The SAYR (32.2°–36.1°N, 95.9°–103.4°E; 2,629–6,235 m a. s. l.) is the up-reach catchment of the Yellow River basin above the Tanag hydrological station (35.5˚N, 100.2˚E; 2,629 m a. s. l.), on the northeastern QTP, China (Figure 1), covering an area of 1.212 × 10⁵ km². The climate in the SAYR is characterized by long winters (November–April) and short summers (June–August). Mean annual air temperature (MAAT) and annual precipitation in the SAYR are 0.2°C and 507.7 mm (1960–2012), respectively (Lan et al., 2013). The climate in the up-reach SAYR above the Huanghe’yan hydrological station (34.9°N, 98.2°E; 4,215 m a. s. l.) is characterized by a dry–cold climate with a MAAT less than −3.5°C and annual precipitation of 300–400 mm (Jin et al., 2010). The growing season varies with elevation but mainly concentrates from May to September. With an areal extent of permafrost at 34% in the SAYR, permafrost occurs extensively (86%) in the up-reach SAYR, with rich manifestations in periglacial landforms and MAAT (1961–2019) varying from −3.5 to about 0°C. Alpine permafrost can be as thick as up to about 360 m on very high mountain tops on the peak of Mt. Maqên Gangri (6,235 m a. s. l.) in the Anye Maqên Mountains (Wang and French, 1995; Jin et al., 2009, 2020; Luo et al., 2020; Sheng et al., 2020). The lower reaches/basins of the SAYR between Huanghe’yan and Tanag hydrological stations are dominated by a relatively warm–wet climate, particularly in the southeastern SAYR (mainly on the Zoîgé Peat Plateau), where MAAT is at about 7.5°C and annual precipitation, about 654 mm (Yu et al., 2017). The snow-melt season generally extends from April to October in the lower reaches/basins of the SAYR. Glaciers are minimally presented in the SAYR (mainly in the Anye Maqên Mountains) (126.7 km²), or 2% of the SAYR in the areal extent (Blue et al., 2013; Jin et al., 2019).

The SAYR is characterized by a high-mountain surrounded by flat valleys and plains. The baseflow accounts for a large portion of annual streamflow (∼50% or more) (Ma et al., 2019a). Numerous meandering rivers in the valleys on the Zoîgé Peat Plateau, which are conducive to water conservancy, in addition to rich precipitation, have facilitated the accumulation of thick peat layers (0.5∼10.5 m) in the Holocene (mainly since 11.7 ka B.P.), particularly in the Megathermal period in the mid-Holocene at 8.5–4.0 ka B. P. (Yu et al., 2017; Jin et al., 2019, 2020). Alpine steppes and alpine meadows dominate alpine vegetation in the up-reach SAYR and the lower reaches/basins of SAYR; The SAYR is sporadically dotted with shrubs and forests at lower elevations. In alpine wetlands on the Zoîgé Peat Plateau, aquatic vegetation (e.g., Kobresia pygmaea and K. tibetica) is extensively observed (Brierley et al., 2016).

DOC concentrations in aquatic systems show strong variations among individual water types in the SAYR (Figure 2A). The highest DOC concentration was found in thermokarst lakes with an average of 38.4 ± 48.5 mg C·L⁻¹ (n = 171 samples; data are expressed as average ± one standard deviation) and a maximum of 234.4 mg C·L⁻¹. Permafrost-related water types showed similarly high DOC concentrations (ground-ice meltwater: 32.94 mg C·L⁻¹, supra-permafrost water: 20.54 ± 3.6 mg C·L⁻¹, and sub-permafrost water: 16.8 ± 1.2 mg C·L⁻¹) (Figure 2B). Thermokarst lakes/ponds (21.3 ± 34.1 mg C·L⁻¹) were characterized by almost five times higher average DOC concentrations than those of rivers/streams (4.3 ± 3.7 mg C·L⁻¹) and with a wide DOC range of ∼0.5–234.4 mg C·L⁻¹. Other water types (hail/snow/rain, springs/icings, and groundwater) had low average DOC concentrations between 0.8 ± 1.0 and 1.8 ± 1.4 mg C·L⁻¹, and their ranges were much narrower than those lakes/ponds and rivers/streams (Table 2). There are marked seasonal variations in in-stream DOC concentrations in the SAYR. The SAYR-average of DOC concentrations in winter is generally lower than those in spring (snow-melt season), summer, and autumn. In 2016, the greatest in-stream DOC concentration (4.5 ± 2.2 mg C·L⁻¹) occurred during the spring snow-melt season, and decreased into summer–autumn (2.4 ± 3.7 mg C·L⁻¹), and winter (0.8 ± 0.6 mg C·L⁻¹) (Figure 2C). In 2017, DOC concentrations were the greatest in summer–autumn (7.0 ± 4.8 mg C·L⁻¹), with a range of 0.5–13.4 mg C·L⁻¹, followed by those of the snow-melt season (3.6 ± 6.0 mg C·L⁻¹) and the winter season (1.1 ± 0.5 mg C·L⁻¹) (Figure 2D).

In the SAYR, in-stream DOC concentrations are averaged at 4.2 ± 3.7 mg C·L⁻¹, with a range of 0.7–21.0 mg C·L⁻¹, comparable to DOC ranges in northern permafrost regions (1–32 mg C·L⁻¹) and some other catchments on the QTP (0.4–30.8 mg C·L⁻¹). In the SAYR, in-stream DOC concentrations were characterized by the greatest values (4.5 ± 2.2 mg C·L⁻¹) during the spring snow-melt season, and decreased in summer–autumn (2.4 ± 3.7 mg C·L⁻¹) and declined further in winter (0.8 ± 0.6 mg C·L⁻¹) in 2016 (Figure 2C). In 2017, DOC concentrations were the greatest in summer–autumn (7.0 ± 4.8 mg C·L⁻¹), followed by those in the snow-melt season (3.6 ± 6.0 mg C·L⁻¹) and winter season (1.1 ± 0.5 mg C·L⁻¹) (Figure 2D). In northern permafrost regions, such as the upper Yukon River Basin, seasonal DOC concentrations are characterized by the greatest average in-stream DOC (36.12 mg C·L⁻¹) in the snow-melt season and decreased to that in summer–autumn (7.3 mg C·L⁻¹) (Guo and Macdonald, 2006; Guo et al., 2018). It is generally proven that in-stream DOC concentrations are positively related with discharge. In the upper Yukon River, the yearly peak flow occurs in spring freshet and the discharge declines in the following seasons (Guo and Macdonald, 2006; Guo et al., 2018).

The Yellow River flow in the SAYR is dominated by the monsoonal climate and mainly recharged by summer precipitation from July to September. In 2017, the summer peak flow induced high in-stream DOC concentrations from July to September. However, in 2016, the SAYR underwent a dryer summer, and no peak flow was observed from July to September. Therefore, accumulated snowfall in winter merely induced a small spring freshet and relatively high snow-melt DOC concentrations. In winter, in-stream DOC concentrations in the SAYR were smaller than those in the summer–autumn and snow-melt season in both 2016 and 2017. They are consistent with those existing observations in northern permafrost regions. However, winter in-stream DOC concentrations in the SAYR (0.9 ± 0.6 mg C·L⁻¹), comparable with those of the upper Yarlung Zangbo River (0.5 ± 0.2 mg C·L⁻¹) on the southern Tibet Plateau, are smaller than those in northern permafrost regions (e.g., Yukon: 3.9 ± 0.5 mg C·L⁻¹). Lacustrine DOC concentrations (0.5–234.4 mg C·L⁻¹) in the SAYR are generally lower than those in northern permafrost regions (0–1,130 mg C·L⁻¹), while comparable to those on the QTP (21.3 ± 34.1 mg C·L⁻¹). DOC concentrations in the supra-permafrost water (20.5 ± 3.6 mg C·L⁻¹) are higher than those in the sub-permafrost water (16.8 ± 1.2 mg C·L⁻¹) (n = 4; north bank of the Wanlong Worma Lake in the Wanlong Worma River Basin of the Duo Qu River in the south-central SAYR). Along the Alaska Arctic coast, DOC concentrations of supra-permafrost water are reported as 33 ± 2 mg C·L⁻¹ (Connolly et al., 2020), which is higher than those observed in the SAYR. DOC concentrations in ground-ice melt-water in the up-reach SAYR are much lower than those in northern permafrost regions, which may be resulted from the limited dataset. DOC concentrations in springs/icings waters, precipitation, and groundwater are generally low (∼1 mg C·L⁻¹), with DOC in groundwater a bit higher than the other two water types.

The site-specific data of environmental elements (PREP, NDVI, SOC, and PP) were extracted and the principle environmental controllers for aquatic DOC concentrations were examined by employing the standard PCA method. Data were binned into in-stream water samples collected in the snow-melt season, summer/autumn and winter, groundwater samples, lacustrine samples, and altogether binning of various aquatic samples for the PCAs (Table 3; Figure 3). In different water types, the sources and formations of DOC may differ, such as seasonally hydrological processes bringing in different DOC sources into streams/rivers. By binning environmental variables and DOC concentrations in different water types, the key environmental controllers can be determined for each water type.

PC1, PC2, and PC3 together explained 81% of the variance among all environmental factors in the snow-melt season (Table 3). PC1 was moderately correlated with PREP (r = 0.61) and PP (r = 0.59), PC2 was moderately correlated with DOC (r = 0.63) and SOC (r = −0.69), and PC3 was strongly correlated with SOC (r = 0.71) and moderately with NDVI (r = 0.52) (Table 3). Due to the fact that we took accumulated solid precipitation in winter as precipitation in the snow-melt season, the positive correlation between PP and PREP illustrates higher accumulated snow in winter in permafrost regions than that in areas of seasonal frost or talik. In-stream DOC concentrations were negatively correlated with the SOC content, obviously inconsistent with those previous observations (Prokushkin et al., 2009; Kindler et al., 2011; Kicklighter et al., 2013). When the ground surface starts to thaw in the snow-melt season, SOC and litterfalls dominate the in-stream DOC (Prokushkin et al., 2010; Guo et al., 2018), possibly due to representativeness of sampling locations. Three components (PC1, PC2, and PC3) together (Table 3) explained 77% variance of all environmental factors in summer–autumn. PC1 was moderately correlated with PP (r = 0.65) and negatively correlated with NDVI (r = −0.57); PC2 was strongly correlated with DOC (r = 0.71); and PC3 was strongly correlated with PREP (r = 0.80) and moderately with SOC (r = 0.55). PP was negatively correlated with NDVI, possibly due to warmer climate and higher NDVI in the downstream SAYR compared to alpine steppes and meadows in upstream SAYR extensively underlain by alpine permafrost. PC1, PC2, and PC3 explained 84% of the variance of environmental factors in winter (Table 3). PC1 was moderately correlated with SOC (r = 0.57). PC2 was closely correlated with PREP (r = 0.77) and moderately with PP (r = 0.51). PC3 was closely correlated with DOC (r = 0.70) and moderately with NDVI (r = −0.65). The negative correlation of winter DOC with NDVI demonstrates relatively higher NDVI in downstream SAYR characterized by lower winter in-stream DOC concentrations.

Two principal components explained 75% of the variance in environmental factors in groundwater sampling sites (Table 3). PC1 was moderately correlated with DOC (r = 0.64) and negatively with PP (r = 0.56). PC2 was moderately correlated with PREP (r = −0.66) and NDVI (r = 0.55). The correlation of groundwater DOC with PP proves that groundwater has become an increasing source of DOC upon permafrost thaw (Moore, 2010). The negative correlation of PREP with NDVI disagrees with the environmental characteristics in the SAYR, likely due to specific sampling locations. PC1, PC2, and PC3 together explained 78% variations of environmental factors in lacustrine sampling sites (Table 3). PC1 was moderately correlated with PREP (r = 0.63), PC2 was moderately correlated with SOC (r = 0.60) and NDVI (r = 0.57), and PC3 was negatively correlated with SOC (r = -0.68) and moderately with NDVI (r = 0.50) and DOC (r = 0.50). Correlations between SOC and NDVI illustrate that a higher NDVI can accumulate more SOC. The correlation of NDVI and lacustrine DOC indicates that sampling plots with higher NDVI are likely to have higher DOC concentrations in lakes/ponds, similar to that reported in earlier studies (Schindler et al., 1997). However, the fact that they were negatively correlated with SOC contradicted those previous observations (De March et al., 1975; Schindler et al., 1997). This might have resulted from samples mostly collected in the up-reach SAYR and lower reaches of SAYR around the Zoîgé wetlands, where SOC was negatively correlated with PP and lacustrine DOC. PC1 explained 38% of the variation of all aquatic samples, with important contributions of PP (r = 0.52), PREP (r = 0.59), and NDVI (r = −0.52) (Table 3). The PC2 explained 22% of the variations, correlated with SOC (r = 0.82) and DOC (r = −0.52) (Table 3). SOC appeared to be grouped with PP, whereas NDVI was on the opposite side of the plot (Figure 3), implying the positive and negative correlations between SOC and NDVI, respectively, with PP. DOC and PREP were grouped together in the down-right quadrant of the plot, indicating the positive correlation between DOC and PREP among all aquatic types. Groundwater and summer streams/rivers were clustered on the left side of the plot, while other water types showed no clear separation.

Aquatic organic carbon is also an important atmospheric carbon source. A meta-analysis of results for decomposability of DOC in varied aquatic systems (e.g., rivers/streams, groundwater, lakes/ponds, and soil solutions) and seasonal rivers/streams was performed for identifying the potentials of transformation from aquatic DOC into atmospheric carbon. Combined with results of DOC concentrations of varied water types and seasonal rivers/streams in northern and other permafrost regions through the method of literature synthesis, hotspots of atmospheric carbon emission from aquatic systems were preliminarily explored (Figure 4).

The ranges of in-stream DOC concentrations are similar in the snow-melt season (2.4–33.6 mg C·L⁻¹) and summer–autumn (6–44.0 mg C·L⁻¹) across 10 sampling sites, in the Yukon River Basin and northeastern Siberia (Finlay et al., 2006, 2009; Guo and Macdonald, 2006; Petrone et al., 2006; Olefeldt and Roulet, 2014) (Figure 4A). In the SAYR, ranges of in-stream DOC concentrations in the snow-melt season (0.1–25.4 mg C·L⁻¹) and summer/autumn (0.5–21.0 mg C·L⁻¹) fell in the existing ranges. In-stream biodegradable DOC (BDOC) in the snow-melt season is 20–40% (Abbott et al., 2014), yet in-stream BDOC in summer–autumn is much smaller ranging from 1 to 10% (Wickland et al., 2012) (Figure 4A). In-stream DOC concentrations in winter range from 1 to 20 mg C·L⁻¹, relatively lower than those in both the snow-melt seasons and summer–autumn (O’Donnell et al., 2012) (Figure 4A). Winter in-stream DOC concentrations in the SAYR ranged from 0.5 to 1.5 mg C·L⁻¹ were also smaller than those in other seasons. In the winter season, in-stream BDOC reported across the Yukon River Basin ranges from 10 to 53%, higher than in-stream BDOC in other seasons (O'Donnell et al., 2012; Wickland et al., 2012) (Figure 4A). Thus, in-stream BDOC in the spring snow-melt season and winter with relatively higher DOC concentrations or BDOC are potentially important sources of atmospheric carbon emissions from aquatic systems. DOC concentrations in different water types have shown featured distributions. The ranges of DOC concentrations are the lowest in groundwater (0.5–3.9 mg C·L⁻¹), and the highest in ground-ice melt-water (48–1,548 mg C·L⁻¹) (Ma et al., 2019b) (Figure 4B). DOC ranges in lakes/ponds (0.5–49 mg C·L⁻¹) are generally higher than those in rivers/streams (1–32 mg C·L⁻¹) and soil solutions (2–30 mg C·L⁻¹) (Abbott et al., 2014; Ma et al., 2019b). The distributive features of aquatic DOC in the SAYR are consistent with existing and latest research results. Correspondingly, the BDOC of ground-ice meltwater is the highest among all water types (23–71%), followed by lakes/ponds (1–71%), rivers/streams (1–40%), and soil solutions (0–10%) (Mann et al., 2015; Payandi-Rolland et al., 2020; Selvam et al., 2017; Textor et al., 2019; Vonk et al., 2013, 2015b) (Figure 4B). Therefore, DOC associated with ground-ice melt-water, lakes/ponds, and rivers/streams are relatively labile and potentially a major atmospheric carbon source among other aquatic water types in permafrost regions.