We present newly acquired paleomagnetic data from Bandelier Tuff exposures in the Jemez Mountains (New Mexico) that show no statistically significant tectonic rotation over Quaternary time. Cooling units of the tuff were mapped in detail and correlated using new geochemical data, allowing us to confidently sample isochronous units for paleomagnetic remanence directions. In total, 410 specimens were subjected to step-wise thermal and alternating field demagnetization. Of the 40 accepted site means, 30 have α95 values ≤5°. Analysis of the geographic distribution of the site-mean declinations of the data set reveals no statistically significant tectonic rotation either across (northwest/southeast) the northeast-striking Jemez fault or across (east/west) the north-striking Pajarito fault zone. Similarly, our data do not record any measurable relative change in declination difference (−1.1° ± 1.6°) that could be interpreted as a rotation over the ∼0.36 m.y. time duration between deposition of the two principal stratigraphic members of the Bandelier Tuff. The step-over discussed in this paper is an area of exceptional structural complexity and, as such, meets the definition of “accommodation zone.” We propose the name “Jemez-Embudo accommodation zone” for this composite of structural and volcanic features in recognition of its regional importance in the evolution of the Rio Grande rift.

In this part of the rift, where Proterozoic- and Laramide-age faults have preconditioned the crust, idealized relay ramps, prevalent locally, do not occur at the regional scale. Instead, transfer fault zones have developed between half grabens dominated by preexisting faults. The pattern of faulting and accommodation of strain in the right-relayed step-over of the rift has been more or less invariant since the onset of rifting. From a global perspective, the difference between areas of modest crustal extension dominated by distributed deformation and those regions that develop transfer fault zones may ultimately be diagnostic of crustal conditioning and fault strength, such that weak fault systems focus strain within narrow zones.


Over a wide range of scales and tectonic settings, en echelon normal fault systems define areas of complex structure that hold critical clues to understanding the initiation and growth of continental rifts. Structural analysis of the step-overs between faults in these systems can provide information on the strength (or weakness) of the crustal volume between the bounding fault systems and the processes by which faults propagate and link together to form connected rift basins. In the last 15 yr, many investigations of en echelon fault systems have been pursued, both in the field (e.g., Trudgill and Cartwright, 1994) and in the laboratory (e.g., Acocella et al., 2005). These studies at the scale of the mechanical model and the outcrop, while providing valuable information on the geometry and kinematics of en echelon fault systems, are not straightforward to apply to the regional scale, where continental crust is preconditioned by previous deformational phases. At the regional (or rift) scale, the role of preexisting faults in concentrating deformation is thought to be significant (e.g., Acocella et al., 2005). It is our purpose to examine an excellent example of a regional-scale step-over with known faults of long ancestry in the Rio Grande rift (New Mexico), and determine the way in which the step-over accommodated strain in Quaternary time and in earlier periods of the rift's evolution.

Deformation of the crustal volume between en echelon faults, or fault systems, has been described as two end members that define a continuum of possibilities: the relay ramp and the transfer fault. A relay ramp is a broad area of ductile strain between normal faults (e.g., Peacock et al., 2000; Acocella et al., 2005), whereas a transfer fault is a discrete, subvertical fault oblique to a rift that transfers displacement between two adjacent basins undergoing differential extension (Gibbs, 1984). Transfer faults may be confined to the upper plate of a detachment or transect the detachment faults, if not the entire crust (e.g., Lister et al., 1986). The term “accommodation zone” was originally suggested by Bosworth (1985) to denote transfer fault systems of great structural complexity that separate half grabens of opposing detachment polarity (e.g., Gregory rift, Africa). The term was later expanded to include areas of differential extension but similar detachment polarity (e.g., Gulf extensional province; Stock and Hodges, 1989). Accommodation zones are now considered to be a fundamental element of rift architecture (Rosendahl, 1987) and are present at all scales of rifting.

Conceptual models of relay ramp evolution predict that the crustal volume between the bounding fault systems should show evidence for vertical-axis rotation (e.g., Ferrill and Morris, 2001). For example, strata deposited between two normal faults of the San Marcos fault system in northeast Mexico record ∼30° of clockwise vertical-axis rotation that occurred during deposition, with rotation magnitude increasing with lithologic age (González-Naranjo et al., 2008). Evidence for vertical-axis rotation is an important test of relay ramp formation.

Therefore, the primary objective of this investigation was to determine the temporal and spatial distribution of the rotational component (about a vertical axis) of deformation along the western boundary of the Española Basin, which lies in the step-over between the Albuquerque and San Luis extensional basins (Fig. 1). This complex zone of faulting is commonly considered to be a relay ramp (e.g., Kelley, 1982; Acocella et al., 2005). However, it has features that typify transfer fault zones (e.g., Gulf of Aden and northern Red Sea [Tamsett, 1984] and northern Gulf extensional province [Stock, 2000]), including important strike-slip faults at a high angle to the rift (Embudo, Jemez, and Tijeras fault zones; e.g., Muehlberger, 1979; Aldrich and Dethier, 1990; Kelley et al., 2003), a major locus of synextensional volcanism (the Jemez volcanic field; e.g., Smith et al., 1970), and an aborted, early rift basin decoupled from the main rift basins that have been active since ca. 7 Ma (Baldridge et al., 1994). For the modest amount of extension that occurred in the Rio Grande rift (∼10%; Golombek et al., 1983), mechanical models predict formation of relay ramps rather than transfer fault zones (Acocella et al., 2005). However, Acocella et al. (2005) also pointed out that in extensional settings where extension is <21%, reactivation of preexisting structures, subparallel to the extension direction and along rift margins, may encourage the formation of transfer faults.

To investigate the rotational component of deformation, we studied the regionally extensive Bandelier Tuff. East of the Valles caldera, on the Pajarito Plateau, the Bandelier Tuff cooling unit stratigraphy has been mapped in detail and correlated with geochemical data (e.g., Broxton and Reneau, 1995; Goff, 1995). Beyond the Pajarito Plateau, it is challenging to map the Bandelier Tuff based on cooling unit stratigraphy alone because welding and vapor-phase alteration vary greatly with radial distance from the Valles caldera, the source of the tuff, as well as with paleotopography. Geochemical correlation is an important means of distinguishing among the cooling units and is an essential component of this investigation. We integrated new stratigraphic and geochemical data for the Bandelier Tuff with the paleomagnetism of its different ignimbrite sheets across the Jemez Mountains in order to test the possibility of geographic variability of vertical-axis rotation. In addition to determining geographic variability of vertical-axis rotations, we also tested for sequential variability in rotation by using more than one isochronous unit (discrete ignimbrite sheet) in our paleomagnetic analysis. Our study reveals no statistically significant magnitude of relative rotation over the area covered by outflow facies of the Bandelier Tuff and has direct implications for understanding temporal and spatial patterns of deformation and evolution of the western margin of the Rio Grande rift. Specifically, our work demonstrates that the complex structure of the Jemez Mountains area does not constitute a regional relay ramp, but rather it is a major transfer fault zone, or accommodation zone. Our work highlights the significance of preexisting structures and fault strength on concentrating deformation within narrow zones.


The Rio Grande rift, ∼1000 km long, consists of a series of right-stepping en echelon basins (Kelley, 1982; Chapin and Cather, 1994), including the Albuquerque, Española, and San Luis Basins (Fig. 1). The Jemez lineament, a broad zone where volcanism and localized deformation have affected all levels of the shallow crust (e.g., Aldrich et al., 1986; Lutter et al., 1995; Lewis et al., 2009), crosses the Rio Grande rift in the Jemez Mountains and corresponds with a right step-over between the Albuquerque and San Luis Basins. The Valles caldera, source of the Bandelier Tuff, is located at the junction of the Rio Grande rift and the Jemez lineament. In our study area, a local manifestation of this lineament (Fig. 2) is the Jemez fault, a broad southwest-northeast–trending zone of locally concentrated, shallow crustal deformation (e.g., Lutter et al., 1995).

Previous paleomagnetic investigations in sedimentary and igneous rocks of the Española Basin have established vertical-axis rotations since middle Miocene time (Brown and Golombek, 1985, 1986; Salyards et al., 1994; Hudson et al., 2004). A recently published paper questioned whether some of these data sets adequately averaged paleosecular variation of the geomagnetic field and whether they could be used to provide an accurate, absolute estimate of rotations (Harlan and Geissman, 2009). Those investigations in which secular variation has been adequately averaged document modest (∼7°–9°) counterclockwise vertical-axis rotation in Pliocene basalt flows (Hudson et al., 2004) and no to modest counterclockwise vertical-axis rotation in Oligocene intrusive and volcaniclastic rocks (Harlan and Geissman, 2009), providing equivocal support for studies done in sedimentary rocks (Brown and Golombek, 1985, 1986; Salyards et al., 1994). Until our study, no one had examined the extensive ignimbrite sheets of the Bandelier Tuff for vertical-axis rotations.

The Quaternary ignimbrite sheets that comprise the ca. 1.85, 1.61 Ma, and 1.25 Ma cooling unit members of the Bandelier Tuff (dates from Izett and Obradovich, 1994; Phillips, 2004; Phillips et al., 2007; Gardner et al., 2010) form a semicontinuous cover from the eastern edge of the Colorado Plateau (eastern flank of the Sierra Nacimiento), across the western margin of the Rio Grande rift and toward the axis of the rift at the latitude of the Española Basin. Unaltered ignimbrite or ash-flow tuff sheets (including welded and partly welded to nonwelded units) are noted for the geologic stability of their magnetic remanence (e.g., Grommé et al., 1972; Reynolds, 1977; Geissman et al., 1980; Weiss and Noble, 1989).

Paleomagnetic data from ash-flow tuffs, in particular, from outflow facies, where accurate reference of the remanence to the paleohorizontal can be accomplished, have been successfully used both to correlate such pyroclastic deposits (e.g., McIntosh, 1983; Best et al., 1995; Maughan et al., 2002) and to quantify rotations about vertical axes (e.g., Hudson and Geissman, 1987; Wells and Hillhouse, 1989; Lewis and Stock, 1998b; Petronis et al., 2009). Intracaldera facies of ash-flow tuffs often present complexities in that compaction fabrics are not always a faithful representation of the paleohorizontal or because intercaldera hydrothermal alteration has affected the paleomagnetic signal. Because ash-flow tuffs typically cool at a rate that is considerably faster (e.g., Riehle, 1973; Riehle et al., 1995) than field directional changes reflecting secular variation of the geomagnetic field, they are usually high-fidelity recorders of the geomagnetic field over short periods of time. Their paleomagnetic directions typically provide single virtual geomagnetic poles (VGPs), and they cannot be a priori assumed to reflect any form of long-term averaging of the geomagnetic field (e.g., Reynolds, 1977; Geissman et al., 2010).

Barring potential complexities associated with emplacement on irregular topography, the inclination of the magnetization characteristic of the tuff should be fixed with respect to the paleohorizontal (Byrd et al., 1995) as identified through field relationships, regardless of the magnitude of tilting or vertical-axis rotation. In addition, compaction/welding processes can result in significant inclination flattening, as has been observed in very thick ash-flow tuff deposits (e.g., Rosenbaum, 1986), or the development of such a strongly anisotropic magnetic fabric that the resulting thermoremnant magnetism is not a high-fidelity recorder of the geomagnetic field (e.g., Rochette, 1987). After accounting for these issues, deviations in declination from locality to locality, as recorded in tuffs, are likely to reflect vertical-axis rotations (e.g., Wells and Hillhouse, 1989; Petronis et al., 2009).

As a possible contributor to crustal deformation, vertical-axis rotation of parts of the Earth's crust can be quantified by comparing the directions (specifically declination) of remanent magnetizations of rocks to the geomagnetic field direction at the time the rocks acquired their remanence. This comparison is done using one of two approaches. In the first, if the collective data from numerous temporally distinct rock units in a part of the crust suspected of rotation provide an adequate long-term (±106 yr) average of the geomagnetic field, then the mean of such data can be compared with an expected field direction for the appropriate time that is determined from the paleomagnetic apparent polar wander path from an adjacent (stable) craton. In the second, directional data from a single isochronous unit (e.g., lava flow or ash-flow tuff cooling unit) at a locality suspected of rotation can be compared with data from the identical single unit at a locality outside of the suspected deformation zone to estimate a magnitude of relative rotation (e.g., Wells and Hillhouse, 1989; Petronis et al., 2009).


Previous Work

The Bandelier Tuff consists of three members erupted as a series of ash flows at ca. 1.85 Ma, 1.61 Ma, and 1.25 Ma (e.g., Izett and Obradovich, 1994; Spell et al., 1996; Phillips, 2004; Phillips et al., 2007; Gardner et al., 2010). While the eruption mechanism of the oldest member of the Bandelier Tuff, the La Cueva Member, is currently unknown (Gardner et al., 2010), both the older Otowi Member and the overlying Tshirege Member of the Bandelier Tuff erupted during enormous, caldera-forming events and are extensively exposed throughout the Jemez Mountains (Smith et al., 1970). The Otowi and Tshirege Members, the stratigraphic foci of this investigation, each consist of a sequence of ash-rich ignimbrites containing crystals, pumice, and lithic fragments (Fig. 3). The commonly accepted age of the Otowi Member is 1.613 ± 0.011 Ma (2σ weighted mean of four samples), obtained by the 40Ar/39Ar method from pumice lumps of the Guaje pumice bed (Izett and Obradovich, 1994; their sample 91G35, close to this study's site AN1).

For the Tshirege Member, Phillips et al. (2007) used single crystal laser fusion on four samples to obtain high-resolution 40Ar/39Ar dates. An outflow-facies ignimbrite from the Pajarito Plateau yielded their preferred date of 1.256 ± 0.010 Ma (Phillips et al., 2007; sample EP-44), as the sample had no anomalously aged crystals and gave a more precise 40Ar/39Ar result than their other three samples. This date is statistically indistinguishable from published ages of Izett and Obradovich (1994) and Spell et al. (1996) that have been recalculated using the same sanidine flux monitor age.

The dates of individual eruptive units of the Bandelier Tuff are indistinguishable within uncertainties from one another, such that eruption of the entire Tshirege Member is thought to have occurred on a time scale shorter than a few thousand years (e.g., Phillips et al., 2007). On the other hand, data based on predictive modeling studies from the Bishop Tuff of California indicate an eruptive cycle lasting approximately 3 yr, with periods of quiescence between eruptive units in various sections of 5–800 d (Sheridan and Wang, 2005). Given that no evidence for unconformities has been found within the Tshirege Member and that the estimated eruptive volume of the Bishop Tuff is 600 km3 (Wilson, 2008), as compared to a Tshirege volume of 250 km3, it is likely that the Tshirege was erupted over a very short time interval.

Ignimbrite sequences are typically subdivided into “cooling units” (Smith, 1960), which are unique to these volcanic rocks and represent single or multiple pyroclastic flows in which the juvenile material (pumice and shards) was welded during compaction and cooling. The Otowi Member is a single cooling unit, whereas the Tshirege Member is a compound cooling unit, showing at least four distinct welding reversals attributed to intervals of cooling between emplacement of successive packages of pyroclastic flows.

Cooling units in the Tshirege Member, which may or may not coincide with contacts between eruptive units (e.g., Gardner et al., 2001), have been mapped based on primary and secondary features, including welding, devitrification, vapor-phase alteration, oxidation, and relative content of crystals, pumice, and lithic fragments. The cooling unit stratigraphy on the Pajarito Plateau has been well documented (e.g., Vaniman and Wohletz, 1990; Broxton and Reneau, 1995; Gardner et al., 1999, 2001; Dethier et al., 2007). In other parts of the Jemez Mountains beyond the Pajarito Plateau, the cooling unit stratigraphy is less well determined, and that was a primary motivation for the stratigraphic and related geochemical studies reported here.

Because several features in ash-flow tuffs, including degree of welding, devitrification, and vapor-phase alteration, can vary laterally and may depend on several factors, including temperature at time of deposition or input of meteoric water during cooling (Lipman and Friedman, 1975), major- and trace-element data are often used to provide definitive identification of eruptive units, and we take this approach in this study. The Tshirege Member is stratigraphically zoned from high-silica rhyolite (76–77 wt % SiO2) at its base to low-silica rhyolite (72 wt% SiO2) at its top (Smith and Bailey, 1966), consistent with a progressively deeper source within a zoned magma body during eruption (e.g., Smith, 1979). Major- and trace-element concentrations vary in a predictable manner in the Tshirege Member (see following), and eruptive units can be readily distinguished (e.g., Caress, 1995; Broxton et al., 1996).


We collected samples of Bandelier Tuff for whole-rock X-ray fluorescence (XRF) analysis to aid in unit identification and facilitate stratigraphic correlations (e.g., Krier et al. 1998; Gardner et al., 1999, 2001). We sampled sections (measured using a Suunto hand level) at Cat Mesa, Eagle Canyon, San Juan Mesa, Sawyer Mesa, and Seven Springs, as well as individual cooling units at San Antonio and Vallecitos (Fig. 4). At all these sites, we took samples from an outcrop depth of 5–10 cm to avoid weathering features. The Bandelier Tuff stratigraphy at Pajarito Mesa and Mortandad Canyon (Fig. 4) is well described from detailed geologic studies (e.g., Lavine et al., 2003; Gardner et al., 2008; Lewis et al., 2009); no additional geochemical analysis was performed at those localities.

Major and trace elements were analyzed in 55 bulk samples using an automated Rigaku wavelength-dispersive XRF spectrometer. Samples were crushed and homogenized in 15–20 g parts in a tungsten-carbide shatterbox. Sample splits were dehydrated at 110 °C for 4 h and then allowed to equilibrate with the ambient atmosphere for 12 h. One-gram splits were fused at 1100 °C with 9 g of lithium tetraborate flux to obtain fusion disks. Additional one-gram splits were heated at 1000 °C to obtain the loss-on-ignition (LOI) measurements. Elemental concentrations were calculated by comparing X-ray intensities for the samples to those for 21 standards of known composition, after correcting for absorption. The XRF method we employed calculates the concentrations of ten major oxides (SiO2, TiO2, Al2O3, Fe2O3, MnO, MgO, CaO, Na2O, K2O, and P2O5), ten trace elements (V, Cr, Ni, Zn, Rb, Sr, Y, Zr, Nb, and Ba), and weight loss on ignition (Table 1). Elemental concentrations of V, Cr, and Ni in the Bandelier Tuff are generally below detection limits and therefore are not reported.

Unit Descriptions

Otowi Member of the Bandelier Tuff (Qbo)

Ash flows of the Otowi Member were erupted and emplaced during the volcanic and structural events that formed the Toledo caldera (Smith and Bailey, 1966). On the Pajarito Plateau, the Otowi Member is a mostly nonwelded ignimbrite, up to 125 m thick, with an 8-m-thick moderately welded zone near the top (borehole SHB-3; Gardner et al., 1993). West of the caldera, however, much thicker sections of densely welded Otowi Member are common, especially in the Seven Springs area (Kelley and Kelley, 2004). Other than the general lack of welding in the Otowi Member on the Pajarito Plateau, even in thick sequences relatively close to the caldera source, it is petrographically similar to the Tshirege Member. Quartz, commonly in bipyramidal form, and sanidine are the most conspicuous crystal phases in the Otowi Member. In contrast to the Tshirege Member, the Otowi tends to contain more lithic fragments, with significant contributions from Precambrian lithologies (Eichelberger and Koch, 1979; Gardner et al., 1993).

Tshirege Member of the Bandelier Tuff (Qbt)

Ash flows of the Tshirege Member were erupted and emplaced during the volcanic and structural events that formed the Valles caldera, which is the modern physiographic caldera (Smith and Bailey, 1966; Gardner et al., 2010). Tshirege Member thicknesses range up to ∼1000 m on the Pajarito Plateau, but the typical thickness of outflow facies of Qbt is ∼300 m. The highest volume and most laterally extensive cooling units are Qbt1 and Qbt2. In the following paragraphs, we describe those cooling units from which we collected data for our geochemical and paleomagnetic investigations; these units are: Qbt1, Qbt2, Qbt3, and Qbt4. Due to its absence in most of the sampled sections, data from Qbt3t (Gardner et al., 1999) were not collected.


Qbt1 is the basal ignimbrite of the Tshirege Member. Qbt1 ranges from entirely vitric to entirely vapor-phase altered, with a near continuity of variations in between. On the Pajarito Plateau, Qbt1 consists of a sequence of variably welded and altered subunits, Qbt1g (glassy), Qbt1v-c (vapor-phase altered, colonnade), and Qbt1v-u (upper vapor-phase altered). Generally, Qbt1 contains 15%–30% pumice, 10%–20% quartz and sanidine phenocrysts, and 1%–5% lithic fragments. Crystals are primarily quartz and sanidine in subequal proportions, with trace amounts of clinopyroxene, hornblende, and fayalite. Qbt1 is widespread across the Pajarito Plateau.


This cooling unit is also widespread across the Pajarito Plateau. The greatest thicknesses and highest degree of welding in Qbt2 occur in the west, closer to the source, whereas eastern outcrops are thin and poorly welded. The base of Qbt2 is commonly marked by a pumice swarm (typically <1 m thick) containing ∼30% pumice lapilli up to 15 cm in diameter. Otherwise, Qbt2 contains ∼10%–15% pumice, 15%–25% quartz and sanidine phenocrysts, and <1%–2% accidental lithic fragments; no mafic accessory minerals were observed. The top of Qbt2 is marked by a gradational decrease in welding (over ∼1 m) from moderately welded Qbt2 to nonwelded Qbt3.


Qbt3 is rich in pumice (∼30%) compared to the underlying Qbt2 and contains relatively abundant (locally up to 5%) accidental lithic fragments, most of which are around 5 cm in diameter. Qbt3 is enriched in phenocrysts compared to other units, with at least 30% crystals. Quartz and sanidine occur in subequal amounts. A distinctive feature of the phenocryst population of Qbt3 is the relatively coarse crystal size—most crystals reach 4–6 mm in diameter. Crystals are also commonly euhedral, as single crystals or single-crystal fragments. The top of Qbt3 is locally characterized by gas escape pipes up to 10 cm across (Crowe et al., 1978) and funnel-shaped fossil fumaroles associated with small-displacement faults and fissures (Gardner et al., 2008).


This unit includes two distinct cooling units. The lower subunit consists of poorly to moderately welded, crystal-poor ignimbrite (e.g., Gardner et al., 1998). This basal part of Qbt4 is relatively poor in pumice (less than 5%) and crystals (10%–15% phenocrysts of mostly quartz and sanidine to 2–3 mm across in diameter) with rare accidental lithic fragments. Ferromagnesian silicate crystals are extensively altered and in many localities are yellow in color and surrounded by distinct white alteration halos (Lewis et al., 2002). The upper subunit of Qbt4 (including the geochemically and mineralogically defined Qbt5 of Warren et al., 2007) consists of nonwelded to densely welded tuff with conspicuous crystal agglomerations, which are relatively large (up to 2 cm in diameter) and consist of intergrown feldspar (anorthoclase) and quartz with minor amphibole and pyroxene. Thin sections show that the anorthoclase in these agglomerations is commonly cored with plagioclase (Gardner et al., 2001). This subunit has been referred to as “the anorthoclase unit” (e.g., Doell et al., 1968).

Geochemical Results

A compilation of XRF data on bulk rock samples from the Otowi and Tshirege Members is given in Table 1. Variations in major-element oxides SiO2 and TiO2 and trace elements Rb and Nb show distinct trends in both members (Fig. 5). The earliest erupted ash flows in both Otowi and Tshirege are higher in silica and are more radiogenic than subsequent deposits (e.g., Stimac et al., 1996).

The SiO2 content in sampled Otowi units ranges from 73.4 to 78.1 wt%. TiO2 ranges from 0.08 to 0.19 wt%. These trends are similar to samples recovered from borehole SHB-3 on the Pajarito Plateau (Gardner et al., 1993), although the ranges determined in this study are wider (Fig. 5). Rb concentration ranges from 105 to 317 ppm, and Nb ranges from 53 to 172 ppm, showing a trend of increasing depletion of incompatible elements from older to younger cooling units. Rb and Nb trends fit with SHB-3 trends as well. This is consistent with systematic depletion of incompatible elements upward in the ignimbrite (Stimac et al., 1996).

The SiO2 content in the Tshirege units we sampled ranges from 75.2 to 78.1 wt%, and TiO2 ranges from 0.05 to 0.13 wt%. SiO2 and TiO2 contents are inversely correlated, with SiO2 decreasing up section and TiO2 increasing up section. This is the typical pattern observed in the Tshirege Member on the Pajarito Plateau (e.g., Gardner et al., 1999). Rb concentration (84–228 ppm range) and Nb concentration (50–150 ppm range) both decrease up section, but there is some overlap between cooling units.

Based on cooling unit characteristics (as described here) and geochemical data (Table 1; Fig. 5), we correlated our measured sections from areas in the Jemez Mountains with respect to well-characterized Pajarito Plateau trends. The Cat Mesa section spans Qbo through Qbt3. The uppermost site collected is well within the Qbt3 compositional field described by the Pajarito Plateau data (Figs. 4 and 5). The lowermost sample from Eagle Canyon is Qbo, and the upper sample may be either Qbt1 or Qbt2; its Rb/Nb signature is in the transitional zone between the two units. The sample from San Antonio is either Qbt1 or Qbt2; on both the Rb/Nb and SiO2/TiO2 graphs, it plots in a zone of overlap between Qbt1 and Qbt2 data. The San Juan Mesa samples fall into two groups, those that match with Qbt1 (the lower part of the section) and those that fit with Qbt2 (the upper part). The two geochemical samples from Sawyer Mesa are in Qbt2 and Qbt3, well within those fields on the geochemical variation plots. Even though the lower sample could be in Qbt1 based on geochemistry, its stratigraphic position is more consistent with Qbt2. All of the Seven Springs section is in Qbo and Qbt1, except for the topmost sample, a small mesa-top remnant, which could be Qbt2.

Summary of Stratigraphic Data

East and southeast of the caldera (Sawyer Mesa and Eagle Canyon), the Bandelier Tuff stratigraphic section is basically the same as on the Pajarito Plateau. West and southwest of the caldera, Tshirege subunits Qbt1 through Qbt3 are present. However, the thickness, degree of welding, and vapor-phase alteration vary considerably from relations on the Pajarito Plateau. Qbt1 at San Juan Mesa is almost entirely vitric. Qbt2 on Cat Mesa is a sequence of thin ignimbrites with partial cooling breaks between them, compared to the densely welded Qbt2 typical of the Pajarito Plateau (e.g., Lavine et al., 2003). In other studies of Qbt1 and Qbt2, X-ray diffraction data were used for correlation purposes due to the similarity in the mineralogy of these units (e.g., Broxton et al., 1996; Broxton and Reneau, 1995; Lavine et al., 2003). The mineralogical variations among deposits on the Pajarito Plateau result from the distinctive pattern of vapor-phase alteration, which is not confined to a particular stratigraphic level (e.g., Stimac et al., 1996). Qbt3 is exposed locally west and south of the caldera, but it does not have the lateral continuity that is common on the Pajarito Plateau. Nonetheless, outcrop characteristics that distinguish these units on the Pajarito Plateau (e.g., high crystal content in Qbt3) are also observed in these other areas.


Previous Work

Numerous paleomagnetic investigations of ash-flow tuffs have demonstrated that these volcanic deposits can provide faithful, isochronous records of the geomagnetic field acquired at the time of emplacement, cooling, and compaction (e.g., Riehle, 1973; Reynolds, 1977; Byrd et al., 1995; Geissman et al., 2010). The characteristic magnetization (ChRM) of most ash-flow tuffs is typically a thermoremanent magnetization, carried by a low-titanium magnetite and/or maghemite. Isolation of the ChRM is usually accomplished with progressive alternating-field (AF) demagnetization to peak fields of 120–160 mT. In some cases, because of a relatively high concentration of high-coercivity, single-domain magnetite, a combination of AF followed by progressive thermal demagnetization has been used to fully isolate the ChRM. Maximum laboratory unblocking temperature ranges between ∼610 °C and 630 °C have been interpreted to suggest that a low-titanium maghemite is an important contributor to the remanence. Because an ash-flow tuff cooling unit cools over a range of temperatures typical of those associated with thermoremnant magnetism blocking, a single, discrete ash-flow tuff unit should acquire a uniform ChRM direction, with an inclination that is fixed relative to the paleohorizontal unless affected by differential flattening due to extreme compaction and/or welding. Thus, data from the same outflow facies cooling unit collected from different geographic locations allow for the examination of differential, or relative, rotation expressed as statistically significant differences among paleomagnetic declinations.

Paleomagnetic investigations in the Bandelier Tuff have been previously conducted by Doell et al. (1968) and MacDonald and Palmer (1990). In their classic investigation of Quaternary volcanic rocks in the Jemez Mountains to elucidate details of the geomagnetic polarity time scale, Doell et al. (1968) collected samples from six sites in the Otowi Member (Qbo) at two geographic localities and six sites in the Tshirege Member (Qbt) at two different locations. In their study of the anisotropy of magnetic susceptibility of the Bandelier Tuff with the goal of inferring pyroclastic transport directions, MacDonald and Palmer (1990) concentrated on the Tshirege Member and reported data from 18 sites around the Valles Caldera, collected at 14 distinct localities. The paleomagnetic data from these studies are discussed later herein.


Using a gas-powered, water-cooled drill, we obtained 8–20 independent samples within the five cooling units of the Bandelier Tuff at a total of 40 sites. Standard oriented cores were drilled from adequately welded intervals, with oriented block samples collected at seven sites where only nonwelded layers were present. Field cores were oriented using both magnetic and sun compasses, with differences in orientation that were less than 2° between determinations. After preparation, all specimens were stored in a magnetically shielded room with average field intensities below 200 nT. Specimens were demagnetized using both AF and thermal methods. Demagnetization procedures were performed in low-magnetic-field environments in paleomagnetic laboratories at the University of Michigan and the University of New Mexico using 2-G Enterprises cryogenic magnetometers. In total, 410 specimens were demagnetized; of these, ∼20% were thermally treated, and ∼80% were subjected to AF demagnetization. Thermal demagnetization employed 15–23 temperature steps ranging from 70 °C to 680 ° C using a furnace with magnetic fields less than 10 nT in the sample region. Alternating field demagnetization employed 14–25 steps ranging from 2 to 120 mT. Average intensities of natural remanent magnetization (NRM) were ∼0.95 A/m. Illustrative vector component diagrams of thermal and alternating field demagnetization response (Fig. 6) show peak laboratory unblocking temperatures between 570 °C and 620 °C and progressive randomization of the NRM over a range of alternating peak fields, indicating that a relatively moderate-coercivity mineral (possibly both magnetite and maghemite) is the remanence carrier.

The ChRM direction for each specimen was computed using principal component analysis (Kirschvink, 1980). Following determination of sample ChRM directions (Table 2), site-mean characteristic directions were calculated using Fisher (1953) statistics, due to the straightforward nature of Fisher's approach to calculating confidence limits and applying significance tests (Butler, 1992). In order to assess data quality, we chose to use α95 values (as opposed to the precision parameter, k) because it provides a tool for visualization, data comparison, and calculation of any vertical-axis rotations. For our investigation, only site means with a 95% confidence limit (α95) ≤10° were considered acceptable. Two site means were eliminated from further study (EC5 and SJ2l), and of the remaining 40 site means, 30 had α95 values ≤5°. Site mean ChRM directions are listed in Table 2, and examples of paleomagnetic data at the site level (Fig. 6) show typical distributions of sample directions used to estimate site mean directions. After calculating the mean direction for each site, we corrected for local magnetic declination by adding 10° to the site mean; all site mean directions are of reverse polarity. All of the paleomagnetic data obtained in this study are given in geographic (in situ) coordinates. Tectonic corrections were not applied at any of the sampling localities. Throughout most of the Jemez Mountains where the Bandelier Tuff is exposed, orientations of eutaxitic structures and/or contacts between cooling units are essentially horizontal. In a few cases, we measured eutaxitic structures and/or contacts between cooling units with dips up to 5°, yet it is difficult to assess whether such inclined compaction fabrics are tectonic, and thus should be corrected for, or if they are primary features associated with initial emplacement and compaction and/or welding.

Although a single ash-flow tuff cooling unit cools over too short an interval to average secular variation of the geomagnetic field, it is possible that subtle, small-magnitude directional changes are recorded within a single cooling unit, or among cooling units that are closely emplaced over time; a good example of this is the Miocene Peach Springs Tuff (Wells and Hillhouse, 1989). Concentrated sampling at specific stratigraphic levels within a single ash-flow tuff offers a way to define a single, instantaneous field direction, and our collection of data from multiple sites in the Bandelier Tuff provides a robust approach with which to evaluate directional variations within isochronous cooling units.

Paleomagnetic Results

Our 40 paleomagnetic sites were distributed in the five distinct cooling units of the Bandelier Tuff (described previously) at 14 localities around the Valles caldera (Fig. 4). Mean directions were considered in temporal and spatial contexts to look for any systematic distribution of paleomagnetic declinations that could be interpreted as a record of vertical-axis rotations. First, we combined site mean directions into stratigraphically ordered data sets, to test for any systematic temporal variation in directions. This was accomplished by correlating paleomagnetic sites with measured sections and geochemical sample locations to be certain directional data were from the same cooling unit. We compared results from units within the Tshirege Member and between the Tshirege and Otowi Members from sites where a vertical section was exposed. Second, we evaluated the possibility for spatial variability of rotation. This was achieved by combining site mean directions from specific geographic localities and grouping locality means with respect to major tectonic structures (e.g., the Jemez and Pajarito faults).

Temporal and Spatial Relationships

We collected data from each of the isochronous cooling units in the Bandelier Tuff, including Qbo, the Otowi Member, and Qbt1, Qbt2, Qbt3, and Qbt4 in the Tshirege Member, to determine whether relative vertical-axis rotations have affected the western margin of the Española Basin since ca. 1.2 Ma, as well as over the time interval from ca. 1.6 to 1.2 Ma between major eruptions. By correlating our paleomagnetic sites with measured sections and geochemical data (Tables 1 and 2; Fig. 4), we identified each specific Bandelier Tuff cooling unit from which all paleomagnetic data were sampled (with the exception of site AN-1, which may be either Qbt1 or Qbt2). Where possible, we incorporated previously published paleomagnetic data from Doell et al. (1968) and MacDonald and Palmer (1990) to increase the overall confidence for each estimate of a specific cooling unit mean direction.

Our paleomagnetic data were obtained from measured sections of multiple cooling units at Cat Mesa, Eagle Canyon, San Juan Mesa, and Seven Springs, as well as from individual cooling units at Los Alamos Canyon, Dome Road, Mortandad Canyon, Pajarito Mesa, Peralta Canyon, San Antonio Canyon, Sawyer Mesa, upper Sandia Canyon, Vallecitos, and Young Ranch (Fig. 4).

First, we consider the four cooling units of the Tshirege; overall unit mean directions from selected localities where we were able to gather data in vertical and lateral sections are shown in Figure 7. Vertical section data were collected at San Juan Mesa and Sawyer Mesa, and data comprising a lateral section were collected at Pajarito Mesa. For San Juan Mesa (Fig. 7A), three sites (28 samples) were located in Qbt1, yielding mean values of D = 177.8°, I = –34.1°, and α95 = 8.7°; two sites were located in Qbt2 (14 samples), yielding mean values of D = 174°, I = –34.0°, and α95 = 3.6°. For Sawyer Mesa (Fig. 7B), one site was located in Qbt2 (10 samples), and it gave a mean of D = 175.0°, I = –36.0°, and α95 = 2.0°, and two sites were located in Obt3 (20 samples), giving a mean of D = 172.5°, I = –38.0°, and α95 = 13.2°. We included one Doell et al. (1968) site (D12) in Qbt4 from Sawyer Mesa because it was less than 3 km away, and it allowed us to compare another unit within the Tshirege. The mean for the Sawyer Mesa Doell et al. (1968) site is: D = 173.0°, I = –38.0°, α95 = 4.4° (7 samples). For the lateral section shown in Figure 7C, D = 179.3°, I = –29.7°, and α95 = 9.6°. This is based on four site means (61 samples).

Given the consistency of site means among the Tshirege cooling units (Qbt1–Qbt4) from the vertical and lateral sections described here, we then considered the Tshirege as a whole member (Fig. 8A). In total, 321 samples from 45 sites provided a Qbt member mean of D = 175.6°, I = –35.7°, α95 = 2.2°. While data quality is well within acceptability criteria as indicated by the low α95 value (Butler, 1992), the dispersion in inclination and declination visually represented in Figure 8A is undeniable. Since a temporal change spanning the four cooling units of the Tshirege is not the source of the dispersion (Fig. 7), we next look at the influence of faults in causing spatial variation in declinations.

Here, we present our Tshirege data in terms of locality means with respect to the two major tectonic features in the study area: the Jemez and Pajarito faults (Figs. 4, 8B, and 8C). Our data show no statistically significant declination change across (northwest/southeast) the Jemez fault (4.0° ± 2.7°) or across (east/west) the Pajarito fault zone (1.9° ± 1.7°). Similar results were obtained for the Otowi Member from a total of 62 samples distributed among 14 sites (Fig. 9). The whole member mean yields the following orientation: D = 176.7°, I = –45.0°, α95 = 3.0°, with no statistically significant declination change across (northwest/southeast) the Jemez fault (−3.9° ± 2.6°) or across (east/west) the Pajarito fault (7.9° ± 2.3°).

We also consider whether any localities show a systematic shift in declination between the Otowi and Tshirege Members that could indicate a phase of rotation between ca. 1.6 and 1.2 Ma. The overall Otowi and Tshirege Member means differ by ∼9° in inclination yet less than 2° in declination (–1.1° ± 1.6°). Our data therefore do not show evidence for vertical-axis rotation between the time of Qbo and Qbt eruptions. This is a simplified assessment, however, as the difference in inclination is significant, and part of it probably reflects paleosecular variation of the field over the ∼0.36 m.y. time interval between major eruptive events, as previously noted by Doell et al. (1968).

Other Potential Causes of Data Dispersion

Our data (Figs. 7–9) show dispersion that cannot be explained within a tectonic framework. Other parameters that may contribute to data outliers in ash-flow tuffs include: variable compaction and welding, paleosecular variation recorded during extended cooling, vapor-phase alteration or fumarolic activity, and identification and/or sampling errors. Palmer et al. (1996) and McIntosh (1991) reported that the Bishop Tuff and the Mogollon-Datil ignimbrites, respectively, show little correlation between welding and inclination and that thermoremnant magnetism directions are generally laterally and vertically consistent within individual ignimbrite outflow sheets, showing no evidence of syncooling secular variation or subblocking temperature flowage. For the Bishop Tuff, the chief source of within-unit variation in site-mean directions is uncertainty in structural corrections, particularly in areas of strong tectonic extension. This is also the case in northeastern Baja California, where tilting caused dispersion in inclinations (Lewis and Stock, 1998a, 1998b). In terms of our data, six sites out of 59 (10%) can be considered outliers; these are MP16, MP17, PM6, PM4, PM5, and SE10 (Table 2; Fig. 4). MP16 and MP17 are the only two sites from a previous investigation (MacDonald and Palmer, 1990) for which we could not assign a member, Otowi or Tshirege. Since proper stratigraphic correlation is paramount when determining a member mean, we were forced to remove these two data points from our analysis. Whereas PM6 has a high inclination, PM4 and PM5 have low inclinations; none of these sites is any more or less welded than the other sites at the Pajarito Mesa locality. However, the presence of fumaroles (Gardner et al., 2008) suggests that protracted cooling, resulting in paleosecular variation, could account for deviations in inclination from other sites. Given that PM6 was a block sample, we removed this data point in the event of a sampling error. Finally, SE10 had a low inclination, which we are unable to explain, as none of the factors associated with data dispersions is relevant for this site; as such, we kept data from SE10 in our analysis. In the end, of the six sites deemed outliers, we removed three and kept three.


The data presented here from isochronous cooling units in the Quaternary Bandelier Tuff reveal no temporal or spatial variations in declination. We conclude that, at the scale of this study, no block rotations about a vertical axis occurred near the major faults (Jemez and Pajarito) that transect the area (Fig. 10). Similarly, data from the Oligocene Espinaso Formation show no significant vertical-axis rotation and indicate that footwall rocks of the La Bajada fault did not rotate throughout the entire period of rifting (Harlan and Geissman, 2009). This is especially notable given the close proximity of the La Bajada fault to the Tijeras fault, a major northeast-striking fault with a long history of strike slip (Sanders et al., 2006). Throughout its Cenozoic history, strike-slip displacement has been closely focused within the Tijeras fault zone, suggesting strain partitioning in this part of the rift. These interpretations contrast with structural models of the Rio Grande rift in north-central New Mexico, which suggest that regional-scale (approximately tens of kilometers) block rotations accommodated an important component of strain (e.g., Brown and Golombek, 1985, 1986; Salyards et al., 1994; Hudson, et al., 2004). For example, the block bounded by the Embudo, Pajarito, Tijeras- Cañoncito, and Pecos-Picuris faults is thought to have rotated counterclockwise (Brown and Golombek, 1985). We suggest that well-documented vertical-axis rotations in this region represent localized effects, rather than a fundamental component of the regional deformation (Fig. 10); work by Goteti (2009) supports this conclusion.

Ingersoll et al. (1990) noted that the Pliocene–Quaternary Rio Grande rift is distinctly different from its precursor that evolved in early Miocene time, especially in structural style. In many regards, this conclusion is robust. However, the pattern of faulting and accommodation of strain in the right-relayed step-over of the rift has been more or less invariant since the onset of rifting. In the Rio Grande rift, where Late Cretaceous to early Tertiary Laramide style structures, and perhaps much earlier phases of crustal shortening of late Paleozoic and even Proterozoic age, conditioned everything that followed (e.g., Kelley et al., 1992; Bauer and Ralser, 1995; Pazzaglia and Kelley, 1998; Ingersoll, 2001; Magnani et al., 2004; Cather et al., 2006; Sanders et al., 2006; Koning et al., 2007), idealized relay ramps do not occur at the regional scale, although they are prevalent locally (e.g., Kelley, 1979; Goteti, 2009; Lewis et al., 2009). Instead, transfer zones (e.g., Morley et al., 1990) between half grabens are zones of distributed faulting dominated by north- and northeast-striking faults inherited from previous deformational events. Subvertical faults such as the Tijeras fault, and probably the Jemez and Embudo faults, strongly influenced the orientations and locations of transfer zones during extension (cf. Ingersoll et al., 1990).

In the northern part of the Rio Grande rift, three half-graben basins (Albuquerque, Española, and San Luis) developed beginning in the middle Miocene (Ingersoll et al., 1990; House et al., 2003). Exposures of early rift-fill deposits (Santa Fe Group, including basalts dated between 16.5 and 13.0 Ma; e.g., Hulen et al., 1991; Gardner and Goff, 1984) form a northeast-trending belt that marks the local boundary of the rift during the Miocene. In addition, basalts were erupted along the northeast-striking faults that were active during that period (Gardner and Goff, 1984). Between 10 and ca. 7 Ma, the rift boundary evolved as a swath of north-striking normal fault zones (Gardner and Goff, 1984; Baldridge et al., 1994) (Fig. 2). By ca. 4 Ma, the primary locus of extension in the Española Basin had shifted eastward to the Pajarito fault system and associated structures, directly linking the Albuquerque and San Luis Basins. During this time, displacement along the northeast-striking Embudo fault changed from dominantly dip slip to dominantly strike slip (Aldrich and Dethier, 1990).

Our paleomagnetic results strongly suggest that all these events occurred during one principal phase of transtensional shear deformation influenced by ancestral faults. Ancient faults that strike SW-NE may have facilitated early development of pull-apart basins of the same trend, accounting for thick sections of Santa Fe Group rocks in and near the Valles caldera (e.g., Hulen et al., 1991). Variation through time in fault rakes presented by Minor et al. (2006) may reflect a changing balance of strain partitioning between normal and strike-slip faults. Clearly, more work is needed to understand strain partitioning and the role of fault ancestry in this complex step-over.

The step-over we discuss in this paper is a fundamental element of the Rio Grande rift and an area of exceptional structural complexity, dominated by wrench tectonics. The location of ancestral faults determined the shape of the rift at this latitude and the boundaries of individual basins. We propose the name “Jemez-Embudo accommodation zone” (JEAZ) for this composite of structural and volcanic features in recognition of its regional importance in the evolution of the Rio Grande rift.

On a regional scale, a further implication of the paleomagnetic results we report in this study concerns interpretations involving large magnitudes of dextral or sinistral shear between the Colorado Plateau and Rio Grande rift. Some authors have argued for a time period during which significant sinistral offset characterized rift-age structures in northern New Mexico (e.g., Kelley, 1982; Chapin and Cather, 1994), whereas others have argued that dextral shear, related to plate-boundary deformation, was the dominant shear-sense component during rift formation (Wawrzyniec et al., 2002). In addition to our results, paleomagnetic data from Oligocene tuffs that crop out across the Colorado Plateau–Rio Grande rift margin in southern Colorado show no vertical-axis rotations since the time the tuffs were deposited (Mason, 2011). Thus, any dextral or sinistral shear within this region may have been accommodated by oblique slip on faults without vertical-axis block rotations and should be factored into any models of the evolution of the Rio Grande rift.

Finally, to contribute to the understanding of large-scale rift systems globally, we note that in contrast to our study, vertical-axis rotations have been documented in the northern Gulf extensional province of Baja California, Mexico (Lewis and Stock, 1998b). Distributed deformation appears to be characteristic of that region and perhaps more widespread than previously considered (Seiler et al., 2010). The pronounced difference in deformation style between the two regions may ultimately be diagnostic of fault strength. For example, whereas the relative isotropy of the Cretaceous batholithic rocks in northern Baja may have favored distributed deformation, the crustal anisotropy in north-central New Mexico favored concentrated deformation along preexisting faults. The Rio Grande rift, then, may be analogous to the San Andreas and Altyn Tagh fault systems, where weak faults focus strain (Zoback, 2000; Dupont-Nivet et al., 2004).


In this contribution, we concentrate on the rotational component of the deformation field and assess paleomagnetic data from the Quaternary Bandelier Tuff sampled within and near the relay zone that bounds the western margin of the Española Basin. These data are interpreted to show the absence of appreciable Quaternary-age vertical-axis rotations within this zone. The ignimbrite sheets of the Bandelier Tuff were correlated using new geochemical data, allowing us to confidently sample isochronous units. The absence of statistically significant differences among paleomagnetic declination data from the Otowi and Tshirege Members of the Bandelier Tuff exposed in the vicinity of the Pajarito fault suggests that post–ca. 1.2 Ma vertical-axis rotation of crustal blocks along this part of the western margin of the Rio Grande rift was negligible and cannot be measured within the uncertainties of the paleomagnetic data assembled here. Also, the data do not show any measurable rotations within the ∼0.36 m.y. time frame between emplacement of the Otowi and Tshirege Members of the Bandelier Tuff. Similarly, analysis of the data in terms of geographic location with respect to major structures reveals negligible rotation with respect to the Jemez fault, the Pajarito fault zone, or along the boundary of the Colorado Plateau.

In this part of the Rio Grande rift, where Proterozoic- and Laramide-age faults have preconditioned the crust, idealized relay ramps, prevalent locally (Lewis et al., 2009), do not occur at the regional scale. Instead, transfer fault zones were developed between half grabens as preexisting faults were reactivated. We suggest that modest-magnitude counterclockwise vertical-axis rotations since middle Miocene time, as documented in some previous studies, represent highly localized phenomena. The pattern of faulting and accommodation of strain in the right-relayed step-over of the rift has been more or less invariant since the onset of rifting. The difference between areas of modest crustal extension dominated by distributed deformation and those regions that develop transfer zones may ultimately be diagnostic of crustal conditioning and fault strength, such that weak fault systems, like the Rio Grande rift, focus strain within narrow zones.

Insightful discussions with Mark Hudson, Steve Harlan, and Gautam Mitra, and comments by Stephen Self and two anonymous reviewers greatly improved this manuscript. We appreciate field assistance from: Ruth Soto, Rajesh Goteti, Troy Held, Scott Broome, and the 2008 Earthwatch Institute Student Challenge Award Program participants. We are grateful for funding from the Institute for Geophysics and Planetary Physics and the Earthwatch Institute.