Introduction

Karst landscapes are distributed worldwide and cover ~ 18% of the Earth’s land areas1. The largest continuous karst outcrop globally (~ 54 × 104 km2) is located in Southwest China, and most of this area has experienced severe land degradation in the form of rocky desertification2,3,4. Ecological projects, such as Grain for Green, have promoted vegetation restoration in desertified areas5,6, making them one of the globally largest carbon sink hotspots7. However, nitrogen (N) limitation has hindered natural succession8, leaving an area of ~ 18.3 × 104 km2 in the grass and grass–shrub stage even after ~ 30 years of natural recovery9, severely impacting the carbon sink and ecological services in southwest China10,11.

Karst landscapes and aquifers consist of carbonate rocks wherein the fractures have been enlarged by chemical dissolution, forming a special soil–epikarst–bedrock structure12. They are characterized by unique geomorphological and hydrogeological features, such as shallow soils, exposed bedrock, developed fractures, conduits, and caves13. The soil–epikarst–bedrock structure in karst areas forms a dual hydrological system of ground and underground drainage resulting in complex and rapid hydrologic processes12,14,15,16, which may lead to a large amount of soluble nutrient loss during rainfall17, once considered a key factor in N limitation18,19.

Rainfall net N gain is a balance of N input from rainfall and soil-dissolved N loss during hydrologic processes. Rainfall provides N to the soil but also causes soil N loss through leaching20,21. Therefore, N-related hydrologic processes determine the trade-off between rainfall N input and soil N loss. However, systematic studies on these processes are lacking, limiting the analysis of the effect of rainfall on N supply in karst areas. This study was conducted to analyse changes in N from rainfall and hydrological N loss between 2018 and 2020 and determine whether rainfall induced hydrologic N losses are the key process of N limitation in the early karst succession stage. Our findings will support ecological restoration of rocky desertified areas from a rainfall N budget perspective.

Materials and methods

Site description

The study was conducted in the Duokan catchment northwest of Guangxi (23° 9′ 45″–23° 9′ 47″ N, 107° 14′ 13″–107° 14′ 30″ E), a typical karst peak cluster depression with a mean slope of 25°–35° and an elevation of 448 m asl. It has a subtropical monsoon climate, distinct dry and rainy seasons, average annual temperature of 20 °C, and rainfall amount of 1245 mm, that mainly occurs during the rainy season (76%, April–September, ~ 945 mm), with ~ 300 mm in the dry season (October–March, Fig. S1). The soil, developed from limestone, is shallow and discontinuous, with an average thickness of 10–15 cm on sloping land. The soil type is alkaline limestone soil with a texture between that of clay and clay–loam (4.1–49.7% silt and 31.8–72.4% clay). More soil properties are listed in Table 1. Previously, severe rocky desertification occurred in the area affecting 1997.9 hm2 of land, accounting for 62.9% of the total area22. Since 1993, the region has been closed for vegetation restoration through natural succession. The plant community succession has still maintained the grass–shrub stage. The dominant shrubs are Berchemia sinica, Leptodermis glomerata, and Sageretia thea, the dominant grasses are Pogonatherum crinitum, Hypolytrum nemorum, Eragrostis pilosa, and Digitaria sanguinalis. Karst endemic plants are Tirpitzia sinensis and Viburnum triplinerve23.

Table 1 Soil (0–10 cm) properties in karst grass and grass–shrub ecosystems.
Full size table

Experimental design and sampling

The study was conducted for a period of 3 years beginning in June 2018. Two hillslopes (100 ha each) 3 km apart were selected, in which six 3 m × 3 m plots were randomly set in the grass and grass–shrub ecosystems on each hillside, yielding a total of 24 plots. Leach trays (30 cm × 25 cm) were installed at a 10-cm soil depth on the downslope edge of each plot, connected to 15 L sampling bottles for soil leachate collection. Three rain gauges (RG3, Onset, America) were randomly placed on the hillslopes, connected to 30 L sampling bottles for rainfall collection. From August 2018 to April 2020, rainfall and leaching solutions were collected after each rainfall/every 0.5 month/1 month, depending on rainfall amount. Volumes were measured, and 100 mL samples were filtered, sealed in polyethylene bottles, and frozen at − 15 °C. Soil samples were collected in June 2018. Five soil subsamples (0–10 cm) were randomly collected from each plot after removing the organic layer (if available) then fully mixed and sieved (< 2 mm). Soil samples were divided into two parts: one part was air-dried to analyse soil physicochemical properties (Table 1), such as soil organic carbon (SOC), total N (TN), total phosphorus (TP), exchangeable Ca, and pH; and the second part was stored at 4 °C for analysing soil NH4+, NO3−, dissolved organic C (DOC), and dissolved organic N (DON).

Rainfall and leachate N concentration analysis

NH4+ and NO3− concentrations were measured using a flow analyser (Bran–Luebbe Inc., Germany). Total soluble N (TDN) was determined using a TOC/N analyser (Multi N/C 2100, Jena, Germany). DON was calculated by subtracting dissolved inorganic N from TDN.

Soil physicochemical analysis

Fresh soil (20 g) was extracted with 100 ml 2 M KCl shaking for 60 min at 200 rpm and 25 °C. The extracts were filtered and 10 ml of each extract was used to determine the concentrations of NH4+ and NO3− with a Continuous Flow Analyzer (Bran-Luebbe Inc., Germany). Fresh soil (20 g) was extracted with 100 ml 0.5 M K2SO4 and shaken for 60 min at 200 rpm and 25 °C. Soil TDN and DOC were determined using a TOC/N analyser (Multi N/C 2100, Jena, Germany). Soil DON was calculated by subtracting dissolved inorganic N from TDN.

A proportion of air-dried soil samples were washed using 0.5 M HCl to remove carbonate24. HCl-washed and normal soils were used to analyse SOC and TN concentrations, respectively, using a Vario EL cube element analyser (Elementar, Germany). The ground air-dried soil (5 g) was placed on slow filter paper and washed with 200 ml of 1 M ammonium acetate, and the extract was calibrated to 250 ml. The exchangeable Ca was measured using an atomic absorption spectrophotometer (4530F, China). The ground air-dried soil (250 mg) was placed in a silver crucible, moistened with alcohol, covered with 2.0 g NaOH, and placed in a Muffle furnace for alkali melting. The frit was then dissolved with 10 ml of 3 mol L−1 H2SO4 solution, transferred to the volumetric bottle for cooling, and calibrated to 100 ml. The extract was filtered with phosphorus-free qualitative filter paper, and soil total phosphorus was determined with a Continuous Flow Analyzer. Air-dried soil (5 g, sieved 2 mm) was placed in 250 ml conical bottle, extracted with 50 ml of 0.5 M sodium bicarbonate shaking for 30 min at 180 rpm and 25 °C; the extract was filtered and soil available phosphorus was determined with a Continuous Flow Analyser.

Soil pH was measured at a 1:5 (w/v) soil to water (CO2-free) ratio using a pH detector (SevenExcellence™, METTLER TOLEDO, USA). Soil texture was determined using a laser particle size analyser (Mastersizer 2000, Malvern, UK). Soil bulk density was measured using the cutting rings method.

Statistical analysis

Surface runoff N loss

$${\text{SRN}} = \frac{{{\text{R}}_{{{\text{subsurface}}}} /{\text{k}}_{{{\text{subsurface}}}} \times {\text{k}}_{{{\text{surface}}}} \times {\text{N}}_{{{\text{rainfall}}}} }}{{\text{A}}}$$
(1)

where SRN is the surface runoff N loss (mg m−2), Rsubsurface is leachate volume of 10-cm soil depth (L), and ksubsurface is the proportion of rainfall distribution in the subsurface leaching. Values of 21.5% and 41.5% for the dry and wet season (mean 31.5%), respectively, were derived from experimental results at a long-term trench runoff plot in the nearby karst region25. The ksubsurface was consistent with previous estimates of \({\text{32.1\%}}_{\text{12.4\%}}^{\text{54.4\%}}\) (\({\text{mean}}_{\text{minimum}}^{\text{maximum}}\)) in karst areas26,27. ksurface is the proportion of rainfall distribution in surface runoff, and was 1% and 7.5% in the dry and wet season (mean 4.3%), respectively25. The ksurface was consistent with previous estimates of \({\text{4.2\%}}_{\text{2.0\%}}^{\text{9.5\%}}\) in karst areas26,27. Nrainfall is N concentration of surface runoff (mg L−1). We hypothesized that the N concentration of surface runoff is equal to rainfall N concentration because surface runoff is mainly directly derived from rainfall and hardly interacts with the soil according to the mechanism of surface runoff generation28. A is the area of leach tray, which was 0.075 m2.

Subsurface leaching N loss

$${\text{SLN}} = \frac{{{\text{R}}_{{{\text{subsurface}}}} \times {\text{N}}_{{{\text{leaching}}}} }}{{\text{A}}}$$
(2)

where SLN is the subsurface leaching N loss (mg m−2) and Nleaching is the N concentration of subsurface leaching (mg L−1).

Underground leakage N loss

$${\text{ULN}} = \frac{{{\text{R}}_{{{\text{subsurface}}}} /{\text{k}}_{{{\text{subsurface}}}} \times {\text{k}}_{{{\text{underground}}}} \times {\text{N}}_{{{\text{underground}}}} }}{{\text{A}}}$$
(3)

where ULN is the underground leakage N loss (mg m−2) and kunderground is the proportion of rainfall distribution in underground leakage, which was 51% and 28% in the dry and wet season, respectively25. The kunderground was consistent with previous estimates, which range from 23.7% to 59.2%, deducting the proportion of soil storage and interception in karst areas26,27. Nunderground is the N concentration of underground leakage N loss. We hypothesized that Nunderground is equal to Nleaching because Nleaching is the initial N state losing from the soil layer to underground.

Hydrologic N loss

$${\text{HL }} = {\text{ SRN}} + {\text{SLN}} + {\text{ULN}}$$
(4)

where HL is the hydrologic N loss (mg m−2).

Rainfall net N gain

$${\text{NG }} = {\text{ NI}} - {\text{HL}}$$
(5)

where NG is the net N gain (mg m−2) and NI is the rainfall N input (mg m−2).

Statistical analyses were performed using SPSS (version 26, IBM SPSS statistics), and differences were considered significant at p < 0.05. LSD post hoc multiple comparisons were performed using one-way analysis of variance (ANOVA), if normality was indicated. For non-normal distributions, the nonparametric Kruskal–Wallis test was used. We used linear and nonlinear regressions to detect the relationships between rainfall and hydrologic N losses. Linear regression was used to analyse the relationships between rainfall N input and net N gain.

Results

Changes in soil properties

SOC, TN, and soil N/P were significantly higher in the grass − shrub ecosystem than those in the grass ecosystem, while other soil properties were similar (Table 1). Soil DON concentration was significantly higher than inorganic N concentration (p < 0.05), and NH4+ concentration was significantly higher than NO3− concentration (p < 0.05).

Changes in rainfall N input

The rainfall N input for the study period was 12.0 kg N ha−1 y−1 (Table 2). The average annual inputs of NH4+, NO3−, and DON were 3.7, 2.9, and 5.4 kg N ha−1 y−1, respectively, of which inorganic N accounted for 56.1%. NH4+, NO3−, DON, and TDN inputs all increased with increasing rainfall (Fig. S1). Rainfall N input was higher in the rainy season than that in the dry season (Table 2), and it was significant in NO3− and TDN (Fig. 1).

Table 2 Temporal dynamics of rainfall N input, hydrologic N losses, and N gain from 2018 to 2020 during the initial phase of natural recovery following karst desertification.
Full size table
Fig. 1
Fig. 1
Full size image

Rainfall N inputs in the dry and rainy seasons during the initial phase of natural recovery following karst desertification. nr and nd indicates the sample size of each N form in the rainy and dry seasons, respectively. *p < 0.05.

Changes in hydrologic N losses

Annual hydrologic TDN losses were 8.1 and 9.0 kg N ha−1 y−1 in the karst grass and grass–shrub ecosystems, respectively, and the difference was insignificant (Table S1). Taking the grass and grass–shrub ecosystems together, the average annual hydrologic TDN losses were 8.5 kg N ha−1 y−1 (Table 2), with surface runoff N loss of 0.6 kg N ha−1 y−1, subsurface N leaching of 3.5 kg N ha−1 y−1, and underground N leakage of 4.5 kg N ha−1 y−1, accounting for 7.1%, 41.2%, and 52.9% of the total N losses, respectively (Table S2). The hydrologic N losses of NH4+, NO3−, and DON were 1.2, 1.0, and 6.3 kg N ha−1 y−1, respectively, and DON accounted for 74.1% of the total N losses (Table 2). The changes were consistent between ecosystems and across seasons (Table S1). DON was the main loss component via subsurface leaching and underground leakage, but not surface runoff (Table S2).

Hydrologic TDN losses were 4.7 and 3.4 kg N ha−1 y−1 in the rainy and dry season, respectively, in karst grassland (Table S1), and NH4+, NO3−, DON, and TDN were all significantly higher in the rainy season than they were in the dry season (Fig. 2). Hydrologic TDN losses were 4.9 and 4.1 kg N ha−1 y−1 in the rainy and dry season, respectively, in the karst grass − shrub ecosystem (Table S1), and NH4+, NO3−, and TDN were significantly higher in the rainy season than they were in the dry season (Fig. 2).

Fig. 2
Fig. 2
Full size image

Hydrologic N losses in the dry and rainy seasons during the initial phase of natural recovery following karst desertification. These mainly include surface runoff N loss, N leaching of 10-cm soil depth, and underground N leakage. The difference between the grass and grass–shrub ecosystems is insignificant. **p < 0.01, *p < 0.05.

Hydrologic TDN losses were 0.5, 2.3, and 1.8 kg N ha−1 y−1 via surface runoff, subsurface leaching and underground leakage, respectively, in the rainy season (Table S2), which accounted for 10.9%, 50.0%, and 39.1% of hydrologic TDN losses. Hydrologic TDN losses were 0.1, 1.2, and 2.4 kg N ha−1 y−1 via surface runoff, subsurface leaching, and underground leakage, respectively, in the dry season (Table S2), which accounted for 2.7%, 32.4%, and 64.9% of hydrologic TDN losses. Hydrologic N losses, subsurface N leaching, and underground N leakage had a significantly correlation with rainfall, with a maximum loss at cumulative ~ 180 mm/month rainfall, after which it gradually decreased (Fig. 3). Surface runoff N loss was significantly positively correlated to rainfall (Fig. S2).

Fig. 3
Fig. 3
Full size image

Relationships between hydrologic N losses and rainfall.

Changes in rainfall N gain

Rainfall TDN gains were 3.9 and 3.0 kg N ha−1 y−1 in karst grass and grass–shrub ecosystems, respectively, and the difference was insignificant (Table S1). Taking the grass and grass–shrub ecosystems together, the average annual rainfall TDN gain was 3.4 kg N ha−1 y−1, and those of NH4+, NO3−, and DON were 2.5, 1.9, and − 0.9 kg N ha−1 y−1, respectively (Table 2). Inorganic N was mainly N gain, while DON was net N loss (Table 2). The changes were consistent between the ecosystems and across seasons (Table S1). Rainfall TDN gain was 2.1 and 1.8 kg N ha−1 y−1 in the rainy and dry season, respectively, for the grass ecosystem (Table S1), and NH4+, NO3−, and TDN were significantly higher in the rainy season than they were in the dry season (Fig. 4). Rainfall TDN gain was 1.8 and 1.2 kg N ha−1 y−1 in the rainy and dry season, respectively, for the karst grass − shrub ecosystem (Table S1), and NO3− and TDN were significantly higher in the rainy season than they were in the dry season (Fig. 4). Rainfall N input could explain 73% and 87% of the variation in net NH4+ and NO3− gain, but the relationship between rainfall N input and net DON gain was weak (Fig. S3).

Fig. 4
Fig. 4
Full size image

Rainfall N gain in the dry and rainy seasons during the initial phase of natural recovery following karst desertification. The difference between the grass and grass–shrub ecosystems is insignificant. **p < 0.01, *p < 0.05.

Discussion

Rainfall is the main N input source in natural ecosystems29, and can alleviate N limitations. Here, the annual N input by rainfall into karst grass and grass–shrub ecosystems was estimated to be 12.0 kg N ha−1 y−1 (Table 2), consistent with the results of Chen et al.30 and Zhu et al.31. The estimated atmospheric wet N deposition in the study area was ~ 12.0 kg N ha−1 y−1. N input gradually increased with increasing rainfall (Fig. S1), inducing higher N input in the rainy season compared with that in the dry season (Fig. 1). However, rainfall not only increased N input in karst ecosystems (Fig. S1) but also increased N losses through hydrologic processes (Fig. 2). We found the mean hydrologic N loss was 8.5 kg N ha−1 y−1 in the karst grass and grass–shrub ecosystems, which is consistent with previous estimates of \({8.1}_{1.2}^{26.8}\) kg N ha−1 yr−1 among grass, shrub, and forest ecosystems in karst areas26,32,33. This is substantially lower than that in non-karst ecosystems with thick soil layers in similar climate regions (50–71 kg N ha−1 yr−1)34,35. The karst soil layer is shallow and underground fissures often develop, forming an aboveground–underground binary hydrologic structure13,36,37, which was once regarded as conducive to hydrologic N losses. However, contrary to our expectations, rainfall net N gain was 3.4 kg N ha−1 y−1, and was 33.3% higher in the rainy season than it was in the dry season (Table 2, Fig. 4). Moreover, we found a strong positive correlation between rainfall N input and rainfall N gain, especially for NH4+ and NO3− (Fig. S3). This indicates that rainfall N input is a net N gain process in karst grass and grass–shrub ecosystems. Similarly, Dirnböck et al. also found that 70–83% of atmospheric N deposition was retained in karst forest ecosystem33. Noticeably, rainfall N input was mainly inorganic N, while hydrologic N losses were mainly DON (Table 2). At our study site, soil-dissolved N was dominated by DON at ~ 78.8% (Table 1), which was similar to the hydrologic N losses (DON was ~ 74.1% of TDN). This indicates that most rainfall inorganic N was retained and more DON losses originated from the soil (Table 2, Fig. S3). Thus, hydrologic processes partly controlled the changes in N exchange between rainfall and soil.

Increased rainfall generally results in more soil nutrition loss38,39, and we indeed found that surface runoff N loss increased with increased accumulative rainfall (Fig. S2). However, the special karst aboveground–underground binary hydrologic structure hinders the generation of surface runoff, which occurs only when the rainfall intensity reaches steady infiltration rate threshold of the soil − epikarst interface at ~ 40 mm h−1 (infiltration-excess), meanwhile cumulative rainfall amount reaches the epikarst water regulation capacity threshold at ~ 180 mm (saturation-excess, Fig. 5)25,28. Thus, surface runoff was only 0%–9.5% of total runoff25,26,27, and surface runoff N loss only contributed to 7.1% of hydrologic TDN loss (Table S2). Additionally, the dominated hydrologic N loss processes, such as subsurface leaching and underground leakage (Table S2), occurred a maximum N loss threshold (at ~ 180 mm) as rainfall increased (Fig. 3) due to the special karst binary hydrologic processes (Fig. 5)28,36. Meanwhile, the three-layer vertical structure of soil–epikarst–bedrock induces complex hydrologic processes in karst areas28, which can prolong the time of water–soil interaction17. These are important for reducing hydrologic N losses, especially for inorganic N (Table 2). In our study region, 76% of rainfall occurred in the rainy season (Fig. S1), while disproportionately contributing to 56.5% of hydrologic N loss (Table 2).

Fig. 5
Fig. 5
Full size image

Hydrologic N loss during the initial phase of natural recovery following karst desertification. ESR of vertical infiltration is the dominant hydrologic process under low cumulative rainfall (< ~ 40 mm), SSR of interflow is dominant under moderate cumulative rainfall (40–180 mm), and SR of surface runoff is dominant under high cumulative rainfall (> 180 mm). ESR and SSR are the main hydrologic N loss processes, which reach the maximum threshold at accumulative ~ 180 mm/month rainfall. X-axis indicates the three different dominant hydrologic processes corresponding with increasing rainfall. Y-axis indicates the special three-layer soil–epikarst–bedrock structure in karst ecosystems. The soil layer is shallow with a high conduit. The epikarst layer is highly weathered, contains multiple fissures or conduits, and is highly porous and permeable. The bedrock layer is mainly compact with low permeability. Rainfall + : rainfall increase; ESR: epikarst seepage runoff; SSR: soil subsurface flow (interflow); SR: soil surface runoff (overland flow). Arrow thickness and direction indicates the strength of hydrologic processes and water flow direction, respectively.

The geological structure of karst leads to hydrologic processes transitioning from epikarst seepage runoff (ESR), to subsurface runoff (SSR), to surface runoff (SR) with increasing rainfall (Fig. 5)28,36,40. The ESR of vertical infiltration is the dominant hydrologic process under low rainfall owing to the many fissures, conduits, and caves in epikarst (resulting in preferential flow)17,36. Rainfall infiltrates into the epikarst through the soil layer and interacts with soil-dissolved N. The epikarst is gradually saturated with water owing to the continuous increase in rainfall, and excess runoff infiltration causes underground N leakage36. Consistently, underground leakage was the dominated hydrologic N loss process in the dry season, which contributed to 64.9% of hydrologic TDN losses (Table S2). When rainfall continues to increase by > 40 mm, it accumulates in the soil subsurface and produces interflow (fill and spill)28; at this point, the dominant hydrologic process changes from ESR to SSR, which mainly involves lateral flow. Karst slopes can change the rainfall reallocation and affect N exchange between rainfall and soil17,28. When rainfall is < 180 mm, ESR and SSR are the main parts of runoff in karst areas25,41. The lag time of runoff production in ESR and SSR was long28, which prolonged water–soil interaction and increased soluble N retention, especially inorganic N (Fig. 4, Table 2). Consistently, subsurface leaching and underground leakage were the dominated hydrologic N loss processes in the rainy season, contributing to 50.0% and 39.1% of hydrologic TDN losses, respectively (Table S2). When cumulative rainfall amount reaches ~ 180 mm/month, the threshold of SR generation has been attained25, and SR becomes dominant runoff (saturation-excess)17,28,39. Surface seals decreased the infiltration rates, and rainfall does not penetrate in time, and mainly runs off through SR28,40, shortening the water–soil interactions and decreasing hydrologic N losses derived from the soil.

Our results showed that rainfall DON gain was − 0.9 kg N ha−1 y−1 (Table 2), indicating net DON loss in karst grass and grass − shrub ecosystems. Thus, rainfall can cause the risk of soil net N loss in karst areas as previous expected. However, prolonged water − soil interaction combined with the threshold of hydrologic N losses were observed to effectively protect soil soluble inorganic N under high rainfall during the initial phase of natural recovery following karst desertification. However, the karst hydrologic process is complex, and it is difficult to precisely distinguish of SR, SSR, and ESR25,28,36,40; as such, our results represent rough estimates of N loss deriving from different hydrologic processes. Together, according to our limited results, we speculated that rainfall N input may be an important N supplier in karst grass and grass–shrub ecosystems, and alleviates N limitation at the early karst succession stage.

Conclusion

Our results show that rainfall N input was 12.0 kg N ha−1 y−1, with inorganic N contributing 55.0%. Hydrologic TDN loss was 8.5 kg N ha−1 y−1, with DON contributing 74.1%. These result in a 3.4 kg N ha−1 y−1 net N gain in the initial phase of natural recovery following karst desertification, with inorganic N contributing to net N gain and DON contributing to net N loss. Complex hydrologic processes in the soil–epikarst–bedrock prolong soluble N exchange between rainfall and soil, which combined with the maximum threshold of hydrologic N losses under high rainfall can effectively protect soil inorganic N. Together, rainfall may be an important N supplier in karst grass and grass–shrub ecosystems, and may alleviate N limitation during the early karst succession stage.