Introduction

Grasslands are great ecosystems around the world and occupy 22.10% of the global land area, excluding areas of Greenland and Antarctica (Suttie et al. 2005; White et al. 2000). In temperate grasslands, only 4.6% of native temperate grasslands are conserved globally within protected areas (Carbutt et al. 2017), which is lower than the percentage of those in all other ecosystems (Hoekstra et al. 2005). A large number of grasslands have been converted to other land covers, such as crops, mixed farming, artificial pastures and desert, due to climate change and increased human activity (Suttie et al. 2005). In addition, grassland reclamation has gradually become one of the greatest threats to grassland degradation (Stephens et al. 2008).

Grassland reclamation occurred prior to the 1950s in temperate grasslands (Assessment 2005). At present, grassland reclamation mainly occurs in the tropics, with many tropical grasslands and savannas undergoing change. The expansion of grassland reclamation in North America (Wright and Wimberly 2013), South America (Gavier-Pizarro et al. 2012), Asia (Qi et al. 2012) and Southern Africa (Maeda et al. 2010) creates new pressures on global grassland ecosystems.

Crop fields that are transformed from grassland reclamation are generally of larger dimension and exceed 1.00 km2, a large proportion of which is fallow during the windy spring season (Hoffmann et al. 2011). Bare and ploughed soils are more susceptible to wind erosion than grasslands (Reiche et al. 2015). For example, 75% of the total dust emissions originated from cropland (5% of the area) and only 25% of dust emissions originated from grassland (95% of the area) in the Xilingol steppe in 2006 based on Hoffmann et al.’s (2011) estimation (Reiche et al. 2015). Therefore, croplands are considered to contribute extraordinarily to total dust emissions and soil degradation in temperate grasslands (Hoffmann et al. 2011).

Grasslands are one of the most dominant land use types in China, covering approximately 2.15 × 106 km2 and accounting for 22.39% of the total land area (Bureau, 2016). The organized reclamation of grassland can be traced back to the Qin and Han Dynasties (221 BC–220 AD) in Northern China (Liu et al. 2003, 2010). Since the 1980s, the intensity of grassland reclamation has reached a peak with the increase in population and the improvement in production technology in China. In addition, cropland increased by 2.83 × 104 km2, with reclaimed cropland concentrated mainly in the farming–forestry ecotones of Northern China from the late 1980s to 2000 (Liu et al. 2014). Various natural disasters began to emerge in this period, such as soil wind erosion, dust storms, desertification and droughts. Consequently, the Grain-to-Green Program (GTGP) was implemented with the aim of reducing wind erosion by restoring grassland and forest (Ouyang et al. 2016), and cropland decreased by 1.02 × 104 km2 from 2000 to 2010 in China (Liu et al. 2014). However, over the last few years (since 2010), grasslands (accounting for 64.00%) and unused lands (accounting for 21.90%) have been reclaimed, causing the cropland area to rapidly increase in Northern China (Ning et al. 2018). In recent decades, grassland reclamation has become one of the main types of land use change in the temperate grasslands of Northern China.

Similar fluctuations from grassland to cropland in Northern China also occurred from 1965 to 2010 in the Mu Us Desert (MUD) (Li et al. 2017). However, we have discovered through field investigations over the past 5 years in the MUD that grassland reclamation has increased dramatically. Such reclamation may become the greatest threat to the steppe area of the MUD. However, the areal, pattern and potential effects of grassland reclamation on the MUD in the past 5 years are unclear.

Remote sensing is an effective and convenient tool for monitoring and assessing land use and land cover across a large region. Effective and real-time satellite images can be easily acquired. In particular, multispectral satellite imagery, such as that provided by Landsat series images (with a re-visit time of 16 days), is a precious resource for the timely monitoring of surface processes (Wang et al. 2012). In the present study, we use Landsat TM/OLI data from 2 years (2010 and 2015) covering the entire MUD to detect and evaluate the latest trend in the conversion from grassland to cropland. The objectives of this study are to quantify the latest area of grassland reclamation and analyse the possible ecological hazards in the MUD.

Method and materials

Study area

The MUD covers approximately 48,288 km2 (Li et al. 2017) and is located in a transitional zone between typical steppe and desert steppe regions in Northern China (Fig. 1a). Elevations vary from 902 to 1684 m above mean sea level (Yan et al. 2015) (Fig. 1b). Three types of landforms dominate the MUD: hard hills caused by erosion and the ageing of bedrock, sandy uplands consisting of sediments accumulated during the Quaternary period, and lower wetlands due to cuts in Quaternary sediments by rivers and streams (Wu and Ci 2002). There is an abundance of surface water and groundwater in the MUD. Most rivers eventually flow into the Yellow River or sodic lakes in sandy land. The groundwater table is shallow, approximately 1–3 m below the ground, especially in the inter-dune regions (Karnieli et al. 2014; Liang and Yang 2016). The annual mean temperature ranges from 6.0 to 8.5 °C, and the mean annual precipitation is approximately 250–400 mm, 60–80% of which falls in summer (Li et al. 2015).

Fig. 1
figure 1

Overview of the study area. a Location, sample sites, and land use (2015) of Mu Us Desert. b SRTM DEM map of the study area, original data source from Consortium for Spatial Information (CGIAR-CSI) (http://srtm.csi.cgiar.org/srtmdata/)

Full size image

Grassland and desert are the main land use types in the study area. Historically, grassland reclamation has been known as the “fertile land of million hectares”, with advanced farming and grazing as early as the Qin and Han Dynasties (Liu et al. 2003). The policy “provide for people with other lands” caused the conversion of naturally vegetated land to cropland in the mid-Qing Dynasty (1644–1911 AD) (Liu et al. 2003). Cropland continued to encroach on grassland until the GTGP was implemented in 1999 (Li et al. 2017). However, the phenomenon of grassland reclamation has been found through field investigations by the author in recent years to have increased significantly after the second stage of the GTGP.

Data and methods

TM images for the 2010 data and OLI images for the 2015 data (nominal resolution of 30 × 30 m) were acquired from https://glovis.usgs.gov/. All images were obtained during the region’s growing season. The images were georeferenced and orthorectified by ground control points (GCPs) derived from a 1:100,000 topographic map developed by the Chinese Mapping Agency in the early 1980s. The mean location error after georectification was less than 1 pixel.

An object-based approach was chosen to interpret grassland and farmland information due to the unique feature of the object-based approach that image objects—not individual pixels—become the basic units of analysis, as they represent meaningful geographic entities or phenomena at multiple scales (Blaschke 2010; Chen et al. 2018).

The operation process is as follows (Fig. 2). After a trial-and-error method, the meaningful and homogeneous objects in the OLI images for 2015 are segmented using a multiresolution segmentation method, and the parameters of scale, shape, and compactness are 50, 0.1 and 0.5, respectively (Jia et al. 2014), based on the eCognition Developer 8.64 software. Then, using the algorithm “assign class”, the cropland is extracted by the ratio vegetation index (RVI) (Jordan 1969), which is set to be greater than 3.0. There are two types of classification errors in the process of classification. One is that a small percentage of high-cover grassland is misclassified as cropland. The other is that newly cultivated land is not classified as cropland because its RVI value is similar to that of bare ground. Therefore, the tool Manual Editing Toolbar was used to reclassify the error classification according to visual interpretation. The cropland information in 2015 is fully extracted. Subsequently, the Landsat TM image from 2010 and the farmland information from 2015 are simultaneously loaded into the eCognition Developer 8.64, and the object information is again assigned by the tool Manual Editing Toolbar. Finally, the verification sample data obtained from the field survey and Google Earth are imported into eCognition. The classification results are assessed through the available Google Earth image sampling points and field survey data. A total of 120 cropland verification points are accurately obtained using a HOLUX GPS M241 in April and June 2017, on which 118 patches are correctly classified. The classification accuracy is 98.33%.

Fig. 2
figure 2

Scheme of grassland reclamation information extraction using the object-based approach

Full size image

Results and discussion

Spatial distribution of grassland reclamation

Table 1 and Fig. 3 present the profile and geographical patterns of grassland reclamation in the MUD, showing that the trend of conversion from cropland to grassland was reversed dramatically after the end of the second stage of the GTGP. The area of grassland reclamation increased by 411.29 km2 from 2010 to 2015, accounting for 9.78% of the total cropland area in 2015. The majority of grassland reclamation occurred in the southeastern and southwestern parts of the MUD, and the area of grassland reclamation was relatively rare northwest of the MUD because the moving sand dunes of the MUD were mainly distributed in the northwest, which was unsuitable for tillage (Fig. 1). According to the administrative division (Fig. 1a), the MUD is located at the border of Shaanxi Province, Ningxia Hui Autonomous Region, and the Inner Mongolia Autonomous Region and includes 12 counties. The majority of grassland reclamation was distributed within Inner Mongolia and Shanxi, accounting for 211.37 km2 and 196.20 km2, respectively. When analysing at the county scale, grassland reclamation was mainly distributed within the Yuyang District, Otog Qian Banner, and Uxin Banner, accounting for 118.27 km2, 89.00 km2, and 82.96 km2, respectively.

Table 1 The area of grassland reclamation in the study region
Full size table
Fig. 3
figure 3

The grassland reclamation map of the Mu Us Desert from 2010 to 2015. The green circle marks the position of Fig. 5

Full size image

Two patterns of grassland reclamation

According to the field investigation in April and June 2017, we determined that there were two patterns of reclaimed grassland (Fig. 4). One pattern is that herdsmen cultivated pasture or feed corn on their own pasture. The characteristics of this method are small area, self-sufficient, and rough field management and irrigation through pumping groundwater by means of flooding or sprinklers. The management of leaving standing crop residues is carried out after gaining the corn or artificial pastures. This fractured patch of reclaimed land is mainly found in the Inner Mongolia Autonomous Region of the MUD (Fig. 3). The other pattern is that grassland reclamation projects are implemented by enterprises under the direction of local government. The features are large area, maximum interest, normative management (with little residue or field ploughing retained after harvesting potato or corn), and irrigation through pumping groundwater by means of sprinklers. The large patch of reclaimed land mainly appears in the Shanxi Province region of the MUD (Fig. 3). For example, Fig. 5 shows the largest grassland reclamation patches in the MUD, which were implemented by the Shaanxi Provincial Land Engineering Construction Group Co., Ltd., with the Land and Resources in Shaanxi Province (Fig. 5a). Reclamation of the region began in 2010; the area used to be a sand sagebrush (Artemisia filifolia) shrub ecosystem (Fig. 5b). By the end of 2015, the area of grassland reclamation was 16.24 km2 (Fig. 4c, d). Obviously, the second pattern can produce more economic benefits. However, some study results show that wind erosion over ploughed fields was significantly higher than that over stubble corn fields (Sharratt and Collins 2018). Therefore, the second pattern may cause more serious wind erosion than the first pattern.

Fig. 4
figure 4

The two method patterns of grassland reclamation in the MUD. a Grassland was reclaimed by herdsmen, and the crop was alfalfa (Medicago sativa L.), which was used to make silage; b the grassland reclamation project was implemented by enterprises, and the crop was potato (Solanum tuberosum L.), which was used to gain income. The blue boxes indicate two different irrigation methods. a A permanent pipeline sprinkler irrigation system; b a sprinkler system with a travelling pipe. The red lines mark the boundaries between the grassland and cropland

Full size image
Fig. 5
figure 5

The grassland reclamation region in the study. a Area of grassland reclamation in Mengjiawan Town, Yulin City, Shanxi Province (longitude: 109°38′15″; latitude: 109°38′15″; the green circle in Fig. 1.). b Image from Google Earth dated September 25, 2009. c Image from Google Earth dated January 31, 2016. d Photograph taken by the DJI Phantom 3 UAV on June 9, 2017

Full size image

Negative effects that may be caused by grassland reclamation

Grassland reclamation destroys the topsoil structure and native vegetation, and a large proportion of areas are fallow during the windy spring season (Hoffmann et al. 2011). Consequently, bare and ploughed soils are more susceptible to wind erosion than grassland. Many simulative wind tunnel experiments of cultivated and non-cultivated soil have shown that wind erosion may be accelerated 2–15 times by cultivation in northern China during the windy spring season (Cai and Wang 1999; Dong et al. 1987; Hoffmann et al. 2011; Zhang and Dong 2014). Zhang and Dong (2014) analysed the total sediment transport for five surface types, including cultivated land planted with wheat or oilseed crops, and fixed dunes, with approximately 35% vegetation cover of sand sagebrush in the Yanchi County region of the MUD (Fig. 6). The amount of cultivated land was five times that of the fixed dunes during the entire year. In addition, Du et al. (2018) built a wind erosion model with a high spatio-temporal resolution that was developed and employed to identify the sand/dust emissions in areas within northern China from 2001 to 2014. Simulations show that the proportions of dust emissions were related to the fractions of areas of different land cover types and that the total dust emission fraction of cropland first displayed a decreasing trend from 2000 to 2008, followed by a rapidly increasing trend from 2008 to 2013 (Du et al. 2018) (Fig. 7). The results also indicated that an increase in cropland leads to an increase in sand/dust emissions in the MUD. Ultimately, wind erosion may decrease agricultural productivity (Buerkert and Lamers, 1999) and trigger desertification (Zhao et al. 2006) with the selective removal of fine particles, nutrients and organic matter, and the accumulation of coarse mineral grains (Hoffmann et al. 2011; Houyou et al. 2016; Li et al. 2005; Wiesmeier et al. 2015; Zhang and Dong 2014). Previous studies (Su et al. 2004; Zhang et al. 2013; Zhang and Dong, 2014) have suggested large amounts of silt and clay (diameter < 0.05 mm) losses by wind erosion, and the remaining particles tend to be coarse, which is similar to those of shifted sand in the near-surface layers in the grassland reclamation areas. Meanwhile, soil organic carbon (SOC) and nitrogen (N) were lost with the removal of silt and clay under the cultivation of sandy grasslands and subsequent short-term tillage practices (Su et al. 2004; Zhang et al. 2013). According to the studies of Wu et al. (2003) and Xie et al. (2007), Wang et al. (2011) determined that the conversion of freely grazed grassland to cropland led to an approximate 36% SOC loss in surface soils (0–20 cm) in grassland in Northern China. Su et al. (2004) and He et al. (2012) studied the N storage of reclaimed lands at different times in the Horqin Sandy Land and at the Inner Mongolia Grassland Ecosystem Research Station (43°33′N, 116°40′E), respectively. These results showed that cultivation significantly reduced N storage in the 0–15 cm ploughed layer.

Fig. 6
figure 6

Total sediment transport in different months above cultivated land and fixed dunes in 2006. Data are from Zhang and Dong (2014)

Full size image
Fig. 7
figure 7

Fractions of dust emissions from cropland between 2001 and 2013. Data are from Du et al. (2018)

Full size image

Strong motivations and effective measures to prevent grassland erosion

Strategies to reduce wind erosion on reclaimed grassland in arid and semi-arid grasslands have been carried out (Hoffmann et al. 2011), such as leaving standing crop residues (Hagen 1996; Wang et al. 2002), reducing tillage intensity (Mendez and Buschiazzo 2010), using several alternative cultivation systems (Li et al. 2002), and applying ridge tillage (Liu et al. 2006). However, as the authors were investigating grassland reclamation in the MUD, the first two strategies were unsuccessful in the study area because the soil surface layers had to be disturbed when digging the potatoes, which destroyed the non-tillage systems and crop residues. Generally, it was learnt through interviews with local farmers that two alternative rotations, mainly using potatoes and maize, were established to reduce wind erosion in the study area. Ridge tillage is worth using as an alternative practice for reducing wind erosion during the fallow period and can decrease wind erosion by 20–60% (Liu et al. 2006). However, regardless of the kind of protective measure taken, reclaimed grasslands cannot be restored to their original condition.

Superficially, grassland reclamation is related to national policy and the prompting of short-term economic benefits. The main reason is that the government regards large sandy lands and grasslands as reserve resources for cultivated land, and there are even reports that the MUD would be transformed into a new base for regional commercial grain production. The sandy soil area, with less than 400 mm of precipitation, is not suitable for large-scale grassland reclamation north of the Great Wall because of fragile soil conditions and the strong wind environment. Therefore, the study area should be classified as a high-risk ecological region for dry farming, and tillage should be prohibited (Hoffmann et al. 2011).

Conclusions

The conversion from grassland to cropland was effectively curbed with the implementation of the Grain-to-Green Program in northern China from 1999 to 2010, and soil properties were significantly improved in a short time (Zhang et al. 2013). However, the phenomenon of grassland reclamation has significantly increased since 2010. The latest grassland reclamation trend is worthy of close attention in the Mu Us Desert and even the whole grassland region of northern China.

In this paper, we use an object-based approach to detect and evaluate the latest trend in the conversion from grassland to cropland. The results show that the area of grassland reclamation increased by 411.29 km2 from 2010 to 2015 and was mainly distributed in the southeastern and southwestern parts of the Mu Us Desert, accounting for 9.78% of the total cropland area in 2015. Interviews with local farmers revealed two ways to reclaim grassland, including herdsmen cultivating grasslands on their own pastures and grassland reclamation projects implemented by enterprises under the direction of local governments. Exposed cultivated land is more susceptible to wind erosion, which eventually leads to decreased agricultural productivity and desertification with the selective removal of fine particles, nutrients, and organic matter and the accumulation of coarse mineral grains. Appropriate tillage strategies should be selected according to local water and soil conditions and crop types.