Orogen‐parallel deformation of the Himalayan midcrust: Insights from structural and magnetic fabric analyses of the Greater Himalayan Sequence, Annapurna‐Dhaulagiri Himalaya, central Nepal

The metamorphic core of the Himalaya (Greater Himalayan Sequence, GHS), in the Annapurna‐Dhaulagiri region, central Nepal, recorded orogen‐parallel stretching during midcrustal evolution. Anisotropy of magnetic susceptibility and field‐based structural analyses suggest that midcrustal deformation of the amphibolite facies core of the GHS occurred under an oblate/suboblate strain regime with associated formation of low‐angle northward dipping foliation. Magnetic and mineral stretching lineations lying within this foliation from the top of the GHS record right‐lateral orogen‐parallel stretching. We propose that oblate strain within a midcrustal flow accommodated oblique convergence between India and the arcuate orogenic front without the need for strain partitioning in the upper crust. Oblate flattening may have also promoted orogen‐parallel melt migration and development of melt‐depleted regions between km3 scale leucogranite culminations at ~50–100 km intervals along orogen strike. Following the cessation of flow, continued oblique convergence led to upper crustal strain partitioning between orogen‐perpendicular convergence on thrust faults and orogen‐parallel extension on normal and strike‐slip faults. In the Annapurna‐Dhaulagiri Himalaya, orogen‐parallel stretching lineations are interpreted as a record of transition from midcrustal orogen‐perpendicular extrusion to upper crustal orogen‐parallel stretching. Our findings suggest that midcrustal flow and upper crustal extension could not be maintained simultaneously and support other studies from across the Himalaya, which propose an orogen‐wide transition from midcrustal orogen‐perpendicular extrusion to upper crustal orogen‐parallel extension during the mid‐Miocene. The 3‐D nature of oblate strain and orogen‐parallel stretching cannot be replicated by 2‐D numerical simulations of the Himalayan orogen.

The MCTZ was previously defined as the Lower Greater Himalayan Sequence (LGHS) [Parsons et al., 2016c]. Recent work from Parsons et al. [2016a] demonstrated that the structural evolution of this section of the GHS reflects that of a crustal-scale top-SW shear zone, synonymous to the Main Central Thrust Zone (MCTZ) as described by Bouchez and Pêcher [1981], Grasemann et al. [1999], Mottram et al. [2015], , and others. Based on these studies, the LGHS of Parsons et al. [2016c] is more appropriately defined as the MCTZ.

Anisotropy of Magnetic Susceptibility (AMS)
Magnetic susceptibility (K) is defined as the ratio between magnitudes of an externally applied magnetic field (H) and an object's corresponding induced magnetization (M).
Anisotropy of magnetic susceptibility (AMS) describes directional variation of an object's magnetic susceptibility (K) as a second rank tensor. In geological materials, AMS is controlled by mineral content, mineral shape fabric, and crystallographic fabric [Borradaile and Jackson, 2004]. AMS can be used to investigate deformation kinematics if contributions of controlling factors are determined [Borradaile and Jackson, 2010]. The AMS ellipsoid is defined by three mutually orthogonal principal susceptibility axes, K 1 (maximum), K 2 (intermediate), and K 3 (minimum) [Tarling and Hrouda, 1993]. K 1 is commonly referred to as the magnetic lineation. K 3 represents the pole to magnetic foliation (K 1 -K 2 plane). Bulk susceptibility (K m ) is the mean of K 1 , K 2, and K 3 [Janák, 1965].
The corrected degree of anisotropy (P′-magnetic anisotropy from hereafter) describes strength of AMS in terms of deviation from a sphere (P′ = 1) to an ellipsoid (P′ > 1) [Jelínek, 1981].

Controlling Factors of AMS
In order to make meaningful interpretations from AMS data, factors controlling the AMS of individual samples must be determined [e.g., Wallis et al., 2014a]. Potential controlling factors are briefly outlined below and reviewed in detail by Tarling and Hrouda [1993], Tauxe [ 2002], and Borradaile andJackson [2004, 2010].
Magnetic materials may be classed as diamagnetic, paramagnetic, or ferromagnetic, listed in order of strength of induced magnetization [Tarling and Hrouda, 1993;Tauxe, 2002]. Induced magnetization of diamagnetic materials is diametrically opposed to the external field (Àve K). Induced magnetization of paramagnetic and ferromagnetic (sensu lato) materials occurs in the same direction as the external field (+ve K). Diamagnetic (e.g., quartz and calcite) and paramagnetic (e.g., phyllosilicate) materials lose induced magnetization instantaneously upon removal of the external field. Most ferromagnetic minerals possess strong (e.g., magnetite) or weak (e.g., hematite) remanent magnetism [Tarling and Hrouda, 1993].
Grain-scale AMS is affected by grain shape (magnetostatic anisotropy) and crystal structure (magnetocrystalline anisotropy), whereby AMS axes align parallel or orthogonal to grain shape or crystal axes. Magnetization of ferromagnetic minerals is affected by grain size, which determines the number of magnetic domains (regions of uniform magnetism with a single dipole) within a grain [Tauxe, 2002]. For magnetite, grain size from smallest to largest is classified as single domain (SD, typically <0.1 μm), pseudo single domain (PSD), and multiple domain (MD). Very small SD grains (≤0.05 μm), which lose their remanent magnetism almost instantaneously, are superparamagnetic (SP) [Tarling and Hrouda, 1993;Tauxe, 2002].
Whole-rock AMS fabrics reflect summation of grain-scale anisotropies of all constituent grains, plus effects of grain shape preferred orientation (SPO) and crystallographic preferred orientation (CPO). AMS fabrics are controlled by the magnetically predominant mineral phase(s), referred to as magnetic carrier(s) [Borradaile and Jackson, 2004]. Strong SPO of ferrimagnetic minerals and CPO of all other minerals may produce a magnetic lineation and/or foliation that mimics mineral fabrics [Borradaile, 1991;Borradaile and Jackson, 2010]. AMS fabrics may be influenced by magnetostatic interactions between ferromagnetic grains. Significance of such interactions depends on the spatial distribution and concentration of ferromagnetic grains [Hargraves et al., 1991;Stephenson, 1994;Muxworthy et al., 2004]. AMS fabrics controlled by SPO and/or CPO fabrics may correlate with 3-D strain, such that principal susceptibility axes (K 1 ≥ K 2 ≥ K 3 ) and finite strain axes (X ≥ Y ≥ Z) are parallel [Borradaile, 1991;Borradaile and Jackson, 2010;Kruckenberg et al., 2010Kruckenberg et al., , 2011Ferré et al., 2014Ferré et al., , 2016. Correlations may also exist between P′ and finite strain magnitude [e.g., Benn, 1994;Tripathy, 2009] and between T and strain geometry [e.g., Sidman et al., 2005]. Magnetic carriers must be determined, and competing factors that control/influence AMS must be investigated before such correlations can be made [Borradaile and Jackson, 2010;Ferré et al., 2014;Wallis et al., 2014a].

Analytical Methods
Magnetic fabric analyses were conducted at the Department of Geology, Southern Illinois University, following procedures recommended by Ferré et al. [2003Ferré et al. [ , 2004 and Kruckenberg et al. [2010]. Forty fieldorientated samples were analyzed from the Kali Gandaki Valley and surrounding foothills ( Figure 2 and Table 1). Cubes (17-18 mm) cut from each sample (614 cubes) were analyzed using an AGICO KLY-4S Kappabridge susceptometer at 300 A/m [Pokorný et al., 2004]. A minority of samples with low susceptibility were analyzed at 450 A/m. AMS data were acquired using SUFAR 1.2 and processed with Anisoft 4.2 to a Samples listed in order of vertical structural height above the MCT. Mean magnetic properties of each sample and the orientation of the resulting magnetic fabrics and magnetic lineations, plus the locally measured structural foliation and mineral lineation orientations. Magnetic carrier of each sample is also given. K m , bulk susceptibility; P′, corrected degree of anisotropy; T, shape parameter-see the supporting information Data Set S1 for full AMS data set. See supporting information Table S1 for detailed summary of all magnetic carriers.
Magnetic hysteresis analyses were conducted on one to three cubes from each sample to determine magnetic carrier types [e.g., Dunlop, 2002a]. First-Order Reversal Curve (FORC) analyses were conducted on samples with ferromagnetic carriers to measure magnetostatic grain interactions [e.g., Muxworthy et al., 2004]. These analyses were conducted with a Princeton 3900-04 vibrating sample magnetometer up to a field of 7.94 × 10 5 A/m (1 T). FORC data were processed using FORCinel 1.21 [Harrison and Feinberg, 2008]. Scanning electron microscopy (SEM), electron dispersive spectroscopy (EDS), and electron backscatter diffraction (EBSD) analyses aided identification of magnetic carriers (Text S1 and Figures S3 and S4 in the supporting information).

AMS Parameters (K m , P′, and T)
The full AMS data set is presented in Data Set S1. Sample mean bulk susceptibility (K m ) varies between 3.70 × 10 À6 and 3639.41 × 10 À6 (SI) ( Table 1), reflecting variety of diamagnetic, paramagnetic, and  (Figure 3). Mean P′ varies between 1.01 and 3.43 and does not correlate with tectonostratigraphy ( Figure 3a and Table 1).
The maximum magnetic anisotropy (P′) that can be obtained from well-oriented paramagnetic silicates reflects the intrinsic magnetic anisotropy of such minerals (P′ ≤ 1.36) and could only be attained by a monomineralic assemblage of perfectly oriented grains [Martín-Hernández and Hirt, 2003]. As this is not the case, maximum magnetic anisotropy attributed to paramagnetic silicates in schists and mylonites culminates typically at P′ ≈ 1.15 [Martín-Hernández and Ferré, 2007]. P′ > 1.15 is likely to arise from ferromagnetic (sensu lato) contributions to AMS. Figure 3a provides valuable information on the origin of AMS in our samples. Samples with K m < 20 × 10 À6 (SI) and P′ > 1.15 display AMS typically dominated by diamagnetic and, to a minor extent, other minerals. These samples display negative correlation between K m and P′; however, these values are too close to the sensitivity of the Kappabridge instrument to be interpreted with confidence. Samples with 20 × 10 À6 (SI) < K m < 100 × 10 À6 (SI) and P′ < 1.15 are characteristic of paramagnetic fabrics dominated by magnetocrystalline phyllosilicates. There is no correlation between P′ and K m between these ranges, most likely because the maximum fabric strength is bound by the intrinsic magnetic anisotropy. Samples with K m > 100 × 10 À6 (SI) and P′ > 1.15 display AMS dominated by magnetostatic anisotropy of ferromagnetic (sensu lato) carriers such as magnetite or pyrrhotite. Correlation between P′ and K m at this range is typical of such ferromagnetic phases [e.g., Ferré et al., 1997].
Mean T ranges between À0.85 (prolate) and 0.92 (oblate) (Figure 3b). T correlates roughly with tectonostratigraphy, with mean T of À0.06 for the THS, 0.41 for the STDS, 0.73 for the UGHS, 0.44 for the MCTZ, and 0.32 for the LHS. T is dominantly oblate for most samples in the UGHS, while samples from the STDS produce suboblate fabrics (Figure 3b). In the MCTZ, specimens dominated by paramagnetic minerals have oblate ellipsoids, whereas specimens dominated by ferromagnetic carriers display more triaxial ellipsoids.
Mean principal axis orientations for K 1 , K 2 , and K 3 are presented in Table 1. K 1 plunge and azimuth typically ranges between 10 and 50°northwest to east. K 3 typical plunges 40-80°toward west to south.

Magnetic Carriers
To constrain the origin of AMS (i.e., magnetic carrier(s)), we supplement AMS analyses with magnetic hysteresis and FORC analyses (Figures S1 and S2) to assess the abundance of magnetic mineral species and determine which phases are present and likely to contribute to AMS. Only complete, well-defined hysteresis loops are used in interpretation of magnetic carriers ( Figure S1 and Table S1). Contributions of specific phases to AMS are further evaluated through EBSD-CPO analysis of paramagnetic phases (i.e., phyllosilicates) ( Figure S3) and electron microscopy and EDS of minor ferromagnetic phases (i.e., pyrrhotite and magnetite) ( Figure S4). Hysteresis and FORC diagrams (Data Set S2) are included in Figures S1 and S2.
A range of magnetic responses indicative of diamagnetic, paramagnetic, and ferromagnetic (sensu lato) carriers were recorded. In general, results produced by multiple specimen cubes of the same sample were consistent, suggesting homogenous ferromagnetic phase and grain size distributions. A few samples produced nonself-consistent results that may reflect uneven ferromagnetic phase and grain size distributions across the sample. Identified magnetic carriers are listed in Tables 1 and S1.
Microscopy of mineral assemblages indicates that phyllosilicates form magnetic carriers in 21 paramagnetic samples from the LHS, GHS, and THS (Table 1). Correlation between paramagnetic AMS fabric orientations and EBSD-derived phyllosilicate CPOs suggests that phyllosilicate CPO controls the AMS of most of these samples ( Figure S3). Diopside and/or phyllosilicate CPO/SPO form magnetic carriers in P12/052M and P13/041M (calc-silicate gneisses, UGHS- Table 1). Paramagnetic slope correction of hysteresis loops typically reveals minor volumes (<0.1%) of ferromagnetic phases within paramagnetic and diamagnetic samples ( Figure S1 and Table S1).

Correlation Between AMS Fabrics and Deformation in the Annapurna-Dhaulagiri Himalaya
The magnetic carrier(s) of every sample cannot be identified with absolute certainty; however, consistent trends in AMS ellipsoid orientation and shape, and strong correlation between macrostructural and magnetic  (Table 1 and Figures 4 and S5) of all samples form well-defined magnetic foliations and, in most cases, magnetic lineations (except P12/087M, LHS). A lack of consistent correlation between K m , P′, and T ( Figure 3) or between magnetic material types, P′ and T ( Figure S6), suggests that AMS varies independently of mineral assemblages. Close correlation is observed between AMS fabric orientation and locally measured deformation fabrics (Table 1 and Figures 2, 4, and 5).
Plotting AMS foliation and lineation orientations on a geological map ( Figure 2) and cross sections ( Figure S7) strengthens correlation between AMS fabrics and regional structure. AMS fabrics correlate with macroscopic S3 foliation and L3 lineation populations from the GHS [Godin, 2003;Parsons et al., 2016c]. These are shear-related fabrics which overprint and transpose earlier deformation fabrics ( Figure S8). Quartz and feldspar CPO fabrics and vorticity analyses of high-temperature S3-L3 fabrics (550°C to >650°C) in the UGHS and base of the STDS record general shear with components of both oblate and plane strain coaxial flattening [Larson and Godin, 2009;Parsons et al., 2016aParsons et al., , 2016b. AMS fabrics derived from migmatitic samples in the STDS and UGHS may have recorded deformation during partial melting. These correlations between AMS fabrics, S3-L3 macroscopic fabrics, and CPO fabrics [Parsons et al., 2016a[Parsons et al., , 2016b[Parsons et al., , 2016c suggest that AMS fabrics recorded midcrustal deformation kinematics.

10.1002/2016TC004244
The AMS fabrics also appear to record variations in 3-D strain geometry. Some authors warn against the use of T as a proxy for strain geometry as T also depends on magnetic carrier properties, which is especially the case when measuring AMS of a single specimen cube [e.g., Borradaile and Jackson, 2010;Ferré et al., 2014]. However, sample mean T was obtained using Anisoft 4.2 [Chadima and Jelínek, 2008] with calculations derived from directional statistics of Jelínek [1978Jelínek [ , 1981. These statistics consider the symmetry of the whole-rock AMS fabric (prolate versus oblate) as a composite of all specimen cube fabrics and are less dependent on mineral properties.
In the MCTZ and LHS, magnetic carrier and T show a possible correlation (Figures 3b and 6). As such, kinematic interpretations are not made from T in these units. Samples from the THS, STDS, and UGHS lack correlation between T and K m (Figure 3b) or between magnetic material type and T ( Figure S6b), which suggests that T varies independently of magnetic mineral assemblage. Importantly, both ferromagnetic and paramagnetic samples in these units produce the same trends in T over a range of T = 0.0-0.9 (Table 1). This agreement is significant as the magnetic carriers of different samples vary between phyllosilicate, magnetite, and pyrrhotite, all of which have distinct intrinsic magnetic anisotropies. For example, undeformed magnetite single crystals have an intrinsic T of À0.3, while phyllosilicate single crystals have an intrinsic T ranging from 0.7 to 1.0 [Tarling and Hrouda, 1993;Martín-Hernández and Hirt, 2003]. The absence of macrostructural and microstructural evidence for two or more crosscutting deformation fabrics indicates that AMS fabrics are not the product of superposition of multiple structural fabrics. Agreement between values of T from individual specimen cubes of a single sample and the corresponding sample mean T indicates that planar sample fabrics are not produced by superposition of variably oriented linear fabrics measured in cube specimens. As such, correlation between T and tectonostratigraphy ( Figure 6), independent of magnetic carrier types, suggests that observed variations in T correspond to variations in strain geometry. This hypothesis is strengthened by correlation between AMS fabric orientations and macroscopic deformation kinematics. We therefore propose that the measured AMS fabrics provide a proxy for 3-D strain geometries and their kinematic interpretation.

10.1002/2016TC004244
Unit II and Unit III (Figure 7a). Mean lineation azimuths equate to (a) orogen-perpendicular stretching in the MCTZ and Units I and II and (b) orogen-parallel stretching in Unit III and the STDS.   (4)). C = fabric strength [Woodcock, 1977]. See text for explanation.

10.1002/2016TC004244
magnetic carriers [Tarling and Hrouda, 1993;Martín-Hernández and Hirt, 2003]. In the UGHS, mean T = 0.73, suggesting an oblate strain geometry (Figures 3b and 6). In the STDS, mean T = 0.41 (Figure 6), suggesting a suboblate strain geometry (transitional oblate-plane strain; smaller magnitude of stretching parallel to Y direction, relative to X). T varies little within the UGHS, STDS, and THS but changes sharply across the CT, AD, and STD, suggesting that these units deformed under different 3-D strain geometries.
In some cases, magnetic foliation of oblate AMS sample fabrics is defined by a girdle distribution of cube specimen K 1 and K 2 axes (Figures 4c and 4d, e.g., P13/024M and P13/033M). Within our data set, this type of oblate AMS fabric is found in migmatitic samples containing leucosomes. We suggest that these samples, which yield the most oblate AMS fabrics of the whole sample suite (T = 0.9) record synmigmatitic heterogeneous deformation during partial melting and provide the strongest evidence for oblate flattening within the GHS during midcrustal deformation.

Eigenvalue Analysis
The proportionality of cluster to girdle distribution tendencies of a population of linear orientation data can be defined by the eigenvalue statistic, K [Woodcock, 1977]. To avoid confusion with the magnetic susceptibility (also K), we refer to this eigenvalue statistic as S K , where S 1 , S 2 , and S 3 define the maximum, intermediate, and minimum eigenvalues with orthogonal eigenvectors. S K = 0 defines an axially symmetric girdle distribution. S K ➔ ∞ defines an axially symmetric cluster distribution. S K = 1 defines a distribution with equal cluster and girdle tendencies [Woodcock, 1977]. Calculating S K for populations of macroscopic mineral stretching lineation and pole-to-foliation populations provides constraint on regional strain geometry [e.g., Flinn, 1978;Ramsay and Huber, 1983].
Magnetic and mineral stretching lineations in the MCTZ (Figure 7) have a strong cluster distribution (S K = 4.3, Figure 7d) orientated within a common regional foliation plane orientation. This is in agreement with microstructural analyses from the MCTZ that record noncoaxial plane strain [Larson and Godin, 2009;Parsons et al., 2016a]. In contrast, strong girdle distributions of magnetic and mineral stretching lineation populations (Figures 7b and 7c) in the UGHS and STDS (S K = 0.4-0.5, Figure 7d) are suggestive of oblate flattening (i.e., nonplane strain). Correlation between microstructural and magnetic fabrics in these units (see section 5) suggest that deformation in the UGHS and STDS reflects a general shear with components of both oblate/suboblate and plane strain coaxial flattening [Larson and Godin, 2009;Parsons et al., 2016a].

Orogen-Parallel Stretching in the GHS
The consistent orientation of pole-to-foliation populations within different tectonostratigraphic units of the GHS (Figures 5, 7b, and 7c) indicates that orogen-parallel lineation azimuths are not the result of late-stage folding or tilting. Correlation with previously published deformation temperature constraints [Parsons et al., 2016a, and references therein] suggests that lineation populations correspond to midcrustal deformation during which pervasive shearing in the STDS and Unit III involved a component of orogen-parallel stretching.
In the STDS, deformation had an oblique-top-down-to-ENE normal shear sense associated with a component of right-lateral orogen-parallel stretching within the plane of the low-angle S3 foliation.

Oblate Strain in the GHS
High-temperature macroscopic and microstructural deformation fabrics from the UGHS and STDS are interpreted as a record of midcrustal channel flow [Larson and Godin, 2009;Searle, 2010;Parsons et al., 2016aParsons et al., , 2016b. As such, our record of inferred strain geometries, some of which correspond to synmigmatitic deformation, suggests that crustal flow in the Annapurna-Dhaulagiri Himalaya occurred under an oblate/suboblate strain regime, involving an orogen-parallel component of stretching. Parsons et al.
Three-dimensional crustal flow is proposed in several models for the crustal evolution of the Tibetan Plateau [Dewey et al., 1988;Westaway, 1995;Clark and Royden, 2000;Chen and Gerya, 2016]. Three-dimensional flattening during channel flow in the Himalayan orogen has not been simulated by thermomechanical models as their two-dimensional construct necessities a plane strain regime [Beaumont et al., 2001[Beaumont et al., , 2004Jamieson and Beaumont, 2013]. However, the authors of these and similar models have speculated that channel flow could involve a component of oblate strain resulting in simultaneous orogen-perpendicular and orogenparallel lateral crustal flow [Beaumont et al., 2006, p.135;Culshaw et al., 2006, p. 734].

Orogen-Parallel Midcrustal Deformation During the Himalayan Orogeny
Within the THS of the Kali Gandaki Valley, E-W extension across the Thakkhola graben (Figures 2 and S9) is well documented [Hurtado et al., 2001;Hurtado, 2002;Garzione et al., 2003;Godin, 2003]. The earliest extension recorded along the basin-bounding Dangardzong fault in the upper reaches of the Kali Gandaki Valley occurred prior to 17-18 Ma, as indicated by indistinguishable 40 Ar/ 39 Ar muscovite ages from amphibolite facies footwall rocks and adjacent weakly metamorphosed hanging wall rocks [Hurtado, 2002]. Importantly, this earliest record of E-W orogen-parallel extension before 17-18 Ma overlaps with the latest record of top-SW orogen-perpendicular synmigmatitic shearing in the UGHS of the Annapurna-Dhaulagiri Himalaya at 18-22 Ma [Nazarchuk, 1993;Hodges et al., 1996;Hurtado, 2002;Iaccarino et al., 2015;. These time constraints are similar to the earliest record of E-W extensional faulting and fracture development in the THS and STDS of the Manaslu Himalaya at 14-17.5 Ma,~70 km east of the Kali Gandaki [Coleman and Hodges, 1998]. Approximately 170 km northwest of the Kali Gandaki, a transition from orogen-perpendicular to orogen-parallel deformation is recorded in the upper structural levels of the GHS and the overlying THS in the upper Karnali Valley between 13 and 15 Ma [Nagy et al., 2015]. Similarly, 230 km northwest of the Kali Gandaki Valley in the Pulan region of southern Tibet (Figure 1), midcrustal orogen-parallel stretching initiated in the upper structural levels of UGHS-equivalent strata between 15 and 22 Ma [Xu et al., 2013]. In both regions (Figure 1), orogen-parallel deformation initiated at or close to peak metamorphic temperature [Xu et al., 2013;Nagy et al., 2015].

Accommodation of Orogen-Parallel Stretching During Continued Convergence
Radial spreading of the Tibetan plateau during gravitational collapse provides one explanation for E-W extension of southern Tibetan Plateau and Himalayan upper crust [Styron et al., 2011]. An additional or alternative explanation is provided by models of strain partitioning in response to obliquity between northward motion of the Indian Plate and the arcuate orogenic strike of the Himalaya [McCaffrey and Nabelek, 1998;Styron et al., 2011]. Earthquake focal mechanisms along the length of the Himalayan frontal thrust system record thrust motions perpendicular to local orogenic strike, while GPS vectors indicate that the motion of the Indian Plate is only normal to the orogenic strike in the Everest region (Figure 1) [Styron et al., 2011]. Obliquity between the Indian Plate motion and slip vectors on orogen-perpendicular thrust faults increases along strike from the Everest region toward the east and west syntaxes [Styron et al., 2011]. In order to maintain strain compatibility during oblique convergence, displacement vectors are partitioned into orogen-perpendicular convergence on the frontal thrust system and orogen-parallel stretching and extension on hinterland strike-slip and normal faults [McCaffrey and Nabelek, 1998;Styron et al., 2011;

10.1002/2016TC004244
In the Annapurna-Dhaulagiri region, obliquity between the Indian Plate motion and orogenic strike is~15°[ cf. Styron et al., 2011]. Seventy kilometers west of the Annapurna-Dhaulagiri region, strain partitioning is suggested to be responsible for development of the Western Nepal Fault System (WNFS), which links the hinterland Karakoram Fault Zone to the frontal thrust system south of Annapurna (Figure 1) [Murphy et al., 2014;Wallis et al., 2014b;Silver et al., 2015]. As yet, a kinematic link between the WNFS and the Thakkhola graben has not been identified at ground level; however, given the proximity of these structures, it is possible that the two systems formed in response to the same stress regime.
Current channel flow models are implicitly two-dimensional [e.g., Beaumont et al., 2001Beaumont et al., , 2004Grujic, 2006]. In these models, the direction of the lateral pressure gradient produced by the overburden of the Tibetan Plateau is diametrically opposed to the motion of the underthrusting Indian Plate (i.e., pressure gradient and plate motion are parallel but with opposite directions). This, by default, simulates channel flow in plane strain. In reality, the Indian Plate motion remains constant along the length of the orogen, while the direction of the opposing overburden-induced lateral pressure gradient rotates to maintain an orogen-perpendicular orientation, reflecting the radial gravitational spreading of the Tibetan Plateau [e.g., Brun and Merle, 1985;Merle, 1989]. In regions where the Indian Plate motion and the overburden-induced pressure gradient are obliquely opposed (i.e., nonparallel), the combined shear of these two forces will not produce a plane strain but will instead produce an oblate/suboblate strain. We suggest that in these regions, prior to upper crustal orogen-parallel extension, pervasive oblate flattening of the active channel flow accommodated oblique convergence within the channel, without the need for orogen-scale partitioning of deformation into orogen-perpendicular and orogen-parallel displacement vectors.
Evidence for midcrustal orogen-parallel stretching of the GHS continues to grow; however, it remains unclear as to how this deformation was/is accommodated at the orogen scale. Some authors suggest that orogenparallel stretching within the GHS is accommodated by orogen-parallel transportation of material into the east and west syntaxes where deformation is characterized by horizontal shortening and vertical stretching [McCaffrey and Nabelek, 1998;Butler et al., 2002;Whipp et al., 2014]. Alternatively, orogen-parallel stretching may have been and may still be restricted on a regional scale to oblate strain domains separated by transport-parallel prolate/plane strain domains [e.g., Sylvester and Janecky, 1988;Law, 2010], in response to radial spreading of the GHS and Tibetan midcrust [e.g., Brun and Merle, 1985;Merle, 1989]. Alternation between oblate and prolate/plane strain domains may occur along orogenic strike. Alternatively, oblate strain domains may be confined to the top of the GHS (i.e., the UGHS), while prolate/plane strain domains may be confined to the base of the GHS (i.e., the MCTZ) [e.g., Merle, 1989, Figure 8]. Similar models of orogen-parallel stretching and oblate flattening driven by gravitational radial spreading or oroclinal bending have been proposed for the Hudson Highlands [Gates, 1996], European Alps [Ratschbacher et al., 1991Ring, 1992], Carpathians [Schmid et al., 1998;Jeřábek et al., 2007], Betics [Williams and Platt, 2013], Sveconorwegian Orogen [Viola and Henderson, 2010], and Caledonides [Hossack and Cooper, 1986;Ellis and Watkinson, 1987;Sylvester and Janecky, 1988].
In the Himalaya, along-strike alternating oblate and prolate/plane strain domains provide plausible explanation for the regional-scale along-strike distribution of km 3 scale accumulations of leucogranite at intervals of 50-100 km (Figure 1), separated by regions that are comparatively depleted in leucogranite [Parsons et al., 2016b, and references therein]. In this situation, regional domains of oblate flattening strains may have driven orogen-parallel melt migration toward prolate/plane strain domains. Orogen-parallel magmatic lineations in the Manaslu leucogranite support the concept of orogen-parallel melt migration [Guillot et al., 1993;Coleman, 1996]. In the Garhwal Himalaya, Scaillet et al. [1995] made similar interpretations from field-macrostructural and AMS analyses of the Gangotri granite (Figure 1). These authors suggest that km 3 scale tabular laccoliths of the Gangotri granite represent crustal-scale boudins, which formed during oblate coaxial deformation [Rochette et al., 1994;Scaillet et al., 1995]. Additionally, kinematically distinct regional-scale deformation domains are proposed for the present-day crustal structure of the Tibetan Plateau, based on the spatial variation of seismic anisotropy [Sherrington et al., 2004].

Transition of Deformation Modes
Numerous studies from across the Himalaya have recorded a transition from orogen-perpendicular extrusion of the GHS to orogen-parallel stretching and extensional and strike-slip faulting between the THS and GHS at~13-22 Ma [Coleman and Hodges, 1995;Coleman, 1996;Coleman and Hodges, 1998; Tectonics 10.1002/2016TC004244 Hurtado, 2002;Jessup et al., 2008;Jessup and Cottle, 2010;Styron et al., 2011;Xu et al., 2013;Langille et al., 2014;Nagy et al., 2015]. It is likely that mineral stretching lineations in the UGHS and the STDS correspond only to the final stage of midcrustal deformation during which high-temperature deformation microstructures are set in [e.g., Knipe and Law, 1987;Parsons et al., 2016a]. As such, orogen-parallel stretching lineations in Unit III and the STDS may record the initial transition from midcrustal orogen-perpendicular extrusion of the UGHS to middle to upper crustal orogen-parallel extension and dextral transtension of the THS and upper portion of the GHS, during the mid-Miocene [Styron et al., 2011;Xu et al., 2013;Nagy et al., 2015].
We note that in the Annapurna-Dhaulagiri-Manaslu Himalaya and neighboring regions, the earliest records of E-W extension [e.g., Coleman and Hodges, 1998;Hurtado et al., 2001;Hurtado, 2002;Xu et al., 2013;Nagy et al., 2015] overlap with the cessation of midcrustal flow/pervasive shearing of the GHS which came to an end between 25 and 18 Ma [Parsons et al., 2016b, and references therein] but possibly began to shut down as early as~30 Ma [Carosi et al., 2016]. We also note that the timing of this transition overlaps with emplacement of the Mustang and Mugu granite plutons in the northern reaches of the Kali Gandaki Valley, ( Figure S9) at 24-23 Ma and~21-17 Ma, respectively, and emplacement of the Manaslu leucogranite pluton at 22-19 Ma [Harrison et al., 1997;Guillot et al., 1999;Harrison et al., 1999;Hurtado, 2002;Hurtado et al., 2007]. The Mustang and Mugu granites ( Figure S9) form the mylonitized footwall to the Dangardzong fault along the western margin of the Thakkhola graben [Harrison et al., 1997;Guillot et al., 1999;Hurtado et al., 2001;Hurtado, 2002]. The Manaslu leucogranite pluton and surrounding country rock preserve orogen-parallel magmatic and solid-state mineral stretching lineations, as identified from field macrostructures and rock magnetic fabrics [Guillot et al., 1993;Coleman, 1996]. Consequently, we suggest that in this region, the cessation of midcrustal flow, leucogranite pluton emplacement, and initiation of orogen-parallel stretching at midcrustal levels were kinematically linked. We propose that existing timing constraints indicate a cause and effect relationship between the cessation of crustal flow and the initiation of orogen-parallel extension, such that the two processes could not be effectively maintained simultaneously.
During cessation of flow, the channel strengthened to form a channel plug which mechanically recoupled the upper, middle, and lower crustal units by removing any rheological contrast between the channel and channel walls [Parsons et al., 2016b]. Orogen-parallel melt extraction from the Annapurna-Dhaulagiri Himalaya may have promoted the cessation of channel flow in this region and also provides an explanation for the lower-than-typical thickness of the UGHS (7 km), relative to other regions (10-30 km) in the Himalaya [Parsons et al., 2016b]. Following mechanical strengthening and recoupling, oblique convergence could no longer be accommodated through pervasive oblate strain of the UGHS. Subsequently, continued oblique convergence required strain partitioning in the upper crust between thrust faults and normal and strike-slip faults to maintain strain compatibility [e.g., Styron et al., 2011].
We note that the position of the kinematic discontinuity between Unit II and Unit III lineation orientations is coincident with the top-SW Kalopani Shear Zone [Vannay and Hodges, 1996;Carosi et al., 2016]. We also recognize a potential discontinuity between north to east plunging lineations in Unit I and north to northwest plunging lineations in Unit II (Figure 7). Similar kinematic discontinuities between orogen-perpendicular and orogen-parallel stretching directions within the GHS have been mapped elsewhere along the Himalaya Xu et al., 2013]. In many cases, these discontinuities are also metamorphic in nature and can be observed in the field as reverse sense shear zones, some of which lie coincident with the sillimanite-in isograd Xu et al., 2013;Montomoli et al., 2015;. It is possible that these discontinuities aided the partitioning of orogen-perpendicular and orogen-parallel stretching following the cessation of crustal flow and pervasive shearing of the GHS [cf. Montomoli et al., 2015;Carosi et al., 2016].
We suggest that similar deformation sequences occurred elsewhere in the Himalaya, where the GHS can be divided into portions of orogen-perpendicular and orogen-parallel stretching [e.g., Brun et al., 1985;Pêcher, 1991;Pêcher et al., 1991;Xu et al., 2013]. These models require further investigation to assess their validity, but they do provide coherent reasoning for why the two deformation modes (midcrustal flow/extrusion and upper crustal extension) were not sustained simultaneously. Additionally, while a cause and effect relationship appears to exist between the cessation of flow and the initiation of orogen-parallel extension, it is not clear which is the cause and which is the effect. It is possible that an external factor provided the catalysts to start the transition in deformation modes. This transition is reported from multiple locations along the Tectonics 10.1002/2016TC004244 Himalaya between 22 and 13 Ma [see Nagy et al., 2015, and references therein] and may reflect a major change in the boundary conditions governing the Himalaya-Tibet orogenic system. We support the proposal of Nagy et al. [2015] who suggest that this transition in deformation modes reflects a change in the balance of forces across the orogenic system between~20 and 10 Ma, caused by one or more of the following events: (1) an increase in the mean elevation of the Tibetan Plateau, (2) the removal of mantle lithosphere beneath the Tibetan Plateau, (3) the onset of eastward crustal flow of the Tibetan lower crust, and (4) a 35-45% decrease in convergence rate between Indian and Eurasia [Westaway, 1995;Clark and Royden, 2000;Royden et al., 2008;Molnar and Stock, 2009;Searle et al., 2011;Iaffaldano et al., 2013]. Such events highlight the importance of considering the rheological and mechanical boundary conditions of Composite Orogenic Systems in three dimensions [Parsons et al., 2016a].

Conclusions
Integrated AMS and structural analyses have been conducted across the Greater Himalayan Sequence (GHS) and bounding units in the Kali Gandaki Valley of the Annapurna-Dhaulagiri Himalaya, central Nepal. AMS analysis of 40 samples, accompanied by magnetic hysteresis and FORC analyses, reveals that magnetite, pyrrhotite, and phyllosilicate form the magnetic carriers of most samples. Consideration of magnetic carrier properties demonstrates a clear correlation between AMS and deformation fabrics and supports the use of these AMS fabrics as proxies for 3-D strain geometries and their kinematic interpretation. Correlation with previously published constraints indicates that AMS fabrics from the UGHS and base of the STDS provide a record of high-temperature, synmigmatitic to postmigmatitic deformation (>550-650°C).
Magnetic and mineral stretching lineations record orogen-perpendicular stretching in the MCTZ and Units I and II of the UGHS and are structurally overlain by orogen-parallel stretching in the STDS and Unit III of the UGHS. Shape parameter (T) analyses of AMS ellipsoids and eigenvalue analyses of lineation populations suggest that these data represent plane strain proxies in the MCTZ and oblate/suboblate strain proxies in the UGHS and STDS. The most oblate AMS ellipsoids (T = 0.9), which are recorded in migmatitic samples, correspond to synmigmatitic heterogeneous deformation under an oblate strain regime.
We interpret these data as an indication that channel flow in the Annapurna-Dhaulagiri Himalaya occurred under an oblate/suboblate strain regime. We propose that prior to upper crustal orogen-parallel extension, midcrustal oblate/suboblate strain during channel flow accommodated the obliquity between northward Indian Plate motion and the lateral pressure gradient induced by the overburden of the Tibetan Plateau and orientated perpendicular to the arcuate orogenic front.
Midcrustal orogen-parallel stretching and oblate flattening may have been accommodated through transportation of crustal material into the east and west syntaxes [e.g., Whipp et al., 2014] or by development of regional-scale oblate strain domains between prolate/plane strain domains [e.g., Sylvester and Janecky, 1988;Merle, 1989;Law, 2010]. The latter hypothesis may have promoted development of melt-depleted regions between km 3 scale leucogranite plutons emplaced at 50-100 km intervals along the length of orogen, via orogen-parallel melt migration.
During cessation of crustal flow, rheological strengthening of the UGHS and mechanical recoupling of the upper and lower crusts resulted in the initiation of upper crustal orogen-parallel extension as a means to maintain strain compatibility during continued oblique convergence. Timing constraints from the Annapurna-Dhaulagiri Himalaya and neighboring regions suggest that midcrustal flow and upper crustal orogen-parallel extension could not be effectively sustained simultaneously. We suggest that cessation of midcrustal flow/pervasive shearing and initiation of orogen-parallel extension share a "cause and effect" relationship such that the occurrence of one promoted the occurrence of the other. It is unclear which process occurred first, and it is possible that an external factor provided the catalyst to start the transition in deformation modes. We favor the proposal of Nagy et al. [2015] who suggest that an orogen-wide transition in deformation modes occurred in response to a change in balance of forces across the orogenic system between~10 and 20 Ma. Hightemperature orogen-parallel magnetic and mineral stretching lineations in the STDS and upper UGHS record the initial transition between deformation modes in the Annapurna-Dhaulagiri Himalaya.
Lastly, our 3-D strain observations cannot be accounted for by current numerical simulations of Himalayan midcrustal deformation due to their 2-D nature that implicitly simulates plane strain deformation.

10.1002/2016TC004244
Importantly, midcrustal orogen-parallel deformation within Composite Orogenic Systems [Parsons et al., 2016a] will have significant implications for rates of convergence, extrusion, exhumation, fluid flow, melt migration, heat advection, and the forces that balance them. Further efforts should be made to incorporate orogen-parallel deformation into tectonic models of Himalayan orogenesis.

Acknowledgments
Reviews and comments from an anonymous reviewer, Teresa Román-Berdiel, and Associate Editor Augusto Rapalini helped improve this manuscript. Suka Ghale, Basan Sherpa, and their Nepalese colleagues are thanked for their assistance during fieldwork. Dave Wallis (University of Oxford) provided helpful discussion during write-up. Research was supported by the Natural Environment Research Council (training grant NE/J50001X/1) and the Geological Society of London (Elspeth Matthews Research Fund) to A.J.P. and National Science Foundation grant EAR0711207 to R.D.L. Rock magnetism analyses were conducted on instrumentation acquired through National Science Foundation grant EAR0521558 to E.C.F. Research Councils UK funded open access publication. Data used are listed in the references, tables, and supporting information.