确定华南湿润亚热带地区坡耕地径流和土壤流失的不同降雨类型下的最佳山脊做法

A R T I C L E I N F O

Handling Editor Dr R. Thompson

Keywords:

Water erosion

K-means clustering

Soil and water conservation Red soil region

A B S T R A C T

The relationship between different rainfall types, different soil management practices, and soil erosion is not yet fully understood. In-situ observations of soil and water loss on sloping farmland in the red soil region of southern China with a subtropical monsoon environment were taken at 12 runoff plots with four treatments, i.e, down- slope ridges, downslope ridges with hedgerow intercropping, contour ridges, and bare flat land as control, over a seven-year period from 2012 to 2018. During this time, 253 natural rainfall events were classified into three rainfall types by K-means clustering according to the rainfall depth, maximum-30 min rainfall intensity and rainfall duration, and surface runoff and soil erosion processes in relation to the rainfall types under different ridge practices were analyzed. The results show that water-induced soil erosion on the flat land control was

significant, with average annual soil loss of 76.73 t⋅ha—1⋅yr—1, reaching the “intense erosion” classification, and

ridge practices were confirmed to reduce annual runoff and soil loss in all rainfall events by 18.9–62.0% and 68.9–86.3%, respectively. On the whole, rainfall events can be divided into three types: intense, normal, and long-duration. Among them, intense and normal rainfall cause the majority of soil (88.5–93.7%) and water (75.0–83.8%) loss in this area, but the efficiencies in runoff and soil reduction during long-duration rainfall events were the lowest, or even negative on farmlands with only downslope ridges. 20% of the total rainfall events, in which 84.3–92.2% were intense and normal rainfall events, contributed to 29–33% of the total rainfall depth, 68–89% of the total runoff depth, and 94–98% of the total soil loss. Rainfall depth played a dominant role in generating runoff, while runoff accumulation was a main factor influencing on soil loss. Findings from our

study indicate that by choosing a more appropriate ridge practice according to different rainfall types, there can be a positive effect on soil and water conservation.

1. Introduction

Maintaining healthy soil is the foundation of agriculture and is an essential resource to ensure human needs (Amundson et al., 2015), such as food, feed, fiber, clean water and clean air. Soil is a vital part of functioning earth systems that support the delivery of primary ecosystem services (Borrelli et al., 2017; Keesstra et al., 2016a), but is threatened by accelerated soil erosion (Borrelli et al., 2017). Water-induced soil erosion is a main cause of land degradation and declines in productivity around the world (Prosdocimi et al., 2017; Wei et al., 2007). In the last 10 years, the total amount of global soil erosion increased by 2.5% due to changes in land use/cover, with soil loss

ranging from 1 to 2 orders of magnitude higher than the soil formation rate (Borrelli et al., 2017; Amundson et al., 2015). Soil erosion seriously limits agricultural productivity (Mekonnen et al., 2015; Teshome et al., 2016), causes declining soil fertility (Teressa, 2017; Moges et al., 2012) and significantly reduces crop yields (Erkossa et al., 2015; Haileslassie et al., 2005). Soil and water conservation practices (SWCPs) play a vital role in reducing soil erosion and enhancing sustainable land manage- ment, thereby helping to achieve the United Nations sustainable development goals and to deal with the land degradation neutrality challenges (Guadie et al., 2020). The red soil region in southern China is an important agricultural production area (Huang and Zhao, 2014), accounting for 36% of the total cultivated land and for more than 50% of

* Corresponding author at: Jiangxi Academy of Water Science and Engineering, Nanchang, Jiangxi 330029, China.

E-mail address: zhj@jxsl.gov.cn (H. Zheng).

1 These authors contributed equally to this work.

https://doi.org/10.1016/j.agwat.2021.107043

Received 19 January 2021; Received in revised form 16 May 2021; Accepted 18 May 2021

0378-3774/© 2021 Elsevier B.V. All rights reserved.

the total oil-bearing crops, grain, and agricultural outputs from China (Huang, 2017). Due to high precipitation, hilly landforms, and unsus- tainable farming practices, the red soil region has been facing severe soil and water loss (Barton et al., 2004; Fang et al., 2017). According to Liang et al. (2010) and Zheng et al. (2021), sloping farmland is the

primary source of soil erosion, and the rate of soil loss can reach 98.78 t ha—1 yr—1 in the red soil region. Currently, research examining soil

erosion and suitable control measures for sloping farmland have been successfully undertaken in the region (Shi et al., 2014; Yang et al., 2013; Fang et al., 2017; Dai et al., 2018). However, due to obvious differences in topography, soil types, crop types and farming habits in different study areas, findings from previous investigations report different re- sults, and long-term field observation experiments are still insufficient. Soil erosion processes on sloping farmland are affected by different ridge practices that change the surface micro-topography. At present, research findings about the effect of surface morphology on erosion process are basically consistent. The surface morphology intercepts rainwater through the hollows, changing the source and sink processes and the paths by which water is gathered and dispersed, in turn affecting the surface runoff and soil loss (Appels et al., 2016). Furthermore, compared with flat terrain, concave terrain makes the surface runoff and sediment generation process have a more obvious boundary effect. The greater the undulation and roughness of the surfaces, the more signifi- cant the time delays of surface runoff and sediment generation (Zhao et al., 2018). Many studies have shown that by applying ridge practices on sloping farmland (e.g. ridge hayrick field, tied ridges, and downslope ridges plus contour hedgerows), soil erosion is significantly reduced (Ngetich et al., 2014; Xie et al., 2015; Grum et al., 2017; Yang et al., 2013; Dai et al., 2018). For example, Grum et al. (2017) carried out field experiments during the rainy seasons in arid and semi-arid environ- ments and showed that tied ridges significantly reduced runoff by 56% and improved soil-moisture content compared with normal agricultural practices without ridge treatments. Ngetich et al. (2014) reported that tied ridging was the most efficient measure among selected tillage practices in reducing runoff and sediment yield in semi-arid regions. Compared with flat tillage, ridge tillage is widely used in the red soil region of southern China with humid subtropical environments, where the alternation of dry and wet soil is frequent, and can improve the temperature, light, water, and fertilizer conditions during crop growth. However, poorly-chosen ridging practices are an important cause of high surface runoff and soil erosion. Even with maximum vegetation coverage of 80%, poor choices in ridging can still cause serious soil erosion (Liang et al., 2010). A comparison between different ridge management strategies on sloping farmland need to be examined in order to determine best practices for reducing erosion and runoff

damage.

Rainfall is another primary factor influencing runoff generation and soil erosion (Truman et al., 2007; Girmay et al., 2009; Huo et al., 2020). The impacts of rainfall characteristics (e.g. rainfall intensity, duration, depth and type) on runoff and soil erosion have been and continue to be extensively studied (de Lima et al., 2003; Kinnell, 2005; dos Santos et al., 2017; Chen et al., 2018). Among important rainfall characteristics, rainfall type is the complex primary factor influencing hydrological and erosion processes (Ran et al., 2012). Results from past studies show that due to differences in climate, rainfall in different regions is classified differently and has unique relationships with soil erosion. The signifi- cant relationship between rainfall-runoff-erosion processes is also significantly impacted by management practices (Mchugh et al., 2007; Dagnew et al., 2015). At present, the classification of rainfall types for the study of soil erosion laws in China is mostly found in the Northwest Loess Plateau, which has a semi-arid environment (Wei et al., 2007, 2014; Ran et al., 2012; Fang et al., 2008). The rainfall in the Loess Plateau can be divided into three regimes: short duration and high in- tensity rainfall, medium duration and medium intensity rainfall, and long duration and low intensity rainfall. Among them, short duration and high intensity rainfall is the main cause of soil erosion in this area.

The red soil region of southern China experiences a humid environment with annual average rainfall between 1200 and 2000 mm, and has the greatest rainfall erosivity in China, but there are few detailed studies about the effect of rainfall types on soil and water loss (Zhang et al., 2011; Qin et al., 2015). For example, Huang et al. (2010) used the rapid clustering method to divide erosive rainfall into four types, and selected three of those types to compare the runoff and sediment yield changes of different vegetation restoration models. Another example of relevant research is that Liu et al. (2016) classified 393 rainfall events into five regimes by the K-means clustering method, and analyzed surface runoff, subsurface flow, and sediment loss processes in relation to the rainfall regimes under three land cover types. As previously stated, the red soil region in southern China has the highest rate of rainfall induced erosion in the country, and the management practices used on the sloping farmland are diverse. The relationship between different rainfall types, different soil management measures, and soil erosion remains unclear. The goal of this study was to evaluate the effects of various ridge practices and rainfall types on surface runoff and soil loss on sloping red soil farmland with a humid subtropical environment in Southern China. The specific objectives were: (1) to compare the differences in surface runoff and soil loss under different ridge practices; (2) to determine the responses of surface runoff and soil loss under different rainfall types; and (3) to study the role of different ridge practices on soil erosion control under different rainfall types. Findings from this study could ultimately lead to a better management of agricultural systems in the red

soil region of southern China.

2. Materials and methods

2.1. Study area

The study was carried out at the Jiangxi Ecological Park of Soil and Water Conservation in the Yangou watershed (115◦42′38–115◦43′06′′E, 29◦16′37–29◦17′40′′N), De’an County, Jiangxi province, China. The

study area was 15 km away from Poyang Lake, which is the biggest

freshwater lake in China. The area has a subtropical monsoon climate, with a mean annual temperature of 16.9 ℃ and a mean annual frost-free period of 249 d. Mean annual precipitation in this area is 1449 mm, with precipitation between April and September accounting for about 70% of the total annual precipitation. The topography is generally characterized

by low hills with elevations of 30–100 m. The slopes across the area range from 5◦ to 25◦. This watershed belongs to the typical plain land-

form of a lake area, with a concentrated distribution of sloping farmland. Crop rotations of winter rapeseed (Brassica napus) and summer peanut (Arachis hypogaea L.) account for the use of the sloping farmland. The soil, dominated by a red soil developed from quaternary red clay, is acidic to slightly acidic. According to the U.S. Soil Taxonomy, the dominant soil is generally classified as Ultisol. The total organic matter, total nitrogen, total phosphorus and total potassium contents of the

tested soil (0–20 cm) are 2.32, 0.24, 0.25, and 12.53 g kg—1, respec-

tively; its clay, silt, and sand contents are 46.85%, 47.39%, and 5.76%, respectively (Zheng et al., 2021). Natural resources and land use types in the Yangou watershed are typical of the red soil region in southern China.

2.2. Experimental design and agricultural practices

Twelve in-situ runoff plots were constructed on a waste grassland with a 10◦ gradient with similar soil conditions. Each plot was 20-m-

long 5-m-wide, and all plots were adjacent and parallel to the slope. In order to retain external runoff, each plot was surrounded by brick walls that were plastered with cement mortar. These walls were inserted 30 cm into the ground and protruded 20 cm above the surface. A rect- angular drainage trough and a set of circular collecting buckets were installed on the lower slope of each plot to collect surface runoff and eroded sediment. A water-level gauge installed in each bucket was used

to record the runoff volume for each rainfall event. The construction of runoff plots was finished at the beginning of 2011.

The twelve runoff plots were arranged in a randomized complete block design with the following treatments: downslope ridge tillage (DT), downslope ridge tillage with hedgerows planted in contour (DT HG), contour ridge tillage (CT), and bare flat land as control (CK) (Fig. 1). There were five planted ridges along the slope of each DT and DT HG plot, and 20 planted ridges that contoured the slope of each CT plot. Each planted ridge was 70 cm wide, 20 cm high, and 200 cm long. In each DT HG plot, Day-lily (Hemerocallis citrina) hedgerow was established every 5 m along the slope by transplanting seedlings on October 26, 2011. Day-lily was planted in two rows at a spacing of 20 30 cm in each belt. In order to reduce surface disturbance, the hedge belts were maintained without tillage during the study period. Bare flat land was employed as the control treatment and the weeds in the runoff plots were manually removed every month.

According to local planting practices, peanut (summer) and oil rapeseed (winter) were selected as crops for annual rotation. Peanut (Arachis hypogaea L.) was planted in each plot in late-April and har-

vested in late-August. Oil rapseed (Brassica napus) was transplanted in late-October and harvested in late-April. The ridge’s surface was tilled in ridged plots and on the whole surface in CK plots. Soil was ploughed using a hoe to a depth of about 20 cm. Peanuts were sown at a density of about 100,000 holes per hectare with two seeds per hole, and oil rape-

and sediment concentration. Notably, if the drainage trough had accu- mulated bed load, the dry weight of bed load was calculated after sampling and drying, which was included in the total sediment loss of the erosive rain event. Precipitation was recorded using an automatic meteorological situated station 20 m from the experimental plots. The start and end time of a rainfall event, the rainfall intensity and the rainfall depth were recorded. A total of 264 observations were recorded during the entire experiment. Among them, 11 rainfall events were excluded from analysis as they occurred across holidays or festivals during which observations were not made consistently and were diffi- cult to distinguish from independent erosive rainfall events.

2.4. Data processing

In order to assess the effects of ridge practices and rainfall types, surface runoff and sediment yield data (runoff depth and soil loss) were collected for 253 erosive rainfall events from 2012 to 2018. In this study, K-means clustering was used to classify rainfall types according to rainfall depth (P), maximum-30 min rainfall intensity (I30), and rainfall

duration (D). The classification met the ANOVA criterion for signifi- cance (P < 0.05).

Runoff depth (R) and soil loss (S) were calculated using the following formulas:

seed was transplanted at a density of about 67,000 plants per hectare. The cropping system and cultivation practices used were applied ac- cording to local typical agricultural management in the red soil region of southern China.

2.3. In-situ observations

The observations of runoff and sediment under natural rainfall began in April 2011. To moderate the disturbances caused by plot construction, surface runoff and sediment yield data (runoff depth and soil loss) used in this paper began in 2012. A rainfall event was not considered to be independent unless the rainfall interval was over 6 h (Renard et al., 1997). Surface runoff volume of each runoff event was calculated ac- cording to a pre-determined water level-volume ratio using the water-level gauge installed in the runoff bucket. After each rainfall event, a runoff sample mixed with sediment was collected in an

aluminum box, and sediment concentration was determined by drying the sample in an oven to a constant weight at 105 ◦C. The sediment loss

of each rainfall event was calculated by multiplying the runoff volume

Fig. 1. Study site and runoff plots of different treatments. Note: a and b represent collecting buckets and drainage trough, respectively; CK, DT, DT+HG and CT represent bare flat land as control, downslope ridge tillage, downslope

ridge tillage with hedgerow intercropping, and contour ridge tillage, respectively.

R = A × 1000 (1)

where, R was runoff depth (mm), Q was the runoff volume (m3), and A

was the runoff plot area (100 m2).

S = A × 10 (2)

where, S was soil loss (t ha—1), and E was erosive sediment (kg).

Runoff and sediment reduction of ridge management were calculated according to the Comprehensive Control of Soil and Water

Conservation-Method of Benefit Calculation (GB / T 15774–2008) of

China, as:

ηR = R0 — Rm × 100% (3)

ηs = S0 — Sm × 100% (4)

where, ηRand ηs are runoff reduction and sediment reduction, respec- tively; Rm and R0 are the runoff depth of ridged management plots and bare control plots (mm), respectively; Sm and S0 are the soil loss from

ridged management plots and bare control plots (t ha—1), respectively.

The Kolmogorov-Smirnov test showed that the data were normally or log-normally distributed. The one-way ANOVA test was used to test for differences in surface runoff and sediment yield between different management types and rainfall types. Spearman correlation was used to illustrate the correlations between the runoff depth, the soil loss, and rainfall characteristics. SPSS 17.0 was used to perform statistical anal- ysis, and OriginPro 2019b was used to draw charts.

3. Results

3.1. Description of erosive rainfall events

The 253 erosive rainfall events used in this study were divided into three groups by K-means clustering based on the rainfall depth, the maximum-30 min rainfall intensity, and the rainfall duration (Table 1). Rainfall type III was characterized by intense rainfall with the shortest duration (325 min) and the lowest rainfall depth (22.6 mm), but the

highest maximum-30 min rainfall intensity (22.7 mm h—1). Rainfall

type III occurred 108 times, which was the highest frequency of any

Table 1

Statistical characteristics of different rainfall types.

1.15

0.98

1.04

D(min) 325 188 1.72 35,080

I30(mm⋅h—1) 22.7 16.7 1.36

Note: P, rainfall depth; D, rainfall duration; I30, maximum 30-min rainfall intensity; SD, standard deviation.

rainfall type. Rainfall type I, long-duration rainfall, included rainfall events that had the greatest depth (51.7 mm), the longest duration (1888 min), and the lowest maximum-30 min rainfall intensity

(15.5 mm h—1), but also the lowest frequency, with only 53 such events

observed. Rainfall type II, characterized by normal rainfall, had a moderate depth, moderate duration, and moderate maximum-30 min rainfall intensity; this rainfall type occurred 92 times during the seven years. Considering seasonal distributions (Fig. 2), rainfall type I mainly occurred in spring (from March to May) and autumn (from September to November) with the total occurrence of 31 times and 58.5% probability; rainfall type II mainly occurred in spring with the highest seasonal occurrence of 47 times and 51.1% probability; and rainfall type III occurred both in spring and summer (from June to August), with 46 and 45 times and probabilities of 42.6% and 41.7%, respectively.

3.2. Runoff and soil loss under different ridge practices

Fig. 3 presents the results of average annual runoff and soil loss from

the four treatments from 2012 to 2018. It was clear that the runoff and soil loss were strongly affected by ridge practices. Runoff depth was the highest and the lowest in CK and CT, with average annual amounts of

179.86 and 68.40 mm, respectively. The average annual runoff depth in

DT and DT HG was 145.79 and 84.34 mm, respectively. Runoff depth was significantly (P < 0.05) lower in CT and DT GH compared with CK and DT. The highest average annual soil loss was found in CK

(76.73 t ha—1 yr—1), which had the poorest surface cover conditions,

followed by DT (23.86 t ha—1 yr—1). The lowest average annual soil loss was found in CT and DT HG (both with 10.49 t ha—1 yr—1). Soil loss was

significantly (P < 0.05) lower in all ridge treatments than from the bare flat control. Significant (P < 0.05) differences were also shown in the average annual soil loss between DT and DT+HG, CT.

Fig. 3. Average annual runoff depth and soil loss on the sloping farmland for four treatments in 2012–2018. Note: Different letters indicate significant dif- ferences among four treatments according to one way ANOVA and LSD test (p < 0.05). CK, DT, DT+HG and CT represent bare flat land as control, down- slope ridge tillage, downslope ridge tillage with hedgerow intercropping, and

contour ridge tillage, respectively.

3.3. Runoff and sediment loss from different rainfall types

Fig. 4a shows that mean annual runoff depth for the CK treatment demonstrated a significant decreasing trend (P < 0.05) from rainfall type II > III > I. Rainfall types II and III created significantly (P < 0.05)

Fig. 2. Frequency (a), rainfall depth (b) and I30 (c) of different rainfall types under different seasons. Note: Rainfall type I, II, III and All represent rainfall events of long-duration rainfall, normal rainfall, intense rainfall, and all rainfall, respectively.

Fig. 4. Average annual runoff depth (a) and soil loss (b) on the sloping farmland for four treatments under three typical rainfall types in

2012–2018. Note: Different letters indicate sig-

nificant differences among rainfall types for the four treatments according to one way ANOVA

and LSD test (p < 0.05). CK, DT, DT+HG and

CT represent bare flat land as control, down- slope ridge tillage, downslope ridge tillage with hedgerow intercropping, and contour ridge tillage, respectively. Rainfall type I, II and III represent rainfall events of long-duration rain- fall, normal rainfall, and intense rainfall, respectively.

more runoff depth than that created in rainfall type I for the DT and DT HG treatments. Rainfall type II created significantly (P < 0.05) more runoff depth from CT treatments compared with rainfall type I. As

shown in Fig. 4b, rainfall type III had the highest soil erosion for all treatments, with mean annual soil loss ranging from 5.02 to 43.33 t ha—1 yr—1. Rainfall type I showed the lowest soil erosion for the four treat- ments, with mean annual soil loss ranging from 0.66 to 6.33 t ha—1 yr—1.

Soil loss across the three rainfall types showed significant (P < 0.05)

differences under the CK treatment. Soil loss in rainfall type I showed a significant (P < 0.05) difference from the other rainfall types for DT, CT, and DT HG treatments. On the whole, it can be concluded from Fig. 4 that intense rainfall (type III) events caused more soil loss, whereas normal rainfall (type II) events generated more runoff on the sloping

farmland.

4. Discussion

4.1. Effects of ridge practices on water-induced erosion

Water-induced soil erosion is a serious issue in the red soil region of southern China (Liang et al., 2010; Zheng et al., 2021). Due to the in- crease in cultivated vegetation coverage using a peanut/rape rotation compared to that of bare land, runoff generation and soil loss were reduced by an average of more than 18.9% and 68.9%, respectively, during the seven years of the study (Fig. 3). However, due to poor choice of soil management practices implemented in cultivated areas, there are serious risks of soil and water losses. Experimental results (Fig. 5) showed that annual soil loss from cultivated farmland with a conven-

tional downslope tillage in the first year reached 55.99 t ha—1 yr—1,

reaching classification as “intense erosion” (50–80 t ha—1 yr—1) ac- cording to the Standards for Classification and Gradation of Soil Erosion (SCGSE) issued by the Ministry of Water Resources of the People’s Re- public of China. As the years since the land was reclaimed increased, soil loss from the sloping farmland decreased, but due to disturbances in the

crop rotations, annual soil loss continued to be large (above 15 t ha—1

yr—1). Tu et al. (2018) analyzed the erosion characteristics under an

orchard planting mode in the same study area, and found that soil erosion during the first year after planting was the largest. However, after the trees reached maturity, soil erosion decreased and 64% of the

study years recorded annual soil loss rate with a tolerance of < 5 t ha—1

yr—1. This study showed that soil erosion from farmland was severe and

that crop farmland experienced longer erosion times than the orchard farmland. These were consistent with other research results. Wei et al. (2007) determined that crop land experienced the highest surface runoff and soil erosion in the loess hilly areas of China, far more than those in other land use types. Cropland is strongly affected by human-caused disturbances, which disturbs the soil layer and reduces land coverage (Bagarello and Ferro, 2010). Keesstra et al. (2016b) reported that the highest runoff and erosion amounts were seen in uncovered plots, and the lowest values were identified in covered plots. Shui et al. (2001)

found annual soil loss rate from sloped red soil farmland in the first two years after development reached above 50 t ha—1 yr—1 with higher losses