Carbon accumulation and sequestration of lakes in China during the Holocene

Study regions

There are five major lake regions in China: lakes in East Plain of subtropical China (EP), Yunnan-Guizhou Plateau in southwest China (YG), Mongolia-Xinjiang Plateau (MX), the Qinghai-Tibetan Plateau (TB), and northeast China (NEC) (Wang & Dou, 1998). The climate differs among these regions, with the warmest and wettest in EP, followed by YG. The other regions have cold and dry climate, especially that in TB and MX (Table S1). Moreover, lakes in China show regional difference in evolution and formation. In general, lake formation started much later in EP than in other regions (Wang & Dou, 1998), mostly beginning to form in the middle or late Holocene; in contrast, lakes in other regions mostly initiated before Holocene (Wang & Dou, 1998). Among the five major lake regions in China, TB, EP, and MX are lake-rich regions, with lake area accounting for ~94% of the total lake area of China (Table 2).

Data set and analysis

We compiled lake sediment records with multiple radiocarbon (¹⁴C) ages from available published sources across China (Table S1). The lakes selected for synthesis represent a wide range of lake types, lake areas, catchment sizes, water depths, and climate conditions (Table S1). The lakes span a latitudinal range from 21.17°N to 47.70°N, a longitudinal range from 81.17°E to 132.42°E, and an altitudinal range from 5 m to 5051 m a.s.l. (Table S1). We used 82 lake sediment records with multiple radiocarbon ages (mostly more than 7 dates) to investigate the temporal and spatial patterns of sediment accumulation rates during the Holocene (Table S1). The exception was lakes in EP, where multiple radiocarbon ages mostly did not cover the early Holocene as a large portion of lakes there initiated in the middle or late Holocene.

All ¹⁴C dates were converted to calibrated dates using program CALIB Rev. 6.0.1 with the IntCa104 calibration data set (Reimer et al., 2004). Lake sediment accumulation rate (SAR) at 1000-year bins for each profile was calculated based on dated depths of the sediment profiles, as been done on peatland accumulation records (Yu et al., 2010).

We used 58 lake sediment records with direct measurements of OC content among 82 lakes to investigate the temporal and spatial patterns of lake C accumulation rate (CAR). Only Qinghai Lake and Erlongwan had dry bulk density (DBD) data (Fig. 1). For those without DBD data, we determined DBD using the empirical relationship with C content as reported in the literature (Dean & Gorham, 1998; Avnimelech et al., 2001; Kastowski et al., 2011). In case of high organic carbon (OC) content (>6%), we used the relationship: DBD = 1.665*(OC)^−0.887 given by Dean and Gorham (1998). In case of low OC content (≤6%), the DBD for sediments was calculated using the formula: DBD = 1.776 − 0.363*ln (10*OC) as reported by Avnimelech et al. (2001), where OC is in (%). We also generated the 1000-year bin TOC and DBD. The CAR was generated in 1000-year bins using the following formula. CAR = (SAR × OC)/(1000 × DBD), where SAR is in (mm yr⁻¹); DBD is in (g m⁻³); and OC is in (%).

The total lake area for each region was from an authoritative limnology book in China (Wang & Dou, 1998). The total lake area along with CAR was used to estimate the regional lake C pool. To estimate lake C pools at 1000-year bins in different regions, we multiplied CAR for each 1000-year bin by lake area. The C pool intervals were added up for cumulative C pools since 12 ka.

We also carried out regression analysis between the lake C accumulation rate and other lake characteristics or environmental variables. For each potential relationship, we included data from any lakes in the set where both CAR and the relevant lake characteristic were available, except for a few outliers. The annual temperature was available for 50 lakes, annual precipitation for 52, average water depth for 28, maximum water depth for 44, lake area for 49 and the ratio of catchment area to lake area for 36 lakes. We also studied the relationship between CAR and SAR. Lake area and the ratio of catchment area to lake area were log-transformed before regression analysis. All regression analyses were conducted with the SigmaPlot 12.0 (SPSS Inc., Chinago, IL, USA). One-way analysis of variance (anova) was used to determine whether the mean Holocene sediment total organic carbon (TOC) content, SAR, and CAR differed significantly among regions and whether the three variables of 1000-year bins differed over times. Multiple comparisons among means were obtained by Duncan's multiple comparison procedure at the significance level of 0.05. These analyses were conducted using SAS 9.1/PROC GLM (SAS Institute, Cary, NC, USA).

Temporal and spatial pattern of the sediment and carbon accumulation

The temporal dynamic patterns of sediment OC content, and sediment and C accumulation varied significantly among regions (Figs 2-4). Lakes in TB, MX, and NEC showed an increasing trend in sediment OC content (Fig. 2a–c); the sediment and C accumulation pattern suggested a growing trend for lakes in most regions, except for NEC and EP during the last 12 ka (Figs 3 and 4). No obvious increasing trend of CAR was found in lakes of NEC (Fig. 4c), nor was there any of either SAR or CAR for those in EP (Figs 3d and 4d). Overall, the sediment and C accumulation pattern of China's lakes in this study set showed a growing trend (Figs 3f and 4f), but the sediment OC content lacked such trend during the past 12 ka (Fig. 2f). The SAR of 1 ka was 0.75 ± 0.13 mm yr⁻¹, higher than that of other time periods; SARs of 2 ka and 3 ka were same as 0.55 ± 0.05 mm yr⁻¹ higher than that of 11 ka (0.34 ± 0.04 mm yr⁻¹) and 12 ka (0.35 ± 0.04 mm yr⁻¹) (P < 0.05) (Fig. 3f).

The sediment OC content was 4.27 ± 0.68% for lakes across China, with a significant higher value of lakes in YG than its counterparts in TB, MX, and EP (P < 0.05) (Fig. 5a). The overall SAR of lakes in China was 0.47 ± 0.05 mm yr⁻¹ (Fig. 5b). The mean lake SAR in EP was 1.05 ± 0.28 mm yr⁻¹, which was higher than that of other regions (P < 0. 05); the SAR of the other regions differed insignificantly (Fig. 5b). The mean lake CAR in China was 7.7 ± 1.4 g C m⁻² yr⁻¹, with the highest of 11.5 ± 5.6 g C m⁻² yr⁻¹in EP and the lowest of 5.8 ± 1.6 g C m⁻² yr⁻¹in MX. However, the CAR did not differ significantly among regions (Fig. 5c).

Lake C storage and total C burial in China

The C storage accumulated since 12 ka averaged 135 kg C m⁻² (ranging from 24 to 343 kg C m⁻²) in EP, followed by 96 (5–273) kg C m⁻² in NEC, 92 (16–142) kg C m⁻² in YG, 73 (16–274) kg C m⁻² in TB and 68 (7–192) kg C m⁻² in MX. The lake sediment C has been accumulated mostly since the mid-Holocene in all regions (Table 1).

The total lake C burial in China was estimated to be 7.95 ± 0.97 Pg C since the last 12 ka, with nearly 72% of this C stock accumulating since 8 ka (Table 2). TB has the largest lake C stock in China, followed by the EP. The total lake C stock of these two regions accounted for nearly 77% of the total lake C storage in China (Table 2).

Relationship between Holocene lake sediment CAR and lake characteristics

The average CAR of past 12 ka for all lakes in China was positively correlated with annual temperature (r = 0.28; P = 0.047) (Fig. 6a), annual precipitation (r = 0.43; P = 0.002) (Fig. 6b), and the ratio of catchment area to lake area (r = 0.28; P = 0.047) (Fig. 6f). The lake average Holocene CAR was not significantly related to the lake area (Fig. 6e) or the average or maximum lake water depth (Fig. 6c, d). However, the correlation between CAR and lake catchment characteristics varied among regions. The lake CAR in TB showed a negative correlation to maximum water depth (r = 0.53; P = 0.05) (Fig. 6d) and lake area (r = 0.49; P = 0.03) (Fig. 6e); the Holocene average CAR of lakes in EP had positive correlation with maximum water depth (r = 0.8; P = 0.03) (Fig. 6d) and the ratio of catchment area to lake area (r = 0.98; P = 0.02) (Fig. 6f). In addition, we found a significant correlation between the lake CAR and SAR (r = 0.65; P < 0.0001) (Fig. 7).

Correlation of lake sediment and C accumulation at different scales

This study found an overall increasing trend of SAR and CAR for lakes in China, which was mainly contributed by lakes in TB, MX, and YG (Figs 3 and 4). The SAR and CAR of a large portion of lakes in these regions increased since 12 ka, with relatively high accumulation rate in the middle or late Holocene (Figs. S6-7, S10-12, S15). The SAR and CAR in EP, and CAR in NEC did not follow a clear increasing trend during the Holocene (Figs 3 and 4). For lakes in NEC, this can be explained by the relatively high sediment and C accumulation in the early Holocene. Most of lakes there had an accumulation peak in the late Holocene due to human activities and a subpeak in the early Holocene as a result of warm and wet climate (Liu et al., 2005, 2010; You & Liu, 2012) (Figs S3, S8). For lakes in EP, the sediment and C accumulation did not have an obvious increasing trend may be due to the lack of the early Holocene data, as most of them initiated in the middle or late Holocene (Figs. S9, S14).

Controls on lake sediment and C accumulation dynamics in China

This study demonstrated that temperature and precipitation were positively correlated to CAR of lakes across China, suggesting that regional climate difference can partly explain the variability in CAR of lakes among different regions. Indeed, EP, the warmest and wettest among the five regions in China (Table S1), had the highest sediment and C accumulation rate, which supports our hypothesis that higher C sequestration occurs at warmer and wetter region. However, climate did not serve as a strong factor in explaining within-region CAR variability, as neither temperature nor precipitation was found linked to CAR. Moreover, climate can act as an important controller over the temporal dynamics of lake sediment and C accumulation. The sediment and C accumulation pattern of a couple of individual lakes showed that the accumulation peak occurred at early–middle Holocene (Figs. S6–S15), consistent with the Holocene warm and wet climate.

Lake sediment accumulation reflects not only the climate history but also human activities, as the matter output from catchments constitutes an important portion of matter input of lake ecosystems (Whitmore et al., 1994). Our result showed CAR was significantly related with the ratio of catchment area to lake area for lakes in China, suggesting that human activity has a huge effect on lake C accumulation. The relatively high SAR in EP can probably be partly explained by the intensive human activities. EP is responsible for a large portion of the whole national agricultural production (National Bureau of Statistics of China, 2011). Increasing evidence suggests that intensified agricultural activities lead to pronounced eutrophication of many of the lakes there (Dong et al., 2008; Yang et al., 2008; Zhang et al., 2010; Chen et al., 2011). Similarly, anthropogenic interference was suggested to be responsible for the relatively high CAR of a few individual lakes, for example, Daihai, Gonghai, Xingyun Lake, Gucheng Lake, and Poyang Lake (Wu, 1999; Xiao et al., 2006; Chen et al., 2013; Hillman et al., 2014; Zhu et al., 2015).

Moreover, the temporal dynamics of sediment and C accumulation of lakes in China was observed to be closely linked to human activities. The sediment and C accumulation pattern of lakes in China showed an increasing trend, with an accumulation peak at 3–1 ka, probably as a result of intensification of human activity. Also, the SAR and CAR of a large portion of individual lakes in this study set were relatively low in the early and middle Holocene, but increased rapidly since 3 ka (Figs. S6–S15). This could be related to the beginning of intensifying anthropogenic impacts at ~3–1 ka in all regions of China, based on analysis of pollen, magnetic susceptibility, grain size, and charcoal (Ren, 2000; Han, 2004; Shen et al., 2005; Atahan et al., 2008; Mischke et al., 2010; Miehe et al., 2014). Human interference was mainly characterized by pastoralism in the Qinghai-Tibetan Plateau (Schlütz & Lehmkuhl, 2009; Mischke et al., 2010; Wischnewski et al., 2014), deforestation and agricultural cultivation in MX (Yi et al., 2003; Han, 2004), YG (Shen et al., 2005; Zhao et al., 2007), EP (Ren, 2000; Wang et al., 2008), and NEC (Ren, 2000; Makohonienko et al., 2004). Human-induced deforestation, cultivation, and pastoralism destroyed the stability of catchment soils, resulting in accelerating erosion, transportation (Whitmore et al., 1994) and sediment accumulation of water ecosystems. It was true that human impact was found to increase the sediment accumulation of lakes (Whitmore et al., 1994) and rivers (Saito et al., 2001) in China. Mulholland & Elwood (1982) also reported that human activities increased the C accumulation of lakes significantly in Central US.

Besides, other characteristics including lake area, catchment area and slope can significantly affect lake sediment and C sequestration. It was suggested the high CAR of Tiancai Lake was probably due to the small lake area (0.02 km²) (Han et al., 2011), although we did not find a significant correlation between lake area and CAR; the high CAR of Bosten Lake was probably attributable to the large ratio of catchment area to lake area (Wünnemann et al., 2006), which is supported by our result that CAR was positively related to the ratio of catchment area to lake area; the high CAR of Chen Co was probably due to steep lake shore (Feng et al., 2004). However, we did not study the relationship between slope and CAR though it was proved to be an important factor in affecting CAR (Kastowski et al., 2011).

Comparison of lake C sequestration in China and other regions in the world

The long-term CAR of lakes in China was 7.7 g ± 1.4 g C m⁻² yr⁻¹, comparable to its counterparts in southwest Greenland (Anderson et al., 2009), Europe (Kastowski et al., 2011), Finland (Kortelainen et al., Kortelainen et al., 2004), and northern Quebec (Ferland et al., 2012) (Table 3). This suggests that lakes in China play a crucial role in regional C cycle as lakes elsewhere.

The lake CAR did not show simple variation along latitude (Fig. 8). Instead, this study found the relatively low value appearing at 30–35°N, besides at 60–70°N. The relatively low CAR for lakes at 30–35°N was mainly related to their high altitude and corresponding cold and dry climate as most of these studied lakes are located in the Qinghai-Tibetan Plateau with an average altitude of 4266 m (Table S1). The highest CAR at 55–60°N was probably due to intensive human activities in the lake catchment (Kastowski et al., 2011). For example, three lakes including Schwarzsee in Switzerland (Dapples et al., 2002), Barton Broad in UK (Bennion et al., 2001), and Lake Dudinghausen in Germany (Dreßler et al., 2006) with high C accumulation rate of ~100 g C m⁻² yr⁻¹ were significantly affected by human activities.

However, a couple of lakes in China which had been intensively impacted by human activity did not have high CAR as their counterparts elsewhere (Mulholland & Elwood, 1982; Bennion et al., 2001; Dapples et al., 2002; Dreßler et al., 2006; Kastowski et al., 2011). For example, lakes such as Chaohu and Taihu have obviously been affected by human activities significantly (Qin et al., 2007; Zhang et al., 2010; Yang et al., 2013), but the long-term CAR was only 2.26 g C m⁻² yr⁻¹ in Taihu and 5.04 g C m⁻² yr⁻¹ in Chaohu. Similarly, Dong et al. (2012) estimated the recent lake C accumulation of lakes in middle and lower reaches of Yangtze River as ~15 g C m⁻² yr⁻¹, much smaller than that of lakes in subtropical regions elsewhere. The recent CAR was estimated to be ~100 g C m⁻² yr⁻¹ in an Amazon floodplain lake (Moreira-Turcq et al., 2004) and 60 g C m⁻² yr⁻¹in a Florida lake (Brezonik & Engstrom, 1998). The relatively low CAR of these lakes was mainly due to high mineralization, as suggested by a previous study (Dong et al., 2012). Similarly, the TOC content of sediments in EP is relatively low (~2.3%) compared with lakes elsewhere. This was probably related to the large area and small depth of most lakes in EP (average depth: ~2 m; average lake area: ~1500 km²) (Table S1), which can promote oxygenic mineralization of lake sediments through facilitating wind-induced resuspension (Sobek et al., 2009). Indeed, we found significant relationship between CAR and maximum lake water depth in EP (Fig. 6d). High water temperature of lakes in EP was probably another important factor contributing to the enhancement in mineralization (Gudasz et al., 2010), although temperature was indicated to increase lake primary productivity by this study. In addition, temperature was suggested to promote the growth of Cyanobacterial crops (Joehnk et al., 2008) and the switch of dominant primary production from Diatom-macrophyte to Cyanobacterial crops was found to promote the mineralization of organic matter of lake sediments (Qin et al., 2007).

The role of lakes of China in regional C cycling

With ongoing climate change, the C sink of natural ecosystem has attracted a lot of attention in China. Previous studies have focused on estimation of the C storage of forests (Xie et al., 2007; Yu et al., 2007), grasslands (Fang et al., 2010), crop lands (Tang et al., 2006), peatlands (Wang et al., 2014) and so on. However, the C stock of lakes in China still remains unknown. This study suggested that China's lakes which are important component of C stock (~8.0 ± 1.0 Pg) should be also taken into consideration when estimating the regional C stock of natural ecosystems. Moreover, on a unit area basis, the Holocene C density of lakes was more than eightfolds of other soil types (Fang, 1996; Ni, 2001; Li et al., 2003; Wu et al., 2003; Xie et al., 2007; Yu et al., 2007).Therefore, lakes are obviously C-rich ecosystems and thus an important participant in C cycling in China.

We recognize that this first estimate of long-term lake C accumulation rate and its C stock on regional and national scales in China as reported here is still limited by some unavoidable uncertainties. The largest error source is the general lacking of high-resolution ¹⁴C dating controls, especially for lakes in East Plain of subtropical China, and lacking data for dry bulk density measurements. Moreover, like the previous studies (Anderson et al., 2009; Ferland et al., 2012), we did not consider the lake area dynamics during the past 12 ka, which may cause bias in estimation of the lake C stock in China. Overall, this study updates our knowledge about how mass and carbon sequestration of China's lakes change under the past climate change and human activities and improves our understanding about the role of lakes in regional C cycling in monsoon-influenced regions.