INTRODUCTION

Soil erosion is a global environmental disaster that threatens the survival and continued development of mankind. It is considered to be the greatest threat to agriculture and ecosystems [1]. According to the investigation results of the “Comprehensive Scientific Investigation on Soil Erosion and Ecological Security in China” from 2005 to 2007, the soil erosion area in China was 3.57 × 106 km2, accounting for approximately 14.2% of the soil erosion area of the entire world [2, 3], particularly in the Three-North region (northeast, north, northwest). The Three-North region is the most serious area of wind disturbance and soil erosion in China and is affected by many factors, such as long-term population growth, excessive cultivation, overgrazing, firewood collection, deforestation, and climate change. The total area of soil erosion in northern of China is 554 000 km2, accounting for 14.4% of the surface area. Among the most serious and concentrated areas of soil erosion throughout the world, the Loess Plateau, is in the study region [4]. The data illustrated that the average water erosion rate was 391.61 t/(km2 a) during 2000 [5] in the Three-North region, while the average soil erosion rates were 3362 t/km2 a during 2000 and 2405 t/(km2 a) during 2008 on the Loess Plateau. The Three-North Shelter Forest Project (TNSFP), known as the ‘‘Green Great Wall,’’ is an ecological construction project in China. It is the largest and longest term afforestation reconstruction project in human history, with approximately 2/3 of the area considered arid or semi-arid, and aims to control sand storms and soil erosion via increasing forest coverage [6]. The TNSFP was planned for implementation from 1978 to 2050. The project has three stages (1978–2000, 2001–2020, and 2021–2050) following eight engineering schedules. During 2010, the fourth phase of the TNSFSP (there are a total of seven phases) and the first construction phase of the Beijing–Tianjin Sand Source Control Program were completed (2001–2010). In addition, the Grain for Green Project finished in 2010 (1999–2010). 2015 was the middle period of the TNSFSP over the past 37 years. At this time, there is an urgent need to assess the effects of these programs and some scientific problems need to be solved, such as how to scientifically and objectively assess the soil erosion effects of the TNSFSP, and how to quantitatively measure the soil loss changes with shelter construction, so as to make the ecological environment of the TNSFSP development efficient, stable, and sustainable.

According to soil erosion levels, the types of soil erosion mainly include water erosion, wind erosion and freezing and thawing erosion in the Three-North region, but water erosion is the most serious type of soil erosion [7]. Existing water erosion assessment methods include quantitative methods for measuring erosion rates and qualitative methods for determining erosional risk. Quantitative methods include multi-temporal digital elevation models [8], manpower neural networks [9], the universal soil loss equation (USLE) [10], the revised universal soil loss equation (RUSLE) [11], chemicals, runoff, and erosion from agricultural management systems (CREAMS) [12], the Water Erosion Prediction Project (WEPP) [13, 14], and the European Soil Erosion Model (EUROSEM) [15]. Qualitative methods include visual interpretation [16, 17], image classification [18, 19] and index integration [20, 21]. Among these, the most popular empirical model are USLE, which estimates long-term average annual soil loss rates at different spatial scales and different environments, and RUSLE, which is a revised method of USLE incorporating additional data [22, 23]. RUSLE represents how climate, soil, topography, vegetation, and land use affect rill and inter-rill soil erosion caused by raindrop impacts and surface runoff [18, 24]. In this regard, RUSLE is meaningful used in assessment of water erosion dynamics and thus the soil erosion control service evolution of ecosystems, but more accurate input data are needed and seasonal variation in vegetation has been ignored in previous research.

However, a remote sensing technique periodically receiving data provides a large area, ​​uniform dataset, which greatly contributes to regional erosion assessment [25]. Other countries of the world have already created soil erosion risk and soil erosion rate assessments over a long period based on remote sensing, for example in Italy [22], Cyprus [26], India [27], Bangladesh [28], Russia [29–33], USA [15, 34], and Australia [35–37]. However, in the Three-North region, studies on land degradation and vegetation cover change have been reported [38–40], while there are few reports regarding monitoring the change in water the erosion rate for the whole Three-North region over the last 30 years, and the monitoring time frequency is very low. In the past, soil erosion rate assessment research has been concentrated in local regions, such as the Loess Plateau or one county or one basin [4, 24, 41].

The objective of this study was to assess the spatial temporal changes in soil erosion control service following the implementation of the TNSFP, combined with remote sensing, to effectively and quickly monitor the water erosion risk and the erosion rate dynamics from 1980 to 2015. This paper first describes the rationale and reliability of a new approach to calculate the C-factor value for RUSLE, which fully considers rainfall intensity during the year as a weight to the average monthly C-factor, and then assesses the RUSLE model in estimating the water erosion rate in the study area. Subsequently, the paper discusses the relationship between water erosion and land use and assesses the effects of the TNSFP during the middle period of the project.

MATERIALS AND METHODS

Study area

The TNSFP region is in northern China (33°30′–50°12′ N, 73°26′–127°50′ E) and covers the northeastern, northern, and northwestern regions of China. The TNSFP region covers 551 county-level administrative regions in 13 provinces. The total area of this region is approximately 4.069 million km2, accounting for 42.4% of the total land area of China [42]. From west to east, it passes through arid, semi-arid, and semi-arid semi-humid climates, including the Gobi Desert; Loess Plateau; and arid desert, mountain saline land, forest steppe, and other ecological types. Arid and semi-arid regions account for two-thirds of the study region, which seriously constrains ecological restoration. The mean altitude is between 100 and 5000 meters. The annual precipitation decreases moving from east to west and from south to north, ranging from less than 50 mm to more than 400 mm [19]. The annual average temperature is between –2…–14 °C, and the annual average wind speed ranges from 9 to 30 m/s. The soil and vegetation types among the different regions are very different. The soil types show a horizontal distribution from east to west with the moisture condition changing from wet to dry. In the same soil distribution zone, a vertical band spectrum forms with the change in elevation [39]. In the study area, soil and water conservation regionalization including five districts, as shown in Fig. 1, has been incorporated according to National professional standards for classification and gradation of soil erosion [43]. The area includes the Black soil region with Histosols and Chernozems, the Loess Plateau region with Kastanozems, Chemozems and Calcisols, the Qinghai-Tibet Plateau region with Histosols, the Rocky mountainous region with Kastanozems and greyzems, the Sandstorm region with Kastanozems, Podzoluvisols and Arenosols. Mountains and hills, plains, and plateau account for 27, 11 and 25% of the area, respectively. Before 1977, the proportion of agricultural land, forestry land, grassland and other land-use types was 8.21, 5.61, 43.86, and 42.52%, respectively, of which 84% of the desert and desertification land in China was distributed in the area During the last 30 years, afforestation of 2918.53 km2 has been completed in the area. The average water soil erosion rate was 391.61 t/(km2 a) during 2000 [5].The concentrated population and barren land seriously restrict ecological management and economic development in the study area.

Fig. 1.
figure 1

Location and soil and water conservation regionalization map of the study area.

Full size image

Datasets

Rainfall data were collected by a network of approximately 127 weather stations of the National Meteorological Information Center of China from 1980 to 2015. Data were averaged for every 10-day period of the year at each station. The flood season rainfall (June–October) was used here as a measurement because it had a strong relationship with rainfall intensity. The stations are shown in Fig. 2.

Fig. 2.
figure 2

Location of precipitation observational stations in the study area.

Full size image

The data from Digital Elevation Models (DEMs) were preprocessed by radar images acquired by the Shuttle Radar Topography Mission (SRTM) system. SRTM2 data was adopted in the study with a spatial resolution of 90 m. They were provided by the international scientific data mirror site of the Computer Network Information Center, Chinese Academy of Sciences (http://datamirror.csdb.cn/).

The soil composition data, which included the soil types and the spatial distribution of soil physical and chemical properties (sand, silt, clay, organic matter, and organic carbon content) was collected from “Chinese soil” [44] and “Chinese soil species database” [45] of the Second Soil Survey at a 1 : 1 million scale. The soil types were classified according to the “national second soil census technology specification.”

The land use data for 1980 was provided by National Earth System Science Data Sharing Infrastructure (http://www.geodata.cn) in 1-km spatial resolution. The “land use classification” national standard used the first grade and the second grade of the classification system. The land use for 2015 was provided by the Data Center for Resources and Environmental Sciences, Chinese Academy of Sciences (RESDC) (http://www.resdc.cn) in 30-m spatial resolution.

The normalized difference vegetation index (NDVI) data in 1980 was provided by the global inventory modeling and mapping studies (GIMMS) in a spatial resolution of 8 km over the study area (https://nex.nasa.gov/nex/projects/). To be clear, during the retrieval of soil erosion status during the end of the 1980s, because it was restricted by remote sensing image acquisition, it was difficult to maintain the acquisition time of NDVI for 1980. Therefore, the average NDVI from 1981 to 1985 was used to replace the year 1980. The NDVI data during 2015 was extracted from a moderate resolution imaging spectroradiometer (MODIS) image (MCD43A3, 16-day synthesis). MCD43A3 data was provided by the National Aeronautics and Space Administration (NASA) (https://wist.echo.nasa.gov/api/) in a spatial resolution of 1 km.

Because it is difficult to obtain and identify water erosion rate, we collected water erosion grade based on 499 sample points in the study area between 2008 and 2010 to check the water erosion rate by reclassify them to water erosion grade [41]. The 499 sample points were selected in different land use types by a stratified random process using a Global Positioning System (GPS) (±15 m spatial accuracy) distributed in the study area and with the help of specialists working in local soil and water conservation departments or water conservancy committees. Soil erosion grade, land use, vegetative fraction, topographic information, and physiognomic features were recorded. The erosion grades were mainly determined based on expert knowledge and accumulation of water erosion knowledge in the study area, and all the erosion grades were reclassified to six levels (Fig. 3) according to the field survey. The accuracy of the erosional risk level results was assessed by the erosion grade at the location of the 499 sampling points. The correct rate was confirmed by the same erosional risk grade samples between the field samples and their calculated erosional risk level. The accuracy of the research results was defined by dividing the number of correctly evaluated cells by the total number of samples. Therefore, 499 sampling points were analyzed in terms of the accuracy of the erosional risk grade using the methods described in this study.

Fig. 3.
figure 3

Sample points of each water erosion grade in the study area: Slight: (a) and (b); Light: (c) and (d); Moderate: (e) and (f); Severe: (g) and (h); More severe: (i) and (j); and Extremely severe: (k) and (l).

Full size image

Methodology

Annual soil erosion using the water estimation method

Because vegetative restoration mainly occurred in areas of the TNSFP region dominated by water erosion, wind erosion and freeze-thaw erosion were not considered in this assessment. The RUSLE model was used to assess the amount of annual soil erosion. According to Wischmeier and Smith [46], the formula is defined as follows:

$$A = R~\, \times ~\,K~\, \times ~\,LS~\, \times ~\,C~\, \times ~\,P,$$
((1))

where A is the amount of soil loss (t/(hm2 a)); R is the rainfall erosivity factor (MJ mm/(hm2 h a)); K is the soil erodibility factor (t hm2 h/(MJ mm hm2)); L is the slope length factor (dimensionless); S is the slope factor; C is a dimensionless vegetative cover factor; and a dimensionless P refers to soil conservation practice. The key to applying RUSLE is determining each factor of the equation. The key to the prediction of water erosion using a Geographic Information System (GIS) and RUSLE is also the generation of each factor graph.

Rainfall erosivity facto r(R)

The R factor indicates the erosive force of a specific rainfall [47]. The relationship between rainfall erosivity and rainfall depth developed by Wischmeier and Smith [46] and modified by Arnoldus et al. [48] was used to translate the rainfall depth to rainfall erosivity. Such a method considers both the total annual precipitation and the annual distribution of precipitation. The data is easily obtained [49–57]. The calculation formula is as follows:

$$R = \mathop \sum \limits_{i = 1}^{12} \left\{ {1.735 \times {{{\text{e}}}^{{\left( {1.5 \times {\text{log}}\frac{{P_{i}^{2}}}{P} - 0.8188} \right)}}}} \right\},$$
((2))

where Pi (mm) is the average monthly rainfall during the flood season and P is the average annual rainfall (mm). The unit of rainfall (R) is in metric units, MJ mm/(hm2 h a).The average rainfall intensity distribution surrounding the rainfall stations was processed using the inverse distance weighted (IDW) interpolation [58] method. The rainfall data were resampled to match the NDVI data using a bilinear resampling algorithm.

Soil erodibility factor (K)

K factor describes the vulnerability of a soil to raindrop detachment and runoff wash. The K factor in this study area was determined using the Environmental Policy Integrated Climate (EPIC) model [46]. Only the soil organic carbon and particle composition data were needed in the model to estimate K as follows:

$$\begin{gathered} {\text{K}} = \left\{ {0.2 + 0.3\exp \left[ {0.0256SAN\left( {1 - \frac{{SIL}}{{100}}} \right)} \right]} \right\}~{{\left( {\frac{{SIL}}{{CLA + SIL}}} \right)}^{{0.3}}} \\ \times \,\,\left( {1.0 - \frac{{0.25C}}{{C + {\text{exp}}\left( {3.72 - 2.95C} \right)}}} \right)\left( {1.0 - \frac{{0.7SN1}}{{SN1 + {\text{exp}}\left( { - 5.51 + 22.9SN1} \right)}}} \right), \\ \end{gathered} $$
((3))

where SAN, SIL, and CLA are sandy, silt, and clay content (%), respectively; and C is the organic carbon (%), SN1 = 1 – SAN/100.

Based on the soil map and soil species data at a 1 : 1 million scale, Zhang et al. [59] validated that the estimated K factor was higher than actual observations, and revised the estimated K factor for all of China using Equation 4 as follows:

$$K = -0.01383 + 0.51575{{k}_{{{\text{epic}}}}},$$
((4))

where kepic is the soil erodibility factor estimated using the EPIC model. The unit of the K factor is an American unit, t acre h/(100 acre.ft⋅tonf⋅in), although metric units are typically used in most of the world t ha h/(ha MJ mm). The unit conversion from metric units to American units is multiplication by 0.1317.

Topographical factors (L and S)

The effect of topographic features on soil erosion is concentrated on both the slope length (L) and the slope (S) factors. Both accelerate rainfall erosivity [60]. The slope affects the rainfall area and the accumulated rainfall, which affects the runoff, infiltration, and runoff kinetic energy and in turn soil erosion [61]. The influence of slope length determines the principle of flow and sediment transport, and the evolutionary process of the erosion form, thus affecting the means and amount of soil erosion [62]. With an increase in slope length, water flow energy is more consumed by sediment transport because of the increase in sediment concentration, and erosion is weakened [5, 62, 63]. In this study, we used the improved estimators of L proposed by Wischmeier and Smith [46] as follows:

$$L = {{\left( {{\lambda \mathord{\left/ {\vphantom {\lambda {22.13}}} \right. \kern-0em} {22.13}}} \right)}^{m}},$$
((5))

where, λ is the slope length, m is the slope-dependent index, and L is the slope length factor.

$$\left\{ {\begin{array}{*{20}{c}} {m = {\beta \mathord{\left/ {\vphantom {\beta {\left( {1 + \beta } \right)}}} \right. \kern-0em} {\left( {1 + \beta } \right)}}} \\ {\beta = {{\left( {{{\sin \theta } \mathord{\left/ {\vphantom {{\sin \theta } {0.0896}}} \right. \kern-0em} {0.0896}}} \right)} \mathord{\left/ {\vphantom {{\left( {{{\sin \theta } \mathord{\left/ {\vphantom {{\sin \theta } {0.0896}}} \right. \kern-0em} {0.0896}}} \right)} {[3.0}}} \right. \kern-0em} {[3.0}}{{{(\sin \theta )}}^{{0.8}}} + 0.56]} \end{array}} \right..$$
((6))

In this study, the flow length is estimated by flow accumulation as follows:

$$\begin{gathered} \lambda = {D \mathord{\left/ {\vphantom {D {\cos \theta }}} \right. \kern-0em} {\cos \theta }} \\ = \,\,Flow{\text{ }}Accumulation \times Cell{\text{ }}Size, \\ \end{gathered} $$
((7))

where θ is the slope (%), D is the slope length using a pixel scale, and cell Size is the pixel size.

The aforementioned formula for the L factor was proposed by McCool and colleagues [64], and the S factor for conditions with steeper slopes was first proposed by Liu and colleagues [64, 65]. The formulas for S factor are combined as follows:

$${\text{S}} = \left\{ {\begin{array}{*{20}{c}} {10.8\sin {\alpha\text{}} + 0.03~\,\,\,\,\,{\alpha\text{}} < 5^\circ } \\ {16.8\sin {\alpha\text{}} - 0.5~\,\,\,\,\,5^\circ \leqslant {\alpha\text{}} < 10^\circ } \\ {21.9\sin {\alpha\text{}} - 0.96~\,\,\,\,\,{\alpha\text{}} \geqslant 10^\circ } \end{array}} \right.,$$
((8))

where S is the slope factor and α is the slope. The topographic factor LS is obtained by DEMs using ARCGIS10.3 software.

Cover management factor (C)

The cover management factor represents the effect of plants, crop sequences, and other cover surfaces on soil erosion. The value of the C-factor is defined as the ratio of soil loss from certain types of land-surface-cover conditions. According to Prasannakumar et al. [47], the NDVI can be used as an indicator of land vegetation vigor and health. The following formula was used to generate a C factor surface using the NDVI [22]:

$${{{\text{C}}}_{i}} = {\text{exp}}\left( { - \alpha \frac{{{{{\left( {{\text{NDVI}}} \right)}}_{i}}}}{{\beta - {{{\left( {{\text{NDVI}}} \right)}}_{i}}}}} \right),$$
((9))

where C is the calculated cover management factor, and α and β are two scaling factors. Van and Knijff et al. [22] suggested that Equation 9 was better than using a linear model. They suggest the values for the two scaling factors α and β should be 2 and 1, respectively.

Here, we propose a new approach to calculate the C-factor values fully considering the distribution of rainfall during the year, given the appropriate weight to the average monthly C-factor. The annual C-factor is defined as follows:

$${{{\text{C}}}_{{{\text{year}}}}} = \mathop \sum \limits_{i = 1}^{12} \frac{{{{R}_{i}}}}{{{{R}_{{{\text{year}}}}}}}{{{\text{C}}}_{i}},\,\,\,I = 1 \ldots 12,$$
((10))

where Ri and Ci, are the ith month of rainfall erosivity factor and cover management factor, respectively. Ryear and Cyear are the average annual rainfall erosivity and the average annual cover management factor, respectively.

In the study, the NDVI in 1980 was replaced by the average NDVI of the flood seasons (June to October) [66] from 1981 to 1985 using advanced very-high-resolution radiometer (AVHHR) GIMMS data, while the NDVI in 2015 was extracted from MODIS data during the flood seasons of 2015 in the study area. The weighted average of the C-factor was calculated because of the difference of the months of the flood seasons in the TNSFP region, instead of the full year C-factor.

Soil erosion control practice factor (P)

The soil erosion control practice factor (P) is also termed a support factor. It represents the soil-loss ratio after performing a specific support practice to the corresponding soil loss, which can be treated as the factor to represent the effect of soil-water conservation practices [67]. The range of the P factor varies from 0 to 1. The P factor corresponds to the different land-use types [68–70] during 1980 and 2015 as shown in Table 1. A total number of 27 sub-classes of land-cover subtypes for the study area were further grouped into eight aggregated land cover types: Paddy field, Dry land, Forest, Other forest, Shrub & Grassland, Construction, Water, and Bare rock.

Table 1.   P value of the different land-use types
Full size table

RESULTS AND DISCUSSION

Water Erosion Risk Assessment

Water erosion risk was assessed according to the national professional standard SL190-2007 (The Ministry of Water Resources of the People’s Republic of China, 2008) [71]. Water bodies or snow-covered and construction land were considered to have no soil erosion risk. Table 2 shows the different levels corresponding to the average erosion rate among the different regions. The TNSFP region with water erosion is mainly divided into five regions, each of which has a different average tolerance value for the soil erosion rate. The water erosion maps for 1980 (Fig. 4a) and 2015 (Fig. 4b) were reclassified into different soil-erosion grades that ranged from slight to extremely severe according to indices of soil erosion reported in the standards for classification and gradation of soil erosion in China.

Table 2.   Classes for water erosion intensity using the national professional standard SL190-2007
Full size table
Fig. 4.
figure 4

Soil erosion risk by water maps of the study area (a) during 1980 and (b) during 2015.

Full size image

In this study, when comparing the field erosional risk samples to the calculated erosional risk of 2015, the overall accuracy was 92%. In this regard, RUSLE can be meaningfully used to assess water erosion dynamics and the soil erosion control service of ecosystems [26, 72] in the TNSFP region, when the C-factor for RUSLE fully considers the distribution of rainfall during the year, given the appropriate weight to the average monthly C-factor during the flood season. However, RUSLE is an effective model for predicting erosion estimating the long-term average annual soil erosion amounts [73]. However, the terrain is complex, and the physiognomy type is diverse in the study area, and additional research is needed to improve the accuracy of the soil erosion rate calculation.

The average water erosion rate in the Three-North region during 1980 and 2015 was 530 and 230 t/(km2 a), respectively. By calculating the different erosion levels of the two periods, the overall risk of erosion could be analyzed from a macroscopic point of view by comparing the regional dynamic changes. As shown in Fig. 5, the erosional risk levels between 1980 and 2015 were compared. The erosion regions worse than “Slight” are considered to be eroded. According to the statistical results of the erosional area during 1980 and 2015, the erosional area decreased from 783 400 km2 (20.36% of the total study area) to 368 700 km2 (9.38% of the study area). The light, moderate, severe, more severe, and extremely severe erosion areas decreased from 549 909 km2 (14.29% of the total study area), 137 210 km2 (3.57% of the total study area), 56 047 km2 (1.46% of the total study area), 30 849 km2 (0.80% of the total study area) and 9348 km2 (0.24% of the total study area) to 284 592 km2 (7.24% of the total study area), 45 839 km2 (1.17% of the total study area), 18 759 km2 (0.48% of the total study area), 13 332 km2 (0.34% of the total study area), and 6207 km2 (0.16% of the total study area), respectively. The area of slight erosion increased from 306 3531 km2 (79.64% of the total study area) to 3 561 387 km2 (90.62% of the total study area). These findings suggest an overall decline in erosional area and improvement of the erosion status.

Fig. 5.
figure 5

A comparison of the area of each water erosion risk grade during 1980 and 2015.

Full size image

Comparison of Water Erosion Risk

By overlaying the erosional risk results for 1980 and 2015 pixel-by-pixel, an area change matrix of erosion in the TNSFP region from 1980 to 2015 was obtained. In general, the unchanged erosional risk area is the largest region (309.08 million km2, or 77.82% of the total study area). The improved erosional area is obviously larger than the deteriorated erosional area, indicating that the TNSFP has been significantly effective in improving water erosion. However, there 14.87 million km2 of the region still shows a deteriorating trend, accounting for 4.97% of the total area. Although most of the area with deteriorated erosion grades is only within two levels, management measurements cannot be ignored in this region in the future. The accurate development of erosional management is determined by the spatial distribution of regions with a deteriorating erosional trend. Consequently, soil erosion status overall is improving. The unchanged regions are mainly of low erosional grade, which indicates that the area of lower erosional risk is more stable.

Trend Analysis of Erosion Transformation

By analyzing the water erosion risk grades pixel-by-pixel during 1980 and 2015, a map of changes in water erosion risk was developed by grouping the changing trends into 11 classes (Fig. 6).

Fig. 6.
figure 6

Erosion grade change map of the study area between 1980 and 2015.

Full size image

The change area of water erosion risk is shown in Table 3. The area of erosional risk increased by 148 696 km2, accounting for 3.89% of the study area. The area of erosional risk decreased by 583 506 km2, accounting for 15.26% of the study area. The decreased erosional risk area was greater than that of the increased area, which indicated that the erosional area has generally decreased and the risk of erosion has as well. According to the distribution of the erosional risk, the areas of increased erosional risk are mainly distributed in the middle of the Three-North region, extending to Kunlun Mountain and the southern part of Qilian Mountain in the Qinghai and Gansu provinces. The most significant area of decreased erosional risk was mainly distributed on the Loess Plateau, which accounted for 6.64% of the study area and 43.49% of the decreased erosional risk area. The most significant area of increased erosional risk was mainly distributed in the Qinghai-Tibet Plateau region, accounting for 2.72% of the study area and 70.03% of the increased erosional risk area.

Table 3.   Area (km2) and percentage (%) of increased and decreased erosional risk grade from 1980 to 2015
Full size table

Above all, it can be seen that the more serious the water erosion, the less stable the soil. The slight, light, moderate, and severe erosion risks tended to change to a higher level of erosion, indicating that the trend of erosional risk as a priority should be fully considered.

Relationship between Soil Erosion and Land Use

Soil erosion by water is occurring under certain environment conditions. Vegetative cover and land use are significant erosivity factors for water erosion, because they can quickly change by developing governance measures in a short period of time. Land use in the study area was divided into nine types: paddy land, dry land, forest, open forest, other forest, shrub & grass, waters, construction, and bare rock. Superimposing land use on an erosion grade map, the erosion grade of different land uses is shown in Table 4. Forest is forest with tree cover >20%, except shrub forest and artificial orchards. Open forest is forest with tree cover ≤20%, except shrub forest and artificial orchards. Other forest is artificial gardens, for example, orchards, tea gardens, etc. The differences between open forest and other forest from the erosion point of view are that open forest cause the erosion by the nature process, while the erosion in the other forest could be controlled by the human being active. Because the area of other forest in the study area is small, we put it with shrub and grass together.

Table 4.   Area (million km2) and percentage (%) of each erosion grade by land use during 1980 and 2015
Full size table

As seen in Table 4, the statistical results illustrate that the erosional area of shrub & grass land was the largest, 623 700 and 333 700 km2 during 1980 and 2015, respectively, accounting for 79.62 and 92.85% of the total erosional area. The land use of the second erosional area was dry land. The erosional area was 7.89 million km2 and 10  400 km2 during 1980 and 2015, accounting for 10.08 and 2.89% of the total erosional area, respectively. The land use of the third erosional area was forest land, 60 700 nd 14 700 km2 during 1980 and 2015, respectively, accounting for 7.75 and 4.09% of the total erosional area in the study region. The land use with the least erosional area was mainly open forest. Comparing the two periods, the soil erosion area and proportion of woodland and open forest during 2015 decreased more rapidly than that during 1980, followed by shrub & grassland. Overall, it is safely concluded that the intensity of soil erosion by water among the land use types is as follows: open forest < forest < dry land < shrub & grassland.

Discussion

Water erosion estimated using RUSLE demonstrated a change in soil erosion control service driven by vegetative cover and land use. Land-use change is a key driving force for soil and water loss [4]. Kateb et al. [74] showed that soil erosion in shrub & grassland was worse than that in forest in Shaanxi Province, China. Some scholars [75, 76] have illustrated that forest converted from farmland may have better soil conversation than grass. However, forest conversion has been considered as the most effective land-use conversion type to reduce water erosion in the TNSFP region. Otherwise, vegetative cover restoration [77], as a result of land-use change, plays a significant role in enhancing soil erosion control services of ecosystems.

A dramatic reduction in soil erosion has been recorded during field experiments in China [4, 78] and other regions of the world [29, 79]. During 2010, the total area of soil erosion control in the TNSFP region was approximately 607 hm2, and the sediment loss was approximately 4.1 million tons. The area of desert land and sand land significantly decreased [80]. Before 1978, the Loess Plateau lost 1 000 000–10 000 t/(km2 a). Given the Grain-to-Green Program, the average soil erosion rate decreased from 3362 t/(km2 a) during 2000 to 2405 t/(km2 a) during 2008 [4]. Therefore, all the data illustrate that the trend in water erosion rate has been obviously declining following initiation of the TNSFP. Particularly, in black soil, rocky mountainous, and the Loess Plateau regions, a series of ecological restoration projects and soil and water loss control projects, such as land-use structural adjustments, returning farmland to forest (grass), etc., were measured such that the vegetative coverage rapidly increased and the soil erosion rate remarkably decreased in the regions [4, 81]. However, the most significant erosion rate increase during 2015 was mainly distributed in the Qinghai-Tibet Plateau region, because of the fragility of the ecological system and grassland degradation, cultivated land reclamation, thin soils, low water storage capacity, and increased rainfall erosivity. These have resulted in an obvious increase in soil erosion in the alpine grassland ecosystem of the Qinghai-Tibet Plateau region [81–83]. For the same reasons the soil erosion rate is higher in the southwestern Sandstorm region than that of northwestern Sandstorm region; otherwise, a higher runoff coefficient and a larger area of cultivated land reclamation, as well as more rainfall, are the main reasons for the north-south difference in soil erosion rate change [84, 85]. A detailed trend analysis was obtained by analyzing the time series of erosional risk in the region, providing more detailed information to support conservation priorities. From this research, we found that the major difficulty in assessing soil erosion at a regional scale was the availability and the quality of data. However, the shortcomings of datasets used in this study were that there were only two periods of remote-sensing data with a 30-year time span. To fully analyze the dynamic risk of soil erosion, remote-sensing data during the vegetative growing season of higher spatial and temporal resolution is needed.

CONCLUSIONS

Weather station data, multi-sensor, and multi-temporal remote-sensing data; Chinese soil mapping at a 1 : 1 million scale; and ecological environment background data for extracting and calculating water erosion factors were collected, including soil erosivity, rainfall erodibility, topography, vegetative cover, and land use. Combined with the RUSLE model, we systemically studied the risk and trend of water erosion in the TNSFP region between 1980 and 2015. The main conclusions are as follows:

First, the water erosion rate of the TNSFP region was effectively calculated using the quantitative assessment model of RUSLE based on remote-sensing data. The average soil erosion rate decreased from 530 to 230 t/(km2 a) from 1980 to 2015. The erosional area of the study area decreased from 783 400 km2 (20.36% of the total area) during 1980 to 368 700 km2 (9.38% of the total study area) during 2015. The erosion rate and area significantly decreased as did the water erosion risk. Otherwise, between 1980 and 2015, the area of improved erosion was significantly more than that of deteriorated area, which indicates that the Three-North Shelter Forest Project and Grain for Green Project significantly improved the ecological environment. However, there is still 14.87 million km2 of the region trending toward deterioration, which should be a priority for measures in the future. Furthermore, the most significant erosion rate increase from 1980 to 2015 was mainly distributed in the Qinghai-Tibet Plateau region and the southwestern Sandstorm region, a result of grassland degradation, cultivated land reclamation, increased rainfall erosivity and runoff, etc.

In addition, our new approach to calculate the C‑factor values for RUSLE fully consider the rainfall intensity during the year as the appropriate weight to the average monthly C-factor. This innovative method provides a rational and reliable estimate of the C-factor on a pixel-by-pixel basis and thus represents significant progress for spatial modeling of soil erosion by water using the USLE or RUSLE model. Finally, a comprehensive interrelationship between land use and water erosion was found. Generally, shrub & grassland resulted in worse water erosion than did dry land and forest land. The effectiveness of forest conservation for reducing water erosion was illustrated as a significant priority.

The soil erosion by water map indicated the relative probability of erosion and contributes to effective allocation of scarce conservation resources and development of policies and regulations. The quantitative evaluation of spatial and temporal distribution of erosional area and erosion factor analysis are an important basis for decision making. According to the study results, great efforts are needed to enhance ecological restoration in the Qinghai-Tibet Plateau region and the southwestern Sandstorm region, such as afforestation or enclosures for grassland to reduce soil erosion.