Introduction

Nitrous oxide (N2O) is a long-lived (about 120 years) trace gas that has 310 times the ability of greenhouse effect in the atmosphere compared to carbon dioxide (CO2)1,2. Nitrous oxide accounted for 7.9% of the global anthropogenic greenhouse gas emissions in 20042. It has also been the most dominant ozone-depleting substance since 20093,4. The atmospheric N2O concentration has increased by about 120% since 1750, from around 270 ppb, to 324 ppb in 20114. Human activities are responsible for 40–50% of the annual increase in N2O emissions over its pre-industrial levels, with about 57% of the global atmospheric sources of N2O are estimated to be related to emissions from soils5. Thus, there is little doubt that terrestrial soils are the most important sources of atmospheric N2O.

Nitrous oxide was mainly produced by the microbial processes of nitrification and denitrification in soils4. Nitrification and denitrification can occur simultaneously in soils, although the relative rates of the two processes depend on soil aeration and micro-site availability of substrates6. Soil N2O emissions from nitrification and denitrification react sensitively to soil water content, soil temperature and available N and C in the soils, as well as some other physical and chemical properties of the soil such as texture and pH7,8. Increased soil N availability as a result of N inputs by atmospheric deposition has greatly increased N2O emissions from soils9. In a meta-analysis, N addition (mainly NH4+ and NO3−), ranging from 10 to 562 kg N ha−1yr−1, significantly increased N2O emission by an average of 216% across various ecosystems10.

Tropical forests currently represents the largest terrestrial sink for anthropogenic CO2 emissions, accounting for 59% of the global C pools in forests11,12. Recent studies suggested that tropical forests have a key role not only in the C cycle but also in the global atmospheric balance of N2O7,13,14. IPCC2 in 2007 estimated that N2O emission in tropical forests was 4.4 Tg N yr−1, accounting for 14 to 23% of all atmospheric N2O sources. A biogeochemical model estimated 0.9–2.4 Tg N yr−1 N2O emission in tropical rainforests7. Recent observations of the N2O atmospheric column suggested that 50–64% of the atmospheric N2O was derived from the tropical zone with large temporal variation15,16. Furthermore, N addition to a tropical montane forest in Panama increased the N losses via nitrate leaching and N2O emission17.

Although our knowledge of sources and sinks of N2O in tropical forests is increasing, there still remain major knowledge gaps in our understanding of N2O emission from tropical forests. Differing in soil nutrient status from temperate forests which are usually limited by N supply18, tropical forests lying on highly weathered soils are relatively rich in available N and poor in available P19,20. As a result, the productivity is likely limited by P20,21. If input of N exceeded the biotic demands, the ecosystem N retention capacity would be exceeded and gaseous losses of N as N2O would increase22. P-limited forests thus could emit more N2O than the N-limited forest after N fertilization23. Hall and Matson23 observed a 54 times higher soil N2O emission in P-limited site than N-limited site after N addition in two tropical forests. Furthermore, recent studies have shown that anthropogenic N deposition is increasing dramatically in tropical regions24,25 and P deposition may already play an important role as a source of P in tropical regions26.

The increased nutrient deposition may alleviate nutritional constraints on aboveground plants and soil microbial activities20,21,27, which would alter soil N2O emissions. Recent work indicated that P application reduced N2O emissions by reducing soil inorganic N due to increased N uptake by plants in a tropical plantation28 and in a P limited soils8. However, Mori et al.29 found that P addition in a tropical plantation soil increased N2O emission under high soil water content. Some researchers argued that P-enrichment stimulated soil denitrifiers and nitrifiers which would enhance the N2O emission from soils8,30,31. White and Reddy32 found that P addition increased microbial biomass and denitrifying enzyme activity in soil. Moreover, previous study has observed that P addition in tropical forests could alter the composition of soil microbial community27. These results suggested that both aboveground plants and belowground microbes would affect the response of soil N2O emission to increased nutrients availability in tropical forests. However, direct evidence of soil N and P availability on soil N2O emission in tropical forests is still rare33 and most of the studies neglected the interaction of N and P availability on soil N cycling23,28,29.

In this study, we conducted a manipulative experiment with a randomized block design to investigate N and P addition and their interaction on soil N2O emission and N transformations in a secondary tropical forest in southern China. We hypothesized that: (1) N addition would increase soil N2O emission and nitrification due to increased substrates for nitrification and denitrification and (2) P addition would increase soil N2O emission and soil N transformation rates due to stimulated activities of soil microbes from the relief of P limitation in the secondary tropical forest.

Results

Soil available nutrients and N transformations

The general soil properties before fertilization were shown in Table 1. Over the study period, soil NH4+ concentrations were greatly increased by N addition (P < 0.001, Fig. 1A). There was a gradual increase of soil NH4+ with the time of N addition. In Sep. 2010, one year after the beginning of nutrients addition, the NH4+ concentration in N-addition plots (aN and aNP) was 2.52 mg/kg, which was about 20% higher than the value in control plots. This percentage in June 2011 and June 2012 became 65% and 67%, respectively (Fig. 1A). Similar to soil NH4+, soil NO3− was significantly increased by N addition since September 2010 (P < 0.001. Fig. 1B). Soil NO3− concentration under aN and aNP plots increased with the duration of fertilization until June 2011. In June 2011, aN and aNP plots had nearly 3 times higher NO3− values than control and aP plots. After that sampling, a decline of soil NO3− was observed in aN and aNP plots (Fig. 1B). Soil available P was greatly enhanced by P addition, while N addition and their interactions has no effect on available P. There was also a gradual increase of soil available P with the duration of P addition. In June 2012, soil available P in aP and aNP plots were approximately 40 mg/kg, which was nearly ten times as much as the values in control and aN (Fig. 1C).

Table 1 Soil physical and chemical characteristics (0–10 cm) in the tropical forest before the start of fertilization (conducted in September 2009)
Full size table
Figure 1
figure 1

Concentrations of soil extractable NH4-N, NO3-N and available P under different treatments in the secondary tropical forest of Xiaoliang station.

Data are illustrated as means; Error bars represent 1 SE, n = 5.

Full size image

Neither soil net nitrification nor net N mineralization rates were significantly affected by nutrients addition or their interactions in the 3 yrs experiment (Fig. 2A&B). However, the seasonal pattern of soil N transformations was obvious. In the dry seasons (Dec. 2010, Mar. 2011, Dec. 2011), soil N transformation rates were always lower than those in the wet season. Furthermore, we observed a high rate of soil nitrification in comparison with N mineralization rate, with the ratio of nitrification to N mineralization was over one. Soil nitrification rate accounted for nearly 88.9% variation of soil N mineralization rate (r = 0.943, P < 0.001).

Figure 2
figure 2

Soil net nitrification and N mineralization rates in the four experiment treatments in the secondary tropical forest of Xiaoliang station.

Data are illustrated as means; Error bars represent 1 SE, n = 5.

Full size image

Soil microbial biomass

The seasonal variation of soil MBC and MBN was very obvious (Fig. 3). In the wet seasons, soil MBC and MBN usually had two times values than the corresponding data in the dry seasons. We also found that the response of soil microbe to N and P addition differed among seasons. P addition significantly increased soil MBC and MBN only in wet seasons (P = 0.02 and P = 0.03, respectively, Fig. 3A&C). However, neither N addition nor N×P interactions significantly affected soil MBC or MBN in either season.

Figure 3
figure 3

The wet season and dry season soil microbial biomass C (MBC) and N (MBN) in the four experiment treatments of the secondary tropical forest of Xiaoliang station.

Error bars represent 1 SE, n = 5.

Full size image

Soil N2O emissions, temperature and Water Filled Pore Space (WFPS)

Nitrogen addition significantly increased soil N2O emission in this tropical forest (P < 0.001, Fig. 4). The mean soil N2O emission in aN and aNP were 60.2 and 69.4 μg N2O-N m−2h−1, which were 39% and 60% higher than that in the control (43.3 μg N2O-N m−2h−1), respectively (Table 2). The lowest N2O emission was observed in aP (41.8 μg N2O-N m−2h−1), which was similar to the control, but significantly lower than aN and aNP (Table 2). Moreover, the temporal variation of soil N2O emission was obvious. Soil N2O emission was higher in wet seasons than that in dry seasons (Table 2). In wet seasons, N addition significantly increased N2O emission (P < 0.001, Table 2) and the interaction of N and P addition marginally significantly affected soil N2O emission (P = 0.092, Table 2). aNP had the highest N2O emission (97.4 μg N2O-N m−2h−1) in wet seasons, which was significantly higher than the second highest - aN (78.8 μg N2O-N m−2h−1). However, in dry seasons, we only observed a marginally significant effect of N addition on N2O emission (P = 0.06, Table 2).

Table 2 Mean soil N2O emission from 2010 to 2012 after N and P additions in the secondary tropical forest of South China (Mean ± SE, n = 5)
Full size table
Figure 4
figure 4

Monthly soil N2O emission, temperature and WFPS in the secondary tropical forests of Xiaoliang station from October 2010 to September 2012.

Full size image

Soil temperature had a clear seasonal pattern with the highest temperature in wet seasons and the lowest in dry seasons (Fig. 4). However, N and P addition neither affected soil temperature nor soil WFPS. Soil WFPS also shows a seasonal variation with higher soil WFPS in wet seasons and lower in dry seasons (Fig. 4). Both soil temperature and WFPS was positively correlated with soil N2O emission in the control plots along the two years duration (Soil temperature vs N2O: R2 = 0.18, P < 0.01; Soil WFPS vs N2O: R2 = 0.26, P < 0.01).

Pathways determining soil N2O emission by N and P addition

Due to the large seasonal difference of the soil microbial and N2O emission variables, we conducted SEM (structural equation model) analyses separately for wet season and dry season data (Fig. 5). Consistent with the ANOVA results (Fig. 1), both of wet and dry season SEM showed that N addition significantly increased soil inorganic N concentrations and P addition significantly increased soil available P (P < 0.05 for all, Fig. 5A&B). In wet seasons, the final SEM adequately fitted the data describing interaction pathways analysis of N and P addition on soil N2O emission and nitrification (Probability level = 0.416, Fig. 5A). The final wet season SEM explained 58% of the variation in soil N2O emission and 21% of the variation in soil nitrification. The increased soil NO3− by N addition significantly increased N2O emission (P < 0.05, Fig. 5A), while soil NH4+ had no direct effect on N2O emission. The increased soil available P after P addition marginally significantly increased soil MBC (P = 0.06, Table S1), which further increased soil N2O emission (P < 0.05, Fig. 5A). In the dry season, P addition did not have clear pathway to affect soil N transformations and the final SEM only explained 7% of the variation in soil N2O emission and 23% of the variation in soil nitrification. Soil NO3− was negatively correlated with nitrification rate in dry season (P < 0.05, Fig. 5B). The relationships between the remaining exogenous and endogenous variables were not significant, but improved the model fit (Table S1).

Figure 5
figure 5

The final structural equation model of N and P addition effects on soil N transformations and N2O emission in the secondary tropical forest in wet seasons (X2 = 16.53; df = 16, Probability level = 0.417; RMSEA = 0.04; AIC = 56.5) and dry seasons (X2 = 11.51; df = 13, Probability level = 0.568; RMSEA = 0.00; AIC = 57.5).

Numbers on arrows are standardized path coefficients (equivalent to correlation coefficients). Width of the arrows indicates the strength of the causal influence. Green arrows indicate significant positive relationships and red arrows indicate significant negative relationships (P < 0.05). Black arrows indicate nonsignificant positive relationships and grey arrows indicate nonsignificant negative relationships (P > 0.05). Grey dashed arrows indicate paths removed to improve model fits (see Methods). Percentages close to Nitrification and N2O flux indicate the variance explained by the model (R2).

Full size image

Discussion

The soil N2O emission in the secondary tropical forest ranged from 28.6 μg N2O-N m−2h−1 in dry season to 54.6 μg N2O-N m−2 h−1 in wet season, with an average N2O emission of 43.3 μg N2O-N m−2h−1. This value is lower than the N2O emission reported by Tang et al.34 for a nearby old-growth tropical forest (84 μg N2O-N m−2 h−1) and by Konda et al.35 for an Acacia mangium plantation in Indonesia (77.01 μg N2O-N m−2 h−1 in wet season), but is higher than the data reported in many African, South American and Australian tropical forests. Castaldi et al.14 reported an average 26.6 μg N2O-N m−2 h−1 emission for a rainforest in Africa. Werner et al.36 found that the N2O emission in a Kenya rainforest was 42.9 μg N2O-N m−2 h−1, which was comparable to our data. Two other studies estimated lower N2O emission than our study, i.e., 24.2 μg N2O-N m−2 h−1 in a moist tropical forest in the Amazon Basin37 and 16.3–24.3 μg N2O-N m−2 h−1 in a lowland tropical rainforest in Australia during wet season38.

The high atmospheric N deposition might explain the relatively high N2O emission data in this secondary tropical forest. South China is a region suffering from serious N deposition due to the rapid industrial growth. Bulk N deposition (NH4+ + NO3−) ranged from 16.2 to 38.2 kg N ha−1yr−1 in 14 tropical and subtropical forests of South China39. There was also an approximately 40 kg N ha−1yr−1 atmospheric wet N deposition (DON+ NH4++ NO3−) in this forest (Wang et al. unpublished). In a nearby old-growth forests with 84 μg N2O-N m−2 h−1 emission34, the bulk N deposition was measured at 38 kg N ha−1yr−1in 199940. Moreover, we have found that N addition have greatly enhanced N2O emissions from soils in this tropical forest (See below).

In the present study, N addition enhanced soil N2O emission by 39–60% in comparison to control. The increased N2O emission following NH4+ or NO3− addition was observed in many N-saturated ecosystem6,33,41,42. In this study, the ratio of net nitrification to net N mineralization was >1, which was a common phenomenon in many P-limited forests33. The value of this ratio suggested that the majority of N that is mineralized is also nitrified and the soil of this tropical forest is not a strong consumer of N (N saturated)33. In N saturated forests, if inputs of N exceeded the biotic demands for N, the ecosystem N retention capacity would be exceeded, with the potential for increased leaching losses of N as NO3− and DON and elevated gaseous losses of N as N2O and N222. Liu and Greaver10 reported that tropical forest emitted more N2O (on average +739%) under N enrichment compared with other ecosystem types. Hall and Matson33 suggested that N2O emissions from P-limited tropical forests would be especially sensitive to N inputs, since they might respond to even small initial N additions with larger losses than predicted by models developed for N-limited temperate forests. In a P-limited Hawaiian forest, N2O emissions increased rapidly and by much larger amounts than in a nearby N-limited forest after N additions23.

Nitrous oxide was mainly produced by the microbial processes of nitrification and denitrification in soils4. The increased soil NH4+ and NO3− after N addition would increase N substrates to nitrification and denitrification processes. However, N addition did not increased soil net nitrification in the present study, which is contrary to our hypothesis. In the final SEM models, no significant relationships were observed between soil net nitrification and soil NH4+ and the final model only explained 21%–23% variation of soil net nitrification in wet and dry seasons. This finding supports Hall and Matson33 who also found that N addition did not change soil net nitrification in a P limited tropical forest. One explanation was that the method we used to quantity nitrification here was net nitrification rate, which was basically calculated from the change of the soil NO3− pool across time by incubation, thereby disregarding consumption of nitrate by assimilation and denitrification. Direct comparisons of net and gross nitrification rates reveal that net nitrification rates can be an order of magnitude lower than gross nitrification rates and, thus, changes in the NO3− pool do not necessarily reflect the total N turnover from NH4+ to NO3−38.

Besides nitrification, denitrification process is also an important source of N2O flux. From a kinetic perspective, denitrification is assumed to be most efficient in soils with prevailing warm and moist conditions7,24. This secondary tropical forest in wet seasons thus has favorable conditions for denitrification to emit considerable amounts of N2O, i.e., low oxygen content due to high water-filled pore-space (WFPS) and temperature high enough to allow microbial activity to occur. In some tropical and subtropical forests, denitrification contributed more than 70% of soil N2O emission43. In the present study, the SEM shows that soil NO3− was significantly associated with soil N2O emission and the model totally explained 58% of the variation of soil N2O emission in wet seasons. It is thus highly possible that in this tropical forest, denitrification greatly contributed to the increased N2O emission in wet seasons.

Although P addition did not greatly affected N2O emission, the interaction of N and P marginally significantly increased N2O fluxes during wet seasons. The result indicated a stimulation of P addition on soil N2O after N addition. Consistent with our results, Mori et al.29 also observed increased N2O emissions after P addition under high soil water content in an N-fixing Acacia mangium plantation6. In aNP plots, substrate N was high even though the increased plant uptake of N (Fig. 1). The stimulated nitrifying and/or denitrifying bacteria by P thus could enhance N2O emission without substrate N limitation. The wet season SEM model also suggested that soil available P increased soil MBC and MBC significantly increased N2O emission. Furthermore, White and Reddy32 found that P enrichments increased soil microbial biomass and the denitrifiying enzyme activity. Since major part of soil N2O production originated from denitrification process under high soil WFPS43, we assumed that increased soil N2O emission as a result of P addition under elevated N soils in wet seasons was mainly through stimulated soil microbial denitrification process in this forest. The increased N2O emission in wet season of 2011 also coincided with the decline of NO3− between June 2011 and December 2011 (Fig. 1B), which could support above assumption.

To date, most work has focused on the role of N status in N2O emission23, with very little attention on how P availability may interact with plant and soil microbes on N2O emission8. The potential influence of P availability and N×P interaction on the tropical N2O emission remains largely ignored. Increased P availability would increase NPP (net primary productivity) and biomass C sequestration in tropical forests44, which could mitigate the global warming effects28. In the secondary tropical forest, our vegetation investigation data also support this assumption: from 2010 to 2013, aP had the highest tree growth rate, followed by aNP, aN and control (Wang et al. unpublished data). In this study, P addition thus could increase biomass C sequestration and has no significant effect on soil N2O emission under ambient atmosphere N deposition. However, under expectedly elevated N deposition conditions in the future, P addition will enhance the emission of N2O in wet seasons. Some of the potential mechanisms behind our results remain unclear, but the implications are not. Given the dramatically increasing of anthropogenic N deposition in tropical regions24,25, P addition in this tropical soils will stimulate soil microbial activities in wet seasons, which will further enhance soil N2O emission and would offset the plant biomass C sequestration.

Methods

Site description

This study was conducted at the Xiaoliang Tropical Coastal Ecosystem Research Station of the Chinese Academy of Science (CAS) (21°27′N, 110°54′E), southwest of Guangdong Province, China. This region experiences a tropical monsoon climate, with mean annual temperature of 23°C. The annual rainfall ranges from 1400 to 1700 mm with a distinct variation of dry and wet seasons. The wet season is from April to October and the dry season is from November to March. The soil is a latosol developed from granite45.

Our experimental site was a secondary mixed forest. The forest started as Eucalyptus exserta plantation in 1959, then 312 native tress species were introduced in 1960s45,46. Now, the most common tree species are: Castanopsis fissa, Cinnamomum camphora, Carallia brachiata, Aphanamixis polystachya, Ternstroemia pseudoverticillata, Acacia auriculaiformis, Cassia siamea, Albizia procera, Albizia odoratissima, Leucaena leucocephala, Aquilaria sinensis and Chakrasia tabularis. The forest is considered a typical tropical secondary forest with regards to biodiversity and structure complexity of the forest community47.

Experimental design

An N and P addition experiment was designed as a randomized complete block (n = 5) and established in September 2009. Each block was located in a site more than 50 meters apart in the forest. Within each block, four 10 × 10 m plots were established and each plot was surrounded by a 2-m-wide buffer strip in each site. The treatments, N addition (aN), P addition (aP), N and P co-addition (aNP) and control (no addition of mineral nutrients) were assigned randomly to the four plots within each block. Both N and P were applied at 100 kg ha−1yr−1. Briefly, 476.6 g NH4NO3 (equal to 166.6 g N) and/or 808 g NaH2PO4 (equal to 166.6 g P) were dissolved in 30 L groundwater and, then, applied to the corresponding plots near soil surface using a backpack sprayer in each two months since September 2009 for three years. Thirty liters of groundwater was applied to control plots in each treatment event.

Soil N transformations

An in situ soil-core technique48,49 was used to estimate soil net nitrogen mineralization in September 2010 (12 months), December 2010 (15 months), March 2011 (18 months), September 2011 (24 months) and July 2012 (34 months). Briefly, in each replicate plot, 4 points were randomly located. In each of these points, two PVC (polyvinyl chloride) tubes of 4.6 cm in diameter and 15 cm in height were hammered into the soil to a depth of 10 cm. Before sampling, forest floor litter was removed. One of the two tubes from each subplot was retrieved and sent to the lab. The other tube, with a lid on the top and some holes on the sidewall for aeration, was retained in situ for one month, 30 days, before being retrieved.

All soil cores were transported to the lab immediately and stored at 4°C and extracted for mineral N within 48 hrs. Before extraction, each of the four cores from the same plot was manually mixed thoroughly. Visible roots and stones were removed manually. Twenty grams of fresh soil from each layer were extracted with 100 ml of 2M KCl solution (1:5 ratio) and filtered (Shuangquan quantitative filter paper 202#). Concentrations of ammonium and nitrate in the extraction solution were determined by a flow injection autoanalyzer (FIA) (Lachat Instruments, USA), ammonium by the salicylate-nitroprusside method and nitrate by sulfanilamide colorimetry after the Cd-core reduction to nitrite. Soil moisture was determined by weight loss after oven drying at 105°C for 24 hrs. Bulk density (dry soil) was calculated based on soil weight in all tubes and soil moisture. Net N mineralization was calculated as the increase in ammonium plus nitrate N between the initial soil sample and the incubated sample, while net nitrification was the increase in nitrate and net ammonification was the increase in ammonia.

Soil general properties

Soil chemical properties (i.e. soil pH, organic matter, total N, total P,) were determined using the soil samples obtained in September 2009 (Table 1), before the fertilization started. All soil samples were air-dried and passed through a 2 mm sieve. Soil pH was measured in a 1:2.5 mixture of soil:deionized water. Soil available P was extracted with Bray-2 solution48 and determined by the molybdate blue colorimetric method. Soils for analyses of total N (TN) and organic matter were grounded to pass through a sieve of 60 mesh. Total N concentration was determined by micro-Kjeldahl digestion followed by salicylate-nitroprusside colorimetric determination on the Lachat FIA. The measured total N thus only included organic N and ammonium. Soil organic C (SOC) was determined by the wet combustion method50, with SOM calculated as with Van Bemmelen's factor, following the guidelines of Liu, et al.51.

Soil microbial biomass C and N (MBC and MBN) were determined by the chloroform fumigation extraction method52 in the March, September, December 2011 and June 2012.

N2O flux measurement

Nitrous oxide fluxes were measured monthly, from October 2010, one year after nutrients addition started. Gas fluxes were monitored once every month using the static chamber and a gas chromatograph (Agilent 4890D). The static chamber was a 25 cm diameter by 16 cm tall PVC pipe permanently anchored 8 cm into the soil. During gas collection, a 30 cm tall removable cover chamber was attached tightly to the anchor ring with a rubber band. Gas samples were collected from each chamber from 9:00–10:00 local time. Gas samples were taken with a 60 ml plastic syringe at 0 min, 15 min and 30 min after the chamber closure. Before each sampling, syringes were flushed three times with chamber gas to mix the headspace. Laboratory tests showed that chambers and syringes were inert to N2O53,54. Gas samples were analyzed within 12 h in a gas chromatograph (Agilent 4890D) fitted with an electron capture detector (ECD) for N2O. Calibration gases (N2O at 321ppbv, bottle's No. 070811) were obtained from the Institute of Atmospheric Physics, Chinese Academy of Sciences.

The calculation of N2O flux followed that described in Zhang et al.55, based on a linear regression of chamber gas concentration versus time. Atmospheric pressure was measured at the sampling site using an air pressure gauge (Model THOMMEN 2000, Switzerland). Air temperature (enclosure), soil temperature (at 5 cm depth) and moisture (0–10 cm depth) were measured during each sampling. Soil moisture content was quantified using a TDR-probe (Model Top TZS-I, China). Soil moisture (0–10 cm depth) values were converted to WFPS (Water Filled Pore Space) according to the following formula:

Where SBD is soil bulk density, Vol is volumetric water moisture and 2.65 is the density of quartz.

Statistical analyses

Repeated measures ANOVA was used to examine the effect of N and P additions on soil N2O fluxes, soil temperature and WFPS from October 2010 to September 2012, Soil NH4+, NO3−, available P and soil N transformations were also analyzed by the repeated measures ANOVA. Soil microbial biomass C and N were separately analyzed in wet seasons and dry seasons.

Structural equation modelling (SEM) was performed to analyze different hypothetical pathways that may explain soil N2O emission and N transformations in wet seasons and dry seasons (Fig. S1). Because of a positive correlation between net N mineralization and nitrification, we only used net nitrification rate in the SEM analysis. We used soil NH4+, NO3−, available P as the soil abiotic variables and soil MBC as soil microbial variables. In the SEM analysis, data were fit to the models using the maximum likelihood estimation method. Adequacy of the models was determined using X2 tests, Akaike Information Criteria (AIC) and root square mean errors of approximation (RMSEA). Adequate model fits are indicated by a nonsignificant X2 test (P > 0.05), low AIC and low RMSEA (<0.05). We improved the adequacy of the model by removing relationships between observed variables in the prior models based on Modification Indices. Repeated-measures ANOVA were performed in SPSS 18.0 (SPSS Inc., USA) and SEM was performed in Amos 21.0 (SPSS Inc., USA).