The Wujiang River is the largest tributary of the upper Changjiang River in its south bank. It originates from the eastern Wumeng Ranges in Yunnan–Guizhou Plateau and flows through Yunnan Province, Guizhou Province, Sichuan Province, and Hubei Province. It flows into the Three Gorges Reservoir in Fuling District, Chongqing Municipality. The drainage area of Wujiang River is 87 920 km2, of which 76 % drains in Guizhou Province. Its average annual discharge is 534 × 108 m3 yr-1. The watershed belongs to a warm subtropical climate. The mean annual air temperature varies from 10 to 18 ∘C with mean annual precipitation from 800 to 1400 mm. The altitudes of Wujiang River range from 500 to 2400 m with a decreasing trend from west to east. Plant species are diverse with the upper reach dominated by broadleaf evergreen forest and dryland crop, the middle reach dominated by evergreen broadleaf forest and deciduous broadleaf mixed forest, and the lower reach dominated by subtropical evergreen Castanopsis forest. Yellow soil and limestone soil are dominant in the watershed (Zhang et al., 1995). The land use is dominated by forest land, cultivated land, and grass land, which account for 50, 31, and 18 %, respectively. The soil erosion rate decreased from 2678 t km-2 yr-1 in 1980s to 2313 t km-2 yr-1 in 1990s due to sustainable soil conservation measures (Wang, 2011).

The Wujiang River is a typical karst watershed. In the upper reaches, Permian and Triassic carbonate rocks and basalt are dominant with coal-bearing formations outcropping in the western area. In the midstream area, Permian and Triassic limestones, dolomitic limestones, and dolomites are dominant whereas the lower reaches are dominated by carbonate rocks intercalated with shales, sandy shales, and siltstones (Zhang et al., 1995).

In all, 25 samples of suspended particulate matter (SPM) were collected from the mainstream and major tributaries of the Wujiang River (Fig. 1) in May and August 2013, respectively. In all, 18 surface sediments were sampled in August 2013. An extreme drought occurred in most Guizhou Province during June and August. Figure 2 showed the monthly water discharge and suspended sediment load in 2013 at the Wulong Hydrological Station, located at the Wujiang River mouth. The data were taken from Changjiang Sediment Bulletin (2013). As plotted in the Fig. 2, water discharge and suspended sediment load decreased abruptly from June to August due to the drought event.

Sampling of SPM were conducted by filtration through precombusted (450∘ for 6 h) and preweighted 47 mm glass fiber filters for SPM weight concentrations and stable isotopic analyses of carbon and nitrogen. Surface sediments were collected using a 0.05 m2 Van Veen grab (Jiang and Ji, 2013). All samples were stored in a freezer (≤ 20∘) prior to laboratory analyses.

The filter samples were freeze-dried before the particulate substance was scraped from the filter. The freeze-dried suspended matter and sediment were sieved to 200 mesh, treated with 1 M HCl to remove inorganic carbon, while nitrogen isotopes of particulate matter were measured on the bulk samples without acidification. Organic carbon (OC) and total nitrogen (TN) contents as well as isotopes of carbon and nitrogen were determined by using an elemental analyzer (Flash EA 1112HT, Thermo Fisher Scientific, Inc., USA) coupled with an isotope-ratio mass spectrometer (Finnigan Delta V Advantage, Thermo Fisher Scientific, Inc., USA) in the Laboratory of Stable Isotope Ratio Mass Spectrometry, Chinese Academy of Forestry (Beijing 100091, China). Stable isotope ratios are reported in δ-unit notation as follows: δX(‰)=(Rsample/Rstandard-1)×1000, where Rsample is the δ13C / 12C or 15N / 14N ratios of the sample, and Rstandard is the corresponding ratios of sample standard. δ13C values are reported relative to Pee Dee Belemnite (PDB) and δ15N values are reported relative to N2 in atmospheric air (AIR). Precision for δ13C is 0.2 ‰ and for δ15N is 0.2 ‰.

The minerals of suspended particulate and surface sediment were analyzed in State Key Laboratory for Advanced Metals and Materials, University of Science and Technology, Beijing. To test the relationships between total suspended sediment (TSS) concentrations, elemental compositions (POC %, TN % and C / N ratios), and isotopic compositions, Person's correlation coefficient (Table S2 in the Supplement) was calculated using Statistical Package for Social Science (SPSS) software (version 16.0 for Windows).

As presented in the Fig. S1, the major minerals of SPM in the Wujiang River included detrital minerals (quartz, calcite, and dolomite), clay minerals (illite, kaolinite, and smectite), magnetite, and ilmenite. In contrast, the surface sediments contained few clay minerals, reflecting the preferential enrichment of clay minerals in SPM. The enrichment of clastic carbonate minerals in surface sediment indicated the preferential sedimentation of clastic carbonate relative to clay minerals. This was similar to the study in the Changjiang River by Ding et al. (2014). The dolomite in the SPM and surface sediment was observed in the middle and lower reaches, which was in agreement with the catchment lithology. The mineral composition of SPM and surface sediment reflected the process of physical and chemical weathering.

The ratios of carbon to nitrogen showed the wide range of 2.8–29.3, with a mean value of 13.6 in May and 8.8 in August (Table S1), indicating a source with high C / N ratios in May and low C / N ratios in August. Figure 3 showed the spatial and seasonal variations of C / N ratios, δ13CPOC, and δ15CTN in the Wujiang River. Compared with August, more samples in May had C / N ratios higher than 15, whereas more samples had C / N ratios < 8 in August, which were distributed in sites near or in reservoirs, for example sites 1, 2, and 19 (Fig. 3). Considering the cascade of reservoirs along the Wujiang River, the impact of the reservoir should be examined. The sampling sites were divided into two types based on the relation with reservoirs: sites directly affected by reservoirs and less affected by reservoirs. Table 1 showed the comparison of elemental and isotopic parameters in these two types of sites. As shown in the Table 1, C / N ratios were lower in sites directly affected by reservoirs than those in sites less affected by reservoirs.

The δ13CPOC of SPM in May ranged from -30.18 to -21.09 ‰ and averaged -26.30 ‰. Differently, δ13CPOC in August displayed relatively depleted values with an average value of -27.23 ‰. δ15NTN of SPM ranged from 1.88 to 12.93 ‰ and averaged 6.82 ‰, with higher values in August (7.58 ‰) than those in May (6.05 ‰). Spatially, δ13CPOC values, especially in August, were more depleted in reservoirs and sites downstream of the reservoirs (-28.65±1.22 ‰; Table 1 and Fig. 3) than those less affected by reservoirs (-26.68 ± 3.23 ‰; Table 1 and Fig. 3). In contrast, δ15NTN values were more enriched in sites directly affected by reservoirs than other sites. Chen and Jia (2009) obtained the similar trend of δ13CPOC and δ15NTN in a dam-controlled subtropical river. As presented by Fig. 3, the heavier δ15NTN values (> 10 ‰) in August corresponded to lighter δ13CPOC values and lower C / N ratios (Fig. 3).

As shown in Table S2, POC %, TN %, and δ15NTN were significantly negatively correlated with TSS concentrations in May; i.e., POC %, TN %, and δ15NTN decreased when TSS concentrations increased. In contrast, δ13CPOC was significantly positively correlated with TSS concentrations; i.e., δ13CPOC increased when TSS concentrations increased. Similar results are obtained in the freshwater part of the Scheldt Estuary (Hellings et al., 1999). However, the relationships between isotopic parameters and TSS concentrations in August were not significant. This might be due to the low concentrations of TSS in August, when extreme drought occurred in most Guizhou Province. No significant correlation of δ15NTN and δ13CPOC was observed, which could be related to the inorganic nitrogen in samples (Guerra et al., 2013; Hu et al., 2006).

Carbon to nitrogen ratios in the surface sediments presented higher values ranging from 7.7 to 41.1 compared with those of SPM. Like the SPM, POC % and TN % in surface sediments displayed a relatively significant positive correlation (R=0.595, P<0.01), which indicated that some nitrogen in the sediments was inorganic nitrogen. Compared with the C / N ratios in SPM, surface sediments had higher C / N ratios, indicating different sources of POC in surface sediments.

δ13CPOC of surface sediments showed a relatively narrow range of -26.40 to -22.73 ‰ with an average value of -24.76 ‰ . The enriched δ13CPOC of surface sediments in contrast with SPM indicated a source with elevated δ13CPOC values. The variation trend of δ13CPOC in surface sediments was similar to that in SPM, which indicated these two carbon pools might be coupled. δ15NTN of surface sediments ranged from 2.88 to 9.39 ‰ with a mean value of 6.01 ‰. Spatially, the mean values of δ13CPOC and δ15NTN in sites directly affected by reservoirs and sites less affected by reservoirs were -24.85 ± 0.91 ‰ vs. -24.73 ± 1.04 ‰ and 6.23 ± 2.22 ‰ vs. 5.93 ± 1.55 ‰ (Table 1), respectively. Compared with the SPM, the difference of isotopic values were smaller in surface sediment between sites directly affected by reservoirs and sites less affected by reservoirs.

Knowledge of the POC transformation is useful to get a better understanding of the riverine carbon cycle. In-stream processes, such as assimilation and respiration of phytoplankton, affect the isotopic compositions and element contents of carbon and nitrogen. Hence, δ13CPOC, δ15NTN, and C / N can be utilized to trace transformation processes of organic matter. The trend of increasing TSS concentrations with decreasing POC contents (%, Fig. 6) indicated that POC contents (%) of SPM were diluted with the inorganic constituents derived from soil erosion. This was similar to other rivers (Ludwig et al., 1996; Jiang and Ji, 2013; Zhang et al., 2009). The positive correlation between TSS concentrations and POC contents (µmol L-1, Table S2) indicated that terrestrial organic matter was an important source of POC in SPM, which confirmed the erosion process. Although the erosion rate catchment decreased significantly in the Wujiang River (Wang, 2011), soil erosion had remarkable influence on the riverine carbon cycles.

δ15N is a potential tracer to identify aquatic biogeochemical processes. High δ15N may be caused by anthropogenic activities and transformation processes, such as denitrification and assimilation. Denitrification was excluded due to the weak correlation between δ15N–NO3- and δ18O–NO3- (Li and Ji, 2016). Kendall et al. (2007) reported that animal waste and domestic effluents had typical values of δ15N–NO3- > 10 ‰. Some samples had high δ15NTN values with elevated δ15N–NO3- values (> 8 %), indicating the inputs of manure and domestic sewage. Particulate organic carbon could be influenced by sewage water through the uptake of NH4+ and NO3- by phytoplankton. Uptake of NO3- with high δ15N by phytoplankton might result in elevated δ15NTN values (Kendall et al., 2001; Jiang and Ji, 2013). This process was confirmed by the significant positive correlation of δ15NTN–SPM vs. δ15N–NO3- and δ15NTN-SPM vs. NO3- in May (Fig. 7a). However, some samples with δ15NTN–SPM > 10 ‰ (δ15N–NO3- < 8 ‰) (Fig. 7a) deviated from other data in August, indicating other influencing factors. One mechanism might be related to uptake of other forms of dissolved inorganic nitrogen. The drought event during June to August created a longer time for the mineralization of soil organic matter, favorable to the production of ammonia with heavy δ15N values. The elevated δ15NTN–SPM might be caused by the uptake of nitrification-derived NH4+, as nitrifiers preferentially consumed 14N, leading to an increase in δ15N–NH4+ of the remaining ammonia. Consumption of such enriched 15N–NH4+ by phytoplankton resulted in the scattered trend of δ15NTN–SPM in August. A similar result was observed in the study by Sarma et al. (2012).

Positive correlations of δ13CPOC (Fig. 7b) and δ15NTN (Fig. 7c) between suspended and surface sediments indicated that an intense exchange might exist in these two carbon pools (Jiang and Ji, 2013; Sarma et al., 2012). The resuspension/deposition of suspended matter with mixing of different organic matter sources might result in the significant correlation of δ13CPOC (Fig. 7b) and δ15NTN (Fig. 7c). Compared with suspended matter, the heavier δ13CPOC and lighter δ15NTN in sediments indicated an enriched source of refractory allochthonous organic matter. The good correlations between river water and SPM as well as surface sediment indicated that these carbon pools were tightly connected.

Flux of POC (FPOC) was estimated based on a simple method from Tao et al. (2009): FPOC=[POC]H×Discharge× 66%+[POC]L×Discharge× 34%, where [POC]H and [POC]L is the average POC concentration of SPM in high-water season and low-water season, respectively. The POC concentration of river mouth (sample 18 in Fig. 1) was used to calculate the POC flux. Water discharge of the Wujiang River in 2013 (website: www.cjw.gov.cn) was taken from the Wulong hydrologic station, which was close to the river mouth. Discharge in high-flow and low-flow season accounted for 66 and 34 % of the annual discharge in 2013, respectively (Changjiang Sediment Bulletin, 2013). Since the water discharge in May was the almost its highest compared with other months (Fig. 2), the POC concentration in May collected at the river mouth (site 18 in Fig. 1) was used to calculate the POC flux in high-water season. Similarly, POC concentration in August was used to calculate the POC flux in low-water season since water discharge in August was close to that in low-water season (Fig. 2). The estimated POC flux was 1.17 × 1010 g in 2013, lower than that in the Wujiang River determined by Tao et al. (2009) in 2002. The decrease in POC flux might be due to measures of soil and water conservation, dam construction (Wu et al., 2007) as well as decreased discharge due to extreme drought in 2013 in Guizhou Province.

In Table 2 water discharge, TSS concentration, POC %, POC flux, and yield were compared with 15 world rivers. The total drainage area of the 15 rivers amounted to 30.3 × 106 km2 (Table 2), accounting for 55 % of the 60 world rivers' area from the study of Gaillardet et al. (1999). As shown in Table 2, the very low POC yield in the Wujiang River was the forth lowest observed in the rivers documented before Yenisey, Ob, and St. Lawrence. In comparison, the first two rivers in terms of POC yield were Zhujiang and Ganges–Brahmaputra with POC yield higher than 3 t km-2 yr-1. Both Zhujiang and Ganges–Brahmaputra were located in the mid-latitude, subject to tropical and subtropical climate. The latitudinal distribution patterns of POC yield were examined for the 15 world rivers and Wujiang River (Fig. 8a). As shown in the Fig. 8a, POC yields were higher in the mid-latitude rivers and tended toward the subtropical rivers. This was similar to the distribution feature of HCO3- yield for the world rivers in the study of Cai et al. (2008). According to Amiotte Suchet et al. (2003), carbonate rocks were mainly distributed between 20 and 50∘ N. The carbonated area in the documented 15 world rivers accounted for 8 % of the 60 world rivers' area, while POC flux accounted for 13 % of the global POC flux of 0.17 Gt (1015 g) estimated by Ludwig et al. (1996). It appeared that more carbonate would result in elevated POC export. Unfortunately, no clear correlation was found between carbonate percentage and POC yields (Fig. 8b), which indicated that percentage of carbonate area was not a significant factor of riverine POC transport. This was not similar to Mackenzie River, where organic-rich sedimentary rocks contributed a significant particulate organic matter (Carrie et al., 2009). However, the influence of carbonate might be underestimated as discussed by Cai et al. (2008) since carbonate rocks were defined as those that contained up to 50 % of carbonate minerals (Amiotte Suchet et al., 2003).

Compared with the previous study in Wujiang River, the POC yield decreased from 0.47 t km-2 yr-1 (Tao et al., 2009) to 0.13 t km-2 yr-1 (Table 2). It was noted that five dams were constructed in the lower reaches of the Wujiang River after the study of Tao et al. (2009). Moreover, POC yield in the Wujiang River was much lower than that in Xijiang and Zhujiang with a high carbonate percentage. This could be related to the smaller watershed area and extensive water reservoirs in the Wujiang River (Zhang et al., 2006). The impacts of reservoirs will be discussed below.

There are eleven artificial dams along the mainstream of Wujiang River (Fig. 1). The cascade of reservoirs created by dams may exert significant impacts on source and transport of riverine POC. In addition, extreme drought must be noted when considering the meteorological characteristics of the Wujiang River in 2013.

δ13C and δ15N proved to be potentially useful indicators for qualitative estimation of reservoirs and climatic impact on POC in dam-affected rivers (Chen and Jia, 2009; Zhang et al., 2009). In order to analyze the reservoir impact on riverine POC, δ13C and δ15N values were compared in sites directly affected by dams with those less affected in the Wujiang River (Table 1). As shown in Table 1, δ13C were more depleted in sites affected by dam than those less affected by dam. This was similar to the study by Chen and Jia (2009) in a dam-controlled river. Compared with sites far from the dam, the more depleted values of δ13C in sites close to dam were attributed to increasing phytoplankton contribution. This was confirmed by the higher phytoplankton contribution to POC of SPM in sites directly affected by dam with an average of 47 % relative to those less affected with an average of 27 %. Two mechanisms could explain this elevated phytoplankton contribution: (1) extended water retention time in reservoirs with low flow; (2) increasing light availability due to the low TSS concentrations (Table 1) in reservoirs. In contrast with δ13C values, δ15N values were heavier in sites close to reservoirs than those far from reservoirs. Delong and Thorp (2006) reported that δ15N values in aquatic algal were more enriched than detrital fraction. Thus, the heavier δ15N values in sites directly affected by dams might result from the aquatic algal and plankton. This hypothesis was verified by the concurrent lighter δ13C values and lower C / N ratios in the sites directly influenced by reservoirs (Table 1). Moreover, denitrification could be another cause for the heavier δ15N values in the reservoirs since hypoxic environments might occur in the deep water. Further study of δ15N variations in different water depths could help to trace the denitrification process. Seasonally, the average δ13C values in August at sites directly affected by reservoirs (-28.65 ‰) were much higher than other sites (-26.68 ‰). This large difference could be related to the higher temperature and extreme drought during June to August, which was favorable to in situ phytoplankton production. According to Chen and Jia (2009), accumulated terrestrial organic matters in winter were flushed during the first heavy rain, which resulted in increasing terrestrial input in the onset of the wet season. This could also be the reason for the Wujiang River in May, when the water discharge increased abruptly.

It is difficult to quantitatively evaluate the impact of dams and extreme drought. The method described in the study of Yu et al. (2011) made it possible to distinguish the impact of dams and climate. This method was based on the significant correlation between POC flux and suspended sediment load. Thus, the variations of suspended sediment load could reflect the POC flux variations under the condition of dam and extreme drought. The comparison of suspended sediment loads was made between normal years and the drought year 2013 in the Wulong hydrologic station, at the mouth of Wujiang River (Table S3). As shown in Table S3, the suspended sediment loads in 2013 at Wulong station reduced by 80 % compared with normal years, which could be due to the combined impacts of cascade of dams and extreme drought. Because suspended sediment at Wulong station directly flowed into Three Gorges Dam (TGD), the impact of climate on TGD sediment revealed the similar impacts on Wulong station. The impacts of extreme drought can be estimated based on the comparison of sediment inputs to TGD between normal and drought years. The reduction 41 % of sediment inputs to TGD was obtained in 2013 compared with normal years, which was attributed to the climate impact. Thus, the impact of cascade of dams was 39 %. The normalized impacts of cascade of dams and extreme drought were 49 and 51 %, respectively. This result was similar to the impacts of extreme drought in 2006 on TGD determined by Yu et al. (2011). This result indicated that extreme drought and dams were important factors of suspended sediment load. Considering the significant correlation between TSS concentrations and POC concentrations (Table S2), reservoirs and extreme drought had a similar impact on the POC transport. However, this estimation based on the Wujiang River mouth was limited relative to the whole basin.