Comparing watershed black locust afforestation and natural revegetation impacts on soil nitrogen on the Loess Plateau of China

This study examined a pair of neighbouring small watersheds with contrasting vegetations: artificial forestland and natural grassland. Since 1954, afforestation which mainly planted with black locust has been conducted in one of these watersheds and natural revegetation in the other. The differences in soil total N, nitrate, ammonium, foliar litterfall δ15N and dual stable isotopes of δ15N and δ18O in soil nitrate were investigated in the two ecosystems. Results showed that there was no significant difference in soil total N storage between the two ecosystems, but the black locust forestland presented higher soil nitrate than the grassland. Moreover, the foliar litterfall N content and δ15N of the forestland were significant higher than the grassland. These results indicate that 60 years of watershed black locust afforestation have increased soil N availability. The higher nitrate in the forestland was attributed to the biological N fixation of black locust and difference in ecosystem hydrology. The dual stable isotopes of δ15N and δ18O revealed that the two ecosystems had different sources of soil nitrate. The soil nitrate in the forestland was likely derived from soil N nitrification, while the soil nitrate in the grassland was probably derived from the legacy of NO3− fertiliser.

areas could lead to progressive N limitation. Drylands occupy 47 percent of the surface of the earth 20 . However, information on the effects of the natural and artificial management practices on soil N in drylands is still limited. Moreover, few studies have used the dual stable isotopes of δ 15 N and δ 18 O to characterise the changes in soil N cycling during drylands afforestation or natural revegetation.
Recently, several studies have demonstrated that foliar δ 15 N is a useful indicator of ecosystem N availability [21][22][23] . The basic premise is that δ 15 N measured in organic material characterises the fractionation processes 24,25 . When the N supply is high relative to biotic demand, N is lost through fractionating pathways; the remaining ecosystem N is enriched in 15 N 26 . Therefore, ecosystems with high N availability typically exhibit high δ 15 N values in plant tissues 27 . Moreover, dual stable isotopes (δ 15 N and δ 18 O) of nitrate have been used successfully to identify the sources and transformation processes of nitrate in water and terrestrial ecosystems [28][29][30][31] . In many cases, the dual stable isotopes offer a direct means of source identification because different sources of nitrate often have different isotopic compositions 32 . For example, nitrate and ammonium fertilisers, animal and human wastes, and soils have distinct δ 15 N and δ 18 O levels, which can be used to distinguish the sources of nitrate 32 .
The Loess Plateau of China is a unique geographical area that is characterised by an extensive loess distribution, severe soil erosion and low vegetation coverage. Since the 1950s, the Chinese government has made great efforts to control soil erosion and restore vegetation, including implementing large-scale tree plantation in the 1970s, integrated soil erosion control in the 1980s and 1990s and the Grain for Green Project in the 2000s 33,34 . The most recent study by Lü et al. 35 demonstrated that a total of 8.69 × 10 5 ha of cropland was converted to forestland on the Loess Plateau between 2000 and 2008. This extensive afforestation has been reported to decrease regional water yield and deplete soil water resources 35,36 . However, it is unknown whether several decades of afforestation will lead to more serious N limitation along with soil water deficits. The resulting information will greatly assist in understanding the effects of dryland afforestation on soil N cycling.
In this study, a pair of neighbouring small watersheds with similar topographical and geological backgrounds on the Loess Plateau were selected and used to compare the effects of artificial affestatation and natural revegetation on soil N storage and availability (Fig. 1). Since 1954, afforestation which mainly planted with black locust has been conducted in one of these watersheds and natural revegetation in the other. The two watersheds have formed completely different ecosystems: black locust forestland and natural grassland. In a previous study, we found that the two ecosystems have different patterns of soil carbon cycling and the grassland is more beneficial to soil surface organic and inorganic carbon sequestration than the forestland 37 . The objectives of this study were (1) to examine the difference in soil N storage and availability between the two ecosystems, and (2) to use the dual stable isotopes of δ 15 N and δ 18 O to identify the soil nitrate source.

Results
Soil physical and chemical properties in the two ecosystems. Significant differences were identified in the physical and chemical properties of the soils in the two ecosystems ( Table 1). The black locust forestland had higher soil bulk density and lower soil organic carbon (SOC), soil moisture and C/N compared with the natural grassland. However, no significant difference between the two ecosystems could be identified with respect to the soil pH and total N concentrations (STN).
Soil total N, nitrate and ammonium content in the two ecosystems. The soil total N content exhibited no significant difference between the forestland and grassland ecosystems (Fig. 2a,b, P = 0.702). The storages of soil total N down to a depth of 1 m were 6.93 and 7.65 Mg ha −1 for the black locust forestland and natural grassland, respectively. The concentrations of soil nitrate were significantly different between the two ecosystems, with the content of soil nitrate being significantly higher in the black locust forestland than in the natural grassland ( Fig. 2d, P = 0.003), with mean values at a depth of 1 m of 3.95 and 0.79 mg kg −1 , respectively. Moreover, the soil ammonium content exhibited no significant difference between the forestland and grassland ecosystems ( Fig. 2c, P = 0.499).
Foliar N content and natural abundance of 15 N in the two ecosystems. In this study, the black locust forestland exhibited a higher foliar litterfall N content and δ 15 N than the natural grassland ( Fig. 3, P < 0.0001). The mean foliar N content and δ 15 N in the black locust forestland were 3.31% and −1.14‰, respectively, and 0.84% and − 6.20‰, respectively, for the natural grassland.
Natural abundance of soil nitrate 15 N and 18 O in the two ecosystems. In this study, soil nitrate δ 15 N and δ 18 O also exhibited a large difference between the forestland and grassland ecosystems (Fig. 4). The nitrate δ 15 N level ranged from − 1.37‰ to 2.57‰ in the black locust forestland and from − 2.44‰ to 1.96‰ in the natural grassland. The δ 18 O level ranged from 3.05‰ to 13.16‰ in the black locust forestland and from 10.80‰ to 32.61‰ in the natural grassland.

Discussion
In this study, the artificial forestland and natural grassland provides a unique opportunity to examine the effects of long-term dryland afforestation and natural revegetation on soil N storage and availability. Our results showed that the two ecosystems exhibited no significant difference in soil total N storage ( Fig. 2b), while the content of soil nitrate in the black locust forestland was significantly higher than in the natural grassland (Fig. 2d). Davidson et al. 21,38,39 demonstrated that higher nitrate in ecosystems typically indicates excess available N relative to plant demand and higher soil N availability. Rather than decreasing soil N availability, the higher nitrate content of the black locust forestland investigated in this study indicates that 60 years of watershed black locust afforestation have increased soil N availability. Moreover, the results of the foliar litterfall δ 15 N analysis demonstrated that the black locust forestland had a higher foliar N content and δ 15 N level than the natural grassland (Fig. 3), which further proved that the forestland had higher soil N availability than the grassland.
In a previous study, Jiao et al. 6 assessed the ecological success of restoration by afforestation on the northern Loess Plateau and found that afforestation with black locust offered few additional advantages when compared with natural recovery sites. Moreover, Wei et al. 40 showed that the conversion of grasslands to pine woodlands on the northern Loess Plateau reduced soil N availability. However, Liu et al. 41 evaluated the ecological functions of black locust on the Loess Plateau and found that black locust had the potential to improve soil N availability. Besides, Qiu et al. 42 demonstrated that afforestation with black locust in loessial gully region of the Loess Plateau obviously improved soil N levels and the improvements were greater in long-term than middle-term black locust stands. In this study, we conclude that two factors have led to the increased soil N availability in the forestland. One factor is the excess N input due to biological N fixation of black locust. In the forestland, the main planted  45 showed that approximately 10 years after black locust establishment, soil N had already been enriched by black locust N. In this study, the plantation age is approximately 60 years; thus, the forestland N has probably been enriched by black locust N. Therefore, the biological N fixation by black locust could be the dominant factor contributing to the increased soil N availability in the forestland. The second factor is the difference in ecosystem hydrology. Many studies have demonstrated that the two ecosystems exhibit a large difference in hydrology 46,47 . The annual runoff in the forestland was observed to be significantly lower than in the grassland 37,48,49 . Moreover, the soil water content of the forestland was lower than in the grassland, and soil desiccation had clearly occurred in the forestland due to higher forest transpiration and canopy interception 37 ) represents a very small pool while it is more important than the larger organic pool to maintain ecosystem N availability. The δ 15 N of soil nitrate ranges from about − 10 to + 15‰, with most soils having δ 15 N NO3 values in the range of + 2 to + 5‰ 32,50 .
In this study, the δ 15 N and δ 18 O levels in soil nitrate suggested that the two ecosystems were influenced by different sources of nitrate. The nitrate δ 15 N levels in the forestland and grassland were nearly zero. However, the soil nitrate was more enriched with heavy oxygen isotopes in the grassland than in the forestland (Fig. 4). In comparison with the typical nitrate δ 15 N and δ 18 O levels derived or nitrified from various N sources 32 , we found that the soil nitrate of the grassland was more influenced by NO 3 − fertilisers because δ 18 O NO3 values of the grassland   (Fig. 5). However, the soil nitrate of the forestland was more influenced by soil N transformation and nitrification (Fig. 5). In the forestland, most of the excessive N derives from the biological N fixation of black locust; thus, the nitrate δ 15 N and δ 18 O signals represent the characteristics of soil N transformation and nitrification; while in the grassland, the dual stable isotopes are indicative of NO 3 − fertiliser, suggesting that NO 3 − fertilisers had been applied to the grassland or the former farmland.

Materials and Methods
Study site. This study was conducted in the Nanxiaohe Basin, which is located in the Xifeng District of Qingyang city, Gansu province. The region has a semi-arid continental climate; the mean annual temperature and precipitation are 9.3 °C and 556.5 mm, respectively. Approximately 67.3% of the annual precipitation occurs from In the two watersheds, the area of gully slopes occupies more than 65% of the total area of the watersheds 49 and the measures of black locust afforestation and natural revegetation have been mainly conducted in the area. Therefore, the sampling sites were randomly distributed on the gully slopes, which could represent the pure ecosystems of black locust forestland and natural grassland. In the study, soil samples were collected to a depth of 1 m. The soils were sampled at intervals of 10 cm using a hand-held auger (6 cm in diameter), and 10 soil samples were obtained at each site. Accordingly, 140 soil samples were collected in the forestland watershed and 140 in the grassland watershed. To determine the soil nitrate and ammonium contents, bulk density and pH, three soil profiles at a depth of 0-100 cm were established in each of the forestland and grassland watersheds. For the soil bulk density analysis, three replicate samples were collected at intervals of 10 cm for each profile using a soil corer (a stainless steel cylinder with a volume of 100 cm 3 ). Soil samples were collected at the same distance intervals and used to analyse the nitrate and ammonium contents, pH and the dual stable isotopes of nitrate δ 15 N and δ 18 O.
All of the collected soil samples were air-dried in the laboratory; gravel and roots were carefully removed from the soil. The air-dried soil samples were ground in an agate mortar and passed through a 0.15 mm sieve. Soil total N contents were measured through micro-Kjeldahl digestion, followed by distillation and titration 53 . Moreover, SOC was determined using soil samples digested in K 2 Cr 2 O 7 -H 2 SO 4 solution using a heated oil bath, and the organic carbon concentration was subsequently determined via titration 53 . Soil nitrate and ammonium were extracted with a 2 M KCl solution (soil:solution, 1:5) and filtered through a 0.45 μ m filter 13 . A portion of each solution was prepared to determine the nitrate and ammonium concentrations, while the other portion was prepared to determine the nitrate δ 15 N and δ 18 O levels. The nitrate and ammonium concentrations were analysed using a continuous flow analyser (Skalar San+ + System, Skalar Analytical B.V., Netherlands). The soil samples used for the bulk density analysis were dried at 115 °C for 24 h. STN storage (Mg ha −1 ) values were calculated as follows: where D i , BD i and TN i represent the soil thickness (cm), bulk density (g cm −3 ) and soil total N content (g kg −1 ), respectively, for the ith level of the soil profile. Foliar δ 15 N and nitrate δ 15 N and δ 18 O analyses. The fresh foliar litterfall was sampled in the forestland and grassland in September 5-14, 2013 and returned to the laboratory. Dust and soil were carefully removed from the surfaces of the foliar samples. The samples were air-dried and ground to a powder. The foliar N content and δ 15 N were analysed using a Vario PYRO cube element analyser and an EA-IsoPrime100 stable isotope ratio mass spectrometer (Isoprime Ltd, U.K). The soil nitrate δ 15 N and δ 18 O levels were prepared via quantitative bacterial reduction of nitrate to nitrous oxide. Nitrous oxide was extracted and purified using a trace gas pre-concentrator unit; the product was analysed using an EA-IsoPrime100 stable isotope ratio mass spectrometer (Yue et al., 2014). Three international materials (USGS-32, USGS-34 and USGS-35) were used to calibrate the measured sample data. Each sample was measured in duplicate, and the standard error was 0.3% for nitrate δ 15 N and 0.4% for nitrate δ 18 O.
The isotope ratios (δ 15 N and δ 18 O) are expressed in δ notation as parts per thousand deviations (‰): Statistical analysis. An independent-sample t-test was performed to test the significance of the soil property differences, soil N storage and availability at an alpha level of 0.05 (a = 0.05) between the forestland and grassland. All statistical analyses were performed with the Statistical Program for Social Sciences (SPSS 11.0, SPSS Inc., 2001).