Classification and Evaluation of Marginal Land for Potential Cultivation in Northwest China Based on Contiguity and Restrictive Factors

Abstract

Water, land, and other environmental conditions restrict marginal land (ML) conversion into newly cultivated land. Accurately evaluating ML’s development and utilization potential (DUP) can provide critical support for increasing new cultivated land and ensuring food security. This study focuses on Northwest China, using spatial identification of different types of ML based on remote sensing images, and constructs a county-level DUP evaluation model through contiguous characteristics and restrictive factors to determine new cultivated-land potential, water demand, and liftable grain production. The results show that the DUP of ML in Northwest China is 12.59 million ha, with low-efficiency cultivated land (LCL) and two types of restoration land (TTRL) accounting for 3.29% and 5.95%, and other marginal land (OML) making up 90.76%. The total water demand for ML development and utilization is 69.87 billion cubic meters, which can increase grain production by 62.31 million tons. The coordinated development of water, land, and food promotes an increase in grain production, with water resources being the main restrictive factor. This model effectively evaluates DUP and provides a scientific basis for promoting the rational use of water and land resources. Further research should set up more detailed water resource utilization strategies and scenarios as well as find more development and utilization techniques.

1. Introduction

Marginal land (ML) is an essential supplementary cultivated-land resource for the Chinese “Storing grain in the ground, storing grain in technology” strategy based on cultivated-land management and technological application. To ensure national food security, it is essential to simultaneously improve the quality of existing cultivated land and rationally develop reserve cultivated-land resources. Given the complex and volatile global situation, with the global food supply chain facing pressure from geopolitical tensions and climate change, ensuring food security is strategically significant for China, a populous nation [1,2]. The Ministry of Natural Resources initiated research on increasing new cultivated land in early 2024, with ML being the primary source of new cultivated land. China’s northwest region is vast and sparsely populated, with abundant ML resources. However, due to the limitation of water resources, most ML is challenging to develop [3]. In order to solve the problem of water shortage and balance regional differences, in 2002, China launched a major national water diversion project [4]. The South-to-North Water Diversion Project, a key cross-regional water resource allocation project [5,6], provides new opportunities for ML development in the northwest.

The definition of ML has been a constant concern for global political and research institutions and remains controversial [7]. The concept and its application can generally vary from region to region, country, and organization to organization [8]. ML refers to land with low agricultural productivity and economic benefits and fragile ecosystems due to significant soil obstacles, strong constraints of water and thermal resources, and severe topographical limitations. These typically include low-efficiency cultivated land, non-cultivated ML, and other types [9,10]. The concept of “marginal land” is analyzed from two primary perspectives: from an economic perspective, it is defined by the relationship between agricultural land income and input costs; from an ecological perspective, it is defined by natural endowment conditions and crop-planting restrictions.

Based on the results of the Third National Land Survey, cultivated-land resource quality classification, and agricultural land classification, this paper classifies ML into three categories: low-efficiency cultivated land (LCL), two types of restoration land (TTRL), and other marginal land (OML). LCL refers to cultivated land with low output due to natural conditions, land degradation, reduced soil fertility, and insufficient irrigation facilities. In China, LCL is identified according to the national standard of Cultivated Land Quality Grade, in which cultivated-land quality grades are classified from high to low as 1 to 10, of which cultivated land that scores 7 to 10 is usually considered LCL [11]. TTRL includes “immediate restoration” and “engineering restoration” land, which can be directly restored to cultivation or require engineering measures before being restored [12]. It belongs to the reserve resources of cultivated land. OML refers to the portion of reserve cultivated-land resources that is difficult to develop into cultivated land, including other grasslands, saline–alkali land, sandy land, bare land, and similar types [13].

The development and utilization of ML is an important measure to ensure food security [14]. With these new opportunities, the development and utilization potential (DUP) of ML in the northwest has attracted considerable attention. Accurately assessing the DUP of ML is critically essential. Faisa et al. quantified the DUP of ML in different regions of China, focusing on low-carbon environmental protection, and evaluated the potential for bioenergy crop planting [15]. Mellor et al. distinguished ML types and found that bioenergy can avoid food competition and direct land-use conflicts [16]. Esch et al. evaluated ML development and utilization in Canada and found that planting crops on ML is often unprofitable [17]. Yu et al. quantified the potential for ML development for growing energy crops and analyzed the changes in agricultural greenhouse gas emissions after ML development [18]. In most cases, ML is linked to economic costs and ecosystem services. However, studies have yet to focus on quantifying the potential area of ML that could be converted into cultivated land.

At present, many scholars have explored the evaluation index system and methods for assessing the DUP of ML. Zhang et al. considered the land types included in ML and used crop models to evaluate the DUP of ML for energy crops [19]. Scordia et al. conducted research using experimental stations in three climate zones in Europe to compare the effects of crop types and restrictive factors on the DUP of ML [20]. Zhao et al. evaluated and classified the suitability of ML in Shaanxi Province using the maximum entropy model [21]. Song et al. considered engineering conditions and restrictive factors, constructed an evaluation model for the DUP of ML, and divided the potential into theoretical and actual potential [22]. Akram et al. evaluated the relationship between the yield and water demand of oil palm planted on ML in Indonesia [23]. Csikós et al. reviewed indicators for evaluating the DUP of ML and found that the indicators could be grouped into soil properties, environmental conditions, and economic factors [24]. Although various factors often restrict the development and utilization of ML, and some scholars have evaluated its potential, there is a lack of quantification based on the contiguity and natural attributes of ML.

Based on the above issues, this paper focuses on the following aspects: (1) classifying and identifying the quantity and distribution of ML, considering the contiguous characteristics and restrictive factors of ML, and constructing a potential evaluation model; (2) quantifying the DUP of ML in Northwest China, along with the water demand and the potential increase in grain production; and (3) analyzing the characteristics of restrictive factors in the development and utilization of ML while comprehensively considering the synergistic relationship between water, land, and food and also proposing development and utilization strategies. The study innovatively classified ML into three categories—LCL, TTRL, and OML—to quantify their potential regarding contiguity and restrictive factors. The study aims to provide a scientific basis for the rational use of water and land resources as well as support the preliminary demonstration of the West Route Project of the South-to-North Water Diversion Project, offering practical implementation suggestions for ML development and national food security protection.

2. Materials and Methods

2.1. Methods

From an engineering perspective, this study divided ML into LCL, TTRL, and OML, using these three types of ML as the research objects to evaluate ML’s DUP. There are several methods for evaluating the DUP of ML, with common approaches including machine learning [25], maxent [21], and SWOT [17]. On this basis, the study established a DUP evaluation model. The model considered the scale, contiguity characteristics, and restrictive factors of ML to comprehensively evaluate its DUP and quantify the potential increase in grain production and water demand after the development of ML. The operation process of the evaluation model (Figure 1) was as follows: (1) based on existing land-use data, we identified ML of different types and categorized ML in the northwest region into three groups: LCL, TTRL, and OML; (2) based on the type identification results, we conducted spatial agglomeration analysis and restrictive factors analysis of ML in the northwest to obtain its contiguity characteristics and restrictive factor attributes; (3) using the type identification and spatial analysis results, we calculated the DUP of ML (virtual) newly added cultivated land in the northwest, where the potential of LCL is treated as the potential of virtual cultivated land, and we quantified the water demand and grain yield that can be increased through the development and utilization of ML.

2.1.1. Identification of Marginal Land Types

The study used 30 m resolution raster data for land class identification based on the GEE platform for radiometric calibration, atmospheric correction, image cropping, and the spectral characteristics of typical ground objects analysis [26]. The kth land class in the remote sensing image R was classified and identified, and the data of low-efficiency cultivated land, other grassland, saline land, sandy land, bare land, low-efficiency garden land, and residual forest land were obtained, and the land class identification formula was as follows:

$$C_{mk}=f\left(R_k\right)$$

(1)

where C_mk represents the value of the k th land class in the m th administrative district, such as low-efficiency cropland, other grassland, etc., and f (R_k) denotes the function that recognizes the land class from the remotely sensed image R.

The results of land category identification were overlaid with the spatial data of TTRL to obtain the OML. The area of different types of ML was statistically summarized using county-level administrative divisions as statistical units to obtain the scale of ML in each county S_i, which was calculated by the following formula:

$$S_i={\sum }_k{C}_{mk}$$

(2)

where S_i represents the size of the ML of category i.

2.1.2. Potential Evaluation Model

Characterization of Contiguity

Cold and hot spot analysis, as a local autocorrelation analysis method, can be used to identify the hot and cold spots in ML’s spatial distribution in this paper [27]. In general, hot spots with higher GIE and are surrounded by other hot spots with higher GIE. The Getis–Ord G* index is a commonly used approach to distinguish cold and hot spots for a region [28,29]. Hot spot analysis is commonly used to obtain the contiguity characteristics of ML. The formula for hot spot analysis is given below:

$${Gi}^{\ast }={\displaystyle \frac{\sum _{j=1}^n{w}_{h,j}{x}_j-\overline{X}\sum _{j=1}^n{w}_{h,j}}{s\sqrt{\frac{\left[n\sum _{j=1}^n{w_{h,j}}^2-{\left(\sum _{j=1}^n{w}_{h,j}\right)}^2\right]}{n-1}}}}$$

(3)

where x_j is the attribute value of element j, w_h,j is the spatial weight between elements h and j, n is the total number of elements, Gi* is the score value indicating the spatial attribute, $\overline{X}$ is the mean of x_j, and S is the standard deviation of all point attribute values:

$$\overline{X}={\displaystyle \frac{\sum _{j=1}^n{x}_j}{n}}$$

(4)

$$S=\sqrt{{\displaystyle \frac{\sum _{j=1}^n{x_j}^2}{n}}-{\left(\overline{X}\right)}^2}$$

(5)

Taking the scale of different types of ML in each county as the object, based on the calculation results of hot spot analysis to obtain the characteristics of different types of ML agglomeration in each county A_i, the calculation formula is as follows:

$$A_i={\displaystyle \frac{{Gi}^{\ast }-{Gi}_{min}^{\ast }}{{Gi}_{max}^{\ast }-{Gi}_{min}^{\ast }}}$$

(6)

where A_i represents the agglomeration eigenvalue of the ML of category i, Gi*_min is the minimum value in Gi*, and Gi*_max is the maximum value in Gi*.

Analysis of Restriction

The natural breakpoint method suggests that there are turning points and breakpoints within any sequence that occur naturally rather than being set by humans, and these breakpoints serve as effective boundaries for classification. This study proposed that in a given area, when the proportion of a third-level (worst level) area of a factor exceeds a certain threshold, the factor negatively impacts the quality of cultivated land in the region and is considered a restrictive factor. How was the threshold for this ratio determined? By taking each county as a sample, a binary natural breakpoint was applied to the third-level area ratio for all samples of a certain factor, and the two resulting breakpoint values were used as thresholds. The degree of constraint of a factor in a region was then characterized by the ratio of the area of tertiary cropland in a region to the threshold value [30]. When the degree of constraint was more than 0.8, the factor was considered to be a restrictive factor in the region.

$$C_j={AP}_j/T_j$$

(7)

where C_j is the constraint degree of the j th element, AP_j is the proportion of the third level of cropland area for the j th element, and T_j is the natural breakpoint value of the third level of data for the j th element.

After obtaining the restrictive factors of ML development and utilization in each county, counties with the number of restrictive factors less than 1 were classified as no restrictive category and mildly restrictive category, counties with the number of restrictive factors between 2 and 3 were classified as moderately restrictive category, and counties with the number of restrictive factors of 4 and above were classified as severely restrictive category. The NDVI under restrictive conditions was categorized into three categories, and the coefficients Li for the unrestricted or mildly restricted, moderately restricted, and severely restricted categories were determined by comparing the differences in NDVI under restricted and unrestricted conditions.

Result Evaluation

ML can only partially be converted into reclaimable cultivated land, and this process often requires many factors [24]. From an engineering perspective, land agglomeration significantly impacts cultivated land’s systematic and standardized development [25]. The more agglomerated the land, the lower the development cost, and the easier it is to develop. Therefore, this study took the agglomeration characteristics of ML and the restrictions caused by the natural endowment of the land itself as the two main factors affecting the development of ML into cultivated land. Based on the ML size S_i, agglomeration characteristics A_i, and restrictive coefficient L_i obtained from the potential identification operator for each county, the ML development potential D_i was calculated according to Equation (5), and the formula for the ML development potential is as follows:

$$D_i=S_i\ast A_i\ast L_i$$

(8)

Based on the calculated ML development potential D_i, the area of newly cultivated land ACLA_i for TTRL, the area of newly cultivated land for OML, and the area of virtual newly cultivated land VACLA_i for LCL were calculated based on the conversion rate of newly cultivated land CE. The food yield Y_i and water requirement W_i of crops were calculated based on the area of newly added cultivated land:

$$\left(V\right) {ACLA}_i=D_i\ast {CE}_i$$

(9)

$$Y_i=D_i\ast y_i$$

(10)

$$W_i=D_i\ast w_i$$

(11)

where (V) ACLA_i is the (virtual) area of newly cultivated land added to the ML of category i, CE_i is the conversion rate of newly cultivated land added to the ML of category i, and Y_i is the grain yield of the ML of category i that $y_i$ represents the grain yield (wheat/corn) per unit in each county. W_i is the water demand of the ML of category i, in which $w_i$ represents the irrigation water consumption per unit of cultivated land for wheat/corn in each county.

In China, corn, wheat, and rice are the three main food crops, meaning China’s grain production revolves around these three crops. As can be seen from Figure 2, the study area was located in the northwest of China, most of which belongs to arid and semi-arid areas that do not have the conditions for large-scale rice production. Despite limited water resources, corn, wheat, and cotton are the main crops in the region [31]. Therefore, although other crops such as rice, potatoes, and cotton are also grown in the region, the study considered the planting patterns as corn and wheat [32,33,34].

2.2. Study Area

In this paper, 233 counties in six provinces (autonomous regions), including Xinjiang, Qinghai, Gansu, Inner Mongolia, Shaanxi, and Ningxia, were selected as the study area (Figure 2). Based on the National Water Network Construction Planning Outline and other strategic requirements, this study selected the counties and districts along the project as the study object based on the South-to-North Water Diversion Master Plan and the route selection program of the Western Route Project. According to the national administrative boundaries and the results of natural zoning, the study area was carefully adjusted to ensure that the study area had clear boundaries and a moderate scale and that the selected area was representative and similar in terms of climate, topography, hydrology, and other natural conditions.

The study area is located in northwest China (33°30′ N–43°38′ N, 78°0′ E–111°14′ E), with large elevation differences, generally a high south and low north, complex topography, and diverse landforms, including plateaus, plains, mountains, basins, deserts, and other terrains. Rainfall is relatively low, with most areas receiving less than 200 mm of precipitation: mainly arid and semi-arid. The number of sunshine hours reaches 3400 h in some areas of Xinjiang and 1800 h in Shaanxi, and the average annual temperature is low, with a large difference in temperature from day to night and with a cold winter and hot summer. According to the Ecological Geographic Zoning of China, the study area contains 11 ecological zone types, as shown in

Table 1. The eco-geographical divisions involved in the study area.

Primary Region	Temperature Zone	Wet and Dry Region	Secondary Region
Arid northwest region	I. Medium temperate zone	C Semi-arid region	Eastern steppe of Inner Mongolia (I.C1)
		D Arid region	Desert steppe of the western part of the Ordos and Inner Mongolia Plateau (I.D1)
			Alxa and Hexi Corridor desert area (I.D2)
	II. Warm temperate zone	D Arid region	Tarim Basin desert area (II.D1)
Loess Plateau region	II. Warm temperate zone	B Semi-humid region	Deciduous broad-leaved forest and artificial vegetation area in Fenwei Basin (II.B1)
		C Semi-arid region	Grassland in the north-central part of the Loess Plateau (II.C1)
	III. Northern subtropics zone	A Humid region	Evergreen and deciduous broad-leaved mixed forest area in Qinba Mountain (III.A1)
Tibetan Plateau region	IV. Plateau temperate zone	C Semi-arid region	Coniferous forest and grassland area of Qingdong Alpine Basin, Qilian (IV.C1)
		D Arid region	Desert area of Qaidam Basin (IV.D1)
	V. Plateau subarctic zone	C Semi-arid region	Wide valley alpine meadow steppe area of Qingnan Plateau (V.C1)
		D Arid region	Alpine desert area of Kunlun Alpine Plateau (V.D1)

Note: The table is from the 2023 edition of China’s eco-geographical divisions [35].

. Soil types are mainly brown desert soils, wind sandy soils, etc. According to the remote sensing monitoring data of land-use status quo in 2020, the main land-use types in the study area are unutilized land, grassland, and cropland, accounting for 57.9%, 27.6%, and 7.6% of the total area of the study area, respectively.

2.3. Data Sources and Processing

2.3.1. Data Sources

The 2021 county-level administrative division data, DEM data, and precipitation data from 2018 to 2020 used in the study were obtained from the Resource and Environmental Science and Data Platform of the Chinese Academy of Sciences (https://www.resdc.cn, accessed on 7 May 2024). Additionally, 30 m resolution land-use data (2020) were obtained from the China Geographic Information Resource Catalog Service System (https://www.webmap.cn, accessed on 17 March 2024). Slope data were calculated from DEM data. The soil condition data used in the study, including soil thickness, soil texture, soil organic matter content, and soil pH, were obtained from the results of the survey and evaluation of the cultivated-land quality grade by the agricultural and rural sector as well as from supplemental sampling in the field. These results mainly include the data of the sample points of the survey, the evaluation of cultivated-land quality grades, and the database of main traits of the evaluation unit. The data on natural areas refer to “Ecological Geographic Regions of China” (https://www.zgbk.com, accessed on 28 April 2024). The fifth-grade river vector and runoff data are from the National Glacial Permafrost Desert Science Data Center (http://www.ncdc.ac.cn, accessed on 13 May 2024). Microbial diversity data were obtained from field sampling and institutional testing. Cultivated-land quality classification data were obtained from the results of the Third National Land Survey on Cultivated Land Resource Quality Classification. The irrigation water consumption per unit area of cultivated land for wheat or corn was derived from the “Agricultural Irrigation Water Quota: Wheat” and “Agricultural Irrigation Water Quota: Corn” issued by the Ministry of Water Resources of the People’s Republic of China (http://www.mwr.gov.cn, accessed on 6 June 2024). The grain production data are from the National Bureau of Statistics (https://data.stats.gov.cn, accessed on 6 June 2024).

2.3.2. Data Processing

This study graded and categorized the data on slope, soil thickness, soil texture, soil organic matter content, soil pH, and biodiversity. Concerning the results of the survey on soil physical and chemical properties in the national detailed investigation of soil pollution status on agricultural land, this study took them as the source for determining the values of classification indexes such as soil thickness, soil texture, soil organic matter content, soil pH, and biodiversity. The study determined the slope index value based on the slope distribution map of cultivated land in the Third National Land Survey. The paper summarizes and analyzes the area and distribution of cultivated land under different classification conditions, forms the summary results according to counties, and connects the data of natural areas and cultivated-land quality classification to form the database required for the study. The results of the graded classification of indicators are shown in

Table 2. Classification standards and codes for indicator classification.

Categorical Metrics	Indicator Connotation	Indicator Access	Grading and Classification Standards and Codes
			1	2	3
Slope	The degree of steepness of the surface unit to which the cultivated land belongs	Remote sensing-based DEM data	≤6°	6~15°	>15°
Soil thickness	The sum of the soil layer and the loose parent material layer (cm)	Soil profile survey	≥100	60~100	<60
Soil texture	Assemblage of mineral particles of different sizes and diameters in cultivated soils	Particle size analysis	Loam	Clay	Sand
Soil organic matter content	The amount of organic matter per unit volume of soil (g/kg)	Chemical oxidation or spectroscopy analysis (for example, potassium dichromate oxidation to measure organic carbon)	≥20	10~20	<10
Soil pH	The degree of acidity and alkalinity of the cultivated soil	pH meter for on-site measurement or laboratory measurement	6.5~7.5	5.5~6.5 or 7.5~8.5	<5.5 or ≥8.5
Biodiversity	The abundance of biological species	Field survey of soil fauna and microbial diversity sequencing to obtain chao1 index	Abundant	General	Not abundant

Note: 1 is the best grade; 2 is the second best; 3 is the worst.

.

The study extracted the 3-year average precipitation from 2018 to 2020, which was counted in counties. The study obtained five-level river vector data in the study area. Then, it multiplied each level of river data with river runoff, summarized them by county, corrected the summarized values according to the area, and obtained surface water resource data for each county. The study used ArcGIS 10.8 and IBM SPSS Statistics 25.0, Stata MP 17, Google Earth Engine, and Excel 2021 as the primary analysis tools.

3. Results

3.1. Analysis of Marginal Land Type Identification

The study identified ML based on land-use type data and classified ML in the northwest region into three categories: LCL, TTRL, and OML, counting the scale of different types of ML. The study results (Figure 3) show that the total ML scale of the study area is 37,965,600 ha, and the larger counties are mainly distributed in the southwestern and northeastern parts of the study area. Among these, 435,900 ha, i.e., 2.28%, were classified as LCL; 1,461,800 ha, i.e., 3.85%, as TTRL; and 35,637,800 ha, i.e., 93.87%, as OML.

Regarding different ML types, the LCL scale gradually decreased from east to west. The counties with larger scales are mainly distributed in the districts and counties such as Li and Gangu Counties in Gansu Province, Hengshan District, Shenmu City, and Suide County in Shaanxi Province and YijinhuoLuo Banner in Inner Mongolia Province, which accounted for more than 3% of the total scale of LCL and are mainly distributed in the eastern part of the study area. The counties with lower scales are mainly distributed in the central part of the study area. The scale of the TTRL gradually decreased from east to west. The counties with larger scales are mainly distributed in the districts and counties of Dali, Chengcheng, Mei, and Lantian Counties in Shaanxi Province and Qin’an County in Gansu Province, where the scale of the TTRL accounted for more than 2% of the total scale of the TTRL, and they are mainly distributed in the eastern part of the study area; the counties with lower scales are mainly distributed in the central part of the study area. The scale of OML then gradually decreased from west to east. The counties with larger sizes are mainly distributed in several counties, including Ruoqiang, Qiemo, and Hetian Counties in Xinjiang Province, Gelmu and Dulan Counties in Qinghai Province, and Alashan Left Banner and Ejina Banner in Inner Mongolia Province, where the size of OML accounted for more than 3% of the total size of OML, and it is mainly distributed in the central and western parts of the study area; the counties with lower sizes are mainly distributed in the southeast.

3.2. Characteristics and Factors of Marginal Land DUP

3.2.1. Analysis of Contiguity Characteristics

The study used hot spot analysis to obtain the contiguity characteristics of LCL, TTRL, and OML in the study area, and the spatial agglomeration of ML in the study area was analyzed based on Gi* values. The results of the hot spot analysis showed (Figure 4) that the hot spot areas of ML in the study area are mainly distributed in the southwest, and the cold spot areas are concentrated in the southeast.

From the perspective of different types of ML, the cold spot and hot spot patterns of LCL and TTRL are relatively close. The hot spot areas are concentrated in the easternmost part of the northwest region, while the cold spot areas are concentrated in the central and eastern parts of the study area. There is no apparent clustering phenomenon in the western region. The proportion of the LCL’s cold spot area and hot spot area was 9.89% and 52.11%, respectively. The cold and hot spot areas of the TTRL accounted for 14.44% and 56.05%, respectively. The cold spot area of the OML is concentrated in the easternmost part of the study area, and the hot spot area is widely distributed in the eastern and central parts of the study area. The proportion of cold and hot spot areas to OML was 13.04% and 55.15%, respectively.

3.2.2. Analysis of Restrictive Factors

This study used eight indicators, namely soil thickness, soil pH, soil organic matter content, soil texture, biodiversity, slope, average annual precipitation, and surface water resources (Figure 5), to identify the restrictive factors of ML in the northwest region. The number of restrictive factors of ML in each county in the study area was identified, and the counties were divided into mild or no restrictive areas, moderately restricted areas, and severely restricted areas according to the number of restrictive factors in each county.

The results of the analysis of the spatial distribution of restrictive characteristics are shown in Figure 6. The results show that the restrictive characteristics of different ML types in the study area are relatively close. The mild or no restriction areas are concentrated in the southeast of the study area, and the severely restricted areas are mainly distributed in the northeast. In different counties; the number of restrictive factors in LCL and TTRL is lesser, and the number of restrictive factors is 1–5. The number of restrictive factors in OML is greater, and the number of restrictive factors is 1–6.

From the perspective of different types of ML, the proportions of mild or no restriction area, moderately restricted area, and severely restricted area in LCL were 5.15%, 39.91%, and 54.94%, respectively. The counties with more restrictive factors are mainly distributed in Alashan Right Banner, Guazhou County, Bohu County, Yanchi County, etc. The proportions of mild or no restriction areas, moderately restricted areas, and severely restricted areas in the TTRL were 4.29%, 41.20%, and 54.51%, respectively. The counties with more restrictive factors are mainly distributed in Yijinhuoluo Banner, Qilihe District, Yutian County, Hongsibu District, and so on. The proportions of mild or no restriction area, moderately restricted area, and severely restricted area in OML were 9.44%, 39.48%, and 51.08%, respectively. The counties with more restrictive factors are mainly distributed in Alashan Right Banner, Guazhou County, Bohu County, Yanchi County, and other counties.

3.3. Analysis of Marginal Land DUP and Results Evaluation

Based on the results of the ML scale, contiguous characteristics, and restrictive factors, the DUP of (virtual) newly cultivated land in the northwest region was calculated. The concept of “virtual new cultivated land” was introduced to LCL. By calculating the DUP of “virtual new cultivated land”, the water demand and the liftable grain production for the development and utilization of LCL were quantified. The results show that (Figure 7) the counties with greater DUP are mainly distributed in the central and western parts of the study area, and the counties with lesser DUP are mainly distributed in the southeast. The DUP of most counties was below 100,000 ha, and counties with DUP above 100,000 ha are widely distributed in the study area’s central, western, and northern parts.

From the perspective of different ML types, the DUP of virtual new cultivated land based on LCL is about 435,900 ha (

Table 3. Marginal land DUP, water demand, and liftable yield.

	Low-Efficiency Cultivated Land (LCL) (Virtual) New Cultivated Land		Two Types of Restoration Land (TTRL) New Cultivated Land		Other Marginal Land (OML) New Cultivated Land		Summary
Development and utilization potential (DUP) (104 ha)	43.59		78.50		1137.08		1259.17
Current area (104 ha)	86.60		146.18		3563.78		3796.56
DUP/current area (%)	50.34		53.70		31.91		33.17
DUP/current cultivated land (%)	4.39		7.90		114.48		126.77
Crop type	wheat	corn	wheat	corn	wheat	corn
Water demand (108 m3)	7.26	11.21	18.00	24.22	259.35	378.68	698.72
Liftable yield (104 tons)	63.58	124.54	135.55	204.82	2500.92	3201.94	6231.35
Liftable/current yield (%)	5.39	4.18	11.49	6.88	211.92	107.58	149.92

Note: The current cultivated land area is 9,932,800 ha, and the current yield is 11,801,200 tons of wheat and 29,764,400 tons of corn.

), accounting for about 50.34% of the scale of LCL. The counties with greater DUP are Shenmu City, Li County, Gangu County, Jia County, etc., concentrated in the study area’s eastern part. The DUP of newly cultivated land based on the TTRL is about 785,000 ha, accounting for about 53.70% of the scale of the TTRL. The counties with greater DUP are Dali County, Zhouzhi County, Pucheng County, Heyang County, etc., mainly distributed in the west and southeast of the study area. The DUP of newly cultivated land based on OML is 11,370,800 ha, accounting for 31.91% of the scale of OML. The counties with greater DUP are Ruoqiang County, Gelmu City, Qiemo County, Dulan County, etc., mainly distributed in the western and central parts of the study area.

Based on the results of the newly cultivated-land potential of ML (virtual) in Northwest China, according to the proportion of planting structures in Northwest China, the water demand and grain yield that can be increased in the development and utilization of ML in Northwest China were quantified. The potential research results show that the proportions of virtual new cultivated land of LCL, new cultivated land of TTRL, and new cultivated land DUP of OML to the current cultivated land area were 4.39%, 7.90%, and 114.48%, respectively. Among them, the conversion rate of new cultivated land in TTRL was the highest (accounting for 53.70% of the current area).

The results of water demand research show that the water demand of ML development and utilization is related to the DUP, and the total water demand is 69.87 billion cubic meters. Among them, the water demand for developing and utilizing newly cultivated land in OML was the largest, which is 63.80 billion cubic meters. The proportion of wheat water demand to total water demand was 37.12%, and the proportion of corn water demand to total water demand was 54.20%. The second is the newly cultivated land of the TTRL, which is 4.22 billion cubic meters, and the proportion of wheat and corn water demand to the total water demand was 2–4%. The water demand for newly cultivated land in LCL is the lowest, at 1.85 billion cubic meters, and the water demand for wheat and corn accounted for less than 2% of the total water demand.

The study’s results on the increase in grain yield show that the development and utilization of ML in the study area can increase the grain yield by 62.31 million tons, accounting for 149.92% of the current yield, among them 27.01 million tons of wheat and 35.31 million tons of corn. LCL can increase wheat yield by 63.58 million tons and corn yield by 1.25 million tons, accounting for 5.39% and 4.18% of the current yield, respectively. The TTRL can increase wheat yield by 1.36 million tons and corn yield by 2.05 million tons, accounting for 11.49% and 6.88% of the current yield, respectively. OML can increase wheat yield by 25.01 million tons and corn yield by 32.02 million tons, accounting for 211.92% and 107.58% of the current yield, respectively.

4. Discussion

4.1. Marginal Land Development and Utilization Methods

Due to the different objectives and management methods of ML in different regions, the concept of ML is more extensive and has different types [36,37,38]. Based on the perspective of engineering construction, the study divided the ML in the northwest region into LCL, TTRL, and OML. It used the restrictive factors analysis method to obtain the restrictive factors of different types of ML. Through the characteristic analysis of the classification of restrictive factors, the results show that there are 146 counties in Northwest China where the ML development and utilization are mainly restricted by water resources, 8 counties that are mainly restricted by soil conditions, and 79 counties that are mainly restricted by water resources, soil, and environment. Water resources mainly restrict the ML development and utilization in Northwest China. From the perspective of different ML types, the restrictive factors of LCL and TTRL are fewer, and the restrictive factors of OML types are more. This result is similar to the research results of Feng et al. [39]. LCL and TTRL belong to cultivated land or former cultivated land. After long-term cultivation and utilization, the impact of these two types of ML constraints is reduced [40]. OML is not cultivated land, so OML is more challenging to develop and utilize.

Improving the restrictive conditions of ML is the primary means of developing and utilizing ML. In the study of improving water resource restriction, water-saving irrigation measures are an effective means to improve the utilization efficiency of water resources. Ishfaq et al. proposed measures such as dry–wet alternation and upland rice system, reducing production costs and saving water resources [41]. Crop planting structure and planting methods also affect water-use efficiency. Wang et al. evaluated and compared various indicators of output, water use, groundwater, and water production performance and selected the best planting system and irrigation system suitable for different situations [42]. Li et al. considered the water resources system’s resilience and optimized Heilongjiang’s agricultural planting structure [43]. Zhou et al. proposed different row-spacing planting techniques and found that water-use efficiency increased by about 10% [44]. In the development and utilization of ML in the northwest region, the irrigation water-saving technology suitable for the region should be selected, and the planting structure should be adjusted to improve the utilization efficiency of water resources and reduce the limitation of water resources.

In the study of improving soil conditions, the restrictive factors of ML are mainly due to soil desertification and salinization. Soil desertification affects the soil texture, soil thickness, soil organic matter content, and other soil conditions of ML. Li et al. conducted a study on the coupling relationship between cultivated-land use and desertification. They found that the proportion of cultivated land and desertification are inverted U-shaped curves [45], and the development of ML for desertification should be moderate. Gao et al. studied the relationship between rocky desertification, desertification, and land development and utilization. The results show that slope and land reclamation rates are essential factors affecting land desertification and rocky desertification. They believe that engineering measures should be combined to improve the slope of ML [46], such as soil improvement projects, biological measures, etc. Soil salinization affects the soil type and utilization of ML more. Xu et al. used composite materials to improve coastal saline–alkali soil in China. They found that composite materials containing desulfurized gypsum and biochar could effectively reduce soil salt content by about 93.7% [47]. Goswami et al. reviewed the latest research progress of using salt-tolerant rhizosphere microorganisms to alleviate crop salt stress, expounded on the interaction mechanism of soil salinity with microorganisms and crop roots, and believed that it was feasible to alleviate soil salinization based on microorganisms [48]. Zhou et al. used limestone conditioner, planting of Jerusalem artichoke, and other measures to improve saline–alkali land and found that both measures reduced soil salt content and constructed an improved method of planting Jerusalem artichoke and applying conditioner and soil microorganisms [49]. In developing and utilizing ML in Northwest China, desertification control and saline–alkali land improvement technologies suitable for the region should be selected to increase soil fertility, improve soil quality, and reduce soil conditions.

In summary, ML development in the northwest region should be adapted to local conditions, and different measures should be adopted for different restrictive factors. Engineering measures such as soil improvement and irrigation and drainage management can be used to reduce the water, soil, and environmental restrictive factors of ML and moderately develop ML.

4.2. Analysis of Marginal Land Development and Utilization Based on Water–Land–Food Framework

The water demand for developing ML is calculated based on the DUP of ML and the irrigation quota data in the northwest region. The results show that the total water demand for developing ML in the northwest region is 69.87 billion cubic meters. According to statistics, from 2014 to 2024, the cumulative water transfer volume of the East and Middle Routes of the South-to-North Water Transfer Project accumulated 70 billion cubic meters. The annual water transfer volume is about 7 billion cubic meters [6], far lower than the total water demand for the development of ML in the northwest region [50]. However, the northwest region is perennially dry, and the surface water source is limited. The water used for developing ML mainly comes from the South-to-North Water Transfer Project, which means that there is a priority in the development and utilization of ML. Regarding restrictive water resources, priority should be given to developing LCL and TTRL with higher grain production efficiency. However, this study calculated the water demand of the northwest region based on the potential of ML (virtual) newly cultivated land but lacks comprehensive consideration of water resource loss and utilization methods. Further research should set up more detailed water resource utilization strategies and scenarios to determine the water demand for the development and utilization of ML in the northwest region.

The study calculated the grain yield of different types of ML. The results show that the yield of wheat and corn could be doubled after the development and utilization of ML compared with the current grain yield of cultivated land. Some scholars have put forward a similar point of view. Han et al. evaluated grain production in Northwest China by considering virtual aquatic products. They found that Northwest China can increase grain production after replenishing water resources [51]. The development and utilization of ML should consider the grain production efficiency of different types of ML. The grain production efficiency of LCL is higher than that of TTRL and OML. The TTRL and OML have more restrictive factors when planting grain crops, such as being vulnerable to wind and water erosion, terrain slope more significant than 25°, or environmental problems [52]. The grain production efficiency of TTRL is higher than that of OML. The TTRL refers to land whose land-use type is initially cultivated, but the current land-use type is not cultivated land due to the behavior of returning cultivated land to forest and abandoning cultivated land. In general, these MLs can be immediately restored to cultivated land or restored to cultivated land through engineering, so the soil fertility of TTRL is higher than that of OMLs [53]. Although the grain production efficiency of LCL and TTRL is higher, the proportion of the DUP of these two types of ML in the study area was only 9%, and the proportion of OML was higher. In the development and utilization of ML, more consideration should be given to the constraints and development and utilization techniques of OML.

At present, grain production no longer only depends on a single factor or condition. However, it should focus on the link between water, land, and food and turn the thinking of single-factor governance to the synergistic improvement of the efficiency of the “water–land–food” system. Many scholars have studied the relationship between water, land, and food [54,55,56]. Although this study did not conduct scenario analysis or systematic quantification of the water–soil–grain relationship, the synergistic relationship can still be seen from the connection between the newly added cultivated-land area, water demand, and the increase in grain production (Figure 8). Figure 8a displays a ternary-phase diagram normalized based on water demand, cultivated-land area, and grain production. Cui et al. quantitatively analyzed the relationship between water, energy, and grain in a pumping irrigation system in related studies. They found that improving irrigation system resource-use efficiency contributes to the sustainability of the regional water–energy–grain relationship [57]. Buttinelli et al. similarly found that irrigation water positively correlates with the productivity of other inputs and grain production [58]. Yue et al. considered factors such as precipitation and agricultural water demand; simulated the allocation of water, land, and food resources under different scenarios; and proposed corresponding management strategies [59]. Zhang et al. analyzed the impact of crop type, climate conditions, and irrigation technology on crop water demand and examined how water resources influence land development and grain production [49]. These studies demonstrate the complex relationship between water, land, food, and other subsystems such as energy, they and suggest that water, land, and food often exhibit a synergistic effect. Future research on ML development should consider different scenarios influenced by multiple factors, such as water, land, and food. Establishing ML development models for different scenarios can improve efficiency. In short, when considering the development and utilization of ML under the framework of water, land, and food, the relationships between these three factors should be comprehensively analyzed to optimize water and soil allocation and promote grain production through coordinated development.

4.3. Limitation

The study was conducted at the county level. This is a larger scale for type identification and restrictive factor analysis. While analysis at a smaller scale (such as the plot scale) could provide more accurate results, obtaining comprehensive data at that level is relatively tricky. Future research should improve data acquisition methods to enhance the accuracy of hot spot analysis. The current analysis method is relatively simple, relying mainly on spatial analysis and data calculations. A more sophisticated model should be employed to incorporate factors such as local ecosystems and climate change into the calculations.

This research examined ML development and utilization from the engineering construction perspective. In addition to considering the constraints of natural resources, contiguous characteristics are important for engineering construction. However, the study did not address the food competition and land-use conflicts that may arise from the development of ML [16], nor did it analyze the constraints imposed by various economic and social factors, such as farmers’ willingness and labor availability [24]. Moreover, changing the land-use type and large-scale development of ML may have immeasurable impacts on ecosystems and climate [15,18]. ML and its surrounding environments typically have low carrying capacities and poor ecosystem stability. Future research should focus on a more extensive and in-depth exploration of the potential risks associated with the development of ML.

5. Conclusions

This study focused on the ML in Northwest China, assessing the DUP of LCL, TTRL, and OML. The scale of three ML types was determined through type identification. Hot spot analysis was used to obtain the contiguity of ML in the study area, and the restrictive factors analysis method was employed to determine the restrictive factors characteristics of these lands. Consequently, the DUP of ML in Northwest China was calculated. To provide a scientific basis for further development and utilization of ML in Northwest China, the main conclusions of this paper are as follows:

(1) The total scale of ML in Northwest China is 37.97 million ha, including 0.87 million ha of LCL, 1.46 million ha of TTRL, and approximately 35.64 million ha of OMLs. The DUP of virtual new cultivated land based on LCL is about 0.44 million hectares, accounting for 50.34% of the LCL. The DUP of the TTRL is about 0.79 million ha, accounting for 53.70% of the restoration land area. The DUP for newly cultivated-land development of OMLs is 11.37 million ha, accounting for 31.91% of the OML area;

(2) The total water demand for the development and utilization of ML is 69.87 billion cubic meters. Among this, the new cultivated-land development water demand of OMLs is 63.80 billion cubic meters, the TTRL requires 4.22 billion cubic meters, and the LCL requires 1.85 billion cubic meters. The development and utilization of ML can increase the total grain yield by 62.31 million tons, accounting for 149.92% of the current yield. This includes 27 million tons of wheat and 35.31 million tons of corn. LCL can increase wheat yield by 0.64 million tons and corn yield by 1.25 million tons; the TTRL can increase wheat yield by 1.36 million tons and corn yield by 2.05 million tons; OMLs can increase wheat yield by 25.01 million tons and corn yield by 32.02 million tons;

(3) The synergistic development of water, land, and food resources is beneficial for increasing grain yield, with water resources being the main restrictive factor for such development in Northwest China. The restrictive factors analysis results show that the development and utilization of ML in Northwest China are primarily constrained by water resources, followed by soil conditions. Environmental conditions pose a smaller constraint on the development and utilization of ML. In some counties, ML is affected by a combination of water resources, soil, and environmental constraints. The development and utilization of ML should be tailored to local conditions, adopting different measures for different restrictive factors. From the perspective of the synergistic development and utilization of water, land, and food, priority should be given to developing LCL and the TTRL with fewer constraints to achieve optimal allocation of water, land, and food resources.

In summary, the research results indicate that ML in Northwest China holds great potential for development and utilization, which could significantly boost regional grain production. However, due to the limited resources in the Northwest, the harsh climate and poor soil conditions are not conducive to ML development, and the water demand for such development is substantial. Therefore, to develop ML in Northwest China, it is necessary to fundamentally alter the region’s water resource conditions through significant water diversion projects, such as the West Route Project of the South-to-North Water Diversion Project. Additionally, it is crucial to consider the order of priority of ML development and the synergistic relationship between water, land, and food as well as to optimize resource allocation.