Abstract
Soil and water conservation (SWC) measures are among the most important ways to control soil erosion in mountainous areas. However, little is known about the application of SWC measures on soil erosion over large areas. We developed a new framework for investigating the distribution of engineering measures by using high-resolution remote sensing images at a large scale, and we assessed the soil conservation effects of those measures on the Yunnan Plateau (YP) in China using the Chinese Soil Loss Equation (CSLE). We detected 43,398 km2 of soil conservation measures on the YP, accounting for 53% of farmland. The area of conservation measures decreased with an increase in slope gradient. Different types of engineering measures had different spatial distribution patterns, and together they conserved 5.08 × 108 t of soil each year on the YP. The effects of engineering measures on soil erosion decreased with an increase in slope. Furthermore, more than 82% of soil loss from farmlands was attributed to areas without engineering measures, and steep-slope farmland without such measures was the main source of soil erosion on the YP. Our results suggest that engineering measures were effective in soil conservation, but the effects varied with topography on the YP. We suggest that further studies focus on the relationships between the soil erosion and grain supply of farmlands without engineering measures to provide a basis for the formulation of land use planning policy in China.
Introduction
Soil erosion can result in a decrease in soil fertility and crop yield, an increase in the pollution of ground and surface waters, and an increased threat to sustainable development (Lal 2001). Without effective prevention and control measures, it is estimated that by 2050, the total grain loss will exceed 2.53 × 1011 t, which is equivalent to reducing the crop production area by 1.5 × 107 km2 (FAO 2015). Soil and water conservation (SWC) has received strong support (UNCED 1992; UNFCCC 1992; UNFCD 1996; UNFCC 1997), and in many countries, conservation measures such as terracing, soil bunds, and area closures have been designed to prevent soil erosion (Anley et al. 2007). In China, soil erosion has received considerable attention, and a series of SWC projects have been implemented since the 1950s (Mu et al. 2007). Every year, the government dedicates a large amount of funds to establish various soil conservation measures throughout the country (Zhang et al. 2017). According to the latest report of the National Water and Soil Conservation Program of China, large areas of eroded land must be restored every year (Ministry of Water Resources of the People's Republic of China 2016). Thus, it is critical to identify the factors controlling the effectiveness of such measures on soil erosion.
SWC measures are the most effective ways to decrease soil erosion (Fang and Sun 2017). SWC measures often include engineering measures, defined as constructed changes in topography that reduce runoff and soil loss with terraces, fish-scale pits, and level ditches (Liu et al. 2002). According to the First National Census for Water, engineering measures are one of the most important soil conservation measures used in China (Liu et al. 2013), and these measures are undertaken by the government or by farmers during long-term cultivation. The effects of engineering measures on SWC have been studied by monitoring (Barton et al. 2004; Zhang et al. 2008), investigation (Lü et al. 2009; Andrea et al. 2017), and modeling (Shen et al. 2010; Fang and Sun 2017). Numerous runoff plots with and without conservation measures have been built so that soil losses can be compared (Liu et al. 2013; Zhao et al. 2019). However, most of these studies have focused on the plot (Zhao et al. 2019), hillslope, or small watershed scale (De Graaff et al. 2008). This is because SWC measures were not considered in current land use classification (Anderson et al. 1976; Liu et al. 2000), rendering it difficult to obtain SWC data over large areas. In farmlands, conservation measures such as sloping terraces (Gardner and Gerrard 2003), interval terraces (Fu 1989), contour bunds, inward (reverse)-sloping bench terraces (Sang-Arun et al. 2006), hill-side/level ditches, and fish-scale pits (Wang et al. 2014) have not been included in land use data. In addition, it is difficult to perform field investigations of these measures due to high labor costs and their time-consuming nature. As a result, soil conservation engineering measures have seldom been adopted over large areas, and this lack of information is detrimental to regional SWC planning and sustainable development.
Recently, high-resolution satellite data (0.3 to 0.5 m) have been used to obtain features of soil conservation measures (Eckert et al. 2017). Diaz-Varela et al. (2014) presented a method to use high-resolution ortho-image and digital surface models to identify terraces in a very small experimental agricultural setting located in Spain, and the authors achieved highly accurate terrace classification. Mekuriaw et al. (2017) used a model to map SWC structures with very high spatial resolution imagery obtained from Google Earth (Google, Mountain View, California, United States), and their model was calibrated and validated on the Ethiopian highlands. Eckert et al. (2017) developed an object-based classification approach to identify SWC measures by using 0.5 m resolution stereo satellite data in a case study area of 23 km2 located north of Asmara; their method identified most (79%) terraces and bunds and hillside terraces (81%). Guyassa et al. (2018) also used Google Earth images to interpret the SWC construction of stone bunds and terraces in a catchment located in the northern Ethiopian highlands. These studies were all conducted in small areas. However, few studies have been conducted to assess the effects of SWC measures on soil erosion using high-resolution satellite data over large areas.
The Yunnan Plateau (YP) is a typical mountainous region in southwest China (Development and Planning Committee of Yunnan Province, Land Resource Bureau of Yunnan Province 2004; Duan et al. 2016). The steep slopes and abundant precipitation usually result in severe soil erosion. SWC measures have been widely used to control soil erosion and sustain agricultural production in the YP (Barton et al. 2004). The major objectives of the current study were to (1) develop a new framework for investigating large-scale engineering measures by using high-resolution remote sensing images, and (2) assess the effectiveness of soil conservation measures on soil erosion over a region.
Materials and Methods
Study Areas. The YP is located in southwest China. The eastern YP belongs to the Yunnan-Guizhou Plateau, and the northwestern YP is a part of the southern extension of the Qinghai-Tibet Plateau (Duan et al. 2016). The YP covers an area of 383,000 km2, of which more than 80% is mountainous (Development and Planning Committee of Yunnan Province and Land Resource Bureau of Yunnan Province 2004). The elevation ranges from less than 100 m in the southeast to more than 6,000 m in the northwest, with a mean elevation of 1,879 m. Yunnan's climate is influenced by the interactions of several circulation systems (Duan et al. 2016). The mean annual temperature is 14.7°C, and the mean annual precipitation is 1,102 mm (Gu et al. 2016). Six geographic subregions were defined, i.e., the northwest, west, southwest, central, northeast, and southeast subregions, based on the comprehensive physical geographical regionalization in the YP (Li and Walker 1986). The dominant soil type is red soil according to the Genetic Soil Classification of China (NSSO of Yunnan Province 1996).
Interpretation of Engineering Measures. We established a new framework to visualize the engineering measures by using high-resolution remote sensing images at a 1:10,000 scale. First, we obtained 0.5 m high-resolution remote sensing images of the province (3,570 scenes from Worldview Satellite [DigitalGlobe, Westminster, Colorado] image from 2011 to 2014). Second, we conducted geometric accurate correction and atmospheric correction for all original images. Third, we established visual interpretation of Peugeots for all engineering measures (table 1). Fourth, we overlaid a land-use map (scale of 1:10,000, obtained from the Second National Land Survey) (Lu and Yang 2014) and remote sensing images in ArcGIS 10.3 (ESRI, Redland, California, United States) to interpret engineering techniques with a human-computer interaction method (Dix et al. 2004). The minimal polygon was 100 m2, with an interpretation accuracy of 10 × 10 m. Additionally, field in situ validations were conducted. The total areas of farmland with engineering measures (FWEM) and without engineering measures (FWOEM) were analyzed by ArcGIS version 10.3 (ESRI Inc., Redlands, California, United States).
Descriptions of different engineering measures.
Calculation of Erosion Modulus. The Universal Soil Loss Equation (USLE) and its modified version are widely used in predicting soil loss (Nearing et al. 1994; Bagarello et al. 2013). The original version of the USLE was developed in the United States (Nearing et al. 1994), but the data for assessing the effects of slope gradient on soil erosion at steep slopes are limited (Liu et al. 1994). Considering that much of the farmland is located on slopes of greater than 10° and the complicated SWC measures in China, Liu et al. (2002) developed the Chinese version of USLE, namely, the Chinese Soil Loss Equation (CSLE), in which the slope steepness factors were calculated by using an equation established by Liu et al. (1994) when the slope gradient is greater than 10°. In addition, the cover-management (C) and support practice (P) factors were reclassified into vegetation cover and biological practice (B), as well as engineering techniques (E) and tillage (T) practices in the CSLE. CSLE model has been widely proven and successfully applied in China (Liu et al. 2013; Yin et al. 2018), and the model was not only adopted in the First National Census for Water in China (Liu et al. 2013), but also documented in the technical manual of soil erosion investigation in China by the Chinese Government (Monitoring Center of Soil and Water Conservation in Ministry of Water Resources of the People's Republic of China 2018). Therefore, we adopted the CSLE to calculate the erosion modulus in the present study. The detailed equation is as follows:
1
where A is the predicted soil loss, representing the amount of soil loss per unit area and per unit time (t ha−1 y−1); R is the rainfall erosivity factor representing the average annual rainfall erosivity (MJ mm ha−1 h−1 y−1); K is the soil erodibility factor (t ha h ha−1 MJ−1 mm−1); L is the slope length factor, which is a ratio of soil loss from a plot-slope length to that from a length of 22.13 m in the same soil type and gradient; S is the slope gradient factor expressed as the ratio of soil loss from the plot gradient to that from a 9% slope; B is the vegetation cover (%) and biological measure factor, which is the ratio of soil loss from a field with a specified vegetation type, cover, or measure preventing soil erosion, to that from fallow conditions; E is the engineering factor expressed as the ratio of soil loss from a field with a specified engineering measure to that from an unprotected field with farming in straight rows down the slope; and T is the tillage and management factor.
The R factor was calculated with the algorithm developed by Zhang et al. (2002, 2003). In the current study, the daily precipitation data were obtained from the National Meteorological Observatory (NMO) stations in the YP. The data were provided by the Yunnan Climate Center. To make the best use of the available data and obtain the best spatial coverage, data from 115 stations were selected to calculate rainfall erosivity from 1960 to 2012.
The K factor was calculated using the soil attribute data established by Wischmeier and Manning (1969). First, the soil attributes (soil type, location, parent material, vegetation, horizon depth, bulk density, organic matter, particle size distribution, and soil structure) of 347 soil series were obtained from the Second National Soil Survey of China (NSSO 1998, 1996; NSSO of Yunnan Province 1989, 1994). These collected soil series were classified into 42 soil families and 18 groups. Soil classification was based on 1:750,000 Yunnan soil map (NSSO of Yunnan Province 1994). A field campaign was conducted to determine the soil physicochemical properties in 274 soil series covering 42 soil families and 18 groups, which was used to compensate for some missing information in the databases of the Second National Soil Survey. The primary physicochemical properties—particle size distribution and organic matter (OM) content—were determined according to standard soil analysis methods (Carter and Gregorich 2008). The K values of the 347 soil series were calculated using the equations developed by Wischmeier and Manning (1969).
The formula to calculate the slope length (L) factor was proposed by Wischmeier and Smith (1978) and then modified by Liu et al. (2000) for steep slopes (S) as the following exponent values. 1:10,000 topographic maps were available for 73.41% of the YP area, but only 1:50,000 terrain maps were available for the other areas. Based on projection, edge, correction, resampling, and other technical processes, we generated a regional topographic digital elevation model (DEM) at a 10 m resolution to calculate the L and S factor. Furthermore, to analyze the spatial distribution of engineering measures and their soil conservation effects, the slopes were reclassified to the following eight levels: <5°, 5° to 10°, 10° to 15°, 15° to 20°, 20° to 25°, 25° to 30°, 30° to 35°, and >35°.
The calculation of the B factors involved four steps. First, to generate successive (using a two-week interval) NDVI data for different landmarks (The Leading Group Office of First National Census for Water of the State Council of the People's Republic of China 2010), a succession-correction method was applied to fuse MODIS NDVI (each two-week period) and TM/ETM NDVI (30 m spatial resolution) data. The landmarks (evergreen coniferous trees, evergreen broadleaf trees, deciduous conifers, deciduous broadleaf trees, shrubs, meadows, cereal crops, broadleaf crops, towns, water, etc.) were extracted from 1:10,000 land use maps (Lu and Yang 2014) and MODIS land cover data. Second, the NDVI was converted into canopy cover, and the conversion coefficients and formula indices were determined for different climatic vegetation types and regions. Third, canopy cover raster data for the 24 two-week periods were established using a smoothing process. Finally, the B factors were calculated with the formula established by Liu et al. (2013). The detailed equation is as follows:
2
where 
is the proportion of the mean annual half-month rainfall erosivity in the k-th half-month to the mean annual rainfall erosivity, %; Bk is the soil loss ratio for the k-th half-month with various vegetation types and coverage, and the value can be obtained as in Liu et al. (2002); and B is the annual average B value.
The E value is the ratio of the average annual soil loss from plots with engineering measures to the average annual soil loss from the control (fallow) plot with a specified slope length and slope gradient. The slope length and slope gradient were normalized to the unit plot slope by multiplying by the slope length L and steepness S (Liu et al. 2013). The data of the plots used to calculate the E factor were derived from the team of the First National Census for Water (Guo et al. 2015) and the Soil and Water Conservation Monitoring Station of Yunnan Province, which involved 9 and 42 plots, respectively, for a total of 51 plots. Twenty-six plots have a sloping terrace engineering measure, 11 plots have intermittent/interval terraces, 3 plots have level bench terraces, and the remaining plots have inward (reverse)-sloping bench terraces. The projected areas of these plots were 100 m2 (20 × 5 m) or 150 m2 (30 × 5 m).The monitoring periods ranged from three to five years, and the slope gradient was from 5° to 20°. The estimated E values for four engineering measures (table 1) can be found in table 2. An E value of 1 was assigned to areas without engineering measures.
In the CSLE, the T factor includes the methods of cultivation that shorten the length of slope or change the direction of runoff, thus minimizing runoff and soil erosion (Liu et al. 2002). The current study focuses on the effects of agricultural rotation on soil erosion. To evaluate the T factors of different crop rotations, we first collated the regional map of cropping systems in the YP (Liu and Han 1988); then, we collated crop data by questionnaires and field investigations for counties. The county crop rotation map and the T factor values for different crop rotation systems were obtained from the First National Census for Water (Liu et al. 2013), and were calculated using measured plot data from different crop rotation systems.
Validation of the Chinese Soil Loss Equation Model. To validate the CSLE model, we used the measured soil erosion data from cropland plots. First, we collected the data from 23 cropland plots monitored by the Soil and Water Conservation Monitoring Station of Yunnan Province (table 3), which cannot be used in the calculation of E and other factors. Additionally, the chosen plots should be distributed in different regions of YP, and involve the three engineering measures of level bench terrace, sloping terrace, intermittent/interval terrace, and inward (reverse)-sloping bench terrace, in which the factors affecting soil erosion, such as precipitation and vegetation coverage, were monitored. The monitoring periods of these plots ranged from three years to five years, and the projected areas were all 100 m2 (20 × 5 m). Second, the soil erosion modulus was calculated using the CSLE model based on the monitoring data of precipitation, slope gradient, slope length, vegetation type and coverage, engineering measures, and tillage practices from 23 cropland plots. Finally, the coefficient of determination (R2) (De Vente et al. 2013) and the root mean square error (RMSE) (Cort and Kenji 2005) between the measured erosion modulus and the calculated erosion modulus were applied to validate the investigation results.
E factor values for four engineering measures.
Monitoring cropland runoff plots used to validate the Chinese Soil Loss Equation model.
Assessing the Effects of Engineering Measures on Soil Conservation. To assess the effects of engineering measures on soil conservation, we calculated the soil erosion modulus (A′), which is the predicted soil loss per unit area (t ha−1 y−1) under the assumption of “without engineering measures” (the E factor in the CSLE), and the formula is as follows:
3
with the same parameters as those in the formula presented for the CSLE model. The effect of engineering measures on reducing the soil erosion modulus is assessed by the soil conservation ratio (SCR), which is calculated by the following equation:
4
As the SCR increases, the conservation effect of the engineering measures increases, and vice versa. Furthermore, the ratios of farmland with engineering measures to total farmland (RFWEM), farmland without engineering measures to total farmland (RFWOEM), and farmland area to total land (RF) were calculated with the following equations, respectively:
5
6
7
where AFWEM is the area of farmland with engineering measures (km2), AFWOEM is the area of farmland without engineering measures (km2), AF is the area of total farmland (km2), and AT is the area of total land (km2).
The soil erosion intensity was classified into six levels according to national standards for classification and gradation of soil erosion (Ministry of Water Resources of the People's Republic of China 1997). The erosion intensity levels were as follows: slight at <5 t ha−1 y−1, light at 5 to 25 t ha−1 y−1, medium at 25 to 50 ha−1 y−1, strong at 50 to 80 t ha−1 y−1, very strong at 80 to 150 t ha−1 y−1, and severe at >150 t ha−1 y−1. The areas where the erosion intensity level was above the slight level were regarded as erosion areas.
Results and Discussion
Spatial Distribution of Engineering Measures. Farmland occupied an area of 82,008 km2 and accounted for 17% of the total land area in the YP. Fifty-three percent of the farmland had soil conservation engineering measures. The farmlands that had soil conservation engineering measures were mainly distributed in the central and eastern YP, including Kunming, Yuxi, and Qujing districts, while the farmland without soil conservation engineering measures was found mainly in Zhaotong in the northeastern YP, Wenshang in the southwestern YP, and Lincang and Puer in the southwestern YP (figure 1).
The four engineering measures considered in this study (table 1) exhibited different spatial distribution patterns, with the area changing in the following order: sloping terrace > level bench terrace > inward (reverse)-sloping bench terrace > intermittent/interval terrace. The sloping terrace was most widely distributed with an area of 27,538 km2 and was mainly found on slopes less than 20°. Additionally, its area decreased gradually on slopes greater than 20° (figure 2). Level bench terrace had an area of 15,193 km2 and was mainly distributed on slopes less than 5°. The total area of the inward (reverse)-sloping bench terrace was 499 km2, and this type mainly covered slopes ranging from 15° to 25° (figure 2). The intermittent/interval terrace covered the smallest area (168 km2), and was also distributed mainly on slopes ranging from 15° to 25° (figure 2).
Spatial distribution of farmland and soil conservation engineering measures in the Yunnan Plateau.
Both farmland and engineering measures tended to be distributed on low slopes in this mountainous region. The area of farmland with engineering measures (FWEM) and the ratio of FWEM to farmland both decreased gradually with an increase in slope. The area of farmland without engineering measures (FWOEM) increased with an increase in slope from 5° to 25° and then decreased with a further increase in the slope gradient; the ratio of FWOEM to farmland showed a consistent increasing trend with increasing slope (figure 3b). On the other hand, as slope increased from less than 5° to greater than 35°, the ratio of farmland to total land area (RF) decreased from 52% to 7% (figure 3b).
Slope-dependence of farmland area under different soil and water conservation (SWC) measures.
Validation of the Chinese Soil Loss Equation Model. Compared with the measured soil erosion modulus from the 23 cropland runoff plots from the soil conservation monitoring stations in the YP, the CSLE overpredicted the soil erosion modulus, especially in the plots with low erosion rate. The coefficient of determination was 0.93, and the RMSE was 6 t ha−1 y−1 (figure 4), which indicated a high simulation precision by the CSLE in the YP.
The (a) FWEM and FWOEM and (b) RFWEM, RFWOEM, and RF change along slope gradient. FWEM is the area of farmland with engineering measures; FWOEM is the area of farmland without engineering measures; RFWEM is the ratio of farmland with engineering measures to farmland; RFWOEM is the ratio of farmland without engineering measures to farmland; and RF is the ratio of farmland area to land area.
Effects of Soil Conservation Engineering Measures on Soil Erosion. The average soil loss modulus in farmlands with engineering measures (A) in the YP was 47 t ha−1 y−1. A higher A was found mainly in the cities of Zhaotong (northeastern YP), Wenshang (southeastern YP), Honghe (southern YP), and Puer and Lincang (southwestern YP) (figure 5a), which had areas with a large percentage of steep slopes. More than 3.80 × 108 t of soil was eroded every year from farmlands in the YP, with more than 82% of the total soil loss coming from FWOEM. The average soil loss modulus in farmlands without engineering measures (A′) in the YP was 109 t ha−1 y−1, a 133% increase over that in A. Approximately 5.08 × 108 t of soil, equivalent to an average of 62 t ha−1, was protected by engineering measures every year, accounting for 165% of the actual soil loss from farmland. The spatial distribution of A′ was similar to that of A (figure 5b). The soil conservation ratio (SCR) ranged from 0% to 99%, with an average value of 42%. Higher SCR values were mainly found in the central and eastern areas of the cities of Kunming, Yuxi, and Qujing, which had high percentages of soil conservation measures (figure 5c). As the slope increased from less than 5° to greater than 35°, the average erosion modulus (A) exhibited a consistently increasing trend from 2 to 114 t ha−1 y−1. The erosion modulus A′ exhibited an increasing trend when the slopes were less than 35°, and then decreased slightly. A′ was greater than A on all slope gradients. With an increase in slope, the SCR consistently decreased from 69% to 18% (figure 5c).
According to the national standards for classification and gradation of soil erosion intensity, the total soil erosion area was 48,178 km2 for FWEM, accounting for 59% of total cropland (figure 6a). Among the total erosion area, 25% was at light erosion level, 15% was at the middle level, 9% was at the strong level, 13% was at the very strong level, and 12% was at the severe erosion level. The area of high erosion intensity increased with slope (figure 6a). Without engineering measures, the total soil erosion area was 59,134 km2 for cropland in the YP, accounting for 72% of total cropland. Of that land, 17% was at a light erosion level, 11% at a middle level, 10% at a strong level, 16% at a very strong level, and 17% at a severe erosion level (figure 6b). The area that exhibited low erosion levels (erosion intensities lower than the middle level) was smaller in FWOEM than that in FWEM, while the area of high erosion levels (erosion intensities greater than the middle level) was higher in FWOEM than that in FWEM in all slope gradients (figure 6). This result indicated that engineering measures decreased soil erosion in cropland, and that the soil conservation effects decreased as the slope increased (figure 7).
The linear relationship between measured soil erosion modulus and predicted soil erosion modulus in the 23 cropland runoff plots in the Yunnan Plateau. Solid line signifies the line fitted by linear regression; dashed line signifies the straight line with a slope of 1 and constant of 0.
The Spatial Distribution of Engineering Measures. Soil conservation measures are the most important means of controlling soil erosion, especially in mountainous regions (Marques et al. 2010; Maetens et al. 2012; Tenge et al. 2005). However, physical (Bekele and Drake 2003), economic (Shiferaw and Holden 1999), and policy (Baidu-Forson 1999) factors influence the investment in soil conservation measures (Anley et al. 2007). Our results showed that different types of engineering measures had different distribution patterns due to topographic or other natural conditions. For example, level bench terrace was mainly found on slopes less than 5° because it was mainly used for rice (Oryza sativa L.) cultivation, which requires stable geological and water conditions (Bhushan 1979). It has been suggested that level bench terraces should be built in relatively flat areas to increase the water holding capacity and reduce soil erosion (Zhang et al. 2017). This suggestion is supported by the current study. Thomas et al. (1980) suggested that terraces were most effective in locations where slopes were <15% (8.5°), which was similar to our results. Sloping terraces were the most widely distributed soil conservation measure in the YP, and they were mainly built on local farms for the cultivation of dryland crops (Zhang et al. 2015). Less environmental and technical requirements are required for the construction of sloping terraces than for level bench terraces. This level of input likely explains why sloping terraces were mainly distributed on slopes less than 20°. Inward (reverse)-sloping bench terraces have been recommended for areas with steep slopes (Zhang et al. 2017; Liu et al. 2013). Our results showed that the inward (reverse)-sloping bench terraces were mainly found in areas with slopes ranging from 15° to 25°. Anthropogenic effects, such as the distance to settlements or to roads, were found to be the major drivers for affecting the spatial distribution of terrace types (Schönbrodt-Stitt et al. 2013). In the mountainous plateau region, geological stability and traffic accessibility are relatively low (Development and Planning Committee of Yunnan Province, Land Resource Bureau of Yunnan Province 2004). Subsequently, the construction costs of SWC measures are higher in areas with higher slopes than in areas with lower slopes. This difference explains why the ratio of soil conservation measures to farmland decreased gradually as the slope increased in the current study.
Spatial distribution of (a) current erosion modulus (t ha−1 y−1) in farmland, (b) modeling erosion modulus (t ha−1 y−1) without engineering measures in farmland, and (c) the soil conservation ratio (SCR) of engineering measures in farmland in the Yunnan Plateau.
The Effects of Engineering Measures on Soil Erosion. It was found that soil conservation measures can effectively reduce soil loss in the current study. This result is consistent with those of other studies (Bekele and Drake 2003; Derpsch et al. 2014; Keesstra et al. 2016; Liu et al. 2013; Diyabalanage et al. 2017). Terracing reduced soil losses from 20 to 1 t ha−1 y−1 in New Brunswick, Canada (Chow et al. 1999). A conservation bench terrace system was effective in reducing soil losses by 90% in Dehradun, India (Sharda et al. 2002). Flow-diversion terraces reduced sediment yields by on average of 56% in New Brunswick, Canada (Yang et al. 2009). The average soil loss, ranging between 86 and 230 t ha−1, decreased to values between 260 and 537 t ha−1 after building level terraces in Tuscany and Emilia-Romagna, Italy (Bazzoffi et al. 2006). The sediment yield was reduced from 14 to 9 t ha−1 y−1 after terracing in Tigray, Ethiopia (Nyssen et al. 2009).
However, our results indicated that the soil conservation effects of engineering measures on soil erosion varied at different spatial scales. To date, assessments at regional scales have been sparse due to the lack of data. The effects of conservation measures on soil erosion in the YP decreased with slope, possibly due to three reasons. First, more soil conservation projects were constructed at lower slopes than at higher slopes (figure 3a). On one hand, there are lower costs to establish engineering measures in the gentle slope regions. Thus, local farmers and the government are prone to establishing engineering measures in these gentle regions (Zhao et al. 2016). On the other hand, establishing engineering measures in the gentle slope regions is advantageous for increasing the farmland yield (Xu et al. 2010; Zhao et al. 2019). Second, erosion control decreased in the order of level bench terrace > inward (reverse)-sloping bench terrace > sloping terrace > intermittent/interval terrace in this study. The most effective measure, the level bench terrace, was mainly found in areas with low slope gradients (figure 2). Third, SWC measures for areas with steep slopes require special consideration because these areas are prone to potential ecological risks, such as landslides and debris flows (Tarolli et al. 2014). Koulouri and Giourga (2007) concluded that when the slope gradient reached 25%, soil erosion increased significantly after terrace abandonment due to changes in vegetation cover.
The soil erosion intensity of farmland (a) with engineering measures and (b) without engineering measures.
Mountain plateaus are characterized by high altitude and widely distributed mountains (Fan et al. 2011), where steep-slope farmland and high population density make the ecological environment extremely fragile (Duan et al. 2015). Our results show that the ratio of farmland to total land was highest on low-gradient slopes and decreased as slope increased, indicating that most areas with low slopes have been reclaimed into farmland. To produce more food, some steep-slope areas that may not be suitable for cropland have also been reclaimed. It has been widely accepted that deforestation in steep areas can cause severe soil erosion (Albaladejo et al. 2000), and engineering measures, mainly distributed on low slopes in the YP, cannot address soil erosion issues on steep slopes. We find that soil loss from farmland without engineering measures increased with slope (figure 7). Although the farmland distributed on steep slopes (>25°) without engineering measures accounted for only 16% of the total farmland, this farmland contributed to more than 45% of the total soil loss from farmland (figure 7). This indicated that steep-slope farmland without protective measures was the main source of soil erosion in the mountain plateau of the YP.
The “grain for green” or constructed SWC measures should be applied to steep-slope farmland to reduce soil erosion in this region. However, constructing engineering measures on steep slopes is difficult and represents a high ecological risk (Koulouri and Giourga 2007), and “grain for green” may result in a decrease in crop yields and threaten regional food security. Further studies need to address the balance between soil erosion risk and grain supply in sloping farm fields to provide a basis for formulating a reasonable land use policy.
Namely, the E factor of each engineering technique was assigned a reference to previous findings and based on monitoring results from a runoff plot in the YP; in fact, the effects of different conservation engineering measures varied with the quality of the project, and with the ground-cover conditions (Ramos et al. 2015), indicating that the assessment of the effects of SWC engineering measures with the CSLE model in this study requires further research.
Summary and Conclusions
A large area with SWC engineering measures was detected in a typical mountainous region of the YP in southwest China. The engineering measures decreased as the slope increased. Different types of engineering measures showed different spatial distribution patterns, with level bench terraces mainly distributed in low-slope areas, while other measures were not. The engineering measures considered in the current study efficiently reduced soil erosion by an average of 62 t ha−1 y−1. The effects of engineering measures on soil erosion decreased as slope increased. Steep-slope farmland without soil conservation measures was the main source of soil erosion. However, it should be noted that constructing engineering measures in steep-slope terrain is a challenge. Further studies are needed to address the relationship between soil erosion risk and grain production in this region.
Changes in erosion conditions with slope in Yunnan Plateau. A is modeling erosion modulus with engineering measures in farmland; A′ is modeling erosion modulus without engineering measures in farmland; SCR is soil conservation ratio of engineering measures in farmland.
Acknowledgements
We would like to thank Kathryn B. Piatek (chief science editor, Edit My Science, Morgantown, West Virginia) for language editing and helpful suggestions on the manuscript. This work was funded by the NSFC-ICIMOD Joint Research Program (Grant number: 41661144044), the National Natural Science Foundation Project of China (Grant number: 41561063, and 41101267), and the Nonprofit Industry Research Project of the Chinese Ministry of Water Resources (Grant Number: 201501045).
- © 2020 by the Soil and Water Conservation Society
References
In this issue
