1 Introduction

Soil aggregate stability is a key property that affects the water storage, soil aeration, crop production, and soil erosion [2]. Red soils (Ultisols) are highly weathered, low organic matter soil with low native soil fertility. The dominant clays found in these soils are kaolinite, and they are found in Africa, South America and the subtropical regions of China [39]. The soil structure of these soils has a high proportion of microaggregates due to relatively low organic matter content [13, 40]. The large number of microaggregates makes these soils susceptible to high erosion when left uncovered [12, 20, 29, 32].

Soil aggregate stability is influenced by many factors and can be impacted by seasonal and vegetation changes [9, 13]. As a response to changes in soil water retention, water infiltration, and soil organic matter aggregate stability can vary over the short and a long-term seasonal scales [5, 10, 21, 36]. Short-term changes can occur during a rainfall while long-term changes can occur across seasons or decades. Some studies attributed the aggregate stability variation to changes in soil organic carbon (SOC) stock under different land uses [36]. However, these changes may occur very slowly in tropical and subtropical climates [3]. Other studies suggested that soil aggregate dynamics corresponded to the variations of soil water properties [10, 21]. For example, short-term decreases in soil water content (SWC) increased soil aggregate stability on bare soil [1]. The SWC dynamics (wetting–drying cycles, wetting directions) [22, 27], and soil water infiltration [19] also influenced aggregate pore size distribution. However, SWC properties and aggregate stability can become complicated when plants are introduced into the soil.

The pattern of soil water content properties and aggregate stability varied after land use change, which depended on the impact of plants [34]. For example, Leite et al. [18] reported that soil aggregate stability was improved by switching from an annual crop to a forest. However, an opposite pattern of aggregate stability was found after reforestation from a cropland in semiarid areas, where the aggregate stability declined and the risk of soil erosion increased [20]. Differences of the relationship between aggregate stability and SWC in these studies might be attributed to the different impact of plants on soil water repellency and rainfall interception. First, plant litters (methoxyl C) and root exudate affected aggregate stability by decreasing the aggregate wetting rate due to high water repellency [6]. In addition, plants protect the soil surface by reducing splash erosion [7]. But the influence of SWC change rate under rainfall on aggregate stability was often neglected, even though [14] reported that soil aggregation breakdown was initiated through changes in SWC.

According to the aforementioned information, different pattern of SWC might occur under different rainfalls and vegetation types, however, their effect on short-term aggregate stability is not clear. Therefore, we hypothesize that transitioning from an annual cropped system to a woodland would increase soil aggregates stability. The objective of this study was to investigate the mechanism of aggregate variations in response to SWC under different short-term rainfall events under different plants (fir and Osmanthus fragrans woodlands transformed from cropland and croplands planted with rapeseed). The results will be effective to control soil and water erosion in ecological restoration projects.

2 Materials and methods

2.1 Study sites and soil description

The research site is located in Xianning County, southeast of Hubei province China. The mean annual temperature in the local region was approximately 16.8 °C and the mean annual precipitation was approximately 1300 mm, with more than 50% of the annual rainfall being received from March to August. Previously to 2004 the study area was mainly dominated by croplands. After that, many croplands were converted to woodland because rural exodus produced in the 90s. Three types of lands were selected in the research site (Fig. 1): (1) rapeseed (Brassica campestris L.) land had an area of 503 m2, and the rapeseed grew from early October to mid of May in the following year with the rest time for fallow; (2) 13-year old fir (Taxodiaceae) woodland (conversion from 2005) had an area of 1386 m2; (3) 7-year old Osmanthus fragrans woodland (conversion from 2010) had an area of 536 m2.

Fig. 1
figure 1

Schematic diagram of the research site land use

Full size image

The soil in the research site was a red soil which developed from the Quaternary red clay and was classified as Ultisols using the U. S. Soil Taxonomy system. Soil samples at 0–20 cm were collected under each plant type in fields in Oct. 2016 for basic soil characteristics analysis before the start of study (Table 1).

Table 1 The soil properties under different vegetation type in each site in 2016
Full size table
Table 2 Sampling pattern under the rainfall event and variability of the rainfall event characteristics
Full size table

2.2 Soil aggregate collection and measurement

Soil aggregate samples under each plant type were collected within a range of 2 m in diameter around the position where the soil moisture probes were installed. At least three aggregate samples were collected from a depth of 0–20 cm for each field. The aggregate samples were collected before rainfall, during rainfall and after small (0.0–10 mm day−1), medium (10.1–25 mm day−1), large (25.1–50 mm day−1), and storm (> 50.1 mm day−1) precipitation events (Table 2). The rainfall events were classified according to Precipitation Level-National Standard (GB/T 28592-2012).

Aggregate samples were kept in boxes, air-dried and broken along natural cracks for use. Aggregate samples were dry-sieved through a series of sieves (5, 2, 1, 0.5, 0.25, and 0.1 mm) to obtain each fraction of dry aggregate. Soil aggregate samples were also divided into subsamples to measure the wet aggregate stability by following the method in Le Bissonnais [17]. 10 g of soil aggregates (2–5 mm) were oven-dried at 40 °C for about 24 h to a constant weight for subsequent wet-sieving aggregate analysis. The aggregate water stability was determined by the fast wetting method using an aggregate analyzer (XY-100, Beijing) for 5 min (amplitude 2 cm, 20 oscillations min−1) following the method of Le Bissonnais [17]. The mean weight diameter (MWDfw) and the relative slaking index (RSI) were defined as shown in Eqs. 1 and 2.

$$MWD = \sum\limits_{i = 1}^{n + 1} {} \frac{{r_{i - 1} + r_{i} }}{2} \times m_{i}$$
(1)

where r = aperture of the ith mesh (mm), r0 = r1, and rn= rn+1; mi = mass fraction of aggregates remaining on ith sieve; n = number of sieves. The results of the test for fast wetting were refereed as MWDfw.

$$RSI = \frac{{\mathop {MWD}\nolimits_{SW} - \mathop {MWD}\nolimits_{FW} }}{{\mathop {MWD}\nolimits_{SW} }} \times 100$$
(2)

where MWDSW and MWDFW are the MWD in the treatment of slow wetting and fast wetting methods, respectively.

2.3 Soil water content measurement

The volumetric SWC measurement in all fields was monitored from Nov. 2016 to June 2018 to display the variation of SWC. In Oct. 2016, moisture probes (Decagon Devices Inc., Pullman, WA) were installed in the middle of each field and were connected to a datalogger (Watchdog 2400, Spectrum Technologies, Inc.). The moisture probes were installed to different depths, but only 0–20 cm SWC data in a frequency of 30 min was used in this study. Before the sensors were installed, calibrations were conducted. SWC variations across five different rainfall events were displayed by coefficient of variation (CV). Further SWC conditions were expressed by two indices which included the mean SWC for a duration t (in days) prior to the soil aggregate sampling (θt) and the differences in SWC between the beginning and the end of that period (Δθt). In this study, θ0 (at the sampling day), θ0.5 (half day prior to sampling), θ2, θ4, Δθ0.5, Δθ2, and Δθ4 were also analyzed. Meteorological data were obtained from a weather station 1.5 km from the study site. A separate rainfall gauge was installed under the fir field to monitor difference of the amount of rainfall under trees and the bare field.

2.4 Plant root analysis

To measure the water uptake ability of plants, the root distributions of different types of plants were determined in April 2016 and Aug. 2017. Roots were taken to a soil depth of 0–20, 20–40, and 40–60 cm using cores (200 cm3) for fir and Osmanthus fragrans fields at the position of 25 cm and 50 cm to the base of trees. Rapeseed roots were also taken to soil depth of 0–20, 20–40, and 40–60 cm at a point with row distance of 25 cm. Roots were washed by water, stored in − 20 °C refrigerator and scanned by the LA2400 Scanner for images (Regent instruments Canada Inc.) [28]. After that, root length and diameter were analyzed by WinRhizo, and root length density (RLD) was calculated. Finally, relative root absorption of soil water rate (RASWr) was calculated based on the relative absorption of soil water (RASW) as in Eq. 4 [38].

$$RASW_{{\left( {\text{z}} \right)}} = \frac{{\theta_{actual} - \theta_{pw} }}{{\theta_{fc} - \theta_{pw} }}$$
(3)
$$RASW_{{r\left( {\text{z}} \right)}} = RASW_{{\left( {\text{z}} \right)}} \times \frac{{RLD_{z} }}{{RLD_{\hbox{max} } }}$$
(4)

where z was a certain soil depth, θactual was the actual measured SWC by moisture probe, θpw was the wilting point of soil, θfc was the field capacity of the soil. RLDz/RLDmax was the relative root length abundance (ratio of RLD at depth of z to maximum RLD).

2.5 Statistical analysis

Soil aggregate MWDfw differences with the time steps of each rainfall event for all plant types were tested using the Fisher’s least significant difference (LSD) test (p < 0.05) in SPSS 17.0. The Pearson correlations between MWDfw, rainfall duration, rainfall amount, the maximum rainfall intensity, θ0, θ0.5, θ2, θ4, Δθ0.5, Δθ2, and Δθ4 were all performed. All the Figures were plotted by Origin 8.0.

3 Results

3.1 Short-term variation of soil aggregate stability under rainfall events

Soil MWDfw varied with time under the five different rainfall events (Fig. 2). Soil MWDfw dropped to a minimum value on the rainfall event day and became stable or slightly returned to the original value with time after the rainfall stopped. For example, under the large rainstorm on Aug. 2017, MWDfw declined to minimum values of 0.22 mm, 0.26 mm and 0.32 mm for Osmanthus fragrans, rapeseed, and fir, respectively, which displayed a relative decline of 37.9% (Osmanthus fragrans), 38.7% (rapeseed), and 13.9% (fir) compared to their original MWDfw values (Fig. 2). For the large rainfall event on Sep. 2017, MWDfw displayed a similar pattern as the August large rainstorm except that the fir soil had a greater decrease (30%). MWDfw exhibited a smaller degree of change with time at medium and small rainfalls. The decline pattern of MWDfw with rainfall events was confirmed by the significant correlation between rainfall duration, rainfall amount, maximum rainfall intensity, and aggregate size fraction (Supplemental Table 1).

Fig. 2
figure 2

Variation of aggregate MWDfw with time under different rainfall events. Different lower-case letters indicate significant differences of MWDfw with time steps under each plant field

Full size image

The best correlation coefficient was found between the rainfall amount and aggregate size fraction (0.5–0.25 mm) as r of − 0.66, − 0.91, and − 0.84 for Osmanthus fragrans, rapeseed, and fir, respectively (Supplemental Table 1). This indicates that rainfall amount negatively impacted macroaggregate (> 0.25 mm) after rainfalls. For example, the soil macroaggregate proportion under fir decreased from 66.5% (before rainfall: 17 Sep., 2017) to 45.5% (after rainfall: 19 Sep., 2017). Correspondingly, the microaggregate fraction significantly increased during the same period. The aggregate MWDfw change with rainfall effect can be indicated by different RSI values for each plant field (Fig. 3). Significant increase of RSI with time after rainfall was in agreement with the decline pattern of MWDfw with time under rainfall in Fig. 2 for all lands. MWDfw change in a short-term among plant types was similar as that in a year time (Fig. 4).

Fig. 3
figure 3

Soil RSI under different rainfall events. Different lower-case letters indicate significant differences of RSI with time steps under each plant field

Full size image
Fig. 4
figure 4

Soil aggregate properties: a aggregate size distribution for example samples taken on 1 Nov. 2017, b soil MWDfw comparison among the three plant types over a year time, different letters indicate significant differences of MWDfw within type of plant at each time

Full size image

3.2 Soil water dynamics with rainfall under different plants

3.2.1 Soil water content dynamics with rainfall

The SWC values also varied in response to rainfall events, which displayed an opposite trend compared to the trend of MWDfw for all plants (Fig. 5). The SWC values increased sharply with the beginning of rainfall followed by a gradual decline to relatively stable values after rainfall stopped, but SWC displayed different extent of variation among rainfall types. The variation of SWC was largest in a rainstorm (i.e., rapeseed coefficient of variation (CV) = 11.1%) and the smallest in a small rainfall (CV = 0.56%). Variation of SWC in response to the rainfall event was also expressed by Δθ0.5, Δθ2, and Δθ4. Considering all these parameters, variations of SWC were also different among the land use types, with the most for rapeseed and the least for fir under the same rainfall.

Fig. 5
figure 5

Soil water content variation with time under different rainfall events

Full size image

3.2.2 Plant root properties relationship with SWC

Root length distribution and root length density (RLD) displayed differently within depths among the three plant types (Fig. 6) and were responsible for the difference of SWC among plants. Among the root length, root diameter < 1 mm was dominant in different plants, where rapeseed had a higher percentage (> 70%) of this type of root than fir and Osmanthus fragrans. However, root length (d = 1–3 mm) of rapeseed was approximately 20% lower than that the other plants. Besides, high RLD appeared at surface 20 cm of soils for all plants (Fig. 6). Such root length distribution and RLD resulted in different root relative absorption of soil water (RASWr) values among soils. For example, RASWr was 41.6%, 79.1%, and 19.5% for Osmanthus fragrans, rapeseed, and fir, respectively in 2017. The effect of plant root properties on SWC was especially obvious for the percentage of plant roots (d = 1–3 mm) (0–20 cm) which was indicated by a significant correlation with SWC with r values of 0.99 (fir), 0.99 (Osmanthus fragrans), and 0.63 (rapeseed) (data not shown).

Fig. 6
figure 6

Plant root length distribution and root length density (RLD) at different soil depths for three types of plants

Full size image

3.3 Relationships between aggregate stability and soil water content

Above soil water indices including soil water content (θ) values and dynamics (Δθ) were dominant factors of soil aggregate stability, because they played different roles on aggregate fractions (Table 3). For example, θ0 and θ0.5 were significantly and negatively correlated with aggregate fraction (0.5–0.25 mm) for rapeseed and Osmanthus fragrans lands. However, θ0 and θ0.5 were significantly and positively correlated with aggregate (0.25–0.1 mm, 0.1–0.053 mm, < 0.053 mm) fraction for rapeseed and Osmanthus fragrans lands. Especially, θ0.5 was correlated with aggregate (0.5–0.25 mm%) as − 0.66 and − 0.66 for rapeseed and Osmanthus fragrans lands, respectively. In addition, Δθ2 and Δθ4 positively contributed to the formation of microaggregate (< 0.1 mm) for fir land, but was not significantly correlated with any of aggregate fractions for rapeseed and Osmanthus fragrans lands. After the variation of aggregate size distribution due to SWC, soil MWDfw became different among plants.

Table 3 Correlations between aggregate parameters and soil water indices
Full size table

4 Discussion

4.1 Soil aggregate varied before and after rainfall event

Temporal change in MWDfw can occur over a few days, depending on rainfall and vegetation. Similar short-term change of aggregate stability (up to 46%) was also observed over a 7-days period on bare soils (0–0.5 cm) in Algayer et al. [1]. Rainfall generally played important roles in inducing aggregate destruction, soil crusting and soil loss due to throughfall kinetic energy [16], however, differences in the amplitude of aggregate change under rainfall in studies were probably associated with soil coverage. For example, in our study, rain duration and rainfall amount appeared to be the dominant factors controlling aggregate stability over all rainfall events, instead of the rain intensity in Algayer et al. [1] and Sajjadi and Mahmoodabadi [25]. The differences in studies were attributed to the different soil coverage. Study in Algayer et al. [1] was conducted on bare field where soil was more susceptible to the splash erosion from high velocity of raindrops [14]. Our present study underlined the role of plants to reduce rainfall splash effects on aggregation, for example, rainfall water that was received under fir (measured by a rainfall gauge at site) can be reduced by 28–68% compared to that on bare soils.

The change pattern of rapid decline of aggregate (MWDfw) following rainfall initiation and then gradual increase following rainfall termination (drying process) was due to below mechanisms. Firstly, MWDfw declined at the start of the rainfall, because it provoked a wetting process of soil with larger rainfall amount resulting in more slaking destruction of soil aggregates. This was confirmed by a decrease of aggregate (2–0.25 mm) fraction and an increase of microaggregates (< 0.25 mm) in our study, which was in agreement with [25]. Destruction of macroaggregate from rainfall effect outweighed the aggregate formation, resulting in the net decline of MWD during the rainfall day [21]. Secondly, aggregate MWDfw improved again with time after rainfall stopped, and almost returned to the original value before rainfall event. At this time, the aggregates that were dominated by particles < 0.25 mm were generally more susceptible to be released in next rainfalls to block soil pores, affecting soil permeability and surface runoff [11], and soil sensitivity to erosion [17, 31]. These results confirmed the negative effect of rainfall duration and amount on aggregate size and the positive effect of after-rainfall periods on aggregate size.

4.2 Soil aggregate as controlled by soil water content and plants

Variation of aggregate MWDfw in the above rainfall events can be attributed to the significant change of SWC as adjusted by the plant types. Firstly, the significantly negative relationship between the fraction of mesoaggregate (1–0.25 mm) and SWC (θ0 and θ0.5) confirmed that SWC values at the sampling time (θ0) and half day prior to sampling time (θ0.5) might break the extent of the mesoaggregate (1–0.25 mm). The results were in agreement with [13] who also attributed the aggregate destruction to the antecedent SWC. When the antecedent SWC was low, the slaking was the major reason for aggregate breakdown, but when the antecedent SWC was high, non-uniform expansion of soil minerals was the dominant factor for aggregate breakdown [22]. In addition, the dynamics of SWC (Δθ) also broke mesoaggregate probably due to the wetting–drying cycle effect [22]. Moreover, dynamics of SWC as influenced by the rainfall type was responsible for different MWDfw under each plant field. For example, the improvement of SWC after rainfalls resulted in the decline in aggregate stability through the loss of interparticle cohesion [26] or slaking [37]. In contrast, the decline of SWC (drying process) increased the aggregate stability due to the formation of bonds between particles [15]. Generally, the SWC had dominant effect on the soil aggregation, which was different among plant types.

Plants impacted aggregation also through modifying the SWC distribution as below approaches. Plant regulated SWC distribution through functions of interception of rainfall by vegetation coverage, absorption of water by roots, and change of water infiltration. Firstly, the canopy intercepted large amount of rainfall and reduced the SWC dynamics [7, 16]. In our study, high canopy of fir with small leaves reduced 28–68% of rainfall water compared to that on the bare soil surface, responsible for its lowest SWC variation and the highest aggregate MWD. Similar plant effect on the relationship between SWC and aggregate was reported in Linsler et al. [21] that permanent grassland soils (11 years) with low SWC in dry season resulted in high large water-stable macroaggregate proportion in surface soils, while cropland transformed from grassland with high SWC variations leaded to more aggregate destruction. Secondly, different root distribution among three plants fields also displayed different RLD and RASWr to regulate SWC regimes (r between RLD and SWC > 0.99). Different RASWr among plants (rapeseed > Osmanthu fragrans > fir at surface 20 cm soils), together with different plant roots exudates and litters properties probably exhibited different soil water repellency to control aggregate stability [6]. Thirdly, different soil water infiltration due to different cushions of rainwater, velocity of raindrops, and soil pore distribution, was responsible for different SWC dynamics. For example, Pan et al. [24] found that the tree cover systems with 86% of soil water infiltration coefficient caused less SWC variation than the control (68%). In our study, large fraction of small pores (< 30 μm) existed in the fir soils, while many 30–500 μm transmission pores existed in the other two soils, which may result in different magnitude of infiltration and SWC variation during wetting and drying process (Supplemental Fig S1). Generally, all these factors may explain the most intensive SWC variations under rapeseed and smallest variations under fir, and therefore, these resulted in the least MWD under rapeseed while highest MWD under fir.

Except for SWC effect on aggregation, average soil organic carbon (SOC), as a binding agent of aggregate [8], displayed no significant differences with the time of rainfall among each aggregate fraction (data not shown). Our study was different from Yu et al. [36] that reforestation increased C sequestration, which was probably because short-time land transformation was not yet effective in changing SOC in subtropical climate in our study. Hence, SOC was not the dominant factor in determining aggregate stability in short-time rainfall in our study. Besides, free Fe-oxides, as another important binding agent of aggregate in red soil [35], was significantly greater in fir field than rapeseed field (2.1 vs. 1.1 g kg−1), but will be difficult to change in a few days. Therefore, soil aggregate stability was dominantly determined by above SWC properties under short-term rainfall as influenced by plant types instead of SOC and oxides.

Large rural-to-urban mitigation occurred in 60.2% of rural lands in China in 2016 [23], creating a great land use change in rural regions. Similar land use change during rural exodus was also reported in other developing countries such as Brazil, Laos, Kenya etc., influencing environmental healthy [16, 33]. During the land use change, other authors also found that the water erosion was generally lower over afforested areas [16], aggregate stability was higher in afforested areas [18], and that soil microbial biomass, microbial C, N, P increased when a similar situation was produced in a mixed forest areas in India [30]. Soil aggregate stability was inversely related with the soil erodibility index K [4]. Therefore, understanding the aggregate variation mechanism with SWC under different plant type and time of land use transformation will be important to control soil and water erosion in rainy season. But the effect of single woodland was still limited to resist the soil erosion, high coverage (tress with high canopy and small leaves) combined with minimum height of grass underneath trees was suggested to be the most effective structure to conserve soils [16].

5 Conclusions

Soil aggregate stability varied in response to SWC as influenced by short-term rainfall events at all plant fields. Variation of aggregate MWDfw with SWC was due to different change of aggregate fraction and RSI under rainfall event. The SWC (θ0 and θ0.5, on the sampling time and half day prior to the sampling time) was the dominant factor to determine the aggregate stability which was confirmed by the significantly negative relationship between θ0 and θ0.5 and aggregate fraction (0.5–0.25 mm). Such SWC effect on aggregation with time depended on the process of rainfall and was significantly different under rainfall duration and amount. The SWC effect on aggregation was also adjusted by different plant root distribution and RASWr which was confirmed by significant correlation between SWC and plant roots percentage (d = 1–3 mm) at 0–20 cm. Generally, aggregate stability was determined by the variation of SWC, with the highest aggregate stability for fir field (least SWC variation) and the least for Osmanthu fragrans field (highest SWC variation). The results will provide theoretical basis during land use transformation considering the soil and water protection. Coniferous woodland will be suggested to improve the aggregate stability to prevent water erosion during ecological restoration for policymakers.