Sustained wood burial in the Bengal Fan over the last 19 My

Wood Burial in the Bengal Fan

Turbidites dominate the sedimentary accumulation of the Bengal Fan and millimeter- to centimeter-sized wood debris was picked from the coarse-grained layers of turbidites beds, especially subunits Tb and Td of the Bouma sequence (14) during visual description of the split cores onboard the R/V Joides Resolution (Fig. 1). Wood was mainly found as 1) isolated large fragments within thick sandy deposits, 2) small fragments dispersed throughout the coarse basal units of turbidite layers and, 3) (sub)horizontal layers of fragments concentrated at the transition between coarse and fine sediments (i.e., the subunit Td of the Bouma sequence). The wood fragments varied in color and size but were most commonly gray or black in color, in thin sheets (likely as a result of compaction and dewatering) or small amorphous fragments. Bigger pieces (up to 3 cm × 2 cm) were mostly brown to black, but some layers yielded abundant and apparently unaltered (pale-colored) wood fragments.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Five examples of wood deposits in IODP Expedition 354 cores from sites U1451 and U1455. Buried wood sizes vary from millimeter- to centimeter-scale and are denoted by black fragments on core images (IODP LIMS Reports). Samples with readily visible wood pieces such as in (A) U1451B-23R-1 (47 to 49 cm, red box), (B) U1455C-44R-3 (85 to 87 cm), and (D) U1451A-56F-2 (4 to 7 cm) were picked directly from the face of the split cores. Other layers with visible specks, thought to be wood, were sieved to isolate fragments, including samples of (C) U1451B-19R-3 (36 to 40 cm) and (E) U1451B-58R-1 (50 to 56 cm). All rulers show centimeters.

Our survey of woody horizons spanned 19 My of sediments (Fig. 2), and in total 60 coarse horizons yielded visible wood (up to ∼6 g) that was picked from the core surface (Fig. 2A). In order to further survey for cryptic occurrences of wood, we sieved (>1 mm) 73 additional coarse horizons across the 19 My. We recovered wood debris from 37 out of 73 sieved samples, with up to 4.21 g of wood or 8.64 weight % (hereafter %) of sediment (Fig. 2A). In total, we recovered up to ∼6 g wood per 1 My (Fig. 2B), with 97 wood samples spanning the last 19 My (Fig. 2C). We found wood at each of the drill sites that we studied (U1450, U1451, U1452, U1453, U1454, and U1455). The age–depth distribution of samples reveals the episodic deposition of woody horizons and the extent of our core sampling coverage (Fig. 2D), including overlapping sediment recovery, that attempted to account for spatial variations in deposits as the channel–levee complexes migrate. However, we emphasize neither the coring, nor our sampling of the cores, allows for comprehensive accounting of all wood buried at this location. A critical obstacle is that coarse sediments have the lowest sediment recoveries (SI Appendix, Fig. S2), and thus wood-bearing horizons have been disproportionately missed in drilling. In addition, thousands of coarse horizons were recovered—many of which contain wood debris based on visual inspection of core section photographs—and we did not sample all of them. Nevertheless, we captured evidence of wood deposits throughout the last 19 My.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Semiquantitative accounting of wood abundance. (A) Wood mass normalized to sediment mass including picked (solid) and sieved samples (open symbols). (B) Wood mass summed in 1-My bins (dark blue). One high-abundance picked sample at 0.05 Ma, highlighted in light purple in A and B, is estimated >5 g. (C) Count of wood-yielding samples (cyan) and total samples (white). (D) The age–depth distribution of wood fragments compared to the sampled span of cores to the right. Age estimates for sample depths are provided in Dataset S1 with explanation in Materials and Methods. Note core recovery is discontinuous and sedimentation rates vary (SI Appendix, Fig. S2). For wood mass per sample by core site see SI Appendix, Fig. S3.

All 97 wood-bearing turbidites were sampled in triplicate to determine the C concentration of wood fragments (in weight %, hereafter %). We found wood C concentration ranging from 3.89 to 81.43% (mean = 33.49%, 1σ =14.02%, n = 257; Dataset S3). The mean is close to that expected for living wood, typically 40 to 55% C (15). Low C concentration (<40%) may result from degradation processes or dilution by silicate mineral matrix embedded within wood fragments. Three samples with 80% C likely represent charred wood. We find no systematic trend in wood C content over the 19-My record, suggesting limited degradation of wood debris following burial.

Normalizing for the mass of sediment, we found that wood concentration was 0.31 ± 1.13% (1σ, n = 70), and wood C concentration was 0.15 ± 0.43% (1σ, n = 34) for the sieved samples (Dataset S3). For the picked samples, the sediment mass relating to core surface sampling was not well constrained. We hence assign an average mass of sediment (30 g) in order to estimate concentrations of wood, at 0.94 ± 3.27% (1σ, n = 60) and wood C as 0.17 ± 0.62% (1σ, n = 59) of sediment mass. This assumption suggests that sieved samples had on average lower wood mass recovery than the picked samples, which is reasonable as the picked samples targeted large fragments and/or high abundance horizons, with visibly apparent wood on the split core surface. Across all of the studied coarse horizons, we find that wood mass is 0.60 ± 2.39% (n = 130) of sediment mass and wood C is 0.16 ± 0.55% (n = 93) of sediment mass. These OC concentrations are higher than that of the coarse load of the modern G-B Rivers (typically ca. 0.05%), where OC is almost entirely petrogenic (16). This observation requires the addition of wood fragments, above the OC concentrations observed over the last few decades in the G-B Rivers, and this likely occurs during extreme sediment export events in rivers. The addition of wood to the coarse sediment load is significant with respect to the carbon cycle. Indeed, wood photosynthetic C represents a net carbon sequestration mechanism, whereas the background petrogenic C present in coarse river sediments is only being transferred from terrestrial to marine long-term storage (with potential for loss in transit). As such, wood-laden coarse sediment transport and deposition becomes an effective carbon sequestration mechanism.

Although large wood debris has been analyzed in rivers (9, 10, 17⇓–19), there are only a few studies of wood export to the ocean (11, 20) and even fewer investigating wood preservation and burial in marine turbidites (6, 7, 12). Intense physical erosion events such as landslides may export wood to the oceans in small mountainous rivers. For example, Typhoon Morakot triggered landslides in Taiwan in August 2009 delivering ∼3.8 to 8.4 Tg of coarse woody debris to the oceans, representing 30 to 60% of the river-exported terrestrial biospheric carbon (21). The resulting turbidites (7) further transported the wood offshore where it was buried with up to 70% preservation efficiency (6), which is attributed in part to the short distances in small mountain rivers (22). In the very large G-B catchment (1.6 × 10⁶ km²), extreme erosion events and associated sediment transport have been observed in the high-relief regions of the modern system and/or documented in the geological record, during earthquake, landslide, and glacial-dam outburst floods (23⇓⇓⇓–27). Although some events originate above the tree line, land sliding also occurs within forested zones and high-flow conditions entrain coarse woody debris downstream. In such high-discharge events, wood erosion may follow the “pulse–shunt” concept proposed for dissolved OC (28), with a pulse of input and a shunt of increased velocity, leading to efficient carbon export. However, we are aware of no prior studies that have tried to quantify wood export fluxes to the oceans from large mountainous river catchments. Infrequent, high-flow export events are challenging to capture both in modern river systems and in ancient fan deposits where the locus of deposition shifts over time.

Our sampling of the transect of IODP Expedition 354 cores, focused on locating visible wood fragments within the coarse turbidites to establish the abundance and characteristics of such deposits. Given the length of the G-B Rivers and the >2,000-km distance from the river mouth to deposition, the presence of wood in deep sea sediments of the mid Bengal Fan throughout the last 19 My demonstrates wood export and burial efficiency. Our discovery of wood distributed throughout the 19-My record indicates the need to account for buried wood in carbon budgets by surveying carbon throughout turbidites in order to quantify the contribution of wood to the long-term (10⁷-y) sequestration of atmospheric CO₂. However, accurate accounting of such fluxes is challenging as the locus of deposition changes as the canyon–levee systems migrate across the fan (29) and low recovery of coarse-grained sediments during drilling (SI Appendix, Fig. S2) may lead to underestimates of wood burial.

Galy et al. (16) documented a negative relationship between grain size and OC content in the G-B sediments. Using this relationship, Acoustic Doppler Current Profiler to achieve depth integrations of sediment composition and flux, and ¹⁴C-based estimates of petrogenic C content, they estimated the average biospheric OC concentration of suspended sediments exported by the Lower Meghna River (the confluence of the Ganges and Brahmaputra Rivers) to be 0.30% (30). We estimate that coarse-grained sediments deposited in the Bengal Fan contain half as much biospheric OC as the average G-B River suspended sediment and thus wood should be considered in global C burial budgets and petroleum source rock characterization. However, further partitioning of the relative importance of OC burial flux in fine and coarse sediments is not possible for several reasons. First, in river studies, coarse-grained fluvial export is less well studied than the fine particulate load. Second, in marine cores, the low recovery of sands compared to high recovery of clay-rich sediments, biases attempts at overall quantification. Third, IODP Expedition 354 drill sites sit along a 300-km transect orthogonal to the main direction of sediment delivery in efforts to capture the migrating channels supplying turbiditic sediments to the fan (13). However, we do not yet have a view of transport of wood from the subaerial delta to the fan’s terminal lobes (across 3 × 10⁶ km²), as would be needed to obtain a comprehensive budget of the overall burial of woody debris. Nevertheless, the results of IODP Expedition 354 show that coarse sediments represent a large proportion of the overall sediment deposition in the distal Bengal Fan (13). Combined with our observations of frequent wood debris preservation in coarse sediments, we demonstrate that the contribution of wood debris to the overall burial of terrestrial biospheric OC is significant.

Wood Composition Reveals Sourcing

Next, we studied the chemical structure and isotopic composition of buried wood in order to understand its ecological and geographic origin (Fig. 3). Lignin, a large heterogeneous aromatic polymer and the second-most-abundant biochemical in vascular plants, carries information about wood type and degradation in its phenolic structures (31). A subset of wood horizons (n = 15) were analyzed for lignin phenol composition across the 19-My record, with multiple analyses (n = 9) possible from one abundant horizon at 0.05 Ma (Materials and Methods). The relative abundance of syringyl (S) versus vanillyl (V) phenols indicated that almost all wood fragments were derived from angiosperms (S/V > 0.02) across the 19-My record (31, 32) (Fig. 3A). As angiosperms dominate below 2,500 m elevation in the G-B catchment, and conifers dominate above 3,000 m (33), phenolic composition constrains the elevation source to below 2,500 m. Predominantly lowland provenance has previously been documented for fine particulates in the modern G-B River system for bulk OC (16) and vascular plant biomarkers (34). Similarly, a persistent lowland provenance was inferred for OC associated with fine-grained sediments deposited in the Bengal Fan over the last 20 ky (35). Only in the Pleistocene, at ∼0.05 Ma, did we find gymnosperm wood (S/V = 0.0), for 6 of 9 aliquots of a single horizon (Fig. 3A and Dataset S2), representing a distal, high-Himalayan origin. This may reflect a rare mass wasting event such as may occur with earthquakes, floods, or glacier retreat in high-relief mountain areas. For example glacial-lake outburst failures (23, 24) have been shown to move high sediment loads and large boulders tens of kilometers downstream. However, no modern observations have shown scenarios in which Himalayan coarse OC can be exported rapidly to the Bay of Bengal, as is required to explain the deposit we have documented at 0.05 Ma.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Isotopic composition and phenol ratio of wood fragments in the Bengal Fan. (A) Syringyl/vanillyl, where S/V > 0.2 (gray circle) are angiosperms and arrow indicates degradation. One horizon at 0.05 Ma contains the only gymnosperms (black circle, 6 overlying datapoints). (B) Bulk wood δ¹³C and (C) lignin methoxy δD. Samples with phenol measurements (gray); C₄ plants (blue circle). Isotopic data differentiated by IODP site in SI Appendix, Fig. S3).

In addition to visual observations of compaction and color change in the fragments of wood, lignin phenol analysis indicates progressive degradation in wood fragments from 6 to 18 Ma (Miocene). With increasing age, we found declining concentrations of S, V and p-hydroxyl (P) phenols and increasing acid-to-aldehyde ratios for the S and V groups (Sd/Sl and Vd/Vl, respectively; SI Appendix, Fig. S4), indicative of degradation (36). As S phenols are preferentially lost compared to V phenols during wood degradation, the directional trend in S/V over the Miocene is ascribed to progressive degradation (Fig. 3). However, lignin degradation is not extensive as both Sd/Sl (mean = 0.32, 1σ = 0.15, n = 23) and Vd/Vl (mean = 0.37, 1σ = 0.15, n = 23) fall largely within the range (<0.4) expected for fresh wood (15).

Stable carbon isotopic composition was measured to identify variations in source wood as may result from altitude, humidity, and use of the C₃ versus C₄ photosynthetic pathway. Bulk wood δ¹³C (δ¹³C_wood) values range from −30.4 to −11.4‰ in 97 turbidites (each horizon sampled in triplicate, yielding 257 individual measurements spanning the last 19 My; Fig. 3B). We find no correlation between wood C content and δ¹³C_wood (SI Appendix, Fig. S5 and SI Extended Supplementary Data Discussion), which implies that degradation and/or matrix dilution does not systematically affect δ¹³C_wood. From 18.4 to 5.8 Ma, we find that δ¹³C_wood values in discrete turbiditic horizons range between −29.9 and −23.8‰ (mean = −26.6‰, 1σ = 1.2‰, n = 187). Aliquots within a single horizon have an average variability of 0.7‰ (1σ) that represents heterogeneity within a single coarse-grained horizon of the turbiditic deposit, which could readily derive from variations within a localized plant community (e.g., species differences or microhabitat). The 6.1‰ range between turbidites suggests deposits retain the character of input events (floods, cyclones, etc.) with varied sourcing within the large catchment. This further suggests minimal mixing during fluvial and submarine transport to the edge of the shelf and during resuspension in turbiditic transport to the midfan core location. The apparent stationarity of the range of δ¹³C_wood values from 18.4 to 5.8 Ma implies no systematic change in climate, ecosystem, or sourcing across this 12-My period of the Miocene (Fig. 3B and SI Appendix, SI Extended Supplementary Data Discussion). After a gap in wood recovery from 5.8 to 4.42 Ma, a systematic shift begins at 4.42 Ma, with a δ¹³C_wood measurement of −20.5‰, which is 3.3‰ more positive than the most ¹³C-enriched sample during the Miocene, which may suggest a mixture of C₃ and C₄ fragments within the aliquot. Even more ¹³C-enriched values appear in the last 1 My, where δ¹³C_wood values are bimodal with a C₄-like cluster (mean = −13.1‰, σ = 1.0‰, n = 17) as well as the typical C₃-like values (mean = −26.3‰, σ = 1.7‰, n = 51). This suggests a change in the ecosystems from which the wood is being exported (e.g., plant type and climate), with one-third of the buried wood derived from C₄ plants in the last 1 My.

As bulk tree wood is infrequently reported in modern ecosystems, we report δ¹³C_wood from a survey of the woody tissues of trees, bamboo-like grass, and ferns collected along 2 elevation transects in Arunachal Pradesh (India) and Nepal within the G-B catchment in 2004 to 2007 (Dataset S4) and find that modern wood δ¹³C values range from −23.1 to −33.2‰. We account for the year of sampling to remove the Suess effect and determine preindustrial equivalent values (δ¹³C_pre-I) relative to an atmospheric δ¹³C_atm value of −6.4‰ (37). δ¹³C_pre-I range from −21.4 to −31.4‰ with an average of −26.8‰, similar to buried wood samples from 18.4 to 5.8 Ma. We find no systematic difference between plant type (gymnosperm, angiosperm tree or shrub, bamboo, or fern) but we do observe an overall tendency to ¹³C-enrichment with elevation of +0.8‰ km⁻¹ (r² = 0.22, P < 0.001, n = 53; SI Appendix, Fig. S6) within C₃ plants. Much more enriched values are associated with the use of the C₄ pathway: One tall grass collected at 1,500 m yielded a δ¹³C value of −12.7‰; another was collected in the lowlands, from the bank of the Ganges, where C₄ are dominant (−12.4‰; preindustrial equivalent −10.6‰).

C₄ Expansion in the G-B Catchment and Wood Sourcing

In the late Miocene (after ∼6.5 Ma) there was an ecosystem transformation associated with a C₄ grassland expansion detected in the Siwalik Formation in the Himalayan foreland (38, 39) and in bulk OC and biomarkers in the fine sediments of the Bengal Fan (40, 41). This region was the center of radiation in the C₄ Andropogoneae subfamily (42). Today the catchment includes many tall C₄ grasses (Miscanthus, Saccharum, Hyparrhenia, and Heteropogon), with stems that include woody tissues that may be strong enough to survive fluvial transport. Although expansion of the C₄ pathway is most commonly associated with grasslands (monocots), the C₄ pathway also appeared in some dicotyledenous plants (dicots), mostly shrubs that occupy ephemeral streams and saline habitats (43, 44) and Euphorbiaceae trees (43, 45). Only in glacial periods with lowered pCO₂ and drier conditions in the region may C₄ dicots have a competitive advantage over other plants (44). Pollen from the Siwalik Formation indicate the Miocene appearance of xeric and halophilic shrubs (Amaranthaceae and Chenopodaceae) (39). Those taxa are also represented along with Euphorbiaceae pollen in the last 1 My in the distal Bengal Fan, south of Sri Lanka (46). However, the C₄ pathway is most prevalent within tropical grasses in the catchment today, and tall grasses are therefore also assumed to be the most likely source of C₄ “wood” delivered in the last 5 My. Whether grass, shrub, or tree, C₄ plants generally grow below elevations of 1,500 m and thus add to the evidence for lowland sourcing of woody debris found in the Bengal Fan.

After the C₄ expansion we find signs of reduced burial of woody material in the Bengal Fan, based on barren horizons between 5.8 and 4.42 Ma (Fig. 2A). The grasslands that transformed ecosystems across the floodplains may have altered fluvial geomorphology and erosive processes (e.g., via changing rooting structures and plant lifecycle). Alternatively, this gap may reflect a shift in fan sedimentation processes as channel–levee deposition increases around this time (47).

During glacial cycles of the last 1 My, wood deposits have a bimodal δ¹³C distribution likely reflecting wet, closed lowland forest and C₄ grass/xeric taxa at interglacial and glacial extremes, respectively (46, 48). Enhanced physical erosion likely explains the observed increase in wood export in the last 1 My, and this included export of C₄ with some aliquots yielding δ¹³C values characteristic of pure C₄ plants (mean −13.0‰, σ = 1.0‰, n = 17) between 0.4 and 0.1 Ma. We hypothesize that these fragments were mainly deposited during glacial periods when aridity and low pCO₂ favored C₄ plants in lowland part of the G-B catchment (35). Supporting evidence comes from Pleistocene pollen in Deep Sea Drilling Project site 717 which were dominated by Gramineae (grass) pollen (presumably C₄), with enhanced transport of Pinus pollen from Himalayan sources in cold intervals (46). The buried wood record from the Bengal Fan indicates that C₄ woody debris is at times very abundant in the Pleistocene and provides catchment-scale evidence for the timing and magnitude of their expansion. Very low δ¹³C_wood values are also observed during the Pleistocene, suggesting wood transport and burial was also effective for wet, closed forest wood, perhaps growing during interglacial periods. The Pleistocene stands out as an interval of extreme variability in wood composition that reflects cyclical climate variability and the resulting changes in vegetation and erosion as conditions change, including one instance of high-elevation sourcing, in contrast to dominantly lowland sourcing throughout the long record.

Constraints on the Isotopic Composition of Precipitation

We additionally sought constraints on isotopes in precipitation from the buried wood samples. The methoxy groups on lignin have the advantage of nonexchangeable hydrogen (unlike bulk wood, cellulose, or lignin) such that their hydrogen isotopic composition (δD_methoxy) records the isotopic composition of precipitation (δD_precip). A global survey of modern trees reported the fractionation (²ε_{methoxy/precip}) between δD_precip and δD_methoxy to be −216 ± 17‰ (1σ, n = 222) (49, 50). We measured buried wood samples from the Bengal Fan spanning 19 My, finding δD_methoxy values range from −300 to −150‰ (mean = −253‰, 1σ = 30.5‰, n = 59; Fig. 3C). Applying the global fractionation, we estimated δD_precip values from the buried wood samples to range from −108 to +68‰ (mean = −48‰, σ = 39‰, n = 59; Dataset S6).

Modern δD_precip in the catchment ranges from −81 to −28‰ (amount-weighted mean annual precipitation from 25 sites all situated below 2,000 m; SI Appendix, Fig. S6 and Dataset S5). Most buried wood δD_precip estimates overlap with the range of modern precipitation range in the catchment and this suggests broadly comparable monsoonal precipitation patterns during the last 19 My. A few samples are slightly more D-depleted and suggest a high elevation source (>2,000 m) or a period of intensified monsoonal rainfall (51, 52) compared to the modern data. The power of dual-isotope analyses to diagnose altitude of sourcing has previously been demonstrated for plant wax biomarkers in another large, very wet tropical montane catchment, where both ¹³C enrichment and D depletion are expected with increasing elevation (53). Here we extend that concept to buried wood measuring δ¹³C_wood and δD_methoxy. Accordingly, samples with δD_methoxy values below −280‰ and δ¹³C_wood > −26.0‰ (Fig. 4) potentially derived from high altitude. However, overall the wood compositional distribution is not dominated by the expected altitude trends in each isotope system (SI Appendix, Figs. S6 and S7), indicating other factors contribute to the variability, in particular C₄ inputs, compositional variability within the lowland G-B catchment, and uncertain fractionation factors.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Comparison of δD_methoxy and δ¹³C_wood for paired analyses on individual horizons (open symbols). The 0.05-Ma horizon (black symbols) was parsed into 9 aliquots of which 6 appear to be upland-derived based on dual isotopes (consistent with lignin evidence that they are gymnosperm) and 3 lowland-derived (consistent with their angiosperm identification); see Discussion. Annotations show expected altitudinal pattern within C₃ plants, ¹³C-enrichment with mixed and pure C₄ horizons (blue symbols, n = 3 horizons), and anomalously D-enriched values (right of dashed line).

Three horizons have elevated δD_methoxy (>200‰), and while we calculate δD_precip of +42 to +68‰ for these horizons, such enriched values are unlikely from meteoric water or evaporated surface waters. Instead, we suspect smaller fractionation factors may be appropriate for those plants, and we note that the buried wood dataset includes C₄ plants which have not yet been calibrated for ²ε_{methoxy/precip}. Combining information from carbon and hydrogen isotope systems can also help to diagnose any such complications associated with C₄ inputs (Fig. 4). C₄ plant-derived wood samples are clearly differentiated by carbon isotopes and overall carry more D-enriched values typical of more lowland sources, although some yield values that are anomalously high. However, several C₃ plants are unexpectedly D-enriched as well, suggesting a need for local calibration of δD_methoxy among C₃ as well as C₄ plants.

One abundant deposit, at 0.05 Ma, included large fragments that allowed for multiple aliquots for multiple analyses representing individual trees. In that sample, 6 fragments yielded lignin phenolic evidence for gymnosperms (conifers), thus derived from elevations above 2,500 m. Dual isotopes confirm that interpretation (Fig. 4). In those gymnosperm fragments we measured δD_methoxy (mean = −291‰, 1σ = 14‰, n = 6) yielding δD_precip estimates of −95‰, independently corroborating a high-elevation provenance. In the same samples δ¹³C values (mean = −25.9‰, 1σ = 0.2‰, n = 6) also support a high-elevation source. In contrast, 3 angiosperm fragments from this same horizon yield high δD_precip estimates (mean = +3‰, σ = 6‰, n = 3) and δ¹³C values (mean = −27.8‰, 1σ = 1.7‰, n = 3) indicative of lowland sourcing (Fig. 4), suggesting mixing in transit. As discussed previously, there are several possible mechanisms to explain high-elevation sourcing including earthquake-triggered landslides or glacial damming, either of which can trigger lake outburst flooding with sediment loads and velocities well above normal monsoon flood conditions. Such an event could mobilize a large amount of woody debris from a high-elevation “point source” that can be transported rapidly to lower elevations, entraining wood from the lowlands, and perhaps triggering hyperpycnal flow facilitating rapid delivery to the deep sea. Although such events are reported in the Himalaya, and described offshore of small mountain rivers, a Himalaya-to-Bengal Fan transport event has not been observed in the modern system that is heavily altered with damming, channelization, and agriculture.