To clarify the DOC distributions in ground ice, two sites presenting different landscapes (Figure 2) were designed at the SCG basin to investigate cryostratigraphy and obtain ground ice. Before excavation, the vegetation cover, topography, and hydrological conditions were investigated and photographically recorded (Figure 2). Two profiles were excavated manually using a shovel and an electric pick. The P-1 profile was excavated on top of a palsa-like frozen mound (Figure 2A; Figure 2B) with a height of ∼2 m. Moreover, a thermal-erosion gully located 15 km apart from P-1 was chosen, and a profile (P-2) was dug from an exposed slump crack (Figure 2C; Figure 2D). Two profiles were excavated to a depth of 4.8 m. The lithology, cryostructures, and ice conditions were documented and photographed before sampling. The frozen sediments and ice layers were cut using a chain saw and a chisel (Figure 2B; Figure 2D). The superficial layer of each sample was discarded to avoid any contamination. The profiles were cut in 5-cm intervals for P-1 (every 10 cm below 3 m) and 3-cm intervals for P-2.62, and 125 samples were collected in P-1 and P-2, respectively. In addition, the soil samples from the active layer and frozen layers were retrieved synchronously. All the samples were numbered in terms of depth, preserved in HDPE bottles, and kept frozen at –4°C in the field. In addition, five sites in the SCG basin were selected to collect the active layer water. Five pits were dug to a depth of 1.0 m, and the laterally and vertically permeated soil water were obtained in September (2014–2015) during our fieldwork.

The frozen samples were thawed entirely at 4°C, left until sediment settled, and then the liquid water above was filtered with pre-combusted GF/F filters and acidified (HCL, pH < 2) to prevent microbial conversion and to remove the carbonates. The DOC concentrations were measured with the solid and liquid modules of OI Analytical Analyzer (OI-Picarro, CA, United States), with a standard deviation of less than 0.5‰. The δ¹³C_DOC values were analyzed with a Picarro Isotope Analyzer (Picarro G1102). Stable isotope results were expressed as δ values relative to the Vienna Peedee belemnite (VPDB) standard. The precision of δ¹³C is <0.5‰, and the detection limit of DOC is 0.3 mg/L.

Total organic carbon (TOC) in sediments was measured using an OI Analytical model 1030 TOC analyzer (OI Analytical, United States). Each dry and homogenized sample was put into a small quartz boat after weighing. The 1 M HCl was added to the sample for 12 h to remove the carbonates. Then, the samples were heated to 900°C in the combustion chamber, and the organic matter in the samples was oxidized to CO₂ to measure the carbon contents.

In order to clarify the potential origins of DOC in ground ice, the published data from different water components over the TP were collected and synthesized. Specifically, thermokarst lake water (TLW), glacial meltwater (GMW), small stream water (SSW), large river water (LRW), large lake water (LLW), precipitation water (PW), snowpit water (SPW), and groundwater (GW) at different sites were comprehensively investigated. The DOC and δ¹³C (not available in GMW and GW) data were reanalyzed for comparison (Table 1; Figure 5).

According to the field investigations, fine sandy soil is predominant in the upper layer and fine sandy loam dominates in the frozen layer in P-1 (Figure 3). Substantial soil-bearing ice layers are identified with volumetric ice contents ranging from 40% to 90% (Figure 3). The concentration of DOC ranges between 5.6 mg/L and 24.6 mg/L, with a mean of 9.7 mg/L. The peak DOC (24.6 mg/L) presents at 3.0 m (Figure 3). The δ¹³C_DOC varies from −27.4‰ to −36.1‰. Three higher DOC peaks are present at 0.9 m (13.8 mg/L), 1.8 m (15.8 mg/L), and 3.0 m (24.6 mg/L), corresponding to the relatively enriched δ¹³C_DOC (pink bars in Figure 3). In contrast, the most negative δ¹³C_DOC peaks appear at 1.6 m (−36.1‰), 3.1 m (−33.3‰), 3.4 m (−34.6‰), and 4.5 m (−34.3‰), which are consistent with the lower DOC contents (purple bars in Figure 3). However, higher DOC contents appear at 3.1 m, corresponding to the most negative δ¹³C_DOC (green bar in Figure 3). In addition, the total organic carbon of soil (TOC_soil) in different depths was determined for comparison (Figure 3), the range of which is between 0.65% and 8.14%, with a mean value of 3.31%. The upper layer (0–1.6 m) contains much higher (mean level of 5.25%) and fluctuating TOC_soil than those in the deeper layers (Figure 3), suggesting the important influence of freeze–thaw cycle and evolution of permafrost/ground ice.

For comparison, fine sandy loam layers dominated in the full profile for P-2 (Figure 4). Cryostratigraphy in this profile consists of ice-saturated permafrost (0.78–1.6 m) and soil-bearing ice layers (below 1.6 m), with volumetric ice contents ranging between 50% and 90%. The ground ice in P-2 presents higher DOC compared with P-1 (Figure 4). The DOC ranges between 8.8 and 70.1 mg/L, with a median of 21.5 mg/L. Moreover, the δ¹³C_DOC ranges from −28.1‰ to −36.9‰, which is similar to that in P-1. There are three positive DOC peaks, located at 1.8 m, 2.2 m, and 4.1 m, respectively. Furthermore, δ¹³C_DOC also shows some negative peaks (Figure 4). As for the ground ice in P-1, the DOC and δ¹³C_DOC in P-2 show coincident trends, with higher DOC corresponding to the relatively enriched δ¹³C_DOC and vice versa. However, similarly opposite relations between DOC and δ¹³C_DOC are identified at approximately 3.0 m, at which the lowest DOC concentrations correspond to the most enriched (median: −28.8‰) value of δ¹³C_DOC. Generally, the contents of TOC_soil in this profile exhibit remarkable variations with depth (Figure 4), ranging between 0.23% and 6.74%. Higher TOC_soil values (mean: 4.11%) appeared in the upper layer (0–1.7 m), which was similar to that in P-1, reflecting repeated freeze–thaw cycles and frequent exchange with external environment, resulting in carbon decomposition by microorganism. The frozen soil in the deep layer presents lower TOC_soil (mean: 1.95%).

In order to determine the possible carbon sources of DOC in ground ice, the DOC data of different water components (including precipitation, snowpit, glacial meltwater, lakes, rivers, and groundwater) on the TP were collected for comparison (Table 1).

Remarkably, the ground ice in both profiles exhibits higher DOC concentrations (mean: 9.7–21.5 mg/L) than those in large river water (LRW) (mean: 1.2–2.3 mg/L), LLW (mean: 2.8 mg/L), snowpit water (SPW) (mean: 0.6 mg/L), precipitation water (PW) (mean: 1.0 mg/L), groundwater (GW) in seasonal frozen ground regions (mean: 1.3 mg/L), and glacial meltwater (GMW) (mean: 1.2 mg/L; Table 1 and Figure 5). In comparison, the DOC of active layer water (ALW) (mean: 6.4 mg/L), small stream water (SSW) (mean: 9.5 mg/L), thermokarst lake water (TLW) (mean: 10.5 mg/L), and groundwater (GW) in permafrost regions (mean: 11.9 mg/L) is higher and comparable to that of ground ice (Figure 5), indicating connections between them and similar carbon behaviors with regard to the influence of permafrost.

As shown in Figure 6, the highest contents of DOC (from 9.5 to 347.0 mg/L; mean: 58.7 mg/L) appear in the non-massive intrasedimental ice (NMI) in the arctic (Tanski et al., 2016; Figure 6 and Table 2). However, the pore ice (PI) in Alaska (7.3 mg/L; Abbott et al., 2014) exhibited the lowest DOC concentration. Although the higher ice contents and huge volumes are found in ice wedges (IW) in the Arctic and Alaska, the DOC concentrations are not quite high as expected (Figure 6). By comparison, the layered ice and lenticular ice (LI) in the SAYR exhibit higher DOC concentrations, which are consistent with those in permafrost meltwater (PMW) and ALW. Specifically, the DOC concentrations in P-1 are consistent with the ice wedges, massive ice, and thermokarst exposed ice in the Arctic regions (Vonk et al., 2013a; Fritz et al., 2015; Tanski et al., 2016). However, the much higher concentrations of DOC in P-2 are close to those found in a localized thermal erosion ice wedge in Alaska (mean: 28.8 mg/L; Douglas et al., 2011).

The carbon sources and accumulation processes of DOC in ground ice exert essential roles in the magnitude and bioavailability of DOC (Fritz et al., 2015). Generally, lower DOC concentrations in these types of ice were probably due to limited carbon inputs (Raymond and Bauer, 2000; Spencer et al., 2015), lower ice contents, and rapid in situ freezing (French and Shur, 2010). By contrast, the higher DOC concentrations of these ices are assumed to strongly interact with the higher ice contents and lacustrine origin of host sediments (Tanski et al., 2016). Importantly, the stable carbon isotopes (δ¹³C_DOC) for the initial source water of ground ice vary substantially. It was potentially due to various carbon input, local vegetation conditions, freeze fractionation (Cristea et al., 2014), ice formation patterns (Tanski et al., 2016), and different decomposition processes of organic carbon (Alewell et al., 2011). Accordingly, the δ¹³C_DOC in ground ice can also be used to infer the carbon sources of ground ice.

As shown in Figure 5, the DOC concentrations in ground ice are similar to those of ALW, and the ¹³C values are alike, indicating that the DOC in ground ice was closely related to the dissolving of organic matter from the active layer (Yu et al., 2017). The previous study also reported a similar DOC concentration of the soil solution in the upper active layer (26.3–39.5 mg/L) on the Northeastern Tibetan Plateau (NETP) (Luo et al., 2009) and substantial input (as high as 70%–94%) of OC from the active layer leachate to the permafrost streams (Wang et al., 2018). These kinds of DOC were usually originated from modern terrestrial biomass and surface organic layers (Guo et al., 2007). Due to the high hydraulic conductivity, low mineral content, and low DOC sorption capacity of the active layer soil (Striegl et al., 2005), the younger and biodegradable DOC (Drake et al., 2015) could migrate quickly to the permafrost table with limited microbial transformation. The DOC was easily sequestrated into the ground ice by freezing.

More depleted δ¹³C_DOC values of ground ice in the SAYR (Figure 7; Table 1) are different from those of other water components on the TP. For comparison, the δ¹³C_DOC values of ground ice (–31.6‰ to –29.8‰) are similar to ALW (–30.2‰), suggesting similar decomposition mechanisms and carbon sources. However, the δ¹³C_DOC values of ground ice are much lower than the SSW and TLW (Figure 7), although melting ground ice was an important source of water for SSW and TLW (Yang et al., 2016; 2019). It suggests differences in carbon sources and decomposition processes under changing environments and distinct freezing conditions, which greatly influenced the stable isotopes of DOC (Krüger et al., 2014).

Notably, the mean values of measured δ¹³C_DOC in the ground ice lie within the range of plants on the TP (Li et al., 2007) and are similar (slightly depleted) to the values measured in δ¹³C for moist soils on the northwestern TP (Mu et al., 2014). This suggests that all of the plants performed C₃ photosynthetic pathways with δ¹³C values ranging from approximately −32‰ to −20‰ (Boutton, 1991; Farquhar et al., 1989; Raymond and Bauer., 2001). In addition, the consistent negative δ¹³C_DOC values of ground ice and ALW (Figure 7) also suggest the leaching of organic matter from active layer soil. During water migration through the sediments of the active layer, the organic carbon of sediments and roots of plants melted out and subsequently dissolved into the supra-permafrost water, which sequestrated into ice during cold seasons (Guo and Macdonald, 2006; Guo et al., 2007; Tanski et al., 2016).

We thus conclude that the substantial leaching of DOC from organic layers and surrounding permafrost sediments are the important carbon sources of ground ice in the SAYR. Both freezing–thawing processes and permafrost evolution would influence the embedding and sequestrating of DOC into ice. In addition, the microorganisms heterogeneously decomposed the organic carbon in permafrost and plants.

The DOC and δ¹³C_DOC composition show remarkable variations with depths (Figure 3; Figure 4), indicating different permafrost evolution processes and the resultant drivers of DOC incorporating into ice (Fritz et al., 2015), and the decomposition processes influenced by climate transition (Krüger et al., 2014). The depths with enriched δ¹³C_DOC may reflect warm and wet climate conditions; in contrast, the lower δ¹³C_DOC values reflect dry and cold conditions (Cristea et al., 2014). The higher DOC suggests quick freezing, and the lower DOC points towards the slow segregation, which excluded the solute in the water during ice formation.

As shown in Figure 8, the DOC in ground ice exhibits changing trends along depths, indicating different sequestration mechanisms of DOC during ice formation processes. As one of the most effective mechanisms for long-term carbon fixation in permafrost, freezing complicated the sequestrate process of DOC into ground ice. Meanwhile, the active layer properties, vegetation characteristics, and permafrost aggregation rates are also critical factors for DOC sequestration (Fritz et al., 2015).

Generally, the DOC of ground ice in the upper layers (1.6 m) presents gradually increasing DOC in both profiles, suggesting control of the freeze–thaw cycle. Below 1.6 m, the DOC and δ¹³C_DOC exhibit fluctuating trends, suggesting alternating controls of permafrost evolution and the resultant ice formation mechanism under changing climate conditions (French and Pollar, 1986; Ewing et al., 2015a).

Specifically, the low DOC at P-1 and higher DOC at P-2 at 1.3 m (Figure 8) correspond to different ice formation mechanisms. The ice segregation process at P-1 and quick in situ freezing processes at P-2 are proposed. However, the second highest DOC peaks and enriched δ¹³C_DOC at both P-1 and P-2 at approximately 1.8–2.0 m (Figure 8) are inferred to be affected by the warmer and wetter climate conditions (Cristea et al., 2014). The subsequent cold events at 1.9 ka (∼2.2 m; Wang et al., 2018) promoted the upward aggradation of permafrost and quickly locked the DOC into ground ice. The much higher deposition rate at 2.0 m (Wang et al., 2018) also confirms the quick freezing processes.

At the depth of 3.1 m, the ground ice shows positive DOC peaks in both profiles, which follow depleted δ¹³C_DOC values (Figure 8). As reported in Luo et al. (2018), the mean active layer thickness (ALT) in the SAYR is 2.8 m, and the repeated freeze–thaw cycles and microorganism activities in the active layer could significantly alter the DOC (Yu et al., 2017; Fuss et al., 2016). Also, the diverse carbon input and water availability from the external environment can influence the carbon isotopes in general (Cristea et al., 2014).

Between 3.5 and 4.3 m, the identical δ¹³C_DOC trends in both profiles (Figure 8) suggest similar sequestration processes and climate conditions (Porter et al., 2019). However, a high density of DOC at P-2 and a lower one at P-1 were observed, suggesting distinct ice formation mechanisms and organic carbon behaviors (i.e., decomposition, accumulation, and mixing). The P-1 is a palsa-like frozen mound; it was formed by an aggradation process (Krüger et al., 2014). The slow segregation processes at P-1 excluded the DOC from solid ice due to self-purification (Cheng, 1983). In contrast, P-2 exhibits an open-system feature, which could receive external water and produce a waterlogged and anaerobic environment. Thus, it reveals substantial accumulation of DOC in the ground ice, suggesting quick-freezing and efficiently locking of the DOC into the ice.

As shown, the DOC concentrations of ground ice, TLW, SSW, and GW in permafrost regions are much higher than those of LLW and LRW in seasonal frozen ground regions (Figure 9), which may be significantly related to the degradation of permafrost with high-content ground ice. For comparison, the ice contents in the seasonally frozen ground are usually lower (Zhou et al., 2000; Zou et al., 2017) and are commonly surrounded by thick unfrozen sediments, so the influence of thawing permafrost is thus negligible. Previous studies on the TP suggested that the contribution of melting ground ice to thermokarst lakes were as high as 37.75%–51.75% (Yang et al., 2016), and the melting ground ice similarly provided 13.2%–16.7% of total discharges to the streams (Yang et al., 2019). It has been suggested that permafrost thaw has released substantial melt water, bringing massive labile DOC into streams (Gao et al., 2019; Song et al., 2019) and thermokarst lakes (Mu et al., 2016) on the TP.

During recent decades, the SAYR has been undergoing remarkable permafrost degradation due to continuous temperature rise (Jin et al., 2009). A gradual increase in ALT at a rate of 2.2 cm/a in the SAYR was determined (Luo et al., 2014). Note that these rates of permafrost degradation in the SAYR are sizably higher than those in the interior TP (Jin et al., 2009), which resulted in massive melting of ground ice. On the one hand, the excess ground ice would release a large amount of meltwater to recharge the surface runoff and lakes (Lee et al., 2014) and simultaneously release a considerable DOC flux into the freshwater system (Vonk et al., 2015), which can be used by plant roots (Becker et al., 2016; Figure 9). On the other hand, the melting of ground ice increased the soil pore (Walvoord and Kurylyk, 2016) and allowed much more supra-permafrost water, precipitation, and surface water to infiltrate into the deep permafrost (Yang et al., 2017). These waters could penetrate deeply through taliks to connect with the sub-permafrost water, and further leached labile carbon from the ambient frozen sediments (Figure 9). During this process, the DOC in shallow layers could enter the sub-permafrost water via taliks. These activities substantially altered the DOC dynamics in permafrost (ground ice) and aquatic systems (Figure 9).

Importantly, continuous permafrost would degrade into discontinuous permafrost and finally change into seasonally frozen ground. The GW in permafrost regions presents higher DOC levels than that in seasonally frozen ground regions (Figure 5), which is closely related to the occurrence of ground ice. Accordingly, the transition from permafrost to seasonal frozen ground will greatly influence the thermal regimes of large lakes and large rivers distributing in seasonal frozen ground regions, ultimately leading to a great increase in DOC of LLW and LRW (Figure 9). Because the DOC from ground ice is chemically labile (Vonk et al., 2013a), it can be decomposed easily by microbial and photochemical processes after entering into lakes and rivers (Battin et al., 2008; Vonk et al., 2013a) and finally released to the atmosphere (Figure 9).