3.a. Suture zone units and associated assemblages

The timing of the India–Asia collision is often cited as occurring during the Paleocene (Rowley, Reference Rowley1996; Yin & Harrison, Reference Yin and Harrison2000; Hu et al. Reference Hu, Garzanti, Wang, Huang, An and Webb2016; Najman et al. Reference Najman, Jenks, Godin, Boudagher-Fadel, Millar, Garzanti, Horstwood and Bracciali2017; Parsons et al. Reference Parsons, Hosseini, Palin and Sigloch2020). The Indus–Tsangpo suture zone (Indus-Yarlung-Tsangpo suture zone, Yarlung Zangbo Ophiolite Zone, Liu et al. Reference Liu, Ju, Wei and Li2010) is the collisional boundary between rocks of Indian and Asian affinities to the south and north, respectively (Figures 1 and 2). Suture zones often involve multiple fault systems, recording high-strain and incorporating a wide range of deformed rock materials (Dewey, Reference Dewey1977; O’Brien, Reference O’Brien2001; Catlos & Çemen, Reference Catlos, Çemen, Catlos and Çemen2023). Metamorphosed igneous rocks within the Indus–Tsangpo suture zone and its associated mélange include ophiolites, serpentinite gabbro, volcanic rocks, blueschists and syn-tectonic high-Si, peraluminous granites (e.g. Thakur, Reference Thakur1981; DiPietro et al. Reference DiPietro, Pogue, Hussain, Ahmad, Macfarlane, Sorkhabi and Quade1999). Several studies have reported garnet and its associated conditions from suture zone high-pressure amphibolites, deformed blueschist-facies quartz schists and eclogite-facies metapelites (Honegger et al. Reference Honegger, Le Fort, Mascle and Zimmermann1989; Guilmette et al. Reference Guilmette, Hébert, Dupuis, Wang and Li2008, Reference Guilmette, Hébert, Wang and Villeneuve2009, Reference Guilmette, Hébert, Dostal, Indares, Ullrich, Bédard and Wang2012; Cai & Cao, Reference Cai and Cao2013; Chen et al. Reference Chen, Schertl, Gu, Zheng, Xu, Zhang, Cai and Lin2021; Li et al. Reference Li, Yin, Yakymchuk, Ding, Li, Qian, Gao and Zhang2024).

Garnet-bearing assemblages are also found in deformed island arc rocks located between the Eurasian plate and the Indian plate along the Indus–Tsangpo suture zone. For example, the Kohistan–Ladakh Arc lies adjacent to the suture zone and represents an island arc that initially formed above a subduction zone within the Neo-Tethys Ocean (Fig. 1) (Shah et al. Reference Shah, Sayab, Aerden and Khan2011; Petterson, Reference Petterson2019). Garnet-bearing assemblages within the Kohistan-Ladakh Arc include those associated with magmatic rocks, garnet–kyanite–staurolite gneisses and mafic–ultramafic gabbroic assemblages (Raz & Honegger, Reference Raz and Honegger1989; DiPietro et al. Reference DiPietro, Pogue, Hussain, Ahmad, Macfarlane, Sorkhabi and Quade1999; Petterson, Reference Petterson2010; Thanh et al. Reference Thanh, Sajeev, Itaya and Windley2011; Sayab et al. Reference Sayab, Shah and Aerden2016; Jagoutz et al. Reference Jagoutz, Bouilhol, Schaltegger and Müntener2019; Petterson, Reference Petterson2019; George et al. Reference George, Waters, Gough, Searle and Forshaw2022).

Garnet-bearing magmatic rocks are also exposed near the Indus–Tsangpo suture zone due to subduction-related magmatism before and after collision. These include garnets documented in anorthosites, two-mica granites, leucocratic dikes and leucosomes associated with the Kohistan, Karakoram, Ladakh and Gangdese batholiths (Fig. 1) (Reichardt et al. Reference Reichardt, Weinberg, Andersson and Fanning2010; St-Onge et al. Reference St-Onge, Rayner and Searle2010; Ma et al. Reference Ma, Wang, Kerr, Yang, Xia, Ou, Yang and Sun2017; Ding et al. Reference Ding, Zhang, Palin, Kohn, Niu, Chen, Qin, Jiang and Li2022; Dong et al. Reference Dong, Weinberg, Zhu, Green, Yi, Cawood, Li and Chen2024).

Himalayan rocks exposed directly south of the suture zone are part of the Tethyan Himalayan Sequence, which consists of Paleoproterozoic to Eocene Indian shelf sedimentary rocks interbedded with Paleozoic and Mesozoic volcanic assemblages (Jadoul et al. Reference Jadoul, Berra and Garzanti1998; Yin & Harrison, Reference Yin and Harrison2000; Yin, Reference Yin2006; Bhargava & Singh, Reference Bhargava and Singh2020). The Tso Morari Crystalline Complex in NW India is often spatially associated with the Tethyan Himalayan Sequence (e.g. Steck et al. Reference Steck, Epard, Vannay, Hunziker, Girard, Morard and Robyr1998; Rao & Rai, Reference Rao and Rai2006) (Fig. 1). Garnet-bearing rocks in eclogite-facies assemblages record ultrahigh-pressure conditions closely associated with tectonic processes associated with collision (e.g. Wilke et al. Reference Wilke, O’Brien, Schmidt and Ziemann2015; Jonnalagadda et al. Reference Jonnalagadda, Karmalkar and Duraiswami2019). In contrast, upper structural levels of the Tethyan Himalayan Sequence in NW India contain garnet-bearing phyllites, schists, amphibolites and kyanite-biotite migmatites (Searle et al. Reference Searle, Metcalfe, Rex and Norry1993; Catlos et al. Reference Catlos, Perez, Lovera, Dubey, Schmitt and Etzel2020; Kawabata et al. Reference Kawabata, Imayama, Bose, Yi and Kouketsu2021; Sen et al. Reference Sen, Dey and Sen2023). Some Oligocene to Miocene-aged North Himalayan granitic bodies and gneiss domes formed within the Tethyan Himalayan Sequence are garnet-bearing (Fig. 2) (Lee et al. Reference Lee, Hacker, Dinklage, Wang, Gans, Calvert, Wan, Chen, Blythe and McClelland2000; Zhang et al. Reference Zhang, Harris, Parrish, Kelley, Zhang, Rogers, Argles and King2004; Smit et al. Reference Smit, Hacker and Lee2014; Wang et al. Reference Wang, Zeng, Gao, Chen and Li2022a).

3.b. Himalayan core

The Tethyan Himalayan Sequence is separated from the underlying higher-grade schists and gneisses of the Greater Himalayan Crystallines Complex (GHC) by the north-dipping South Tibet Detachment (STD) (Figs. 1 and 2) (e.g., Burchfiel et al. Reference Burchfiel, Zhiliang, Hodges, Yuping, Royden, Changrong and Jiene1992; Hodges et al. Reference Hodges, Parrish, Housh, Lux, Burchfiel, Royden and Chen1992; Kohn, Reference Kohn2014; Carosi et al. Reference Carosi, Montomoli and Iaccarino2018; Kellett et al. Reference Kellett, Cottle and Larson2019; Wang et al. Reference Wang, Larson, Zhang, Zhao and Wu2024). The GHC is a unit of primarily kyanite to sillimanite grade gneisses intruded by High Himalayan leucogranites (HHL) in its upper portion (Upreti, Reference Upreti1999; Wu et al. Reference Wu, Liu, Liu, Wang, Xie, Wang, Ji, Yang, Liu, Khanal and He2020; Cao et al. Reference Cao, Pei, Santosh, Li, Zhang, Zhang, Zhang, Zou, Dai, Lin, Tang and Yu2022; Zhang, Reference Zhang2024). Both lithologies are garnet-bearing. The GHC protolith is suggested to be the pre-Himalayan, Indian plate margin sediments intruded by Cambrian–Ordovician granitoids that were thrust beneath Tibet during the early stages of the Himalayan collision (Catlos, Reference Catlos, Catlos and Çemen2023; Robinson & Martin, Reference Robinson, Martin, Catlos and Çemen2023). Garnets are also present in some HHL (Searle & Fryer, Reference Searle and Fryer1986; Weinberg, Reference Weinberg2016; Wu et al. Reference Wu, Liu, Liu, Wang, Xie, Wang, Ji, Yang, Liu, Khanal and He2020; Shi et al. Reference Shi, He, Zhao, Liu, Harris and Zhu2021).

The highest-grade assemblages associated with the GHC are granulite metapelites and coesite-bearing eclogites (O’Brien et al. Reference O’Brien, Zotov, Law, Khan and Jan2001; Borghi et al. Reference Borghi, Castelli, Lombardo and Vison2003; Zhang et al. Reference Zhang, Ding, Palin, Dong, Tian, Kang, Jiang, Qin and Li2022). The GHC may have experienced two metamorphic episodes (see review in Catlos, Reference Catlos, Catlos and Çemen2023). The first Eohimalayan event (phase) occurred during the Eocene–Oligocene (ca. 52–30 Ma) and is marked by accretion and metamorphism (e.g., Kelly et al. Reference Kelly, Beaumont and Jamieson2022; Wang et al. Reference Wang, Wu, Zhang, Khanal and Yang2022b). During the subsequent Miocene Neohimalayan event (phase), slip along the Main Central Thrust (MCT) was initiated, and Miocene HHL were generated (e.g., Wu et al. Reference Wu, Liu, Liu, Wang, Xie, Wang, Ji, Yang, Liu, Khanal and He2020).

At its base, the GHC is thrust over lower-grade metasedimentary rocks of the Lesser Himalayan Sequence (LHS) along the MCT (Figs. 1 and 2) (Upreti, Reference Upreti1999; Robinson & Martin, Reference Robinson and Martin2014). The lack of an apparent break in metamorphic grade between the GHC and LHS makes it challenging to place the boundaries of the MCT zone (e.g., Martin, Reference Martin2017). The MCT zone is characterized by inverted metamorphism, where metamorphic grade increases toward structurally shallower levels (e.g., Catlos et al. Reference Catlos, Harrison, Kohn, Grove, Ryerson, Manning and Upreti2001; Larson et al. Reference Larson, Kellett, Cottle, King, Lederer and Rai2016; Carosi et al. Reference Carosi, Montomoli and Iaccarino2018; Pant et al. Reference Pant, Singh and Jain2020). The LHS lies in the hangingwall of the MBT (Meigs et al. Reference Meigs, Burbank and Beck1995; Mugnier et al. Reference Mugnier, Huyghe, Chalaron and Mascle1994), which places it structurally above the Siwalik Group. The LHS is mainly comprised of Paleoproterozoic sedimentary rocks associated with Gondwana and is intruded by granite dated to ∼1.84 Ga and a narrow belt of younger (Permian to pre-Middle Miocene) rocks distributed in its southern margin (e.g., Mukhopadhyay et al. Reference Mukhopadhyay, Bhadra, Ghosh and Srivastava1996; Miller et al. Reference Miller, Klötzli, Frank, Thöni and Grasemann2000; Mishra et al. Reference Mishra, Singh, Slabunov, Nainwal, Singh, Chaudhary and Nainwal2019). To the south, the Main Frontal Thrust (MFT) forms the boundary between the Siwalik Group and the Indo-Gangetic Plain (e.g., Burgess et al. Reference Burgess, Yin, Dubey, Shen and Kelty2012; Srivastava et al. Reference Srivastava, Mukul and Barnes2016).

The Siwalik Group is a thick succession of dominantly fluvial, coarsening upward, sedimentary rocks located along the entire 2400 km length of the Himalaya from Pakistan’s Potwar plateau on the west to the Brahmaputra valley in the east (Fig. 1) (Burbank et al. Reference Burbank, Beck, Mulder, Yin and Harrison1996; Bora & Shukla, Reference Bora and Shukla2005; Bernet et al. Reference Bernet, Van Der Beek, Pik, Huyghe, Mugnier, Labrin and Szulc2006; Sanyal & Sinha, Reference Sanyal and Sinha2010; Khan et al. Reference Khan, Bera, Spicer, Spicer and Bera2019; Dhamodharan et al. Reference Dhamodharan, Rawat, Kumar and Bagri2020). Siwalik sedimentary rocks primarily originated in the Himalaya, and sedimentation occurred in a foreland basin with alluvial fans, fluvial mega-cones, braided channels and flood plains, similar to those that developed the present-day Indo-Gangetic Plain (Parkash et al. Reference Parkash, Sharma and Roy1980; Jain & Sinha, Reference Jain and Sinha2003). Its formation occurred due to the evolution of large river systems analogous to those associated with the Ganga River system (Jain & Sinha, Reference Jain and Sinha2003; Bora & Shukla, Reference Bora and Shukla2005; Sanyal & Sinha, Reference Sanyal and Sinha2010; Taral & Chakraborty, Reference Taral and Chakraborty2018; Khan et al. Reference Khan, Bera, Spicer, Spicer and Bera2019; Dhamodharan et al. Reference Dhamodharan, Rawat, Kumar and Bagri2020). As Fig. 3 demonstrates major river systems, such as the Ganges, Indus and Brahmaputra, along with their tributaries, currently drain extensive areas of the Indian subcontinent. The deposits record a range of environments, including piedmonts, outwash plains, channels, floodplains and oxbow lakes, some of which have marine influence in older sections (Taral et al. Reference Taral, Chakraborty, Huyghe, Van Der Beek, Vögeli and Dupont-Nivet2019; Debnath et al. Reference Debnath, Taral, Mullick and Chakraborty2021; Khan et al. Reference Khan, Mahato, Spicer, Spicer, Ali, Hazra and Bera2023). Siwalik Group sediment accumulation began in the Middle Miocene in a long foredeep near sea level (Bora & Shukla, Reference Bora and Shukla2005; Chakrabarti, Reference Chakrabarti and Chakrabarti2016). The foreland basin developed as the subducting Indian plate flexed under the crustal load of the rising Himalaya (Lavé & Avouac, Reference Lavé and Avouac2000; Rai, Reference Rai2003; Valdiya, Reference Valdiya2016; Dutta et al. Reference Dutta, Biswas and Mukherjee2019).

Figure 3.

Primary basement structural features of the Indo-Gangetic Plain, after Catlos (Reference Catlos, Catlos and Çemen2023).

3.c. Potential Proterozoic mobile belt sources of the Indian plate

Although the dominant documented source of sediment in the Siwalik Group is from Himalayan core assemblages, rocks from the Indian Shield also have the potential to contribute through erosion and river transport (Khan & Tewari, Reference Khan and Tewari2015; Chakraborty et al. Reference Chakraborty, Taral, More, Bera, Gupta and Tandon2020). Figure 3 illustrates the Indo-Gangetic Plain, which lies south of the Himalaya and spans northern and eastern India, Bangladesh, parts of Pakistan and southern Nepal (Pant & Sharma, Reference Pant and Sharma1993; Pathak et al. Reference Pathak, Dagar, Kaushal, Chaturvedi, Dagar, Singh and Arunachalam2014). The region is classified into distinct sections based on geography, river systems and climate (Pathak et al. Reference Pathak, Dagar, Kaushal, Chaturvedi, Dagar, Singh and Arunachalam2014; Catlos, Reference Catlos, Catlos and Çemen2023). If contributions to the Siwalik Group include sediments from proximal sources south of the Himalaya, mixing Himalayan detritus with garnet-bearing Proterozoic mobile belt sources would reflect a dynamic interplay between interconnected river systems.

Ancient rivers may have had axial components that transported sediments eroded from Precambrian basement rocks, including granites, gneisses and schists, into the Himalayan foreland basin, where Siwalik sediments accumulated (Burbank et al. Reference Burbank, Beck, Mulder, Yin and Harrison1996; Ulak, Reference Ulak2005; Mandal et al. Reference Mandal, Sarkar, Chakraborty, Bose, Liu and Wang2014). While the primary sources of garnet-bearing material in Siwalik exposures are linked to the Himalaya, some units in northwestern India contain similar mineralogical signatures. These include the Chotanagpur Granite Gneiss Complex (CGGC) (Sanyal & Sengupta, Reference Sanyal and Sengupta2012; Dey et al. Reference Dey, Choudhury, Mukherjee, Sanyal and Sengupta2019), Aravalli-Delhi Fold Belt (Prakash et al. Reference Prakash, Saha, Petrik, Janak and Bhattacharya2018) and Shillong-Meghalaya Plateau (Chatterjee et al. Reference Chatterjee, Bhattacharya, Duarah and Mazumdar2011; Chatterjee, Reference Chatterjee2017). These regions contain garnets that record ultra-high temperature conditions and share Proterozoic zircon dates in the Siwalik Group (Mukherjee et al. Reference Mukherjee, Dey, Sanyal, Ibanez-Mejia and Sengupta2019; Singh et al. Reference Singh, De Waele, Shukla, Umasankar and Biswal2021; Nag et al. Reference Nag, Hrushikesh, Cogné and Prabhakar2024). While these potential sources are considered, our primary focus is on assessing garnet compositions in the context of Himalayan provenance.

6.a. Garnet grain size and comparison to palaeohydrological indicators

Table 2 summarizes palaeohydrological results from samples collected from locations equivalent to those in this study (Ulak, Reference Ulak2005). These data provide insights into sediment transport dynamics and offer an understanding of the environmental conditions during deposition. We include our measured garnet grain sizes from these samples, allowing for a direct comparison to the D50 and D95 values from equivalent samples. D50 represents the median grain size, where 50% of the sediment sample (by weight) is fine-grained, and 50% is coarse-grained, while D95 represents the fine end-member grain, where 95% of the sediment sample is finer and only 5% is coarser.

Table 2.

Summary of paleohydrology estimates for sampled locations (data after Ulak, Reference Ulak2005)

* PCB = Planar Cross-Bedding, TCB = Trough Cross-Bedding, RL = Ripple Lamination

† Ucr = Upper Critical Velocity, Urd = Upper Regime Deposition Velocity, Uup = Upper Unidirectional Flow Velocity, AV = Average velocity

^§ D50 = Median grain size, in which 50% of the sediment sample (by weight) is finer and 50% is coarser. D95 = Coarse end-member, the grain size at which 95% of the sediment sample is finer and only 5% is coarser.

This study separated garnet grains using a sieve-based method, with a maximum observable size of 260 µm. However, only two samples (SK17 from the Dobata Formation and SK11 from the Surai Khola Formation) contained garnets that reached this size, suggesting that garnets from these rocks were large enough to be transported alongside the coarsest fraction of the sediment. In contrast, the maximum garnet size in other samples was significantly smaller than the D50 and D95 values of equivalent samples, which have implications for sediment sorting, transport behaviour and provenance interpretations.

For example, the sample from the same location as SK7, within the Shivagarhi Member, records the highest-energy conditions, characterized by the greatest discharge (3,430 m³/s), steepest slope (5.49%) and highest upper-velocity limit (2.8 m/s). These conditions are consistent with an environment capable of transporting coarse sediments from an actively uplifting hinterland. This sample has the greatest flow depth (4.7 m), indicative of substantial fluvial power and a well-developed system. However, in SK7, the largest observed garnet was 163 µm, significantly finer than the equivalent sample’s median grain size (D50 = 390 µm) and much smaller than the coarsest transported grains (D95 = 670 µm). This size discrepancy suggests that garnets were deposited with finer-grained material rather than the dominant sand fraction. Possible explanations include selective sorting due to the garnet’s higher density and differential transport behaviour compared to lower-density grains (quartz, feldspar or lithic fragments) or a provenance signal indicating a limited supply of coarser-sized garnets.

Sample SK17, collected from our uppermost stratigraphic level of the Dobata Formation, retains high-energy transport characteristics, but with more moderate discharge, slope and velocity than SK7 (Table 2). The channel depth is lower than that of SK7 but still indicates a moderate fluvial system capable of transporting sand-sized sediments. Garnets from SK17 reached the maximum sieve size (260 µm), suggesting they were transported alongside the dominant sand fraction rather than being selectively sorted into finer sediments. Notably, SK17 was the only sample to exhibit planar cross-bedding structures, consistent with the migration of large sandbars or dunes under moderate- to high-energy conditions. This observation supports the interpretation that SK17 represents a dynamic but less extreme depositional setting than SK7.

Data from samples between these stratigraphic levels, including those equivalent to SK8 (Shivagarhi Member), SK11 (Surai Khola Formation) and SK16 (Dobata Formation), indicate lower and more variable transport energies compared to SK17 and SK7 (Table 2). Ulak (Reference Ulak2005) reports data from two samples at each location, suggesting episodic variations in flow energy. For instance, one sample from the SK16 location indicates fine-grained deposition, whereas another is associated with coarser sediments and more variable transport energy. This variability could reflect deposition in different fluvial sub-environments, such as floodplains, channels, levees or crevasse splays. The reported discharge values for these locations are 128 and 339 m³/s, significantly lower than those of SK7 and SK17, further supporting the interpretation of reduced flow energy. In sample SK16, the largest observed garnet was 194 µm (Table 2). When compared to bulk sediment grain-size distributions, garnets were slightly coarser than the median (D50 = 170 µm) in one equivalent sample but significantly finer than the median (D50 = 520 µm) in another. Similarly, garnet size approached the D95 value (230 µm) in the finer-grained dataset, whereas it was much finer than the D95 value (990 µm) in the coarser-grained dataset. These results indicate that garnets in SK16 were transported under variable hydrodynamic conditions, depending on local depositional energy and grain-size distributions. In finer-grained sediments, garnets would have been among the coarser components. However, garnets were significantly finer than expected based on the D50 and D95 values characterizing the coarser-grained sediments. This pattern suggests that garnets were deposited under flow conditions capable of moving medium- to fine-grained sand but not necessarily the coarsest fraction. Such variability may reflect differences in local depositional energy, sediment input or selective sorting processes.

Discharge estimates for two samples from the SK11 location varied significantly (16 and 492 m³/s), consistent with the episodic flow energy variations observed. This sample also contained garnets that reached the maximum sieve size, suggesting that at least some garnets were transported alongside the dominant sand fraction. In contrast, data from rocks from the SK8 location exhibited the lowest transport energy in the dataset. These samples contained finer-grained sediments, with the lowest discharge rates and slopes recorded. Additionally, one equivalent sample displayed ripple laminations commonly associated with overbank deposits or shallow water environments (Taral et al. Reference Taral, Kar and Chakraborty2017; Rai & Yoshida, Reference Rai and Yoshida2021). In sample SK8, the largest observed garnet was only 70 µm in diameter. Compared to bulk sediment grain-size distributions, garnets were significantly finer than the median grain size (D50 = 170–190 µm) and much finer than the D95 values (290–360 µm). This observation suggests that garnets were deposited under conditions favouring fine-particle transport, potentially due to selective sorting or differences in the original grain size of the source material.

Palaeocurrent directions in the Surai Khola section indicate that transverse rivers, flowing perpendicular to the mountain front (N-S), became the primary sediment suppliers in the later stages of the Middle Miocene (Burbank et al. Reference Burbank, Beck, Mulder, Yin and Harrison1996). Szulc et al. (Reference Szulc, Najman, Sinclair, Pringle, Bickle, Chapman, Garzanti, Andò, Huyghe, Mugnier, Ojha and DeCelles2006) suggest that the Surai Khola lacks evidence for sustained axial flow (WNW or ESE). However, their dataset includes only one direction from the Dobata Formation, which showed NE-directed flow. This observation suggests that a transverse river was possible at that specific location. The dataset from Nakayama & Ulak (Reference Nakayama and Ulak1999) and Ulak (Reference Ulak2005) includes numerous samples with paleoflow indicators that support the presence of an axial drainage system in the Dobata and Dhan Khola units, as well as intermittently throughout the section, particularly in the lower part of the Surai Khola Formation. Their results are consistent with a possible transverse river input by ∼4 Ma, persisting into the Dhan Khola Formation.

Quade et al. (Reference Quade, Cater, Ojha, Adam and Mark Harrison1995) and Hoorn et al. (Reference Hoorn, Ohja and Quade2000) documented a major environmental shift from C₃- to C₄-dominated ecosystems in the upper part of the Middle Siwalik (MS1) around 6.5 Ma. This transition occurs at a stratigraphic level slightly above the Surai Khola marker, as shown in Fig. 4b. The more recent identification of a second phase of vegetation change, characterized by the expansion of C₃ plants during the last 3 Myr, appears unique to the Surai Khola section and may have been influenced by sediment recycling at 3–4 Ma (Charreau et al. Reference Charreau, Lavé, France-Lanord, Puchol, Blard, Pik and Gajurel2021; Roy et al. Reference Roy, Ghosh and Sanyal2020). This recycling could be linked to a shift from an extensive, trans-Himalayan river system to one primarily draining the LHS and a Siwalik piedmont (Charreau et al. Reference Charreau, Lavé, France-Lanord, Puchol, Blard, Pik and Gajurel2021).

Overall, the reported paleohydrology results indicate that sediment transport within the Surai Khola Siwalik section was episodic and spatially variable (Table 2). Specific samples (e.g., SK7 and SK17) record high-energy transport, likely linked to hinterland exhumation and localized shifts in discharge or slope. In contrast, others (e.g., SK8 and SK16) reflect lower-energy conditions. Given the depositional ages of SK7 and SK17 (9.5 Ma and 3.8 Ma, respectively), these events likely occurred under differing climatic and tectonic conditions. The 6.5 Ma vegetation shift (e.g., Quade et al. Reference Quade, Cater, Ojha, Adam and Mark Harrison1995; Hoorn et al. Reference Hoorn, Ohja and Quade2000) marks an important climatic transition; however, sediment transport intensity may also be modulated by local slope dynamics, discharge variability or tectonic uplift. These variations may reflect transient sediment routing responses to changing sediment supply and transport capacity within a dynamic fluvial system. This episodic behaviour aligns with that of modern Himalayan rivers, where peak discharges, often driven by the monsoon, can periodically flush large volumes of sediment downstream (Singh et al. Reference Singh, Singh and Müller2007; Clift, Reference Clift2020). While regional climate likely influenced discharge patterns, the garnet-bearing samples appear to reflect tectonic controls on sediment availability and grain size. Larger garnets are generally found in samples associated with higher-energy, coarser-grained environments (e.g., SK17), while smaller grains dominate finer-grained, lower-energy settings (e.g., SK8 and SK7).

6.b. Geochronological constraints on sediment provenance and exhumation

Detrital white mica 40Ar/39Ar, zircon U–Pb, and zircon and apatite fission track (ZFT and AFT) dates from equivalent samples of the Surai Khola section have been previously reported in isolation (Bernet et al. Reference Bernet, Van Der Beek, Pik, Huyghe, Mugnier, Labrin and Szulc2006; Szulc et al. Reference Szulc, Najman, Sinclair, Pringle, Bickle, Chapman, Garzanti, Andò, Huyghe, Mugnier, Ojha and DeCelles2006; van der Beek et al. Reference Van Der Beek, Robert, Mugnier, Bernet, Huyghe and Labrin2006; Baral et al. Reference Baral, Lin and Chamlagain2016). This section synthesizes these independent datasets relevant to the samples in this study. By integrating existing age constraints with palaeohydrological records, we can better assess their provenance and how fluvial processes modulated sediment flux from the hinterland to the foreland basin.

Szulc et al. (Reference Szulc, Najman, Sinclair, Pringle, Bickle, Chapman, Garzanti, Andò, Huyghe, Mugnier, Ojha and DeCelles2006) report detrital white mica 40Ar/39Ar dates from the same stratigraphic level as SK16 (4 Ma) and samples closely matching the depositional ages of SK11 (6.8 Ma; equivalent sample at 7 Ma), SK8 (8.8 Ma; equivalent sample at 8.9 Ma) and SK7 (9.5 Ma; equivalent sample at 10 Ma). The youngest mica 40Ar/39Ar dates for each sample fall within the Middle to Late Miocene at 16.1±0.3 Ma (SK8), 13.8±3.4 Ma (SK11), 12.1±0.1 Ma (SK16) and 11.4±1.2 Ma (SK7). These results indicate that no detrital white mica grains record active exhumation synchronous with deposition. Instead, the sediment was derived from previously exhumed and cooled source terrains. Lag times, estimated as the difference between the youngest mica cooling age and depositional age, vary among the samples: ∼2 Myr (SK7), ∼7 Myr (SK11), ∼7.2 Myr (SK8) and ∼8 Myr (SK16). The shorter lag time of SK7 is consistent with the observation that it was deposited under the highest-energy conditions. The longer lag times in the other samples reflect the variable depositional energy conditions, as described in the previous section.

These lag times differ from those estimated using AFT dates, which suggest a constant lag time of 0.8±0.5 Myr due to rapid source-area exhumation rates of ∼1.8 km/Myr since ∼7 Ma (van der Beek et al. Reference Van Der Beek, Robert, Mugnier, Bernet, Huyghe and Labrin2006). The difference likely reflects the sensitivity of these thermochronometers to different closure temperatures, with AFT recording more recent cooling histories. Detrital white mica grains were sourced from previously exhumed rocks, whereas AFT records more recent cooling closer to the depositional period. The AFT and mica geochronology studies indicate that the Surai Khola section underwent sediment recycling, with the Upper Siwalik sedimentary rocks more strongly affected than the underlying units (Szulc et al. Reference Szulc, Najman, Sinclair, Pringle, Bickle, Chapman, Garzanti, Andò, Huyghe, Mugnier, Ojha and DeCelles2006; van der Beek et al. Reference Van Der Beek, Robert, Mugnier, Bernet, Huyghe and Labrin2006).

Figure 7 presents the detrital 40Ar/39Ar white mica dates from equivalent samples in a radial plot, illustrating their distribution, age peaks, and central and weighted mean dates. A radial plot enables the comparison of age estimates of varying precision and the visualization of mixed-age populations while accounting for individual uncertainties. It is beneficial for identifying outliers (Galbraith, Reference Galbraith1990; Vermeesch, Reference Vermeesch2009). Central peak and calculated weighted mean dates highlight dominant exhumation phases. Data from samples SK16, SK11 and SK7 indicate a primary sediment source with exhumation pulses in the Middle Miocene, whereas SK8 suggests derivation from regions that cooled during the Late to Middle Miocene. The consistent Middle Miocene dates (∼14–19 Ma) across all samples suggest a common source region. The dates also align with ZFT dates from the section (∼16 Ma) (Bernet et al. Reference Bernet, Van Der Beek, Pik, Huyghe, Mugnier, Labrin and Szulc2006). The absence of younger white mica 40Ar/39Ar dates (<10 Ma) also indicates that by ∼9.5–4 Ma, actively exhuming regions were not significantly contributing to the sediment load. Instead, sediment was sourced from terrains that had already undergone exhumation and cooling by the Middle Miocene.

Figure 7.

Radial plots for ⁴⁰Ar/³⁹Ar mica dates on data from Szulc et al. (Reference Szulc, Najman, Sinclair, Pringle, Bickle, Chapman, Garzanti, Andò, Huyghe, Mugnier, Ojha and DeCelles2006). Sample numbers include both SK identifiers and those from original publications. Central and peak ages are labelled. WMA = Weighted Mean Ages.

Older 40Ar/39Ar age peaks in each sample (e.g., 21–26 Ma, 31–45 Ma) may reflect contributions from additional sources or excess argon, yielding artificially older apparent dates. Caution is needed in interpreting detrital white mica 40Ar/39Ar dates. However, these older ages are not uncommon in the Himalaya, and muscovite is generally less prone to incorporating atmospheric Ar than other minerals (Stuart, Reference Stuart2002).

U–Pb zircon dates are also reported from equivalent sample locations in the Surai Khola section (Bernet et al. Reference Bernet, Van Der Beek, Pik, Huyghe, Mugnier, Labrin and Szulc2006; Baral et al. Reference Baral, Lin and Chamlagain2016). We focus on data from Baral et al. (Reference Baral, Lin and Chamlagain2016) due to the large number of zircon dates from samples at the same stratigraphic level as SK11 as well as the same formation and nearby locations for SK8 and SK16. Fig. 8a–c shows radial plots for these samples that show dispersion consistent with a mixed population. However, each sample exhibits consistent age peaks at ∼550 Ma, 900 Ma, 1.2 Ga, 1.8 Ga and 2.5 Ga. These age distributions closely resemble those of previous studies, which show that Siwalik sedimentary rocks in western Nepal yield U–Pb zircon populations of 460–530 Ma, ∼850–1200 Ma, ∼1.8–2.0 Ga and ∼2.5 Ga (DeCelles et al. Reference DeCelles, Gehrels, Quade, Ojha, Kapp and Upreti1998). Zircons from the Tethyan Himalayan Sequence are typically associated with U–Pb age peaks at ∼500 Ma and 1 Ga, while those from the GHC sequence cluster around ∼1.1 Ga, with minor peaks at ∼1.50 Ga, 1.7 Ga and 2.5 Ga. Upper formations of the GHC and granitic rocks from the LHS also yield U–Pb zircon dates ∼500 Ma (DeCelles et al. Reference DeCelles, Gehrels, Quade, LaReau and Spurlin2000, Reference DeCelles, Gehrels, Najman, Martin, Carter and Garzanti2004; Gehrels et al. Reference Gehrels, Kapp, DeCelles, Pullen, Blakey, Weislogel, Ding, Guynn, Martin, McQuarrie and Yin2011).

Figure 8.

(a–d) Radial plots for U–Pb zircon dates from Baral et al. (Reference Baral, Lin and Chamlagain2016), showing central and peak ages. Both our sample numbers and previously reported sample identifiers are included for reference. (d) U–Pb zircon dates from High Himalayan leucogranites, after Liu et al. Reference Liu, Zhu, Wang, Cawood, Stockli, Stockli, Lin, Zhang, Zhang and Zhao(2022). All analyses yield p(χ²) = 0.

Overall, past geochronological work in the Surai Khola on samples taken from stratigraphic levels similar to those in this study (<9.5 Ma) suggests that sediment source regions remained broadly consistent over time. This indicates long-term stability in sediment routing, with shifts primarily reflecting changes in the relative mixing proportions of different source terrains rather than abrupt provenance changes.

7.a. Garnet compositions and groupings

Garnet compositional profiles are categorized into nine groups based on both manual classification (zoning-based) and PCA clustering. Most garnets exhibit low XMg values (<0.1) and display flat zoning. Those garnet groups that show higher XMg (>0.1) are denoted by an asterisk (*). Figure 5 is intended to be representative of the range of zoning profiles and, therefore, highlights garnets with compositional variation. Specifically, garnets that do not exhibit flat zoning are from Groups 2–6 and include samples SK11-24, SK11-17, SK11-1, SK11-9, SK11-14, SK17-10, SK16C-3, SK16B-14, SK17-13 and SK17-19. These garnets were selected for modelling their pressure–temperature conditions and paths.

Groups 1 and 2 are Ca-rich garnets found exclusively in SK11, SK16 and SK17 (Fig. 5a and b). Group 1 garnets have XCa > XFe with low XMn + XMg, while Group 2 garnets show more variability in their XFe, XMn and XMg contents. These Ca-rich garnets are low in Cr and Ti, consistent with previous findings that disregarded the uvarovite and hydrogrossular end-members (Yoshida et al. Reference Yoshida, Nakajima, Matsumoto, Osaki, Rai, Cruz and Sakai2021). Group 3 garnets are spessartine (SK11), characterized by high MnO (26–28 wt%), low CaO and MgO contents and mixed zoning, with one grain exhibiting flat zoning and the other displaying a bell-shaped profile (35μm in diameter).

Groups 4–7 garnets are almandine with varying XMn, XCa and XMg. Most have low XMg, except for those designated in Group 6* (n=1) and 7* (n=8) (Fig. 5). Group 4 and Group 5 garnets are almandine-spessartine but differ in XCa. Group 4 garnets were found in samples SK7, SK11 and SK17. Group 5 and 7 garnets were present in all samples. Group 6 garnets were found in all samples except SK8. Group 6 and Group 7 garnets are similar but differ in their XCa and XMn compositions. These are almandine and possess higher XFe contents than the other garnet groups.

Figure 5 provides examples of the relationship between clustering and grouping for specific garnet analyses, whereas Figure 6c shows the percentage distribution of garnets across the classification approaches. Some garnet clusters were exclusively linked to manual groups (i.e. Cluster 4 and 6 are linked to Groups 7* and 1, respectively, Fig. 6c). However, garnet groups with compositional zoning from rim-to-rim are part of multiple clusters (i.e. Group 4 garnets are assigned in Clusters 0, 3, 4, 5 and 7, Fig. 5d). These observations indicate that the PCA prioritizes compositional variance over zoning trends that are best observed manually.

7.b. Garnet discrimination approaches

Figure 9 illustrates the results of garnet groups using the approach of Schönig et al. (Reference Schönig, Von Eynatten, Tolosana-Delgado and Meinhold2021). The ‘setting’ scheme discriminates for garnet sourced from mantle rocks (MA), igneous rocks (IG), metasomatic rocks (MS) and metamorphic rocks (MM) based on the votes for each class. Siwalik garnets exhibit votes from the MM, IG and MS categories (Fig. 9a). The metamorphic category dominates in all samples except SK8, which shows more igneous affinities. The classification scheme also classifies some metasomatic garnets in samples SK11, SK16 and SK17, although at lower amounts (2.6%, 23.4% and 2.0%, respectively).

Figure 9.

Kernel density plots showing distribution of classification votes for: (a) Setting classes; (b) Metamorphic facies; (c) Composition types, based on Schönig et al. (Reference Schönig, Von Eynatten, Tolosana-Delgado and Meinhold2021). Abbreviations: MS = metasomatic; IG = igneous; MA = mantle; MM = metamorphic; BS/GS = blueschist/greenschist; AM = amphibolite; GR = granulite; EC/UHP = eclogite/ultrahigh-pressure; CS = calc-silicates; IF/S = intermediate–felsic/metasedimentary; UM = ultramafic; M = mafic. (d) Bar plot showing maximum percentage of each manual group per sample.

The Schönig et al. (Reference Schönig, Von Eynatten, Tolosana-Delgado and Meinhold2021) metamorphic classes classification scheme shows that a wide variety of possible metamorphic classes are present in Siwalik samples (Fig. 9b). This scheme allocates analyses in the MS and IG categories but also includes possible sources from granulite (GR), eclogite/ultrahigh-pressure terranes (EC/UHP), blueschist/greenschist rocks (BS/GS) and amphibolite-facies rocks (AM). This scheme suggests we should anticipate finding igneous garnets in all samples and metasomatic garnets in samples SK11, SK16 and SK17. It further suggests that AM and BS/GS-sourced garnets are present in all samples. GR and EC/UHP garnets appear in all samples except SK7. In the compositional classification, garnets with intermediate-felsic/metasedimentary (IF/S) dominate in all samples, with calc-silicate garnets appearing at higher structural levels in samples SK11, SK16 and SK17 (Fig. 9c). We also see the potential for mafic garnets in samples SK11 and SK17, although at low abundance (1.8% and 9.5% respectively).

7.c. Other minerals and lithological context

Bulk mineralogical analyses were performed on select Surai Khola samples using XRD (Fig. 4c). These data provide context on whole-rock mineralogy, support general interpretations of provenance and depositional environments, and help differentiate quartz- and carbonate-rich units. The XRD mineralogical analysis reveals the presence of quartz, K-feldspar, calcite, illite/mica and illite/smectite across all analysed samples. Notably, certain mineral phases are absent in specific samples: dolomite is absent in sample SK1, chlorite is absent in sample SK17 and both plagioclase and kaolinite are absent in sample SK18. Apatite is uniquely detected in sample SK17. Based on their mineralogical compositions, samples SK1, SK2, SK3, SK4, SK7, SK8 and SK11 are classified as sandstones. In contrast, samples SK5, SK17 and SK18 are classified as calcareous sandstones due to their significantly elevated calcite contents.

Observations of the mineralogical assemblage align with findings from previous studies on Surai Khola sedimentary rocks. For example, Szulc et al. (Reference Szulc, Najman, Sinclair, Pringle, Bickle, Chapman, Garzanti, Andò, Huyghe, Mugnier, Ojha and DeCelles2006) found zircon, tourmaline, rutile, sphene, garnet and staurolite in all equivalent samples. However, kyanite and sillimanite appeared at stratigraphically higher levels in samples equivalent to SK16 and SK11. Baral et al. (Reference Baral, Lin and Chamlagain2016) reported that samples from the Lower and Middle Siwalik sedimentary rocks are dominated by quartz (91% and 85%, respectively), with feldspar, lithic fragments and phyllosilicates, including muscovite, biotite and chlorite, comprising the matrix. Lithic fragments with calcite cement included carbonate, chert, phyllite, schist and gneiss. Critelli and Ingersoll (Reference Critelli and Ingersoll1994) attributed this detritus to a recycled orogenic source, primarily derived from low- to medium-grade metamorphic rocks within the suture belt with contributions from volcanic and ophiolitic rocks. However, they reported ophiolitic material only in the Siwalik Group exposures in Pakistan, which are sourced from the palaeo-Indus drainage system and, therefore, expected to contain detritus from the suture zone. Although current petrographic and detrital zircon datasets do not support significant suture zone input in the Siwalik Group of central Nepal (Szulc et al. Reference Szulc, Najman, Sinclair, Pringle, Bickle, Chapman, Garzanti, Andò, Huyghe, Mugnier, Ojha and DeCelles2006; Bernet et al. Reference Bernet, Van Der Beek, Pik, Huyghe, Mugnier, Labrin and Szulc2006), the presence of chromium spinel and high XCa garnets in the eastern Nepal Siwalik Group has been linked to ophiolites of the Indus–Tsangpo suture zone provenance (Rai et al. Reference Rai, Yoshida and Kuritani2021).

7.d. Detrital garnet pressure–temperature conditions and paths

Figures 10 and 11 show estimated pressure–temperature conditions from Siwalik Group garnets that preserve higher Mn in their central sections and hinterland examples with similar zoning patterns. The coloured bars in the figures are isopleths of Siwalik garnet compositions ±0.2 XMn, XCa, XMg and XFe from the central section. The isopleth intersection provides our best estimates of the central section pressure–temperature conditions.

Figure 10.

Pressure–temperature diagrams showing the central section conditions and paths for samples (a) SK11-1, (b) SK11-9, (c) SK11-14 and (d) SK11-24. Coloured bars are isopleths and indicate the garnet compositions ±0.2 mole fraction of spessartine, grossular, pyrope and almandine from the central section. Where they intersect is the best estimate of the garnet central section pressure–temperature condition. Some fields are labelled with the relevant mineral assemblages, and the garnet-in reaction boundary is indicated in bold. The volume % of garnet growth is also provided in 0.5 vol %. See Table 3 for the bulk composition used to create the diagrams. Fig. 5b and d show the zoning profiles for these garnets. We include examples of conditions from garnets with similar zoning in each panel.

Figure 11.

Pressure–temperature diagrams showing the central section conditions and paths for samples (a) SK16C-3, (b) SK16B-14, (c) SK17-13 and (d) SK17-19. The coloured bars are isopleths and indicate the garnet compositions ±0.2 mole fraction of spessartine, grossular, pyrope and almandine from the central section. Where they intersect is the best estimate of the garnet central section pressure–temperature condition. Some fields are labelled with the relevant mineral assemblages, and the garnet-in reaction boundary is indicated in bold. The volume % of garnet growth is also provided in 0.5 vol %. See Table 3 for the bulk composition used to create the diagrams. Fig. 5e and f show the zoning profiles for these garnets. We include examples of conditions from garnets with similar zoning in panels (a), (b) and (d).

Table 3 summarizes the effective bulk compositions used for the samples adapted from an LHS rock (sample MA43; Catlos et al. Reference Catlos, Harrison, Kohn, Grove, Ryerson, Manning and Upreti2001, Reference Catlos, Lovera, Kelly, Ashley, Harrison and Etzel2018). The MA43 composition was adjusted for the garnet groups, except Group 6. For instance, adjustments for Group 4 garnets involved decreasing the Fe₂O₃ (−2 mol%) and CaO (−0.3mol%) and increasing the MnO (+0.1 mol%) and MgO (+2 mol%) content of the MA43 composition. Group 2 garnets are the most Ca-rich garnets, and the MA43 composition was modified by adding CaO (+1 mol%), MnO (+0.1 mol%), SiO₂ (+0.5 mol%) and MgO (+1.5 mol%) while decreasing the Fe₂O₃ (−2.8 mol%). Changes in the MA43 composition were required for the intersection of central section XFe, XMg, XCa and XMn isopleths. Without these adjustments, intersections would not occur, impeding the estimation of central section pressure–temperature conditions. Notably, this challenge was encountered with spessartine Group 3 garnet SK11-17 and Group 4 garnet SK17-10. The chosen MA43 effective bulk composition and solution models were unsuitable for garnets with higher XMn contents.

Table 3.

Bulk compositional data (mol%) used to generate the core phase diagrams

* Sample MA43 composition was taken from Catlos et al. (Reference Catlos, Lovera, Kelly, Ashley, Harrison and Etzel2018).

Table 4 outlines the estimated pressure–temperature conditions and mineral assemblages for the garnet central sections. Notably, Group 2 garnet SK11-24 exhibited overlapping isopleths at the lowest temperature (480°C), with pressure similar to that of Group 4 and Group 6 garnets (∼6 kbar). Sample SK11-24 also suggested higher CaO contents and uniquely included titanite in its mineral assemblage.

Table 4.

Summary of the central section and edge pressure–temperature conditions and mineral assemblages

* All mineral assemblages have Qz + H₂O. Mineral abbreviations after Whitney and Evans (Reference Whitney and Evans2010).

– not determined.

Group 4 garnets share the mineral assemblage of plagioclase + garnet + biotite + muscovite + ilmenite + chlorite + quartz + H₂O and similar initial thermal conditions of 520–535°C. However, modelled pressure conditions recorded by the garnets differ by 2 kbar among the samples. Sample SK11-1 had the lowest pressure of 4.6 kbar, whereas SK11-9 and SK11-14 were similar, with core conditions of ∼6 kbar (Fig. 10a–c).

Group 5 garnets and sample SK17-19 from Group 6 share the Group 4 mineral assemblage. The central section of sample SK17-13 lacks plagioclase but contains the other minerals (Table 4). Group 5 garnets share the central section conditions of 520–525°C and a lower pressure of 3.2–3.6 kbar (Fig. 11a, b). Despite a different mineral assemblage, Group 6 garnets are of similar temperature (505°C) but differ in pressure (6.8 vs. 5.9 kbar) (Fig. 11c, d).

Due to limitations in zoning preservation for some smaller samples, pressure–temperature paths were constructed for only six garnets. Most samples exhibited isothermal burial over 0.5–2 kbar, except for Group 2 garnet SK11-24, which showed an N-shaped path fluctuating over 0.5 kbar (Fig. 10d). This path type may reflect small-scale pressure fluctuations during garnet growth, possibly caused by local tectonic or erosional processes (Catlos et al. Reference Catlos, Lovera, Kelly, Ashley, Harrison and Etzel2018, Reference Catlos, Dubey and Etzel2022). The shape could indicate burial followed by rapid unloading due to erosion, followed by renewed burial or tectonic compression. Alternatively, it may reflect minor variations in the bulk composition of the garnet-bearing assemblage during growth, leading to apparent variations in the modelled path. Fluctuations are small and within the uncertainty of the modelling.

8.a. Classification and compositional trends of Siwalik Garnets

High-resolution transects across individual grains is an approach commonly used in metamorphic studies to obtain detailed insights into garnet growth histories in both metamorphic and igneous rocks (Fig. 5). Although these zoning profiles are from garnet fragments and likely record only portions of their overall compositional distributions, the consistency in zoning within groups suggests that they could be as credible or valuable as the options traditionally provided in ternary diagrams.

Our interpretation of the garnet compositions incorporates manual grouping and PCA-based clustering. While both approaches identify nine major groups and clusters, they differ in how the individual analyses are assigned. Some agreements exist between groups and clusters, as Group 3 maps exclusively to Cluster 5, and the analyses designated as Group 6* are entirely placed in Cluster 8. Group 5 mostly belongs to Cluster 0 (95.8%), with a small fraction in Cluster 3 (4.2%). However, more complex distributions occur among the strongly zoned garnets (Figs. 5 and 6). For example, Group 4 garnets are divided among Cluster 7 (77%), Cluster 0 (17.6%), Cluster 5 (4.1%) and Cluster 3 (1.4%), and Group 7 garnet spans Cluster 8 (84.2%), Cluster 3 (8.6%) and Cluster 2 (7.2%).

The relationship between PCA clusters and manual groups is also seen within single crystals. For example, zoned garnet SK17-10, characterized by a high XMn and lower XFe core, and a high XFe rim and lower XMn core, falls into our Group 4 category (Fig. 5d). However, the PCA separates the analyses into Clusters 5 and 7. Similarly, two zoned garnets, which we classify as Group 6 (SK17-13 and SK17-19), exhibit low XFe cores and higher XFe rims and would be placed in Clusters 3 and 7 by the PCA (Fig. 5f).

Both manual clustering and PCA rely on the same dataset, but PCA rigorously quantifies variance, identifying specific compositions without regard for crystal-scale compositional changes. The PCA approach treats each microprobe spot as a separate data point, which may seem appropriate for garnets lacking compositional zoning but risks interpreting zoned garnets as multiple distinct ‘sources’ rather than unified crystals with internally evolving compositions. In this way, PCA effectively treats each composition as if it were an entirely separate provenance signal.

The same issue was found with the approach of Schönig et al. (Reference Schönig, Von Eynatten, Tolosana-Delgado and Meinhold2021), which uses a random forest machine-learning algorithm for garnet host-rock discrimination (Fig. 8). By integrating this machine-learning-based method, we gained greater confidence in our provenance interpretations but observed that minor compositional variations occasionally pushed individual spot analyses into different classification categories, even in broadly ‘flat’ zoned garnets (Fig. 5). Overall, ∼ 35% of individual garnet grains contained spot analyses falling into different metamorphic settings, facies or composition categories, even though they were taken from the same crystal. Examples are shown in Fig. 5. The category shifts reflect subtle compositional heterogeneities rather than indicating multiple distinct events within a single garnet. Variability can arise from subtle zoning, borderline compositional values, or simply the thresholds used by the classification algorithm.

These findings highlight the complexity of garnet growth histories and underscore the importance of the manual classification approach, which considers an entire garnet’s compositional profile and geological context rather than treating each microprobe spot as a separate metamorphic or provenance signal. The manual approach captures patterns, allowing us to consider a garnet’s entire growth history as a single entity linked to a potential provenance. Combining manual and statistical methods preserves the nuances of garnet zoning while still benefiting from a rigorous classification.

8.b. Pressure–Temperature conditions and metamorphic history

Many Surai Khola garnet grains exhibit flat zoning profiles, suggesting that the metamorphism of Himalayan units reached thermal conditions high enough to facilitate diffusive zoning or that garnet growth occurred rapidly (Fig. 5) (Ague & Carlson, Reference Ague and Carlson2013). Since flat zoning was observed in larger garnet grains, we favour the interpretation of diffusive modification. Garnets that deviate from this trend provide valuable constraints for pressure–temperature modelling (Figs. 10 and 11). Table 5 lists Himalayan outcrop garnets used as comparisons for possible provenance for each group. Hinterland garnets were also analysed for their pressure–temperature conditions, which can be compared to the results generated for the Siwalik garnets or provide potential options for those in which conditions could not be ascertained using the modelling approach applied in the paper. In addition, the isochemical phase diagrams provide options for the mineral assemblage of the host rock from which the garnet eroded. Table 4 lists these assemblages, which can be compared to those reported for their potential source region.

Table 5.

Options for provenance for Siwalik garnet groups

* Abbreviations: Indus–Tsangpo suture zone = ITSZ; MCT = Main Central Thrust; GHC = Greater Himalayan Crystallines; NHG = North Himalayan Granite; HHL = High Himalayan Leucogranite; LHS = Lesser Himalayan Sequence.

Group 2 garnets exhibit zoning patterns where XCa + XFe + XMn > XMg (Fig. 5). The pressure–temperature conditions for one representative Group 2 SK11 garnet indicate a central section temperature of 480°C at 6.0 kbar and an edge temperature of 515°C at 6.5 kbar (Fig. 10d). The pressure–temperature path is N-shaped with minor pressure variations, which may reflect erosional exhumation rather than tectonic activity (Catlos et al. Reference Catlos, Lovera, Kelly, Ashley, Harrison and Etzel2018, Reference Catlos, Dubey and Etzel2022). Himalayan garnets with similar zoning patterns and compositions have been identified in rim analyses of garnet-bearing blueschists and granite enclaves and xenoliths within a North Himalayan granite in the Tethyan Himalayan Sequence (Honegger et al. Reference Honegger, Le Fort, Mascle and Zimmermann1989; Thakur et al. Reference Thakur, Singh, Rao, Sharma, Pandey and Ao2018). The mineral assemblage inferred from the isochemical phase diagram suggests that the original Group 2 garnet host rock contained plagioclase + garnet + biotite + muscovite + chlorite + quartz + H₂O, with titanite or rutile as possible accessory phases (Table 4). We did not incorporate solutions for epidote or spinel, which are present in the North Himalayan granite enclave assemblage. The estimated conditions for the Siwalik Group 2 garnet align with those from the enclaves, suggesting a potential contribution from North Himalayan granites (Fig. 8d). Additional constraints are needed to evaluate alternative sources, such as garnets from blueschists along the Indus Suture Zone in Ladakh, NW Himalaya, that show similarities in terms of zoning (Honegger et al. Reference Honegger, Le Fort, Mascle and Zimmermann1989).

We generated pressure–temperature conditions and paths for three Group 4 garnets from sample SK11 (Fig. 10a–c). Group 4 garnets are almandine-rich, with higher spessartine and grossular contents and low pyrope content. They are found only in samples SK7, SK11 and SK17 (Fig. 9d). These garnets exhibit prograde zoning with paths that increase pressure over small temperature changes, supporting a burial-driven metamorphic history. Himalayan garnets with similar zoning patterns to those of Group 4 occur in schists from the Bhimphedi Group (central Nepal) and gneisses of the Pangong metamorphic complex (NW India). The origin of Bhimphedi Group garnets remains debated, as they could be associated with GHC or LHS (see discussion in Webb et al. Reference Webb, Schmitt, He and Weigand2011). In contrast, the Pangong metamorphic complex in Ladakh is linked to the Karakoram Fault in the northwestern Himalaya-Karakoram belt (Streule et al. Reference Streule, Phillips, Searle, Waters and Horstwood2009). Peak conditions for garnet-bearing assemblages with zoning patterns similar to Group 4 range from 535–680°C and 8.1–8.5 kbar (Table 5). The core conditions of all Group 4 garnets overlap in temperature with the Bhimphedi Group samples, but their pressures are ∼2 kbar lower. The inferred mineral assemblage for all Group 4 garnets (plagioclase + garnet + biotite + muscovite + ilmenite + chlorite + quartz + H₂O) is broadly consistent with that of the Bhimphedi Group, increasing confidence in their potential host rock protolith.

Group 5 garnets are almandine-spessartine with low XCa + XMg. These garnets are present in all SK samples but are most dominant at lower stratigraphic levels (Fig. 9d). In the Himalaya, garnets with similar compositions are reported from magmatic rocks, including HHL, that record high temperatures (∼700°C) and low emplacement pressures (∼3.8 kbar) (LingSen et al. Reference LingSen, LingHao, LiE, KeJun and Qian2019; Xie et al. Reference Xie, Tao, Wang, Wu, Liu, Liu, Li and Zhang2020; Yan et al. Reference Yan, Yu, Wang and Ma2022). A garnet core composition within a pelitic xenolith inside a north Himalayan granite also yields a similar composition within 0.5 wt% of some of the SK16C-3 analyses and yields a higher P-T condition of 7.5 kbar and 588°C (Thakur & Patel, Reference Thakur and Patel2012). Outside the Himalaya, almandine-spessartine garnets are characteristic of evolved melts, where their distinct zoning patterns are associated with granitic differentiation and fractionation (Nabelek et al. Reference Nabelek, Russ-Nabelek and Denison1992; Diella et al. Reference Diella, Bocchio, Marinoni, Langone, Adamo and Rotiroti2018). Conditions from a Group 5 garnet in sample SK16 show that the core records the lowest pressure of all garnets analysed (3.2–3.6 kbar) at 520–528°C. A pressure–temperature path from this garnet shows isothermal burial over 1 kbar (Fig. 11d). Based on their zoning, Group 5 garnets are likely magmatic and share the low emplacement pressures seen in some HHL outcrop garnets (Table 5). The mineral assemblage suggested by the phase diagram is consistent with what is observed in HHL rocks (Pl + Grt + Bt + Ms + Ilm + Chl + Qz + H₂O) (Table 4).

Group 6 garnets are almandine-rich with low pyrope and exhibit similar XCa and XMn contents. Himalayan outcrop garnets with comparable zoning occur in a variety of lithologies, including garnet blueschists and eclogites from the Indus–Tsangpo suture zone, augen gneiss and schists of the GHC, and graphitic schists from higher structural levels of the LHS near the MCT (Table 5). Two analysed Group 6 garnets record conditions of 505°C at 5.9 and 6.8 kbar, with one exhibiting a slight pressure–temperature increase to 513°C and 6.3 kbar (Fig. 11d, e). These temperatures are comparable to other garnets with Group 6 compositional profiles, and the isochemical phase diagram suggests a felsic protolith (Table 4). A pressure–temperature path from a graphitic schist/pelite in the Kohistan-Ladakh Arc, developed using garnets with compositions similar to Group 6, is shown in Fig. 11d (Thanh et al. Reference Thanh, Sajeev, Itaya and Windley2011). This path closely parallels the Siwalik Group 6 garnet but at lower pressures. While the mineral assemblages broadly overlap, the Kohistan-Ladakh garnet likely crystallized in the presence of chloritoid and zoisite, which were not included as solution models for the Siwalik garnets.

Although rooted in several assumptions, the Siwalik garnet conditions and paths are tentative and testable frameworks for linking detrital garnets to potential source lithologies and metamorphic histories. The following section discusses provenance and tectonic implications for all Siwalik garnet groups, using these conditions to interpret sedimentary inputs, exhumation processes and regional metamorphic evolution.

8.c. Provenance and tectonic implications of Siwalik Group Garnets

Although the flat zoning and high-pressure conditions recorded in many Siwalik garnets imply derivation from deep crustal levels, this does not necessarily reflect continuous rapid exhumation of the hinterland. Instead, these garnets may record episodic, short-lived but intense phases of uplift and erosion, interspersed with more extended periods of tectonic quiescence (Thiede et al. Reference Thiede, Bookhagen, Arrowsmith, Sobel and Strecker2004; Adlakha et al. Reference Adlakha, Patel and Lal2013; Thiede & Ehlers, Reference Thiede and Ehlers2013). Such punctuated exhumation histories may better explain the transport and preservation of high-grade minerals in foreland basin deposits as high-energy transport processes (landslides, debris flows or major flooding events) likely affected their transportation into the Siwalik Group.

Garnet is prone to chemical weathering in outcrop samples (e.g., Baidya et al. Reference Baidya, Pal and Upadhyay2019). The widespread presence of garnet in the Siwalik Group is consistent with efficient sediment routing and minimized exposure to chemical weathering. High-energy mass-wasting events could transport large amounts of previously stored material from uplifted regions into the foreland basin in pulses rather than through steady erosion and exhumation. In the Surai Khola section, this interpretation is consistent with 40Ar/39Ar detrital white mica dates that support lag times of 2–8 Myr.

Siwalik garnets within similar classification groups likely share the bulk rock compositions in which they crystallized, allowing for potential correlation with hinterland rock formations that contain garnets of comparable chemistry. The consistently low XMg contents observed in most Siwalik Group garnets support a derivation from a crustal rather than a mantle source, indicating growth in Fe-, Ca- or Mn-rich and Mg-poor environments, depending on the garnet group. In granulites, lower garnet rim XMg has been linked to proximity to biotite rather than grain size, emphasizing the role of localized chemical equilibrium rather than grain-scale diffusion (O’Brien, Reference O’Brien1999). Additionally, XMg is more susceptible to diffusion along sub-grain boundaries than elements like Ca, which diffuses more slowly (Konrad-Schmolke et al., Reference Konrad-Schmolke, O’Brien and Heidelbach2007). Most Siwalik garnets that exhibit flat zoning show the trend in all components (XCa, XFe, XMn and XMg) (Fig. 5), indicating that post-growth modification affected all cations to some extent. The near-universal presence of low XMg also suggests that broader environmental controls, rather than localized conditions, likely governed their compositions.

The relatively low MgO contents of the Siwalik garnets observed by us and others (average 1.7 wt% MgO, with most <2.5 wt% MgO; our data; see also Nakajima et al. Reference Nakajima, Matsumoto, Rai and Yoshida2020; Rai et al. Reference Rai, Yoshida and Kuritani2021; Yoshida et al. Reference Yoshida, Nakajima, Matsumoto, Osaki, Rai, Cruz and Sakai2021) helped narrow the search for potential hinterland sources. Similar low-MgO garnets have also been reported in other Himalayan fluvial deposits (e.g., Jamuna River sands, Rahman et al. Reference Rahman, Pownceby and Rana2020; Indus Basin, Alizai et al. Reference Alizai, Clift and Still2016). This disparity further necessitated a broader search that extended beyond the traditionally emphasized GHC and LHS units. Unlike previous studies, our findings suggest additional provenance hypotheses that warrant further investigation.

Garnets with similar zoning as Group 1 have been reported from calc-silicate rocks from the MCT shear zone, GHC and calc-silicate granulites and veins of the CGGC in the Indian Proterozoic mobile belt (Fig. 3, Table 5). The Schönig et al. (Reference Schönig, Von Eynatten, Tolosana-Delgado and Meinhold2021) classification scheme would place the analyses almost evenly divided between MM and MS settings (58% and 42%, respectively), with most (93%) in the CS/MS category and some (7%) in the IF/S setting. Rocks that exhibit Group 1 garnet zoning experienced higher-grade metamorphic conditions (650–890°C, 5.5–12 kbar; see references in Table 5).

As noted in the previous section, conditions estimated for Siwalik Group 2 garnet align with those estimated from enclaves in North Himalayan granites (Fig. 10d). Like the Group 1 garnets, Schönig et al. (Reference Schönig, Von Eynatten, Tolosana-Delgado and Meinhold2021) classification scheme places the majority of Group 2 garnet analyses in the MM (83% setting, followed by MS at 29%). Most analyses are categorized as CS/MS (60%) followed by IF/S settings (40%). In terms of classes, both Group 1 and Group 2 garnets share metamorphic classes with AM, EC/UHP and GR classes represented. Group 2 garnets have the most analyses in the GR setting out of all the analysed garnets (55%).

Spessartine-almandine garnets like those in Group 3 are reported from a high-pressure (14 kbar at <600°C) metagreywacke located in the Indus–Tsangpo suture zone (Laskowski et al. Reference Laskowski, Kapp, Vervoort and Ding2016) and a magmatic garnet within a highly fractionated rare-metal-bearing aplite from southern Tibet (Xie et al. Reference Xie, Tao, Wang, Wu, Liu, Liu, Li and Zhang2020). An igneous origin is also supported by analyses of garnet found in the HHL (640°C, 3.5 kbar, Visonà & Lombardo, Reference Visonà and Lombardo2002) and pegmatite in the Gangdese batholith (∼634°C, Yu et al. Reference Yu, Xia, Zheng, Zhao, Chen, Chen, Luo, Li and Xu2021). The Schönig et al. (Reference Schönig, Von Eynatten, Tolosana-Delgado and Meinhold2021) classification scheme places most Group 3 garnet compositions in the IG category (79%) followed by MM (21%).

Groups 1 and 2 garnets with higher XCa contents are only present later in the Surai Khola section, possibly reflecting a shift in sediment sources due to transient changes in erosion patterns or sediment routing. If sourced from the GHC or MCT shear zone (Table 5), the presence of these garnets at upper stratigraphic levels of the Siwalik Group aligns with rapid erosion rates and shorter lag times during the Pliocene (van der Beek et al. Reference Van Der Beek, Robert, Mugnier, Bernet, Huyghe and Labrin2006). A five-fold increase in apparent erosion between 2.5 and 0.9 Ma has been documented in MCT shear zone rocks (Huntington et al. Reference Huntington, Blythe and Hodges2006). Such rapid erosion is consistent with the exhumation of calc-silicate rocks from the MCT zone, contributing detrital garnets to the foreland basin. However, lag times may be as long as ∼8 Myr (SK16) based on 40Ar/39Ar detrital white mica. Based on garnets with similar zoning, Group 1 garnets may originate from mobile belt sources (Table 5). Their presence raises intriguing questions about sediment transport mechanisms and tectonic processes that can deliver this material into the Himalayan foreland basin. Potential transport pathways could include major river systems draining the Indian Proterozoic mobile belt or tectonic processes facilitating the reworking of mobile belt sediments into the basin. Paleohydrological indicators consistent with axial rivers suggest this option should be considered (Burbank et al. Reference Burbank, Beck, Mulder, Yin and Harrison1996; Ulak, Reference Ulak2005; Mandal et al. Reference Mandal, Sarkar, Chakraborty, Bose, Liu and Wang2014).

The core conditions of all almandine Group 4 garnets overlap with the thermal conditions of the Bhimphedi Group samples, which are located closer to the Surai Khola section. Group 4 garnets are only found in samples SK17, SK11 and SK7. The Schönig et al. (Reference Schönig, Von Eynatten, Tolosana-Delgado and Meinhold2021) classification scheme places all Group 4 garnets in the MM setting class. The group has 96% of all analyses in the BS/GS metamorphic class (96%), followed by 4% in the AM class. Most analyses were in the IF/S composition class (99%), with only 1% in the M class.

We have the most confidence in the origin of Group 5 almandine-spessartine garnets as having a magmatic origin, likely crystallized in rocks similar in composition to the HHL. The Schönig et al. (Reference Schönig, Von Eynatten, Tolosana-Delgado and Meinhold2021) classification scheme anticipates that 76% of Group 5 garnets would fall within the IG setting, followed by 25% in the M class. All were in the IF/S compositional category. Group 5 garnet compositions are consistent with those formed in highly fractionated, felsic magmas. Today, the HHL are located at the highest elevations of the Himalayan range (Figs. 1 and 2). These granites are thought to have crystallized over a prolonged period between 7 and 46 Ma reflecting a dynamic interplay of crustal melting, tectonic exhumation, and episodic magmatism associated with the ongoing Indo–Asia collision (see review in Wu et al. Reference Wu, Liu, Liu, Wang, Xie, Wang, Ji, Yang, Liu, Khanal and He2020). Group 5 garnets are present in all Siwalik samples (Fig. 8d), highlighting their widespread distribution and persistence within the foreland basin sediments. These garnets, now buried as deep as 4 km in the Siwalik strata (e.g., SK7 depth, 9.5 Ma, Fig. 9d), provide evidence of significant erosion and sediment transport from the Himalayan crystalline core. Their preservation across the Siwalik stratigraphy suggests that detritus derived from HHL exposures contributed consistently to sedimentary deposition over time.

Group 6 garnets are almandine with low pyrope contents but have XCa and XMn at roughly similar levels. Himalayan outcrop samples with similar zoning as those seen in Group 6 are found in a variety of lithologies, from garnet blueschists and eclogites of the Indus–Tsangpo suture zone, augen gneiss and schists within the GHC and graphitic schist collected from higher structural levels of the LHS, near the MCT (Table 5).

Group 6* garnets with similar XCa and higher XMg as those in Group 6 are exclusively found in various lithologies associated with the GHC (Table 5). These garnets show a wide range of metamorphic conditions. Group 6 garnets are present in all samples except SK8, but only one Group 6* garnet was found in sample SK17. The Schönig et al. (Reference Schönig, Von Eynatten, Tolosana-Delgado and Meinhold2021) classification scheme places the majority of Group 6 and 6* garnets into the MM class (90% and 100%), except a few Group 6 analyses in the IG class (10%). Most analyses are IF/S, with a few exceptions in Group 6 that are in the M compositional setting (11%).

We classified almost 1/3 of all samples analysed in this study in the Group 7 category, which are almandine-rich and have lower XCa + XMg + XMn. Group 7* is similar to those in Group 7 but has slightly higher XMg contents. Group 7 garnets are found in all samples, and Group 7* garnets are found in all samples except SK11 (Fig. 9d). Like Group 6 garnets, outcrop Group 7 garnets are found in a range of lithologies but are dominated by LHS and GHC rocks. Outcrop garnets in the Group 7* category are found in the LHS and GHC (Table 5). The Schönig et al. (Reference Schönig, Von Eynatten, Tolosana-Delgado and Meinhold2021) classification scheme would place all Group 7 and 7* garnets into the IF/S composition class and MM settings, except 2% in the IG category for Group 7.

The Schönig et al. (Reference Schönig, Von Eynatten, Tolosana-Delgado and Meinhold2021) metamorphic category identifies garnets with BS/GS-sourced garnets as present in all samples, and an EC/UHP origin is present in all samples except SK7. The EC/UHP classified garnets are the same as four of those we placed in Group 1 (SK17_5, SK17_9, SK16B_scattered and SK16_24) and two in Group 2 (SK11_15 and SK24), and one in Group 7* (SK8_9).

We examined the comparison of Surai Khola garnet compositions with those from stratigraphic sections in Nepal and India. Figure 12 shows analyses reported for Lower Siwalik units along the Karnali River in western Nepal and Tinau Khola in central Nepal, the Muksar Khola in eastern Nepal (Nakajima et al. Reference Nakajima, Matsumoto, Rai and Yoshida2020; Rai et al. Reference Rai, Yoshida and Kuritani2021) and pre-Siwalik sedimentary rocks from the Kasauli Formation in NW India (Najman & Garzanti, Reference Najman and Garzanti2000). Most garnets reported from these sections fall into the Group 7 and 7* categories, except for one garnet resembling Group 2 from the Middle Siwalik of the Muksar Khola section and four analyses from the Tinau Khola, Muksar Khola and Kasauli Formation that resemble Group 6. One garnet from the Lower Siwalik units along the Karnali River exhibits a very high Ca content (35.6 wt% CaO), which was not observed in our studied section (Fig. 12a) but is characteristic of garnets from calc-silicate rocks in the Greater Himalayan Crystallines (GHC) or the CGGC (Neogi et al. Reference Neogi, Dasgupta and Fukuoka1998; Dey et al. Reference Dey, Choudhury, Mukherjee, Sanyal and Sengupta2019). Using the Schönig et al. (Reference Schönig, Von Eynatten, Tolosana-Delgado and Meinhold2021) classification scheme, garnet would be classified as having a granulite origin. This scheme also suggests that most garnet compositions reported by others are metamorphic in origin with minor igneous input. Fig. 12f shows the presence of GR and EC/UHP garnets in both the Middle and Lower Siwalik sedimentary rocks along the Muksar and Karnali Khola transects. However, the figure likely does not reflect any critical heterogeneity that may be present. Our analysis is based solely on selected garnet compositions reported in published datasets and available data repositories.

Figure 12.

Garnet compositions from Siwalik Group samples are shown for (a) the Karnali River section (Lower Siwalik, 15.8–9.6 Ma), (b) Tinau Khola (Lower Siwalik, 13.2–9.2 Ma) and (c) Muksar Khola, including Upper (<3.5), Middle (10.0–3.5 Ma) and Lower (>10 Ma) Siwalik intervals (Nakajima et al. Reference Nakajima, Matsumoto, Rai and Yoshida2020; Rai & Yoshida, Reference Rai and Yoshida2020; Rai et al. Reference Rai, Yoshida and Kuritani2021). Panel (d) presents garnet compositions from the Kasauli Formation, a pre-Siwalik (Oligocene–Miocene) sedimentary unit (Najman & Garzanti, Reference Najman and Garzanti2000). Garnet group classifications are noted above each analysis. See Fig. 1 for sample locations. Panel (e) shows garnet classification by tectonic setting, (f) by metamorphic class and (g) by compositional fields after Schönig et al. (Reference Schönig, Von Eynatten, Tolosana-Delgado and Meinhold2021).

Baral et al. (Reference Baral, Lin and Chamlagain2016) identified the Tethyan Himalayan Sequence and the upper LHS as primary sources for the Surai Khola section, with minor input from GHC zircon. However, 40Ar/39Ar white mica dates indicate a significant contribution from the GHC (Szulc et al. Reference Szulc, Najman, Sinclair, Pringle, Bickle, Chapman, Garzanti, Andò, Huyghe, Mugnier, Ojha and DeCelles2006). The oldest sample in our study is 9.5 Ma, placing it well within the timeframe when GHC exhumation was likely underway. The presence of Groups 4–7 garnets supports the interpretation that GHC-derived garnets were actively supplied to the basin by 9.5 Ma, consistent with ongoing hinterland exhumation. The presence of Group 6 and 7 garnets, consistent with LHS sources, also supports previous findings that LHS material contributed significantly to Siwalik deposition in this region (Baral et al. Reference Baral, Lin and Chamlagain2016). While our study does not provide direct exhumation rate estimates, our findings are consistent with a model of progressive and rapid exhumation of the GHC exhumation (Szulc et al. Reference Szulc, Najman, Sinclair, Pringle, Bickle, Chapman, Garzanti, Andò, Huyghe, Mugnier, Ojha and DeCelles2006; van der Beek et al. Reference Van Der Beek, Robert, Mugnier, Bernet, Huyghe and Labrin2006).

Tourmaline in the Siwalik Group has been linked to HHL provenance in central Nepal (Rai, Reference Rai2003), while geochemical and petrological studies of the Middle and Upper Siwalik sedimentary rocks in NW India indicate increased contributions from plutonic rocks and granitoids (Ranjan & Banerjee, Reference Ranjan and Banerjee2009). The presence of sediment sourced from leucogranites in the Siwalik Group and Bengal Fan has been discounted due to the absence of zircon dates linked to the HHL (Bernet et al. Reference Bernet, Van Der Beek, Pik, Huyghe, Mugnier, Labrin and Szulc2006; Blum et al. Reference Blum, Rogers, Gleason, Najman, Cruz and Fox2018). However, many HHL zircon dates contain inherited cores that may be mistaken for GHC, THS or LHS affinity (Fig. 8d) (Schärer et al. Reference Schärer, Xu and Allègre1986; Fan et al. Reference Fan, Zhang, Lin, Wang and Zhang2021; Liu et al. Reference Liu, Zhu, Wang, Cawood, Stockli, Stockli, Lin, Zhang, Zhang and Zhao2022). These inherited cores from older magmatic episodes may obscure the recognition of HHL contributions in detrital zircon datasets. The rims that typically record Cenozoic crystallization could be selectively removed through abrasion or omitted in provenance studies due to analytical approaches that target core analysis. The HHL represents a significant Himalayan lithotectonic component, but its exposures account for only ∼2% of the belt (Scaillet et al. Reference Scaillet, Holtz, Pichavant and Schmidt1996; Searle, Reference Searle1999; Searle et al. Reference Searle, Simpson, Law, Parrish and Waters2003). Group 5 garnet compositions can only be linked to a magmatic environment, underscoring the critical role of the HHL as a sediment source for the Siwalik Group.

8.d. Limitations and cautions

Detrital garnets can survive recycling (Morton & Hallsworth, Reference Morton and Hallsworth2007; Baldwin et al. Reference Baldwin, Schönig, Gonzalez, Davies and Von Eynatten2021). However, we might anticipate that garnets more susceptible to weathering and erosion would disappear from the section, and recycled grains would appear smaller. For example, grossular garnets are considered less stable than their almandine counterparts (Morton & Hallsworth, Reference Morton and Hallsworth2007; Tolosana-Delgado et al. Reference Tolosana-Delgado, Von Eynatten, Krippner and Meinhold2018), and in this study, we only identified grossular compositions in samples from higher stratigraphic levels (Fig. 9d). While low sampling density (ca. n = 20 grains per sample) may have contributed to the absence of grossular-rich garnet, similar trends have been observed in other studies (Yoshida et al. Reference Yoshida, Matsumoto and Sakai2015; Nakajima et al. Reference Nakajima, Matsumoto, Rai and Yoshida2020; Yoshida et al. Reference Yoshida, Nakajima, Matsumoto, Osaki, Rai, Cruz and Sakai2021), suggesting that this pattern is consistent and may not be solely an artefact of sample size (Fig. 12).

We correlate Siwalik garnet compositions to those from potentially analogous bedrock samples (Table 5), which requires understanding all possible sources of Himalayan garnet. This approach has been applied elsewhere, including the central Swiss Alps (Stutenbecker et al. Reference Stutenbecker, Berger and Schlunegger2017). We are limited by what is reported in the literature and recognize that other options may exist. We are conservative in our interpretations, including identifying multiple possibilities for each group. Our interpretations are similar to those using ternary diagrams for garnet provenance, which have long been used to relate to lithologies, pressure-temperature conditions and even specific rock units. In addition, it is evidence-based, as we seek to identify options for Siwalik garnet sources based on those found in Himalayan outcrops with similar zoning patterns despite unavoidable limitations. Of the garnet groups identified, we are most confident that Group 5 almandine-spessartine garnets are from an igneous provenance, as these compositions are commonly reported from HHL assemblages in the Himalaya and elsewhere.

We acknowledge that previous studies have highlighted the challenges in generating pressure–temperature conditions and paths from detrital garnets, with some even describing it as ‘impossible’ (Baldwin et al. Reference Baldwin, Schönig, Gonzalez, Davies and Von Eynatten2021). Some researchers discard detrital garnet analyses if core and rim compositions differ by more than 5 mol% (Nakajima et al. Reference Nakajima, Matsumoto, Rai and Yoshida2020). These grains, however, are precisely those we identified as necessary for thermobarometric analyses. We assume a pelitic lithology, which is reasonable for the Himalayan hinterland, and use specific solution models and effective bulk rock compositions to identify where garnet isopleths intersect. Central section isopleth intersections are also unique solutions. We made what may appear as minor changes in the MA43 bulk rock composition (Table 3), and isopleth intersections only result if these specific changes are made within the context of equilibrium conditions, specific solution models, assumed mineralogy, and garnet composition. Pressure–temperature conditions and paths for central sections of Siwalik detrital garnets were approximated through phase equilibria modelling based on the assumption that a garnet of a specific composition existed within an effective bulk rock composition and underwent metamorphism under closed-system equilibrium conditions.

The fact that garnets from the same groups from which we can generate pressure–temperature results yield similar conditions lends some confidence that the groupings are reasonably meaningful. We acknowledge that only the central regions of the detrital garnets were analysed, and the outer rims, which may have recorded peak metamorphic conditions, are likely absent. As such, the final pressure–temperature estimates from modelling should not be taken as representative of peak metamorphism. In addition, the central sections should not be considered garnet cores. In the pressure–temperature diagrams, intersection regions are slightly off the garnet-in isograd, suggesting that we are not measuring the true garnet core but our best approximation.