Estimating N₂O emissions from soils under natural vegetation in China

Model description

DyN-LPJ (Xu-Ri and Prentice 2008) is based on the Lund–Potsdam–Jena (LPJ) dynamic global vegetation modelling framework (Sitch et al. 2003) (Fig. 1). In addition to the coupled carbon and water cycling and vegetation dynamics simulated in LPJ, DyN includes following sub-models: (1) plant N uptake, (2) plant N allocation, turnover, reproduction, and mortality, (3) plant and soil N mineralization, (4) biological N₂-fixation, (5) nitrification, (6) NH₃ volatilization, (7) nitrate leaching, (8) denitrification, and (9) N₂, N₂O and NO production and emission. Sub-model (2) uses an annual time step; the others use a daily time step. The model runs at a spatial resolution of 0.5° × 0.5° (Xu-Ri and Prentice 2008). N₂O fluxes were modelled as byproducts of nitrification and denitrification. Nitrification is assumed to occur in aerobic microsites within the top 50 cm of soil, while denitrification is assumed to occur in anaerobic microsites within the same soil layer. Soil moisture (water filled pore space, WFPS) was used as an indicator to allocate substrates into two soil fractions with different aeration status. Nitrification and denitrification rates were then calculated directly from substrate availability (NH₄⁺ for nitrification; labile C, NO₃⁻ and NO₂⁻ for denitrification) and soil environmental factors (soil temperature and WFPS). The rate of each transformation was regulated by the substrate concentrations (NH₄⁺, labile C, NO₃⁻ or NO₂⁻) following Michaelis–Menten kinetics and soil temperature. The detailed functions and parameters for nitrification and denitrification are documented in Xu-Ri and Prentice (2008). We applied the maximum fraction of N₂O losses from nitrification and denitrification as 0.05% (R_N2ON) and 1.8%(R_N2ODN), respectively (Xu-Ri et al. 2012).

Open image in new window

Model evaluation

Twenty-eight sets of measurements, covering 18 sites were used in this study for model evaluation (Fig. 2). As shown in Table 1, the first 5 sites have been used in a previous global-scale evaluation (Xu-Ri et al. 2012), the additional 23 sets of measurements of annual N₂O emissions (covering 13 sites), conducted during the period 2003–2006, were used in this study for model calibration for grassland, forest, desert and alpine steppe ecosystems in China (Table 1, Fig. 2). These measurements were all conducted in the field, using closed chamber techniques and gas chromatography. Three gas flux measurement locations within each of the above sites were established randomly and N₂O fluxes were typically measured at the same time period (usually mid-morning, 10:00–12:00 h) of each sampling day and 1–2 times per week during the growing season and 1–2 times per month during winter.

Open image in new window

Table 1

Observed annual N₂O emissions from China for comparison with the model results

No.	Long.	Lat.	Vegetation type	Year observed	N2O fluxes (uncorrected data) kg N ha−1 yr.−1 ± SD	N2O fluxes (corrected data1) kg N ha−1 yr.−1 ± SD	Location	Sources
Soils under natural vegetation, used for model calibration in Xu-Ri et al. (2012)
1*	93	35	Alpine steppe	2000		0.07	Wudaoliang, Tibet	(Pei 2003)
2*	116.5	39.5	Temperate forest	1997–1998		0.28	Beijing	(Sun and Xu 2001)
3*	127	41.5	Alpine tundra	1994–1995		0.28	ChangBai Mountain	(Chen et al. 2000)
4*	116.0	43.5	Temperate steppe	1995		0.27	Inner Mongolia	(Chen et al. 2000)
5*	116.5	43.5	Temperate steppe	1998		0.37	Inner Mongolia	(Xu-Ri et al. 2003)
Soils under natural vegetation
6	101.5	37.5	Alpine meadow	2003	2.55 ± 1.91	0.23 ± 0.31	Qinghai, Haibei	Data provided in this study
7	101.5	37.5		2004	1.84 ± 1.71	0.16 ± 0.29
8	101.5	37.5		2005	2.67 ± 2.76	0.92 ± 1.03
9	101.5	38	Alpine shrub	2003	2.37 ± 1.25	0.70 ± 0.47
10	101.5	38		2004	1.90 ± 1.80	0.51 ± 0.46
11	102.5	33	Alpine swamp	2004	0.67 ± 1.15	0.30 ± 0.53	Ruoergai, Sichuan
12	102.5	33		2005	0.73 ± 0.77	0.31 ± 0.45
13	87.5	44.5	Temperate desert	2004	0.83 ± 1.68	0.31 ± 1.24	Fukang, Xinjiang
14	87.5	44.5		2005	0.25 ± 1.66	0.05 ± 0.96
15	113	22.5	Subtropical forest	2004	3.14 ± 3.83	1.60 ± 3.26	HeShan, Guangdong
16	113	22.5		2005	3.78 ± 4.20	2.26 ± 3.66
17	102	22	Tropical forest	2003	3.01 ± 2.66	1.22 ± 0.90	Xishuang banna, Yunnan
18	102	22		2004	1.82 ± 1.47	0.84 ± 0.92
19	102	22		2005	1.61 ± 1.88	1.05 ± 0.92
20	112.5	23	Tropical forest	2003	3.46 ± 2.67	1.51 ± 1.20	Dinghushan, Guangdong
21	112.5	23		2004	3.80 ± 2.86	1.39 ± 0.89
22	112.5	23		2005	4.72 ± 2.70	2.06 ± 1.06
23	90.5	30.5	Alpine steppe	2009		0.01 ± 0.02	Namco, Tibet	(Wei et al. 2012)
Background emissions from croplands
24*	123.5	41.5	Maize–spring wheat	2004–2005		1.12	Shenyang Liaoning	(Gu et al. 2007) and subsequently corrected by (Gu et al. 2009)
25*	114.5	35	Winter wheat–maize	2004–2006		0.67	Fengqiu Henan
26*	119	32	Winter wheat–rice	2004–2006		1.00	Jiangdu Jiangsu
27*	120.5	31.5	Winter wheat–rice	2002–2004		1.70	Wuxi, Jiangsu
28*	105	31	Winter wheat–rice	2005–2006		1.28	Yanting Sichuan

*the value provided for site 1–5 and site 24–28 are directly cited from the original literature, SD values were not available

The N₂O fluxes were determined using opaque static chambers with subsequent analysis on a gas chromatograph (GC) fitted with an electron capture detector (ECD). In studies 6–22 and 24–28 the carrier gas used was nitrogen gas (N₂). However, Zheng et al. (2008) showed that only using N₂ as carrier gas can overestimate the N₂O concentration. The reason is an interference of the presence of large concentrations of carbon dioxide (CO₂) in the gas samples collected. To combat this problem ascarite was added to the carrier gas stream to remove the CO₂ from the air samples. By comparing the N₂ only (the DN method) with the N₂-Ascarite method, Zheng et al. (2008) proposed correction terms for N₂O flux measurements. Using the correction terms, we adjusted the N₂O fluxes for studies No. 6–22 (Table 1). Emissions in study 24–28 had already been corrected by Gu et al. (2009), using the same method, and studies 1–5 and 23 did not require any correction.

For comparison, annual N₂O emissions were simulated at the geographic location of each site following the transient simulation protocol described below. The simulated results were compared with the specific grid cell location and specific year of each measurement. The simulated cropland background N₂O emissions represent the potential natural soil N₂O emissions in cropland areas.

Historic simulations for China

The transient model runs were set up using identical parameter values, model protocols and forcing to the simulations described by Xu-Ri et al. (2012) except that the transient runs were extended to 2009. TS 3.10.1 climate data (http://www.cru.uea.ac.uk/cru/data/hrg/) from the Climatic Research Unit and updated atmospheric CO₂ concentration data from Keeling et al. (2009) were used. For a model spin-up period of 2000 years, we used climate input from 1901 to 1930. Thereafter, the model was run with transient monthly climate data from 1901 to 2009. These transient simulations were performed at both site level and national-scale.

Six historic simulations were performed: (1) all factors including CO₂, climate and N deposition varied throughout the twentieth century, (2) CO₂ and climate varied without N deposition, (3) climate varied but kept CO₂ constant, (4) only temperature varied, (5) only precipitation varied, (6) only cloud cover varied.

In simulation (4), (5) and (6), CO₂ was kept at 296 ppm, while the “fixed” climate variables were held at their 1901–1920 averages. Simulations were conducted for the period of 1901 to 2009. The simulations for the period 1970 to 2009 were analyzed.

Comparison between simulated and observed N₂O fluxes

Our 28 sets of measurements cover 18 sites (Fig. 2), comprising alpine steppe, alpine tundra, alpine meadow, alpine shrubland, alpine swamp, temperate desert, temperate steppe and temperate forest, tropical forest and subtropical forest (Table 1, No. 1–23). The remaining five sites were background emissions from cultivated soils (Table 1, No. 24–28).

Simulated and observed annual N₂O fluxes were highly correlated either with (R² = 0.75, n = 28, p < 0.001) or without (R² = 0.70, n = 28, p < 0.001) N deposition. However, the slope was closer to the 1:1 line when atmospheric N deposition was considered (slope = 0.94, as shown in Fig. 3), especially for the tropical and subtropical forest areas, and the background emissions from cultivated soils, where the N₂O emissions rate was higher than 1 kg N ha⁻¹ yr.⁻¹(Table 1, Fig. 3).

Open image in new window

We compared the observed and simulated monthly soil N₂O emissions (kg N ha⁻¹ month⁻¹) for two temperate grasslands in Inner Mongolia in 1998 and Haibei in 2004; and two tropical forests sites in Heshan in 2004 and DingHushan in year of 2005. The correlation coefficients of these measurement and model results are 0.93 (Inner Mongolia), 0.86 (Haibei), 0.65 (Heshan), and 0.68 (Dinghushan), respectively (Fig. 4). The model generally captured the seasonal variation of soil N₂O emissions in monsoon climate regions of China, which is highly consistent with the precipitation seasonality (Fig. S2).

Open image in new window

N₂O emissions correlated linearly and exponentially with precipitation and temperature on either monthly or annual time scale, respectively (Fig. 5). The linear and exponential correlation at monthly time step was significant (R² = 0.64, R² = 0.45, n = 96, p < 0.01). The exponential fit showed a Q₁₀ (the factor by which the rate increases with a 10 °C rise in temperature) value of 2.99, with a temperature range between 10 °C and 20 °C. Our model simulations of N₂O fluxes correlated well with temperature and precipitation at monthly and annually time scales. Large soil N₂O emissions occurred when the soil moisture and temperature were high. Some of the discrepancies between modeled and simulated N₂O fluxes on the monthly time scale were expected as the observation were conducted at the site scale while our climate data for driving the model are at 0.5° × 0.5° spatial resolution. Additionally, our model might have failed to capture the enhanced soil N₂O emissions during spring thawing events and the soils N₂O uptake during the non-growing season (Fig.4).

Open image in new window

N₂O emissions from natural terrestrial ecosystems in China in the 2000s

N₂O emissions were highest in the humid tropics but limited by low temperatures in temperate and boreal ecosystems, and by low soil moisture in the desert areas of China (Figs. 6a, b). Tropical and subtropical deciduous forests, shrubland, grassland and coniferous forests were predicted to emit 1.18 ± 0.67, 1.07 ± 0.21, 1.27 ± 1.08, and 0.60 ± 0.46 kg N₂O-N ha⁻¹ yr.⁻¹, respectively (Table 2). The N₂O emissions of temperate deciduous forests, shrubland, grassland, deserts, swamps and coniferous forests were predicted to be 0.28 ± 0.21, 0.41 ± 0.38, 0.47 ± 0.34, 0.16 ± 0.22, 0.21 ± 0.18 and 0.18 ± 0.07 kg N₂O-N ha⁻¹ yr.⁻¹, respectively. Alpine grassland and meadow, with a very low emissions rate, were predicted to be 0.01 ± 0.10 kg N₂O-N ha⁻¹ yr.⁻¹. The potential natural soil N₂O emissions from the cropland area were predicted as 0.78 ± 0.47 kg N₂O-N ha⁻¹ yr.⁻¹ (Fig. S3).

Open image in new window

When we aggregated the fluxes for different ecosystems according to their specific distribution maps (Fig. S1), we found that N₂O emissions from shrubland played an important role either in temperate or in subtropical regions, because of their relatively larger distribution area when compared to forest and grassland (Fig. S4, Table 2). The total emissions from shrublands were 0.14 ± 0.07 Tg N, and N₂O emissions from forest and grassland amounted to 0.10 ± 0.06 and 0.09 ± 0.09 Tg N, respectively (Fig. 7, Table 2). The potential background N₂O emissions from cropland areas were 0.13 ± 0.08 Tg N (as shown in Table 2 and Fig. 7). The total amount of N₂O emissions from the soil of natural terrestrial ecosystems in China in the 2000s was estimated to be 0.46 ± 0.30 Tg N with, and 0.35 ± 0.34 Tg N without N deposition (Table 2).

Open image in new window

Responses of N₂O emissions in China to a changing climate and atmospheric N deposition

The total amount of N₂O emissions from natural terrestrial ecosystems in China increased from 0.28 ± 0.03 Tg N in the 1970s (average for 1970–1979) to 0.46 ± 0.03 and 0.35 ± 0.02 Tg N with and without considering N deposition, respectively, in the 2000s (average for 2000–2009). Warming and atmospheric N deposition accounted for 37% (or 0.07 Tg N) and 63% (or 0.11 Tg N) of the increase in N₂O, respectively, while the effects of precipitation, cloud and atmospheric CO₂ concentrations were negligible (Fig. 8).

Open image in new window

Atmopsheric N deposition rates during the period 2000–2010 (2000s) were estimated to be 8.36 ± 8.56 Tg N. The N deposition rates used were an average of 10.19 kg N ha⁻¹ yr.⁻¹and a maximum of 40 kg N ha⁻¹ yr.⁻¹ (Fig. S5). Their distribution patterns and magnitudes were closer to the observation based estimates of 12.89 kg N ha⁻¹ yr.⁻¹ by Lü and Tian (2007). The N₂O increase resulting from atmospheric N deposition was 0.11 ± 0.01 TgN (Figs. 6a, b), with an average emission factor of 1.19% (Fig. S6). Similarly, the increase of 0.17 Tg N from Lü and Tian (2007) might be due to their higher estimated atmospheric N deposition for China.

N₂O emissions versus net primary productivity

Satellite derived net primary productivity (NPP) has been used as an important variable to simulate global soil N₂O emissions (Potter and Klooster 1998; Potter et al. 1996). In this study, we examined the relationships between the modeled N₂O emissions and NPP on both spatial and temporal scales (Fig. 9). For China, spatially, the modeled soil N₂O emissions (g N m⁻² yr.⁻¹) were highly correlated with NPP (kg C m⁻² yr.⁻¹) with a slope of 0.13 (R² = 0.83, p < 0.001). Temporally, the correlations were also significant with a slope of 0.17 (R² = 0.99, p < 0.001).

Open image in new window

N₂O emissions from grassland ecosystems

N₂O emissions from forest ecosystems

Our model estimates of N₂O emissions from forest ecosystems were 0.10 ± 0.06 Tg N and were similar to the emissions of 0.09 ± 0.09 Tg N from grassland soils (Table 3). But, forest ecosystems cover a smaller area of China and per unit area therefore have a higher emission rate of 0.66 ± 0.42 kg N ha⁻¹ yr.⁻¹ compared to grasslands (0.24 ± 0.23 kg N ha⁻¹ yr.⁻¹). Our estimate is consistent with those of Chen et al. (2000), 0.10 Tg N, but much lower than that of Cai (2012), 0.29 ± 0.19 Tg N. The higher estimate of Cai (2012) may be partly because most of the N₂O emissions in China during the 2000s were measured using the DN-method, as detailed in section “Model evaluation”.

This study includes three observation sites in tropical and subtropical forests: Guangdong Heshan, Guangdong DingHuShan, and Yunnan Xishuangbanna (Table 1). The annual precipitation at these sites ranged between 1300 mm and 1900 mm, the annual air temperature between 17.5 °C and 22.0 °C, and annual N₂O fluxes between 0.84 to 2.26 kg N ha⁻¹ yr.⁻¹(Table 1). These values fall within the range of N₂O fluxes 0.88 to 4.36 kg N ha⁻¹ yr.⁻¹ for tropical forests worldwide (Stehfest and Bouwman 2006; Werner et al. 2007; Xu-Ri et al. 2012). Annual N₂O emissions were generally less than 2 kg N ha⁻¹ yr.⁻¹ when annual precipitation was lower than 2000 mm and annual temperature was lower than 22 °C (Xu-Ri et al. 2012) (Fig. S7). However, there are still large uncertainties in N₂O emission estimates for tropical forests in China. For example, emissions ranging from 2.7 to 3.4 kg N ha⁻¹ yr.⁻¹ have been reported based on short-term observations (Zhu et al. 2013). Future long-terms and year-round continuous measurements are needed to reduce these large uncertainties.

N₂O emissions from shrubland

N₂O emissions from the shrubland ecosystems are not well researched globally and have not previously been estimated for China. In this study, after model calibration against forest and grassland ecosystems in China, N₂O emissions from shrubland ecosystems were calculated at 0.14 ± 0.07 Tg N, in which temperate shrublands contributed 0.05 ± 0.05 Tg N (with an average annual N₂O emissions of 0.41 ± 0.38 kg N ha⁻¹ yr.⁻¹) and tropical/subtropical shrubland contributed approximately 0.09 ± 0.02 Tg N (with annual emissions of 1.07 ± 0.21 kg N ha⁻¹ yr.⁻¹) (Table 3). Recent N₂O flux measurements from an alpine shrub meadow in China ranged from 0.18 to 0.27 kg N ha⁻¹ yr.⁻¹ (Fu et al. 2018) and falls within our uncertainty range for temperate shrubland. The values for the temperate shrublands are also roughly consistent with those from European temperate shrublands (0.14 to 0.42 kg N ha⁻¹ yr.⁻¹; Carter et al. 2012).

Background N₂O emissions from cropland

Background N₂O emissions are defined as emission from cultivated soil that has not received nitrogen fertilizer during the current season or year (Gu et al. 2007; Zheng et al. 2004). The background N₂O emissions from croplands may originate from the following nitrogen sources: natural microbial nitrification and denitrification processes from soils driven by natural N fixation; atmospheric N deposition, mineralization of soil organic matter or crop residues (Gu et al. 2007). Bouwman (1996) reported a background emissions rate of 1.0 kg N ha⁻¹ yr.⁻¹, which was based on five sets of field observations conducted in European and American unfertilized grasslands. The spatial variation of background N₂O emissions proved to be highly climate dependent (Gu et al. 2009; Lu et al. 2006).

In the present study, background N₂O emissions from cropland are referring to the potential natural soil N₂O emissions on cropland areas. After calibrating against the site scale observations (Gu et al. 2009), we estimated an average background N₂O emission from cropland of 0.13 ± 0.08 Tg N (Fig. 7, Table 3), with an average flux rate of 0.79 ± 0.47 kg N ha⁻¹ yr.⁻¹ for the period 2000 to 2009. This result falls within the uncertainty range of 0.72–1.66 kg N ha⁻¹ yr.⁻¹ reported by previous studies (Li et al. 2001; Gu et al. 2009) and is close to the IPCC default value of 1 kg N ha⁻¹ yr.⁻¹ (Lu et al. 2006).

N₂O emissions from natural ecosystems in China

Our estimate of average N₂O emissions from natural terrestrial ecosystems using the model DyN-LPJ was 0.46 ± 0.23 Tg N for the period 2000 to 2009. This estimate is comparable with a previous estimate of 0.44 Tg N, calculated from the difference between total terrestrial ecosystem emissions (0.81 Tg N) and those from N fertilized soils (0.37 Tg N) in China (Lu and Tian 2013). Cai (2012), estimated a similar value of 0.42 Tg N for total emissions from terrestrial ecosystems, but only forest and grassland emissions and not background emissions from croplands were included (Table 3). Since the1970s, the average N₂O emissions from natural ecosystems have increased from 0.28 to 0.46 Tg N yr.⁻¹ in the 2000s. These values are similar to even exceed the changes of agricultural fertilizer induced emissions, which were about, 0.10 in the 1980s (Gao et al. 2011), 0.28 Tg N yr.⁻¹ in the 1990s (Lu et al. 2006; Zheng et al. 2004) and 0.31 Tg N yr.⁻¹ in the 2000s, primarily due to global warming and the increasing rates of atmospheric N deposition (Fig. 8).

Uncertainties in estimating N₂O emissions considering land use change

The N₂O emission estimates calculated from the new land area extent during the period of 2000s for grassland, forest and cropland suggest that the increasing N₂O emissions due to increasing grassland and forest areas, might be offset by decreasing background emissions due to decreasing cropland area (Table 4). However, the overall estimates still fall within the uncertainty range of our estimates based on constant land area for various types of ecosystems.

Here we also examined vegetation area differences between various data sources (Table S1). The vegetation area of shrubland used in this study are 2.07 × 10¹² m² based on the VMC (1:4000000), which is closer to the shrubland area of 2.15 × 10¹² m² used in Piao et al. (2009) and 2.09 × 10¹² m² according to the VMC (1:1000000) in Wang et al. (2017). We observed a large discrepancy in natural vegetation area extent between VMC and GLC 2000. Especially for the shrublands area, a large decrease in shrublands is obviously due to a systematic mismatch between two sources, VMC is mainly based on the field survey and the GLC2000 is mainly derived from the remote sensing production. Consequently, the large uncertainties in shrubland N₂O emissions come from land area uncertainties for various ecosystems, while the limited amount of observational data of N₂O emissions also contributed to the uncertainty of model extrapolation.

N₂O emissions and their controls

The observed soil N₂O fluxes were found to be significantly correlated with precipitation and temperature at seasonal and annual time scales (Fig. 5), which is consistent with other field studies (Smith et al. 2018). Nitrous oxide is produced naturally through nitrification and denitrification by soil microorganisms. Nitrification is an aerobic process in which ammonium (NH₄⁺) is oxidized to nitrate (NO₃⁻) and with nitric oxide (NO) and N₂O as byproducts. Denitrification is an anaerobic process in which NO₃⁻ is reduced to the denitrification products N₂O and N₂. In the DyN-LPJ model, soil moisture (represented as water filled pore space, WFPS) is used as an indicator of N₂O production rates by either nitrification or denitrification pathways (Xu-Ri and Prentice 2008). The close correlation between WFPS and N₂O emission rates is supported by many field observations (Bai et al. 2014; Fu et al. 2018; Gütlein et al. 2018). The fraction of N₂O loss from gross denitrification can be 10 times larger than the maximum rate of N₂O production from nitrification processes (Xu-Ri and Prentice 2008). Both nitrification and denitrification process are highly temperature-dependent, with a Q₁₀ value of ~2.0 to 4.0 over the range 15–25 °C. The exponential relationship between temperature and N₂O production has been well documented in field studies (Griffis et al. 2017; Song et al. 2018; Zhou et al. 2018). The above discussed empirical relations were implemented in the DyN-LPJ model (Xu-Ri and Prentice 2008).

Nitrous oxide fluxes were nonlinearly controlled by both precipitation and temperature (Fig. 9, Fig. S7) at the global scale and for China. However, the N₂O fluxes cannot be simply predicted by either of these single factors (Fig. 9, Fig. S7). The net primary productivity of CO₂ (NPP) was distributed similarly under the control of precipitation and temperature on the spatial scale (Fig. 9). These results verify the findings of a previous study in which satellite-derived NPP was used to predict global N₂O emissions (Potter et al. 1996). More recently, a close correlation between ANPP (Aboveground Net Primary Productivity) and N₂O fluxes was also observed in a field study (Zhang et al. 2018); The underlying mechanism is that ecosystem N loss (e.g., N₂O emissions) is affected by the rate of N fixation, while the NPP is one of the good indicators for the ecosystem scale N fixation (Xu-Ri and Prentice 2017). However, the correlation coefficients need to be calibrated locally according to local measurements when the upscaling of N₂O emissions is based on NPP data. On the temporal scale, however, NPP is a less reliable predictor, both globally and for China, because NPP more strongly responds to the increasing CO₂ concentration (Cramer et al. 1999; Xu-Ri et al. 2012).

On a decadal timescale, temperature appeared to be the dominant factor for natural soil N₂O emissions (Fig. 8). This could be due to the fact that in the twentieth century, there was no significant variation in annual precipitation, whereas the air temperature has significantly increased in China (Fig. 8).