The Mu Us dune field (Figure 1) has an area of ~40,000 km², accounting for ~3.6% of the total desert zone in China. The area has a prevailing northwest wind direction for most of the year. The region falls within a typical temperate continental semiarid monsoon climate zone. The climate shows distinct changes from the northwest to the southeast, transitioning from arid and semi-arid to semi-humid climate conditions. Correspondingly, mean annual temperature, annual evaporation, and annual precipitation of the study region are 6.0-8.5℃, 1800-2500 mm, and 250-440 mm, respectively, with 60%-70% of precipitation falling during June-August. The Mu Us dune field receives higher rainfall compared with other dune field areas in Northern China, and surface water and groundwater resources are relatively plentiful. The natural vegetation of the study area mainly consists of Artemisia ordosica Krasch., Tamarix chinensis Lour., and Hippophae rhamnoides Linn. The study area shows a zonation of vegetation, with desert steppe in the west, typical steppe in the central and southeast parts, and forest steppe in the southeastern marginal area. The northwest part of the study area falls within Erdos City (Otog Banner, Uxin Banner, and Otog Front Banner, etc.) in the Inner Mongolia Autonomous Region in which traditional grazing is the main form of economic activity. Crop agriculture dominates in the southeast, and in particular in Yulin City in the Shaanxi Province (including Yuyang District, Hengshan District, Jingbian County, etc.) (Figure 1b).

The remote sensing images for 2018 showed that few CPI farmland units were present in the northwest of the study area, which can be attributed to this area being affected by erosion due to a higher altitude and less precipitation (Liu et al., 2017). Therefore, the present study focused on the southeastern part of the Mu Us dune field in which traditional agriculture practices dominate.

The original Landsat remote sensing image data used in the study were acquired from the geospatial data cloud (http://www.gscloud.cn). A total of 20 images from 2009 to 2018 were selected. Most of the images were selected from the growing season since the active cultivation of crops allowed the farmland to be easily identified. All selected images had cloud cover of less than 1% to minimize image errors. Table 1 shows a summary of the data chosen. In addition, the study used a Digital Elevation Model (DEM) derived from the LocaSpaceViewer software with a resolution of 9.55 m/per pixel.

Processing of the Landsat remote-sensing images for 2009 to 2018 included radiometric calibration, atmospheric correction, and data processing integration. These data were then used to determine the distribution of modern agricultural zones in the study area as characterized by the presence of CPI farmlands. More specifically, the yearly total amount of CPI farmland units from 2009 to 2012 was less than 300. The visual interpretation method by ENVI and GIS software was used to identify the CPI farmlands units. Since there has been a large increase in the number of CPI farmlands units from 2012 to 2018, both visual and computerized methods were used within their enumeration. The TM 432 and OLI 543 band combinations were chosen to identify the boundaries of the CPI farmland units. The false color composite images were then used to interpret the units.

The remote-sensing images from 2009 to 2018 were used to extract CPI farmland units. These data were used to create a database to represent the status of CPI farmland agriculture in the Mu Us dune field, mainly in the southeastern region. The results showed yearly increases in the numbers and areas of CPI farmland units (Figure 2a). While the numbers and total area of CPI farmland in 2009 were only 24 and ~8.11 km², their numbers and total area had increased dramatically to 2,149 and ~340.72 km², by 2018, respectively. It showed that the CPI farmlands in our study region started from scratch, and then blossom, reflected in the pronounced expansion of numbers and total area (Figure 2).

The annual variations in the total area and number of CPI farmland units (Figure 2b) first showed an increasing trend up until 2015, and then declined. This result indicated an increasing oscillation in the development of modern CPI farmlands in the southeastern part of Mu Us dune field. There was a booming in both the total area and number of CPI farmland units from 2009 to 2015. The maximum growth rate during this period occurred in 2014 when the CPI farmland units and area increased by 428 and 80.80 km², respectively. However, there were rapid declines in the area and number of CPI farmland units during 2015 to 2016, followed by a further increase in their number from 2016 to 2017 during they increased by 293, although their total area continued to rapidly drop. This contradiction could be attributed to a rise in the number of CPI farmlands with a small area, particularly in the 0-0.2 km² and 0.2-0.4 km² ranges (Figures 2c and 2d). As shown in Figure 2c, over the past ten years, CPI farmland units with areas in the range of 0-0.2 km² and 0.2-0.4 km² have become dominant, with few units with an area larger than 0.8 km². During the initial years of the study period, the proportion of CPI farmland units with an area 0.2-0.4 km² was very high (71%), followed by those with an area of 0.4-0.6 km² (29%). The proportions of the CPI farmland units with these sizes gradually dropped with the progression of the study period (Figure 2c). While local people and the local government initially attempted to promote the use of CPI farmland units with larger areas exceeding 0.6 km² or 0.8 km², the results of the present study showed that these larger units were not in expected expansion. It was increasingly realized that small-scale CPI farmland units (0-0.2 km²) were more applicable to the region, with the proportion of this size range progressively increasing during the study period (Figures 2c and 2d). In general, the meso- and micro-scale CPI farmland units, mainly with areas of 0-0.2 km² and 0.2-0.4 km², became increasingly popular.

The total area and number, their variation, and the variation in the proportions of CPI farmland units of different sizes showed that the development processes could possibly be divided into three stages: (1) the period of rapid development during 2009-2015, including the maximum value during 2014-2015; (2) the turning point period during 2015-2016; (3) a possible stable development period after 2016. In addition, it was assumed that the development of modern agriculture, characterized by the use of CPI farmland, will enter a period of stability in the future.

The study identified the spatial variation in the CPI farmland units using density analysis to explore the development of modern agriculture. The results of the interannual variability (Figure 3) in the CPI farmland units during the early period showed that CPI farmlands began to appear in the western part of Erdos City in the Inner Mongolia Autonomous Region and in Yulin City in Shaanxi Province (Figures 3a-3c), with relatively limited areas and numbers. There were extensive increases in the numbers and area of CPI farmland units during 2011 to 2012, particularly in the Otog Front Banner, western Erdos City. The most noticeable change during 2012 to 2015 (Figure 3d-3f) was the centralization of CPI farmland units with scalable development, distinguished from their scattered and sporadic previous distribution. The Yulin City showed prominent increases in the use of CPI farmland units during this time span. In general, western Erdos City constituted the centroid of CPI farmland development in the study area over the study period. As shown in Figure 3e, there were remarkable increases in the number and area of CPI farmland units during 2014 to 2015, coinciding exactly with the variations in total area and total number mentioned in Section 3.1. However, the growth rates in the number and area of CPI farmland units during 2015 to 2016 showed clear declines, as reflected in the degree of scatter and density. The declines in the CPI farmland units in the Erdos region were larger than those during the other years. There was a general recovery in the area and numbers of CPI farmland units from 2016 to 2018, mainly in Otog Front Banner and Uxin Banner, Western Erdos region, whereas there were only minor increases in Yulin City, Shaanxi Province.

Regardless of the development period, CPI farmlands remained dominant in the Western Erdos region in the southeastern Mu Us dune field, primarily with the centralized type. In contrast, the development of CPI farmlands remained limited in the Yulin region. However, the results of a field campaign and image interpretation showed that the CPI farmlands in the Yulin region were marked by the emergence of the form of combinations, exactly as shown in Figure 2d. Therefore, the degree of mechanization was relatively higher than that in the Erdos region, Inner Mongolia.

The development of the CPI farmlands from 2009 to 2018 in the southeastern Mu Us dune field were shown in the section above. To understand these changes involved, we now consider some possible natural influence factors below.

Agricultural development in China remains in a transitional stage (Dcr et al., 2021). A certain spatial disparity in the level of agricultural mechanization in China remains. Consequently, government policies have encouraged the adoption of modern agricultural practices in the Mu Us dune field, such as the CPI farmland units. The question should therefore be asked whether the continued expansion of the CPI farmlands in the high-profile Mu Us dune field is sustainable. Our study indicates that the main natural factors influencing the distribution and development of CPI farmlands, whereas from the perspective of different stakeholders including the government, the field master/enterprise and the local farmer, anthropogenic factors may be vital.

Government perspective. Policy-driven, large ecological restoration projects implemented in recent decades can exert a positive impact on vegetation restoration (Xu et al., 2018). In the Mu Us dune field, government policy plays a large role in regional agricultural development. In particular, in recent years there has been a focus on the construction of high-quality farmland to lay a solid foundation for the modernization of agriculture and to ensure grain security. Agricultural schemes issued by local government (Figure 5, Published) in the study area included the Policies on Supporting the Development of Modern Agriculture (Trial) in 2018 in the city of Yulin, Shaanxi Province and The Enforcement Advice about Further Enhancement of Construction of High Standard Farmland in order to improve national capacity to ensure food security in 2020 in the city of Erdos, Inner Mongolia Autonomous Region (Figure 5, Public Document). These schemes have in particular recommended the development of CPI-dominated water-saving irrigation projects through the awarding of large financial incentives (Figure 5, Bonus) for the construction of high-standard modern agriculture farmlands, thereby providing essential technical (Figure 5, Technical Support) and financial support (Figure 5, Initial Investment) for new large-scale intensive agriculture (area exceeding 7 acres).

Field master/Enterprise perspective. It is clear that substantial investment from field master/enterprise has been made to promote modernization of regional agriculture (Figure 5, Investment). The study found that the development of most of the modernized farmland, mainly represented by vast areas of CPI farmlands, was supported by government policies (Figure 5, Public Document). This result considered with the variations in the areas of CPI farmland units of different sizes (Figure 2c) indicates that decision-makers including the local government and actual investor had identified the optimal size and scale of the CPI farmlands through years of trial and error, with these aspects stabilizing after 2014. In addition, these policies had some negative outcomes, such as the conversion of some sandy area to wasteland.

Local farmer perspective. Supported by favorable policies, measures, and funds, the local farmer as the actual practitioner, made these big and rapid changes in the development of CPI farmlands from decades in the study area, meanwhile they obtained real profits from the technical improvement of modern intensive agriculture.

Advantages. Clearly policies by national and local government in China have served as a guideline for scaling up effective investments in agriculture and in rural areas as part of efforts to eradicate poverty. The CPI technology has facilitated greater agricultural productivity, thereby increased food production and realized food security. Previous studies have shown that the adoption of the CPI systems increase crop yield and improve water-use efficiency (O'Shaughnessy et al., 2016; Darko et al., 2017).

Disadvantages and further adverse effects. The Mu Us dune field has become popular in recent years for research into the prevention and management of desertification and both the government and people have made great effort to control desertification. Nevertheless, the question remains whether the policy regulating modernization of regional agriculture is beneficial from an ecological perspective. It is suggested that the current trend in agricultural modernization in the Mu Us dune field area is unsustainable and will inevitably lead to the destruction of native arenicolous vegetation. Generally, the development of new CPI farmlands required reclaiming sandy land, which lead to the loss of natural vegetation and increase the risk of soil erosion and reactivation of fixed dunes. From November to April, the developed CPI farmland units lie dormant. These large expanses of bare ground experience aggravated wind erosion, leading to land degradation. A previous study identified that serious environmental problems, such as intensified soil erosion, were triggered by the large-scale land consolidation and development in the Mu Us dune field (Shi et al., 2019).

Field observations in the Mu Us dune field show a lack of soil and water conservation measures, such as straw covering and stubble mulch farming. From the angle of irrigation, reasonable irrigation management based on the water balance was needed (Qassim et al., 2008). The optimal pattern of irrigation application could prevent water stress, control salinification, and resulted in high irrigation efficiency (Izquiel et al., 2015; Domínguez et al., 2022). The Mu Us dune field contains the main area of the Three-North Shelter Forest Program and is also a source area of the sand and dust affecting the cities of Beijing and Tianjin in China. Therefore, the unplanned development of CPI farmlands in Mu Us dune field can possibly have a detrimental effect on the acquirement of government ecological and environmental targets. In addition, the results of field surveys (Figure 5) indicated that local farmers utilizing CPI farmlands considered the operating costs to be prohibitive. Therefore, these respondents were not willing to accept the costs associated with the expansion of CPI farmlands.

Proposal. Based on the background in understanding of interactions between regional ecological recovery, desertification combating work, and economic development, although the construction of CPI farmlands in Mu Us dune field was partially policy driven, for the policy-makers, both natural and socioeconomic factors must should be considered. In terms of the development process, the scale of CPI farmland units was adjusted effectively and stabilized at the economical and practical meso- and micro-scale after 2016.

Figure 5 shows a summary of the different stakeholders and their relationships as well as advantages and disadvantages of the development of CPI farmlands. In conclusion, our study indicates that the disadvantages of the CPI farmlands generally outweigh their advantages. Nevertheless, modern agricultural development remains a priority within the sandy land areas, because these areas with unconsolidated sediments could meet the need of relatively flat terrain after the land consolidation, generally (Chen et al., 2019b). Although Chen et al. (2019b) considered that it was not sustainable to vigorously popularize CPI farmland at improving the agricultural production conditions in the Farming-Pastoral Ecotone of Northern China, the Mu Us dune field has its particularity. As a natural resource widely distributed in the southeastern Mu Us dune field, the ancient peat has a high water-retention capacity and high organic matter content, which could improve the agriculture production efficiency (Liu et al., 2018).

Given the ecological significance and risk of further desertification, it is suggested that further modernization of regional agriculture should be more forward-looking and well-planned, which should take many natural and anthropogenic factors into consideration, such as terrain, climate patterns, geological substrate, and other feasible national and regional development strategies. Besides, further study is still needed to uncover the optimal quantity, scale and spatial distribution of CPI farmland.