Human Land Uses Enhance Sediment Denitrification and N 2 O Production in Yangtze Lakes Primarily by Influencing Lake Water Quality

The copyright of individual parts of the supplement might differ from the CC-BY 3.0 licence.

approximately 82 % of the total nitrate loss in lakes (Kreiling et al., 2011). Despite the significant increases in nitrogen loading in lake ecosystems and the growing attention to water quality degradation and toxic algal blooms, our understanding of large-scale sediment denitrification in lakes remains limited (McCrackin and Elser, 2010;Bruesewitz et al., 2011;Liu et al., 2015). 20 N 2 O is an intermediate product of sediment denitrification and may be released into the atmosphere in substantial amounts under certain conditions (Hefting et al., 2006). In Lake Taihu of China, both the littoral and pelagic zones can release large quantities of N 2 O, especially during the summer algal bloom period (Wang et al., 2007). However, N 2 O fluxes in the pelagic regions of shallow boreal lakes are reported to be negligible 25 (Huttunen et al., 2003), which indicates that lakes are only moderate sources of N 2 O (Mengis et al., 1997). N 2 O is a powerful greenhouse gas that is approximately 300 times stronger than CO 2 and is responsible for approximately 6 % of global warming (IPCC, 2007). Thus, it is important to examine the N 2 O production during sediment Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | denitrification. A wide range of relative N 2 O production during denitrification has been reported for wetlands, forests and cultivated soils (e.g., Garcí-Ruiz et al., 1998;Ullah et al., 2005). Seitzinger and Kroeze (1998) have indicated that the relative N 2 O production for heavily polluted river and estuary sediments is approximately 0.03. By contrast, the relative N 2 O production for lakes is still poorly understood (McCrackin and Elser, . At the within-lake scale, overlying water quality and sediment characteristics are recognised as the primary determinants of sediment denitrification and N 2 O production (Saunders and Kalff, 2001;Hasegawa and Okino, 2004;Zhong et al., 2010;Rissanen et al., 2011;Liu et al., 2015). Recently, several studies have found that aquatic plant communities can also greatly influence sediment denitrification in fresh waters (e.g., Forshay and Dodson, 2011;Wang et al., 2013). Plants may affect denitrification primarily through the modification of carbon inputs and soil redox conditions via oxygen excretion from roots (Sutton-Grier et al., 2013). At the watershed scale, few studies have addressed the indirect effects of watershed land uses on lake sediment denitri- 15 fication and N 2 O production (but see McCrackin and Elser, 2010;Bruesewitz et al., 2011;Liu et al., 2015), and no studies have examined whether the indirect effects are mainly mediated through lake water quality or sediment characteristics. A better understanding of multi-scale determinants of sediment denitrification in lakes is necessary to achieve high nitrogen removal efficiency and low N 2 O production rates. 20 In this work, we investigated sediment denitrification and N 2 O production and their relationships to within-lake variables and watershed land uses in 20 lakes from the Yangtze River basin in China. We hypothesised that the human land uses in watersheds would influence sediment denitrification and N 2 O production in lakes, primarily via effects on lake water quality. The objectives of this study were (1)  veal the mechanisms of lake sediment denitrification and N 2 O production in response to human land uses in watersheds for the first time.

Study sites
The Yangtze River, originating from the Qinghai-Tibet Plateau and flowing into the Pa-5 cific Ocean in the city of Shanghai, is the largest river in China and has a drainage area of approximately 1.8 million km 2 and a total length of 6300 km (Fig. 1). There are 648 lakes with areas greater than 1 km 2 in the Yangtze River basin (Wang and Dou, 1998). Most of these lakes are concentrated in the middle and lower Yangtze River basin (i.e., the Yangtze floodplain; Fig. 1), where alluvium predominates. Lakes in the 10 Yangtze floodplain are generally shallow (mean depth < 5 m) with short hydraulic retention times (Liu et al., 2011). Over the past decades, the watershed land uses of many Yangtze lakes have undergone substantial changes (Liu et al., 2012). Large areas of forests, wetlands and other natural landscapes have been transformed into agricultural, industrial, and urban 15 lands. For instance, the percentage of built-up lands in the Chaohu Lake watershed has increased from 6.59 to 9.58 % during the period of 1979-2008 (Wu, 2011). As a result of watershed land use changes, water quality in many lakes has deteriorated considerably in recent years, and up to 86 % of the Yangtze lakes are eutrophic or even hypereutrophic (Yang et al., 2010).

Field sampling
During the summer (from 30 July to 7 August) of 2012, twenty shallow lakes in the middle and lower Yangtze River basin of China were selected non-randomly on the basis of ease of access (Fig. 1). These selected lakes covered a wide range of physical and chemical conditions and vegetation characteristics (Wang and Dou, 1998 geographic location and morphology characteristics of these studied lakes are listed in Table S1. To obtain representative samples, one transect from the littoral zone to the lake centre was randomly established in each lake and 3-4 sampling sites were chosen along this transect at regular intervals based on lake area (Yang et al., 2008). Littoral sampling sites were generally located at a water depth less than 1.5 m, approximately 5 50-100 m away from the lake shore. At each sampling site, three replicate surface sediments were randomly collected within an approximately 50 m 2 area from a boat using a Peterson dredge and then mixed and homogenised to form a composite sample. For each site, approximately 1 kg of sediment was placed in a sealed plastic bag and stored at approximately 5 • C 10 in a refrigerator until return to the laboratory. At each sampling site, a 200 mL unfiltered water sample was collected at a depth of approximately 1 m for use in the denitrification and N 2 O production assays. At the same time, we collected an additional 200 mL water sample at the same depth for water quality analyses in the laboratory. Water sampling was performed before sediment collection to prevent sediments from being 15 resuspended and thereby contaminating the water. At each sampling site, submerged macrophytes were sampled using a pronged grab (25 cm×35 cm) with three replicates, and the species richness (i.e., mean species number recorded in a sampling site) and fresh plant biomass were calculated. 20 The most common method for measuring sediment denitrification is based on the ability of acetylene to inhibit the reduction of N 2 O to N 2 during denitrification (Groffman et al., 1999). Although the acetylene blockage technique has a number of limitations (Yu et al., 2010;Felber et al., 2012), it is still amenable to large-scale comparisons of denitrification, especially for systems with moderate or high NO  , 2006). In this study, we measured potential and background denitrification rates of lake sediments using the acetylene blockage technique. The potential denitrification rate was measured under optimal conditions (by supplying an excess of NO Introduction and organic carbon and ensuring anoxia) and thus provided an upper-bound estimate of in situ denitrification. Background denitrification (i.e., unamended denitrification) rate was a conservative estimate of in situ denitrification without carbon and NO − 3 amendments because acetylene also inhibited nitrification (NO − 3 production). The increase in N 2 O concentrations was always linear throughout the short assay duration, therefore, 5 short-term incubation (2-4 h) was recommended for measuring sediment denitrification of lakes (Bruesewitz et al., 2012).

Denitrification and N 2 O production assays
For potential denitrification rate assays, 50 g of homogenised sediments from each sampling site were slurried with 30 mL of incubation solution (final concentrations: 0.1 g L −1 KNO 3 , 0.18 g L −1 glucose and 1 g L −1 chloramphenicol) in a 250 mL serum 10 bottle. Each bottle was then sealed and purged with N 2 gas for 2 min to induce anoxic conditions. Approximately 10 % of the bottle headspace was replaced with acetylene to block the conversion of N 2 O to N 2 during denitrification. We measured background denitrification and N 2 O production rates using a similar procedure, but with the addition of 30 mL of unfiltered lake water instead of the incubation solution (McCrackin and 15 Elser, 2012). Parallel incubations with and without acetylene (10 % vol/vol in the bottle headspace) were used to differentiate between background denitrification and N 2 O production (McCrackin and Elser, 2010). All bottles were then incubated in the dark for 4 h at 25 • C (the approximate in situ water temperature). At the beginning and end of incubation, 5 mL of headspace gas samples were collected from each bottle (after 20 shaking vigorously) using a syringe. The N 2 O concentrations were measured using a gas chromatograph (Agilent 7890, Santa Clara, CA, USA) equipped with an electron capture detector. Potential denitrification, background denitrification and N 2 O production rates were calculated as the difference between the initial and final headspace N 2 O concentra- Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | N 2 O production rate by the background denitrification rate (García-Ruiz et al., 1998;McCrackin and Elser, 2010).

Sediment characteristics and water quality measurements
Sediment pH was measured in a soil to water ratio of 1 : 5 (v/v) using a pH meter and bulk density was determined by weighing soil cores of known volume after drying 5 for 24 h at 105 • C. Sediment moisture was measured gravimetrically (24 h at 105 • C) from 30 g sediment samples. Sediment total nitrogen (STN) was measured using the Kjeldahl method after digesting samples in a digestor using a sulfuric acid/mercuric oxide catalyst. Sediment total carbon (STC) content of air-dried samples was analysed by a TOC analyser (Vario TOC cube, Elementar, Germany).

10
Conductivity (Cond), oxidation-reduction potential (ORP), dissolved oxygen (DO), nitrate (NO − 3 ) and ammonium (NH + 4 ) were measured on site at the sampling depth using a YSI 6920 multiparameter water quality probe (YSI Inc., Yellow Springs, Ohio, USA). Total nitrogen (TN) concentration was analysed using the micro-Kjeldahl method. The concentrations of total carbon (TC) and total organic carbon (TOC) in waters were 15 measured with a TOC analyser (Vario TOC cube, Elementar, Germany).

Watershed land use calculation
After delineating the watershed boundaries of the 20 studied lakes using a 1 km resolution digital elevation model, overlay functions were used to extract a land use map of each lake's watershed from 100 m resolution national land cover data in ArcGIS 10 20 software (ESRI, Redlands, California, USA). The national land cover data, interpreted from recent Landsat TM images, were obtained from the Data Sharing Infrastructure of Earth System Science in China (http://www.geodata.cn/).
The original land use classes were further grouped into four main categories: (1) vegetation, including forest and grassland; (2) cropland, including dry land and paddy BGD 12,2015 Human land uses enhance sediment denitrification industrial areas, roads, and airports; and (4) water bodies, including lakes, rivers, streams, reservoirs, ponds, and wetlands ( Fig. 1). Bare land was not included in our analysis because its area in watersheds was very small. Cropland and built-up land can be considered as human-dominated land uses (HDL; Abdullah and Nakagoshi, 2006). The percentages of land use types in each lake's watershed were calculated in 5 ArcGIS 10 software. We used the HDL here, because the correlations between in-lake characteristics (i.e., water quality and sediment characteristics) and the percentage of cropland or percentage of built-up land were weak.
To quantify the vegetation cover, we calculated the normalised difference vegetation index (NDVI) of each watershed from the Moderate-resolution Imaging Spectrora-10 diometer (MODIS) red and near-infrared bands. The 250 m resolution MODIS data for the years 2010 and 2011 was obtained from the USGS EROS Data Center. The NDVI is a reflection of the biophysical condition of a watershed's vegetation cover, which, in turn, may affect the water runoff and water quality (Griffith et al., 2002). 15 For each studied lake, we calculated the mean values for sediment denitrification, N 2 O production, water quality, sediment characteristics and plant community structure (Tables S2 and S3). Before statistical analyses, we tested data for normal distribution using a Shapiro-Wilk test. When possible, non-normally distributed data were natural log or square root transformed to reach a normal distribution. To examine the rela-20 tionships among watershed land uses, water chemistry, sediment characteristics, plant community structure, sediment denitrification and N 2 O production at the lake level, we performed Pearson correlation and regression analyses using PASW Statistics 18 software (IBM SPSS Inc., Chicago, USA Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | from our correlation analysis to select the promising explanatory variables to include in the SEM models. Given the positive correlations between NO − 3 and TN (r = 0.86, P < 0.01) and between NO − 3 and NH + 4 (r = 0.47, P < 0.05), we only included the NO − 3 in the SEM analyse to simplify the final models. The software Amos 20 (IBM SPSS Inc., Chicago, USA) was used to design the SEM models and calculate path coeffi-5 cients, squared multiple correlations, direct and indirect effects. Indirect effects, i.e., effects mediated by other variables, were calculated by multiplying the standardised path coefficients (i.e., estimates) of the direct effects of the variables involved in the total pathway. We used the chi-square (χ 2 ) test and the comparative fit index (CFI) to evaluate the overall fit of the SEM models. According to Kline (1998), a good fit was 10 indicated by an insignificant χ 2 statistic (P > 0.05) and a CFI value > 0.90. The χ 2 provides an estimate of how closely the proposed SEM model matches the structure of the actual data. Thus, an insignificant χ 2 statistic (P > 0.05) indicates that the structure of the proposed model is no different than the structure of the data.

Within-lake characteristics and watershed land uses
The highest TC and TN concentrations in water were found in Lake Makouhu (31.40 mg L −1 ) and Lake Yiaihu (2.87 mg L −1 ), respectively (Table S2). The mean TC and TN contents in sediments were 25.18 and 0.51 mg g −1 , respectively (Table 1).
Submerged plants were only found in 5 lakes, with maximum biomass (2438 g m −2 ) 20 recorded in Lake Haikouhu (Table S2). The percentage of HDL in watersheds varied from 33.60 to 80 % with a mean of 60.91 % (Table 1). Among water quality and sediment characteristics, DO, NO − 3 , NH + 4 , TN and STN were significantly correlated with the percentage of HDL (Table 2). There was no significant relationship between plant community structure and watershed land 25 uses (

Denitrification and N 2 O production
Potential denitrification rates varied widely across the 20 lakes and ranged from 11.24 ng N g −1 h −1 in Lake Chaohu to 125.37 ng N g −1 h −1 in Lake Makouhu (Table S3). Background denitrification rates were lowest in Lake Huangdahu at 0.67 µg N g −1 h −1 , and highest in Lake Linghu at 18.95 ng N g −1 h −1 . N 2 O production rates ranged from 5 0.05 to 0.71 ng N g −1 h −1 , with the largest value recorded in Lake Caizihu where the relative N 2 O production was 0.13 (Table S3). There was no significant relationship between potential denitrification rate and background denitrification rates, N 2 O production rate or relative N 2 O production (figures not shown). N 2 O production rates increased with increasing background denitrification 10 rates (R 2 = 0.61, P < 0.01; Fig. 2).

Relationship of denitrification and N 2 O production to within-lake characteristics
Denitrification and N 2 O production rates were found to be significantly related to several water quality parameters (Table 3). However, sediment characteristics and plant 15 community structure had no significant relationships with denitrification and N 2 O production rates. Potential denitrification rate was negatively correlated with DO (Table 3). We determined that both background denitrification rate and N 2 O production rate were enhanced at higher NO − 3 and TN concentrations and lower ORP and DO concentrations (Table 3).

Relationship of denitrification and N 2 O production to watershed land uses
Watershed land uses were not significantly related to potential denitrification rates or relative N 2 O production (figures not shown). However, background denitrification rate was found to be positively correlated with the percentage of HDL in watersheds (R 2 = 0.29, P < 0.05; Fig. 3b). We also found that N 2 O production rate was significantly related to the percentage of water bodies in watersheds (R 2 = 0.27, P < 0.05; Fig. 3a).

Indirect effects of human land uses on denitrification and N 2 O production
The results of the SEM analysis indicated that the fit of the data to the models was acceptable (Fig. 4). In the potential denitrification model (χ 2 = 3.063, df = 4, P = 0.547, 5 CFI = 1.000), the total indirect effect of watershed HDL on potential denitrification (0.26) was small and statistically insignificant (Table 4, Fig. 4a). In the background denitrification model (χ 2 = 1.281, df = 4, P = 0.865, CFI = 1.000), 100 % of the positive indirect effect of HDL on background denitrification (0.55) was mediated through water quality (principally via NO − 3 ; Table 4, Fig. 4b). In both the N 2 O production model (χ 2 = 1.920, 10 df = 4, P = 0.750, CFI = 1.000) and relative N 2 O production model (χ 2 = 2.483, df = 4, P = 0.648, CFI = 1.000), HDL indirect effects were largely transmitted via water quality (Table 4). 15 Numerous studies have revealed that human modification of the land uses plays a key role in determining the water quality of adjacent aquatic ecosystems (e.g., generation of point and non-point source pollutants and expansion of impervious surface area) (Crosbie and Chow-Fraser, 1999; Arbuckle and Downing, 2001;Taranu and Gregory-Eaves, 2008). In this study, human land use in watersheds was strongly and positively 20 related to water N concentrations in lakes (  (Shang and Shang, 2005). Cropland is recognised as the most important source of non-point source pollution affecting lakes in many countries. Approximately half of the area of the middle and lower Yangtze River basin is covered by agriculture land, especially paddy fields (Liu et al., 2011). In the Yangtze River basin, the annual consumption of nitrogen fertiliser on agricultural lands  (Müller et al., 1998;Bruesewitz et al., 2011). Bruesewitz et al. (2011) found that the percentage of catchment agriculture was positively related to sediment carbon but not nitrogen. These researchers found that agricultural landscapes can increase carbon inputs to aquatic ecosystems as a result of increased soil erosion and organic carbon transport 15 or through increased lake productivity resulting in higher rates of organic matter deposition in sediments. However, Müller et al. (1998) indicated that TN concentration in lake sediments strongly increased with increasing agricultural land use in watersheds. Our study also found a significant relationship between watershed human landscapes and STN. It is noteworthy that the correlation between human landscapes and water 20 TN was greater than that between human land uses and STN (Table 2). One possible explanation is that water TN level in lakes primarily reflects recent land use patterns in watersheds. Sediment TN content, however, reflects sedimentation and deposition of nutrients that have occurred over several decades. Any modelling of water and sediment nutrient contents using recent land use data is likely to show stronger correlations Introduction

The role of sediment denitrification in nitrogen removal in Yangtze lakes
Sediment denitrification is recognised as the most important pathway of nitrogen removal in lakes, followed by nitrogen sedimentation and uptake by aquatic macrophytes (Saunders and Kalff, 2001;Lindau et al., 2009). Scaled on an areal basis using the bulk density of the top 7 cm of sediment (Liu et al., 2015), background denitrification 5 rate ranged from 354 to 10 968 kg N km −2 year −1 , with a mean of 1065 kg N km −2 year −1 .
This mean rate is smaller than the global average of lakes (3421 kg N km −2 year −1 ) estimated by a NiRReLa model (Harrison et al., 2009). Background denitrification rate can be considered as a conservative estimate of actual in situ denitrification. Therefore, according to the areal background denitrification rate and the surface area of the 20 10 studied Yangtze lakes (Table S1), we find that Lake Chaohu and Lake Caizihu can at least remove 601 880 and 553 812 kg of N every year, respectively. It is hard to quantify total nitrogen inputs to a lake. To evaluate the proportion of annual nitrogen inputs that are removed via denitrification, Seitzinger et al. (2006) established a relationship between the percentages of N removed by denitrification and 15 water residence time (WRT), detailed below: % N removed by denitrification = 23.4(WRT) 0.204 (1) WRT for most of the lakes we sampled are unknown. Because almost all lakes in the Yangtze floodplain are shallow (mean depth < 5 m), the WRT of these lakes is estimated to be between 1 and 12 months (Wang and Dou, 1998 by McCrackin and Elser (2010). It should be noted that N 2 O production in the absence of acetylene can be derived from both denitrification and nitrification processes, especially under incomplete anaerobic conditions (García-Ruiz et al., 1998;Xu et al., 2008). In the present study, for N 2 O production assays, sediments were slurried with unfiltered lake water, which was always saturated with oxygen. 5 The present study found that the average of relative N 2 O production was 0.17, similar to the value (0.18) reported by García-Ruiz et al. (1998). However, our results contrast with those found in other investigations, the majority of which give values less than 0.05 (e.g., Seitzinger, 1988;Beaulieu et al., 2011). This result might be explained by the fact that oxygen in overlying water of shallow lakes could suppress the N 2 O reductase 10 enzyme, which reduces N 2 O to N 2 during denitrification (García-Ruiz et al., 1998). It is known that this enzyme is the most oxygen sensitive of all enzymes involved in denitrification process, and slurries in the present study were made with unfiltered lake water. The relative N 2 O production > 1 implies that the production of N 2 O through nitrification must have occurred, and nitrification processes is the major source of N 2 O 15 (Xu et al., 2008).

Local and regional determinants of sediment denitrification in lakes
Environmental variables that affect sediment denitrification can be categorised as proximal or distal regulators (Saggar et al., 2013). Proximal regulators, including NO − 3 concentration, oxygen supply or water content, carbon availability and temperature in sed-20 iment and overlying water, affect the instantaneous rate of sediment denitrification. Consistent with previous studies conducted in streams and rivers (García-Ruiz et al., 1998;Inwood et al., 2005), we found that only overlying water quality had significant relationships with sediment denitrification and N 2 O production in Yangtze lakes. Our study suggests that nitrogen availability in water is the primary factor limiting sedi-25 ment denitrification in Yangtze lakes. NO − 3 concentration in water explained approximately 70 % of the variability in denitrification rates in a meta-analysis (Piña-Ochoa and Álvarez-Cobelas, 2006). Some studies also reported a significant relationship be-7829 Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | tween denitrification and sediment carbon concentration (e.g., Bruesewitz et al., 2011). However, this relationship is not significant in the Yangtze lakes, most likely because sediment TC concentration was high (25.18 mg g −1 ) and was not limiting to sediment denitrification. Distal regulators affect sediment denitrification indirectly by acting on the proximal 5 controls (Saggar et al., 2013). These regulators include factors such as land use and climate. The indirect effects of human land uses on stream sediment denitrification and N 2 O emission are mainly mediated through stream water quality or sediment characteristics (Inwood et al., 2007). Only one study has examined the effects of watershed human land uses on sediment denitrification in lakes (Bruesewitz et al., 2011). This 10 study found that potential denitrification but not background denitrification was positively related to the proportion of catchment agriculture. However, consistent with previous observations regarding stream denitrification (e.g., Arango and Tank, 2008), we found a significant positive relationship between background denitrification and human land uses. As we hypothesised, the human land use in watersheds affects the sedi-15 ment denitrification and N 2 O production in Yangtze lakes indirectly, mainly via effects on lake water quality. The positive relationships between watershed human land uses and sediment denitrification were primarily driven by nitrogen loading.

Conclusions
Increased human land uses in watersheds have led to elevated nitrogen concentration 20 in both water and sediments in Yangtze lakes. Among the water quality parameters, ORP, DO, NO − 3 and TN were significantly related to sediment background denitrification and N 2 O production rates. Only the background denitrification rate was found to be positively correlated with the percentage of human land uses in the watersheds. The results of SEM analysis supported the hypothesis that sediment denitrification and N 2 O  Fish. Aquat. Sci., 56, 1781-1791, 1999  Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Liu, W., Zhang, Q., and Liu, G.: Influences of watershed landscape composition and configuration on lake water quality in the Yangtze River basin of China, Hydrol. Process., 26, 570-578, 2012. Liu, W., Wang, Z., Zhang, Q., Cheng, X., Lu, J., and Liu, G.: Sediment denitrification and nitrous oxide production in Chinese plateau lakes with varying watershed land uses, Biogeochem- Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Yu, K., Seo, D. C., and DeLanue, R. D.: Incomplete acetylene inhibition of nitrous oxide reduction in potential denitrification assay as revealed by using 15N -nitrate tracer, Commun. Soil Sci. Plan., 41, 2201-221, 2010. Zhong, J., Fan, C., Liu, G., Zhang, L., Shang, J., andGu, X.: Seasonal variation of potential denitrification rates of surface sediment from Meiliang Bay, Taihu Lake, China, J. Environ. 12,2015 Human land uses enhance sediment denitrification    Ln-Background denitrification rate (ng N g -1 h -1 ) Sqrt-N2O production rate (ng N g