EVALUATION OF LAND-USE CHANGE EFFECTS ON

RUNOFF AND SOIL EROSION OF A HILLY BASIN OF

YANHE RIVER IN THE CHINESE LOESS PLATEAU

Kaijie Yang1,2, Changhe Lu1,2*

1
Key Laboratory of Land Surface Pattern and Simulation, Institute of Geographic Sciences and Natural

Resources Research, Chinese Academy of Sciences, 100101 Beijing, China

2
College of Resources and Environment, University of Chinese Academy of Sciences, 100049 Beijing,

China

Corresponding author:

Changhe Lu (Lu C.H.)

Address:

Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of

Sciences, 11 A, Datun Road, Chaoyang District, Beijing 100101, China

Tel:

+86 010-64889110

Fax:

+86 010-64889110

E-mail:

This article has been accepted for publication and undergone full peer review but has not
been through the copyediting, typesetting, pagination and proofreading process which may
lead to differences between this version and the Version of Record. Please cite this article as
doi: 10.1002/ldr.2873

This article is protected by copyright. All rights reserved.

[email protected]

Running Title:

LAND USE CHANGE EFFECTS ON RUNOFF AND SOIL EROSION

ABSTRACT

Starting in 1999, the Grain-for-Green (GFG) programme has been implemented in the Loess

Plateau to alleviate the severe soil erosion by converting steeply sloping croplands to

forestlands or grasslands. To quantify the effects of these conservation efforts, this study

identified the land-use changes between 2000 and 2015 and quantified their impacts on

runoff and erosion using the Soil and Water Assessment Tools (SWAT) and a typical hilly

watershed of the Yanhe River as a case study. To heighten the applicability of SWAT to the

region, major model parameters were localized and calibrated for the period of 1975–1980

and were then validated for 1981–1987. The R2 and NS validation indices were 0.70 and 0.65

for the monthly runoff and 0.67 and 0.61 for the sediment load, indicating that the model

performance was acceptable. Between 2000 and 2015, the slope croplands were reduced by

39.9%, the forestlands increased by 90.2% and the grasslands increased by 12.9%. These

land-use changes were simulated using SWAT to reduce the watershed runoff by 13.8% and

the sediment load by 50.7%. Spatial analyses using ArcGIS indicated that the simulated

reduction in water yield due to cropland conversion to forestland was more obvious than that

due to the conversion to grassland, but the reductions in the sediment yields were similar. The

results suggest that the GFG practice during this period was effective for preventing soil and

water losses.

This article is protected by copyright. All rights reserved.

KEY WORDS: land-use change; SWAT; runoff; soil erosion; the Loess Plateau

INTRODUCTION

Human activities have caused worldwide land cover changes (Foley et al., 2005), inducing

various environmental problems. The conversion of naturally vegetated areas for agricultural

uses can induce strong increases in runoff (García-Ruiz, 2010) and soil erosion (Tarolli &

Sofia, 2016), whereas vegetation restoration reduces these impacts (Zhang et al., 2015).

However, quantitatively distinguishing the impacts of land-use change from the concurrent

climate variability and other anthropogenic activities remains a challenge (Li et al., 2015).

Recent applications of natural tracers, such as lead-210, proved to be very effective for

detecting the effects of climate change during the past 20-30 years (Porto et al., 2016), and

hydrologic modelling was able to identify the effects of land-use changes (Bieger et al.,

2015). Models like SWAT (the Soil and Water Assessment Tools), which have the

advantages of flexible scenario settings and the ability to detect spatial variations, can

quantify the individual impacts of land-use changes under otherwise identical conditions (Li

et al., 2009). SWAT has been applied to and shown to perform well in various regions of the

world (Gassman et al., 2007), including China’s Loess Plateau (Xu et al., 2013; Yin et al.,

2017).

The Loess Plateau is a semi-arid region with serious soil erosion that contributes more

than 90% of the sediment load of the Yellow River (Yue et al., 2014). To alleviate the severe

land degradation in this region, the Chinese government implemented the Grain-for-Green

This article is protected by copyright. All rights reserved.

(GFG) programme in 1999, which mandated (by grain and financial subsidies to the involved

farmers) converting steeply sloping croplands to forestlands or grasslands as well as

afforestation. Stimulated by this ecological restoration programme, the land-use has changed

significantly (Li & Lu, 2015), providing an excellent opportunity to study the consequent

impacts of these types of change on the hydrologic cycle and soil erosion, which will

contribute to the understanding of the effectiveness of revegetation in reducing water and soil

losses in semi-arid areas. Previously, two studies have estimated the changes in erosion in the

Loess Plateau between 2000 and 2008 with the Universal Soil Loss Equation (USLE) (Fu et

al., 2011) and between 2000 and 2010 with the revised USLE (Sun et al., 2014), both of

which found that the soil erosion was obviously reduced, mainly as a result of land-use

change. Four modelling studies concerning the plateau adopted SWAT to detect the

hydrological effects of land-use change during 1990–2010 in the Fenhe watershed (Lu et al.,

2015), during 1985–2000 in the Heihe (Li et al., 2009) and during 1979–2006 in the Jinghe

(Yin et al., 2017), as well as the runoff and erosion impacts of land-use changes during 1980–

2005 in the Huangfuchuan watershed (Zuo et al., 2016).

The above studies well documented the effects of land-use change on runoff or erosion,

but most did not analyse the effects on both simultaneously and thus were not able to identify

the differences in the responses of runoff and erosion. Additionally, the duration of the GFG

implementation was relatively short, and thus, the effects of the implementation were not

observed thereafter. In 2015, GFG had been in practice for 16 years, so the effects of land-use

change could have stabilized, meaning that now is the right time for further evaluation of the

This article is protected by copyright. All rights reserved.

programme effectiveness. For this purpose, this study selected the areas with observed

land-use changes between 2000 and 2015 to represent the GFG-promoted vegetation

restoration and then estimated the impacts on runoff and erosion using SWAT and the Yanhe

River, which is a typical hilly gully loess watershed in the Loess Plateau, for a case study.

First, SWAT was calibrated and validated using observational data; then, the land-use

changes between 2000 and 2015 were detected, and finally, the calibrated model was used to

quantify the effects of the land-use change on the runoff and sediment yield in the Yanhe

watershed.

MATERIAL AND METHODS

Study area

The Yanhe River [36°21´–37°19´ N, 108°38´–110°29´ E] in northern Shaanxi Province

(Figure 1) is one of the main sediment contribution tributaries of the Yellow River (Yue et

al., 2014), with a total length of 284.3 km, an area of 7687 km2 and an altitude of 479–1787

m asl. The loess landform is fragmented by densely distributed ravines (3–8 km km-2) and

52.1% of the area is characterized as steep-slope lands, with slopes greater than 25°. With a

temperate semi-arid monsoon climate, the mean annual temperature is 10.1 °C, and the mean

annual precipitation is 509 mm (330 to 871 mm during 1961–2010), 78% of which occurs in

the rainy season from June to October. This area is one of the watersheds characterized by the

over-cultivation of slope lands and severe erosion of the Loess Plateau. Since the watershed

was selected as a key area of the GFG programme in 1999, its land use/cover has changed

substantially.

This article is protected by copyright. All rights reserved.

Weather, hydrological and land-use data

Daily weather data, including the air temperature, precipitation, etc., from 1961–2010 and 8

meteorological stations in and around the watershed were collected from the China

Meteorological Data Sharing Service System (http://data.cma.cn/). From the Hydrologic

Yearbook of China, additional daily precipitation data from 1961–2010 and 8 gauging sites

and the monthly runoff and sediment load data from 1975–1987 at the Ganguyi and Yan’an

hydrologic stations were collected. The station locations are presented in Figure 1. Spatial

data including the DEM, land use and soil types were collected from various sources (Table

I).

Modelling application

SWAT model

The ArcGIS extension version of SWAT2012 was used in this study. SWAT is a watershed

hydrologic model that is widely used to predict climate and land-use change impacts on

hydrological cycles and sediment yields in large complex watersheds over long periods

(Gassman et al., 2007). In SWAT, the watershed is divided into subbasins and hydrological

response units (HRU, defined as a unique combination of land use, soil and slope) within

each subbasin. Thus, the simulation comprises two phases: the first calculates the water and

sediment yields of each HRU, and the second traces the amount reaching the channel in each

subbasin (i.e., the summation of the runoff generated and sediment delivered in the HRUs) to

the watershed outlet. For each HRU, the runoff is simulated on a daily basis using the SCS

(Soil Conservation Service) curve number method, taking into account infiltration,

This article is protected by copyright. All rights reserved.

percolation, lateral flow generation, soil evaporation, plant water uptake, and groundwater

recharge. The sediment yield is estimated with the modified USLE, in which the USLE

rainfall erosivity factor is substituted with the surface runoff energy, as computed with the

simulated daily runoff volume and peak flow rate (Neitsch et al., 2011). For instream

processes, water is routed using the variable storage routing method of Williams (1969), in

which the outflow volume for a given river segment is calculated using a continuity equation

of the inflow volume from upstream river segment and storage volume of the previous time

step. The outflow rate is calculated using Manning’s equation. Sediment is routed by

calculating the deposition and degradation in the channel based on the relationship between

the upstream sediment load (USL) and transport capacity (TC). TC is calculated as an

exponential function of the peak flow rate (qch,pk) in a cross-sectional area (Ach) using the

simplified Bagnold equation (Williams, 1980) and is presented as 0.0001 * (qch,pk /Ach) by

taking the SWAT default value of 1.0 for the exponent coefficient. Deposition is calculated as

the product of (USL – TC) and the water volume in the channel. Degradation is the product

of (TC – USL), the water volume, and the channel erodibility and cover factors.

Subbasin delineation and parameter localization

The watershed was divided into 26 subbasins based on the DEM data, with the critical source

area specified to 152 km2 (2% of the watershed). The watershed was then divided into 1338

HRUs according to the classes of land use (6 types), soil (10 types), and slope steepness (5

grades: 0–6°, 6–10°, 10–15°, 15–25° and >25°), by merging the units with areas accounting

for <10% of the subbasin. The slope gradient was computed by SWAT using the DEM data,

This article is protected by copyright. All rights reserved.

but the slope length was calculated by LS-TOOL since SWAT cannot give a reasonable

estimation for steep areas (Yu et al., 2012). The LS-TOOL was developed by Zhang et al.

(2013) for steep terrains and was verified with measured slope lengths of 200 samples in a

small catchment of Xiannangou in the Yanhe River. The soil texture was derived from the

China Soil Scientific Database (http://www.soil.csdb.cn/) and was converted to the USA

standard by cubic spline interpolation (Cai & Zhang, 2003).

A total of 13 major parameters were localized (Table II). For the other parameters,

excepting the crop sowing/harvesting dates that were based on local data, the SWAT default

values were used. GW_DELAY was set to 400, as groundwater is very deep in this region

and thus has a long time lag (Feng et al., 2016). EPCO was set to 1.0 and ESCO was set to

0.4 to allow more transpiration and soil evaporation from deeper soil layers because the loess

soils have a deep active water layer wherein water is actively exchanged among different soil

layers and because plants have the ability to effectively use water as deep as 2 m (Han et al.,

1990). CH_K1 was set to 40 as the extensively distributed check dams in the watershed have

trapped substantial surface runoff (Xu et al., 2013), resulting in a large transmission loss

before the surface runoff reaches the main channel. For the GWQMN that controls the

groundwater generation, the SWAT default value of 500 was used.

Model calibration, validation and evaluation

The localized model parameter values were calibrated by comparing the simulated and

observed monthly data of the runoff and sediment loads during 1975–1980 at Ganguyi station.

The land-use data of 1975 was used, and the daily simulation results were summed to

This article is protected by copyright. All rights reserved.

monthly values. As more than 95% of the sediment in the region is eroded in the rainy season

from June to October (Xu et al., 2012b), the calibration for the sediment load was performed

with the observational data from June to October instead of for all months, as was done for

runoff. The calibration was performed using SWAT-CUP program (Abbaspour et al., 2007)

following the procedure described in Appendix S1. Using intensive calculations, the sensitive

parameters were identified and their optimized values or adjustment coefficients were

obtained and are presented in Table III.

After calibration, the model was used to simulate the runoff and sediment loads with

daily weather data and the land-use data of 1986 for the validation period of 1981–1987. By

summing the daily simulation results to monthly values, the simulated monthly runoffs and

sediment loads were examined with the observed data at the Ganguyi and Yan’an hydrologic

stations using 4 indices, including R2, NS, PBIAS and RMSE (Equations 1–4), based on the

obs
simulated (Yisim ), observed (Yiobs ) and mean observed (Ymean ) monthly data (Moriasi et al.,

2007). R2 and NS describe the goodness of fit of the simulated and observed values, while

PBIAS indicates the percent bias error of the simulated and observed data and RMSE shows

the mean deviation magnitude of all observation-simulation pairs.

2
∑𝑛 𝑜𝑏𝑠 𝑜𝑏𝑠
𝑖=1(Y𝑖 −Y𝑚𝑒𝑎𝑛 )
2
R = 2 (1)
∑𝑛 𝑠𝑖𝑚 −Y𝑜𝑏𝑠 )
𝑖=1(Y𝑖 𝑚𝑒𝑎𝑛

2
∑𝑛 𝑜𝑏𝑠 𝑠𝑖𝑚
𝑖=1(Y𝑖 −Y𝑖 )
NS = 1 − [ 2 ] (2)
∑𝑛 𝑜𝑏𝑠 𝑜𝑏𝑠
𝑖=1(Y𝑖 −Y𝑚𝑒𝑎𝑛 )

∑𝑛 𝑜𝑏𝑠 𝑠𝑖𝑚
𝑖=1(Y𝑖 −Y𝑖 )
PBIAS = 100 ∗ ∑𝑛 𝑜𝑏𝑠 (3)
𝑖=1 Y𝑖

This article is protected by copyright. All rights reserved.

 𝑛 𝑜𝑏𝑠 2
∑ (Y −Y𝑠𝑖𝑚 )
RMSE = √ 𝑖=1 𝑖 𝑖
(4)
𝑛

Estimation of land-use change effects

Two modelling scenarios were defined and are called LU2000 and LU2015. The first was

based on the land-use data of 2000, and the second was based on that of 2015, representing

the land-use conditions at the beginning of the GFG programme and that after 15 years of

implementation, respectively. For both scenarios, aside from the land-use data, the other

inputs and parameters were identical. The evaluation period was set to the 50 years of 1961–

2010 to ensure that the climate reflects the general conditions of the region.

Using the daily weather data from 1961–2010, two simulations were run with the

land-use data of 2000 for the scenario LU2000 and then with that of 2015 for LU2015. The

differences in the 50 year averages of the simulated runoffs and sediment loads of the two

scenarios were deemed as the effects of the land-use changes between 2000 and 2015, given

the identical climate conditions, i.e., those of 1961–2010, and the same watershed conditions

as the 1970s of the calibration period. The mean annual values of the simulation period at the

HRU level were converted with ArcGIS to gridded maps with 30 m resolutions to detect

spatial variations of the water and sediment yield reductions by spatial calculations.

RESULTS

Model performance

A total of 9 sensitive parameters and their ranks were identified (Table III), including 5 for

both the surface runoff and sediment yields and only 2 alone for the separate groundwater

and sediment yields. For CN2, SOL_AWC, USLE_C and USLE_K (Table III) which

This article is protected by copyright. All rights reserved.

significantly affected surface water and sediment yields, the calibrated values were increased

by 0.15 to 0.40 times to account for the spatial heterogeneities and scaling effects, such as the

spatial variations of the slope steepness. The initial values were given based on plot

observations or average regional conditions and thus probably underestimated the values for

steep-slope lands. The other 5 parameters, excepting GWQMN, were calibrated with a new

value.

Figure 2 presents the simulated and observed monthly runoff and sediment load at

Ganguyi station as well as the monthly precipitation. The simulated results were generally

well matched with those observed and varied consistently with precipitation. However,

evident overestimations were observed in few months, such as August 1981 and July 1984,

and underestimations were observed in the years with no flooding events, such as 1980.

Moreover, an evident underestimation in July 1977 was observed during the extreme flooding

event on July 6th.

The R2 and NS values were 0.70 and 0.65 for runoff and 0.67 and 0.61 for the sediment

load during the validation period at Ganguyi station, which are lower than those observed

during the calibration (Table IV). At Yan’an station, the R2 and NS values were 0.63 and 0.67

for the runoff, and 0.65 and 0.51 for the sediment load. The PBIAS indicated that the

simulated runoff and sediment load were underestimated by 11.2% and 9.1%, respectively,

during the calibration period, but were overestimated by 10.0% and 7.8%, respectively,

during the validation period. RMSE was high (Table IV), showing relatively large deviations

between the simulated and observed monthly values. In general, the model performance

This article is protected by copyright. All rights reserved.

could be rated as acceptable according to the criteria of R2>0.5, NSE>0.5 and PBIAS≤±

25%, as summarized from the previous studies by Moriasi et al. (2007).

Land-use change effects

Table V presents the land-use structures of the four selected years and their changes in

different periods. From 1975 to 2000, the slope cropland continuously expanded from 30.8%

of the total area to 41.1% and the forestlands expanded from 5.2% to 11.2%, at the cost of the

grasslands, which shrank from 61.6% to 45.0%. After the implementation of GFG, the slope

cropland area in 2000 shrank by 39.9% to account for 24.7% of the region in 2015. However,

the forestlands nearly doubled and the grasslands were obviously restored to account for 21.3%

and 50.8% of the region, respectively. These land-use changes occurred largely in sloping

areas (Figure 3), implying that the changes were mainly induced by the conversion of slope

croplands to forestlands and grasslands.

The simulated annual runoff and sediment loads of the two scenarios are presented in

Figure 4, showing apparent decreases of LU2015 compared to LU2000 in most years. The

mean annual simulation runoff depth was 31.3 mm with the standard deviation (SD) of 12.5

in LU2000, which decreased by 13.8% to 27 mm (SD=10.7) in LU2015. The mean total

sediment load was 36.5 Mt a-1 (SD=30.4) in LU2000, which was reduced by 50.7% to 18.0

Mt a-1 (SD=16.8) in LU2015. At the monthly scale, the simulated reduction was much greater

in the period of July to September, contributing 66.8% of the annual reduction in runoff and

76.2% of the annual reduction in sediment load.

This article is protected by copyright. All rights reserved.

 Figure 5 presents the spatial distributions of the mean annual water and sediment yields

for the scenarios LU2000 (A, B) and LU2015 (D, E) and the changes between the two

scenarios (C, F), showing obvious spatial differences. The simulated runoff and sediment

yield changes were small, ranging from -5% to 5% in 75% and 62% of the watershed area,

respectively. However, in 12% and 26% of the watershed area, the runoff and sediment yields

were reduced by more than 50%, respectively. In most of the remaining area, the reductions

ranged from 5–50%. In addition, some smaller areas showed increases, such as the area

adjacent to Yan’an City, where vegetation was cleared for urban construction. The greatest

change was mostly found in steep areas where slope croplands in 2000 were converted to

forestlands and grasslands in 2015. The estimated averages from ArcGIS of the simulated

results for the two scenarios indicated that the annual water and sediment yields were 57 mm

and 68 t ha-1 for the slope croplands (including the terraced areas) but were only 13 mm and

1.5 t ha-1 for forestlands and 30 mm and 4.4 t ha-1 for grasslands, respectively.

DISCUSSION

Modelling performance and uncertainties

The R2 and NS values of the validation period were 0.70 and 0.65 for the runoff, and 0.67 and

0.61 for the sediment load, which are comparable to those of previous modelling studies

using SWAT in the Loess Plateau. For the monthly runoff, the R2 and NS values are better

than those recorded in three previous studies, which ranged from 0.43–0.63 (Li et al., 2013;

Li et al., 2009; Zuo et al., 2016), but are lower than those of two other studies, which spanned

0.63–0.83 (Xu et al., 2013; Yin et al., 2017). For the monthly sediment load, the R2 and NS

This article is protected by copyright. All rights reserved.

values are lower than the values of 0.87 and 0.73 reported by Xu et al. (2013), but are better

than the values of 0.47 and 0.43 reported by Zuo et al. (2016) and 0.54 and 0.47 reported by

Li et al. (2013). In general, the calibrated model well reproduced the variation trends and

gave an acceptable estimation of the mean values. However, the model was unable to give an

accurate estimation of extreme rainfall events owing to parameter conditionality, occurrence

of hyperconcentrated flows, inaccurate observations and heterogeneity within HRUs, as

discussed in Appendix S2. As a result, evident errors in some months (Figure 2) and

relatively high RMSE values (Table IV) were observed. However, further evaluation using

the commonly-used ratio of the RMSE to the SD of the measured data indicated that the

value ratios were 0.48/0.72 for runoff and 0.43/0.63 for the sediment load for the

calibration/validation periods, respectively; thus, the results range from very good to

satisfactory compared to the criteria recommended by Moriasi et al. (2007).

The validation results showed obvious overestimations, as reported in the studies of other

watersheds (Li et al., 2013; Li et al., 2009; Zuo et al., 2016) in which the validations were

conducted for a similar period of the 1980s. The reason for this overestimation is partly

attributed to the increase in water and soil retention induced by the constructions of the check

dams during the 1980s (Xu et al., 2013), which were not sufficiently considered due to a lack

of relevant hydrological data. Nevertheless, as this study aimed to detect the changes in the

water and sediment yields induced by the land-use changes during 2000–2015 given identical

climate and engineering conditions, this inefficiency could not have significantly influenced

the final results.

This article is protected by copyright. All rights reserved.

 Parameter equifinality is a major uncertainty source of complex distributed models

(Abbaspour et al., 2007). In this study, the parameter set was not systematically examined to

recognize the possible equifinality since this needs a complex procedure (Beven, 2006). As

an alternative, the feasibility was indirectly verified by the satisfactory model performance

examined at another station of Yan’an (Table IV).

Land-use change impacts on runoff and sediment load

Between 2000 and 2015, the land-use changes were estimated to reduce the sediment load by

50.7%, which is significantly greater than the estimated runoff reduction of 13.8%. A similar

simulation study of the drier watershed of Huangfuchuan in the north of the Yanhe watershed

(Zuo et al., 2016) also found that cropland shrinkage (-3.7%) and forestland expansion

(14.7%) could reduce the sediment load (40.6%) more than the runoff (25.3%). In other

regions, the difference in the reduction was even more significant. For instance, with

significant cropland shrinkage, the modelled sediment load was reduced by 42%, but the

runoff was reduced by only 0.7% in the humid watershed of the Xiangxi River in southern

China (Bieger et al., 2015). With cropland expansion, the modelled runoff and sediment

loads increased by 7% and 60-fold in the Dijle watershed, Belgium (Alexander et al., 2015).

The differences in the responses of runoff and sediment load could be explained by the runoff

response to land-use changes being much more complex than that of sediment yields. For

runoff, the cropland converted to forestland/grassland could significantly reduce the surface

flow as a result of increased soil infiltration (Loch, 2000), but the impact on the base flow

was influenced by both changes in soil infiltration and evapotranspiration (Li et al., 2015).

This article is protected by copyright. All rights reserved.

Owing to the attenuation effects of vegetation on flood peaks due to their associated delays of

flow convergence, vegetation restoration could induce a significant decrease in the high-flow

rate and discharge and thus limit the sediment yield. However, revegetation would induce a

much smaller change in the total runoff (Zheng et al., 2014). In addition, the presence of

strong compensating effects of different land-use types within the hydrological system (Li et

al., 2015) can cause high resistance for runoff generation to land-use changes, particularly for

large scale watersheds and yearly simulations (Lin et al., 2015; Zhang et al., 2016). This

characteristic could also be another reason for why the runoff reduction was lower in the

studied watershed.

Sediment yield is closely related to the erosion susceptibility of the land-use type. The

slope cropland in the loess region is much more sensitive to soil erosion due to tillage

disturbances than forestlands and grasslands are. As a result, the slope cropland conversions

to forestlands and grasslands inevitably cause significant reductions of erosion. This

estimated reduction due to land-use conversion does not sufficiently consider the changes

within individual land-use types for the period of 2000–2015 due to the lack of relevant

spatial data. The 2015 land-use management of croplands has considerably improved

compared to that of 2000, and due to strict forest protections and prohibition of grazing, the

forest and grassland coverages were observed to increase by field surveys and using a series

of high resolution images from Google Earth. Not including these improvements to the land

use of 2015 could lead to underestimations of the protective effects of vegetation restoration.

This article is protected by copyright. All rights reserved.

 However, an overestimation of the soil conservation effects may also exist as the model

cannot simulate gully and gravitational erosion. Observations in the small hilly catchments of

the Loess Plateau found that revegetation can reduce hillslope erosion but cannot effectively

control gully erosion and gravitational erosion; thus, this strategy is incapable of lowering the

sediment concentrations in streams and has the same efficiency in reducing storm runoff and

sediment yield (Zheng et al., 2007; Yan at al., 2015). Modification of soil erosion estimation

method for regions where gully and gravitational erosion prevailed is required.

Spatial variability of water and sediment yield changes

The significant reductions in the simulated water and sediment yields were generally

distributed in accordance with the forestland and grassland expansions and slope cropland

shrinkages (Figures 3, 5). The spatial analysis results showed that with slope cropland

conversion to grasslands, the water yield (mm) was simulated to reduce by 46.8% and with

the slope cropland conversion to forestlands, the water yield was simulated to reduce by 77%,

indicating a considerable difference between the two conversions in associated water yield

reductions. In contrast, slope cropland conversions to forestlands and grasslands were

estimated to greatly reduce the sediment yields (t ha-1) at a similar extent, by an average of

97.8% and 93.5%, respectively.

Several studies based on experimental plot measurements in the region obtained similar

results, confirming that forestlands could reduce the water yields more strongly than

grasslands, but their effects on sediment yield reductions were not very different. For instance,

Xu et al. (2012a) measured the water and sediment yields for 2003–2008 in the experimental

This article is protected by copyright. All rights reserved.

plots in Yangou, a small subwatershed of Yanhe, and found that the water yields were 30–60%

lower in the grasslands and 86–90% lower in the forestlands than those in the slope croplands,

while the sediment yields were more than 90% lower than those in the slope croplands for

both land uses. One reason for these reductions could be that forest transpiration is much

higher than that of grass, resulting in lower water yields (West & Foley, 2011).

Implications

The simulated results indicated that the slope croplands restored to forestlands and grasslands

could have similar mitigation effects on soil erosion, but considerably different mitigation

effects on water yields. Due to the limitation of precipitation in the semi-arid Loess Plateau,

planting trees for land restoration has been debated in recent decades because this strategy

can cause soil desiccation problems (Wang et al., 2008). A recent study by Feng et al. (2016)

indicated that the Loess Plateau is probably reaching its sustainable limit for supporting tree

growth, and thus, further expansion of forest vegetation may exacerbate the water shortage

problem for both ecological and human demands. Therefore, despite the potential benefits of

regional temperature and moisture regime regulation from increased forest coverage (Li & Lu,

2015), further afforestation of the Loess Plateau should be executed with caution, particularly

in the areas with annual precipitations of less than 500 mm.

CONCLUSIONS

SWAT was calibrated and validated for the Yanhe watershed using observational data from

1975–1980 and 1981–1987. The evaluation indices R2 and NS ranged from 0.75–0.87 during

the calibration period and from 0.61–0.70 during the validation period, indicating that the

This article is protected by copyright. All rights reserved.

model performed well. The calibrated model can well estimate the water and sediment yields

of normal rainfall events, but cannot accurately estimate the values of extreme rainfall events.

Between 2000 and 2015 in the studied watershed, the slope cropland decreased by 39.9%,

while the grasslands increased by 12.9% and the forestlands increased by 90.2%. The

land-use changes in the watershed were simulated to reduce the local sediment load by 50.7%

and the local runoff by 13.8%. Spatial analyses revealed that cropland conversions to

forestlands and grasslands might have similar mitigation effects on soil erosion but different

effects on water yields. The results suggest that during the implementation period of the GFG

programme, the local serious land degradation was efficiently alleviated.

ACKNOWLEDGEMENTS

This work was jointly supported by the National Key R&D Plan of China (2017YFA0604701;

2012CB955304) and the National Natural Science Foundation of China (41671093). We

appreciate the critical and constructive comments from the anonymous reviewers and the

English polishing by Wiley Editing Services.

SUPPORTING INFORMATION

Appendix S1. Parameter calibration procedures

Appendix S2. Discussion on deviations between the simulated and observed values

This article is protected by copyright. All rights reserved.

REFERENCES

Abbaspour KC, Vejdani M, Haghighat S. 2007. SWAT-CUP Calibration and Uncertainty Programs for

SWAT. Modsim 2007: International Congress on Modelling and Simulation: Land, Water and

Environmental Management: Integrated Systems for Sustainability: 1603-1609.

Alexander P, Rounsevell M, Dislich C, Dodson J, Engstrom K, Moran D. 2015. Drivers for global

agricultural land use change: The nexus of diet, population, yield and bioenergy. Global

Environmental Change-Human and Policy Dimensions 35: 138-147. DOI:

10.1016/j.gloenvcha.2015.08.011

Beven K. 2006. A manifesto for the equifinality thesis. Journal of Hydrology 320: 18-36. DOI:

10.1016/j.jhydrol.2005.07.007

Bieger K, Hörmann G, Fohrer N. 2015. The impact of land use change in the Xiangxi Catchment (China)

on water balance and sediment transport. Regional Environmental Change 15: 485-498. DOI:

10.1007/s10113-013-0429-3

Chaplot V, Saleh A, Jaynes DB. 2005. Effect of the accuracy of spatial rainfall information on the

modeling of water, sediment, and NO3–N loads at the watershed level. Journal of Hydrology 312:

223-234. DOI: 10.1016/j.jhhydrol.2005.02.019

Feng X, Fu B, Piao S, Wang S, Ciais P, Zeng Z, Lü Y, Zeng Y, Li Y, Jiang X. 2016. Revegetation in China's

Loess Plateau is approaching sustainable water resource limits. Nature Climate Change 6: 1019.

DOI: 10.1038/NCLIMATE3092

Foley JA, Defries R, Asner GP, Barford C, Bonan G, Carpenter SR, Chapin FS, Coe MT, Daily GC, Gibbs

HK. 2005. Global consequences of land use. Science 309: 570-574. DOI: 10.1126/science.1111772

This article is protected by copyright. All rights reserved.

Fu BJ, Liu Y, Lu YH, He CS, Zeng Y, Wu BF. 2011. Assessing the soil erosion control service of

ecosystems change in the Loess Plateau of China. Ecological Complexity 8: 284-293. DOI:

10.1016/j.ecocom.2011.07.003

Fu BJ, Zhao WW, Chen LD, Zhang QJ, Lu YH, Gulinck H, Poesen J. 2005. Assessment of soil erosion at

large watershed scale using RUSLE and GIS: A case study in the Loess Plateau of China. Land

Degradation & Development 16: 73-85. DOI: 10.1002/ldr.646

García-Ruiz JM. 2010. The effects of land uses on soil erosion in Spain: A review. Catena 81: 1-11. DOI:

10.1016/j.catena.2010.01.001

Gassman PW, Reyes MR, Green CH, Arnold JG. 2007. The Soil and Water Assessment Tool: Historical

Development, Applications, and Future Research Directions. Transactions of the ASABE 50:

1211-1250. DOI: 10.13031/2013.23637

Han S, Li Y, Shi Y, Yang X, Zhang X, Shi Z. 1990. The characteristic of soile moisture resources on the

Loess Plateau. Bulletin of Soil and Water Conservation 10: 36-43. (in Chinese with English

abstract)

Li Q, Yu X, Xin Z, Sun Y. 2013. Modeling the Effects of Climate Change and Human Activities on the

Hydrological Processes in a Semiarid Watershed of Loess Plateau. Journal of Hydrologic

Engineering 18: 401-412. DOI: 10.1061/(asce)he.1943-5584.0000629

Li W, Lu C. 2015. Aridity trend and response to vegetation restoration in the loess hilly region of northern

Shaanxi Province. Journal of Geographical Sciences 25: 289-300. DOI:

10.1007/s11442-015-1168-9

Li Z, Liu WZ, Zhang XC, Zheng FL. 2009. Impacts of land use change and climate variability on

This article is protected by copyright. All rights reserved.

 hydrology in an agricultural catchment on the Loess Plateau of China. Journal of Hydrology 377:

35-42. DOI: 10.1016/j.jhydrol.2009.08.007

Li ZH, Deng XZ, Wu F, Hasan SS. 2015. Scenario Analysis for Water Resources in Response to Land Use

Change in the Middle and Upper Reaches of the Heihe River Basin. Sustainability 7: 3086-3108.

DOI: 10.3390/su7033086

Lin B, Chen X, Yao H, Chen Y, Liu M, Gao L, James A. 2015. Analyses of landuse change impacts on

catchment runoff using different time indicators based on SWAT model. Ecological Indicators 58:

55-63. DOI: 10.1016/j.ecolind.2015.05.031

Liu J. 2013. Research of Spatiotemporal Trends of Erosion and Sediment and Driving Factors on the

He-Long Region Recent 30 Years. In Research Center of Soil and Water Conservation and

Ecological Environment. Chinese Academy of Sciences. (in Chinese with English abstract)

Liu J, Kuang W, Zhang Z, Xu X, Qin Y, Ning J, Zhou W, Zhang S, Li R, Yan C, Wu S, Shi X, Jiang N, Yu

D, Pan X, Chi W. 2014. Spatiotemporal characteristics, patterns and causes of land use changes in

China since the late 1980s. Acta Geographica Sinica 69: 3-14. (in Chinese with English abstract)

Loch RJ. 2000. Effects of vegetation cover on runoff and erosion under simulated rain and overland flow

on a rehabilitated site on the Meandu Mine, Tarong, Queensland. Australian Journal of Soil

Research 38: 299-312. DOI: 10.1071/SR99030

Lu ZX, Zou SB, Qin ZD, Yang YG, Xiao HL, Wei YP, Zhang K, Xie JL. 2015. Hydrologic Responses to

Land Use Change in the Loess Plateau: Case Study in the Upper Fenhe River Watershed. Advances

in Meteorology 3: 1-10. DOI: 10.1155/2015/676030

Luo L, Zhang K, Fu S. 2002. Application of Runoff Curve Number Method on Loess Plateau. Bulletin of

This article is protected by copyright. All rights reserved.

 Soil and Water Conservation 22: 58-61, 68. (in Chinese with English abstract)

Luo Y, Yang S, Zhao C, Liu X, Liu C, Wu L, Zhao H, Zhang Y. 2014. The effect of environmental factors

on spatial variability in land use change in the high-sediment region of China's Loess Plateau.

Journal of Geographical Sciences 24: 802-814. DOI: 10.1007/s11442-014-1121-3

Moriasi DN, Arnold JG, Liew MWV, Bingner RL, Harmel RD, Veith TL. 2007. Model Evaluation

Guidelines for Systematic Quantification of Accuracy in Watershed Simulations. Transactions of

the ASABE 50: 885-900. DOI: 10.13031/2013.23153

Moriasi DN, Starks PJ. 2010. Effects of the resolution of soil dataset and precipitation dataset on

SWAT2005 streamflow calibration parameters and simulation accuracy. Journal of Soil & Water

Conservation 65: 1394–1403. DOI: 10.2489/jswc.65.2.63

Neitsch SL, Arnold JG, Kiniry JR, Williams JR. 2011. Soil and water assessment tool theoretical

documentation version 2009. Technical Report No. 406. Grassland. Soil and Water Research

Laboratory, Agricultural Research Service, Blackland Research Center. Texas Agricultural

Experiment Station: Temple, Texas.

Nielsen A, Trolle D, Me W, Luo L, Han BP, Liu Z, Olesen JE, Jeppesen E. 2013. Assessing ways to combat

eutrophication in a Chinese drinking water reservoir using SWAT. Marine & Freshwater Research

64: 475-492. DOI: 10.1071/MF12106

Porto P, Walling DE, Callegari G. 2009. Investigating the effects of afforestation on soil erosion and

sediment mobilisation in two small catchments in Southern Italy. Catena 79: 181-188. DOI:

10.1016/j.catena.2009.01.007

Porto P, Walling DE, Cogliandro V, Callegari G. 2016. Exploring the potential for using Pb-210(ex)

This article is protected by copyright. All rights reserved.

 measurements within a re-sampling approach to document recent changes in soil redistribution

rates within a small catchment in southern Italy. Journal of Environmental Radioactivity 164:

158-168. DOI: 10.1016/j.jenvrad.2016.06.026

Ran D, Yao W, Wu Y, Li X, Lv X. 2014. Effect of forest-grass vegetation on reducing flood and sediment

from 1997 to 2006 in Yanhe basin. Science of Soil and Water Conservation 12: 1-9. (in Chinese

with English abstract)

Saxton KE, Willey PH, Rawls WJ. 2006. Field and pond hydrologic analyses with the SPAW model. In

ASABE Annual International Meeting. ASABE: Portland, Oregon.

Sun L, Fang H, Qi D, Li J, Cai Q. 2013. A review on rill erosion process and its influencing factors.

Chinese Geographical Science 23: 389-402. DOI: 10.1007/s11769-013-0612-y

Sun W, Shao Q, Liu J, Zhai J. 2014. Assessing the effects of land use and topography on soil erosion on the

Loess Plateau in China. Catena 121: 151-163. DOI: 10.1016/j.catena.2014.05.009

Tarolli P, Sofia G. 2016. Human topographic signatures and derived geomorphic processes across

landscapes. Geomorphology 255: 140-161. DOI: 10.1016/j.geomorph.2015.12.007

Wang L, Wang Q, Wei S, Shao MA, Li Y. 2008. Soil desiccation for Loess soils on natural and regrown

areas. Forest Ecology & Management 255: 2467-2477. DOI: 10.1016/j.foreco.2008.01.006

Wei S, Dai Y, Liu B, Zhu A, Duan Q, Wu L, Ji D, Ye A, Yuan H, Zhang Q, Chen D, Chen M, Chu J, Dou Y,

Guo J, Li H, Li J, Liang L, Liang X, Liu H, Liu S, Miao C, Zhang Y. 2013. A China data set of soil

properties for land surface modeling. Journal of Advances in Modeling Earth Systems 5: 212-224.

DOI: 10.1002/jame.20026

West PC, Foley JA. 2011. An alternative approach for quantifying climate regulation by ecosystems.

This article is protected by copyright. All rights reserved.

 Frontiers in Ecology & the Environment 9: 126-133. DOI: 10.1890/090015

Williams JR. 1969. Flood routing with variable travel time or variable storage coefficients. Transactions of

the ASABE 12: 100-103.

Williams JR. 1980. SPNM, a model for predictiong sediment, phosphorus, and nitrogen yields from

agrivultural basins. Water Resources Buletinl 16: 843-848.

Xu J, Liu P, Deng R, Liu D. 2012a. Runoff and Sediment Reductions in the Different Stages of Vegetation

Restoration on a Loess Slope. Scientia Geographica Sinica 32: 1391-1396. (in Chinese with

English abstract)

Xu YD, Fu BJ, He CS. 2013. Assessing the hydrological effect of the check dams in the Loess Plateau,

China, by model simulations. Hydrology and Earth System Sciences 17: 2185-2193. DOI:

10.5194/hess-17-2185-2013

Xu YD, Fu BJ, He CS, Gao GY. 2012b. Watershed discretization based on multiple factors and its

application in the Chinese Loess Plateau. Hydrology and Earth System Sciences 16: 59-68. DOI:

10.5194/hess-16-59-2012

Yan QH, Lei TW, Yuan CP, Lei QX, Yang XS, Zhang ML, Su GX, An LP. 2015. Effects of watershed

management practices on the relationships among rainfall, runoff, and sediment delivery in the

hilly-gully region of the Loess Plateau in China. Geomorphology 228: 735-745. DOI:

10.1016/j.geomorph.2014.10.015

Yin J, He F, Xiong YJ, Qiu GY. 2017. Effects of land use/land cover and climate changes on surface runoff

in a semi-humid and semi-arid transition zone in northwest China. Hydrology and Earth System

Sciences 21: 183-196. DOI: 10.5194/hess-21-183-2017

This article is protected by copyright. All rights reserved.

Yu W, Nan Z, Li S, Li C. 2012. Average Slope Length Calculation and Runoff Simulation. Journal of

Geo-Information Science 14: 41-48. (in Chinese with English abstract)

Yue X, Mu X, Zhao G, Shao H, Gao P. 2014. Dynamic changes of sediment load in the middle reaches of

the Yellow River basin, China and implications for eco-restoration. Ecological Engineering 73:

64-72. DOI: 10.1016/j.ecoleng.2014.09.014

Zhang HM, Yang QK, Li R, Liu QR, Moore D, He P, Ritsema CJ, Geissen V. 2013. Extension of a GIS

procedure for calculating the RUSLE equation LS factor. Computers & Geosciences 52: 177-188.

DOI: 10.1016/j.cageo.2012.09.027

Zhang J, Wang TM, Ge JP. 2015. Assessing Vegetation Cover Dynamics Induced by Policy-Driven

Ecological Restoration and Implication to Soil Erosion in Southern China. Plos One 10: e0131352.

DOI: 10.1371/journal.pone.0131352

Zhang L, Nan Z, Yu W, Ge Y. 2016. Hydrological Responses to Land-Use Change Scenarios under

Constant and Changed Climatic Conditions. Environ Manage 57: 412-31. DOI:

10.1007/s00267-015-0620-z

Zhang Y, Liu BY, Shi PJ, Jiang ZS. 2001. Crop cover factor estimating for soil loss prediction. Acta

Ecologica Sinica 21: 1050-1056. (in Chinese with English abstract)

Zheng MG, Cai QG, Chen H. 2007. Effect of vegetation on runoff-sediment relationship at different spatial

scale levels in Gullied-hilly Area of the Loess Plateau,China. Acta Ecologica Sinica 27: 3572-3581.

DOI: 10.1016/S1872-2032(07)60075-4

Zheng MG, Sun LY, Yan M. 2014. Temporal change of runoff and sediment load and their differential

response to human activities: A case study for a well-vegetated mountain watershed of southern

This article is protected by copyright. All rights reserved.

 China. Journal of Mountain Science 11: 73-85. DOI: 10.1007/s11629-013-2751-0

Zuo DP, Xu ZX, Yao WY, Jin SY, Xiao PQ, Ran DC. 2016. Assessing the effects of changes in land use

and climate on runoff and sediment yields from a watershed in the Loess Plateau of China. Science

of the Total Environment 544: 238-250. DOI: 10.1016/j.scitotenv.2015.11.060

This article is protected by copyright. All rights reserved.

Table I Resolutions and sources of the DEM, soil map and land use maps

Input Resolutio
Source Description
data n

DEM 25 m The National Geomatics Center of Has a scale of 1:50,000

China
(http://ngcc.sbsm.gov.cn/)
Soil The Data Center for Resources and Extracted from Digital Soil Map of China
map Environmental Sciences, Chinese with the scale of 1:1,000,000
Academy of Sciences
(http://www.resdc.cn)
Land- 19 30 m The Loess Plateau Data Center Compiled by the supervised classification
use 75 (http://loess.geodata.cn) of Landsat MSS/TM images in
maps* 19 30 m combination with field surveys (Luo et al.,
86 2014)
20 30 m The Data Center for Resources and Visually interpreted using TM/ETM
00 Environmental Sciences, Chinese images (Liu et al., 2014)
Academy of Sciences
(http://www.resdc.cn)
20 30 m Interpreted by the authors Visually interpreted based on high
15 resolution (1 m) images from Google
Earth
* All land-use maps comprise the six land-use types of flat cropland, slope cropland, grassland, forestland, urban land, and
water body. In addition, the terrace area was estimated from the data of 1979, 1989, 2000 and 2011 as reported in the
literature (Liu, 2013; Ran et al., 2014), such that it covered 5.4%, 8.6%, 9.1% and 18.6% of the slope cropland in 1975, 1986,
2000 and 2015, respectively. For the model simulation, the slope cropland was considered a mixed type and was split into
slope and terraced croplands by specifying the proportion with the HRU definition command in SWAT.

This article is protected by copyright. All rights reserved.

Table II Descriptions and localized values as well as the data sources of the major parameters

Parameter Description Value Source

SOL_BD Soil bulk density (g cm-3) 1.20-1.61 Extracted from the Soil Database of China for Land Surface
SOL_CBN Soil organic carbon content (%) 0.19-0.58 Modelling (http://westdc.westgis.ac.cn) (Wei et al., 2013)
SOL_AWC Soil available water capacity (mm/mm) 0.15-0.17
SOL_K Soil saturation conductivity (mm/hr) 7.5-49.1 Estimated with the SPAW model using soil data (Saxton et al., 2006)

USLE_K Soil erodibility factor ((t h hm2) (hm2 MJ mm)-1) 0.034-0.044 Estimated with the equation suggested by Neitsch et al. (2011)
SOL_ALB Soil albedo 0.09-0.18 Estimated with the regression equation (Nielsen et al., 2013)
CN2 Runoff curve number of moisture condition Ⅱ 48-78 55/48 for grassland/forestland from the plot observations (Luo et al.,
2002), 78/73 for flat/slope cropland as estimated from the literature
(Neitsch et al., 2011)
USLE_C Minimum value of USLE land cover factor 0.004-0.5 0.12/0.5 for grassland/slope cropland from the plot observations (Fu
applicable to the land cover/plants et al., 2005; Zhang et al., 2001); 0.2/0.004* for flat
cropland/forestland estimated from Neitsch et al. (2011)
ESCO Soil evaporation compensation factor 0.4 Based on watershed characteristics (Han et al., 1990)
EPCO Plant uptake compensation factor 1.0 Based on watershed characteristics (Han et al., 1990)
CH_K1 Effective hydraulic conductivity in tributary channel 40 Based on watershed characteristics (Xu et al., 2013)
alluvium (mm/h)
GW_DELAY Groundwater delay (days) 400 Based on watershed characteristics (Feng et al., 2016)
GWQMN Threshold depth of water in the shallow aquifer 500 Taken from the SWAT database
required for return flow to occur (mm)
* The value was multiplied by ten for the forestland established between 2000 and 2015.

This article is protected by copyright. All rights reserved.

Table III Sensitivity ranks and optimized values of sensitive parameters

Sensitivity Ranks
Range Optimized
Parameter
update* value
Streamflow Sediment

CN2 1 1 R 0.15
SOL_AWC 2 4 R 0.40
GW_DELAY 3 - V 300
EPCO 4 7 V 0.92
GWQMN 5 - V 500
CH_K1 6 5 V 30
ESCO 7 6 V 0.28
USLE_C - 2 R 0.34
USLE_K - 3 R 0.15
* R indicates that the
initial parameter value
in Table II is replaced
by multiplying (1 +
optimized value) and V
by the optimized
value.

This article is protected by copyright. All rights reserved.

Table IV Evaluation results of modelling performances

Ganguyi Hydrologic Station Yan’an Hydrologic Station

Calibration Validation Validation

(1975-1980) (1981-1987) (1981-1987)

R2 0.76 0.70 0.63

NS 0.87 0.65 0.67
Streamflow
PBIAS -11.25 9.98 2.33
RMSE 4.92 4.14 3.28

R2 0.81 0.67 0.65

NS 0.75 0.61 0.51
Sediment
PBIAS -9.12 7.76 4.28
RMSE 8.80 3.64 3.05

This article is protected by copyright. All rights reserved.

Table V Land-use structures (% of the catchment area) in 1975, 1986, 2000 and 2015, and the area changes
during 1975–1986, 1986–2000 and 2000–2015

Area change (%)

Land-use type 1975 1986 2000 2015
1975-1986 1986-2000 2000-2015

Flat cropland 2.1 2.4 1.9 1.7 14.3 -20.8 -10.5

Slope cropland 30.8 34.5 41.1 24.7 12.0 19.1 -39.9
Grassland 61.6 54.6 45.0 50.8 -11.4 -17.6 12.9
Forestland 5.2 8.1 11.2 21.3 55.8 38.3 90.2
Urban land 0.2 0.2 0.5 1.1 6.6 181.3 155.3
Water 0.1 0.2 0.3 0.4 181.4 20.9 42.5

This article is protected by copyright. All rights reserved.

Figure 1 Map of the locations of the Yanhe watershed and gauge stations

This article is protected by copyright. All rights reserved.

Figure 2 Variations of the monthly precipitation, runoff (A) and sediment load (B) at the Ganguyi station

during the calibration (1975–1980) and validation (1981–1987) periods

This article is protected by copyright. All rights reserved.

Figure 3 Land-use maps of the Yanhe watershed in 2000 (A) and 2015 (B); spatial distribution of the

expanded forestlands and grasslands (C) as well as that of the reduced slope cropland (D) between 2000

and 2015

This article is protected by copyright. All rights reserved.

Figure 4 Annual simulated runoff depths (A) and sediment moduli (B) in the Yanhe watershed under the

two modelling scenarios

This article is protected by copyright. All rights reserved.

Figure 5 Spatial distributions of the average annual water and sediment yields at the HRU scale: A and B

represent the water yields and D and E represent the sediment yields of the LU2000 and LU2015

scenarios, respectively; C and F represent the changes in water yield and sediment yield of the two

scenarios

This article is protected by copyright. All rights reserved.