ( Fitton & Godard, 2004 ;Herzberg, 2004 ;Herzberg et al., 2007 ;Putirka, 2008 ).
Therefore, the mantle plume paradigm is currently the preferred genetic mechanism for the majority of LIPs (including oceanic plateaus) observed in the geological record (see reviews in Campbell, 2007 ;Mahoney, 2007 andErnst, 2014 ). The model predicts that the high volumes of magmatism associated with LIPs occurs in response to the melting (via decompression) of a cylindrical zone of anomalously hot mantle material, upwelling from significant depths in the mantle ( Campbell & Griffiths, 1990 ). The mantle plume model for LIPs is particularly relevant to those emplaced during the Cenozoic where seismic tomography has frequently detected anomalously low-velocity (presumably high temperature) regions in the mantle beneath the focus of surface volcanism downwards to some boundary layer within the mantle ( Campbell, 2007 ;Montelli et al., 2006 ;Wolfe et al., 1997 ).
In this contribution, we present 271 new whole-rock geochemical analyses from the Siletz Terrane (12 major element oxides and 33 trace elements). These analyses span each constituent suite of the Siletz Terrane, and crucially, represent the first consistent data set analysed by a single laboratory, for the entire magmatic system. We use these geochemical data to assess the petrogenesis of Siletzia, an accreted oceanic plateau composed of the constituent suites of the Siletz Terrane and the Grays River Volcanics, which together, are preserved in the Cascadia forearc region of Vancouver Island, Washington and Oregon in North America ( Fig. 1 ).
Aside from their petrological implications, it is important to understand the evolution of oceanic plateaus because the large volume eruptions over small time periods have environmental consequences (e.g., Kerr, 2005 ). Therefore, we also present 9 new Ar-Ar ages which further constrain the timing and duration of the Siletzia magmatism, and its potential relationship to the Palaeocene-Eocene Thermal Maximum (PETM).
Interpreted tectonic settings for Siletzia magmatic suites are varied. It has been argued that some of the mafic suites represent accreted island chains initially formed via hotspot volcanism ( Duncan, 1982 ), which itself may have been the initial manifestations of the Yellowstone plume ( Haeussler et al., 2003 ;Pyle et al., 2009 ;Wells et al., 2014 ). In contrast, others have suggested Siletzia was emplaced in a rift-related marginal basin ( Brandon, 2014 ;Wells et al., 1984 ), or formed by passive mantle upwelling through a slab window during subduction of a mid-ocean ridge (e.g., Babcock et al., 1992 ).
The new dataset presented here is used to assess the petrogenetic evolution of Siletzia and its constituent formations. In this paper we investigate the nature of the mantle sources and melting regime which produced the primary magmas of the province and use geochemical modelling to constrain the evolution of these primary magmas during their residence in the crust. A potential petrogenetic link between Siletzia and the Yellowstone Magmatic Province is explored, before, finally, we assess the potential effect the formation of Siletzia had on the biosphere.

Siletz terrane
The Siletz terrane represents an oceanic terrane which was accreted to western North America shortly after its formation in the Eocene ( Duncan, 1982 ;McCrory & Wilson, 2013 ;Snavely et al., 1968 ). The ter-rane is composed of 2.6 × 10 6 km 3 of predominately mafic pillow lavas, intrusions and subaerial flows ( Phillips et al., 2017 ;Trehu et al., 1994 ) which are preserved in three formations ( Fig. 1 ). The first of these is the Siletz River Volcanic Formation, which is found in Oregon and crops out near Tillamook and Roseburg. The lavas are mostly comprised of aphyric pillow basalts and massive flows which are interbedded with cherts, turbidites and siltstones ( Phillips et al., 2017 ;Snavely et al., 1968 ). The presence of tuffaceous flows, lapilli-rich horizons and oxidised flow tops indicate that part of the volcanic activity was subaerial ( Phillips et al., 2017 ;Snavely et al., 1968 ).
The second formation is the Crescent Formation which crops out in the Olympic Mountains and Willapa Hills of southern Washington ( Fig. 1 ) and is subdivided into a lower (mostly subaerial) unit and upper (mostly submarine) unit. The lower unit is dominated by pillow basalts with minor massive flows and basaltic breccias that conformably overlie siliciclastic submarine sequences ( Babcock et al., 1992 ) while the upper unit is primarily composed of massive basaltic flows, the tops of which are partially oxidised ( Babcock et al., 1992 ;Glassley, 1974 ).
The third formation is the Metchosin Igneous Complex which is limited in outcrop to the southern tip of Vancouver Island and consists of a lower unit comprising sheeted dykes, pillow basalts, massive flows and an overlying ~1000 m thick upper unit of subaerial massive sheet flows and lesser tuffs and basalt breccias ( Phillips et al., 2017 ). A number of gabbroic and dolerite bodies intrude all levels of the volcanic stratigraphy of the Metchosin Igneous Complex ( Timpa et al., 2005 ).
Detailed petrographic analysis carried out by Phillips et al. (2017) shows that the mineralogy of the mafic rocks is dominated by plagioclase, clinopyroxene and Fe-Ti oxide. The Siletz River Volcanics are the only Siletz Terrane rocks to contain abundant olivine (and pseudomorphs after olivine), while elsewhere, ophitic textures are common. Sub-solidus alteration of the Siletz Terrane rocks is limited, with the Crescent Formation and Siletz River Volcanics only characterised by zeolite to prehnite-pumpellyite facies assemblages. An exception to this is the Metchosin Igneous Complex and some Crescent Formation outcrops which display up to greenschist-facies alteration assemblages. Argon-Argon (Ar-Ar) and Potassium-Argon (K-Ar) has dated the igneous activity of the Siletz Terrane to have occurred between 56-49 Ma ( Duncan, 1982 ;Moothart, 1993 ). Recent work using U-Pb dating ( Eddy et al., 2017 ) has yielded ages of 51.2-50.9 Ma and 53.4-48.1 Ma for the Metchosin Igneous Complex and Crescent Formation respectively, while Haeussler et al. (2003) , has reported U-Pb ages of 53-50 Ma for the Siletz River Volcanic Formation.
Previous workers have shown that the Siletz Terrane lavas contain up to 15 wt. % MgO, with whole-rock major element compositions which range from picritic to rhyolitic (Appendix 1). The trace element signatures of the Siletz Terrane lavas are similarly varied, as shown by the Crescent Formation and Siletz River Volcanics which show moderate to extreme enrichments in the most incompatible elements relative to the least incompatible. The compositions are similar to those observed in modern day hotspot-related ocean island basalts ( Phillips et al., 2017 ;Rarey, 1985 ). In contrast, the basalts of the Metchosin Igneous Complex, Timpa et al. (2005) and Phillips et al. (2017) show trace element signatures characterised by depletions in the most incompatible elements relative to the least incompatible trace elements, and have compositions similar to modern day mid-ocean ridge basalts.
A plume source for the Siletz Terrane has been proposed by Pyle et al. (2009Pyle et al. ( , 2015, on the basis of Sr, Pb and Nd isotopic compositions which they argue are indistinguishable from parts of the Columbia River Basalt (CRB) Province -itself widely considered to be related to the Yellowstone mantle plume ( Brandon & Goles, 1988 ;Geist et al., 2002 ;Prestvik & Goles, 1985 ). This isotopic similarity is used as evidence of a probable cogenetic link between the Siletz Terrane and the CRB, with the Siletz Terrane representing melting of the nascent Yellowstone plume head, and the basalts of the CRB related to melting of the plume tail. However, more recent published isotopic analysis by Phillips et al. (2017) , reveals clear isotopic differences between the Siletz Terrane and the CRB that  Phillips et al., (2017) and McCory & Wilson (2013) ). Also shown are the locations of samples collected by this study for geochemical and geochronological analysis. Full sample location data available in Appendix 1.
are difficult to reconcile with a cogenetic origin for both the Siletz Terrane and CRB. Instead, Phillips et al. (2017) suggest that the variable Siletz Terrane trace element geochemistry and isotopic compositions that fall on mixing trends between HIMU, EM2 and depleted mantle reservoirs show that the Siletz magmatism is derived from a compositionally heterogeneous deep plume source, distinct from the source of the CRB.

Grays river volcanics
The Grays River Volcanics consist of a ~3500 m thick pile of lava flows, volcanoclastic rocks and interbedded sedimentary rocks, preserved within the Coast Range basalt province ( McCutcheon, 2003 ). The volcanic rocks comprise an upper and lower unit ( Kleibacker, 2001 ), with the lower unit being dominated by pillow basalts and hyaloclastites, while the upper unit is made up of subaerially erupted basaltic andesites and dacites. Ar-Ar dating reveals that the entirety of the Grays River Volcanics was emplaced between 40.1 and 36.8 Ma, with the lower basaltic unit emplaced between 40.1 and 39.0 Ma ( Kleibacker, 2001 ).
Previous geochemical analyses of the Grays River Volcanics show that the basalts of the lower unit are tholeiitic, range in MgO from 2.7 to 9.5 wt. %, and are characterised by a moderate enrichment in the Light Rare Earth Elements (LREE) relative to the Heavy Rare Earth Elements (HREE) and incompatible element enrichments similar to ocean island basalts ( Chan et al., 2012 ;Phillips et al., 1989 ). Aside from these trace element signatures, Chan et al. (2012) suggests that the low B/Be ratios and high 206 Pb/ 204 Pb ratios recorded by the Grays River Vol-canics are most likely to have been caused by partial melting of a deep mantle plume source that incorporated lesser amounts of depleted mantle components. Chan et al. (2012) and Pyle et al. (2009) propose that the Grays River Volcanics were derived from the nascent Yellowstone plume, which supplied the primary magmas of the suite.

Major elements and trace elements
Rock sample preparation and geochemical analysis was carried out at Cardiff University. Weathered surfaces, veins and alteration patches were removed from the rock samples using a diamond-bladed rock saw. Samples were then crushed to a coarse grit by a steel jaw crusher. Approximately 80 ml of each sample was reduced to a fine powder in an agate ball mill. Afterwards, approximately 2 g of each powdered sample was ignited for two hours in a furnace at 900 °C to drive off volatile substances and determine loss on ignition (LOI) values.
Samples in this study were prepared for Inductively Coupled Plasma (ICP) analysis using the lithium metaborate fusion method. To prepare the samples, 0.1 ± 0.001 g of each ignited sample was mixed with 0.6 ± 0.004 g of lithium metaborate flux in a platinum crucible. A few drops of lithium iodide wetting agent were added to each mixture which was then fused (at approximately 1000 °C) using the Claisse Fluxy automated fusion system. After the mixture was then dissolved in a 50 ml solution of 4% HNO 3 . After the mixture had fully dissolved, 1 ml of 100 ppm Rh spike was added to the solution which was then made up to 100 ml with 18.2 Ω deionised water obtained using a Milli-Q ® purification system. Approximately 20 ml of each solution was run on ICP-Optical Emission Spectrometry (OES) to obtain major element and some trace element abundances. An aliquot of 1 ml of each solution was added to 1 ml of In and Tl and 8 ml of 2% HNO 3 and run on the ICP-Mass Spectrometer (MS) to obtain trace element abundances. The instruments at Cardiff University used to analyse elemental compositions are a Jobin Yvon Horiba Ultima 2 ICP-OES and a Thermo Elemental X7 series ICP-MS. Full analytical results including repeat runs of standard basalt JB-1A and analyses of other international reference materials, can be found in Appendix 1. The vast majority of elements are within 2 standard deviations ( ) of certified values and those that are not in this range (e.g., Ni and Cu) are not included in the discussion.

Geochronology
Plagioclase separates were initially collected from samples via magnetic separation, followed by light leaching in nitric acid and hand picking. These separates and neutron flux monitors were placed in Cu or Al foil packets and stacked in quartz tubes. The relative positions of the packets were carefully measured for later reconstruction of neutron flux gradients. The sample package was irradiated in the McMaster University, Canada reactor, Cd-shielded facility. Fish Canyon Sanidine (FCs, 28.294 ± 0.036 Ma, Renne et al., 2011 ) was used to monitor 39 Ar production and establish neutron flux values (J) for the samples.
Gas was extracted from samples using an all-metal resistivelyheated furnace with multi-step heating schedules ranging from ~700 to 1500 °C. Liberated argon was purified of active gases (e.g, CO 2 , H 2 O, H 2 , N 2 , CH 4 ) using three Zr-Ti-Al getters; one at 25 °C and two at 400 °C. Data were collected on a GVi instruments ARGUS 5-collector mass spectrometer using a variable sensitivity faraday collector array in static collection (non-peak hopping) mode ( Mark et al., 2009 ;Sparks et al., 2008 ). Time-intensity data were regressed to t 0 with second-order polynomial fits to the data. Mass discrimination was monitored by comparison to running-average values of an air standard. Typical average total system blank for furnace extractions, measured between each sample run, are 2 × 10 − 15 mol 40 Ar, 2 × 10 − 17 mol 39 Ar, 3 × 10 − 17 mol 36 Ar over the temperature range where age information is extracted; these blank levels yielded average sample/blank ratios ranging from 10 to 2000. All data are blank and mass discrimination corrected. Plateaus are calculated with contiguous steps (minimum 5 steps and 50% 39 Ar released), indistinguishable in age at 2 uncertainty. Inverse isochrons are calculated with concordant plateau steps and used to test that trapped components are indistinguishable from the composition of modern air and that resultant age is indistinguishable from the plateau age at 2 uncertainty.

Geochemistry
The 271 new whole-rock major and trace element geochemical analyses presented here represent the single most extensive sampling of the Siletz Terrane to date, both in terms of geographic distribution and number of samples analysed. These new high-quality data have been integrated with older data from the literature, to facilitate for the first time, an assessment of the magmatic evolution of Siletzia as a whole. A representative selection of our new data, including the first trace element data collected from the basalts of the Siletz River Formation, are summarised in Table 1 and given in full in Appendix 1. Also included in Appendix 1 are the 493 individual whole-rock analyses from the constituent suites of the province reported by previous workers.
The total alkali vs. silica (TAS) diagram ( LeBas et al., 1986 ) classifies the Siletzia samples as a continuum of predominantly tholeiitic picrobasalts, basalts, basaltic andesites, andesites and dacites and rhyolites ( Fig. 2 A), with far fewer analyses plotting in the alkaline basalt, phono-tephrite and other regions. On the Zr/Ti vs Nb/Y diagram ( Pearce, 1996 ), which is less susceptible to sub-solidus elemental mobility at greenschist facies, exhibited by some of the Siletz samples ( Phillips et al., 2017 ), the majority of the Siletzia samples plot as an overlapping array between the basalt and alkali basalt field. On this diagram, only a small minority of samples from the Siletz River and Crescent Formations falling on trends towards rhyolitic and tephriphonolitic compositions ( Fig. 2 B). The new samples we present here tend to be more mafic than the literature analyses. This is a product of our sampling strategy which aimed to acquire more primitive samples that would better record mantle source characteristics.
Our new data, taken together with the published data, show that the most magnesian rocks of Siletzia belong to the Crescent Formation, which contain up to 15 wt. % MgO. The most magnesian lavas of the Siletz River formation contain ~14 wt. % MgO, while the most magnesian samples from the Metchosin Igneous Complex and Grays River Volcanics contain ~11 wt. % MgO and ~10 wt. % MgO respectively.
All suites show consistently increasing SiO 2 , Na 2 O and K 2 O with decreasing MgO, while CaO shows a uniform decrease with decreasing MgO. Al 2 O 3 shows no obvious correlation with MgO ( Fig. 3 ). FeO (T) trends for the province are harder to characterise, but seem to record a point of inflection whereby samples with > 4 wt. % MgO, record decreasing FeO (T) with increasing MgO, while samples with < 4 wt. % MgO record decreasing FeO (T) with decreasing MgO. These trends are widely interpreted (e.g., Peterson & Moore, 1984 ) to be the result of the initial fractionation of olivine, clinopyroxene and plagioclase from basaltic parent magmas, followed by subsequent fractionation of Fe-Ti oxides. This potential fractionation history is in agreement with published petrographic analyses ( Phillips et al., 2017 ) which shows that the mineralogy of the mafic rocks of the Siletz Terrane (as focussed on here) is dominated by plagioclase, clinopyroxene and Fe-Ti oxide.
Although the major element bivariate trends are diffuse, this relative scatter cannot be ascribed to element remobilisation during metamorphism, which, for the Siletzia suites is of insufficient grade to have remobilised many of the elements displayed in Fig. 3 ( Phillips et al., 2017 ). Instead, it is likely that some of this apparent scatter within the suites, is the result of their derivation from multiple parent magmas, potentially each with its own subtly different melting and fractionation history. This might be indicated by the bivariate plots involving MgO and CaO and  Fig. 3 C and F). As will be investigated in more detail later, this may indicate that these suites are actually composite suites made up of the products of multiple magmatic systems which experienced different petrogenetic histories.
All of the Siletzia suites show positive correlations between MgO and elements that are compatible during normal basaltic differentiation such as Cr, Ni and Sc ( Fig. 4 ). Incompatible element (La, Sm, Y, Nb) vs. MgO trends all record negative correlations, but for the Siletz River and Crescent Formations, the data show a large degree of scatter, with several samples plotting outside of the main trends defined by the bulk of the samples. Incompatible element vs. Zr trends ( Fig. 5 ) show a similar pattern. Samples from the Grays River Volcanics and Metchosin Igneous Complex plot on linear arrays, while a number of samples from the Siletz River and Crescent formation plot as a diffuse scatter away from the main trend. It is unlikely that this scattering is the result of elemental remobilisation during metamorphism of the Crescent and Siletz River formations as the prehnite-pumpellyite-greenschist metamorphic grade for these suites ( Phillips et al., 2017 ) is usually insufficient to have remobilised the high field strength elements, including the Rare Earth Elements ( Pearce, 1996 ). Instead, the scatter observed in Fig. 5 for the Crescent and Siletz River formations is potentially due to the formations being derived from a number of primary magmas, which is consistent with the bifurcation of major element trends highlighted in Fig The incompatible trace element systematics of the Siletzia suites vary between the constituent suites. The majority of samples from the Siletz River, Crescent Formation and Grays River Volcanics plot as overlapping clusters similar in composition to Oceanic Plateau Basalts ( Fig. 6 ). Fewer samples from these three suites classify as Oceanic Island Basalts according the scheme of Condie (2005) , while only several samples from each classify as Mid Ocean Ridge or Volcanic Arc Basalts. Using the same classification, the Grays River Volcanics show a slightly smaller compositional variability, plotting almost entirely as Oceanic Island Basalts. According to the scheme of Pearce (2008) , the vast majority of samples plot within the mantle array between Normal Mid Ocean Ridge and Ocean Island Basalts. Only very few samples from the Crescent Formation and Siletz River Volcanics plot on vectors towards Upper Continental Crust, which may indicate that contamination processes operated during the petrogenesis of some of these suites.

Geochronology
In total, 14 basalt samples were selected for Ar-Ar dating on the basis of petrographic freshness. Of these samples, 8 yielded acceptable step-heating plateau ages. All ages were obtained on plagioclase separates except sample 116 for which the age was obtained from the groundmass. These data are discussed below, summarised in Table 2 and the stepheating plateau experiments and isochron plots are given in Appendix 2.
Sample 450, from the Black Hills Group, yielded an Ar-Ar plateau age of 34.16 ± 0.40 Ma which is younger than the 56-51 Ma K-Ar reported by Duncan (1982) . This sample is tholeiitic and similar in elemental and isotopic composition to other basalts of the Siletz terrane ( Phillips et al., 2017 ). This age is similar to those from the Cascade Head Volcanics to the south that have been dated to 34.2 Ma (Ar-Ar whole rock plateau, Wells et al., 2014 ) and the Upper Grays River Volcanics at 36.8 ± 0.5 Ma. Hirsch and Babcock (2009) also note a similar age for a sample further to the north, yielding an Ar-Ar plateau age of 31.2 ± 0.2 Ma, originally thought to be correlated with the Crescent Fm.
Basalt sample 441 is from near the base of the upper Crescent Formation in the south of the Main Crescent sequence and has been determined to have an Ar-Ar plateau age of 55.96 ± 0.92 Ma. This age is indistinguishable from published whole rock 40 Ar/ 39 Ar ages for these basalts of 51.0 ± 4.7 Ma ( Hirsch & Babcock, 2009 ), but slightly older than calcareous nanoplankton ages of 53.5-49 Ma, determined for interbedded marine mudstones ( Bukry & Snavely, 1988 ).
Samples 390 and 389 from the Doty Hills Group of basalts, part of the Crescent formation in southern Washington, yield Ar-Ar plateau ages of 47.62 ± 0.28 and 49.49 ± 0.40 Ma respectively. While there are no published ages for this particular region of the Crescent Formation, the Willapa Hills area ~15 km to the south have 40 Ar/ 39 Ar ages ranging from ~54 Ma (at the base of the lava sequence) to 48.7 ± 0.5 Ma (at the top of the lava sequence) ( Duncan, 1982 ;Wells et al., 2014 ) suggesting that the Doty Hills basalts are correlative with late stage Crescent Fm. in this region.
Sample 116 is a basalt from the Siletz River type locality in Oregon and has an Ar-Ar plateau age of 53.35 ± 0.24 Ma. This is similar to the published ages of 53.5-49 Ma (coccolith zones CP10 and CP11) ( Bukry & Snavely, 1988 ) and 53.3-49.5 Ma ( 40 Ar/ 39 Ar step heating ages) (reported by Pyle et al., 2009 ). Sample 162, a similar basalt from the Siletz River Formation type locality, has yielded an Ar-Ar plateau age of 58.84 ± 5.21 Ma which is nearly indistinguishable from sample 116 and the ages reported by Pyle et al. (2009) .
Basalt sample 103 from the Corvallis area, Oregon is part of the Siletz River Volcanic Formation. While there are no specific ages available for the Corvallis area, the ages are likely to be similar to those discussed above for the Siletz River type locality. The Ar-Ar plateau age determined for this sample is 51.48 ± 0.48 Ma, which is again consistent with the published ages.  ( LeBas et al., 1986 ) with data calculated on 100% anhydrous basis -A, and the Zr/Ti vs. Nb.Yb diagram ( Pearce, 1996 ) -B. Alkaline-Tholeiite line in A from Irvine & Baragar (1971) . GRV -Grays River Volcanics, MIC -Metchosin Igneous Complex, CFM -Crescent Formation, SRV -Siletz River Volcanic Formation. * denotes data sourced from literature. Coloured symbols denote new data. White symbols are literature data. Coloured symbols are new data.
Finally, sample 073 is a phyric basalt from the Kentucky Slough Group in southern Oregon and is classified as part of the Roseburg area of the Siletz River Formation in this study. While there are no specific ages published for this Group, the basalts in the Roseburg Group have been dated to be between 56 and 53.3 Ma ( 40 Ar/ 39 Ar and calcareous nanoplankton) ( Pyle et al., 2009 ). The Ar-Ar plateau age for this sample is 48.55 ± 0.29 Ma, which is significantly younger than those of the Roseburg basalts, but consistent with our younger ages in northern parts of the terrane, e.g., samples 389 and 390, Grays River Volcanics analyses in Duncan (1982) , and a dolerite from Willapa Hills with an Ar-Ar plateau of 48.7 ± 0.5 Ma ( Wells et al. 2014 ).

Geochemical selection and screening
Petrogenetic modelling software can be used to calculate primary magma compositions, which could feasibly fractionate to form an observed lava composition. One such piece of software, PRIMELT3 ( Herzberg & Asimow, 2015 ) incrementally adds or subtracts olivine from an observed lava composition, while simultaneously modelling melting of fertile mantle peridotite. This hybrid forward and inverse modelling (which assumes olivine is the only phase to have fractionated from the magma) continues until PRIMELT3 computes a melt fraction capable of: (1) being formed by partial melting of mantle peridotite, and; (2) replicating the major element composition of the observed lava through the accumulation or fractionation of olivine. In addition, PRIMELT3 also calculates the degree of partial melting (by both batch and accumulated-fractional melting), the temperature of the olivine liquidus, the olivine composition in equilibrium with the primary magma, the amount of olivine required to fractionate from/accumulate within the primary magma in order to form the observed lava composition, and the mantle potential temperature ( T P ) required to generate the primary magma.
Although PRIMELT3 is designed to identify samples -particularly those with more evolved compositions -that are unsuitable for the program, datasets must first be filtered to remove evolved samples which may yield spurious thermobarometric results ( Herzberg & Asimow, 2008, 2015. Most LIP magmas, by the time they are erupted or intruded, have experienced significant fractional crystallisation (e.g., Hartley &Thordarson, 2009 andMahoney, 2007 ) and Siletzia is no exception. As shown in the previous section, the vast majority of the lavas sampled from Siletzia contain between 4 and 10 wt. % MgO and so many will have undergone significant fractionation of phases other than olivine and Cr-spinel (e.g., clinopyroxene and plagioclase). Thus, the Siletzia database must be carefully screened prior to modelling. To screen our initial dataset ( n = 778) we first removed all samples with < 7.5 wt. % MgO [which are likely to have experienced clinopyroxene fractionation ( Herzberg & Asimow, 2008 )], or those which fall outside of olivine control lines on major element bivariate diagrams ( Fig. 3 ). Of the 100 analyses left following this filtering, 28 gave permissible PRIMELT3 melting solutions through either batch or accumulated fractional melt-ing models ( Table 3 ). These samples include representatives from each of the constituent formations of Siletzia and their geochemical data are detailed in Table 4 .

Primary magmas
The major element composition of the primary magmas computed by PRIMELT3 for the 28 Siletzia samples are shown in ( Table 3 ). PRIMELT3 does not calculate trace element abundances in the computed primary magmas. However, using the equation for Rayleigh fractional crystallisation ( Eq. 1 ) in conjunction with the estimates of olivine fractionation ( Table 3 ) from PRIMELT3 can yield primary magma trace element compositions ( Hastie et al., 2016 ).
Where C 0 is the concentration of the trace in the primary magma, C l is the concentration of the trace element in the sample, F is the degree of melt remaining (after olivine fractionation -as calculated by PRIMELT3), and D ol is the olivine mineral/melt partition coefficient of the trace element during olivine fractionation (taken from Bédard, 2001 ). The calculated trace compositions of the primary magmas are shown in Table 5 . The primary magmas calculated for the Siletz River Formation and Grays River Volcanics have similar trace element signatures which are characterised by enrichments in the most incompatible trace elements relative to the least incompatible trace elements, e.g., (La/Yb) N > 1, and depleted HREE patterns. Both of these primary magmas have trace element compositions intermediate between EMORB and OIB, though the LREE systematics of the Siletz River Primary magma are more similar to OIB than the Grays River Volcanics primary magma ( Fig. 7 A and 7 B).
The primary magmas of the Metchosin Igneous complex ( Fig. 7 C) all have trace element signatures very similar to NMORB in that they all show depletions in the most incompatible elements relative to the least incompatible elements.
Arguably, the most compositionally complex primary magmas belong to the Crescent Formation. Two (potentially three) primary magma types can be discerned from the primary magma compositions of PRIMELT3-successful Crescent Formation samples. Group 1 primary magmas ( Fig. 7 D) have trace element patterns virtually identical to EMORB, characterised by enrichments in the most incompatible elements relative to the least incompatible elements, (La/Yb) N > 1 and (Nb/La) N > 1. Group 2 primary magmas ( Fig. 7 E) are characterised by depletions in the most incompatible elements relative to the least incompatible elements, (Gd/Yb) N ~1 and (Nb/La) N > 1. A third group ( Fig. 7 F) of Crescent Formation primary magmas, calculated from analyses of Babcock et al. (1992) and Glassley (1974) might be delineated on the basis of (Th/Nb) N ratios > 1, which set these samples apart from the other primary magmas calculated here. However, the analyses from these older papers do not contain the full trace element suite in the primary data. Thus, a full description of their primary magma trace element geochemistry is impossible, and thus, further investigation of their petrogenesis will be limited in this study.
The extent of melting and olivine fractionation calculated by PRIMELT3 for all samples shows large overlap between constituent suites ( Table 3 ). This means that the differences in the trace element signatures of the primary magmas for each suite can only be explained due to their derivation from compositionally distinct mantle sources. The nature of these mantle sources will be explored later.

Magmatic evolution
This section assesses how each of the constituent Siletzia suites evolved to produce the compositional variation observed in the analysed lavas ( Fig. 2 ). MELTS_Excel (MELTS) is thermodynamic modelling  Renne et al., 2011 ) was used to establish neutron flux values (J). Decay constants are listed in Renne et al. (2011) . A: analytical error only. F: full external error including decay constant and standard age uncertainties, as determined by the optimisation method of Renne et al. (2011) . MSWD: mean square weighted deviation (reduced chi-squared statistic). J: Neutron flux parameter. software that can be used to investigate the crystallisation of magmatic systems ( Gualda & Ghiorso, 2015 ). By utilising a database of thousands of experimentally-derived compositions of silicate liquids, MELTS-Excel can calculate the crystallisation sequence, proportions of solid and liquid phases in equilibrium assemblages within a silicate magma. MELTS_Excel can be applied to both hydrous and anhydrous magma compositions over a range of temperature and pressure of 500-2000 °C and < 2 GPa respectively, using a number of different redox buffers. Average major element primary magma compositions were calculated for the Grays River Volcanics, Siletz River, Crescent Formation Group 1 and Group 2 and the Metchosin Igneous Complex by taking the average of the individual primary magmas calculated by PRIMELT3 for each suite or grouping. Each average primary magma was modelled by MELTS_Excel using five scenarios: Model 1, fractional crystallization at 1 kbar (anhydrous); Model 2, fractional crystallization at 1 kbar (1% H 2 O); Model 3, fractional crystallisation at 1 kbar (0.25% H 2 O); Model 4, fractional crystallization at 3 kbar (anhydrous); Model 5, fractional crystallization at 7 kbar (anhydrous). All models use a quartz-fayalitemagnetite (QFM) oxygen buffer and calculate the composition of the evolving silicate liquid at 5% crystallization intervals until 90% crystallisation of the primary magma. Representative bivariate diagrams showing the liquid lines of descent for each model alongside compositional data for each suite are shown in Fig. 8 . Complete model outputs are recorded in Appendix 3.
The majority of the models consistently show that the major element geochemical trends of the constituent suites of Siletzia are best replicated by low pressure (~1 kbar) fractionation occurring in the presence of a small amount (0.25-1 wt. %) of H 2 O. These models predict that olivine is the first mineral to crystallise, followed by clinopyroxene and either plagioclase ± spinel, and then by apatite. Inspection of Fig. 8 shows that the majority of the major element variation of the Siletzia suites can be explained by ≤ 70% crystallisation of the primary magma of each suite.
In general, the trace element trends of the Siletzia suite show decreases in the concentrations of compatible trace elements during basaltic fractionation with decreasing MgO with concomitant increases in concentration of the incompatible trace elements ( Fig. 4 ). However, these trends are diffuse as are the trends on incompatible trace elements vs. Zr diagrams ( Fig. 5 ) and so are unlikely to be the product of a single fractionation process. Further, the suites display a range of trace element ratios ( Fig. 6 ) with some samples displaying negative Nb anomalies on primitive-mantle normalised trace element diagrams (e.g., Fig 7 F) which are often associated with contamination of a fractionating magma by continental crustal materials (e.g., Nelson et al., 1990 ;Ciborowski et al., 2014 ).
Trace elements can be used to test whether the crystallisation of the primary magmas of the suites operated under a purely fractional crystallisation (FC) or assimilation-fractional crystallisation (AFC). To do this, the model that best predicts the major element geochemistry for each Siletzia suite is further tested using trace elements with the equations for Rayleigh fractional crystallisation ( Eq. 1 ) and assimilationfractional crystallisation ( Eq. 2 ). Representative bivariate diagrams showing the liquid lines of descent for each model alongside compositional data for each suite are shown in Fig. 9 . Complete model outputs are recorded in Appendix 3, with preferred modelled parameters for each suite summarised in Table 6 .
Where C 0 , C l and C A are the concentrations of an element in the primary magma, daughter magma and crustal contaminant (Felsic Crust of Rudnick & Fountain, 1995 ) respectively. D is the bulk distribution coefficient for the trace element, themselves calculated from the mineral/melt partition coefficients from Bédard (2001) . F is the fraction of melt remaining after crystallisation. r is ratio of assimilation rate to the Table 3 Primary magma major element compositions (wt. %) and thermobarometric results obtained for Siletzia primary magmas studied here, as calculated by PRIMELT3 and FractionatePT † . Olivine addition = the amount of olivine added to the evolved composition in order to produce a primary magma capable of forming through batch melting of fertile peridodite; T p = mantle potential temperature; F = Degree of melting required to produce a primary magma capable of fractionating olivine in order to produce the evolved composition; Melting pressure = the pressure at which melting occurred at; Melt Model -which melt model (Batch or AFM or both) yielded successful solutions with PRIMELT3.  Table 4 Major and trace element geochemical data for the samples analysed here which yielded successful solutions with PRIMELT3. Data sources: a -this study, b - Glassley (1974) ; c - Babcock et al., (1992) ; d - Moothart (1993) ; e - Chan et al., (2012) ; f - Timpa et al., (2005) , g - Rarey (1985) and h - Yorath et al., (1999) . Full data available in electronic appendix.      fractional crystallisation rate (taken to be 0.2 in this case based on approaches used in similar magmatic systems (e.g., Gunther et al., 2018 ;Hari et al., 2018 ). The consistent (Th/Nb) N vs Zr N observed within the Grays River Volcanics samples indicates that the primary magma of the Grays River Volcanics fractionated without being contaminated by crustal material to any significant degree. Similarly, the trace element systematics of the Metchosin Igneous Complex samples show (La/Yb) N vs. Zr N trends which are most similar to those predicted by FC models, and thus contamination can be ruled out as a significant factor here also.
Fractional crystallisation models involving the primary magma derived from the Group 1 Crescent Formation samples can explain the trace element variation for most of the Crescent Formation as a whole. However, a relatively small number of Crescent Formation samples fall on diffuse trends between both the compositions of the primary magmas calculated from the Group 1 and Group 2 samples, and felsic crustal compositions. This may indicate that the Crescent Formation primary magmas fractionated in a number of shallow magma chambers, some of which interacted with, and assimilated their host rocks (potentially terrestrially derived sediments that were transported to the marine environment).
The crystallisation history of the Siletz River Formation remains enigmatic. Neither FC or AFC models of the calculated Siletz River Formation primary magma produces trends in trace element space that convincingly fit the data for the formation. Instead, it is possible that the majority of the Siletz River Formation were formed by still-unconstrained primary magmas, the derivatives of which are too evolved for the approach detailed here, or that the Siletz River Formation is a composite formation made up of fundamentally different suites which, as yet, remain unresolved.

Temperature of melting
Published modelling results have shown that the temperature within a mantle plume would be significantly higher (100-350 °C) than the surrounding mantle (see Campbell, 2007 for a review and Coogan et al., 2014 ). This prediction has been confirmed by seismic investigation from extant plume-driven magmatic provinces (e.g., Nelson & Grand, 2018 ;Schutt & Humphreys, 2004 ;Wolfe et al., 1997 ).
In order to assess whether the temperature of an ancient mantle source region of a suite of lavas was anomalously hot compared to the ambient upper mantle temperature at that time, it is first necessary to calculate the potential temperature ( T P ) of their mantle source. This can be achieved through comparison of the MgO contents of the primary magma calculated by PRIMELT3 with experimentally derived MgO contents of partial melts and solid-state adiabatic gradients ( Iwamori et al., 1995 ). Once calculated, this T p can be compared to estimates of the temperature of the upper mantle through geological time. This approach is used by Hole and Millett (2016) to estimate the mantle thermal anomaly associated with the emplacement of the North Atlantic Igneous Province who show, crucially, that such an approach yield temperature estimates consistent with seismic interpretations of extant mantle systems.
Various temperature models exist that attempt to show how the upper mantle has cooled since the Archean (e.g., Davies, 2009 ;Herzberg et al., 2010 ;Richter, 1988 ). These models suggest that during the emplacement of Siletzia between ~39 and 56 Ma, upper mantle temperatures would have ranged between 1305 °C and 1330 °C. Fig. 10 A shows the T P calculated for the Siletzia primary magmas, as well as the ambient temperature of the upper mantle during the lifetime of the province, as calculated by Davies (2009) . The PRIMELT3 modelling ( Table 3 ) shows that the largest range of mantle tempera-tures (1512-1682°C) are recorded by the Crescent Formation primary magmas and that primary magmas of the remaining suites fall within error of this range. Further, the data show that the Siletzia primary magmas are all derived from mantle significantly hotter ( > 190 °C) than that calculated for the ambient mantle at the time, as is predicted by mantle plume theory ( Campbell, 2007 ), and confirmed by observation of extant plume systems (e.g., Nelson & Grand, 2018 ;Schutt & Humphreys, 2004 ;Wolfe et al., 1997 ). We therefore conclude that the emplacement of Siletzia was driven by a mantle plume. This finding is supported by the interpretations of others who suggest that Siletzia magmatism formed in response to the impingement of a mantle plume beneath the lithosphere during the early-mid Paleogene ( Duncan, 1982 ;Phillips et al., 2017 ;Wells et al., 2014 ). Conversely, this finding does not support other hypotheses for the petrogenesis of Siletzia such as its emplacement in a rift-related marginal basin ( Brandon, 2014 ;Wells et al., 1984 ), or passive upwelling through a slab window during subduction of a mid-ocean ridge ( Babcock et al., 1992 ), neither of which require (though don't preclude) the input of anomalously hot mantle material.

Depth of melting
Major element ratios such as SiO 2 /MgO, CaO/Al 2 O 3 , and Al 2 O 3 /TiO 2 in a primary magma are sensitive to the conditions of pressure and temperature under which melting occurs. A number of thermobarometers have been developed in order to determine the depth of mantle melting required to form an observed lava composition (e.g., Herzberg, 1995 ;Herzberg & Zhang, 1996 ;Walter, 1998 ). FractionatePT ( Lee et al., 2009 ) is one such thermobarometer which is applicable to melts derived from peridotitic sources and uses the SiO 2 -FeO-MgO-CaO-Na 2 O-K 2 O contents of primary magmas to determine the temperature and pressure conditions required for melting.
Comparison of the temperature estimates from the PRIMELT3 and FractionatePT thermometers (Appendix 4) show that both thermometers give very similar estimates of melting temperature for the Siletzia primary magmas, and are similar, within error for most of the primary magmas calculated during this study. However, there is a systematic difference in the temperatures calculated by the two thermometers, which, in the case of the primary magmas discussed here, never exceeds 5%.
The good agreement between the two thermometers indicates that the estimates of temperature of melting ( Fig. 10 A) for the Siletzia province, are robust, and that the FractionatePT barometer, which calculates the pressure of melting can be used with confidence. Fig. 10 B shows the melting pressures (calculated by FractionatePT) for the Siletzia Primary magmas during the lifetime of the province. The data show that the largest range of melting pressures (2.8-6.4 GPa) are recorded by the Crescent Formation primary magmas and that primary magmas of the remaining suites fall, within error, of this range. These pressures of melting, along with the temperatures of melting indicate that the melting occurred at depths consistent with the garnet-lherzolite stability field ( Fig. 11 ).

Mantle source composition
The behaviour of the REE during partial melting has been quantitatively shown to vary systematically with the degree of partial melting and relative abundance of garnet in the peridotite source (e.g., Song et al., 2009 ). REE trends of the Grays River Volcanics, Siletz River and Group 1 Crescent Formation cluster around trends defined by melts derived from the melting of garnet peridotite. Conversely, the most primitive samples from the Metchosin Igneous Complex and Group 2 Crescent Formation have (Gd/Yb) N ≈ primitive mantle values, which are more indicative of melting occurring within the spinel stability field ( Fig. 7 ).
However, the conditions of melting, as calculated from the Siletzia primary magmas suggest that all of the Siletzia primary magmas were derived from garnet-bearing peridotite ( Fig. 11 ). Typically, magmas derived from the partial melting of garnet-bearing peridotites record trace element compositions characterised, in part, by steeply dipping HREE patterns on primitive mantle-normalised trace element diagrams (e.g., Kushiro, 1996 ;McKenzie & O'Nions, 1991 ). Such patterns are observed in the primary magmas of the Siletz River, Grays River Volcanics and Group 1 of the Crescent Formation ( Fig. 7 ). However, the Metchosin Igneous Complex and Group 2 Crescent Formation have flat HREE patterns on primitive mantle-normalised diagrams, more characteristic of melting occurring in the absence of garnet -despite recording similar degrees of partial melting and depths of melting to the other primary magmas discussed ( Table 3 ).
This apparent contradiction between the conditions of partial melting and the calculated trace element geochemistry of the primary magmas can be most easily explained by differences in the modal mineralogy of the mantle source regions from which the primary magmas were derived. Specifically, the primary magmas of the Siletz River, Grays River Volcanics and Group 1 Crescent Formation were likely derived from garnet-bearing peridotite, where garnet was in sufficient abundance to remain in the residue up to the significant degrees of partial melting ( ≤ 29%) recorded by some of the primary magmas ( Table 3 ). Conversely, the Metchosin Igneous Complex and Group 2 Crescent Formation primary lavas were most likely derived from garnet-bearing peridotite that had sufficiently low abundances of garnet such that the garnet was completely melted during the formation of the primary magmas from these suites (e.g., Walter, 1998 ).

Evolution of the Siletz plume
All of the primary magmas calculated by PRIMELT3 for Siletzia are derived from anomalously hot mantle, as is predicted by plume models ( Campbell, 2007 ), and this is the petrogenetic mechanism we favour for the formation of the province. The calculated primary magma trace element compositions ( Fig. 7 ) suggest that the plume sampled a range of mantle reservoirs common to OIBs (e.g., Jackson & Dasgupta, 2008 ), and potentially a third source (or mixture of sources) with a composition intermediate between that of OIBs and another less incompatible element-enriched reservoir. Therefore, aside from the anomalously high temperature of the Siletzia primary magmas, models invoking a mantle plume origin are further suggested by the identification of an OIB-like mantle (e.g., Condie 2005 ).
The plume paradigm has been argued for by others as a plausible mechanism to explain the origin of constituent suites of Siletzia. Quite often, such models link the proposed Siletz plume with the mantle plume currently driving magmatism within the Yellowstone Plateau, itself widely argued to be responsible for the volcanism preserved in the Columbia River Basalts and Snake River Plain ( Brandon & Goles, 1988 ;Camp et al., 2015 ;Geist et al., 2002 ). Duncan (1982) suggests such a link, partly on the basis on absolute plate motions, and argues that the nascent Yellowstone plume emplaced the Siletz Terrane as an oceanic island chain, proximal to the northwest Pacific coast of ancestral North America. The subsequent accretion of the Siletz Terrane and continuing westward movement of the North American plate over the Yellowstone plume is argued by Duncan (1982) and Murphy et al. (2003) to explain the apparent eastward younging of magmatism associated with the Siletz Terrane, Columbia River Basalt Province and Yellowstone -Snake River Plain. This interpretation is similar to that argued for by Wells et al. (2014) , who suggest that the rapid emplacement, and OIBlike compositional characteristics of the Siletz Terrane are indicative of the terrane being derived from a mantle plume.
Comparison of the trace element composition of the most primitive lavas of Siletzia with similarly primitive lavas of the constituent suites of the Yellowstone Magmatic Province (YMP) ( Fig. 12 A) shows that the highest concentration of analyses from both provinces share an Oceanic Plateau Basalt-like geochemistry. Though the analyses of the YMP span a much broader compositional range than those of Siletzia, the significant overlap in analyses indicate that the two provinces were derived from mantle regions that shared a similar composition. Fig. 7. Primitive mantle-normalised trace element diagrams of the Siletzia primary magmas, estimated by accounting for the effect of the olivine fractionation ( Eq. 1 ) as calculated by PRIMELT3 for each sample ( Table 3 ). Normalisation factors taken from McDonough and Sun (1995) . NMORB, EMORB and OIB compositions taken from Sun and McDonough (1989) . White symbols are literature data. Coloured symbols are new data.
In terms of isotopes, Pyle et al. (2009) argues that parts of the Siletz Terrane are isotopically indistinguishable from parts of the CRBP and uses this as evidence of a probable cogenetic link between the two, with the Siletz Terrane representing melting of the nascent Yellowstone plume head, and the Columbia River Basalts related to melting of the plume tail. Pyle et al. (2009) and Chan et al. (2012) have used similar arguments to suggest that the Grays River Volcanics were also derived from the Yellowstone plume, which supplied the primary magmas of the suite via a slab window within the partially-subducted Juan de Fuca Plate. However, further isotopic analysis by Phillips et al. (2017) , reveals some isotopic differences between the Siletz Terrane and the Columbia River Basalts which are difficult to reconcile with a cogenetic origin for both the Siletz Terrane and CRBP. Instead, Phillips et al. (2017) suggest these compositional differences may be the result of melting of a heterogenous plume head, or alternatively melting of plume(s) unrelated to that which produced the CRBP.
Published isotopic data for Siletzia are shown alongside isotopic data for the YMP in Fig. 12 B and 12 C. In terms of Nd and Sr isotopic space ( Fig. 12 B), the Siletzia lavas cluster around HIMU and PREMA compositions, with very few Siletzia samples trending toward more enriched compositions. The samples from YMP define an expansive trend from PREMA compositions towards EM2. Interestingly, several of the oldest YMP samples (from the Columbia River Basalt Province) overlap with the Siletzia cluster, while the youngest YMP samples (from the Yellowstone Plateau) fall entirely outside of the Siletzia cluster and plot in more enriched isotope space. In terms of Pb isotopes ( Fig. 12 C), a sim- Fig. 8. Representative major element vs. MgO bivariate diagrams of the Siletzia samples, alongside the average major element composition of each primary magma (larger, compound border symbol). Also shown are the liquid lines of descent predicted by Melts_EXCEL for each primary magmas under the model paramters described in the text. Markers on models indicate increments of 5% crystallisation. Note that major element chemistry of each suite is best approximated by low pressure crystallisation (1 kbar) with 0.25-1 wt % H 2 O. Complete model outputs available in the electronic appendix. White symbols are literature data. Coloured symbols are new data. ilar pattern is seen, whereby the Siletzia samples define a tight trend between PREMA and HIMU compositions with a small number of samples trending towards EM2 compositions. In this space, the YMP samples from the Columbia River Basalt Province show partial overlap with the Siletzia grouping, while the youngest YMP samples from the Yellow-stone Plateau plot away from the Siletzia grouping closest towards EM1 compositions.
Therefore, although some trace element compositional similarity exists between the lavas of the two provinces, the wide range of compositions within the YMP makes any such comparison equivocal and cannot, Fig. 9. Primitive mantle-normalised bivariate diagrams showing the trace element evolution of the Siletzia primary magmas (double outline) via modelled FC (dashed line) and AFC (solid line) pathways. AFC models contamination using the Felsic Crust composition of Rudnick and Fountain (1995) and an r value of 0.2. White symbols are literature data. Coloured symbols are new data.
alone, be used to argue for a cogenetic origin. Examination of the YMP and Siletzia isotopic systems better delineate potential mantle sources than trace elements do. These isotopes show that the oldest parts of the YMP share an isotopic composition, similar to (the relatively homogenous) Siletzia. This similarity disappears with the youngest YMP samples, and we interpret this to potentially show a secular change in mantle source region sampled by the Yellowstone Plume over time.

Environmental impact
The formation of LIPs has been shown to have significant impacts on the atmosphere, biosphere and hydrosphere and have been correlated to many first and second order extinction events (e.g., Bond et al., 2014 ;Kerr, 2005 ). The Palaeocene-Eocene Thermal Maximum (PETM) represents an abrupt climate warming of 5-8 °C at around 55.5 Ma ( McInerney & Wing, 2011 ;Westerhold et al., 2012 ). and is marked by a global temperature rise inferred from a > 1 ‰ negative excursion in 18 O of benthic foraminifera, indicating around a 5 °C increase in the temperature of deep waters ( Zachos et al., 2001 ).
The PETM is thought to be the result of a catastrophic release of several thousands of petagrams (Pg) of carbon into the ocean-atmosphere system ( Dickens et al., 1997 ), which is inferred from the large global negative carbon isotopic excursion and the extensive dissolution of deep-ocean carbonates ( McInerney & Wing, 2011 ). The PETM itself was a relatively short lived event that lasted around 200 ka ( McInerney & Wing, 2011 ), with the carbon release spanning less than 20 ka (but it may have occurred in less than a year according to some estimates ( Bowen et al., 2015 ;Higgins & Schrag, 2006 ). With the exception of benthic foraminifera, relatively few groups suffered major extinctions despite most terrestrial and marine organisms experiencing large shifts in geographic distribution, rapid evolution and changes in trophic ecol- Fig. 10. Thermobarometric results for Siletzia samples successfully computed by PRIMELT3. A -Mantle potential temperature ( T P ) as calculated by PRIMELT3. y -axis error bars = 42 °C ( Herzberg and Asimow, 2015 ). Also shown (dashed) is the contemporary temperature of the upper mantle using the mantle cooling curve of Davies (2009) ; B -Melting pressures as calculated by Fraction-atePT. y -axis error bars = 0.2 GPa ( Lee et al. 2009 ); C -Degree of Partial Melting as calculated by PRIMELT3. x -axis error bars for MIC and CFM taken from U-Pb age range and midpoint of Eddy et al. (2017) . x -axis error bars for SRV taken from Ar-Ar age range and midpoint of new data ( Table 2 ). White symbols are literature data. Coloured symbols are new data. Fig. 11. Diagram showing the pressure and temperature conditions of the transition of spinel lherzolite to garnet lherzolite as estimated by Klemme and O'Neill (2000) -solid line and ( Gasparik, 1984 ) -dashed line. Also shown are the melting pressure (FractionatePT) and temperatures (PRIMELT3) for the Siletzia primary magmas respectively. Error bars as in Fig. 10 . White symbols are literature data. Coloured symbols are new data.
ogy ( McInerney & Wing, 2011 ). The age of onset of the PETM has been estimated at around 56 Ma using radiometric dates of marine ash layers and orbital timing of marine sediments ( Westerhold et al., 2009 ), while the upper boundary of the negative carbon isotopic excursion (calculated from zircons in the upper part of the sequence) is thought to be 56.09 ± 0.03 Ma, ( Jaramillo et al., 2010 ).
Estimates for the exact amount of C released into the biosphere during the PETM vary widely from 3000 Pg to ~10,000 Pg ( Cui et al., 2011 ;Zeebe et al., 2009 ), while recent work ( Gutjahr et al., 2017 ), suggests that the carbon source is isotopically heavy likely to be volcanic in origin. Similarly unconstrained are estimates of the flux of this C into the biosphere over time. Recent modelling ( Gutjahr et al., 2017 ) has suggested that a cumulative release of ~10,000-12,000 Pg of C during the PETM would be required in order to drive the carbon isotope excursion observed in foraminiferal records that span the PETM. Arguing for a role for the North Atlantic Igneous Province (NAIP) in the triggering of the PETM, Gutjahr et al. (2017) multiply the estimated magmatic volumes of the NAIP by the 3600-6000 g/m 3 of C that are typically released by degassing magmas ( Saunders, 2015 ). Gutjahr et al. (2017) argue that the resultant 18,000-60,000 Pg of C released by the NAIP eruptions would have been sufficiently large to drive the C isotope excursion observed in the geological record at the onset of the PETM.
Though the majority of the Ar-Ar dating presented here, and other recent high-precision U-Pb analyses ( Eddy et al., 2017 ), rules Siletzia out as a potential initiating mechanism for the onset of the PETM, its potential effect on global climate warrants further investigation. Comparison of the eruptive ages of Siletzia with temperature records at the time ( Fig. 13 ) show that the eruption ages of the suites of the Siletz Terrane coincide with some of the highest sea-surface temperatures recorded during the Eocene ( Evans et al., 2018 ). Following the same approach as Gutjahr et al. (2017) and applying it to the estimated volume (2.6 × 10 6 km 3 ) of the Siletz Terrane ( Phillips et al., 2017 ;Trehu et al., 1994 ), we estimate that a potential 9360-15,600 Pg of C may have been added to the biosphere during emplacement of the terrane. Even before taking into account any potential positive feedbacks this new C flux might have had on atmospheric CO 2 at the time, we propose that the greenhouse effect of the eruption of Siletzia to have been sufficiently large to have contributed to the ~35 °C sea-surface temperature maxima observed in the Eocene climate record ( Evans et al., 2018 ).  ( Condie 2005 ) showing the > 7.5 wt% MgO samples from Siletzia presented here, along with the > 7.5 wt% MgO Yellowstone Magmatic Province (YMP) which itself is composed of the Columbia River Basalt (CRB), Snake River Plain (SRP) and Yellowstone Plateau (YP); B -87 Sr/ 86 Sr vs. 143 Nd/ 144 Nd, and; C -206 Pb/ 204 Pb vs. 207 Pb/ 204 Pb, both showing isotopic compositions of > 7.5 wt% MgO samples from Siletzia ( Phillips et al., 2017 ) and the YMP, showing the main oceanic mantle reservoirs of Zindler and Hart (1986) . White symbols are literature data. Coloured symbols are new data. Data for the YMP extracted from the GeoRoc compilation ( http://georoc.mpch-mainz.gwdg.de/ ).

Conclusions
1. The magmatism preserved within Siletzia was driven by the melting of anomalously hot mantle, several hundred degrees hotter than the ambient mantle at the time, and is consistent with a mantle plume model for the petrogenesis of the province. 2. The Siletzia plume sampled a restricted range of mineralogically diverse mantle reservoirs during the lifetime of the province. Some of these source regions are similar in composition to reservoirs which feed modern Ocean Island, and Mid-Ocean Ridge volcanic systems. 3. The primary magmas formed during partial melting of the Siletzia mantle reservoirs were able to rise through the crust until they ponded in shallow (~1 kbar) magma chambers. Here the majority of  Evans et al. (2018) . Vertical lines indicate ages of Siletzia volcanism using ranges described in Fig. 10 .
the magmas evolved through simple fractional crystallisation, with a minority interacting with their crustal host rocks. 4. Isotopic evidence suggests that most of the Siletzia lavas are distinct from most of the Yellowstone Magmatic Province samples. However, there is a compositional overlap between the Siletzia samples and the oldest parts the Yellowstone Magmatic Province. This may indicate a cogenetic link between the provinces, whereby the mantle source sampled by the plume has changed over time. 5. The onset of Siletzia volcanism is broadly coincident with the highest recorded Eocene sea surface temperatures. This timing, along with the magmatic volume of the province, make Siletzia a plausible contributary cause of the biospheric perturbation observed in the geological record at this time, and thus warrants further study.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.