1 Introduction

Nitrogen (N) is a key element in sustaining primary production in the terrestrial ecosystems including cropland, forest, and grassland systems (Xu et al. 2009; Liu et al. 2010; Sun et al. 2010). Terrestrial ecosystems receive N from divergent sources including biological N2 fixation in forests (Guinto et al. 2000; Bai et al. 2012; Hyodo et al. 2013; Reverchon et al. 2020), grasslands (Conrad et al. 2018), and croplands (Unkovich and Pate 2001), and synthetic fertilizer and manure application in croplands (Xu et al. 1993a, 1993b; Choi et al. 2017) and grasslands (Klaus et al. 2013). With the increased use of fossil fuel and development of the livestock industry, the contribution of atmospheric N deposition to total N input becomes substantial, particularly in certain hot spot areas such as subtropical Asia (Sun et al. 2010; Schwede et al. 2018). Increased soil N availability by fertilizer N application and N deposition leads to increased primary production in N-limited ecosystems (De Vries et al. 2006; Sun et al. 2010; Du and de Vries 2018). However, excessive N addition and subsequent N loss are often related with site degradation due to soil acidification resulting from both the production of protons from nitrification and the loss of non-acidic cations with NO3− leaching (Sun et al. 2010; Tian et al. 2018), which could cause the eutrophication of water bodies (Billen et al. 2013). Gaseous loss of N through NH3 volatilization and N2O emission via nitrification/denitrification is also responsible for atmospheric pollution (Bouwman et al. 2013a; Sutton et al. 2013).

Direct measurements of soil N loss processes are not often possible due to the complexity and spatial and temporal variability of N cycling (Robinson 2001; Denk et al. 2017). In this regard, the natural N isotope abundance (δ15N) of soils is often used as a measure of the site-specific N loss potential, as soil δ15N tends to increase through N isotope fractionation with increased N loss, even though soil δ15N is also related with many factors including the δ15N of the source N, soil N availability, and the type and rate of N transformations occurring in the soil (Robinson 2001; Bai et al. 2012; Craine et al. 2015a). On a global scale, soil δ15N is related with climate (e.g., temperature and precipitation) (Amundson et al. 2003; Craine et al. 2015b); however, within a climate zone, soil δ15N is likely to be determined by land-use type (e.g., cropland, forest, and grassland) and land management (including disturbance) as the source of N and dominant N transformation processes differ with land-use type and land management (Bai et al. 2015a, 2015b, 2015c; Ning et al. 2015; Wang et al. 2015, 2020).

We are aware of many important and valuable reviews on the factors affecting soil δ15N, including soil N processes (Robinson 2001; Wang et al. 2015, 2020; Denk et al. 2017), the global climate (Amundson et al. 2003), ecophysiological N processes in plants (Bai et al. 2012; Craine et al. 2015a; Fu et al. 2020), and agricultural management (Choi et al. 2017). However, no one has systematically analyzed the patterns of soil δ15N as affected by land-use type (e.g., cropland, forest, and grassland) and land management practices; such knowledge would improve our understanding of the effect of land-use type and land management on N processes and thus the global N cycle as croplands, forests, and grasslands account for 1.2, 30.8, and 25.1%, respectively, of the global land area (FAOSTAT 2020).

In this review, we analyzed the variations in the δ15N of total soil N with land-use type and land management including disturbance. Specifically, our review focuses on (1) the overall pattern of the δ15N of various sources of N and the potential changes in soil δ15N with N loss, (2) the δ15N pattern of soils under cropland, forest, and grassland with special emphases on the effects of the source of N (chemical fertilizers with a depleted 15N signature vs. manure and composts with an enriched 15N signature) and water regime (paddy vs. upland) for cropland soils, climate (temperate, tropical, subtropical, and boreal) and tree functional group (evergreen coniferous vs. deciduous) for forest soils, and land type (grazed vs. hayfield) and management (source of N input and grazing density) for grassland soils, and (3) the effects of land-use change (e.g., from forest to cropland and grassland) and site disturbance, including burning and clear-cutting, on soil δ15N. We concluded this review by suggesting further studies that are necessary to better understand the variations in the δ15N of soils with land-use type and land management under environmental changes.

2 Materials and methods

2.1 Definition of δ15N

Since our review and analysis are based on soil δ15N, we would like to briefly define N isotopes and δ15N. There are two kinds of stable N isotopes, 14N and 15N, and 99.6337% of N atoms on earth are the lighter 14N, with the remaining 0.3663% as the heavier 15N (Ingerson 1953). The stable isotope ratio, 15N/(14N+15N), is expressed as 15N atom % (Mariotti 1983), and in the natural environment, 15N atom % varies within a narrow range from 0.355 to 0.377 atom % (Moore 1974). Due to the small variations in atom % in nature, the natural abundance of 15N in a sample is expressed as a relative abundance of 15N compared with the standard (atmospheric N2, 0.3663 atom %) using the δ notation in parts per thousand (‰) as:

$$ {\delta}^{15}N\ \left({\mbox{\fontencoding{U}\fontfamily{wasy}\selectfont\char104}} \right)=\left[\left({R}_{sample}-{R}_{standard}\right)/{R}_{standard}\right]\times 1000 $$
(1)

where Rsample and Rstandard are the 15N atom % of a sample and the standard, respectively. Based on this equation, δ15N of atmospheric N2 is 0‰, and the more 15N-enriched a sample than the standard is, the more positive its δ15N and vice versa.

2.2 Data collection

Data on δ15N of terrestrial N sources (Table S1) and soils (total soil N) under different land uses (Table S2) were collected from the literature by searching on the ISI-Web of Science and Google Scholar databases using combinations of keywords listed in Table S1 and S2. Data were extracted from 134 studies published between 1957 and 2019, with most studies published after 1990.

Land-use types were classified into cropland, forest, and grassland (including both pasture and meadow) following FAOSTAT (2020). For cropland soils, which are intensively managed with fertilization and irrigation, the dataset was further classified into conventional and organic farming based on the type of fertilizer N applied (synthetic fertilizers in conventional and livestock manure and composts in organic farming systems) and into paddy (waterlogged during crop growth) and upland soils based on the water regime. This classification is not mutually exclusive (e.g., conventional–organic vs. paddy–upland) but further division of the classification is not possible due to the limited number of studies available for organic farming in paddy systems and lack of information on the type of N applied in some studies reported in the literature. To investigate the effect of climate on forest soil δ15N, forest sites were classified into tropical, subtropical, temperate, and boreal climate zone. When the climate zone and GPS coordinates were not specified, the data were not included in this review and analysis. Grassland was classified into grazing fields and hayfields based on the purpose of the land use. Grazing fields are used for grazing without grass cutting, while hayfields are fields used for cultivating grass and non-woody plants for hay or silage production with fertilization, and thus, the management and associated N cycling may be different from grazing fields, although grazing fields may be also fertilized (Xu et al. 2019).

Data were collected from tables and figures of the literature. Data presented in figures were extracted using GetData Graph Digitizer (version 2.26) (available at http://getdata-graph-digitizer.com/). Either the mean or the experimental unit-level data were collected depending on data availability. For all land-use types, surface (up to 20 cm) mineral soil was considered, and when the soil depth was not specified, but soil horizon information was available, data for the A horizon was collected. For forest soils, data were also collected from organic layers (O, or L, F, and H layers).

2.3 Data analysis

The data were presented using box plots to display the minimum, first quartile, median, third quartile, and the maximum. For the analysis of variance (ANOVA), data were tested for homogeneity of variance and normality of distribution with Levene’s test and Kolmogorov-Smirnov test, respectively. Our analysis showed that the variance was homogeneous and the distribution was normal. The differences in the δ15N among the N sources and land-use types were tested using ANOVA, and when the difference was significant, means were separated with Tukey’s test. Regression analyses were conducted to explore the relationships between δ15N and other factors. Statistical significance was set at α = 0.05.

3 Results and discussion

3.1 δ15N of N sources in terrestrial N cycling

Soil δ15N in terrestrial ecosystems is affected by the source of N and N loss potential and pathways (e.g., NH3 volatilization, denitrification, and leaching), which are influenced by land-use type and land management practices (Robinson 2001; Amundson et al. 2003). Terrestrial ecosystems receive a total of 313 Tg N of reactive N annually (Fowler et al. 2013) from the atmosphere through industrial N2 fixation (total 120 Tg, 100 Tg as fertilizer N and 20 Tg from the chemical industry; Galloway et al. 2008; Bouwman et al. 2013b), biological N2 fixation (total 118 Tg, 58 Tg from natural ecosystems and 60 Tg from cropland; Vitousek et al. 2013), N deposition (70 Tg; Dentener et al. 2006; Duce et al. 2008), and lightning (5 Tg; Levy et al. 1996; Tie et al. 2002) (Fig. 1). Though most of the N present in terrestrial ecosystems originated from atmospheric N2 with δ15N of 0‰, the δ15N of the N received by the terrestrial ecosystems varies with the source of the N due to the different magnitude of N isotope fractionation when the atmospheric N2 is converted to reactive N (Robinson 2001).

Fig. 1
figure 1

Annual fluxes of nitrogen between the atmosphere and terrestrial ecosystems. Data sources: industrial N2 fixation (Galloway et al. 2008; Bouwman et al. 2013b), biological N2 fixation (Vitousek et al. 2013), N deposition (Dentener et al. 2006; Duce et al. 2008), lightning (Levy et al. 1996; Tie et al. 2002), NOx emission from fossil fuel combustion and biomass burning (van Vuuren et al. 2011), NOx emission from soils (Ganzeveld et al. 2002; Bouwman et al. 2013a; Pilegaard 2013), NH3 volatilization (Sutton et al. 2013), N2 emission via denitrification (Bouwman et al. 2013a), and leaching and runoff (Howarth et al. 1996; Billen et al. 2013)

Full size image

Typical N sources in terrestrial ecosystems are atmospheric N2 fixation, atmospheric NH4+ and NO3− deposition, synthetic fertilizer N, and N from manure and compost; those N all have unique δ15N. Biological atmospheric N2 fixation by the enzyme nitrogenase of diazotrophs is a universal source of N in terrestrial ecosystems (Unkovich 2013). The isotope fractionation factor (α) during N2 fixation is reported to be small (α = 1.000–1.006; Robinson 2001), and thus the δ15N of diazotrophs is close to 0‰, from – 3.7 to + 3.8‰, reflecting the 15N signature of the N source, atmospheric N2 (Delwiche and Steyn 1970; Hoering and Ford 1960; Macko et al. 1987; Karl et al. 1997). The δ15N of nodulated legumes, due to the main source of N being from the atmosphere, is also close to 0‰ with a mean value (± standard error) of – 0.5 ± 0.1‰ (Table S1 and Fig. 2; Boddey et al. 2000; Unkovich 2013; Unkovich and Pate 2001). The δ15N (+ 1.4 and – 0.5‰) of NOx produced by lightning is also very close to 0‰ (Hoering 1957). Atmospheric N deposition such as NH4+ and NO3− also have relatively low δ15N (– 2.1 ± 0.3‰) as the precursors of NH4+ and NO3− are depleted in 15N (Table S1) (Lee et al. 2012). The main precursor of NH4+ is NH3 volatilized from agricultural areas (fertilizers and manure) that is depleted in 15N due to faster volatilization of 14NH3 than 15NH3 (Felix et al. 2013). The NOx compounds emitted from fossil fuel combustion and nitrification/denitrification, which are the main source of wet NO3−, also have low δ15N as atmospheric N2 is converted to NOx during fossil fuel combustion (Felix et al. 2012) and nitrification/denitrification also causes a significant N isotope fractionation, producing 15N-depleted NOx (Choi and Ro 2003).

Fig. 2
figure 2

Box plots of the distribution of δ15N of terrestrial N sources including N2-fixing nodulated legumes (NFL, n = 114), lightning (LT, n = 2), atmospheric N deposition (AND, n = 234), synthetic fertilizers (SF, n = 87), livestock manure (LM, n = 31), and manure compost (MC, n = 59). Boxes represent interquartile ranges (IQRs), and horizontal lines within boxes indicate median values. The upper and lower whiskers indicate 75 percentile plus 1.5 IQR and 25 percentile minus 1.5 IQR, respectively. × is an average value, and ○ is a mild outlier with value > 75 percentile plus 1.5 IQR, but < 75 percentile plus 3.0 IQR. Different lowercase letters above the boxes indicate a statistical difference among N sources. The source of data is provided in Table S1

Full size image

Synthetic fertilizers, livestock manure, and manure composts, which are typical N sources in agricultural systems, have contrasting δ15N (Choi et al. 2017), with – 0.3 ± 0.2‰, + 7.8 ± 0.6‰, and + 16.3 ± 0.8‰, respectively (Table S1 and Fig. 2). Synthetic fertilizers, including CO(NH2)2, (NH4)2SO4, NH4NO3, and KNO3, have a wide range of δ15N (from – 5.9 to + 6.6‰) due to the different magnitude of N isotope fractionation associated with chemical processes following the Haber-Bosch process (Flipse and Bonner 1985). Livestock manure becomes enriched with 15N with time after excretion due to greater loss of 14NH3 than 15NH3 via volatilization, and the δ15N of manure compost is more positive than manure due to the greater loss of 14N during composting (Kim et al. 2008).

3.2 N loss and its effect on soil δ15N

In terrestrial ecosystems, approximately 298 Tg of N is lost annually via NOx emission from fossil fuel combustion and biomass burning (40 Tg; van Vuuren et al. 2011), NOx emission through nitrification/denitrification from soils (18 Tg; Ganzeveld et al. 2002; Bouwman et al. 2013a; Pilegaard 2013), NH3 volatilization (60 Tg; Sutton et al. 2013), N2 emission via denitrification (100 Tg; Bouwman et al. 2013a), and leaching and runoff (80 Tg; Howarth et al. 1996; Billen et al. 2013). Among the N loss pathways, NOx emission through nitrification and/or denitrification and NH3 volatilization critically affect the δ15N of the soils due to the large chemical kinetic isotope fractionation (1.020–1.033 for denitrification and 1.020–1.029 for NH3 volatilization; Högberg 1997). Among the gaseous N loss pathways, NH3 volatilization may be critical in alkaline soils after fertilization, while the closely coupled nitrification-denitrification processes may play a greater role in increasing δ15N of soils in non-alkaline soils (Robinson 2001; Wang et al. 2015, 2020). In particular, nitrification directly increases δ15N of NH4+ via direct emission of 15N-depleted N2O and production of 15N-depleted NO3− (Ibell et al. 2013a, 2013b; Zhang et al. 2018), which is subsequently subject to loss via denitrification as well as leaching and runoff (Craine et al. 2015a; Lim et al. 2015; Wang et al. 2015, 2020). Loss of NO3− through leaching and runoff may not cause chemical kinetic isotope fractionation; however, loss of 15N-depleted NO3−, produced from incomplete nitrification, increases δ15N of bulk soils (Lim et al. 2015).

Under natural and undisturbed conditions, many abiotic and biotic factors including soil type (age and parent materials), topographic position, vegetation, mycorrhizal fungal associations, and climate (temperature and precipitation) affect the closeness of the N cycle and thus the magnitude of N loss, leading to different δ15N (Amundson et al. 2003; Hobbie and Ouimette 2009; Zhou et al. 2014; Craine et al. 2015a). Among those, the global pattern of soil δ15N is often correlated negatively with mean annual precipitation (MAP) and positively with temperature (MAT) (Amundson et al. 2003; Craine et al. 2015b). Though the mechanisms underlying the relationships between soil δ15N and MAP or MAT were not fully understood, wetter and colder ecosystems are thought to be more efficient in recycling NH4+ and NO3−, thereby reducing N loss (Amundson et al. 2003). Such patterns of soil δ15N with climate are also found at the regional scale for temperate rain forests in southern Chile (Boeckx et al. 2005), arid forests in northern Chile (Díaz et al. 2016), temperate forests in southern Argentina (Peri et al. 2012), and grasslands in Inner Mongolia (Cheng et al. 2009).

The direct connection between soil δ15N and climate is, however, still debated. For example, using a data set of δ15N in mineral soils (0–30 cm) across the world (n = 910), soil δ15N has been shown to increase with increasing MAT at a rate of 0.18 ± 0.02‰ °C−1 when MAT is > 9.8 °C and decrease with increasing MAP at a rate of 1.78 ± 0.24‰ per order of magnitude of increase in MAP (log10MAP); however, such relationships were not evident after standardizing soil δ15N for variations in soil C and clay content, which co-vary with the climate gradient (Craine et al. 2015b). Craine et al. (2015b) suggested that the relationship between soil δ15N and climate may be indirect and mediated by the effects of climate on soil C and clay contents through influences on SOC decomposition and subsequent N loss and stabilization of 15N-enriched N in organo-mineral complexes (Craine et al. 2015a). There is a consistent trend of increasing soil δ15N with decreasing soil C/N under different land-use types (Fig. 3a) (Conen et al. 2008; Marín-Spiotta et al. 2009; Stevenson et al. 2010), with a few exceptions (Marty et al. 2019). A high soil C/N indicates that the decomposition of organic matter is slow at the early stage, and thus there is less opportunity of N to leave the soil, resulting in a less 15N enrichment by N loss (Fig. 3b) (Craine et al. 2015a; Weintraub et al. 2016). The soil C/N is also regarded as a soil N status as a lower C/N is often associated with a high N availability, increased soil δ15N with decreasing soil C/N should be attributed to a greater N loss potential and thus a lower N retention capacity under high N availability conditions (Templer et al. 2012). Therefore, it can be postulated that MAT and MAP may influence soil δ15N by affecting microbial decomposition of organic matter, i.e., wetter and colder conditions are unfavorable for microbial activity, including heterotrophs and autotrophic nitrifiers, resulting in lower soil δ15N than drier and hotter conditions (Craine et al. 2015b).

Fig. 3
figure 3

Relationships between δ15N and C/N of ecosystem samples. (a) Changes in δ15N with C/N of organic matter fractions (Conen et al. 2008) and bulk soils (Stevenson et al. 2010; Weintraub et al. 2016) under different land-use types including grassland (Conen et al. 2008; Stevenson et al. 2010) and forest (Stevenson et al. 2010; Weintraub et al. 2016). (b) Progressive increases in δ15N of different soil organic matter fractions (LL, leaf litter; FLF, free light fractions including particulate organic matter; OLF, occluded light fractions including intra-aggregate organic matter; HF, heavy fractions in mineral associations) with decreasing C/N through microbial decomposition and stabilization of organic matter (Data from Marín-Spiotta et al. 2009)

Full size image

3.3 δ15N of soils as affected by land-use type and land management

3.3.1 Overall δ15N patterns of cropland, forest, and grassland soils

Analysis of the soil δ15N under different land uses showed that grassland soils (+ 6.2 ± 0.1‰, n = 624) had the highest δ15N followed by cropland soils (+ 5.0 ± 0.2‰, n = 168), and the mineral (+ 3.1 ± 0.2‰, n = 428) and organic layer of forest soils (– 1.0 ± 0.2‰, n = 227) (Table S2 and Fig. 4). Such differences may reflect the N loss potential of different land uses as well as N sources (Fig. 2) (Robinson 2001). In grasslands, intensive cultivation and harvest of 15N-depleted forage with a high fertilization rate in hayfields and volatilization of 15N-depleted NH3 from livestock excretion in grazing fields may be responsible for increasing soil δ15N (Kriszan et al. 2009; Rui et al. 2011; Klaus et al. 2013; Fornara et al. 2020). Similar to grasslands, removal of 15N-depleted N through harvesting and loss of 15N-depleted N from fertilizers might increase the δ15N of cropland soils (Choi et al. 2017).

Fig. 4
figure 4

Box plots of the distribution of δ15N of soils under different land-uses including cropland (n = 168), forest (organic layer, n = 177; mineral layer, n = 337), and grassland (n = 624). Soil depth for cropland, forest mineral, and grassland soils is limited to the upper 20 cm. Boxes represent interquartile ranges (IQRs), and horizontal lines within boxes indicate median values. The upper and lower whiskers indicate 75 percentile plus 1.5 IQR and 25 percentile minus 1.5 IQR, respectively. × is an average value, and ○ is a mild outlier with the value > 75 percentile plus 1.5 IQR, but < 75 percentile plus 3.0 IQR. Different lowercase letters above the boxes indicate statistical difference among land-use types. The source of data is provided in Table S1

Full size image

The lower δ15N of forest soils than cropland and grassland soils indicates that N cycling in forest soils is more conservative or less open in terms of N loss compared with others due to limited N addition and subsequent N loss, unless forests are disturbed (Högberg and Johannisson 1993; Eshetu and Högberg 2000; Stevenson et al. 2010; Qiu et al. 2015). Lack of removal of 15N-depleted tree tissue via harvest is also responsible for the conservative N cycling; for example, Zhang et al. (2018) reported that long-term retention of harvesting residues on re-established 2nd rotation forest plantations grown on sandy soils in subtropical Australia decreased soil δ15N through reducing N leaching and increasing N immobilization. The greater N retention capacity of forests than other land uses is confirmed by a higher recovery of 15N tracer in forests (74.9%) than in grasslands (51.8%) after 3–18 months of 15N application based on a global a meta-analysis (Templer et al. 2012).

As forest soils rarely receive fertilizers in the way that croplands do, δ15N of forest soils reflects the isotope signal of litterfall, 15N enrichment associated with N loss, and 15N-depleted atmospheric N deposition (Högberg 1997; Pardo et al. 2006; Sun et al. 2010; Fu et al. 2020). The δ15N of tree litterfall is determined by the N acquisition pathway: whether N is directly assimilated from the available N pool in the soil with or without biological N2 fixation, or indirectly from mycorrhizal association (Högberg 1997; Hobbie and Hobbie 2008). The dependence of trees on mycorrhizal association for N acquisition is substantial when soil N availability is low, leading to decreased δ15N in tree tissues due to N isotope fractionation that occurs when N is transported from mycorrhizae to roots compared with trees without mycorrhizal association (Hobbie and Hobbie 2008; Hobbie and Ouimette 2009). With the increased soil N availability towards N saturation, the δ15N soil N increases due to N loss and the dependence of trees on 15N-enriched soil N tends to increase, resulting in increased δ15N of tree tissues and subsequently the δ15N of organic layer that accumulates from litterfall (Takebayashi et al. 2010; Fang et al. 2011b). The higher δ15N of mineral layer than organic layer is generally ascribed to the greater cumulative N loss from the mineral soil that enriches 15N, as well as the selective mycorrhizal assimilation of 15N-depleted N from the mineral soil (Craine et al. 2015a).

3.3.2 Cropland soils: effects of farming system and water regime on δ15N

The δ15N of cropland soils was affected by farming systems and water regimes (Fig. 5). The δ15N of cropland soils under conventional farming (+ 4.6 ± 0.3‰), which is heavily influenced by 15N-depleted synthetic fertilizer input, was lower than that under organic farming (+ 6.8 ± 0.4‰) (Fig. 5a), and this result is consistent with a review by Choi et al. (2017).

Fig. 5
figure 5

Box plots of the distribution of δ15N of cropland soils with different management practices: (a) farming systems (CF, conventional farming, n = 49; OF, organic farming, n = 49) and (b) water regimes (paddy, n = 56; and upland, n = 108). Boxes represent interquartile ranges (IQRs), and horizontal lines within boxes indicate median values. The upper and lower whiskers indicate 75 percentile plus 1.5 IQR and 25 percentile minus 1.5 IQR, respectively. × is an average value, and ○ is a mild outlier with the value > 75 percentile plus 1.5 IQR, but < 75 percentile plus 3.0 IQR. Different lowercase letters above the boxes indicate statistical difference between management. The source of data is provided in Table S1

Full size image

Contrasting δ15N of paddy and upland soils has been reported by Lim et al. (2015). The lower δ15N of paddy soils (+ 3.1 ± 0.2‰) than that of upland soils (+ 6.0 ± 0.2‰) (Fig. 5b) suggests that N cycling in paddy is tighter than that in upland soils. In upland soils, high rate of nitrification of NH4+ derived from fertilizer N and subsequent leaching of 15N-depleted NO3− should increase the δ15N of NH4+ remaining in the soil (Choi et al. 2003a, b). As heterotrophic microbes that assimilate N prefer NH4+ to NO3− due to high energy cost in using NO3− (Rice and Tiedje 1989), microbial assimilation of 15N-enriched NH4+ should increase the δ15N of organic N pool and thus total soil N in upland soils (Lim et al., 2015). In paddy soils, however, the low nitrification rates under waterlogging conditions would not increase the δ15N of NH4+ and thus total soil N as high as in upland soils (Choi et al. 2003a; Wang et al. 2015, 2020). Denitrification in paddy soils might be able to increase δ15N of NO3− and thus contribute to increases in soil δ15N; however, due to limited NO3− availability for denitrifying bacteria under nitrification-restricted conditions as well as a low preference of NO3− for heterotrophs, denitrification-induced increases in δ15N of total N of paddy soils might not be as great as nitrification-induced increases in δ15N of total N of upland soils (Lim et al. 2015).

3.3.3 Forest soils: effects of climate and tree type on δ15N

Among the climate zones, tropical (+ 0.7 ± 0.9‰ and + 7.4 ± 0.5‰ for organic and mineral layer, respectively) forests had greater δ15N than boreal (– 2.2 ± 0.2‰ and + 3.0 ± 0.3‰, respectively), temperate (– 0.3 ± 0.3‰ and + 2.0 ± 0.2‰, respectively), and subtropical (– 3.0 ± 0.3‰ and + 2.9 ± 0.3‰, respectively) forests (Fig. 6 a and b). The higher soil δ15N of tropical forests is believed to reflect high soil N availability due to fast organic matter decomposition under warmer conditions (Martinelli et al. 1999). Under N-rich conditions in tropical forests, N cycling is more open; for example, a high level of NO3− in the stream of tropical forests is often found (Brookshire et al. 2012). The more open N cycling results in increased soil δ15N due to isotope fractionation caused by denitrification in tropical forests (Wang et al. 2018) and loss of 15N-depleted NO3− (Brookshire et al. 2012). However, the lower δ15N in the organic layer in subtropical than in temperate and boreal forests is not consistent with the understanding that temperate and boreal forests have slower N turnover rates and should be more N limited and thus should have lower δ15N compared with subtropical forests (Martinelli et al. 1999; Cheng et al. 2010).

Fig. 6
figure 6

Box plots of the distribution of δ15N of forest soils with climate zone (a and b) and tree type (c and d). (a) Organic layer in different climate zones (n = 167, 3, 36, and 59 for temperature, tropical, subtropical, and boreal forest, respectively). (b) Mineral layer (0–20 cm) in different climate zones (n = 137, 79, 57, and 74, respectively). (c) Organic layer with tree type (n = 40 for coniferous and 16 for deciduous). (d) Mineral layer with tree type (0–20 cm) (n = 190 and 156, respectively). Boxes represent interquartile ranges (IQRs), and horizontal lines within boxes indicate median values. The upper and lower whiskers indicate 75 percentile plus 1.5 IQR and 25 percentile minus 1.5 IQR, respectively. × is an average value, and ○ is a mild outlier with the value > 75 percentile plus 1.5 IQR, but < 75 percentile plus 3.0 IQR. Different lowercase letters above the boxes indicate statistical difference between climate zone and tree type. Comparing between soils of the organic layer and mineral layer within climate zone, δ15N of the organic layer was consistently (P < 0.001) lower than that of the mineral layer across the climate zone. Data sources for climate zone are provided in Table S1, and for tree type, data from Billings and Richter (2006), Boeckx et al. (2005), Callesen et al. (2013), Cheng et al. (2010), Compton et al. (2007), Fujiyoshi et al. (2019), Girona-García et al. (2018), Hobbie and Ouimette (2009), Hou et al. (2015), Koba et al. (1998, 2003), Koopmans et al. (1997), Lemma and Olsson (2006), Marty et al. (2011), Matsushima et al. (2012), Peri et al. (2012), Pörtl et al. (2007), Sah (2005), Scott et al. (2008), Sheng et al. (2014), Templer et al. (2007), Trudell et al. (2004), Vervaet et al. (2002), and Yu et al. (2018)

Full size image

Since the sites of the most studies conducted in the subtropical forests were exposed to high N deposition, we assumed that the lower δ15N in subtropical forests might be a result of high N deposition rates in the sites of these studies (Gurmesa et al. 2017, 2019). Currently, it is estimated that atmospheric N deposition is the highest in subtropical forests (14.6 kg N ha−1) followed by temperate and tropical (both 7.2 kg N ha−1) and boreal (1.2 kg N ha−1) forests (Fig. S1) (Schwede et al. 2018). Therefore, soil δ15N pattern of subtropical soils deviating from the N availability–δ15N relationship can be interpreted as the effect of the high deposition rate of 15N-depleted atmospheric N on subtropical forests. In the studies of Gurmesa et al. (2017, 2019) conducted in subtropical forests of southern China under high N deposition (21–38 kg N ha−1 year−1), the low δ15N of soil, as well as foliage samples, was attributed to the effects of 15N-depleted atmospheric N deposition. Decreases in the δ15N of subtropical forests with increasing atmospheric N deposition has also been demonstrated by Fang et al. (2011a), who reported that foliar δ15N decreased from – 1.5‰ to – 4.4‰ with increasing N deposition from 16.7 to 29.8 kg N ha−1 year−1 followed by increases in δ15N to up to – 2.4‰ when N deposition further increased to 38.2 kg N ha−1. Such responses of subtropical forests to atmospheric N deposition might be related with the high N retention capacity compared with temperate forests as indicated by Yu et al. (2018) who reported that the threshold for N leaching in a subtropical forest in southern China was 26–36 kg N ha−1 year−1, which is higher than in temperate forests by 160%. To address climatic controls on the δ15N of forest soils, however, more studies on less polluted sites are required to investigate the natural variations in soil δ15N of subtropical forests

The δ15N of forest soils also differed with functionally different tree types (evergreen coniferous vs. broad-leaved deciduous trees) (Fig. 6 c and d). As litterfall is a unique N source in forests, the higher δ15N of the organic layer in deciduous trees may reflect 15N-enriched litter of deciduous broadleaves compared with conifer needles (Martinelli et al. 1999; Eshetu 2004; Sah et al. 2006; Pardo et al. 2007; Cheng et al. 2010; Fang et al. 2011b; Marty et al. 2011; Dawes et al. 2013). Many biotic and abiotic factors including rooting depth, tree preference of NH4+ vs. NO3−, mycorrhizal associations, and mineralization–nitrification rates would affect δ15N of tree litter and thus organic layer (Pardo et al. 2007; Templer et al. 2007; Craine et al. 2015a; Park et al. 2019). For example, as deciduous trees have a deeper rooting depth than coniferous trees (Reynolds-Henne et al. 2007; Klein et al. 2013), assimilation of more 15N-enriched N from deeper soils would result in higher δ15N of deciduous litter as δ15N of mineral soils increases with soil depth (Craine et al. 2015a). In addition, faster mobilization and translocation of 15N-depleted N during re-absorption of N from senescing deciduous foliage (Kolb and Evans 2002) should also contribute to the higher δ15N of the organic layer under deciduous trees by producing 15N-enriched litter annually. Besides, faster decomposition of deciduous than conifer litter due to lower C/N and recalcitrant lignin, and higher concentrations of nonstructural carbohydrate and mineral concentrations (Lorenz et al. 2004; Wang and Yang 2007; Chodak et al. 2016; Park et al. 2018, 2019) should be also responsible in part for the 15N-enrichment of δ15N of the organic layer of deciduous forests as increased N availability via N mineralization should lead to 15N enrichment through N loss (Craine et al. 2015a).

It is often reported that the δ15N of mineral layer of deciduous forests is higher than that of coniferous forests due to the faster rate of net mineralization and nitrification of deciduous stands than coniferous stands (Boeckx et al. 2005; Cheng et al. 2010; Callesen et al. 2013). However, an opposite pattern was found in our analysis (Fig. 6d), which is consistent with Sheng et al. (2014). Loss of N depends on NO3− availability and thus a high rate of mineralization–immobilization turnover may reduce N loss by decreasing NO3− availability via retarding nitrification (Stark and Hart 1997; Verchot et al. 2001). In an in situ 15N study conducted on mineral soils (0–10 cm) of oak (Quercus robur L.) and pine (Pinus sylvestris L.) stands, Staelens et al. (2012) reported that gross soil N mineralization was more than two times greater in an oak stand; however, production of NO3− via oxidation of organic N was three times faster in the pine while nitrification rate was not different in both soils, resulting in greater NO3− availability in pine than oak soil. In addition, in the oak soil, dissimilatory reduction of NO3− to NH4+ was also greater than in pine soil (Staelens et al. 2012). Based on the observation of Staelens et al. (2012) together with our analysis, we postulate that the higher δ15N of mineral layer in coniferous than in deciduous forests might be caused by a faster rate of NO3− production through organic N oxidation as well as a lower rate of NO3− immobilization in coniferous soils, resulting in a high availability of NO3− that is subject to loss via leaching and denitrification. However, long-term research on N cycling through organic layer–mineral layer–tree uptake–litterfall by combining both 15N-labeling and natural 15N abundance is necessary to better understand the different patterns of δ15N in organic and mineral soil layers between deciduous and coniferous forests, in association with gross soil N processes.

3.3.4 Grassland soils: effect of grazing and fertilization on δ15N

The overall δ15N values of grassland soils were higher than those of cropland and forest soils (Fig. 4) and were affected by land-use management such as grazing and the type of fertilizer applied (Fig. 7). In grazed grasslands with limited fertilization, biological N2 fixation is an important N source (Conrad et al. 2018) and return of N through excretion of grazing livestock is a critical pathway of N cycling (Dijkstra et al. 2006). Hayfields are often fertilized with N either as synthetic fertilizer and/or livestock manure (Watzka et al. 2006), those would result in more open N cycling and higher soil δ15N in hayfield than in grazed grasslands (Fig. 7a) (Kriszan et al. 2014).

Fig. 7
figure 7

Box plots of the distribution of δ15N of grassland soils with different management. (a) Grassland type: hayfield (n = 492) vs. grazing field (n = 132). (b) Fertilizer type: no fertilizer (n = 26), synthetic fertilizer (n = 34), cattle slurry (n = 20), cattle manure (n = 15), and pig manure (n = 6). (c) Grazing density. Boxes represent interquartile ranges (IQRs), and horizontal lines within boxes indicate median values. The upper and lower whiskers indicate 75 percentile plus 1.5 IQR and 25 percentile minus 1.5 IQR, respectively. × is an average value, and ○ is a mild outlier with the value > 75 percentile plus 1.5 IQR, but < 75 percentile plus 3.0 IQR. Different lowercase letters above the boxes indicate statistical difference between climate zone. Data sources for grassland are provided in Table S2 and data for fertilizer type are from Fornara et al. (2020), Kriszan et al. (2009), Mudge et al. (2013), and Watzka et al. (2006). Variations in the δ15N of pasture soils (y) with grazing density (x). The regression equations between δ15N of pasture soils (y) and grazing density (x) are y = 0.19x + 4.8 (r2 = 0.67, P = 0.067, n = 5) for Cheng et al. (2009), y = − 0.31x + 7.5 (r2 = 0.94, P = 0.010, n = 4) for Du et al. (2017), y = 0.057x + 5.4 (r2 = 0.18, P = 0.470, n = 5) for Xu et al. (2010), and y = 0.92x + 3.7 (r2 = 0.88, P = 0.002, n = 7) for Kriszan et al. (2014)

Full size image

In hayfields, soil δ15N varied with the type of source N applied except for synthetic fertilizer (Fig. 7b). The application of 15N-depleted synthetic fertilizer did not cause a significant change in soil δ15N compared with the control (no fertilizer), probably suggesting that the δ15N of grassland soils receiving synthetic fertilizer is counterbalanced by 15N-enrichment through N losses (Choi et al. 2017). Application of livestock manure increased the δ15N of soils as discussed earlier, and the progressive increases in soil δ15N from cattle slurry to pig manure might reflect the difference in the δ15N of the N source as well as the N application rate and the degree of N availability of the N sources (Choi et al. 2017). Since either δ15N or N application rate or both were missing in the most literature surveyed, it was not possible to analyze the variations in the soil δ15N with manure type. However, the application of manure containing readily available N (NH4+, NO3−, and dissolved organic N) with a high δ15N at a high rate for a longer time should increase soil δ15N over the other cases (Choi et al. 2017).

In grazed grasslands, soil δ15N increased (Kriszan et al. 2014), decreased (Du et al. 2017), or did not change (Cheng et al. 2009; Xu et al. 2010) with grazing density (Fig. 7c). Due to the small number of data sets available, it is not possible to directly attribute the different responses of soil δ15N to grazing density. However, the variations in the soil δ15N might be primarily associated with the amount of livestock excretion related to grazing density, often soil δ15N increased with increasing grazing density (Kriszan et al. 2014). Fresh manure immediately after excretion from livestock is depleted in 15N, ranged from – 3.5 to + 4.8‰ (Kreitler 1975; Gormly and Spalding 1979), as N in animal proteins become enriched with 15N (Gaebler et al. 1966). The processes of NH3 volatilization, leaching, and denitrification will all enrich the remaining N with 15N in the livestock excretion after its deposition (Choi et al. 2017). Among those, NH3 volatilization can be substantial as soil pH increases through hydrolysis of amines (R-NH2 + H2O → CO2 + NH3 + OH−) (Kreitler 1975; Kim et al. 2008). Ammonia NH3 from manure is dependent on temperature (Van der Stelt et al. 2007), and thus, increased soil δ15N with increasing grazing density reported by Kriszan et al. (2014) could be attributed to the warmer conditions (MAT was + 7 to + 10 °C) that might facilitate hydrolysis of amines and subsequent NH3 volatilization compared with the other studies (– 4 to + 7 °C) (Rui et al. 2011; Du et al. 2017). In a similar manner, the low MAT (– 4 °C) in the study area of Du et al. (2017) should be responsible for the decreased soil δ15N with increasing grazing density as manure itself has a low δ15N (Kreitler 1975; Gormly and Spalding 1979). However, to confirm such temperature dependency of the relationship between grazing density and soil δ15N, more experimental data set is required.

3.4 Land-use change and disturbance effects on soil δ15N

3.4.1 Effect of land-use change on δ15N

Forests are often converted to agricultural lands including croplands and grasslands to meet the demand for food production. After conversion of forests to croplands and grasslands, soil δ15N increased over time (Fig. 8a; Awiti et al. 2008; Mudge et al. 2014), due to the loss of N carried from the forest soil and/or addition of fertilizer N (Awiti et al. 2008). Such increases in soil δ15N by conversion of forests to croplands and grasslands are also reported by other studies (Fig. 8b) (Lemma and Olsson 2006). However, contradictory results have been reported; for example, Piccolo et al. (1994) reported that conversion of forests to grassland (pastures) in Brazilian Amazon decreased soil δ15N from + 8.0 to + 3.2‰ for 20 years due to increased contribution of 15N-depleted N2 fixation. Similarly, Billings and Richter (2006) reported that afforestation of croplands in South Carolina increased δ15N of mineral soils by up to 0.6‰ for four decades, due to the incorporation of ectomycorrhizal biomass N into soil N pool as ectomycorrhizal fungi are 15N-enriched due to the preferential transfer of 14N to vegetation.

Fig. 8
figure 8

Changes in soil δ15N as affected by land-use conversion. (a) Chronicle changes in the δ15N of soils after land-use conversion from forest to cropland and grassland relative to the δ15N of the forest soil (δ15Ndiff). (b) The δ15Ndiff between before and after land-use conversion from forest to cropland and grassland, from grassland to cropland, and from cropland to forest. The number beside the data are the reference sources: (1) Mudge et al. (2014), (2) Awiti et al. (2008), (3) Lemma and Olsson (2006), (4) Piccolo et al. (1994), (5) Qiu et al. (2015), (6) Qiao et al. (2015), and (7) Billings and Richter (2006)

Full size image

When grasslands are converted to croplands, soil δ15N may increase or decrease (Qiu et al. 2015), depending on the type of N applied to croplands. Qiao et al. (2015) reported that 10 years after the conversion of grassland to cropland for the cultivation of oat (Avena sativa) with livestock manure in a Tibetan grassland, the δ15N of the top (0–20 cm) soil was increased by 0.4‰. Though the applied manure had a low δ15N (+ 0.5‰), fast NH3 volatilization occurring immediately after manure application should increase the remaining N derived from the manure that is readily immobilized by microbes in the presence of organic C substrate in the manure (Kreitler 1975; Kim et al. 2008). However, Qiu et al. (2015) reported that the δ15N of top soil (0–15 cm) decreased by 0.75 ‰ with repeated (27 years) application of synthetic fertilizer (up to 385 kg N ha−1) to the cropland soils that was converted from grassland in the China Loess Plateau. Such differences in the changes in the δ15N of the cropland soils converted from grasslands as affected by different N sources are in agreement with the general trends of soil δ15N variations with the application of isotopically different N inputs in cropland soils (Fig. 5a) (Choi et al. 2017). Therefore, it is inferred that land-use change alters soil δ15N through changed N retention capacity of the sites, shift of natural N sources, and application of different types of fertilizers.

3.4.2 Effect of site disturbance on soil δ15N

Forest fire (wildfire and prescribed burning) is a typical agent of forest site disturbance, and δ15N of soils and plants of the sites disturbed by fire increases over the unburned sites due to loss of 15N-depleted NO3− during burning and post-fire (Boeckx et al. 2005; Stephan et al. 2015; Tahmasbian et al. 2019). In a temperate rain forest of Chile, Boeckx et al. (2005) found that δ15N of leaves and mineral soil (0–20 cm) was 15N-enriched by 4.7‰ and 5.6‰, respectively, after burning, and Stephan et al. (2015) reported that both wildfire and prescribed burning increased δ15N of soil mineral N by 1.2–3.2‰ and foliar δ15N by 1.3–3.0‰ in Idaho, USA. Although burning initially increases the δ15N of soil and plant compared with the unburned sites, δ15N of plants and soils of the organic layer and top mineral layer at burned sites tends to decrease with vegetation development and succession (Fig. 9). Such time-dependent 15N depletion is primarily attributed to the shift of N source from 15N-enriched soil to 15N-depleted N, such as biological N2 fixation and mycorrhizal association (Hyodo et al. 2013). It is widely reported that biological N2 fixation by either cyanobacteria or nodulated plants is the most important source of N supply during vegetation succession following burning (Markham 2009; Menge and Hedin 2009; Reverchon et al. 2020). In addition, the contribution of mycorrhizal fungi to plant nutrition should also play an important role in decreasing plant δ15N and, subsequently, organic layer and top mineral soils under N-limited conditions (Hobbie and Hobbie 2008; Hyodo et al. 2013).

Fig. 9
figure 9

Changes in the δ15N of tree leaf, the organic and mineral layers of soils with time after burning. The temporal pattern was only significant for leaf samples (r2 = 0.31, P < 0.001). Data from Hyodo et al. (2013) for leaf and organic layer and from Aaltonen et al. (2019) for mineral layer. In the study of Aaltonen et al. (2019), data were provided for 0–50 cm, but 0–5 cm data were selected as the surface soil is more sensitive to burning than deeper soils

Full size image

Other forest-disturbing agents such as clear-cutting (Pardo et al. 2002; Sah and Ilvesniemi 2006), removal of LFH layer (Choi et al. 2005), understory removal (Matsushima et al. 2012), and drainage (Choi et al. 2007), which are associated with forest management practices, are also reported to increase forest δ15N with different mechanisms. For clear-cutting, the increment of soil δ15N is mostly driven by 15N-enrichment of available N through enhanced N loss following increased nitrification (Fisk and Fahey 1990) though the 15N enrichment diminishes with vegetation development (Pardo et al. 2002). In the White Mountains of New Hampshire, USA, Pardo et al. (2002) reported that the δ15N of organic layer soils increased (by 0.8–1.3‰) after clear-cutting followed by gradual decreases until reaching the initial values again in 15 years after clear-cutting. Drainage of peatland also increases the δ15N of soil and vegetation tissue as drainage improves aeration and subsequently N mineralization and nitrification that lead to 15N-enrichment of soil N with progressive N loss (Choi et al. 2007). Understory removal is also shown to increase the δ15N of forest via increased N mineralization and nitrification (Matsushima and Chang 2007; Matsushima et al. 2012); meanwhile, LFH removal increases the δ15N via a shift of N source from 15N-depleted organic layer to 15N-enriched mineral layer (Choi et al. 2005).

4 Directions for future research

Our review proposes interesting research topics that may be of help in enlarging the understanding of the patterns of δ15N of total soil N in terrestrial ecosystem. Though the changes in the soil δ15N across climate are related to soil C/N, the exact mechanisms are not fully understood. Therefore, investigation of the long-term changes in the δ15N and C/N of soils amended with plant litters under different temperature and water regimes in controlled experiments may provide insights into how temperature and water conditions affect soil δ15N through influences on the decomposition of soil organic matter. For cropland, our review shows that the δ15N differs between paddy vs. upland soils and between organic vs. conventional farming; however, it is still unclear how organic and conventional farming in paddy soils affects soil δ15N. Therefore, for mutual exclusive comparison of the soil δ15N between organic vs. conventional farming under different water regimes, more experimental data on the δ15N of paddy soils under organic farming are necessary. For forest, extensive studies on less polluted subtropical forests are required for better understanding of the climate control on the δ15N of forest soils as most relevant studies have been conducted in the areas receiving a high rate of atmospheric N deposition, which might mask the effects of climate on the soil δ15N in the subtropical forests. In addition, to confirm the difference in the δ15N of both organic and mineral soils between deciduous and coniferous forests found in this review, soil N processes and associated changes in the δ15N need to be further explored through a comprehensive study using both 15N tracer and 15N natural abundance. For grassland, the effects of grazing density on the δ15N are not well documented, and thus, further studies are required to find a trend in the changes of δ15N by grazing density under different temperature regimes.

5 Conclusions

The lower soil δ15N of forests compared with croplands and grasslands as well as increased soil δ15N with the conversion of forests to croplands and grasslands reflects difference in N sources as well as the magnitude of openness of N cycling between land-use types. In forest ecosystems, 15N-depleted N2 fixation is an important N source, and N is utilized more conservatively through tree uptake of N with or without mycorrhizal association and N return to the soils via litter decomposition. However, it was also found that atmospheric deposition of 15N-depleted N could shift N source and alter N cycling. Therefore, the δ15N of forest soils should be carefully used in estimating the difference in the closeness of N cycling as the δ15N of soil reflects not only climatic conditions that affect the vulnerability of the soils to N loss but also biological and non-biological N sources such as atmospheric N deposition. Unlike forest soils, the variations in soil δ15N of cropland and grassland were affected by the management practices, particularly with fertilizer types for both cropland and grassland. Though conversion of land-use type changed the soil δ15N, the direction and magnitude of the changes in δ15N were not consistent, highlighting the necessity of more extensive and longer studies to find a general trend. Our review also suggests that though site disturbances such as burning and clear-cutting of forests shift the site to a more open N cycling system in the short term, the N cycling becomes more closed with the progress of site stabilization and vegetation succession. This study provides insights into the changes in soil δ15N with land-use types and management as well as land-use change and site disturbance that may be useful in connecting soil δ15N to N cycling under different land uses.