Large-magnitude Oligo-Miocene extension in southern Sonora: Implications for the tectonic evolution of northwest Mexico

Phillip B Gans

Tectonics, 1997

Introduction
Despite several decades of field and geochronologic studies, the temporal and spatial patterns of Cenozoic extensional tectonism and magmatism in western North America remain poorly understood and controversial. Fundamental questions remain unanswered: Were the distribution, geometry, and magnitude of extension in the Basin and Range province mainly controlled by plate boundary effects (e.g., Atwater, 1970; Stock and Hodges, 1989; Henry and Aranda-Gomez, 1992), or were they mainly controlled by intraplate processes and/or processes in the sublithospheric mantle (e.g., Coney and Harms, 1984; Jones et al, 1996)? Is Cenozoic magmatic activity mainly subduction related, rift related, plume-related, or a consequence of melting of previously thickened crust (see Christiansen et al, 1992 for a review)? Is there a close genetic relationship between Cenozoic magmatic activity and crustal extension and if so, what is the nature of that relationship (cf. Gans et al, 1989; Axen et al, 1993)? The starting point to address these questions is knowing the precise timing and spatial distribution of both magmatic activity and extensional faulting throughout the province.

For the U.S. portion of the Basin and Range province, there now exist reasonably good constraints on the space-time patterns of Cenozoic magmatism and a growing understanding of both the magnitude and timing of extensional tectonism. The inception of voluminous magmatic activity swept generally southward from Idaho and Montana in the Eocene, down through the Great Basin of Nevada and Utah in the Oligocene and Miocene. At the same time, magmatism swept generally northwestward from New Mexico across central Arizona to southern Nevada, such that both sweeps converged on the latitude of Las Vegas at about 10-15 Ma (Armstrong and Ward, 1991). Though most of this early to mid-Cenozoic magmatic activity is calc-alkaline in character and occurred at a time of continued subduction beneath western North America, its complex space-time distribution is in striking contrast to the linear volcanic chains typical of modern, Andean type continental arcs and has made it difficult to assess the role of subduction versus other processes in generating these melts.

The precise timing, magnitude, and distribution of extension within the U.S. Basin and Range province is less well understood. The cumulative amount of extension is probably 50-100% (200-300 km), and is distributed heterogeneously with domains 10-150 km wide of large magnitude extension alternating with areas of little extension (e.g., Gans, 1987; Wernicke et al, 1988). Extension in any given area has been episodic and the absolute timing varies dramatically from place to place. In some areas, large magnitude extension took place as early as the Eocene, whereas in others it did not begin until the middle Miocene. There appears to be a general trend of large magnitude extension beginning earlier in the Pacific northwest and in southern New Mexico and younging toward the Las Vegas-Death Valley portion of the Basin and Range, crudely mimicking the magmatic pattern. However, the degree of linkage between magmatic and extensional processes are still very much matters of debate (cf. Gans et al, 1989; Axen et al, 1993). It is important to note that much of the extension in the northern Basin and Range province occurred inboard of and concurrent with subduction. The relationship, if any, of later phases of extension to the change from a convergent to a transform plate boundary along the western margin is not clear.

A critical database that must be included in any discussion regarding the cause(s) of Cenozoic extensional tectonism in western North America is the geologic record from the Mexican Basin and Range Province. Northern Mexico, including all or most of the states of Sonora, Chihuahua, Durango, Sinaloa, and Nayarit, encompasses nearly half of the total "basin-range" extensional province (Henry and Aranda Gomez, 1992), yet our understanding of the distribution, timing, and magnitude of extension in this region is still in its infancy. Extension appears to be distributed in two broad N-S trending belts separated by the relatively unextended Sierra Madre Occidental. Regional tilt domains of Cenozoic rocks in Sonora delineated by Stewart and Roldan-Quintana (1994) display a pattern similar to that of the southern U.S. Basin and Range and indicate predominantly NE-SW extension. A NNW-trending belt of "metamorphic core complexes" in northern and central Sonora represents a continuation of those better studied in southern Arizona and is similarly interpreted to reflect locally large magnitudes of extension (Nourse et al, 1994). Constraints on the timing and magnitude of extension, however, are poor. Fragmentary evidence exists for normal faulting and tilting as old as Oligocene and as young as Pliocene (see reviews by Stock and Hodges, 1989, Henry and Aranda-Gomez, 1992, Stewart and Roldan-Quintana, 1994, and Lee et al, 1996). Henry and Aranda-Gomez (1992) argue for two principal episodes of extension, beginning at ~23 and 13 Ma, but do not address their respective magnitudes. Stewart and Roldan Quintana (1994) infer three episodes of extension in Sonora: Oligo-Miocene detachment faulting, early (?) to mid Miocene closely spaced normal faulting (pre ~ 10 Ma), and post - 10 Ma widely spaced basin-range faulting. A number of workers have speculated that a substantial amount of extension in northwestern Mexico might be related to Late Miocene "proto-Gulf" transtensional deformation(Stock and Hodges, 1989; Henry and Aranda-Gomez, 1992). For example, in the kinematic model of Stock and Hodges (1989), northwest movement of the Pacific Plate relative to mainland Mexico between ~11 and 5 Ma is hypothesized to have been partitioned between NNW transform motion on the Tosco-Abreojos fault west of Baja (Spencer and Normark, 1979) and ~ 110 ± 80 to 160 ± 80 km of distributed NE-SW extension in the Gulf extensional province.

Additional geologic mapping, construction of restorable geologic cross sections, and precise dating of bracketing units from different areas within the Mexican Basin and Range province is needed to assess these ideas. Mexico is an ideal laboratory for investigating the relationship between Cenozoic tectonism and plate tectonics because it lies inboard of a segment of the Pacific-North America plate boundary whose geometric evolution is increasingly well understood (Atwater, 1989). In addition, an exceptionally complete record of Cenozoic volcanism provides a wealth of dateable units such that rates and distribution of extension may be determined with ˛ 1 Ma resolution.

In this paper, I present new geologic and geochronologic data from a 700 km2 area in southeastern Sonora (Fig. 1) that sheds light on the extensional and magmatic evolution. These data document large magnitude NW-SE extension in an area where little extension was thought to have occurred, and indicates that this extension occurred mainly in the Oligo-Miocene, well before subduction had ceased. This extension was associated with a renewed pulse of (mafic) magmatism, following a prolonged period of tectonic and magmatic quiescence. Reconnaissance examination of a number of areas between the Sierra Madre Occidental and the Sonora coast and a review of work by others suggests that these types of relations might be typical and call into question models that invoke substantial post- 10 Ma extension. A speculative new model for the Cenozoic evolution of northwest Mexico is proposed, whereby extension occurred largely in an intra-arc setting and Baja California is hypothesized to have been transferred abruptly to the Pacific plate at ~ 10 Ma.

Geology of the Santa Rosa Area
The Santa Rosa area is located near the Chihuahua state line in southern Sonora, approximately 200 km southeast of Hermosillo (Fig. 1). It lies at the junction of the east-west Hermosillo-Chihuahua highway and the north-south Obregon-Sahuaripa highway and includes the four small pueblos of Tepoca, Santa Rosa, San Nicolas, and Santa Ana (Fig. 2). The area lies within the transition between the high plateau of the Sierra Madre Occidental and the rugged region of parallel basins and ranges (King 1939).

Field work focused on better defining the local stratigraphy and obtaining structural data regarding the magnitude and direction of tilting, the nature of contacts between the different units, and the location and geometry of major faults. The shaded parts of the map (Fig. 2) correspond to "strips of confidence", areas where most of the observations were made. The geology of the intervening areas is largely interpreted from aerial photographs and may not be accurate in detail.

Previous work
Previous geologic investigations of the Santa Rosa area by Damon et al, (1981), Mead et al (1988), and CochemŽ and Demant (1991) were of a reconnaissance nature and focused on regional stratigraphic correlations and on Laramide intrusive and hydrothermal activity. Regional 1:50,000 mapping by CochemŽ and Demant (1991) delineated a large plutonic complex (the San Nicolas granodiorite) and its volcanic and sedimentary wallrocks exposed as a window beneath extensive cover of presumed mid Tertiary ignimbrite and Neogene basin fill deposits. Most contacts between different units were interpreted as either depositional or intrusive; the only faults shown were high angle faults of relatively little offset. A number of K-Ar and 40Ar/39Ar age determinations on intrusive and volcanic rocks (Damon et al, 1981; Mead et al, 1988) indicate that the plutonic complex is part of the extensive Laramide batholithic belt and that the younger basin fill deposits are early to mid Miocene.

The record of Cenozoic magmatism and sedimentation
Rocks in the Santa Rosa area are divided into six informal mapping units based on their relative ages and lithologic character (Fig. 2). Stratified units include, in ascending order, I: Altered and metamorphosed andesitic volcanic and volcaniclastic rocks, II: Rhyolite ignimbrites and silicic domes and flows, III: Steeply-tilted porphyritic basalt interbedded with tuffaceous sandstone and conglomerate, IV: Younger, gently-tilted basin fill deposits - mainly fanglomerate with minor basalt, and V: Post-tectonic, "capping" basalts and andesites. These are effectively the same divisions used by CochemŽ and Demant (1991), though positions of unit boundaries may differ in detail. The aggregate thickness of the section is estimated to be 3-4 kilometers. The San Nicolas granodiorite demonstrably intrudes and metamorphoses only unit I, and was not unroofed until the early to mid-Miocene. Generalized columnar sections (Fig. 3) for the northern and southern parts of the study area illustrate the approximate thicknesses and proportions of different rock types and the position of dated samples. This stratigraphy and associated geochronologic data (Table 1, Supplementary data: Appendix 1) are crucial to understanding the structural evolution of the area, and are described in some detail below.

40Ar/39Ar Geochronology
40Ar/39Ar age determinations were obtained from fourteen samples (nineteen separates) in order to better understand the detailed chronology of magmatism and to place brackets on the timing of extensional faulting and tilting. This geochronologic data is summarized in Table 1; Additional information, including tabulated analytical data, age and K/Ca spectra, and inverse isochron plots for these samples is available upon request (Supplementary data: Appendix 1). Detailed step heating experiments (5-18 steps) and single grain total fusions (5-14 determinations) were performed on purified separates of sanidine, plagioclase, hornblende, or whole rock basalt from different volcanic units and different geographic areas (Figs. 2, 3). In addition, detailed step heating experiments were performed on coexisting hornblende, biotite, and K-feldspar from a fresh sample of the plutonic complex in order to assess its emplacement age and subsequent cooling history. All the samples yielded results that are readily interpretable. Ages of volcanic rocks within the study area span an impressive range from 60.0 ± 0.5 for andesitic rocks from the top of unit I to 16.7 ± 0.2 Ma for the youngest capping basaltic andesite of unit V (Table 1) and attest to a prolonged but episodic history of Cenozoic magmatism in the area. Errors reported on plateau ages are ± 2 sigma. Plagioclase and sanidine separates generally yielded concordant single grain ages and either simple plateaus or slightly climbing age spectra suggestive of minor loss.

The whole rock basalt samples yielded age spectra that are more complex (Supplementary data: Appendix 1) but the spectra are typical of fresh basalts and basaltic andesites that have been dated from other localities (e.g., Nauert and Gans, 1995). These basalt spectra generally have young apparent ages in the lowest temperature steps, climb abruptly to anomalously old apparent ages, then decrease monotonically and/or flatten out over the middle part of the spectrum, and finally decrease again at the highest temperature steps. The shape of this type of spectrum is interpreted to reflect the combined effects of reactor induced recoil (whereby 39Ar atoms may be dislodged from microcrystals and/or moved from less retentive to more retentive sites) and low-temperature argon loss in nature due to the hydration and/or clay alteration of groundmass glass (e.g. Mankinen and Dalrymple, 1972). Empirical studies by Gans and others (1992) on basalt samples whose ages are tightly bracketed by sanidine-bearing tuffs have shown that such spectra are nevertheless highly reliable. In particular, the reasonably flat, central part of the spectrum, corresponding to the gas released from about 700ˇ to 1000ˇC, generally provides a reliable estimate of the age, and has the advantage of having high radiogenic yields and high K/Ca ratios. Except for those spectra where there is clear evidence for substantial geologic 40Ar loss or reactor induced 39Ar loss, the "plateau" age determined in this fashion closely approximates the integrated total fusion age. Many of these ages are not strictly plateaus because the steps used do not lie within analytical error, so error assignment is somewhat subjective. The uncertainties listed for these ages are one standard deviation of the steps ages used to calculate the plateau - a somewhat arbitrary assessment but one that has proven to be highly conservative in other studies (Nauert and Gans, 1995).

Unit I: Older andesitic volcanic and volcaniclastic rocks
The oldest rocks observed in the study area are a previously undated sequence of fine grained volcaniclastic and andesitic volcanic rocks that are intruded by the San Nicolas granodiorite. These occur in generally NW-trending strips, bounded by faults on one side and intruded by the San Nicolas granodiorite on the other. The lower part of this unit consists mainly of well-laminated volcaniclastic sandstone and siltstone whereas the upper part is dominated by andesite to dacite flows, breccias, and subvolcanic intrusives. The total thickness of this interval is unclear because its base is not exposed and internal faulting is difficult to unravel, but exceeds 1 kilometer. Pervasive low grade metamorphism and/or hydrothermal alteration related to the Laramide intrusive complex has obscured original igneous and sedimentary textures in most of these rocks and largely converted mafic minerals to chlorite ± epidote ± Fe oxides, and plagioclase to albite ± sericite ± epidote. In the upper portion of this interval and distal to the intrusive rocks, hornblende-plagioclase phyric andesite flows are somewhat fresher. Igneous plagioclase from a flow near the top of this interval yielded a slightly climbing age spectrum with high-temperature ages of 59 - 60 Ma (Supplementary data: Appendix 1). In light of the pervasive weak alteration (hornblende replaced by chlorite, plagioclase is weakly sericitized) and the climbing nature of the spectrum, this flow is estimated to have a (minimum) emplacement age of ~60.0 Ma (Table 1). Thus, unit I is inferred to span the late(?) Cretaceous to early Tertiary, and is broadly correlative with the Lower Volcanic Complex of McDowell and Keizer (1977) and the Tarahumara Volcanic Rocks of Wilson and Rocha (1949).

Unit II: Early to Mid-Tertiary Silicic ignimbrites and domes.
Rhyolite ignimbrites and silicic domes disconformably overlie the altered andesitic rocks and underlie the porphyritic basalts and tuffaceous sediments of unit III. In the northern part of the map area, this unit consists of densely welded crystal-rich rhyolite ignimbrite that contains 30-40% crystals of plagioclase (15-20%), quartz (15%), biotite (2-3%), and sanidine (<1%). Pumice lapelli and lithic fragments are inconspicuous. It appears to be a single cooling unit, 400- to 800-m thick and is probably correlative with the widespread quartz-rich, sanidine-poor tuff of Cerro la Cebadilla described by CochemŽ and Demant (1991). A plagioclase concentrate from this tuff yielded a spectrum that climbed from 48.6 to 54.4 Ma, with the older ages generally associated with higher K/Ca ratios. Given the ambiguity in the spectrum and the fact that much of the plagioclase is slightly sericitized, the sample was reproccesed for sanidine. This yielded concordant single crystal total fusion ages and a flat release spectrum with a weighted mean plateau age of 54.33 ± 0.24 Ma (Table 1), which we interpret to be the emplacement age. It is a surprising age in that most reasonably fresh rhyolite ignimbrites in this part of Mexico are assumed to be Oligocene and part of the Sierra Madre Occidental suite (McDowell and Clabaugh, 1979). The impressive thickness for a single cooling unit suggests a near vent, possibly intracaldera, position. Its similarity in age and proximity to the San Nicolas granodiorite (see below) make it tempting to infer that the two are cogenetic. In the northernmost part of the area, a hornblende dacite dome and associated flows and pyroclastic debris intrudes and overlies the Eocene ignimbrite. Replicate analyses of 1-3 grain splits of sanidine and plagioclase yielded simple plateaus with a weighted mean age of 43.8 ± 0.2 Ma for this dome complex (Table 1).

In the southern part of the map area, a crystal-rich rhyolite ignimbrite that occupies approximately the same stratigraphic position, yielded a distinctly younger mean age of 33.07 ± 0.10 Ma based on numerous replicate analyses (Table 1). This tuff is fresher and only partly welded, and contains subequal proportions of sanidine and plagioclase. It may represent a distal outflow sheet from the Sierra Madre Occidental province to the east.

Unit III: Oligo-Miocene conglomerates and basalt
A distinctive sequence of red and tan sandstone and conglomerate, interbedded with basaltic lava and breccia, conformably overlies the rhyolite ignimbrites and records the inception of extensional faulting and tilting in the area. Clastic rocks within this interval generally coarsen upwards, and range from evenly bedded tuffaceous sandstone to thick debris flows and coarse angular sedimentary breccias composed mainly of basaltic clasts. The intercalated basalt and basaltic-andesite flows are fresh and contain olivine ± clinopyroxene ± plagioclase phenocrysts in a trachytic groundmass of plagioclase + Fe-oxides+pyroxene. In the northern part of the map area, complete sections of unit III are relatively thin (500 m) and basaltic lavas represent less than a third of the total section, whereas in the southern part of the area, this interval exceeds 1 km in thickness and is dominated by mafic lavas and breccias (Fig. 3). The contact with overlying gently tilted conglomerates of unit IV is a pronounced angular unconformity in some areas, whereas in other areas, tilts decrease gradually up section and the contact between the two is somewhat arbitrary. Units III and IV together are broadly correlative with the Baucarit Formation as defined by Dumble (1900) and King (1939), but the precise age(s) of the original sections are poorly known. Informal names are applied here because of these uncertainties and to emphasize important lithologic and structural distinctions between these two units.

Incremental heating experiments on groundmass concentrates from several different fresh basalts within unit III yield estimated 40Ar/39Ar ages ranging from 26.4 to 22.3 Ma (Table 1), suggesting that the sequence spans the Oligo-Miocene boundary. Detrital plagioclase separated from a volcaniclastic sandstone bed in the middle part of unit III yielded single grain total fusion ages that are tightly clustered, ranging from 23 to 28 Ma, but with large uncertainties. This single grain data indicates a provenance that was largely limited to coeval basaltic complexes. In contrast, coarse conglomeratic debris near the top of unit III and in the overlying fanglomerates of unit IV includes clasts of all older map units, including the San Nicolas granodiorite and suggest major unroofing at this time.

Unit IV: Early- to Mid-Miocene fanglomerate and basalt
Thick intervals of gently dipping fanglomerate with minor interbedded basalts unconformably overlie the older units and are well exposed in the vicinity of Tepoca and in the northern part of the map area (Fig. 2, 3). These types of successions occur as basin fill deposits related to basin-range type faulting throughout southern and central Sonora, though exact ages and correlations are poorly known. In the Santa Rosa area, unit IV occurs adjacent to and in the hanging wall of the youngest high-angle normal faults (Fig. 2). It consists of at least several hundred meters of evenly bedded polymict conglomerate and sandstone that generally fines upward and contains clasts of all older map units. New 40Ar/39Ar ages of 22.6 ± 0.3 Ma and 18.0 ± 0.4 Ma from mafic flows near its base and middle, and 17.45 ± 0.10 Ma from an overlying andesite flow (Table 1, Fig. 2, 3) indicate that the unit is entirely early Miocene.

Unit V: Middle Miocene capping basalts - The Tepoca Formation
Flat-lying hornblende andesite to aphyric basalt flows unconformably overlie the previously faulted and tilted terrane and form prominent mesas along the western and southeastern margin of the Santa Rosa area. These younger lavas were named the Tepoca Formation by CochemŽ and Demant (1991), who noted that they cover broad areas, extending well to the east and west of the area in Figure 2. This younger volcanic sequence was not examined in detail. A hornblende andesite flow at the base of this sequence and an aphyric andesite near the top yielded well-defined plateau ages of 17.45 Ma and ± 0.10 and 16.7 ± 0.2 Ma respectively (Table 1) and indicate that these post-tectonic deposits are middle Miocene.

Laramide Plutonic Complex
The San Nicolas granodiorite underlies a broad area in the central part of the map area. It has been described by Damon et al (1981, 1983), Means et al, (1988), and CochemŽ and Demant (1991) but has not been mapped in detail. The pluton is a composite mass that includes early batholithic rocks ranging from hornblende-biotite granodiorite to biotite granite, and late stage dikes, sills, and irregular stocks of aplite, tourmaline granite, intrusive breccia, granodiorite porphyry, and microdiorite. The granodiorite demonstrably intrudes only unit I. All other contacts are either known or interpreted to be faults. The pluton occurs at the deepest structural levels and appears to intrude to approximately the same stratigraphic level within each tilted section, suggesting it may have had a subhorizontal top.

Existing geochronologic data from the San Nicolas granodiorite are somewhat contradictory, but demonstrate that it is part of the Sonoran Laramide batholith (Damon et al, 1983). Damon et al (1981) reported concordant hornblende and biotite K-Ar ages of 49.3, 49.5, and 49.6 Ma from two different localities suggesting the pluton cooled quickly and that these approximate the age of emplacement. In contrast, Mead et al, (1988) report K-Ar and 40Ar/39Ar ages of 63 Ma (hornblende), 56.1 Ma (sericite) and 53.7 Ma (biotite) from the Tres Piedras deposit near Santa Rosa - a spread of ages that they attribute to slow cooling and/or reheating of a deep-seated pluton. Plausible explanations for these conflicting results are that the San Nicolas granodiorite includes two or more separate intrusions (> 63 and ~50 Ma) or that the hornblende age obtained by Damon et al (1981) is spuriously young, perhaps a result of biotite contamination. To the extent that the widespread chlorite-epidote alteration can be attributed to the plutonic rocks, field relations argue that the plutonic complex is largely younger than unit I. The very weak alteration of the ~ 54 Ma rhyolite ignimbrite in the northern part of the area could be interpreted as evidence that it is either younger than the pluton or distal to the hydrothermal system.

In order to better constrain the emplacement age(s) of the San Nicolas granodiorite and to shed light on its cooling history, detailed step heating experiments were performed on coexisting hornblende, biotite, and K-feldspar from a fresh outcrop of the granodiorite (sample SR-83), located approximately 3 km south of Santa Rosa (Fig. 2) and within 1-2 km of samples analyzed by Mead et al (1988) and Damon et al (1983). This sample yielded simple plateau ages of 56.7 ± 0.20 Ma for hornblende, 51.63 ± 0.11 Ma for biotite, and an age gradient for K-feldspar that climbs from ~30 Ma at low temperatures to ~49.5 Ma at high temperatures (Fig. 4). The K-feldspar diffusion data was modeled following the theory and numerical approaches of Lovera (1992). The data was best fit by 8 discrete domain sizes and an activation energy E of 44.11 kcal/mol. Iterative calculation of model age spectra and comparison with the observed spectrum were performed to obtain a best fit synthetic cooling history.

Assuming closure temperatures of 525 ± 40ˇC for hornblende and 325 ± 30ˇC for biotite during moderately fast cooling (review by McDougall and Harrison, 1988), the data suggest that the pluton was emplaced prior to 57 Ma, and then cooled rapidly (~40ˇC/m.y) from > 550ˇ C to < 325ˇC (Fig. 4). The K-feldspar age spectrum is best fit by nearly linear slow cooling (~7ˇC/m.y.) from ~300ˇC at 50 Ma to 190ˇC at 30 Ma (Fig 4). Thus, the combined thermochronologic data for the San Nicolas granodiorite are compatible with a simple history of rapid sub-solidus cooling to ~ 300ˇC followed by much slower cooling at lower temperatures. These data do not preclude the existence of younger phases within the plutonic complex, but suggests that their thermal effect was minor. More importantly, the cooling history suggests that the plutonic complex was initially emplaced and then continued to reside at moderate depths (4-8 km) throughout the early to middle Tertiary. A later episode of rapid cooling due to extensional unroofing during the Oligo-Miocene (25-20 Ma) is inferred based on structural observations described below, and by the fact that the plutonic complex was at the surface and shedding debris into local basins by 18-20 Ma.

Structural Geology
Prior to this study, there was little reason to suspect large magnitudes of extension in this part of Sonora. The Santa Rosa area lies far to the south of the belt of metamorphic core complexes (Fig. 1) and is only a few tens of kilometers west of the demonstrably unextended high plateau of the Sierra Madre Occidental. CochemŽ and Demant (1991) identified numerous high-angle normal faults but estimated only ~15% extension. Regional microstructural data collected by Chaulot-Talmon (1984) was interpreted to reflect three successive phases of deformation: (i) Early N40-60ˇE extension; (ii) Intermediate N60-80ˇW extension; and (iii) Late strike-slip faulting. However, no assessment was made of the total magnitude of strain, and the timing of their phases of deformation is not clear. Here, I present geologic mapping and structural observations that indicate large magnitudes of extension beginning in the late Oligocene during deposition of unit III, much of which apparently predates the extension recognized by Chaulot-Talmon (1984). It is important to note that the mapping is still incomplete and that many of the faults are obscured by dense vegetation. Nevertheless, geometric relations from within the "strips of confidence"(Fig 2) are quite systematic and strongly support the overall structural framework outlined below.

Geometric relations
The patchwork distribution of units in the Santa Rosa area (Fig. 2) attests to a complex structure. This map pattern is not simply the consequence of differential erosion, because units are steeply tilted. The central part of the area is underlain by the San Nicolas Granodiorite and represents the deepest exposed structural levels. North and south of the plutonic complex, stratified units are exposed in northwest trending sections that are repeated several times across their general strike. In the better studied northern sections, Units I to III are consistently tilted 30ˇ to 60ˇ NE and are cut by a complex system of high and low-angle faults that have mainly top-to-the-southwest or west displacement. Only the major faults are shown in Figure 2; innumerable smaller scale normal faults were observed in outcrop. Normal faults appear to be divisible into three generations. The oldest generation of faults are subhorizontal and consistently place younger rocks on older. These faults are the most difficult to recognize in the field and it is likely that only a small fraction have been identified. The flat faults are cut by northwest-striking normal faults that dip 20ˇ to 40ˇ to the southwest. These 2nd generation faults appear to be largely responsible for the map scale repetition of units (Fig. 2,5). They are evenly spaced at 2-4 kilometers and have apparent offsets of 3-4 kilometers (Fig 5). The youngest generation of faults are widely spaced NNE- to NNW-trending, high angle (50-75ˇ) normal faults that cut obliquely across the previously faulted and tilted terrane. These faults are the most obvious on satellite imagery and aerial photography because they control the modern topographic grain and the location of depocenters for the Baucarit Formation, but their offsets are estimated to be only 1-2 kilometers. Kinematic indicators on several of these faults indicate both dip slip and late stage(?) strike slip displacement.

The overall geometry of steeply tilted sections cut and offset by successively steeper generations of normal faults that dip opposite to the direction of stratal tilt is a structural style that is characteristic of many highly extended regions and has been described in detail by Proffett (1977), Gans and Miller (1983), and many others. It is not entirely clear whether the faults comprise three distinct generations or whether there is more of a continuous progression of faulting events. The N30-50ˇW axis of tilting for the older units and slickenlines on rarely exposed fault planes indicates that the extension direction during the first two generations of faults was approximately N50ˇE-S50ˇW. In contrast, a more northerly strike for the younger, high angle faults and for the associated basin fill deposits as well as slickenline data from exhumed fault surfaces indicate that the younger phase of extension was oriented approximately ENE-WSW.

South of the plutonic complex, the polarity of tilting and faulting for units I to III is reversed. The area was not studied in detail, but it was clear from both field observations and from inspection of aerial photographs that stratified units dip consistently 30-50ˇ to the southwest and that the major faults have top-to-the-northeast displacement (Fig. 2, 5). Thus, the San Nicolas granodiorite appears to occupy a generally northeast-trending accommodation zone between the two tilt domains. Poor exposure and the lack of stratigraphic markers precludes making any assessment of the detailed geometric relations within this zone. Three major, northeast-trending fault zones (Fig. 2) are evident from the truncation of units on aerial photographs and are interpreted to be left-slip(?) transform-like faults that developed within this accommodation zone during extension associated with the first two generations of faults.

Magnitude of extension
The map scale repetition of sections, steep tilts in Neogene rocks, and multiple generations of normal faults ranging from flat to steep are all suggestive of large magnitudes of extension. A schematic palinspastic reconstruction of part of the northern tilt domain (Fig. 5) indicates approximately 90% (8.5 km) cumulative extension, most of which occurred on the first two generations of faults. This reconstruction has large uncertainties in that only a fraction of the total faults are restored and their exact dips and subsurface geometry is poorly constrained. However, the simple fact that steeply-tilted sections are repeated with large horizontal separations virtually requires a large stretching factor (b Ĺ 2.0). Iterative reconstructions using a range of plausible fault geometries suggest that the cumulative extension is probably no less than ~60% but could exceed 100%. Most of this extension was oriented N50E-S50W; no more than 10-15% (~2 km) of the total extension occurred on the younger high angle normal faults.

Timing of extension
The structural, stratigraphic, and geochronologic data outlined above place tight constraints on the timing of extension in the Santa Rosa area. By analogy with other, better studied parts of the Basin and Range province, tilting of Cenozoic rocks in the area is attributed to "domino-style" rotation during slip on normal faults. Fanning tilts or "growth fault" relationships are interpreted to reflect syntectonic deposition and are used to bracket increments of faulting. In the northern part of the area, units I, II, and III, spanning 60 to 26.4 Ma, are all tilted 30-60ˇ to the northeast and the available data show no evidence for systematic changes in dip upsection. Gently dipping fanglomerate of unit IV, including a basalt flow dated at 22.6 ± 0.3 Ma (sample SR-56), unconformably overlies this tilted section and thus brackets ~ 40ˇ of NE tilting and presumably most of the SW-directed faulting between 26.4 ± 0.5 and 22.6 ± 0.3 Ma.

In the southern tilt domain, a sequence of 26.1 to 22.3 Ma basaltic andesite lavas and breccias (unit III) is uniformly tilted approximately 40ˇ to the southwest and is unconformably overlain by Unit IV fanglomerates that include aphyric basalt lavas dated at 18.0 ± 0.4 (SR-101). These relations indicate that major NE-SW extension may be younger in the southern tilt domain. Interestingly, a small exposure of 33 Ma rhyolitic tuff (Unit II) northeast of the tilted basalt section appears to be only gently tilted (~10ˇ). This gentle dip could either be attributed to local deformation or may indicate an earlier phase (pre-26 Ma) of northeast tilting followed by later southwest tilting. More work is required in the southern tilt domain to establish both the magnitude and precise timing of NE-SW extension.

Unit IV conglomerate is inferred to have been deposited during movement on the younger, ~N-S-trending high-angle faults because it appears to thicken towards them, dips decrease upsection, and strikes generally parallel the bounding faults. Bracketing ages of 22.6 ± 0.3 and 17.45 ± 0.05 on this unit suggest therefore, that even the younger ~ENE-WNW extensional faulting event is largely Early Miocene. Limited slip on some of these high-angle faults subsequent to the 17-16 Ma Tepoca Formation indicates that faulting continued locally after the middle Miocene, albeit at a greatly reduced rate.

Additional evidence bearing on the age of extension in the Santa Rosa area comes from the clast content of sedimentary deposits. Older conglomerates within the Oligo-Miocene sequence contain only volcanic debris derived from units II and III, whereas conglomerates in the unit IV contain clasts of all older rock units, including the San Nicolas granodiorite, and demonstrate that major extensional unroofing had occurred by the early Miocene.

Summary of field and geochronologic data
Though the mapping and analytical studies are still incomplete, several important conclusions can be drawn from the Santa Rosa area.
1. Volcanic and intrusive rocks span most of the Tertiary, yielding ages that range from 60 to 16 Ma, with no hiatus greater than 10 Ma. However, the intensity of magmatic activity fluctuated profoundly, with major episodes of local magmatism inferred for the Laramide (>60 to 50 Ma) and the Oligo-Miocene (27 to 17 Ma).

2. The cumulative amount of extension approaches 100% over a distance of 20 km. Large magnitude N50ˇE-S50ˇW directed extension was followed by minor amounts of ENE-WSW to E-W extension and distributed strike-slip deformation .

3. The age of most of the extension in the Santa Rosa area is tightly bracketed between about 26 and 20 Ma. After about 20 Ma, extension was limited to slip on widely spaced, fairly high angle faults and is no more than 10-15%.

4. A northeast-trending accommodation zone separates southwest-directed normal faulting (NE tilts) to the north and northeast-directed faulting (SW tilts) to the south. The accommodation zone is a structural high that exposes Laramide plutonic basement.

5. The inception of extension coincided with or immediately followed eruption of voluminous basalts and basaltic andesites, as these occur at or near the base of growth fault sequences. This fundamentally basaltic magmatism persisted after rapid, large magnitude extension had ceased.

The regional significance of these observations and the extensional history of some adjacent areas is explored in the following section.

Extensional history of adjacent areas
Stock and Hodges (1989), Stewart and Roldan-Quintana (1994), Henry and Aranda-Gomez (1994), Nourse and others (1994), and Lee et al, (1996) summarized fragmentary evidence for extension in northern Mexico. Though it is now generally recognized that the inception of extension locally dates back to the Oligocene (Henry and Aranda-Gomez, 1994) and that the magnitude of extension is locally great (e.g., Nourse et al, 1994), the paucity of geologic and geochronologic data has hindered efforts to estimate the total magnitude of extension across the Mexican Basin and Range province or to assess how this extension was distributed in time and space. Here, I review data that bears on the magnitude and timing of extension specifically along a transect from the Santa Rosa area to the Sonoran coast at this latitude. This 200 km-wide transect spans more than two thirds of the reconstructed width of the Gulf extensional province at 5.5 Ma, and thus may be representative of the province as a whole.

Suaqui Grande - Onabas area
The Suaqui Grande-Onabas area is located approximately midway between Santa Rosa and the coast (Fig. 1). Several weeks were spent reconnaissance mapping selected areas within a broad 50 x 70 km region, generally south of the Hermosillo-Chihuahua highway and west of the Yaqui River (Fig. 6). The focus of this work was on the stratigraphy and structure of Neogene rocks -i.e., correlations of units, degree of tilting, and geometry and magnitude of extensional faulting. Additional mapping and geochronologic work is in progress.

The stratigraphy of the Suaqui Grande-Onabas area (Fig. 7) broadly resembles that of Santa Rosa . Quartzose sandstone, siltstone, and conglomerate of the Triassic-Jurassic(?) Barranca Formation (Stewart and Roldan-Quintana, 1991) is overlain by a thick assemblage of andesite to rhyolite lavas, tuffs, breccias and volcaniclastic sandstones of the Cretaceous -Tertiary(?) Tarahumara Volcanics (Wilson and Rocha, 1949). These older units are propylitically altered and intruded by extensive masses of hornblende-biotite granodiorite that is part of the Laramide batholithic belt. Neogene rocks include an older suite of steeply to moderately tilted volcanic and volcaniclastic rocks and younger, gently tilted fanglomerate typically assigned to the Baucarit Formation (Fig. 7). The tilted volcanic suite is typically 1.0 to 2.5 km thick and consists, in ascending order, of hornblende andesite flows and breccias, pyroxene basalt interbedded with conglomerate, biotite-hornblende dacite flows and breccias, and conglomerate and tuffaceous sandstone interbedded with olivine basalt. A number of K-Ar ages on volcanic rocks from this group range from 27 to 20 Ma (McDowell and others, submitted), indicating the sequence is time correlative with Unit III in the Santa Rosa area.

Flat-lying to gently-dipping, well-indurated conglomerate and sandstone unconformably overlies the tilted and faulted early Neogene sequence. These younger conglomerates are restricted to three ~N-S trending strips, locally exceed 500 m in thickness, and contain clasts of all older rock types. A distinctive, crystal-poor rhyolite tuff is interbedded with the lower part of the conglomerate section and has yielded a K-Ar age of 12.5 Ma (McDowell and others, submitted).

The present level of geologic mapping is inadequate and the degree of structural complexity too great to portray anything but some of the youngest and most conspicuous faults on a regional scale map (Fig. 6). These youngest faults are an insignificant fraction of the total number of normal faults and may not even be the most important in terms of displacement. Careful examination of widely spaced localities consistently documented steeply to moderately tilted early Neogene rocks cut by closely-spaced (0.5 to 3 km) normal faults that have hundreds of meters to kilometers of displacement. Schematic cross sections for a number of these areas illustrate the typical structural style and variations in degree of tilting (Fig. 8). Overall, the structural style closely resembles that of the Santa Rosa area. At least two generations of normal faults are present and early Neogene sequences are tilted an average of 30-40ˇ, locally as much as 75ˇ, with generally the steepest tilts present in the stratigraphically lower units. The overall extension direction implied by the tilt axis ranges from N50ˇE to N75ˇE (average ~ N60ˇE - S60ˇW), consistent with what has been observed elsewhere (e.g. Henry and Aranda-Gomez, 1994). Northwest of a poorly defined boundary that trends southwest from Onabas, tilts in Neogene sequences are consistently to the northeast (Fig. 8, sections 1-5), and the major normal faults are inferred to have mainly top-to-the-southwest displacement. Southeast of this boundary, the polarity of faulting and tilting is reversed (Fig. 8, section 6). The reconnaissance nature of this structural data precludes making a precise estimate of the magnitude of extension. However, the presence of steeply tilted sections repeated by multiple generations of normal faults and the similarity in structural style to better studied highly extended terranes elsewhere, indicates that the magnitude of extension must be 50-100% and is comparable to the Santa Rosa area.

Gently-dipping conglomerate unconformably overlies the previously faulted and tilted early Neogene volcanic and sedimentary rocks but is cut by a prominent set of younger, NNW to NW trending high angle faults (Fig. 6). These faults control much of the topographic grain and have very linear map traces. A number of features suggest that they have late-stage(?) strike-slip displacement: Exhumed fault planes are steep (60-90ˇ) and commonly display subhorizontal slickenlines. The sense and amount of dip-slip displacement is inconsistent - in some cases even along the same fault. Units that are flat lying or very gently dipping away from the faults are commonly rotated to steep dips adjacent to the faults. Oroclinal bends of sections adjacent to some of these faults indicate predominantly right-lateral shearing. It is likely that a number of these faults were initiated as mainly dip-slip normal faults bounding basins in which the younger conglomerate accumulated, and then were reactivated as strike slip faults, sometime after 12.5 Ma. The magnitude of strike slip displacement on these faults is unknown.

The timing of deformation within this area is broadly constrained by K-Ar ages on some of the Neogene volcanic rocks (McDowell and others, submitted). Local angular unconformities and growth fault relations within some of the 27 to 20 Ma volcanic and volcaniclastic rocks indicate that extension may have begun in the early Miocene. Gently-dipping strata of the Baucarit Formation, including the 12.5 Ma tuff, unconformably overlies these faulted and tilted sections, demonstrating that most of the extension was completed by this time. Deposition of these younger conglomerates is inferred to have been coeval with movement on younger, widely-spaced high-angle normal faults, indicating minor amounts (<15%) of extension during this time. The conglomerates are in turn, cut by the youngest NW-trending strike slip faults, providing a lower age bracket on this deformation. Overall, the structural and geochronologic relations within the Suaqui Grande to Onabas area indicate a complex Neogene structural history. Large magnitude NE-SW extension occurred during the Early to Middle Miocene (mostly 20-13 Ma) and was accomplished by multiple generations of normal faults in two opposing tilt domains separated by an accommodation zone. This extension was largely over by 12.5 Ma and deposition of the younger conglomerates. Extension gave way to right-lateral shearing or mixed-mode faulting during the Middle to Late Miocene (post-12.5 Ma).

Sonoran coastal areas
Between Suaqui Grande and the Sonoran coast, poor exposure due to the deeply eroded landscape, and extensive cover by younger, gently dipping clastic and volcanic rocks (e.g. Bartolini et al, 1991) make it difficult to assess the magnitude and timing of extension. Two areas along the central and southern Sonoran Coast have been studied in detail. On Isla Tiburon (Fig. 1), 19 to 15 Ma volcanic rocks dip 40-50ˇ, and are overlain by progressively less tilted 13-11 Ma conglomerate, and nearly flat-lying 11 to 4 Ma ignimbrites (Neuhaus et al, 1988). These growth fault relations indicate that extension began by at least 15-13 Ma and was largely completed by about 11 Ma. In the Sierra Santa Ursula (Fig. 1), significant extensional faulting and tilting is bracketed between 11.4 and 10.3 Ma, in that 23.5 to 11.4 Ma volcanic rocks are tilted 20-35ˇ east and are unconformably overlain by more gently dipping 10.3 to 8.5 Ma volcanics (Mora Alverez and McDowell, in press). Much further to the south, in southern Sinaloa, Henry and Fredrikson (1987) report westward tilts in early Miocene volcanic rocks of 20-60ˇ and estimate the magnitude of extension to be 20 to 50%, but this extension is only broadly bracketed between 17 and 3 Ma. Finally, Gastil and Krummenacher (1977) describe a number of through-going NW-trending faults in coastal Sonora that have little vertical offset but may have substantial right-lateral offset - as much as 40 kilometers on one fault. The age of these faults is poorly constrained, except that they cut rocks as young as 10 Ma.

Summary
Structural and stratigraphic data between the Sierra Madre Occidental and the Sonoran coast suggest that the entire breadth of the extensional province may be highly extended. An evolution from earlier, large-magnitude NE-SW extension to late, relatively modest ~ENE-WSW extension and right-lateral shearing along N- to NW-trending high angle faults is suggested by structural relations throughout the transect (see also, Stewart and Roldan-Quintana, 1994). The magnitude of early NE-SW extension across this part of Sonora is estimated to be at least 50-100% (or ł 70-100 kilometers); Younger ~ENE-WSW extension is estimated to be no more than 10-15% (˛ 20-30 kilometers), and the magnitude of distributed strike slip deformation is unconstrained, but may be substantial. It is important to emphasize that evidence for large magnitude extension comes primarily from early Neogene volcanic and sedimentary sequences, and is easy to miss; Intense alteration and the complex stratigraphy of pre-Neogene rocks make faulting relations difficult to unravel and extensive cover by younger, gently-dipping volcanic and sedimentary rocks hide much of the evidence.

Existing geochronology on volcanic units that bracket increments of the extensional history indicates that almost all of the extension in Sonora occurred between ~25 and 10 Ma and that it generally youngs toward the Gulf. Similar conclusions regarding the general timing of extension were reached independently by Stewart and Roldan-Quintana (1994) and McDowell et al (submitted). Rapid, large magnitude extension occurred between ~ 26 to 20 Ma in the Santa Rosa area, ~20(?) to 13 Ma in the Suaqui Grande-Onabas area, and ~15-10 Ma along the coast. The only deformation in southern Sonora that is likely to be post ~10 Ma includes modest amounts (˛ 15% ) of ENE-WSW extension associated with slip on widely spaced, high angle normal faults, and possibly substantial amounts of dextral strike-slip on NW trending faults.

Discussion

Early (pre-transform) extension
Neither the large overall stretching factor nor the prolonged history of extension in southern Sonora are particularly surprising when placed in the context of the Basin and Range province as a whole (e.g., Gans et al, 1989). Precise timing constraints are still lacking in many areas, but the data presented here indicates that much of southern Sonora was highly extended during the Oligocene and early to middle Miocene, and only slightly extended during post-middle Miocene. Even accounting for uncertainties in the plate reconstructions (Stock and Molnar, 1989), much of this extension took place while subduction was still occurring to the west, and was broadly synchronous with the westward shift of arc(?) magmatism from the Sierra Madre Occidental to the area now occupied by the Gulf of California. In the Santa Rosa region, the relative timing of extensional tectonism, magmatism, and changes in the plate boundary at this latitude are now well understood (Fig. 9), and shed light on their relationship. The inception of rapid, large-magnitude extension in the Oligo-Miocene predates the termination of subduction at this latitude by at least 10 Ma. Indeed, nearly all of the extension occurred prior to the change in the plate boundary configuration, indicating that this change exerted minimal influence on local tectonism. Local magmatic activity occurred episodically throughout the Tertiary, but was most intense during the Laramide (> 60 to 50 Ma) and the Oligo-Miocene (27 to 17 Ma) (Fig. 9). Rapid extension followed a long period of apparent tectonic quiescence, as evidenced by very slow cooling of Laramide basement rocks and the lack of profound angular unconformities within the early to middle Tertiary sections. The inception of extension did, however, closely follow the inception of renewed magmatic activity (Fig. 9) and is compatible with the idea that mafic magmatism may have thermally weakened the underlying lithosphere and thereby triggered extension (e.g., Gans et al, 1989). This spatial and temporal association of mafic volcanism with the inception of Oligo-Miocene extension has been noted elsewhere in northern Mexico (e.g., Henry and Aranda Gomez, 1989; McDowell and others, submitted) and may suggest a general linkage between the two.

Whether this early extension should be viewed as fore-arc, intra-arc, back-arc or simply unrelated to the arc depends on how one defines the Oligocene to Miocene "arc". The apparent westward shift of the axis of volcanism from the Sierra Madre Occidental in the Oligocene to the Gulf of California region in the early to middle Miocene has been attributed to roll back and progressive steepening of the subducting slab (e.g., Coney and Reynolds, 1977). It is tempting to also attribute extensional tectonism during this time period to a similar mechanism. For example, trench retreat due to the roll back and sinking of the subducting oceanic lithosphere may have caused extension in the arc and back-arc region (e.g., Uyeda and Kanamoori, 1979). However, the slab that was being subducted beneath northwestern Mexico during the Oligocene and early Miocene was getting progressively younger (and more buoyant), as the spreading center between the Guadalupe Plate and the Pacific plate approached the trench. Another explanation for the westward shift of magmatic (and tectonic) activity during the Oligocene to middle Miocene is the retreat of the eastern limit of slab penetration beneath the continental margin (e.g., Severinghouse and Atwater, 1990). As younger and thinner oceanic lithosphere entered the subduction zone, it would require progressively less time to re-equilibrate to upper mantle temperatures. East of the limit of coherent slab, upwelling of asthenosphere may have thermally weakened and elevated the overlying continental lithosphere, thereby providing the impetus for extension and for generating basalts and their contaminated derivatives. It remains unclear to what extent the generation of middle to late Cenozoic magmas in northern Mexico was controlled by subduction versus unrelated processes in the underlying mantle.

Regardless of the mechanism, the magnitude of the westward shift of magmatic and tectonic activity has been greatly exaggerated by subsequent extensional and transform deformation in northwest Mexico. The present distance, measured perpendicular to the continental margin, from the base of the continental slope west of Baja California Sur, across southern Sonora to the axis of Oligocene volcanism in the Sierra Madre Occidental of western Chihuahua is about 650 km. In present geographic coordinates, this yields a slab dip of only 13ˇ if we assume that the depth to the top of the subducting slab beneath the Sierra Madre during the Oligocene was 150 km. However, if Baja California is restored southeastward to a reasonable pre-transform position, and ~ 150 km of NE-SW extension is removed from the Gulf extensional province (see reconstruction below), the arc-trench distance is decreased to only ~ 350 km and the implied Oligocene slab dip is increased to ł 25ˇ. Similarly, the locus of magmatism (and deformation) would need to shift westward no more than 100-150 km.

Evolution from a convergent to a transform plate boundary
A persistent question regarding the Cenozoic evolution of western North America is how did the change from a convergent plate margin to a transform margin influence the distribution, rates, and style of deformation in the overriding plate? Plate tectonic reconstructions, beginning with Atwater(1970) and revised and refined by many others, show the Pacific plate first making contact with the North American plate near the present-day latitude of Los Angeles at about 25 Ma, and then development of a growing transform boundary connected to triple junctions at each end. For northwest Mexico, reconstructions by Stock and Hodges (1989) and Lonsdale (1991) show that the southern (Rivera) triple junction migrated southward a total of 1700 km in a series of jumps related to ridge deaths and microplate capture between about 20 and ~11 Ma. The new transform boundary between the Pacific and North America plates is generally assumed to have initiated in the vicinity of the old trench - i.e., the San Benito-Tosco- Abreojos fault zone (Spencer and Normark, 1979) and maintained this position until about 5.5 Ma when the plate boundary moved into the Gulf of California (Lonsdale, 1991). As first pointed out by Spencer and Normark (1979), the orientation of this early transform is oblique to the late Miocene plate motion and they proposed that extension within the "proto-Gulf" might make up the missing component of the plate vector. This general concept was refined and quantified in a kinematic model by Stock and Hodges (1989), who proposed that between 10.6 and 5.5 Ma, the Pacific-North America relative plate motion was partitioned between NNW strike-slip displacement along the Tosco-Abreojos fault and 110 ± 80 to 160 ± 80 kilometers of distributed NE-SW extension within the Gulf extensional province. Subsequent to about 5 Ma, geologic and geophysical data from the Gulf of California and southern California suggests that Baja California has been attached to the Pacific plate and has moved 250-300 kilometers to the northwest along the Gulf of California- San Andreas transform system (Lonsdale, 1991).

A potential difficulty with the Stock and Hodges (1989) model is the need for fairly large magnitudes of NE-SW extension in the Gulf extensional province during the late Miocene. At the latitude of Guaymas, the reconstructed 5 Ma width of the extensional province is only about 300 kilometers, and 150 ± 80 km of NE-SW extension would correspond to 100% extension (30% to 330% extension at the 95% confidence limits) entirely within the preceding 5 Ma (a spreading rate of 3.0 ± 1.6 cm/yr.). Though local evidence for extension of this age certainly exists in the circum-Gulf region (Stock and Hodges, 1989, Lee et al, 1994), it appears to be generally of small magnitude and largely restricted to areas near the gulf . Evidence for late Miocene extension might lay largely hidden underwater in the southern Gulf of California, but in the northern part, fully 60-80% of the reconstructed 5 Ma width of the extensional province is exposed on land, and evidence for this magnitude of late Miocene extension should be obvious. The observations presented in this paper suggests that the magnitude of post-10 Ma extension across southern Sonora - nearly three quarters of the total width of the province - is probably no more than 20 kilometers. It seems highly unlikely that the missing 130 (± 80) kilometers of NE-SW extension could be absorbed within the remaining 70-80 km west of the Sonoran coast. It is possible that 10-5 Ma extension may also have occurred east of the Sierra Madre Occidental (e.g., Henry and Aranda-Gomez 1992), but existing structural and geochronologic constraints are poor and there is no compelling evidence for large magnitude late Miocene extension in this region.

A speculative new reconstruction for northwest Mexico
A fundamental assumption of existing kinematic models for proto-Gulf deformation is that the transform component of the Pacific-North America plate motion prior to ~ 5 Ma was taken up along the Tosco-Abreojos fault. Indirect evidence for the existence of this fault comes from bathymetric and shallow seismic records west of Baja California (Spencer and Normark, 1979), but there is little direct evidence that constrains either the timing or magnitude of displacement. An speculative new model for the late Cenozoic evolution of northwest Mexico is illustrated in Figure 10 and provides a possible solution to the dilemma of inadequate late Miocene NE-SW extension in the Gulf region. Prior to about 16 Ma, subduction was occurring beneath much of northwestern and central Mexico (Stock and Hodges, 1989). Voluminous magmatism and large scale NE-SW extension occurred inboard of this convergent plate boundary and both generally migrated towards the trench. At about 11 Ma, spreading ceased along a large segment of the Pacific-Guadalupe ridge and the triple junction jumped nearly 1000 km southward (Stock and Hodges,1989; Lonsdale, 1991). I speculate that at this time, both the captured remnant of the Guadalupe plate and Baja California began moving with close to Pacific plate motion (cf. Nicholson et al, 1994). In this scenario, 10 to 5 Ma transform motion between the Pacific and North American plate was accommodated largely by distributed shear along a system of NW-striking, en-echelon, strike slip faults inboard of Baja California, a situation not unlike the modern Gulf. This reconstruction implies that Baja California may have been displaced at least 500 kilometers to the northwest in the last 10 Ma and that the Tosco-Abreojos fault may never have accommodated a significant fraction of the Pacific-North America transform motion.

Implications
If the speculative tectonic history and reconstructions illustrated in Figure 10 are approximately correct, there are a number of important implications:
(1) First and foremost, it suggests that extension in northwest Mexico was initiated and mainly driven by body forces within the overriding plate of a convergent margin and was not directly kinematically linked to Pacific- North America plate motions. The evolution from a convergent to a transform plate boundary was manifested primarily by the inception of distributed strike slip faulting, a substantial reduction in the extensional strain rate, and perhaps a reorientation of the extension direction, from NE-SW (orthogonal to the trench) to more east-west. The spatial and temporal association of early extension with mafic volcanism in many areas is intriguingly similar to relations that have been described in the Basin and Range province (Gans et al, 1989) and suggests that extension may have been triggered by magmatic activity.
(2) The transfer of Baja California from the North American plate to the Pacific plate may have happened earlier than has been previously thought. There may be a fundamental difference in how a plate boundary evolves from convergence to transform motion if the ridge stops spreading before reaching the trench. In such a case, there is no a priori reason why the new transform must develop at the old trench. It may generally be easier to initiate the transform motion within the magmatically and mechanically weakened continental lithosphere- perhaps at the inner limit of stranded oceanic plate, rather than rupture relatively strong oceanic lithosphere at the outer rise. Stock and Hodges (1989) proposed that the plate motion was initially partitioned between outboard transform motion and inboard extension during proto Gulf transtensional deformation, perhaps analogous to the well established partitioning that occurs at obliquely convergent margins (Fitch, 1972). However, in oblique convergence, strain partitioning occurs in order to minimize the work done along the mechanically strong subduction interface. The strain partitioning proposed by Stock and Hodges (1989) for oblique extension appears to increase the work done by requiring the lithosphere to fail in two places rather than one. Put another way, if the gulf extensional province was sufficiently weak to fail in extension, why not in transtension?
(3) The speculative reconstruction presented here implies that Baja may have been displaced to the northwest as much as 500 kilometers in the last 10 Ma. Thus, prior to 10 Ma, the southern tip of Baja might lie adjacent to Puerto Vallarta and Tepic, and San Felipe would lie south of Hermosillo. This reconstruction requires that the continental shelf west of Nayarit consist of highly stretched continental crust and younger sediment so that it can collapse to almost nothing (Fig. 10). An important test of this reconstruction is how well the geology on either side of the gulf matches. Gastil and others (1981) concluded that there is only about 300 km of dextral strike slip across the northern Gulf of California, based on the correlation of distinctive pre-Middle Miocene conglomerate found near San Felipe and east of Isla Tiburon (Fig. 1, 8) If these tie points are valid, then perhaps the additional 200 km of strike slip deformation is distributed across Sonora to the east of the tie points. Strike slip faults of the correct orientation exist in Sonora, but there are virtually no constraints on their magnitude of slip nor on their timing, except that they appear to be largely post 10-12 Ma. Another test of this reconstruction might be to see if distinctive Oligocene ignimbrites in the La Paz region (Hausback, 1984) are present on the mainland midway between Mazatlan and Tepic. This more southerly position for Baja California prior to 10 Ma helps account for arc magmatism persisting along the northeastern side of the peninsula up until 16 Ma (Lee et al, 1996) because northern Baja could still lie inboard of a trench at that time (Fig. 10). Previous reconstructions (e.g., Stock and Hodges, 1989) generally show subduction shutting off up to several million years before arc magmatism ceased for most of the northern half of the peninsula.
(4) All of the postulated 500 km of right lateral displacement, including about 250 km that is pre-5 Ma would have to be taken up to the north - presumably along the southern San Andreas transform system. This may be untenable. The cumulative amount of right lateral shear in southern California and how it was distributed in time and space remain controversial (see Dickinson, 1996 for a review). There is evidence for limited slip on the San Andreas system prior to 5 Ma (e.g., Humphreys and Weldon, 1991;) but whether the data permit sufficient right lateral shear between 10 and 5 Ma and how this strain might have been partitioned between slip on known faults and internal deformation and rotation of fault bounded blocks is beyond the scope of this paper. Another possibility is that much of the inferred pre-5 Ma right lateral motion was accommodated by the Eastern California Shear Zone and the Walker Lane Belt of western Nevada (Jack Stewart, 1996, personal communication).

Additional studies are needed that focus on the timing and magnitude of deformation in northwestern Mexico and a rigorous matching of the geology on either side of the gulf. The most important conclusions to be drawn from the data presented in this paper are that much of northwestern Mexico was highly extended and that most of this extension occurred during the Late Oligocene to Middle Miocene while subduction was ongoing. The inception of rapid, large-magnitude extension was often associated with voluminous mafic to intermediate magmatic activity, suggesting that the input of heat from the mantle may have triggered extensional collapse. An interesting corollary of these conclusions is that it is difficult to come up with sufficient extension that post dates subduction to support the strain partitioning model of Stock and Hodges (1989). Finding the locus of 200-300 km of NW-oriented dextral shear that must have occurred between the Pacific and North American plates during the late Miocene remains an outstanding problem. The choices now are either to hide this displacement offshore along the unfavorably-oriented Tosco-Abreojos lineament and then muster evidence for fairly large magnitude NE-SW extension of this age, or to find this displacement within the proto-Gulf province - perhaps as a system of more favorably oriented en echelon strike slip faults and oblique normal faults (Fig. 10). In either case, the definitive test of these ideas will have to come from the geology.

Acknowledgments
This opportunity to investigate the extensional history of southern Sonora was made possible through the financial and logistical support of Magma Copper Company -now part of Broken Hill Proprietary. In particular, I would like to thank John-Mark Staude, Eric Seedorff, and Vernon Deryter for their encouragement and enthusiastic support. Elizabeth O'Black superbly drafted the illustrations . I thank Tanya Atwater, Andy Calvert, Jeff Lee, Ralph Levy, John-Mark Staude, and Doug Wilson for informal reviews and discussions and Fred McDowell for sharing some of his insights and a preprint of his paper. Thorough and helpful technical reviews were provided by Joann Stock and Jack Stewart.

REFERENCES CITED

Phil's home Phil's Home