The channel-flow model for the Greater Himalayan Sequence (GHS) of the Himalayan orogen involves a partially molten, rheologically weak, mid-crustal layer “flowing” southward relative to the upper and lower crust during late Oligocene–Miocene. Flow was driven by topographic overburden, underthrusting, and focused erosion. We present new structural and thermobarometric analyses from the GHS in the Annapurna-Dhaulagiri Himalaya, central Nepal; these data suggest that during exhumation, the GHS cooled, strengthened, and transformed from a weak “active channel” to a strong “channel plug” at greater depths than elsewhere in the Himalaya. After strengthening, continued convergence resulted in localized top-southwest (top-SW) shortening on the South Tibetan detachment system (STDS). The GHS in the Annapurna-Dhaulagiri Himalaya displays several geological features that distinguish it from other Himalayan regions. These include reduced volumes of leucogranite and migmatite, no evidence for partial melting within the sillimanite stability field, reduced structural thickness, and late-stage top-southwest shortening in the STDS. New and previously published structural and thermobarometric constraints suggest that the channel-flow model can be applied to mid-Eocene–early Miocene mid-crustal evolution of the GHS in the Annapurna-Dhaulagiri Himalaya. However, pressure-temperature-time (PTt) constraints indicate that following peak conditions, the GHS in this region did not undergo rapid isothermal exhumation and widespread sillimanite-grade decompression melting, as commonly recorded elsewhere in the Himalaya. Instead, lower-than-typical structural thickness and melt volumes suggest that the upper part of the GHS (Upper Greater Himalayan Sequence [UGHS]—the proposed channel) had a greater viscosity than in other Himalayan regions. We suggest that viscosity-limited, subdued channel flow prevented exhumation on an isothermal trajectory and forced the UGHS to exhume slowly. These findings are distinct from other regions in the Himalaya. As such, we describe the mid-crustal evolution of the GHS in the Annapurna-Dhaulagiri Himalaya as an atypical example of channel flow during the Himalayan orogeny.
The kinematic and metamorphic evolution of the metamorphic core of the Himalayan orogen (Fig. 1), referred to as the Greater Himalayan Sequence (GHS), is the central focus of all models of Himalayan orogenesis (e.g., Grujic et al., 1996; Beaumont et al., 2001; Bollinger et al., 2006; Robinson et al., 2006; Searle et al., 2006; Kohn, 2008; Mukherjee, 2013b; He et al., 2014; Cottle et al., 2015; Frassi, 2015; Montomoli et al., 2015). The channel-flow model for the Himalayan orogen proposes that the GHS represents a rheologically weak, partially molten, mid-crustal channel that flowed laterally southward, between the upper and lower bounding rigid crust. Flow was driven by lithostatic loading by the Tibetan plateau, underthrusting of the Indian lower crust, and focused erosion at the orogenic front (Fig. 2) (Beaumont et al., 2001; Godin et al., 2006a; Harris, 2007). Extrusion and exhumation of the channel were facilitated by coeval shearing along the top-to-the-south (top-S) Main Central thrust zone (MCTZ) and top-to-the-north (top-N) South Tibetan detachment system (STDS), which bound the GHS below and above, respectively (see reviews by Godin et al., 2006a; Grujic, 2006).
Many of the geological and geophysical constraints on which the channel-flow model is based (e.g., pressure-temperature [PT] conditions and crustal thicknesses) are derived from the Everest, Sikkim, and Bhutan regions (e.g., Grujic et al., 1996, 2002) of the central-eastern Himalaya (Fig. 1A). In these regions, the channel-flow model provides a robust explanation for the kinematic, dynamic, and temporal evolution of the GHS (e.g., Nelson et al., 1996; Beaumont et al., 2001; Searle and Szulc, 2005; Unsworth et al., 2005; Searle et al., 2006; Streule et al., 2010). However, the lithologic, structural, and metamorphic framework of the GHS varies along the ∼2500 km length of the orogen. It thus remains unclear the extent to which the channel-flow model can be applied along the whole orogen (Godin et al., 2006a; Harris, 2007).
In this study, the applicability of the channel-flow model to the GHS in the Annapurna-Dhaulagiri Himalaya in central Nepal (Fig. 1) is tested through a synthesis of new structural and thermobarometric analyses of the metamorphic rocks, combined with previously published geochronometric and thermobarometric constraints. Field observations from the Modi Khola and Kali Gandaki valleys (Fig. 1B) reveal geological features of the GHS that are atypical when compared to other Himalayan regions along strike. These include reduced volumes of leucogranite and migmatite, an absence of evidence for partial melting within the sillimanite stability field, reduced structural thickness, and late-stage top-SW shortening on the STDS following cessation of top-NE extensional shearing. The presence of such features has significant implications for the rheology of the GHS and raises the question of whether or not the GHS in this region was weak enough for mid-crustal flow.
1.2. Channel Flow during the Himalayan Orogeny
Channel flow describes the laminar flow of a viscous fluid between upper and lower rigid plates (Turcotte and Schubert, 2002). Flow can be driven by horizontal pressure gradients (Poiseuille flow) and/or motion of the channel walls (Couette flow). When a pressure gradient and channel-wall motion are applied in opposite directions, a hybrid flow may form (Grujic, 2006) in which a portion of the channel flows in the opposite direction to the moving channel wall (the “return flow” of Mancktelow, 1995).
In applying the channel-flow model to the Himalaya, it is proposed that the GHS is an exhumed portion of a mid-crustal channel (Beaumont et al., 2001; Godin et al., 2006a; Harris, 2007). After initial collision between India and Eurasia, crustal thickening and heating followed by widespread partial melting at mid-crustal levels resulted in the development of a weak, low-viscosity channel located between rigid upper and lower crust (Beaumont et al., 2001; Godin et al., 2006a). Upon weakening below a threshold viscosity, southward return flow of the mid-crustal channel initiated. This flow was driven by a horizontal gradient in lithostatic pressure produced by the relative topographic elevations and crustal thicknesses of the Tibetan plateau and Indian continent, and from shear stresses due to the northward underthrusting of the lower Indian continental crust (Beaumont et al., 2001, 2004; Grujic, 2006). Focused erosion aided exhumation of the channel to the orogenic front (Beaumont et al., 2001). Relatively low viscosities were maintained during exhumation due to continued partial melting along an isothermal exhumation path (Harris and Massey, 1994; Streule et al., 2010; Jamieson et al., 2011; Searle, 2013).
Application of the channel-flow model to the Himalayan orogen places a crucial dependence on widespread partial melting of the mid-crust for the required reduction in viscosity (Beaumont et al., 2001, 2004; Grujic, 2006; Jamieson et al., 2011). Finite element thermomechanical models of channel flow simulate this strength drop by changing the rheology of elements within the modeled crust from a stress- and temperature-dependent power-law rheology based on experimentally derived flow laws (e.g., Blackhill Quartzite; Gleason and Tullis, 1995) to a “melt weakened” linear viscous rheology with a viscosity of 1019 Pa s once the temperature exceeds 700 °C (Beaumont et al., 2001, 2004; see also experimental verification of assumed viscosities by Rutter et al., 2011). This imposed reduction in viscosity simulates the order of magnitude drop in rheological strength expected to occur during widespread partial melting of the crust (Rosenberg and Handy, 2005; Rosenberg et al., 2007). Experimental deformation of rocks shows that an increase in melt fraction from 0.00 to 0.07 (i.e., 7%) is accompanied by a ∼90% reduction in strength, referred to as the melt-connectivity transition (MCT, Rosenberg and Handy, 2005). An intermediate strength drop is also recorded between a melt fraction of 0.20 and 0.50, which corresponds to the solid-liquid transition (SLT; Rosenberg and Handy, 2005). Consequently, determining the extent and duration of partial melting within the GHS is crucial for validating the channel-flow model for the Himalayan orogen.
The extent to which channel-flow models are representative of the whole of the Himalaya has also been questioned (Godin et al., 2006a; Harris, 2007). The original thermomechanical channel-flow model (Beaumont et al., 2001) was based on seismic reflection and magneto-telluric data from the INDEPTH transect (Nelson et al., 1996), conducted across the southern part of the Tibetan plateau to the north and east of Sikkim, India (Fig. 1). The channel was modeled as a homogeneous layer; whereas the GHS comprises a lower portion of greenschist-facies quartzites, marbles, and metapelitic rocks and an upper portion of amphibolite-facies pelitic, psammitic, and calc-silicate paragneisses and schists, orthogneiss, and migmatites. While lithologies in individual tectonostratigraphic units within the GHS are remarkably constant along the length of the Himalaya, the relative proportions of these lithologies do vary (Le Fort, 1975; Upreti, 1999; Yin, 2006) and are likely to have produced orogen-parallel variations in viscosity during mid-crustal evolution. Additionally, thermomechanical channel-flow models suggest that a minimum channel thickness of 10–20 km is required for crustal flow to be an effective form of extrusion, although this thickness may change slightly for different model resolutions (Beaumont et al., 2004; Jamieson et al., 2004, 2006). The thickness of the GHS varies along the length of the orogen, and it remains unclear whether sections with a structural thickness of <10 km are too thin to have pervasively deformed by channel flow (Beaumont et al., 2004; Godin et al., 2006a). In order to test applicability of the model to the whole orogen, it is crucial, therefore, to assess the validity of the channel-flow model in regions of the Himalaya that differ from those where the model is best suited.
1.3. The Annapurna-Dhaulagiri Himalaya
The Himalayan orogeny initiated at ca. 50 Ma during final closure of Neotethys (Searle, 1986; Green et al., 2008; Najman et al., 2010) (Fig. 1A). Since this time, the Himalayan belt has been under a continuous state of convergence, resulting in uplift and erosion of the world’s highest mountain peaks (e.g., Searle et al., 2011; Avouac, 2015; Searle, 2015). In central Nepal, the Himalaya can be divided into four tectonic units separated by orogen-parallel faults and shear zones (Fig. 1B). From SW to NE, these units are the Subhimalayan Zone (SZ), the Lesser Himalayan Sequence (LHS), the Greater Himalayan Sequence (GHS), and the Tethyan Himalayan Sequence (THS) (Le Fort, 1975; see extensive reviews by Dhital, 2015).
The central Himalaya in Western Region, Nepal, is dominated by the peaks of Annapurna I (8091 m) and Dhaulagiri (8167 m) (Fig. 1B). Access through the region is gained along the NE-SW–trending Modi Khola and Kali Gandaki valleys and the foothills between. The geology of the region was first described by Le Fort (1975), was subsequently mapped in detail (Colchen et al., 1981; Hodges et al., 1996; Godin, 2003; Martin et al., 2010; Searle, 2010; Parsons et al., 2016), and is the subject of numerous detailed structural, thermobarometric, and geochronometric studies (Bouchez and Pêcher, 1981; Arita, 1983; Le Fort et al., 1986; Pêcher, 1989; Brown and Nazarchuk, 1993; Nazarchuk, 1993; Kaneko, 1995; Hodges et al., 1996; Vannay and Hodges, 1996; Godin et al., 1999a, 1999b, 2001; Godin, 2003; Bollinger et al., 2004; Martin et al., 2005; Paudel and Arita, 2006; Kellett and Godin, 2009; Larson and Godin, 2009; Martin et al., 2010; Corrie and Kohn, 2011; Kohn and Corrie, 2011; Carosi et al., 2015; Iaccarino et al., 2015; Larson and Cottle, 2015; Martin et al., 2015).
2. TECTONOSTRATIGRAPHY AND STRUCTURAL FRAMEWORK
In the Annapurna-Dhaulagiri Himalaya (Plates 1 and 2), the GHS is bound below by the Main Central Thrust (MCT) and above by the South Tibetan detachment (STD) and is divided here into the Lower GHS (LGHS), Upper GHS (UGHS), and South Tibetan detachment system (STDS). The LGHS and UGHS are separated by the Chomrong thrust (CT), while the UGHS and STDS are separated by the Annapurna detachment (AD) in the Kali Gandaki valley and the Deurali detachment (DD) in the Modi Khola valley. Internally, the UGHS is deformed by the Kalopani shear zone (KSZ) in the Kali Gandaki valley and the Modi Khola shear zone (MKSZ) in the Modi Khola valley (Hodges et al., 1996; Vannay and Hodges, 1996). These structural definitions are in accordance with those proposed by Searle et al. (2008) and used by Larson and Godin (2009) for the same region but differ from other studies in which the CT, MCT, and LGHS as defined here, are referred to as the MCT, Ramgarh thrust, and Lesser Himalayan rocks, respectively (Martin et al., 2005; Corrie and Kohn, 2011; Carosi et al., 2015; Iaccarino et al., 2015). Additionally, the CT and MCT as defined here have also been referred to as the MCT I and MCT II, respectively (Arita, 1983; Searle and Godin, 2003), and the LGHS has also been referred to as the Main Central thrust zone (MCTZ) (Grasemann et al., 1999) and the Lesser Himalaya crystalline series (LHCS) (Caddick et al., 2007). The Tethyan Himalayan Sequence (THS) overlies the GHS and is bound below by the STD and above by the Indus-Yarlung suture zone (IYSZ), north of the study area. Detailed geological maps and cross sections through the region are displayed in Plates 1 and 2 and Figures 3 and 4. Table 1 summarizes six fabric generations observed in the GHS and THS across the region. These definitions are based upon the foliation classification defined by Godin (2003) for deformation fabrics in the UGHS, STDS, and THS in the Kali Gandaki valley. The orientations of these fabrics are summarized in Table 1 and Figure 5. The full structural data set is included in Supplemental File 11.
2.1.1. Lower Greater Himalayan Sequence (LGHS)
The LGHS has a structural thickness of ∼5700–6650 m. At its base, the MCT forms a 10–20-m-thick top-SW shear zone (Plates 1 and 2). The lowermost portion of the LGHS comprises a unit of undifferentiated quartzites, pelites, and semipelites (Unit A; Plates 1 and 2), plus a laterally discontinuous layer of feldspathic orthogneiss (Plate 2—Ulleri augen gneiss; Le Fort, 1975). Above this, Unit B consists of quartzite and dolomitic marble (marble defined as >80% carbonate). Unit C, at the top of the LGHS, consists of interbedded dolomitic metacarbonates (metacarbonate defined as >50%–80% carbonate), metamarls, and metapelitic rocks.
Regional foliation in the LGHS (S3) dips NE, with NE-plunging mineral lineations (L3) (Fig. 5), and transposes an earlier crenulation cleavage. Top-SW shear-sense indicators associated with S3 are frequently developed. A later E-W–striking subvertical crenulation cleavage (S4) and N-S–striking normal fault and fracture set (S5) are sometimes observed (Table 1).
2.1.2. LGHS Metamorphic Constraints (Previous Work)
An up-section increase in temperature in the LGHS from ∼400 °C to ∼500–650 °C has been determined from Raman spectroscopy of carbonaceous material (RSCM, Beyssac et al., 2004), quartz c-axis fabric opening angle thermometry (Larson and Godin, 2009), and garnet-biotite and garnet-ilmenite thermometry (Le Fort et al., 1986; Kaneko, 1995; Vannay and Hodges, 1996; Martin et al., 2010; Corrie and Kohn, 2011). Garnet-plagioclase barometry from Unit C yields pressures of ∼7–12 kbar (Le Fort et al., 1986; Vannay and Hodges, 1996; Martin et al., 2010; Corrie and Kohn, 2011).
2.2.1. Upper Greater Himalayan Sequence (UGHS)
The amphibolite-facies UGHS has a structural thickness of ∼6500–7150 m and is divided into Unit I, Unit II, and Unit III (Plates 1 and 2). At the base of the UGHS, the Chomrong thrust (CT) forms a 50–100-m-thick top-SW shear zone comprising quartz mylonites and metapelitic horizons with a well-defined S-C fabric (Fig. 6E).
The lowermost unit in the UGHS, Unit I, is 3300–3900 m thick and consists of kyanite-grade schists, gneisses, migmatites, and leucogranites. Migmatites typically concentrate into foliation-parallel bands ∼10–50 m thick, while leucosomes and small leucogranite bodies (<1 m thick) are present throughout Unit I. Where present, migmatite comprises 10%–40% leucosome in the Modi Khola valley and 5%–20% leucosome in the Kali Gandaki valley. Bulk mineral assemblages are dominated by biotite, quartz, plagioclase ± garnet ± kyanite ± K-feldspar. Kyanite is first observed at ∼400–450 m above the unit base. Garnet commonly contains inclusions of quartz, plagioclase, biotite, and rutile. In some cases, garnet cores contain ilmenite inclusions with rims containing rutile inclusions. Rutile inclusions are also observed within the largest kyanite crystals. Muscovite is sporadically observed, most commonly as a secondary mineral phase but is also identified as a primary phase at the very base of Unit I in the Kali Gandaki transect. Secondary paragonite is also found in samples from the lowermost Unit I within the Modi Khola transect. Sillimanite has not been observed in Unit I during this study. Carosi et al. (2015) identified sillimanite as rare microscopic fibrolitic needles within fractures in kyanite, and at the rims of biotite and muscovite in contact with garnet and at plagioclase grain boundaries.
Unit II is 2250–2500 m thick and dominated by calc-silicate gneisses that comprise variable proportions of quartz, calcite, clinopyroxene, plagioclase, K-feldspar, phlogopite, scapolite, clinozoisite, rutile, and titanite. Amphibole is also observed in some retrogressed samples. Subordinate kyanite paragneisses have the same mineral assemblages as in Unit I. Leucosomes within Unit II are rare and typically <10 cm thick and <50 cm long. Leucogranite sills are seen in the uppermost layer of Unit II at the top of the UGHS in the Modi Khola valley (Plate 2).
Unit III is 700–1000 m thick and dominated by orthogneiss and leucogranite containing albite, K-feldspar, quartz, biotite, and muscovite. In this study, garnet was only observed in orthogneiss from the Modi Khola transect. Larson and Godin (2009) reported garnet in leucogranites from Unit III in the Kali Gandaki transect. Subordinate pods of calc-silicate gneisses containing plagioclase, biotite, clinozoisite, amphibole, and quartz are observed at the top of Unit III in the Kali Gandaki valley. Previous studies have also reported subordinate pods of sillimanite-biotite-gneiss with kyanite-bearing leucosomes (Arita, 1983; Godin et al., 2001; Larson and Godin, 2009). The top of Unit III is bound by the Annapurna detachment in the Kali Gandaki transect (Plate 1 and Fig. 3), and the Deurali detachment in the Modi Khola transect (Plate 2 and Fig. 4).
Leucogranite bodies across the UGHS are no larger than tens of meters in thickness, and injection complexes are only observed in the Modi Khola transect and are less densely packed with dikes and sills than typically observed in other regions (e.g., Searle et al., 2003).
Regional foliation (S3) dips approximately NE, with mineral lineation trends ranging between NW and E (Fig. 5). A general parallelism observed between the orientation of the regional S3 foliation, C-planes of localized S-C fabrics (sensu Berthé et al., 1979), and the major bounding shear zones suggests that S3 is a shear-related transpositional fabric, as observed in the LGHS and STDS (see below). Top-SW S3 shear indicators are observed throughout the UGHS (Figs. 6A, 6C, and 6D). The upper portion of the UGHS is deformed by the top-SW Kalopani shear zone (KSZ; Vannay and Hodges, 1996) and Modi Khola shear zone (MKSZ; Hodges et al., 1996) in the Kali Gandaki (Plate 1) and Modi Khola valleys, respectively. Subordinate S4 deformation structures associated with deformation on the KSZ and MKSZ are observed locally across the UGHS, deforming the regional S3 foliation (Plate 1 and Fig. 6D). Late-stage, N-S–striking normal faults and fracture sets (S5) are observed in the Kali Gandaki transect (Fig. 5).
2.2.2. UGHS Metamorphic and Geochronometric Constraints (Previous Work)
The UGHS in the Annapurna-Dhaulagiri Himalaya equilibrated within the kyanite stability field (Carosi et al., 2015). No evidence has been found for partial melting within the sillimanite stability field (Nazarchuk, 1993; Hodges et al., 1996; Larson and Godin, 2009). In the Modi Khola transect, peak metamorphic temperatures and pressures of 750–825 °C and 11–14 kbar are recorded (garnet-biotite cation exchange [GARB] thermometry—Kaneko, 1995; GARB and garnet-ilmenite cation exchange thermometry, garnet-plagioclase-quartz-muscovite (GPMQ), and garnet-plagioclase-aluminosilicate-quartz (GASP)–phase equilibria barometry—Martin et al., 2010; GARB and Zr-in-titanite thermometry, GPMQ, GASP, and garnet-plagioclase-biotite-quartz (GPBQ)–phase equilibria barometry—Corrie and Kohn, 2011).
A range of temperatures and pressures is recorded from the UGHS in the Kali Gandaki transect. Iaccarino et al. (2015) recorded peak temperatures and pressures of 710–720 °C and 10–11 kbar from garnet-muscovite-biotite-ilmenite multiphase equilibria thermobarometry and Zr-in-rutile thermometry. Le Fort et al. (1986) recorded temperatures and pressures of ∼650–700 °C and 5–10 kbar from GARB-GASP thermobarometry. Vannay and Hodges (1996) recorded temperatures and pressures of ∼500–750 °C and 6–12 kbar from GARB and garnet-hornblende cation exchange thermometry and GASP, garnet-rutile-aluminosilicate-ilmenite-quartz, garnet-plagioclase-muscovite-biotite, and garnet-amphibole-plagioclase-quartz phase equilibria barometry. It is likely that the use of different thermobarometer calibrations and analytical methods to select appropriate element compositions has contributed to some of the differences in pressure and temperature estimates between these studies (e.g., Hodges and McKenna, 1987; Kohn and Spear, 2000; Kohn, 2014). Quartz c-axis fabric opening angle thermometry (Kruhl, 1998) from Unit I in the Kali Gandaki valley records a deformation temperature of 670 ± 50 °C (Larson and Godin, 2009).
U-Pb zircon and monazite geochronology and titanite thermochronology in the Annapurna-Dhaulagiri Himalaya indicate prograde metamorphism of the UGHS at temperatures exceeding 700 °C, initiated at 48–43 Ma (Carosi et al., 2015; Iaccarino et al., 2015; Larson and Cottle, 2015). Peak metamorphic temperatures were attained between 35 and 30 Ma in Unit III and 28–20 Ma in Unit I and Unit II (Corrie and Kohn, 2011; Kohn and Corrie, 2011; Iaccarino et al., 2015). In Unit I in the Kali Gandaki valley, retrogression from peak conditions began as early as 28 Ma and reached 650–670 °C and 7–8 kbar between 25 and 18 Ma (Iaccarino et al., 2015). This equates to a vertical exhumation from ∼30–33 km to ∼21–24 km at a time-averaged rate of 0.6–4.3 mm yr–1. In the Modi Khola valley, Corrie and Kohn (2011) recorded retrogression of Unit I between 17 and 22 Ma. During metamorphism, partial melting initiated as early as 41 Ma and continued to as late as 18.5 Ma (Nazarchuk, 1993; Hodges et al., 1996; Godin et al., 2001; Corrie and Kohn, 2011; Kohn and Corrie, 2011; Carosi et al., 2015; Larson and Cottle, 2015; Iaccarino et al., 2015).
Crystallization ages of undeformed leucogranite dikes suggest that motion on the CT occurred before 22–18.5 Ma (Nazarchuk, 1993; Hodges et al., 1996). Uniform 40Ar/39Ar muscovite cooling ages from across the UGHS suggest that the sequence was extruded and exhumed to the topographic surface as a coherent single block at ca. 16–12 Ma (Vannay and Hodges, 1996; Godin et al., 2001; Martin et al., 2015).
2.3. South Tibetan Detachment System (STDS)
The STDS has a structural thickness of ∼1700–2200 m and forms the uppermost part of the GHS, consisting of metacarbonates, marble, and subordinate calc-silicate gneiss. Marbles and metacarbonates within the STDS are dominated by calcite and contain variable proportions of quartz, plagioclase, and mica ± K-feldspar ± dolomite. Calc-silicate gneiss at the base of the STDS is composed of clinopyroxene, amphibole, clinozoisite, plagioclase, K-feldspar, quartz, biotite, and calcite ± rutile ± titanite. Leucogranite sills and dikes are found in the lower to middle portions of the STDS. At the base of the STDS, the top-down-E Annapurna detachment (Plate 1, AD—Kali Gandaki valley, Brown and Nazarchuk, 1993) and the top-down-NE Deurali detachment (Plate 2, DD—Modi Khola valley, Hodges et al., 1996) form foliation-parallel, normal-sense mylonitic shear zones with deformed, foliation-parallel leucogranite sills (Fig. 7). At the top of the STDS, the STD forms a more discrete top-E to top-NE brittle-ductile normal-sense detachment between the GHS and THS (Plates 1 and 2; Burg et al., 1984).
The S3 foliation dips to the ENE and NE in the Kali Gandaki and Modi Khola valleys, respectively (Fig. 5). The L3 stretching lineation plunges E and NE in the Kali Gandaki and Modi Khola valleys, respectively (Fig. 5). The S3-parallel transposed calcite veins are common across the STDS. In the Kali Gandaki valley, S3 foliation is locally deformed by S4 foliation (Plate 1 and Fig. 5C) and folds with S3-associated normal (top-E—F3) and S4-associated reverse (top-SW—F4) senses of vergence (Plate 1; Figs. 7 and 8) are observed. Leucogranite intrusions in the STDS in the Kali Gandaki valley are also deformed by F4 folding (Fig. 8). In addition, deformation microstructures from the STDS, including rotated porphyroblasts and grain-shape–preferred orientations, record both top-E (S3) and top-SW shearing (S4) (Fig. 9). The youngest deformation in the STDS is defined by a set of N-S–striking, subvertical normal faults and fractures (S5).
U-Pb geochronology of deformed and undeformed leucogranites suggests that the DD was active at ca. 22.5 Ma (Hodges et al., 1996) and that motion on the AD had ceased by ca. 22 Ma (Godin et al., 2001). In the Modi Khola valley, the STD was active after 18.5 Ma but may have initiated before this time (Hodges et al., 1996).
2.4. Tethyan Himalayan Sequence (THS) and Mid-Miocene to Recent Sedimentation
The THS is bound by the STD (Plates 1 and 2) to the south and the Indus-Yarlung suture zone (IYSZ) to the north (Searle, 2010). The IYSZ is exposed in southern Tibet, north of the study area. In the Annapurna-Dhaulagiri Himalaya, the THS comprises weakly metamorphosed (chlorite-grade or lower) Cambro-Ordovician to Cretaceous limestones, dolostones, and marls belonging to the Annapurna, Sanctuary, Nilgiri, Sombre and Thini Chu, and Lake Tilicho formations (Plates 1 and 2). The full stratigraphic framework of the THS is described in detail by Gradstein et al. (1992), Garzanti (1999), and Godin (2003).
The structure of the THS in the Kali Gandaki valley has been extensively studied by Godin (2003) and is summarized in Table 1. Within the upper Kali Gandaki valley, the Thakkhola graben has deformed the THS since mid-Miocene times during E-W extension (Plate 1 and Fig. 3) and is responsible for the subvertical, N-S–striking S5 foliation recorded in the THS and GHS (Hurtado et al., 2001; Garzione et al., 2003; Godin, 2003). Basin fill within the Thakkhola graben is mid-Miocene to Plio-Pleistocene in age (Garzione et al., 2003; Adhikari and Wagreich, 2011; Baltz, 2012).
3. NEW THERMOBAROMETRIC CONSTRAINTS FOR THE GHS IN THE KALI GANDAKI VALLEY
New estimates of peak metamorphic temperatures and pressures for the UGHS in the Kali Gandaki valley have been determined via GARB and Zr-in-titanite thermometry and GASP barometry. All analyses were conducted using wavelength dispersive X-ray spectroscopy with a JEOL 8230 electron microprobe at the University of Leeds using Probe for EPMA (Probe Software, Inc.). A description of methods used during data acquisition and the analytical conditions for both analyses are provided in Supplemental File 22. The microprobe analysis data sets are included in Supplemental File 33.
Three samples were selected for thermobarometry from Unit I of the UGHS in the Kali Gandaki transect (P12/055, P12/058, and P12/060; see Plate 1 for sample locations). All samples contain garnet, biotite, plagioclase, kyanite, and quartz. Sample P12/060 also contains muscovite. Peak temperature and pressure estimates are based on major-element compositions of garnet, biotite, and plagioclase, calculated simultaneously using the “Thermobarometry With Estimation of EQUilibrium state” method (TWEEQU; Berman, 1991) with the winTWQ v2.32 software (Berman, 2007). This multi-equilibrium approach uses a single internally consistent thermodynamic database (Berman and Aranovich, 1996) to find the point at which multiple phase reactions are at equilibrium (i.e., the intersection between multiple reaction curves in PT space) and provides more robust PT estimates than using individual calculations for pressure and temperature.
It is necessary to consider the effects of reequilibration and retrograde net transfer reactions on garnet, biotite, and plagioclase element compositions when calculating peak PT estimates (e.g., Frost and Chacko, 1989; Kohn and Spear, 2000). Peak mineral compositions were determined using the methods outlined by Kohn et al. (1992), Dasgupta et al. (2004), and Corrie and Kohn (2011) (see Supplemental File 2 [footnote 2] for method details). Samples P12/055, P12/058, and P12/060 produced peak temperature estimates of 747 ± 10 °C, 703 ± 13 °C, and 688 ± 19 °C, respectively (Table 2). These samples respectively yielded peak pressure estimates of 11.4 ± 0.9 kbar, 11.5 ± 1.3 kbar, and 11.3 ± 1.2 kbar.
Four samples of calc-silicate gneiss (P12/053; P12/054–Unit II; P13/032–Unit III; P13/031–STDS; see Plate 1 for sample locations) were selected for Zr-in-titanite thermometry (Table 3). All samples contained calcite, zircon, quartz, and rutile (see Table 3 for full mineral assemblage of each sample). Samples P13/031 and P13/032 were sampled from immediately above and below the AD, respectively (Plate 1). Temperatures were calculated using the Zr-in-titanite thermometer of Hayden et al. (2008) from average titanite Zr concentrations (ppm) in individual samples. The highest temperatures in P12/053 were calculated from rim concentration of Zr and are assumed to reflect peak metamorphic conditions (Kohn and Corrie, 2011). Titanite rim and core temperatures from P13/031 and P13/032 were within error of each other and only titanite core temperatures were calculated for P12/054. These may reflect crystallization temperatures rather than peak metamorphic temperatures and thus provide a minimum bound for peak metamorphic conditions. For a pressure of 11 ± 1 kbar (based on GASP barometry from this study, plus Corrie and Kohn, 2011; Iaccarino et al., 2015), peak temperature estimates of 740 ± 35 °C, 772 ± 33 °C, 744 ± 38 °C, and 720 ± 48 °C are calculated for samples P12/053, P12/054, P13/032, and P13/031, respectively (Table 3).
Our GARB-GASP thermobarometry results are similar to the results of Iaccarino et al. (2015), who record peak temperatures and pressures of 710–720 °C and 10–11 kbar from multi-equilibrium thermobarometry and Zr-in-rutile thermometry of a kyanite-migmatite in Unit I in the Kali Gandaki transect (Fig. 10A). Our Zr-in-titanite thermometry results from Unit II, Unit III, and the STDS are similar to previous temperature estimates from Unit II and Unit III in the Modi Khola valley (750–850 °C, Fig. 10B; Martin et al., 2010; Corrie and Kohn, 2011). Differences in results between this study and Le Fort et al. (1986) and Vannay and Hodges (1996) most likely derive from the use of different analytical procedures and thermobarometric calibrations to those used in this study. Vannay and Hodges (1996) analyzed the rims of garnet-biotite and garnet-plagioclase grain pairs that were in contact with each other, in accordance with the methods of Hodges and McKenna (1987) and Hodges et al. (1993). As stated by Vannay and Hodges (1996) and Hodges et al. (1993), the purpose of their employed analytical method was to determine the temperature and pressure at which these minerals equilibrated for the last time, and this is unlikely to yield PT estimates representative of peak conditions. Le Fort et al. (1986) also use different thermobarometers, and it is not clear what analytical methods they used to select element compositions for PT calculations.
4. ATYPICAL GEOLOGICAL FEATURES OF THE ANNAPURNA-DHAULAGIRI HIMALAYA AND THEIR KINEMATIC-DYNAMIC IMPLICATIONS
The GHS in the Annapurna-Dhaulagiri Himalaya differs from more typical GHS sections elsewhere (e.g., Everest-Makalu Himalaya and Sikkim Himalaya) upon which the channel-flow model was based, with atypically low volumes of leucogranite and migmatite, an absence of evidence for partial melting within the sillimanite stability field, a smaller structural thickness (<10 km), and evidence for late-stage top-SW shortening on the STDS (Brown and Nazarchuk, 1993; Nazarchuk, 1993; Hodges et al., 1996; Godin et al., 1999a; Godin, 2003; Larson and Godin, 2009; Parsons et al., 2016). Additionally, alternative models of underplating and thrust stacking or composite channel-flow–underplating models have been proposed for mid-crustal emplacement of the GHS in the Annapurna-Dhaulagiri Himalaya, following identification of several metamorphic discontinuities within the GHS (Martin et al., 2010; Corrie and Kohn, 2011; Montomoli et al., 2015; see review by Cottle et al., 2015). These atypical features and their implications for applicability of the channel-flow model and kinematic evolution of the Annapurna-Dhaulagiri Himalaya are discussed and assessed below.
4.1. Reduced Volumes of Leucogranite and Migmatite and Absence of Evidence for Partial Melting within the Sillimanite Stability Field
A key feature in the structure of the GHS observed along the length of the Himalaya is a high concentration of leucogranite dikes, sills, and injection complexes throughout UGHS- and STDS-equivalent structural sections. These melt-bearing sections typically have maximum combined thicknesses of tens to thousands of meters and can account for >50% of the total volume of UGHS-equivalent sections (e.g., Garhwal—Scaillet et al., 1990; Langtang—Inger and Harris, 1992; Bhutan—Grujic et al., 1996; Manaslu—Harrison et al., 1999; Zanskar—Walker et al., 1999; Everest—Searle et al., 2003; Sikkim—Searle and Szulc, 2005; Makalu—Streule et al., 2010). Typically, most partial melts and leucogranites are the product of fluid-absent muscovite dehydration and, at higher temperatures and lower pressures, biotite dehydration. These reactions occurred during isothermal to near-isothermal exhumation of the UGHS from high-temperature–high-pressure (>700 °C, 8–14 kbar) to high-temperature–low-pressure (>700 °C, 2–6 kbar) conditions (Fig. 11) (Harris and Massey, 1994; Guillot et al., 1999; Visonà and Lombardo, 2002; Searle et al., 2010; Streule et al., 2010). This exhumation corresponds to the Neohimalayan M2 metamorphic event that is reported from across the orogen within the sillimanite stability field between 22 and 15 Ma and is coincident with activity on the MCT and STD (Guillot et al., 1999; Searle et al., 2010). Importantly, if melt production continued during exhumation, the melt-weakened rheology (i.e., low viscosity) of the UGHS should have allowed crustal flow to continue to lower pressures during exhumation (Beaumont et al., 2004; Jamieson et al., 2004).
In the Annapurna-Dhaulagiri Himalaya, leucogranite bodies are no larger than tens of meters in thickness, and injection complexes are only observed in the Modi Khola transect and are less densely packed with dikes and sills than typically observed in other regions (Nazarchuk, 1993; Hodges et al., 1996; Larson and Godin, 2009). Leucosome content within migmatites ranges between 5% and 20% in the Kali Gandaki valley and 10%–40% in the Modi Khola valley, which is comparable to observations recorded elsewhere in the Himalaya that record ∼10%–40% partial melting within the GHS (e.g., Inger and Harris, 1992; Harris et al., 2004; Larson et al., 2010). However, migmatite volume within the UGHS is comparatively low and accounts for <50% of the structural thickness of Unit I, confined to ≤50-m-thick horizons and is almost negligible in Unit II and Unit III. Elsewhere in the Himalaya, UGHS-equivalent migmatitic sections can be as much as 10 km thick (Searle, 2013).
Pressure-temperature-time (PTt) constraints from the lower portion of the UGHS in the Annapurna-Dhaulagiri Himalaya indicate that burial and heating at kyanite-grade conditions (Fig. 11) initiated at ca. 48–43 Ma (Carosi et al., 2015; Iaccarino et al., 2015; Larson and Cottle, 2015). Subsequently, partial melting initiated as early as 41 Ma (Carosi et al., 2015) and continued almost entirely within the kyanite stability field to as late as 18.5 Ma (Nazarchuk, 1993; Hodges et al., 1996; Godin et al., 2001; Martin et al., 2010; Cottle et al., 2011; Kohn and Corrie, 2011; Carosi et al., 2015; Iaccarino et al., 2015; Larson and Cottle, 2015). As such, it is possible that melt volumes within the UGHS were not always low. If lateral melt migration occurred during deformation, then melt volumes in the exposed UGHS may not directly correspond to the amount of partial melting, especially if melt migration out-paced melt production. However, melt migration observed in the GHS across the Himalaya is dominantly vertical (up-section) (e.g., Harris and Massey, 1994; Hodges et al., 1996). There is no direct evidence of lateral melt migration in the Annapurna-Dhaulagiri Himalaya, although such phenomena are not ruled out.
It is acknowledged that the PTt evolution of different structural positions in the UGHS may vary (Jamieson et al., 2004; Cottle et al., 2015). However, available constraints suggest that the PTt evolution of the upper portion of the UGHS in the Annapurna-Dhaulagiri Himalaya (Fig. 10, Godin et al., 2001; Corrie and Kohn, 2011) may not be that different from the lower portion. In the upper portion of the UGHS, peak temperature metamorphism and associated in situ partial melting are recorded at 35–30 Ma (Godin et al., 2001; Corrie and Kohn, 2011). Leucogranite dikes with crystallization ages of 23–18.5 Ma are also identified from the STDS and upper UGHS (Hodges et al., 1996; Godin et al., 2001). These ages may indicate that partial melting initiated earlier in the lower portion of the UGHS (41 Ma) than in the upper portion (35 Ma) and the STDS. Alternatively, they may imply that multiple phases of melting and crystallization occurred between 41 and 18 Ma.
Field and petrological observations indicate that partial melting and leucogranite production occurred within the kyanite stability field (Hodges et al., 1996; Godin et al., 2001; Corrie and Kohn, 2011; Carosi et al., 2015; Iaccarino et al., 2015; Larson and Cottle, 2015). Evidence for partial melting within the sillimanite stability field is absent, and leucosomes observed within the upper portion of the UGHS contain kyanite, not sillimanite, and remain low in volume (Nazarchuk, 1993; Hodges et al., 1996; Godin et al., 2001; Corrie and Kohn, 2011; Parsons et al., 2016). Where present, sillimanite is identified as a late-stage overgrowth postdating partial melting under lower temperature retrograde conditions (Carosi et al., 2015; Iaccarino et al., 2015). Consequently, it is unlikely that the PTt path for the UGHS entered the sillimanite stability field at high temperatures and may have only resided there for a short period of time during retrogression (Carosi et al., 2015; Iaccarino et al., 2015). The available PTt constraints (Figs. 10 and 11) suggest that exhumation of the UGHS in the Annapurna-Dhaulagiri Himalaya followed a much shallower PTt path (this study; Pêcher, 1989; Iaccarino et al., 2015) than recorded elsewhere in the Himalaya (Fig. 11). Isothermal decompression was not a characteristic feature of metamorphism in this region and explains the reduced volumes of sillimanite-bearing metapelitic rocks and the absence of partial melting in the sillimanite stability field in the Annapurna-Dhaulagiri Himalaya. Isothermal decompression requires rapid exhumation that out-paces rates of cooling so that high temperatures are maintained through to lower pressures (Whitney et al., 2004). Consequently, a less steep exhumation path (i.e., non-isothermal) suggests that the UGHS in the Annapurna-Dhaulagiri Himalaya either exhumed slower or cooled faster relative to equivalent UGHS sections with isothermal exhumation paths. Given the lack of potential explanations for faster cooling, we propose that the shallow gradient exhumation path of the UGHS is indicative of slower exhumation.
The shallow gradient of the PTt exhumation path also suggests that fluid-absent muscovite dehydration may have only occurred during the initial phase of exhumation, while biotite dehydration could not occur at any stage during exhumation, which may have contributed further to the apparently low volumes of melt produced during exhumation. Lateral melt migration out-pacing melt production would also account for the low melt volumes.
4.2. Viscosity of the UGHS during Peak Metamorphism
Widespread partial melting is the proposed mechanism for the viscosity reduction needed to initiate and sustain mid-crustal channel flow during the Himalayan orogeny (Beaumont et al., 2001, 2004; Grujic, 2006; Rosenberg et al., 2007; Jamieson et al., 2011). While the volume of migmatite and leucogranite in the Annapurna-Dhaulagiri Himalaya is lower than elsewhere in the Himalaya, metamorphic and geochronometric constraints suggest that temperatures and melt volumes may still have been high enough for mid-crustal flow during peak conditions prior to the onset of exhumation.
For a flow stress of ∼10 MPa, most of the flow laws yield viscosities (Table 4) that are equal to or lower than the “melt-weakened” viscosity (1019 Pa s) required to numerically simulate mid-crustal channel flow (Beaumont et al., 2001). However, at lower flow stresses (σ = 2 MPa) that may be more representative of conditions toward the center of the channel (Grujic et al., 1996; Grujic, 2006; Law et al., 2013), all melt-free flow laws except for Blackhill quartzite yield a viscosity of 1020 Pa s (Table 4). Additionally, Fourier transform infrared spectroscopy of quartzites from UGHS-equivalent strata in the Sutlje valley of NW India records remarkably low volumes of water (Kronenberg et al., 2014) and suggests that the Heavitree (dry) quartzite flow law should be favored over the Blackhill (wet) quartzite flow law. The flow law for a partially molten synthetic granite, which we use as a proxy for migmatite rheology, yields very low viscosities even at low flow stresses. However, migmatites account for <35% of the UGHS structural thickness and are present in relatively thin bands (<50 m thick). As such, while this flow law clearly demonstrates the rheological weakening effect caused by partial melting, it is not representative of the bulk volume of the UGHS. Instead, the rheology of the UGHS is more likely to have been controlled by the major load-bearing lithologies that account for >65% of its structural thickness (i.e., quartzofeldspathic and calc-silicate gneisses). Viscosity estimates for these lithologies toward the center of the channel are on the order of 1020 Pa s, which is an order of magnitude greater than the “melt-weakened” viscosity (1019 Pa s) simulated by model HT1 of Beaumont et al. (2004).
An additional constraint on flow conditions for the UGHS during its mid-crustal evolution is structural (i.e., channel) thickness (Beaumont et al., 2004; Jamieson et al., 2004; Mukherjee, 2013a). A reduction in channel thickness increases the effective viscosity of the channel material and results in a reduction of the threshold viscosity below which return flow in a hybrid channel flow may occur (Turcotte and Schubert, 2002; Grujic, 2006). The numerically modeled thickness required for a southward-directed return flow to develop in the HT1 model is ∼10–20 km (Beaumont et al., 2004; Jamieson et al., 2004). In most regions, UGHS-equivalent strata match or exceed this thickness (e.g., 10–15 km—Bhutan, Manaslu, Garhwal [Grujic et al., 2002; Searle and Godin, 2003; Webb et al., 2011]; 20 km—Langtang, Sikkim [Inger and Harris, 1992; Searle and Szulc, 2005]; 30 km—Everest-Makalu [Searle et al., 2003; Jessup et al., 2006; Streule et al., 2010]). However, the UGHS of the Annapurna-Dhaulagiri Himalaya has a structural thickness of only ∼7 km and consequently, it is necessary to consider whether this reduced thickness would hinder or even prevent the development of a hybrid channel flow with a southward-directed return flow component (Godin et al., 2006a).
The average channel velocity (um) of the down-going channel wall (u0) and horizontal pressure gradient (dp/dx, where dp is the difference in channel pressure over the horizontal channel distance, dx) must also be known, and a linear viscous rheology is assumed (Turcotte and Schubert, 2002). Suitable parameters may be derived from the return-flow portion of four particle pathways (G1–G4, Table 5) taken from thermomechanical channel-flow model HT1 of Beaumont et al. (2004) and Jamieson et al. (2004; their figure 4). Over a 24 m.y. period of southward-directed channel flow with a convergence rate of 50 mm yr–1, the particle pathways from model HT1 have a time-averaged velocity of 0.017–0.018 m yr–1 and undergo a 1.75–8.75 kbar reduction in pressure over a horizontal distance of 416–430 km (Table 5). In order to maintain this velocity over this distance and pressure decrease, a modeled channel flow of 7 km thickness requires an estimated viscosity of 7.1 × 1018 to 3.7 × 1019 Pa s (Table 5). Larger viscosities would produce slower flow velocities, resulting in a more “sluggish” channel flow than that simulated by model HT1.
Assuming deformation was isochoric, transport-perpendicular shortening estimates of 25%–32% from the GHS in the Kali Gandaki valley (Larson and Godin, 2009) suggest that the UGHS may have had an initial thickness of 9.3–10.3 km. Larson and Godin (2009) state that their shortening estimates are likely to be underestimates. Larger shortening estimates of 35%–50% reported elsewhere in the Himalaya would indicate an initial thickness of up to 14 km (Larson et al., 2010; Law et al., 2013). If volume loss occurred during channel flow due to the lateral migration of magma, then the initial channel thickness may have been even greater. Using the HT1 channel-flow parameters defined above, the estimated viscosity of a larger channel thickness of 9–14 km must not exceed 1.25 × 1019 Pa s to 1.48 × 1020 Pa s in order to replicate the channel flow simulated by model HT1 (Table 5).
Assuming the chosen flow laws provide representative rheologies, much of the UGHS in the Annapurna-Dhaulagiri Himalaya, which consists of melt-free and probably dry quartzites and para-, ortho- and calc-silicate gneisses, was probably too viscous for channel flow to occur at modeled rates simulated by model HT1, if confined to a 7-km-wide channel with no change in channel thickness during flow. In order to maintain a constant channel thickness during channel flow, volume addition must occur to counteract transport-parallel shortening (e.g., Grasemann et al., 2006). Volume addition due to the migration and culmination of large leucogranite bodies and injection complexes is evident in UGHS-equivalent sections elsewhere along the Himalaya (e.g., Harrison et al., 1999; Grasemann et al., 2006; Searle et al., 2010; Searle, 2013). However, the absence of large leucogranite bodies and injection complexes in the Annapurna-Dhaulagiri Himalaya suggests that volume addition during channel flow of the UGHS in this region was negligible. Consequently, shortening estimates suggest that the UGHS may have been 9–14 km thick at the initial onset of channel flow. The possibility of volume loss due to lateral melt migration during channel flow implies that these are minimum initial thickness estimates.
At these larger initial thicknesses, many of the flow-law–derived viscosity estimates are of the same order of magnitude or lower than that required to produce flow velocities simulated by channel-flow model HT1 (Table 5). As such, these data suggest that prior to vertical shortening, the UGHS in the Annapurna-Dhaulagiri Himalaya was at least weak enough for mid-crustal channel flow to initiate. However, without the addition of new material during flow, as suggested by the absence of large culminations of leucogranite, the subsequent decrease in channel thickness due to vertical shortening resulted in a gradual increase in effective viscosity, which decelerated channel flow to slower rates than those predicted by both the thermomechanical models (Beaumont et al., 2001; Jamieson et al., 2004) and elsewhere in the Himalaya where melt content and structural thickness are typically higher. Being limited by a greater-than-typical viscosity, channel flow in the Annapurna-Dhaulagiri Himalaya may have been too viscous to facilitate isothermal exhumation. These calculations highlight the importance of volume addition during channel flow as a means of maintaining low viscosities.
4.3. Top-SW Shortening on the STDS
The STDS forms a top-NE shear zone at the top of the GHS that can be traced almost continuously along the entire orogen (Burg et al., 1984; Burchfiel et al., 1992). Recent estimates of top-NE dip-slip motion on the STDS range from 190 km in western Nepal (Borja et al., 2013) to 100–200 km in eastern Nepal (Searle et al., 2003, 2006; Law et al., 2011). Within the Kali Gandaki valley, field structural observations from the STDS indicate late-stage, top-SW (reverse) shortening that deforms earlier top-NE (normal) shear-related fabrics and leucogranites (Fig. 8). These structures are localized and probably reflect only minor amounts of shortening. Microstructures also indicate a top-SW shear sense in some samples from the STDS (Fig. 9). This late-stage shortening has been reported by other authors (Brown and Nazarchuk, 1993; Hodges et al., 1996; Godin et al., 1999a; Godin, 2003; Larson and Godin, 2009; Parsons et al., 2016) and correlated with D4 deformation in the THS, and with syn- to postpeak metamorphism in the UGHS along the KSZ in the Kali Gandaki valley (Vannay and Hodges, 1996; Godin et al., 1999a; Godin, 2003) and the MKSZ in the Modi Khola valley (Hodges et al., 1996).
Many authors favor the explanation that the STDS is a fixed hanging-wall stretching fault and/or shear zone (sensu Means, 1989) that formed in response to the rheological contrast between the hot, weak southward-flowing UGHS (footwall) and the cold, strong stationary THS (hanging wall) (Searle et al., 2003; Williams et al., 2006; Searle, 2010; Law et al., 2011; Kellett and Grujic, 2012). Such a situation reflects the development of a superstructure (i.e., THS)-infrastructure (i.e., UGHS) association (e.g., Williams et al., 2006; Jamieson and Beaumont, 2013) and implies that the STDS is inherently linked to the rheology of the Himalayan orogen. This is the only explanation for formation of the STDS that has been replicated by thermomechanical models of Himalayan orogenesis (Beaumont et al., 2001, 2004). Other tectonic models for the Himalayan orogeny have failed to produce a mechanically working explanation for formation of the STDS (e.g., Jamieson and Beaumont, 2013). This point is echoed by thermomechanical channel-flow models simulated without a significant rheology contrast between mid- and upper-crustal levels, which were unable to produce an STDS-equivalent structure (Model 3, Beaumont et al., 2004). These findings imply that the STDS formed in response to the rheological weakening of the Himalayan-Tibetan mid-crust (UGHS), which caused the upper (THS) and middle (UGHS) crust to mechanically decouple, allowing southward return flow of the UGHS and northward underthrusting of the Indian lower crust, relative to the overlying stationary THS (e.g., Kellett and Grujic, 2012). Subsequently, above a threshold viscosity, rheological strengthening of the UGHS led to the cessation of channel-flow and mechanical recoupling of crustal units. Thus it follows that late-stage top-SW shortening in the STDS may correspond to northward underthrusting of the Indian lower crust, following the cessation of channel-flow and mechanical recoupling of the THS and GHS.
Godin et al. (2001) postulate that an 40Ar/39Ar muscovite age of ca. 18 Ma from the lowermost THS records the growth and crystallization of muscovite during D4 deformation. Within the UGHS and STDS, D4 deformation structures (Figs. 6 and 8) indicate that the UGHS was still hot and weak enough to deform via pervasive ductile shearing during this deformation phase. As such, the timing of top-SW shortening associated with D4 deformation in the STDS, UGHS, and THS is bracketed between the latest age of D3 top-N shearing on the STDS (22–18.5 Ma; Hodges et al., 1996; Godin et al., 2001) and wholesale exhumation of the UGHS above the muscovite closure temperature for Ar loss (16–13 Ma; Godin et al., 2001; Vannay and Grasemann, 2001; Martin et al., 2015). This is contemporaneous with retrograde metamorphic conditions of 650–670 °C and 7–8 kbar within Unit I at 25–18 Ma (Iaccarino et al., 2015).
Here we suggest that top-SW shortening in the STDS and associated D4 deformation in the UGHS and THS occurred in response to an increase in viscosity of the UGHS due to cooling during exhumation and its subsequent transformation from an “active channel” (weak) to a “channel plug” (strong) (also referred to as a paleochannel; Beaumont et al., 2004; Godin et al., 2006a; Grujic, 2006). Continued northward underthrusting of the Indian lower crust after this viscosity increase resulted in minor amounts of shortening between the THS and GHS. This was synchronous with top-SW shearing on the KSZ and MKSZ sometime after 22–18.5 Ma and before 16–13 Ma, possibly at 18 Ma (Hodges et al., 1996; Godin et al., 2001; Vannay and Grasemann, 2001; Martin et al., 2015). This scenario is comparable to the regional-scale buckling of the THS and GHS reported from the Manaslu Himalaya in central Nepal (Fig. 1B) that occurred in response to the “locking-up” (i.e., cessation) of mid-crustal flow of the GHS and subsequent recoupling of the GHS and THS during continued top-S convergence (Godin et al., 2006b).
4.4. Metamorphic Discontinuities in the Context of Channel Flow
In the Modi Khola valley, metamorphic discontinuities at the base of the UGHS have been interpreted by some authors as faults (the Sinuwa and Bhanuwa faults) between discrete thrust slices (Martin et al., 2010; Corrie and Kohn, 2011). Claims that these discontinuities refute the occurrence of channel flow (Martin et al., 2010; Corrie and Kohn, 2011) can be reconciled if faulting occurred after the cessation of mid-crustal flow (e.g., Cottle et al., 2015). However, there are no field- or micro-structural observations to support the interpretation of these metamorphic discontinuities as faults. Nor is there evidence for significant annealing and static recrystallization that might have overprinted and obscured such structures.
Alternatively, metamorphic discontinuities can be explained by consideration of the relative velocity profile of a channel flow (Mancktelow, 1995; Grujic, 2006). Southward-directed channel flow during the Himalayan orogeny requires a viscosity reduction at mid-crustal levels, so that the overburden-induced, southward lateral pressure gradient (Poiseuille flow) can overcome the northward crustal flow produced by drag of the lower Indian crust (Couette flow) (Mancktelow, 1995; Turcotte and Schubert, 2002; Grujic, 2006). The resulting velocity profile resembles that of a “hybrid flow” (Grujic, 2006) composed of an upper portion of southward-exhuming, pressure-gradient–driven “return flow” and a lower portion of northward-burying flow (Fig. 12) (Mancktelow, 1995; Grujic, 2006). Within the northward-burying flow, the magnitude of northward velocity increases toward the base of the UGHS. Consequently, across a vertical transect through the channel, particles at the base of the northward-burying section have younger ages of prograde metamorphism and partial melting than particles at the top of the northward-burying section (Fig. 12). Pressure-temperature-time discontinuities could also develop if flow within the channel was discontinuous, due to localized variations in viscosity that would not necessarily be identifiable in the field. On this basis, combined with the absence of field structural evidence to support the presence of the inferred Sinuwa and Bhanuwa faults, it is argued that the aforementioned metamorphic discontinuities at the base of the UGHS in the Annapurna-Dhaulagiri Himalaya do not represent discrete faults and thrust slices. The proposed metamorphic discontinuities can be produced during pervasive ductile deformation (e.g., Mottram et al., 2014) and are entirely compatible with the channel-flow model.
5. IMPLICATIONS FOR THE CHANNEL-FLOW MODEL IN THE ANNAPURNA-DHAULAGIRI HIMALAYA
Geochronometric and thermobarometric constraints presented in this study and others indicate that the UGHS in this region underwent a prolonged phase of partial melting and kyanite-grade metamorphism between 41 and 18.5 Ma (Nazarchuk, 1993; Kaneko, 1995; Hodges et al., 1996; Godin et al., 2001; Martin et al., 2010; Corrie and Kohn, 2011; Kohn and Corrie, 2011; Carosi et al., 2015; Iaccarino et al., 2015; Larson and Cottle, 2015). During this time, coeval top-SW and top-NE motion occurred on the CT and AD/DD, respectively (Nazarchuk, 1993; Hodges et al., 1996; Godin et al., 2001). Both coaxial and non-coaxial, synmigmatitic, transpositional deformations (early fabrics transposed into S3 fabrics) are observed across the UGHS. The transpositional nature of D3 deformation between the STDS and THS indicates that the STDS formed as a stretching fault (Means, 1989) and replicates the detachment zone within the superstructure-infrastructure association (e.g., Williams et al., 2006; Jamieson and Beaumont, 2013). The apparent rheological contrast between upper and middle crustal units is likely to be responsible for formation of the STDS between the hot, partially molten, horizontally stretched UGHS (infrastructure) and cold, low-grade, horizontally shortened THS (superstructure) (Williams et al., 2006; Kellett and Godin, 2009; Searle, 2010; Kellett and Grujic, 2012).
Based on the features outlined above, the channel-flow model provides a favorable explanation for mid-crustal evolution of the UGHS in the Annapurna-Dhaulagiri Himalaya, despite its lower-than-typical structural thickness and leucogranite content. However, our findings indicate that the UGHS in the Annapurna-Dhaulagiri Himalaya was more viscous than elsewhere in the Himalaya. Consequently, channel flow in this region was likely to have been more limited (i.e., slower), perhaps with a shorter duration of activity than in other parts of the Himalaya with greater UGHS structural thicknesses and larger melt volumes.
The pervasive and transpositional nature of ductile deformation within the UGHS, which indicates horizontal stretching and vertical shortening, is not compatible with models of thrust stacking and duplex development, which require localization of deformation on discrete thrust planes between competent thrust slices (e.g., Robinson et al., 2006). Likewise, coeval reverse- and normal-sense motion below and above the UGHS, respectively, is yet to be explained with a mechanically working model of any process other than channel flow (e.g., Jamieson and Beaumont, 2013). However, in limiting the rate and/or duration of channel flow, we question whether tectonic processes in the UGHS may reflect a combination of both channel flow and discrete thrusting, particularly when the rheological contrast between different lithologies with different volumes of leucogranite and migmatite is considered. We postulate that in such a situation, pervasive deformation characteristic of channel flow could occur across the whole UGHS, but at different rates in different lithologies. The contacts between different lithologies would represent rheological discontinuities, and in such a situation, it is envisaged that discrete shear zones could form at these boundaries either during or after channel flow. Constraining the PTt evolution of each lithological unit in the UGHS would help further investigation of these concepts (e.g., Cottle et al., 2015). Similar models are presented by Cottle et al. (2015) and Jamieson and Beaumont (2013) who propose that channel-flow processes can evolve into thrust stacking processes due to changes in rheology during extrusion and exhumation.
The shallow-inclined PTt exhumation path (Fig. 11) indicates that the UGHS exhumed at a slower rate (non-isothermal) than typically recorded elsewhere in the Himalaya (isothermal) and than predicted by particle paths G1 and G2 from channel-flow model HT1 (Jamieson et al., 2004). Slow exhumation of the UGHS is explained by the viscosity-limited subdued nature of channel flow in the Annapurna-Dhaulagiri Himalaya. Our viscosity calculations demonstrate the control of channel thickness and partial melting on viscosity. Importantly, without the production of large amounts of leucogranite (i.e., volume addition), as recorded elsewhere in the Himalaya, vertical shortening during deformation will reduce channel thickness over time and gradually decrease the threshold (i.e., maximum) viscosity at which channel flow can be sustained. Without the greater structural thickness and melt volumes typically recorded elsewhere in the Himalaya, the viscosity of the UGHS in the Annapurna-Dhaulagiri Himalaya was probably too high to facilitate rapid isothermal exhumation. Instead, channel flow gradually “seized up” and ground to a halt at significant depth (7–8 kbar). This allowed the UGHS to cool, strengthen, and transform from a weak “active channel” to strong “channel plug” at greater depths than in other UGHS-equivalent sections elsewhere in the Himalaya, resulting in slower exhumation.
Correlation between: (1) the timing of retrograde conditions of 650–670 °C and 7–8 kbar at 25–18 Ma and a quartz deformation temperature of 670 ± 50 °C in Unit I; (2) the youngest timing of deformation on the CT and AD/DD at 22–18.5 Ma; and (3) the minimum timing of D4 top-SW shortening in the STDS and on the KSZ and MKSZ in the UGHS at 18 Ma suggests that this transformation from a weak “active channel” to strong “channel plug” occurred at 22–18 Ma (Nazarchuk, 1993; Hodges et al., 1996; Godin et al., 2001; Larson and Godin, 2009; Iaccarino et al., 2015). We postulate that these conditions correspond to the threshold viscosity for channel flow of the UGHS in the Annapurna-Dhaulagiri Himalaya and record a change in rheology from a pervasively deforming and/or flowing channel (“active channel”) to a coherent block (“channel plug”). Elsewhere along the Himalaya, similar retrograde temperatures recorded in UGHS-equivalent sections were not reached until pressures of 2–6 kbar (Fig. 11). This occurred sometime after 16 Ma in central and eastern Nepal (Searle et al., 2003; Streule et al., 2010; Imayama et al., 2012) and at 14–11 Ma in Bhutan (Daniel et al., 2003; Harris et al., 2004), following exhumation and partial melting well within the sillimanite stability field. Field observations and thermobarometric constraints suggest that following cessation of channel flow in the UGHS, pervasive top-S shearing in the LGHS facilitated the progressive exhumation of the UGHS as a rigid block toward the topographic surface from ∼650 °C equivalent depth. This hypothesis is comparable to models proposed by Mottram et al. (2014, 2015) for LGHS-equivalent strata in Bhutan.
Field observations from the Annapurna-Dhaulagiri Himalaya highlight several geological features across the GHS that are distinct from other Himalayan regions. These include low volumes of leucogranite and migmatite, an absence of evidence for partial melting in the sillimanite stability field, a reduced structural thickness, and late-stage top-SW shortening in the STDS. These features are not readily compatible with proposed models for the evolution of the Himalayan orogen, and their implications for the rheological behavior of the GHS in this region must therefore be considered.
Field structural observations, combined with new and previously published thermobarometric and geochronometric constraints indicate that the mid-crustal evolution of the UGHS in the Annapurna-Dhaulagiri Himalaya is more favorably explained by the channel-flow model than models based on thrust-stacking mechanisms. However, consideration of the rheological properties of the UGHS with a specific focus on the effects of melt volume and channel thickness suggests that channel flow was more limited and subdued (i.e., slower) in the Annapurna-Dhaulagiri Himalaya than in other regions where melt volume and structural thickness were larger. Flow-law–derived viscosity estimates of ∼1020 Pa s for the bulk volume of the UGHS also imply that channel flow was slower and more limited than that simulated by thermomechanical channel-flow model HT1 of Beaumont et al. (2004). Furthermore, viscosity calculations demonstrate that without addition of material during channel flow (i.e., leucogranite generation), vertical shortening led to a gradual decrease in the threshold viscosity for flow within the UGHS.
The PTt exhumation path of the UGHS in the Annapurna-Dhaulagiri Himalaya has a shallower gradient (i.e., slower) than in other parts of the Himalaya where UGHS-equivalent sequences exhumed along isothermal to near-isothermal (i.e., rapid) trajectories. Additionally, the UGHS PTt exhumation path from the Annapurna-Dhaulagiri Himalaya is much shallower than modeled isothermal PTt paths determined from channel-flow model HT1. Our findings suggest that in the Annapurna-Dhaulagiri Himalaya, viscosity-limited, subdued channel flow of the UGHS was not weak enough to facilitate rapid, isothermal exhumation due to its lower-than-average structural thickness and melt volumes. Slower exhumation along a more shallow-inclined PTt path prevented extensive fluid-absent decompression melting within the sillimanite stability field, which compounded the rheological effects of low melt volumes during exhumation even further. Consequently, vertical shortening of ∼25%–50% without the addition of large melt volumes led to a gradual deceleration of channel flow as the threshold viscosity for flow decreased. Eventually, the effective viscosity of the UGHS exceeded the threshold viscosity for flow, at which point mid-crustal channel flow could no longer be sustained.
The initiation of D4 deformation associated with top-SW shortening in the STDS and top-SW motion along the KSZ and MKSZ at ca. 18 Ma is believed to reflect the transformation of the UGHS from a weak “active channel” to a strong “channel plug.” Pressure-temperature-time constraints suggest that in the Annapurna-Dhaulagiri Himalaya this transformation, and thus the threshold viscosity for channel flow, occurred at 650–670 °C and 7–8 kbar. Consequently, the UGHS in this region exhumed slowly, cooling and strengthening 2–7 m.y. earlier and at 1–6 kbar greater pressure (∼3–18 km deeper) than elsewhere in the central and eastern Himalaya.
These findings are distinct from other regions in the Himalaya, and, as such, we consider the mid-crustal evolution of the GHS in the Annapurna-Dhaulagiri Himalaya to be an atypical example of channel flow during the Himalayan orogeny.
We thank Ron Frost, Dawn Kellett, and Claire Warren for their extensive reviews that greatly improved this work, and we also thank Geosphere editors Mike Cheadle and Ray Russo, who oversaw the review and revision process. Mark Caddick and Matt Kohn are thanked for helpful advice concerning the thermobarometry presented in this manuscript. Suka Ghale, Basan Sherpa, and their Nepalese colleagues are thanked for guidance and assistance during fieldwork. The research reported in this paper was supported by Natural Environment Research Council (training grant NE/J50001X/1) to AJP and National Science Foundation grant EAR0711207 to RDL. We thank Research Councils UK for funding the open access publication of this manuscript.