Is afforestation-induced land use change the main contributor to vegetation dynamics in the semiarid region of North China?

Ecological Indicators 88 (2018) 282–291

Contents lists available at ScienceDirect

Ecological Indicators journal homepage: www.elsevier.com/locate/ecolind

Original Articles

Is afforestation-induced land use change the main contributor to vegetation dynamics in the semiarid region of North China?

T

⁎

Fangtian Wanga,b, Pingli Ana,b, , Can Huanga,b, Zhe Zhanga,b, Jinmin Haoa,b a b

College of Land Science and Technology, China Agricultural University, Beijing 100193, China Key Laboratory of Agricultural Land Quality, Ministry of Land and Resources, China

A R T I C L E I N F O

A B S T R A C T

Keywords: Land use and cover change Vegetation dynamics Afforestation Agricultural intensification NDVI

Quantitatively analyzing the response of vegetation dynamics to land use change is very important, especially it relates to gaining a better understanding of the effects of ecological restoration projects. Previous studies have focused on the effects of land use change caused by the Grain for Green Project (GGP) on vegetation dynamics, but the effects from other changes in land use have less been explored. Therefore, in order to bridge this gap, Landsat images and MODIS Normalized Difference Vegetation Index (NDVI) data were used to examine how land use had changed from 2000 to 2014 as well as to study its influence on vegetation growth in Ulanqab City, Inner Mongolia, China. For this, four major land use change processes were identified through land use trajectories analysis: the Grain for Green Program (GGP), agricultural intensification, cropland abandonment, and cropland degradation. The GGP caused 36.60% of the total land use change in our study area, while three other processes caused the remaining of 44.40%. Anthropogenic activities significantly influenced vegetation coverage in 10.47% of the research area based on residual trend analysis. Pixels analysis showed 8.72% of the research area experienced a significant increase in vegetation coverage, where 39.53% of this increase was caused by afforestation while 33.25% was attributable to agricultural intensification. However, vegetation degradation was observed in 1.75% of the research area, of which 12.91% was caused by afforestation, an amount that was lower than that caused by the combined effects of the other three practices. Overall, afforestation can effectively increase vegetation coverage, but the overall effects can be undermined by other unstainable land use. However, although agricultural intensification contributed greatly to an increase in vegetation coverage, it also caused severe land degradation. This study demonstrates that ecological restoration projects and regional ecological systems are facing increasing pressure in Ulanqab City caused by an increase in human activity. Managing and maintaining restoration projects sustainably and appropriately in fragile areas will help land managers to achieve better results during vegetation restoration as well as to contribute to sustainable development.

1. Introduction Anthropogenic land use activities have transformed a large proportion of the planet’s land surface and degraded ecological conditions across the globe (Foley et al., 2005). These endeavors create land use change in dryland ecosystems worldwide that expose grassland and woodland to degradation, deforestation, and soil loss (Zika and Erb, 2009). The loss of vegetation in arid and semiarid zones has increased significantly over the last several decades, resulting in land degradation (Landmann and Dubovyk, 2014; Zewdie and Csaplovics, 2016). As grassland is cleared and exposed to severe landscape-level changes caused by extensive human activities, northwestern China is no different and has witnessed a deterioration of natural environment, resulting from the removal of vegetation, soil erosion, desertification, and

⁎

sandstorms. Afforestation has been used to restore vegetation coverage as part of a major ecological restoration program to control desertification and environmental deterioration in China (Chen et al., 2007; Le Houerou, 2000; Ma et al., 2013). Pushed into action by a series of serious floods in 1998, the Chinese central government launched some ecological restoration programs in the late 1990s and early 2000s. These included the Grain for Green Program (GGP) and the Grazing Withdrawal Program (GWP) that were introduced in northwest of China with particular emphasis on grassland (Yin et al., 2010). The GGP, also known as the “Sloping Land Conversion Program”, is usually explained as a program for “replacing cropping and livestock grazing in fragile areas with trees and grass.” To complement the efforts of the GGP, the national government also initiated the GWP, which aimed to conserve grassland by

Corresponding author. E-mail address: [email protected] (P. An).

https://doi.org/10.1016/j.ecolind.2017.12.061 Received 12 October 2016; Received in revised form 6 November 2017; Accepted 27 December 2017 1470-160X/ © 2018 Elsevier Ltd. All rights reserved.

Ecological Indicators 88 (2018) 282–291

F. Wang et al.

Thus, a combined analysis of land use and cover change with NDVI can help us to both discover the vegetation change of a certain region, and to gain the influent information about LUCC(Li et al., 2017). To evaluate the effects of afforestation on vegetation dynamics effectively, this study examined the types of land use change that have occurred in Ulanqab City, Inner Mongolia, China and the extent on which LUCC influenced vegetation growth and the changes it provoked. An ecological fragile region located in Ulanqab, Inner Mongolia was chosen for the research region. Our study focused on the following three objectives: 1) to quantify LUCC and identify the main land conversion types from 2000 to 2014. This temporal period was broken into two periods, 2000–2007 and 2007–2014; 2) to analyze temporal and spatial human induced change of vegetation coverage following the inception of the GGP; 3) to combine an analysis of LUCC and human induced NDVI changes in order to quantify the effects of LUCC.

banning of grazing and rotational grazing or by converting grazing land to cultivated pasture (Mu et al., 2013). These projects were applied with different ways in North China and created large changes in land use and land cover (Chen et al., 2008; Zhou et al., 2012). The strong connection between land use and land cover change (LUCC) is a well-recognized agent of environmental change and vegetation is well-known to respond accordingly. As a result, extensive studies have been carried out to evaluate the effectiveness of ecological restoration projects based on land use and vegetation dynamics. Currently, the major fields of study mainly include the monitoring the patterns of land cover change induced by the GGP (Wang et al., 2013; Zhang et al., 2014b; Zhou et al., 2012), analyses of vegetation dynamics (Li et al., 2013; Zhang et al., 2012), and studies of the relative contribution of climate and ecological restoration projects to vegetation restoration (Sun et al., 2015b; Tian et al., 2015; Tong et al., 2017). However, rigorous debate is emerging on the ecological effects of afforestation, some studies have shown that these projects have help to restore vegetation coverage (Chen et al., 2014a; Sun et al., 2015a); while other studies indicate that reforestation has a paradoxical effect that did not always lead to recover in vegetation coverage, even may result in increased ecosystem deterioration (Zhang et al., 2014a). These studies provide a solid background for further study in ecological fragile areas. Nevertheless, the main findings vary depending on different individual basins, time periods and data sources. Notably, numerous studies are focused on the effects of land conversions caused by the GGP on vegetation dynamics, however, current empirical evidence is limited regarding the effects that are caused by other types of land use change. Admittedly, land use change induced by the ecological projects has been largely related to vegetation coverage change, but it cannot be further from the truth. Human-environment systems in northern China have been experiencing unprecedented changes, resulting in drastic land use cover change with noticeable vegetation variations (Ge et al., 2015; Li et al., 2016b). For example, some studies have realized that agricultural production, such as irrigation, has great effects on vegetation growth and ecological environment (Li et al., 2017; Yang et al., 2014). However, current studies have documented specifics regarding land cover change and its influence on vegetation dynamics since the end of the program; these studies have neglected that fact that inferences drawn from other types of land use changes may result in over- or under-estimates of the effects of ecological projects. Because of this gap, we suggest that some previous studies have not paid sufficient attention to the latest changes land use that have occurred with the evolution of natural–human systems. Thus, a need exists to quantify the entire scope of land use change and its influence on vegetation coverage. With the development of air- and space-borne sensors over the past four decades, remote sensing technology and related methodologies have been adapted to assessing and monitoring LUCC (Du et al., 2002). Optical satellite data, such as data from Advanced Very High Resolution Radiometers, MODIS and Landsat sensors, have been widely used to detect land surface changes and assess regional environmental change in recent years. In particular, Landsat satellite images are ideal for mapping and monitoring land cover because they have a relatively high resolution of 30 m and the longest temporal record of space-based surface observations that extends over 40 years (Roy et al., 2014). Land cover data based on artificial classifications of these satellite images can show the effects of changes in human activities. This type of data can also allow the detection of difference in classifications and are effective for documenting distinct, abrupt anthropogenic impacts on the land surface (Wright et al., 2012). Therefore, we used high spatial and temporal resolution data to detect land use and cover change, displaying the result of human activities on land use. In addition, MODIS NDVI product are very useful in helping researchers to understand continuous and gradual changes such as land degradation through the use of time series analysis (Wright et al., 2012). The present study used MODIS NDVI data to quantify the effects of different human activities.

2. Methods 2.1. Study area The study area, located in Ulanqab City (39°37′–43°28′N, 109°16′–114°49′E), Inner Mongolia, China (Fig. 1), contains six county level administrative units. The landform types listed from north to south include the Mongolia Plateau, the Yin Mountain Range, and Loess Hills. The Yin Mountain Range runs from east to west, dividing the landscape into the Houshan and Qianshan regions. This region’s arid and semi-arid climate features an annual average temperature that increases from 3.1 °C in the north to 4.2 °C in the south. Similarly, annual average precipitation increases from 307 mm in the north to 376 mm in the south with more than 70% of the rainfall concentrated in August and September. Ulanqab has suffered severe soil erosion and undergone severe desertification over the last half century as a result of its fragile ecosystems and extensive human activity. The GGP and the GWP were implemented in this region in the early 2000s with the goal of restoring fragile environments. 2.2. Data and pre-processing 2.2.1. Landsat imagery Landsat (TM/ETM+) images downloaded from USGS (https:// glovis.usgs.gov/) with a spatial resolution of 30 m were used to interpret land cover data. Taking vegetation coverage and image quality into account, satellite images were captured in three months (June, July and August) on three years (2000, 2007 and 2014). We chose images with that corresponded to the timing of significant changes in policies. The GGP was implemented nationwide in 2000. The central government declared that no new cropland regions could be approved to join the project after the end of 2006 and the most important task of the GGP shifted from sloping land conversion to barren land afforestation at that time (Wang et al., 2013). Four field surveys from 2012 to 2015 were carried out to ascertain the characteristics of the landscapes in the study area, and training samples were constructed for different landscapes. Based on the ability to classify land use/land cover using Landsat images in the study area, the National Land Use Change Database Hierarchical Classification System was adopted and modified to include classes that can be readily identified on Landsat imagery in our study area (Liu et al., 2005). After the process of geometric correction and FLASSH atmospheric correction, in order to best describe the LULC, we reclassified the LULC patterns into seven types of areas based on local geographical conditions. The areas mapped included dryland, irrigated land, grass, forest, water, urbanized, land and unused land (mainly barren sand). Given that some classes have similar spectral characteristics and a single type of object may have different spectral characteristics, the change detection of different classified maps was then obtained in Arcgis 10.0 after the supervised classification. We applied this kind of change vector to the corresponding Landsat images, and 283

Ecological Indicators 88 (2018) 282–291

F. Wang et al.

Fig. 1. Location of the study area.

subsequently made a detailed revision in the post-classification for each classified map together with high-resolution satellite images from Google Earth. A total of 271 independent sample sites were randomly selected to assess classification accuracy. The overall classification accuracy in 2000, 2007, and 2014 was 88.7%, 89.2%, and 89.5%, respectively, and the respective Kappa coefficients were 0.80, 0.84, and 0.87.

checked. Annual precipitation and annual mean temperature for the growing season per station were calculated for each year. All the historical annual surfaces of precipitation and temperature were generated for the study area by using thin plate smoothing splines as implemented using ANUSPLIN software (Hong et al., 2005).

2.2.2. Vegetation coverage index The Normalized Difference Vegetation Index (NDVI) has been widely used as a proxy to analyze the effects of climatic events and human activities on vegetation dynamics because of its close relationship with biophysical and biochemical variables, such as vegetation coverage, green biomass, and growth conditions (Higginbottom and Symeonakis, 2014; Huete et al., 2002). The dataset used in this study was the surface reflectance 8-day L3 Global 250 m MODIS time series data product (MODIS 09Q1), which was acquired for the growing season (defined as from April to October) for the period of 2000–2014 and can better reflect the ground vegetation growth. We calculated the vegetation coverage index using ENVI/Interactive Data Language. NDVI was calculated using Eq. (1):

We first used a land use change transition matrix to reveal the detailed information related to land use conversions. Maps were paired (2000–2007 and 2007–2014) and overlaid in Arcgis software to produce two cross-tabulation matrices. Second, all land use types were coded as follows: 1, dryland; 2, irrigated land; 3, grass; 4, forest; 5, water; 6, urbanized land; 7, unused land. Then a temporal trajectory analysis was used to identify the paths by which land use change occurred through time (Wang et al., 2012). In this paper, based on overlaying vectors for all the classification maps, a distribution map of all trajectories was developed and trajectories codes for every polygon were acquired to trace the land use history. In all the trajectories, trajectories with the same class in each node, such as 111, 222, 333, and so on, stand for polygons with no land use conversion. Other trajectories such as 112, 123, and 144 represent the related change in land use type for a given polygon. For example, 144 shows the conversion from dryland to forest in second period where the land use type remained as forest through the third period. Third, according to the percentage of each land use trajectory in the total area, the main change trajectories for vegetation types were divided into two levels. Level 1 referred to whether the cause of land use change was induced by the GGP or not. Level 2 was related to the reclassification of the change trajectories with same land use result in 2014. In Level 2, four patterns of change were identified: 1) afforestation involving the conversion of cropland and grassland to forest; 2) agricultural intensification resulting from trajectories that converted an area from dryland to irrigated land; 3) cropland abandonment referring to conversion of land use from agricultural land to grassland; and 4) cropland degradation representing a

NDVI = (Pnir−Pred )(Pnir + Pred )

2.3. Land use change detection and reclassification

(1)

where Pnir and Pred are the values of the spectral channels in the near infrared and red wavelengths of the satellite images, respectively. We used NDVI calculated for the growing season (defined as April to October) to analyze vegetation activity in each year. Finally, the SavitzkyGolay filter method was also used to reconstruct the NDVI data set from 2000 to 2014 so as to reduce the interrupt of cloud or data transmission errors (Jönsson and Eklundh, 2004). 2.2.3. Climate data The climatic datasets from 19 climate stations in and around the study area were obtained from the China Meteorological Data Sharing Service Systems and the quality of all records from these stations was 284

Ecological Indicators 88 (2018) 282–291

F. Wang et al.

Table 1 Land use and cover change in the study area (unit:km2). Land use types

2000

Dryland Irrigated land Grassland Forest Water Built-up land Unused land

2007

Area

Proportion/%

Area

Proportion/%

Area

Proportion/%

7301.52 297.35 10345.81 788.00 132.32 350.00 558.17

36.93 1.50 52.32 3.99 0.67 1.77 2.82

5581.49 529.57 9666.07 2837.29 112.16 440.27 606.33

28.23 2.68 48.88 14.35 0.57 2.23 3.07

3623.04 1361.31 10911.88 2846.02 98.94 643.50 288.49

18.32 6.88 55.19 14.39 0.50 3.25 1.46

2.6. Combined analysis of land cover and NDVI changes

state of decreased productivity as land was converted from irrigated land to dryland.

In order to quantify the effects of land use change on vegetation dynamics, a combined analysis of change trends in land cover change and NDVI residuals was further carried out. Based on the analysis of land use change trajectories and human induced NDVI change, we refined our work into analyzing specific human activities that differed by land use change trajectories. We spatially explored information related to the trajectories of land use change on what was the corresponding NDVI residuals change.

2.4. Trends detection of vegetation coverage Based on a pixel-level analysis, the trends of change in vegetation coverage can be obtained by linear regression, where the function is shown in Eq. (2): n

Slope =

n

n

n × ∑i = 1 i × NDVIi− ∑i = 1 ∑i = 1 NDVIi n×

n n ∑i = 1 i 2−( ∑i = 1

i)2

(2)

3. Results

where slope is the change trend of the slope, n is the years of monitored, and NDVIi is the average mean NDVI value of year i. The significance of inter-annual variability of vegetation coverage can be ascertained based on the correlation of the annual time series sequence and the corresponding NDVI values. An F-test was used to test the significance of trends using Eq. (3):

F=U×

2014

N −2 Q

3.1. Changes in the composition of land use and land cover Table 1 illustrated the area of respective land use and cover patterns of the study region in 2000, 2007, and 2014. As shown below, during the period 2000–2014, the study area was totally dominated by grass and cropland. These covered 90.75%, 79.79%, and 80.39% of the total area in 2000, 2007, and 2014, respectively. Grass and cropland included 7301.52 km2 area of dryland, accounting for 36.93% of the total area in 2000, but only 5581.49 km2 (28.23%) remained in 2007 and this decreased to 3623.04 km2 (18.32%) in 2014. Irrigated land expanded significantly during this time period, covering 1.50%, 2.68%, and 6.88% of the landscape in 2000, 2007, and 2014, respectively. Grassland covered the largest landscape area, occupying about half of the total area; this area declined by 679.75 km2 from 2000 to 2007, but then increased by 1245.81 km2 (6.30%) by 2014. The proportion of forest in this region increased from 3.99% to 14.35% and to 14.39% in the same time periods, respectively. In addition, the respective percentages of the landscape covered by water, urbanized land, and unused land in 2000 were 0.67%, 1.77%, and 2.82%, and these percentage changed to 0.50%, 3.25%, and 1.46% by 2014, respectively. All in all, obvious land use and cover change in this region can be summarized as continuous decrease of dryland and water bodies, and noticeable increase of irrigated land, forest and built-up land.

(3) 15

where U equals to the sum of squares, Q = ∑i = 1 (yi−y î )2 is regression sum of squares, and yi is the NDVI value in year i. y î is the average value of 15-year NDVI values. According to the results, a positive or negative slope indicated an increase or decrease in vegetation coverage, respectively. Therefore, variation trends can be divided into the following three types: significant increase (slope > 0, P ≤ 0.05); significant decrease (slope < 0, P ≤ 0.05) and no significant change (P > 0.05).

2.5. Residual trend method The residual trend method has been widely used to distinguish human induced NDVI changes from those driven by climatic variables (Duo et al., 2016; Evans and Geerken, 2004; Wessels et al., 2012). The residual trend method is based on the observation that a strong relationship exists between vegetation production and climatic variables. After removing the influence of climate, human induced vegetation change was assumed to be the dominant cause of the residual trends. We selected precipitation and temperature as the climatic variables. The method essentially had three steps. First, the partial correlation coefficients of annual mean NDVI with precipitation and temperature were calculated, and significance test at a 95% confidence level was identified. Second, a regression model was then established between NDVI and climate factors at per pixel level of analysis, and the difference between the predicted and real value of NDVI was calculated. Third, the residuals were then regressed over time for each pixel. The temporal trend of the residuals was used to monitor human induced vegetation trends. An increasing or decreasing trend suggested improved or degraded vegetation under the influence of human activities, respectively.

3.2. Land use change detection 3.2.1. Transformation matrix The transition matrices during 2000–2007 and 2007–2014 were shown in Table 2 and Fig. 2(a and b). As observed, between 2000 and 2007, the major land use change processes included the expansion of forest, and, to a lesser extent, the expansion of irrigated agriculture. The increase of forest land occurred as a result of conversions to forest from dryland (19.27%), grassland (5.89%), and unused land (6.49%). In addition, irrigated land expanded at the expense of dryland (5.27%). Within this period, a large area of dryland was developed into irrigated (5.27%) and urbanized land (0.58%). The second time period (2007–2014) saw a process of evolution in agricultural systems. As part of this process, irrigated land increased quickly at the expense of dryland (17.22%) and grassland (0.44%). At the same time, 26.41% of the 285

Ecological Indicators 88 (2018) 282–291

F. Wang et al.

Table 2 Transitions in percentages of the total land area during 2000–2014 (%). Land use types

Dryland

Irrigated land

Grassland

Forest

Water

Built-up land

Unused land

2000–2007 Dry land Irrigated land Grassland Forest Water Built-up land Unused land

74.05 46.09 0.28 0.30 0.08 0.27 0.91

5.27 48.07 0.02 0.00 0.00 0.04 0.02

0.75 5.47 90.35 0.39 16.00 0.59 39.68

19.27 0.07 5.89 99.29 0.50 0.31 6.49

0.00 0.00 0.02 0.00 82.52 0.07 0.02

0.58 0.30 0.46 0.01 0.03 98.72 0.60

0.08 0.00 2.97 0.00 0.87 0.01 52.28

2007–2014 Dry land Irrigated land Grassland Forest Water Built-up land Unused land

62.14 26.41 0.02 0.22 0.09 0.05 1.00

17.22 66.60 0.44 0.13 0.03 0.00 0.19

18.76 5.73 98.13 0.22 12.98 0.72 53.69

0.07 0.00 0.21 99.42 0.11 0.00 0.16

0.00 0.00 0.01 0.00 86.56 0.00 0.10

1.80 1.26 1.02 0.01 0.12 99.24 0.07

0.00 0.00 0.17 0.00 0.12 0.00 44.80

provided us specific temporal sequences of land use change. In all the trajectories, the trajectory with land use conversions occurred in 29.44% of the total area. The 11 major change trajectories, which covered 81.00% of the total land use change, were selected for the following type of analysis (Table 3, Fig. 2c). At the policy-induced Level 1, the total afforested area was 2057.66 km2, accounting for 36.60% of the entire area experiencing land use change. The trajectory defined as 144 represented the largest land use change type, covering 68.18% of the afforested areas. Farmer-induced land use change included 44.40% of the total land use change area. That is, human activities had exerted major effects on the agricultural system after returning farmland to ecological land. In this level, agricultural intensification mainly consisted of five change trajectories where trajectory defined as 112 accounted for more than 72% of all LUCC. Large amounts of dryland were transformed to irrigated land from 2000 to 2007, including 19.71% of the newly irrigated land. Cropland abandonment mainly included three land use change trajectories and covered 42.66% of the farmer-induced LUCC area. The 113 type of trajectory shared the absolute proportion,

irrigated land in 2007 was converted into dryland. Concurrently, another important land use change process that was taking place in this period was the abandonment of agricultural land; 18.76% of the total dryland in 2007 was abandoned land. Moreover, a significant conversion from grassland to forested land also occurred with 0.21% of grassland developed into forest. Overall, from 2000 to 2007, the transition area covered 16.90% of the total area, and the major land use change processes caused by the GGP accounted for 61.54% of the land use change area. During the period of 2007–2014, the transformation area included 14.60% of the total area, of which the largest transformation occurred between cropland and grassland which made up 77.00% of the area experiencing LUCC. The results of the present study confirmed that land cover types in Ulanqab City exhibited continual and complex changes both during and after the end of the implementation of the GGP, particularly for cropland.

3.2.2. Reclassification of the major land use trajectories Change trajectory analysis of land use in different time nodes

Fig. 2. Land use shifts in the period 2000–2014: (a) 2000–2007, (b) 2007–2014 and (c) 2000–2014. Note: 1, Dryland; 2, Irrigated land; 3, Grassland; 4, Forest; 5, Water; 6, Built-up land; 7, Unused land.

286

Ecological Indicators 88 (2018) 282–291

F. Wang et al.

Table 3 Spatial pattern metrics of the major land use change trajectories and reclassification. Level 1 classes

Percentage

Level 2 classes

Percentage to level 1

Land use change trajectories

Area/km2

Percentage to level 2

Policy-induced

36.60%

Afforestation

100.00%

144 344 744 334

1403.17 603.22 35.28 16.00

68.18% 29.33% 1.72% 0.78%

Farmer-induced

44.40%

Agricultural Intensification

49.53%

Cropland abandonment

42.66%

Cropland degradation

7.81%

112 122 212 332 232 113 123 213 121 211 221

891.60 243.83 60.83 20.69 19.91 1027.48 21.97 15.92 113.76 55.85 25.52

72.09% 19.71% 4.92% 1.67% 1.61% 96.44% 2.06% 1.49% 58.30% 28.62% 13.08%

Note: 1, Dryland; 2, Irrigated land; 3, Grassland; 4, Forest.

accounting for 96.41% of the total abandonment area. In addition, cropland degradation resulted in 7.81% of the farmer-induced land use change, which was the negative result of excessive use of ground water.

Table 5 Significant decrease and increase in NDVI changes during 2000–2014.

3.3. Human-induced vegetation changes According to the area of changed partial coefficients range (Table 4), for the entire period of 2000–2014 the variation in the annual mean NDVI was significantly correlated with annual precipitation. The percentage of significantly correlated pixels was 68.11% of the total area with 73.02% of these coefficients ranging from 0.40 to 0.86. In contrast to precipitation, the NDVI values were weakly correlated with annual average temperature for most pixels and more than 95% of the coefficients were not statistically significant (P > .05). Overall, the results indicated that precipitation rate was the major limiting climatic factor that influenced vegetation growth in this arid and semiarid region. Further analysis showed that small variations in temperature were observed as a background that cannot be included in multiple linear regressions for the prediction of residuals; therefore, the relationship between NDVI and precipitation for each pixel was obtained following a linear regression model. In addition, the residuals were obtained based on the regression analysis. Significant positive and negative change trends were observed in NDVI and NDVI residuals for the entire study area (Table 5). Significant change trends in the observed NDVI covered 23.31% of the research area. After removing the influence of climate, significant changes in NDVI residuals that showed the proportion of NDVI changes induced by human activities accounted for 10.47% of the

NDVI

Type

Significant decrease

Significant increase

Not significant

The observed NDVI NDVI residuals

Area/km2 Proportion/% Area/km2 Proportion/%

835.82 4.23 344.95 1.75

3768.49 19.08 1722.55 8.72

15144.10 76.68 17678.49 89.53

entire study area. Generally, pixels showing a significant decrease or increase in vegetation coverage were found over 1.75% and 8.72% of the research area, respectively. This indicated that human activities primarily resulted in an improvement in vegetative conditions. 3.4. Combined analysis of land cover and NDVI Based on the residual analysis, Fig. 3 and Table 6 displayed the calculated change trends in NDVI residuals for each location with a concrete land use trajectory. In unchanged vegetated areas, where areas experienced no land use change, vegetation coverage had increased significantly; this type of area made up 8.39% of the entire region. The most notable areas with improved vegetative conditions were observed in forest and irrigated land, where the vegetation coverage increased significantly by 17.52% and 12.30%, respectively. In contrast, dryland and grassland experienced less improvement; in particular, only 4.62% of dryland observed a significant increase in NDVI. In contrast with unchanged dryland and grassland, the NDVI change

Table 4 The area of changed partial correlation coefficient range between NDVI and climate variables during 2000–2014. Variables

Correlation coefficient range

Significant area/km2

Proportion/%

Non-significant area/km2

Proportion/%

NDVI-Precipitation

<0 0–0.2 0.2–0.4 0.4–0.6 > 0.6 Total

74.68 3.33 55.11 5637.41 7683.44 13453.97

0.38 0.02 0.28 28.54 38.90 68.11

838.44 1287.24 3070.24 1094.45 8.83 6299.20

4.24 6.52 15.54 5.54 0.04 31.89

NDVI-Temperature

< (−0.6) −0.6–(−0.4) −04–(−0.2) −0.2–0 0–0.2 0.2–0.4 > 0.4 Total

0.03 0.12 0.94 73.47 3.68 32.61 1132.72 1243.58

0.00 0.00 0.00 0.37 0.02 0.17 5.73 6.30

83.80 1538.76 5674.70 4866.91 3090.31 2720.50 532.01 18506.99

0.42 7.79 28.73 24.64 15.65 13.77 2.69 93.70

287

Ecological Indicators 88 (2018) 282–291

F. Wang et al.

Fig. 3. Land use change trajectories and the human-induced NDVI change in the study area. Note: The same as Fig. 2.

changes in NDVI in the total area of LUCC, afforestation and agricultural intensification accounted for 39.53% and 33.25% of the pixels with positive changes in NDVI residuals, respectively (Table 7). Significantly, afforestation also covered 12.91% of the pixels with a significant decrease in vegetative conditions; however, this was lower than the percentage (18.99%) that was caused by the other three practices. In particular, cropland abandonment accounted for 10.87%. Generally, human-induced changes in vegetation varied among the land use change trajectories. Land use change caused by afforestation and agricultural intensification were the main drivers that contributed to vegetation improvement, but cropland abandonment and cropland degradation led to vegetation degradation.

trends of the afforested area showed an overall upward tendency and exhibited greater improvement than other areas, which indicated that afforestation in this semiarid area served as an important method of restoring vegetation. Vegetation coverage increased significantly in 10.04% of the afforested area and significantly decreased in 1.08% of the area. Generally, among the four trajectories, the 334 trajectory contributed to 39.38% of the increase in vegetation coverage and the 744 trajectory contributed to 19.52%. In contrast, vegetation coverage improved in 8.27% of cropland conversion areas, but vegetation coverage in 1.57% of the converted areas showed a significant decrease, which had the largest area of decreasing NDVI values in all the afforested area. In non-afforested areas, the ecological effects of areas with dramatic land use change on vegetation should not be neglected, especially for the remaining cropland areas. Specifically, the unsustainable removal of groundwater for agriculture contributed to a major improvement in vegetation coverage in the newly developed irrigated cropland. Overall, vegetation coverage showing positive changes that accounted for 14.07% of these newly developed irrigated areas, of which more than 60.00% of the land use trajectories exhibited a significant increase in vegetation coverage. In contrast, however, cropland abandonment and degraded areas showed an obvious degradation in vegetation coverage; significantly stronger negative changes were observed in these two conversion areas than that were observed in other trajectories. Almost certainly, LUCC has had important effects on vegetation dynamics in the study area; the magnitude of those effects depended heavily on the way in which the conversions occurred. For residuals

4. Discussion 4.1. Effects of human activities on LUCC Large-scale agricultural practices in recent years have had severe negative effects on land cover and have transformed the landscape markedly in many areas in China (Li et al., 2016a; Liang and Yang, 2016; Xin et al., 2008). However, previous studies emphasized that land use change was mainly induced by the GGP. This finding was based on land use data with a resolution of 1000 m for a 10-year period, captured from the Data Center of Chinese Academy of Sciences, or other land use data obtained from the MODIS Land Cover Type product (Li et al., 2017; Mu et al., 2013; Yang et al., 2014), which could not clearly capture specific changes in land use under a regional context. Thus, a 288

Ecological Indicators 88 (2018) 282–291

F. Wang et al.

Table 6 The response in the change of vegetation coverage to land use trajectories during 2000–2014. Areas

Land use trajectories

Significant decrease

Significant increase

Area/km2

Proportion/%

Area/km2

Proportion/%

Unchanged Vegetated area

111 444 333 222 Total

23.63 7.92 75.97 2.86 110.38

0.70 1.02 0.83 2.65 0.82

156.41 135.48 824.08 13.27 1129.25

4.62 17.52 8.96 12.30 8.39

Afforestation

144 344 744 334 Total

22.03 4.17 0.13 0.07 26.40

1.57 0.69 0.35 0.46 1.28

116.03 77.38 6.89 6.34 206.64

8.27 12.83 19.52 39.68 10.04

Agricultural intensification

112 122 212 332 132 Total

4.22 2.06 3.40 0.16 0.06 9.90

0.47 0.85 5.60 0.77 0.28 0.80

115.73 50.75 2.44 2.97 1.91 173.79

13.00 20.83 4.01 14.36 9.59 14.07

Cropland abandonment

113 123 213 Total

17.98 3.26 1.00 22.24

1.75 14.88 6.28 2.09

35.31 0.63 0.02 35.95

3.44 2.87 0.12 3.38

Cropland degradation

121 211 221 Total

1.43 2.89 2.39 6.71

1.26 5.17 9.39 3.44

6.81 0.20 0.18 7.18

5.99% 0.35 0.70 3.68

Note: The same as Table 3. Table 7 Percentage of land use change trajectories with vegetation changes in NDVI residuals. Vegetation changes

Afforestation/%A

Agricultural intensification/%A

Cropland abandonment/%A

Cropland degradation/%A

Significant decrease in NDVI residuals Significant increase in NDVI residuals

12.91 39.53

4.84 33.25

10.87 6.88

3.28 1.37

Note: %A is the percentage in the total land use change area.

large-scale farming in this region. Also, the spatial extent of irrigated land had expanded quickly during the study period; this was promoted by changes in technology that allowed the exploitation of groundwater as well as the construction of reservoirs and aqueducts (Chen et al., 2014b). At the same time, rural areas became depopulated and large amounts of cropland were abandoned; this phenomenon largely appeared to be related to droughty farming conditions, climate change, socioeconomic development, and large-scale rural emigration. Moreover, the conversion from irrigated land to dryland not only reflected the degradation of cultivated land productivity but also indicated the development of a severe environmental emergency caused by the unsustainable exploitation of groundwater. Even more seriously, large areas of previously irrigated land have deteriorated directly to grassland and cannot even be used for dryland cultivation.

lack of spatially-detailed data with a consecutive change history has existed with respect to policy and social development; we saw this as a gap that needed to be filled so we could better understand the effects of human activities. Based on the analysis of land use change in Ulanqab City, both national policies and regional development had various effects on land cover change in the study area. With the implementation of the GGP, retiring and converting cropland and wasteland was anticipated to result in a decrease in the amount of the remaining cropland and grassland while the area of forest land increased. Actually, this land use transformation occupied about 10.42% of the total study area, and this transition mainly resulted in a change from cropland to forest during the period 2000–2007. According to government statistics, about 1513 km2 cropland was converted into forest in this period. This value was only a little larger than that of the area detected in our study (1404.16 km2). The newly developed forested area detected in our study generally agreed with government report for data covering from 2000 to 2007. Moreover, it was difficult to effectively detect a minor transition using remote sensing because of the extremely low survival rate of trees or shrubs in some afforested areas, which may shape the actual afforested area and unavoidably overestimate the effects of afforestation on vegetation restoration. Rapid and extensive changes were observed in the remaining cropland during the period of 2007–2014. Pushed by policies of the central government and an excessive pursuit of commercialized and intensified food production, large amounts of agricultural land have been transferred to large commercial farms in an effort to develop

4.2. Comparison of vegetation restoration based on the trajectory of land use change Land use change is a major factor affecting vegetation dynamics in arid areas (Hasler et al., 2009; Huang et al., 2015); any ecological restoration strategy must be assessed by first understanding the dynamics of land use change so as to understand how restoration policies affect the growth of vegetation (Xu et al., 2010). In Ulanqab City, our results showed that vegetation improved in the unchanged vegetated areas; this may be closely related with the implementation of the GWP. Our field investigation found that the prohibition of grazing was one of the important devices that could be used to promote vegetation restoration. 289

Ecological Indicators 88 (2018) 282–291

F. Wang et al.

degradation. To prevent land degradation, land use intensity must stay at a reasonable level by considering the ability of the land to naturally regenerate if ecological restoration projects are to be sustainable, and some land should be set aside for restoration in areas experiencing cropland degradation. The availability of water strongly limits growth and survival of vegetation in China’s arid and semiarid regions (Donohue et al., 2007). Limited precipitation rates and high temperatures in arid and semiarid regions may cause severe water stress for vegetation resulting in a decrease in vegetative growth and leading to low survival rates for native plants (Cao et al., 2014). Annual rainfall of less than 400 mm in some areas may result in the shriveling of cultivated sea buckthorn and alfalfa. In our study, vegetation growth was closely related to precipitation rates in the study area where the average annual precipitation is less than 400 mm. Thus, an excessive emphasis on tree and shrub planting in environmentally fragile areas and a failure to consider climate, soil, hydrology, and landscape should be questioned and will need additional attention. Overall, restoration is a comprehensive issue that is affected by many mechanisms and these vegetation restoration mechanisms call for the careful consideration of field conditions if project management is to be successful.

Notably, pixels with a significant decrease in NDVI residuals still covered a large area of both grassland and dryland. This occurs because grassland and dryland are the main dominant vegetation types and are commonly used for grazing, animal production, and cultivation. Therefore, improved management practices, such as forbidding grazing and using fencing to close grassland and support afforestation, are still important and urgently needed. In the afforested areas, the results of the present study indicate that afforestation can serve as an important method for restoring vegetation, but the vegetation coverage of an afforested area that is converted from dryland to forest can become severely deteriorated. Our fieldwork actually indicated that while shrubs were planted and vegetation was established on the set-aside sites, a striking lack of follow-up tending has existed. These afforested sites also need attention to control competition from undesirable species as well as supplanting and thinning. In addition, areas of degraded vegetation were mainly concentrated in the Houshan region; this degradation was largely associated with the relatively undeveloped socioeconomic conditions, massive rural emigration, and the abandonment of settlements. Therefore, recognizing that restoring degraded ecosystems is a long and arduous process. This process requires sustained effort including land maintenance and management that is very important, and these areas will require adequate attention. Apart from afforestation, our results also suggested that irrigation was responsible for the trend of increased vegetation coverage for both agricultural land and other vegetated areas. However, it was notable that because of an unsustainable use of both land and groundwater, many areas have degenerated from irrigated land to dryland or were abandoned, resulting in a significant decrease in vegetation coverage and degradation of the entire ecosystem. In addition, field investigations showed that a large amount of cropland had been abandoned and grass had thinned near the scale farming areas. Therefore, without restoration management, the severity of land degradation increased in these areas, providing a serious warning sign related to the long-term sustainability environment conditions and of the local communities.

5. Conclusions Based on Landsat images and MODIS NDVI remote sensing data, the present study quantitatively analyzed environmental changes in Ulanqab City, Inner Mongolia, China using the combined analysis of land use and cover change along with human induced NDVI change over the past 15 years since the inception of the GGP. From 2000 to 2014, important land use changes have occurred in the study area involving the GGP, agricultural intensification, as well as cropland abandonment and degradation. The GGP resulted in 36.60% of the total land use change, but the other three practices analyzed here contributed 44.40%. The study demonstrated that human activities significantly influenced vegetation coverage in 10.47% of the study area, and pixels with significant increasing residual trends covered 8.72% of the research area. Afforestation and agricultural intensification primarily improved vegetative conditions while cropland abandonment and cropland degradation were the primary causes of land degradation. Generally, for NDVI residuals in the total area experiencing land use change, the GGP caused 39.53% of the improved vegetative conditions while other land use changes caused 41.50% of the improved land changes; in particular, agricultural intensification contributed 33.25% of the improved areas. In contrast, vegetation degradation occurred in 1.75% of the research area, of which 12.91% was caused by afforestation, which was lower than that caused by the other three land management practices combined. Although afforestation and agricultural intensification produced a favorable effect on vegetation recovery during this period, vegetation coverage in some afforested areas showed significant decreasing trends and the intensification process had caused severe land degradation at the same time. Considering the important role of human activities in restoring vegetation coverage, our results clearly show that ecological restoration efforts were experiencing increasing pressure from human activity that threatens the sustainability of human–natural systems in Ulanqab City. This paper argues that agricultural production should fully consider the effects of agricultural intensification to avoid the unsustainable exploitation of groundwater resources; local governments should fully understand and mandate the sustainable evolution of dryland agriculture systems and rural society. Seeking ways to use regional land resources sustainably will be crucial to the development regional land use schemes. In particular, managers of large-scale afforestation should be more concerned about supervision of afforestation efforts and the maintenance of such areas in the future, thereby improving environmental conservation efforts and allowing them to succeed on both an ecological and a human scale.

4.3. Challenges and implications for ecological restoration projects As many studies have shown, the massive “greening” effort attempted by the GGP and similar projects has been less effective than expected in some geographic areas (Cao et al., 2011). For example, tree survival during afforestation across the Three North Shelter Forest System Projects during this period was only 15% (Cao, 2008). Our study also revealed the condition of vegetation overall experienced no apparent improvement. Afforestation, which may have been largely responsible for the recovery of vegetation during the last decade, had a limited effect on vegetation restoration as a whole; this result was closely related to human activities and climate change. Obviously, the complicated LUCCs in Ulanqab City have exerted major impacts on the regional environment. Scientists tended to acknowledge the positive significant effects of afforestation on vegetation recovery, but human beings should take more responsibility for the sustainability of other land uses. As analyzed in the present study, although agricultural intensification contributed greatly to an increase in vegetation coverage, groundwater throughout this region was being exploited unsustainably for agriculture, resulting in a steady decline in the groundwater table. Under these conditions, ecological problems were generated such as serious water stress for other plants; grass and other vegetation can no longer depend on near-surface soil water and may be unable to survive under the changed conditions with a lowered water table (Cao et al., 2011), thereby impairing the ecological functions of afforestation. In addition, cropland abandonment and degradation has led to serious vegetation degradation. Therefore, our study suggests that continuous treatments of agricultural land use will be necessary to produce sustainable and permanent landscape scale improvement, especially for areas with high agricultural land use intensity and cropland 290

Ecological Indicators 88 (2018) 282–291

F. Wang et al.

Acknowledgments

Prod. 142, 697–715. Li, Z., Wu, W., Liu, X., Fath, B.D., Sun, H., Liu, X., Xiao, X., Cao, J., 2016b. Land use/cover change and regional climate change in an arid grassland ecosystem of Inner Mongolia, China. Ecol. Model. 353, 86–94. Liang, P., Yang, X.P., 2016. Landscape spatial patterns in the Maowusu (Mu Us) Sandy Land, northern China and their impact factors. Catena 145, 321–333. Liu, J.Y., Tian, H.Q., Liu, M.L., Zhuang, D.F., Melillo, J.M., Zhang, Z.X., 2005. China's changing landscape during the 1990s: large-scale land transformations estimated with satellite data. Geophys. Res. Lett. 32. Ma, H., Lv, Y., Li, H.X., 2013. Complexity of ecological restoration in China. Ecol. Eng. 52, 75–78. Mu, S., Zhou, S., Chen, Y., Li, J., Ju, W., Odeh, I.O.A., 2013. Assessing the impact of restoration-induced land conversion and management alternatives on net primary productivity in Inner Mongolian grassland, China. Glob. Planet. Change 108, 29–41. Roy, D.P., Wulder, M.A., Loveland, T.R., Woodcock, C.E., Allen, R.G., Anderson, M.C., Helder, D., Irons, J.R., Johnson, D.M., Kennedy, R., Scambos, T., Schaaf, C.B., Schott, J.R., Sheng, Y., Vermote, E.F., Belward, A.S., Bindschadler, R., Cohen, W.B., Gao, F., Hipple, J.D., Hostert, P., Huntington, J., Justice, C.O., Kilic, A., Kovalskyy, V., Lee, Z.P., Lymbumer, L., Masek, J.G., McCorkel, J., Shuai, Y., Trezza, R., Vogelmann, J., Wynne, R.H., Zhu, Z., 2014. Landsat-8: science and product vision for terrestrial global change research. Remote Sens. Environ. 145, 154–172. Sun, W., Song, X., Mu, X., Gao, P., Wang, F., Zhao, G., 2015a. Spatiotemporal vegetation cover variations associated with climate change and ecological restoration in the Loess Plateau. Agric. For. Entomol. 209–210, 87–99. Sun, Y.L., Yang, Y.L., Zhang, L., Wang, Z.L., 2015b. The relative roles of climate variations and human activities in vegetation change in North China. Phys. Chem. Earth 87–88, 67–78. Tian, H., Cao, C., Chen, W., Bao, S., Yang, B., Myneni, R.B., 2015. Response of vegetation activity dynamic to climatic change and ecological restoration programs in Inner Mongolia from 2000 to 2012. Ecol. Eng. 82, 276–289. Tong, X.W., Wang, K.L., Yue, Y.M., Brandt, M., Liu, B., Zhang, C.H., Liao, C.J., Fensholt, R., 2017. Quantifying the effectiveness of ecological restoration projects on long-term vegetation dynamics in the karst regions of Southwest China. Int. J. Appl. Earth Obs. Geoinf. 54, 105–113. Wang, D.C., Gong, J.H., Chen, L.D., Zhang, L.H., Song, Y.Q., Yue, Y.J., 2012. Spatiotemporal pattern analysis of land use/cover change trajectories in Xihe watershed. Int. J. Appl. Earth Obs. Geoinf. 14, 12–21. Wang, J.Y., Liu, Y.S., Liu, Z.G., 2013. Spatio-temporal patterns of cropland conversion in response to the “grain for green project” in China's Loess Hilly Region of Yanchuan County. Remote Sens. 5, 5642–5661. Wessels, K.J., van den Bergh, F., Scholes, R.J., 2012. Limits to detectability of land degradation by trend analysis of vegetation index data. Remote Sens. Environ. 125, 10–22. Wright, C.K., de Beurs, K.M., Henebry, G.M., 2012. Combined analysis of land cover change and NDVI trends in the Northern Eurasian grain belt. Front. Earth Sci. 6, 177–187. Xin, Z.B., Xu, J.X., Zheng, W., 2008. Spatiotemporal variations of vegetation cover on the Chinese Loess Plateau (1981–2006): impacts of climate changes and human activities. Sci. China Ser. D 51, 67–78. Xu, D.Y., Kang, X.W., Zhuang, D.F., Pan, J.J., 2010. Multi-scale quantitative assessment of the relative roles of climate change and human activities in desertification – a case study of the Ordos Plateau, China. J. Arid Environ. 74, 498–507. Yang, H., Mu, S., Li, J., 2014. Effects of ecological restoration projects on land use and land cover change and its influences on territorial NPP in Xinjiang, China. Catena 115, 85–95. Yin, R.S., Yin, G.P., Li, L.Y., 2010. Assessing China's ecological restoration programs: what's been done and what remains to be done? Environ. Manage. 45, 442–453. Zewdie, W., Csaplovics, E., 2016. Identifying categorical land use transition and land degradation in Northwestern Drylands of Ethiopia. Remote Sens. 8, 408. Zhang, G.L., Dong, J.W., Xiao, X.M., Hu, Z.M., Sheldon, S., 2012. Effectiveness of ecological restoration projects in Horqin Sandy Land, China based on SPOT-VGT NDVI data. Ecol. Eng. 38, 20–29. Zhang, J., Niu, J.M., Bao, T.L.G., Buyantuyev, A., Zhang, Q., Dong, J.J., Zhang, X.F., 2014a. Human induced dryland degradation in Ordos Plateau, China, revealed by multilevel statistical modeling of normalized difference vegetation index and rainfall time-series. J. Arid. Land. 6, 219–229. Zhang, J.L., Pham, T.T.H., Kalacska, M., Turner, S., 2014b. Using landsat thematic mapper records to map land cover change and the impacts of reforestation programmes in the borderlands of southeast Yunnan, China: 1990–2010. Int. J. Appl. Earth Obs. Geoinf. 31, 25–36. Zhou, D., Zhao, S., Zhu, C., 2012. The grain for green project induced land cover change in the loess plateau: a case study with Ansai County, Shanxi Province, China. Ecol. Indic. 23, 88–94. Zika, M., Erb, K.H., 2009. The global loss of net primary production resulting from human-induced soil degradation in drylands. Ecol. Econ. 69, 310–318.

This work was supported by the National Natural Science Foundation of China (Grant No. 41371232 and 41271110). The authors are grateful to editors and three anonymous reviewers for their helpful comments and suggestions. References Duo, A., Zhao, W., Qu, X., Jing, R., Xiong, K., 2016. Spatio-temporal variation of vegetation coverage and its response to climate change in North China plain in the last 33 years. Int. J. Appl. Earth Obs. Geoinf. 53, 103–117. Cao, S., Chen, L., Shankman, D., Wang, C., Wang, X., Zhang, H., 2011. Excessive reliance on afforestation in China's arid and semi-arid regions: lessons in ecological restoration. Earth-Sci. Rev. 104, 240–245. Cao, S., Ma, H., Yuan, W., Wang, X., 2014. Interaction of ecological and social factors affects vegetation recovery in China. Biol. Conserv. 180, 270–277. Cao, S.X., 2008. Why large-scale afforestation efforts in China have failed to solve the desertification problem. Environ. Sci. Technol. 42, 1826–1831. Chen, B., Zhang, X., Tao, J., Wu, J., Wang, J., Shi, P., Zhang, Y., Yu, C., 2014a. The impact of climate change and anthropogenic activities on alpine grassland over the QinghaiTibet Plateau. Agric. For. Entomol. 189–190, 11–18. Chen, L., Huang, Z., Gong, J., Fu, B., Huang, Y., 2007. The effect of land cover/vegetation on soil water dynamic in the hilly area of the loess plateau, China. Catena 70, 200–208. Chen, R., Ye, C., Cai, Y., Xing, X., Chen, Q., 2014b. The impact of rural out-migration on land use transition in China: past, present and trend. Land Use Policy 40, 101–110. Chen, Y., Li, X., Su, W., Li, Y., 2008. Simulating the optimal land-use pattern in the farming-pastoral transitional zone of Northern China. Comput. Environ. Urban Syst. 32, 407–414. Donohue, R.J., Roderick, M.L., McVicar, T.R., 2007. On the importance of including vegetation dynamics in Budyko’s hydrological model. Hydrol. Earth Syst. Sci. 11, 983–995. Du, Y., Teillet, P.M., Cihlar, J., 2002. Radiometric normalization of multitemporal highresolution satellite images with quality control for land cover change detection. Remote Sens. Environ. 82, 123–134. Evans, J., Geerken, R., 2004. Discrimination between climate and human-induced dryland degradation. J. Arid Environ. 57, 535–554. Foley, J.A., DeFries, R., Asner, G.P., Barford, C., Bonan, G., Carpenter, S.R., Chapin, F.S., Coe, M.T., Daily, G.C., Gibbs, H.K., Helkowski, J.H., Holloway, T., Howard, E.A., Kucharik, C.J., Monfreda, C., Patz, J.A., Prentice, I.C., Ramankutty, N., Snyder, P.K., 2005. Global consequences of land use. Science 309, 570–574. Ge, X., Li, Y., Luloff, A.E., Dong, K., Xiao, J., 2015. Effect of agricultural economic growth on sandy desertification in Horqin Sandy Land. Ecol. Econ. 119, 53–63. Hasler, N., Werth, D., Avissar, R., 2009. Effects of tropical deforestation on global hydroclimate: a multimodel ensemble analysis. J. Clim. 22, 1124–1141. Higginbottom, T., Symeonakis, E., 2014. Assessing land degradation and desertification using vegetation index data: current frameworks and future directions. Remote Sens. 6, 9552–9575. Hong, Y., Nix, H.A., Hutchinson, M.F., Booth, T.H., 2005. Spatial interpolation of monthly mean climate data for China. Int. J. Climatol. 25, 1369–1379. Huang, C., Zhang, M., Zou, J., Zhu, A.X., Chen, X., Mi, Y., Wang, Y., Yang, H., Li, Y., 2015. Changes in land use, climate and the environment during a period of rapid economic development in Jiangsu Province, China. Sci. Total Environ. 536, 173–181. Huete, A., Didan, K., Miura, T., Rodriguez, E.P., Gao, X., Ferreira, L.G., 2002. Overview of the radiometric and biophysical performance of the MODIS vegetation indices. Remote Sens. Environ. 83, 195–213. Jönsson, P., Eklundh, L., 2004. TIMESAT—a program for analyzing time-series of satellite sensor data. Comput. Geotech. 30, 833–845. Landmann, T., Dubovyk, O., 2014. Spatial analysis of human-induced vegetation productivity decline over eastern Africa using a decade (2001–2011) of medium resolution MODIS time-series data. Int. J. Appl. Earth Obs. Geoinf. 33, 76–82. Le Houerou, H.N., 2000. Restoration and rehabilitation of arid and semiarid Mediterranean ecosystems in North Africa and west Asia: a review. Arid. Soil. Res Rehab. 14, 3–14. Li, S., Liang, W., Fu, B.J., Lu, Y.H., Fu, S.Y., Wang, S., Su, H.M., 2016a. Vegetation changes in recent large-scale ecological restoration projects and subsequent impact on water resources in China's Loess Plateau. Sci. Total Environ. 569, 1032–1039. Li, S.S., Yan, J.P., Liu, X.Y., Wan, J., 2013. Response of vegetation restoration to climate change and human activities in Shaanxi-Gansu-Ningxia Region. J. Geogr. Sci. 23, 98–112. Li, Y.R., Cao, Z., Long, H.L., Liu, Y.S., Li, W.J., 2017. Dynamic analysis of ecological environment combined with land cover and NDVI changes and implications for sustainable urban-rural development: the case of Mu Us Sandy Land, China. J. Clean.

291