Evolution of Urban–Agricultural–Ecological Spatial Structure Driven by Irrigation and Drainage Projects and Water–Heat–Vegetation Response

Abstract

In the context of global climate change and intensified water resource constraints, studying the evolution of the urban–agricultural–ecological spatial structure and the water–heat–vegetation responses driven by large-scale irrigation and drainage projects in arid and semi-arid regions is of great significance. Based on multitemporal remote sensing data from 1985 to 2015, this study takes the Inner Mongolia Hetao Plain as the research area, constructs a “multifunctionality–dynamic evolution” dual-principle classification system for urban–agricultural–ecological space, and adopts the technical process of “separate interpretation of each single land type using the maximum likelihood algorithm followed by merging with conflict pixel resolution” to improve the classification accuracy to 90.82%. Through a land use transfer matrix, a standard deviation ellipse model, surface temperature (LST) inversion, and vegetation fractional coverage (VFC) analysis, this study systematically reveals the spatiotemporal differentiation patterns of spatial structure evolution and surface parameter responses throughout the project’s life cycle. The results show the following: (1) The spatial structure follows the path of “short-term intense disturbance–long-term stable optimization”, with agricultural space stability increasing by 4.8%, the ecological core area retention rate exceeding 90%, and urban space expanding with a shift from external encroachment to internal filling, realizing “stable grain yield with unchanged cultivated land area and improved ecological quality with controlled green space loss”. (2) The overall VFC shows a trend of “central area stable increase (annual growth rate 0.8%), eastern area fluctuating recovery (cyclic amplitude ±12%), and western area local improvement (key patches increased by 18%)”. (3) The LST-VFC relationship presents spatiotemporal misalignment, with a 0.8–1.2 °C anomalous cooling in the central region during the construction period (despite a 15% VFC decrease), driven by irrigation water thermal inertia, and a disrupted linear correlation after completion due to crop phenology changes and plastic film mulching. (4) Irrigation and drainage projects optimize water resource allocation, constructing a hub regulation model integrated with the Water–Energy–Food (WEF) Nexus, providing a replicable paradigm for ecological effect assessment of major water conservancy projects in arid regions.

1. Introduction

In the context of global climate change and increasing rigid constraints on water resources, large-scale irrigation and drainage projects in arid and semi-arid regions have become strategic infrastructure for maintaining food security and ecological balance. Through reshaping the surface hydrological process, these projects are profoundly altering the land use/land cover (LULC) structure and ecosystem service functions in arid and semi-arid regions [1,2,3]. The Hetao Plain, as the largest unified gravity-fed irrigation area in China, has relied on the Sanshenggong Water Control Project and its supporting canal system (a comprehensive renovation project, not new construction) since 1961. The project has undergone three phases of modernization: 1961–1985 (initial construction of the central trunk canal), 1985–2009 (expansion and renovation of branch canals), and 2009–2015 (water-saving transformation and ecological compensation). Currently, the total length of the Yellow River irrigation canal system is 6873 km, controlling an irrigation area of 574,000 hectares, with an annual water diversion of 47.3 × 10⁸ m³ (equivalent to 4.73 billion cubic meters), supporting 1.412 million people in Bayan Nur City and ensuring food security. However, traditional engineering assessment mostly focuses on hydrological regulation and yield enhancement benefits, and existing studies have prominent methodological gaps: (1) Most focus on single land categories (e.g., farmland or wetlands) and lack cross-type comparative measurements across urban space (US), agricultural space (AS), and ecological space (ES), leading to fragmented engineering impact assessment; (2) most only compare pre- and post-construction data, ignoring the construction disturbance peak, operational lag response, and long-term steady-state evolution of the entire life cycle, making it difficult to identify phased thresholds and reversibility of impacts [4]; and (3) the relationship between LST and VFC is often simplified as a linear negative correlation, without separating the interactive interference of multiple factors such as irrigation thermal effect, planting structure transformation, and hydrological redistribution, failing to reveal the unique “water–heat–green” coupling characteristics in arid areas. Under the dual pressures of increasingly rigid water diversion volume constraints and groundwater overexploitation governance, how to quantify the spatial structure effect throughout the project’s life cycle, reveal the response threshold of surface parameters, and identify the synergistic optimization path of “increased yield–increased greenery–reduced temperature” has become a key scientific issue for regional sustainable development [5,6,7,8]. This study defines the project’s life cycle as three phases: (1) construction phase (1985–1997): including initial construction (1985–1991) and late construction (1991–1997), focusing on canal excavation, reservoir construction, and infrastructure layout, with intense surface disturbance; (2) operational phase (1997–2009): including early operation (1997–2003) and mid-operation (2003–2009), characterized by stable water diversion, improved irrigation guarantee rate, and gradual adaptation of the ecosystem to hydrological changes; and (3) post-operation optimization phase (2009–2015): focusing on water-saving renovation, ecological compensation, and integrated governance of “mountains, rivers, forests, farmlands, lakes, grasslands, and sands”, with balanced development of production and ecological functions.

Current research on the ecological effects of irrigation and drainage projects has formed three main lines. The first is the cutting effect of water conservancy facilities on landscape patterns [9]. Wang Jing et al. [10] established an ecosystem assessment framework covering the entire basin to investigate the impact of water conservancy facilities on the ecosystem in the source of the Tarim River–Hotan River Basin. It was found that the impact of water conservancy facilities on the low shrub ecosystem was the most significant (26.38%) from 1980 to 2020. Hanbing Zhang et al. [11] identified the dominant factors driving urban green space degradation in arid cities, emphasizing the interaction between water conservancy projects and LULC change. The second is the response mechanism of LST and VFC [12]. LST (land surface temperature) refers to the temperature of the Earth’s surface skin, VFC (vegetation fractional coverage) is the proportion of vegetation-covered area in a pixel, and NDVI (Normalized Difference Vegetation Index) is an indicator reflecting vegetation growth status. Bouchelouche Asma [13] found that over time, there is a clear linear negative correlation between LST and NDVI, and the reduction in vegetation leads to an increase in LST and vice versa. Haldar Sarbeswar et al. [14] found that urban industrial expansion is an important factor causing temperature rise and changes in vegetation and surface water bodies. The third is the functional trade-off study of urban–agricultural–ecological space [15,16,17,18]. Zexu Chen et al. [19] examined dynamic changes in the ecological–production–living space in Ganzhou City, Jiangxi Province, over 20 years, finding that the functional score of the ecological space was consistently higher than that of the production and living spaces, highlighting the dominant position of ecological function in this region. Ziqiang Bu et al. [20] found that the same piece of land may have different production, living, and ecological functions; the dominant function of the land will affect and restrict other uses. Unordered urbanization and industrialization have intensified conflicts over land use, livelihoods, and ecological functions, a significant factor that has restricted regional sustainable development. In addition, in terms of technical methods, Ma Jun [21] constructed a comprehensive fragmentation index in the study of global forest fragmentation, confirming that the combination of edge density (ED) and MPA (Maximum Patch Area) can better explain LST variation than the single PD (Patch Density). However, this study did not include the human-driven factor of engineering. The single-window algorithm LST inversion model proposed by Z. Qin et al. [22] achieved a verification accuracy of ±1.8 °C in arid regions and has been widely applied in the study of water and heat processes in irrigation areas.

Although the above research has made significant progress, the aforementioned methodological gaps remain unresolved [23]. To address these deficiencies, this study takes the Inner Mongolia Hetao Irrigation Area as the empirical object. The Hetao Plain is unique for three reasons: (1) As China’s largest gravity-fed irrigation area, its 6873 km canal system and 47.3 × 10⁸ m³ annual water diversion form a typical “engineering-driven water–land coupling system” in arid regions. (2) It has a temperate continental arid climate (annual precipitation 138–185 mm and evaporation 2200–2400 mm) with scarce water resources and fragile ecosystems, making it a sensitive area for testing the ecological effects of water conservancy projects. (3) It presents a typical urban–agricultural–ecological mosaic pattern, providing a representative case for studying spatial structure evolution. This study constructed a classification system for urban–agricultural–ecological space based on the “multifunctionality–dynamic evolution” dual principles and adopted a three-level scoring method to quantitatively depict the temporal changes in land use function intensity, breaking through the limitations of traditional static classification; in remote sensing interpretation, the technical process of “separately interpreting a single land category in different time periods and then merging them” was adopted, which improved the classification accuracy to 90.82%. It revealed the evolution path of the ecological thermal environment “short-term disturbance–long-term steady state” driven by drainage and irrigation projects and the phased reversal and spatial heterogeneity mechanism of the relationship between LST and VFC, providing a replicable paradigm for the ecological effect assessment of major water conservancy projects in arid areas.

2. Overview of the Study Area

The study area is located in the core of the Hetao Plain in Bayan Nur City, Inner Mongolia (40°10′–41°20′ N, 106°20′–109°20′ E), covering Dengkou County, Hangjinhouqi, Wuyuan County, Urad Front Banner, and Linhe District (the municipal government’s location). As shown in Figure 1 (the figure adds the distribution of core irrigation and drainage projects), the total area is approximately 1.68 × 10⁴ km², and the permanent resident population in 2023 was 1,412,000. The region has a temperate continental arid climate, with an average annual temperature of 7.2 °C, annual precipitation ranging from 138 to 185 mm, and annual evaporation of 2200–2400 mm. The sunlight is abundant, but the water resources are scarce. The terrain is a river alluvial plain, with higher elevations in the southwest and lower elevations in the northeast, ranging from 1020 to 1050 m. The irrigation canal system of the Yellow River is densely distributed, and it is the largest self-flowing irrigation area in the country. Since the completion of the Sanshenggong Water Control Project and the main trunk canal in 1961, the engineering system has undergone three phases of modernization and renovation. Currently, there are 13 main canals and 48 branch canals, with a total length of 6873 km, and the annual diversion of Yellow River water is 47.3 × 10⁸ m³, controlling a total irrigation area of 57.4 × 10⁴ hm². The study area presents a typical urban–agricultural–ecological mosaic pattern: Linhe District is a regional central city, with 18.7% of urban construction land; the agricultural space is dominated by cultivated land (accounting for 61.3%, mainly water-irrigated land), concentrated in the area east of the main trunk canal; and the ecological space includes the Ulanbuhesh Desert in the southwest (within Dengkou County, with an area of approximately 0.31 × 10⁴ km²) and the Wulansu Sea Wetland in the northeast (in Urad Front Banner, the largest freshwater lake in the Yellow River Basin), jointly forming a barrier for wind prevention and sand fixation and biodiversity conservation [24]. In agricultural production, the study area widely adopts plastic film mulching and facility agriculture to improve water use efficiency and crop yield, which has a significant impact on surface albedo and turbulent fluxes. In recent years, with the tightening of water diversion from the Yellow River and the advancement of groundwater overexploitation control, the spatial structure evolution driven by the engineering and the ecological thermal environment feedback has become a key scientific issue for the region’s sustainable development.

3. Research Method

3.1. Data Sources and Preprocessing

The remote sensing images in this article mainly come from the Landsat satellite series operated by the National Aeronautics and Space Administration (NASA, Washington, DC, USA) and received by the Remote Sensing Satellite Ground Station of the Chinese Academy of Sciences (Beijing, China). Specifically, the data acquisitions are based on the following sensors: from 1985 to 2013, the Thematic Mapper (TM) sensor onboard LANDSAT 5, manufactured by Hughes Santa Barbara Remote Sensing (Santa Barbara, CA, USA, now part of Raytheon Technologies); from 2011 to 2012, the Enhanced Thematic Mapper Plus (ETM+) sensor onboard LANDSAT 7, manufactured by Raytheon Intelligence & Space (El Segundo, CA, USA) was used to fill the data gap between LANDSAT 5 and 8; and from 2013 to 2015, the Operational Land Imager (OLI) sensor onboard LANDSAT 8, manufactured by Ball Aerospace & Technologies Corp. (Boulder, CO, USA). The spatial resolution is divided into two types: 30 × 30 m (multispectral bands) and 15 × 15 m (panchromatic band) [22]. To obtain 15 m high-resolution multispectral images, the Gram–Schmidt pan-sharpening method was used to fuse panchromatic and multispectral bands, preserving spectral information while improving spatial resolution. To ensure the accuracy of the data and enhance the spectral characteristics of the remote sensing information, the FLAASH model in ENVI 5.6 software is used for image atmospheric correction. Additionally, based on the 1:250,000 basic geographic data (http://www.webmap.cn/main.do?method=index (accessed on 10 September 2025)), the remote sensing data of the study area are geometrically corrected in ENVI 5.3 software. The processing method primarily combines cubic polynomials and nearest-neighbor interpolation. During the correction process, the Krasovsky_1940 projection was selected. At the same time, to ensure the accuracy of the geometric correction, 15 to 16 ground control points were uniformly arranged in each image. Finally, the corrected TM images will be used as the benchmark to geometrically correct the other OLI images to ensure the accuracy and reliability of the data. According to the national standard “National Land Use Status Classification” (GB/T 21010-2017 [25]), the land use data are reclassified into different land use types, which are classified into construction land, cultivated land, forest land, grassland, water area, unused land, etc. At the same time, the land use types are precisely identified based on the identification characteristics of the remote sensing images.

3.2. Classification of Urban–Agricultural–Ecological Space

The classification of urban–agricultural–ecological spaces is the basis for analyzing their spatial patterns [26]. The land use system, as a complex composite of multiple levels of structure and functions, has a causal relationship of “element–structure–function” [27,28,29]. Different land use types carry various dominant functions, and their combination forms differentiated land use structures, thereby achieving multiple land use functions [30]. The spatial aggregation of functions then forms specific functional spaces. Thus, it can be seen that space is the carrier of functions, and the identification and classification of urban–agricultural–ecological spaces based on the dominant functions undertaken by land use types is feasible.

Based on the multifunctionality and dominant function characteristics of the land, this study follows the following two principles for classification:

(1)

Balancing multifunctionality and highlighting the main function principle: Land is a comprehensive entity for the interaction of urban–agricultural–ecological functions. However, different land use methods and intensities will result in differences in the primary and secondary functions of the three types of spaces. Therefore, classification should highlight the main function of the land while also considering the secondary functions. For example, cultivated land, as the foundation of national food production, has the dominant function of agricultural production; at the same time, cultivated land also has ecological functions, and contiguous crops can, to some extent, regulate the surrounding ecological environment. Therefore, the function of cultivated land should be defined as “agricultural–ecological function”.

(2)

Emphasizing the principle of dynamic functional evolution: Land functions have temporal dynamics. For example, grassland initially has a dominant function of livestock production and also takes into account the ecological protection function; however, in the later stage, with the evolution of human–land relations and the implementation of the ban on grazing policy, its ecological function has significantly increased. In contrast, the production function has tended to weaken. Therefore, the function of grassland should be transformed from “agricultural–ecological function” to “ecological function”.

Based on the above classification principles and existing research [31,32,33] combined with the actual land use conditions in the study area during different periods, this paper grades and quantifies the urban–agricultural–ecological space functions of each type according to their intensity (see

Table 1. Classification levels of urban–agricultural–ecological space.

Year	1985–1991			1991–1997			1997–2003			2003–2009			2009–2015
Spatial Type	US	AS	ES	US	AS	ES	US	AS	ES	US	AS	ES	US	AS	ES
Farmland	0	5	3	0	5	3	0	5	3	0	5	1	0	5	1
Garden	0	5	3	0	5	3	0	5	3	0	5	3	0	5	3
Forest land	0	0	5	0	0	5	0	0	5	0	0	5	0	0	5
Grassland	0	1	5	0	3	5	0	3	5	0	1	3	0	1	5
Transportation land	3	1	1	3	1	1	3	1	1	3	1	1	3	1	1
Water body	0	3	5	0	3	5	0	3	5	0	3	5	0	3	5
Construction land	5	0	0	5	0	0	5	0	0	5	0	0	5	0	0
Unutilized land	0	0	1	0	0	1	0	0	1	0	0	1	0	0	1

). Specifically, a three-level scoring system is adopted (the highest score is 5 points, the medium score is 3 points, the lowest score is 1 point, and the absence of function is 0 points). The scoring is based on two empirical bases: (1) expert consultation: where 10 experts in land use planning, remote sensing ecology, and water conservancy engineering scored the function intensity of each land type, with a consistency coefficient of 0.87; and (2) literature calibration: referring to functional scoring standards of similar studies in arid regions [31,33]. Limitation: The scoring is semi-quantitative and may be affected by regional heterogeneity; future research can introduce objective indicators (e.g., net primary productivity for ecological function) for optimization.

3.3. Land Use Information Extraction

To ensure the universality and fairness of the spectral sampling points, this study used the Create Random Points function in ArcGIS 10.0 to generate 1000 random points within the study area. This study adopted a method of separately interpreting each land use type using the maximum likelihood algorithm, followed by merging. The specific process is as follows:

• For each land use type (e.g., cultivated land and water area), select the time period with the most significant spectral difference from other types (e.g., cultivated land in July–August and water area in September) for separate supervised classification using the maximum likelihood algorithm [34,35];
• During the merging phase, for conflicting pixels (pixels classified as different types in separate interpretations), adopt a “spectral similarity + neighborhood analysis” method: calculate the spectral angle matching (SAM) between the pixel and the standard spectrum of each candidate type and combine the dominant type of adjacent 3 × 3 pixels to determine the final type (retaining the type with SAM < 0.1 and neighborhood proportion > 50%).

The evaluation of the accuracy of land use remote sensing mapping is a very important task, and it is generally assessed from two aspects: area consistency and spatial consistency [36,37,38,39]. Area consistency emphasizes the accuracy in terms of quantity, used to verify whether the quantity of each category of the evaluated data is consistent with that of the corresponding categories in the reference data, and it is usually characterized by consistency coefficients or correlation coefficients; spatial consistency compares the classification results of specific locations with the categories of the corresponding points in the reference data and is usually measured by confusion matrices. Based on a large number of sample data, this study conducted a quantitative evaluation of the spatial consistency of land use remote sensing mapping. The specific formula is as follows [38]:

$$O={\displaystyle \frac{{\displaystyle \sum _{i=1}^n{X}_{ii}}}{N}}$$

(1)

$$K={\displaystyle \frac{N{\displaystyle \sum _{i=1}^n{X}_{ii}}-{\displaystyle \sum _{i=1}^n{X}_{i+}{X}_{+i}}}{N^2-{\displaystyle \sum _{i=1}^n{X}_{i+}{X}_{+i}}}}$$

(2)

where O represents the overall accuracy; N is the total number of sample points used for accuracy evaluation; n is the total number of columns in the confusion matrix; X_ii represents the number of sample points in the i-th row and i-th column of the confusion matrix; X_i+ and X_+i are the total number of sample points in the i-th row and i-th column, respectively; and K is the Kappa coefficient.

For historical data (1985–1999) where high-resolution Google Earth images are unavailable, the ground truth was established through two sources: (1) regional land use statistical yearbooks (1985–2000, provided by Bayan Nur Bureau of Natural Resources) and (2) archival high-resolution aerial images (1986 and 1995, provided by the Inner Mongolia Water Conservancy Department). Accuracy was verified using the confusion matrix and Kappa coefficient, with an overall accuracy of 87.6% for 1985–1999 (Kappa = 0.78), meeting research requirements

Using ArcGIS software, 1000 sample points were randomly selected in the study area, and the land use/cover classification results for 2000, 2005, and 2010 were extracted. Combined with field survey data and Google Earth data, the sample point data of the study area were determined, and the accuracy was evaluated using the confusion matrix (

Table 2. The evaluation of LUCC classification accuracy in the study area.

Year	2000	2005	2010	Mean Value
Farmland	94.25	96.45	90.56	93.75
Forest land	90.52	93.56	92.21	92.10
Grassland	92.32	90.21	93.21	91.91
Water body	90.54	90.25	92.14	90.98
Construction land	89.55	90.32	85.45	88.44
Unutilized land	88.25	86.78	88.21	87.75
Overall Accuracy (%)	90.91	91.26	90.30	90.82
Kappa coefficient	0.845	0.881	0.83	0.81

). For 2011–2015, Google Earth high-resolution images (2013, 2015) were used for validation, with an overall accuracy of 89.7% (Kappa = 0.79). Limitation: Historical validation data (1985–1999) lack high-resolution reference images, and accuracy was indirectly verified by consistency with regional land use statistics [38].

The precise evaluation results show that the overall accuracy of separately interpreting different land use types in different time periods reaches 90.82%, with a kappa coefficient of 0.81. Moreover, the interpretation accuracy of different land use types in different time periods is all above 85%, indicating that the remote sensing images in different time periods can objectively and realistically reflect the characteristic differences of vegetation growth and effectively distinguish different land use/cover types.

3.4. Characterization of Surface Temperature

Based on Landsat 5 and 8, the surface temperature was retrieved using the Qin Zhihao single-window algorithm [40]. The formula is as follows [41]:

$$T_s={\displaystyle \frac{a(1-C_0-D_0)+\left[b(1-C_0-D_0)+C_0+D_0\right]T_1-D_0{T}_0}{C_0}}$$

(3)

where T_s represents the surface temperature (K); T₁ represents the brightness temperature (K); T₀ represents the average atmospheric temperature (K), obtained from local meteorological station data; C₀ = ϵτ; D₀ = (1 − τ)[1 + (1 − ϵ)τ]; ϵ is the surface emissivity (estimated based on the NDVI threshold method); and τ is the atmospheric transmittance (calculated based on water vapor content). Through algorithm accuracy verification, when performing linear regression with the synchronous MODIS LST product (MOD11A1), R² = 0.84 and RMSE = 1.8 °C, meeting the research requirements.

3.5. Characterization of Vegetation Coverage

The VFC was retrieved using the pixel binning model [42,43]:

$$VFC={\displaystyle \frac{NDVI-NDV{I}_{soil}}{NDV{I}_{veg}-NDV{I}_{soil}}}$$

(4)

where NDVI_soil and NDVI_veg are taken as the 5% and 95% cumulative frequencies of the image, respectively, as the threshold values for bare soil and pure vegetation pixels. The NDVI calculation formula is [42,43]:

$$NDVI={\displaystyle \frac{{\rho }_{NIR}-{\rho }_{Red}}{{\rho }_{NIR}+{\rho }_{Red}}}$$

(5)

where ρ_NIR represents the reflectance of the near-infrared band of the remote sensing image, and ρ_Red represents the reflectance of the red-light band of the remote sensing image. Through this formula, the normalized vegetation index (NDVI) of each pixel in the study area can be calculated. Its value range is [−1, 1], and the higher the value, the stronger the vegetation coverage. The calculated NDVI data need to be combined with the previously determined NDVI_soil (bare soil pixel threshold) and NDVI_veg (pure vegetation pixel threshold) and then substituted into the pixel binary model (Formula (2)) to inversely calculate the vegetation coverage (VFC) of the study area. This provides key parameter support for subsequent analysis of the response of vegetation to the spatial structure evolution driven by irrigation and drainage engineering.

3.6. Analysis Method for the Evolution of Urban–Agricultural–Ecological Spatial Pattern

3.6.1. Land Transfer Matrix

The land use transfer matrix can identify the transfer directions and quantities of various land use types across different time periods [44] and quantify the transfer rates for each land category. The formula is as follows [45]:

$$S_{ij}=\left[\begin{array}{cccc}{S}_{11}& S_{12}& \dots & S_{i1n}\\ S_{21}& S_{22}& \dots & S_{2n}\\ \dots & \dots & \dots & \dots \\ S_{n1}& S_{n2}& \dots & S_{nm}\end{array}\right]$$

(6)

where n represents the area of each land use type, and ij indicates the transfer amount of each land use type within the time interval.

3.6.2. Standard Deviation Ellipse Model

The standard deviation ellipse method, also known as the “Lefever directional distribution”, mainly analyzes and studies the evolution trajectory of urban–agricultural–ecological space in the research area, the spatio–temporal migration characteristics of the discrete spatial distribution pattern, and the degree of discretization in each direction based on the changes in the coordinates of the ellipse center, the ellipticity, and the direction angle. Its calculation formula is [46]:

$$\tan \theta =(A+B)/C$$

(7)

$$A={\displaystyle \sum _{i=1}^n{\tilde{x}}_i^2}-{\displaystyle \sum _{i=1}^n{\tilde{y}}_i^2}$$

(8)

$$B=\sqrt{({\displaystyle \sum _{i=1}^n{\tilde{x}}_i^2}-{\displaystyle \sum _{i=1}^n{\tilde{y}}_i^2{)}^2+4({\displaystyle \sum _{i=1}^n{\tilde{x}}_i}-{\displaystyle \sum _{i=1}^n{\tilde{y}}_i)}}}$$

(9)

where ${\tilde{x}}_i$ and ${\tilde{y}}_i$ represent the deviation between the spatial coordinates of the mean center.

4. Results Analysis and Discussion

4.1. Urban–Agricultural–Ecological Spatial Structure Change in the Drainage and Irrigation Project

In order to study the changes in the urban–agricultural–ecological spatial structure during each stage of the irrigation and drainage project, this paper classifies the spatial structure of urban–agricultural–ecological space for each stage of the study area through spatial classification, as shown in Figure 2. According to

Table 3. Transfer matrix of urban–agricultural–ecological space in the whole stage of drainage and irrigation engineering.

	US	AS	ES
US	1580.211	0.5616	12.9276
AS	117.0324	39,755.9133	2374.5258
ES	205.5753	2956.2606	41,826.762

, across the five stages, the area of urban space has increased continuously from 249.46 km² in the initial stage to 558.58 km² in the final stage, mainly due to the coordinated transfer of agricultural and ecological space. Among them, the largest scale of ecological-to-urban transfer occurred in the mid-term of completion (88.20 km²), reflecting the acceleration of the urbanization process of the project. From

Table 4. Urban–agricultural–ecological space transfer matrix of each stage of drainage and irrigation engineering.

	At the initial stage of construction
	US	AS	ES
US	247.455	0	2.0052
AS	7.128	8184.384	493.8165
ES	3.8286	238.2498	8589.0888
	During the later stage of construction
	US	AS	ES
US	256.6134	0.0882	1.71
AS	10.7208	7526.745	885.1689
ES	15.6942	422.1702	8647.0461
	At the beginning of its construction
	US	AS	ES
US	280.9791	0.0396	2.0097
AS	15.3999	7562.853	370.7505
ES	46.1979	806.7339	8680.9932
	Mid-term completion
	US	AS	ES
US	338.8347	0.3006	3.4416
AS	33.1902	8036.7516	299.6847
ES	88.1982	783.8262	8181.729
	In the later stage of completion
	US	AS	ES
US	456.3288	0.1332	3.7611
AS	50.5935	8445.1797	325.1052
ES	51.6564	705.2805	7727.9049

and Figure 2, it can be seen that from the initial construction to the later stage, there was a drastic spatial replacement: the conversion volume from ecological space to agricultural space increased from 238.25 km² to 422.17 km², reflecting the temporary occupation of ecological land such as grassland and wetlands by channel excavation. Field surveys found that the soil texture of the canal buffer zone was consistent with the surrounding ecological land, confirming the conversion of ecological land for engineering construction [24]. This was a necessary physical space demand for the implementation of the project. At the same time, the conversion volume from agricultural space to urban space increased from 7.13 km² to 15.69 km², with an average annual increase of 119%, indicating that the improvement of infrastructure enhanced the accessibility of the region, triggering the urbanization process in the peripheral areas. During this stage, the loss of ecological space (net reduction of 942.04 km²) was significant. However, the main conversion entity was marginal ecological land, and the core area’s ecological function was not fundamentally damaged.

From the mid-term to the later stage of completion, a key turning point was presented: the stable volume of agricultural space (AS → AS) increased from 8036.75 km² to 8445.18 km², with an increase of 4.8%, indicating that after the improvement of the irrigation system, the rate of abandoned farmland decreased significantly, and the improvement in the irrigation guarantee rate promoted the stabilization of marginal farmland, while the project directly guaranteed the function of grain production. What is more notable is that the conversion volume from agricultural space to ecological space was 325.11 km² in the later stage, which was a slight increase compared to the mid-term (299.68 km²) but a 54% decrease compared to the construction peak (692.3 km² in 1991–1997). This project effectively suppressed the “land conversion from forest to farmland and grassland to farmland” behavior caused by insufficient irrigation through optimizing water resource allocation.

The stock of urban space (US → US) continued to expand (247.46 → 456.33 km²). However, the growth mode shifted from external encroachment to internal filling: the direct encroachment of urban space on agriculture (AS → US) showed a marginal decreasing trend in the later stage (50.59 km² vs. 33.19 km² in the middle stage), and the occupation of ecological space (ES → US) in the later stage decreased from 88.20 km² to 51.66 km², indicating that the land consolidation policy supporting the project effectively guided the solidification of the urban boundary. In particular, the active transformation from urban space to ecological space (US → ES) increased from 3.44 km² to 3.76 km², indicating the activation of the urban green space compensation mechanism and confirming the ecological benefit feedback driven by the project.

Based on the data of the five periods, after the completion of the project, the net increase of agricultural space was 260.80 km² (8184.38–8445.18), while the ecological space decreased by 862.18 km². However, the conversion intensity from agricultural space to ecological space decreased by 42.3%, and the retention rate of the core ecological area (ES → ES) remained above 90% (7727.90/8589.09). This indicates that the true value of the irrigation and drainage project lies in improving the stability of each unit of agricultural space, reducing the continuous encroachment on ecological space, and achieving the goal of “increased production without increased land/increased efficiency without loss of greenery”. The matrix evolution confirmed the “short-term disturbance–long-term stability” path, providing a replicable land management paradigm for water conservancy construction in arid areas under ecological constraints.

To deeply explore the driving mechanism of irrigation and drainage projects on the urban–agricultural–ecological space, this study employed the standard deviation ellipse model for spatial analysis. In terms of the spatial distribution characteristics of the city, the ratio of the long semi-axis to the short semi-axis, the direction, and the coverage range of the standard deviation ellipse, respectively, represent the directionality, agglomeration, and expansion/contraction trend of the city space. The direction of the long axis of the ellipse (the extension direction in Figure 3a) indicates the dominant expansion direction of the city space (the ellipse shown is oriented northwest–southeast); a shorter short-axis length indicates an enhanced agglomeration degree in the secondary direction. The multi-period ellipse superimposition analysis reveals that the eccentricity of the ellipse was higher in the early periods (1985, 1991) (significant differences in the long and short semi-axes) and gradually approached a circular shape in the later periods, indicating that the city space shifted from a single-direction expansion to multi-directional diffusion, and the spatial agglomeration decreased. The temporal change in the area of the ellipse (coverage range) reflects the degree of expansion of the city space: from 1985 to 2015, the ellipse range continued to expand, indicating an overall expansion trend in the city’s space. The points in each year (1985, 1991, 1997, etc.) represent the center of the city’s spatial distribution, and their migration trajectories reflect the direction of the spatial center’s shift. The superimposed analysis of multiple periods shows that in the early stages (1985–1991), the center position was concentrated and the migration distance was short, indicating that the city space was in a local agglomeration development stage; in the later stages (1991–1997), the center migrated to the southeast, corresponding to the transformation of the city space from agglomeration to local expansion; in the initial and middle stages (1997–2003) and (2003–2009), the center continued to migrate southeastward and the displacement increased, indicating that the city space entered a rapid expansion period, with a clear expansion direction; and in the later periods (2009–2015), the migration rate of the center slowed down (the trajectory became smoother), marking that the city space gradually entered a stable diffusion stage.

The standard deviation ellipse of the agricultural space also presented a northwest–southeast orientation (consistent with the direction of the city space but with a different shape), indicating a significant correlation between its dominant distribution/extension direction and the overall spatial pattern of the region. In the early periods (1985 and 1991), the ellipse shape was compact; in the later periods (2009 and 2015), the range expanded, and the shape became more relaxed, reflecting a transition from a concentrated layout to a more moderate diffusion of the agricultural space. The area of the ellipse expanded over time (1985–2015), confirming the overall outward expansion trend of the agricultural space during the study period; meanwhile, the elliptical flattening (the difference between the long and short semi-axes) decreased, indicating a shift from a single-direction agglomeration to a multi-directional balanced distribution, with a decrease in spatial agglomeration degree. The analysis of the migration trajectory of the center shows that in the early stages (1985–1991), the center position was stable (short migration distance), and the agricultural space was mainly concentrated in the core areas (such as Linhe and Wuyuan local areas), without large-scale expansion; in the later stages (1991–1997), the center migrated to the southeast, corresponding to the expansion of agricultural space to the surrounding suitable areas (such as the southeast of Wuyuan); in the initial and middle stages (1997–2003 and 2003–2009), the center continued to migrate southeastward and the displacement increased, indicating that the agricultural space entered a rapid expansion period, in line with the development rhythm of regional development and the expansion of cultivated land/agricultural land; and in the later periods (2009–2015), the migration rate of the center slowed down (the trajectory became smoother), marking that the layout of the agricultural space became mature and stable.

The standard deviation ellipse of the ecological space also presented a northwest–southeast orientation (consistent with the direction of the city and agricultural space). However, in the early periods (1985 and 1991), the ellipse was located in the west (the Dengkou area). In the later periods (2009 and 2015), it expanded towards the southeast, indicating that the ecological space gradually shifted from a concentrated layout in the west to full-scale diffusion. The sequential expansion of the ellipse area reflects the overall expansion trend of the ecological space; a decrease in the elliptical shape indicates a shift from single-directional concentration to a multi-directional, balanced distribution. The early elliptical shape was compact (with a large difference between the major and minor axes), indicating that the ecological space was concentrated in the western core area. In the later stage, the ellipse became more relaxed, reflecting a decrease in spatial concentration and a trend of more dispersed distribution, which may be related to regional ecological restoration and the expansion of green spaces/eco-land. The characteristics of the center migration trajectory are as follows: In the early construction period (1985–1991), the center remained stable in the west (near Dengkou), and the ecological space was centered on the original ecological area in the west, without significant changes. In the later construction period (1991–1997), the center migrated slightly to the southeast, corresponding to the background of regional construction advancement, and the ecological space was moderately expanded to the southeast (in the direction of Wuyuan). In the initial and mid-stages of completion (1997–2003 and 2003–2009), the center continued to migrate southeastward and the displacement increased, indicating that the ecological space entered a rapid expansion period (possibly driven by ecological protection projects and green space construction). In the later completion period (2009–2015), the center migration rate slowed down, the ecological space layout gradually stabilized, and a comprehensive coverage pattern was formed.

Overall, it indicates that the ecological space was concentrated in the west (far from the urban/agricultural core area) in the early stage and expanded to the southeast in the later stage, forming a spatial complementary relationship with the expansion of urban and agricultural spaces (filling the gaps between urban and agricultural spaces); The consistency of the elliptical paths of the three systems reveals that the regional topography and planning policies jointly shaped the spatial pattern of coordinated development of city, agriculture, and nature.

4.2. Response Characteristics of Surface Parameters (LST and VFC) at Different Stages of Irrigation and Drainage Engineering

Figure 4 shows the spatial variation pattern of vegetation coverage (FVC). As shown in Figure 4a, in the early stage of the engineering construction, the FVC in the main areas of the central and western regions (Linhe, Wuyuan, Hangjinhouqi, and Dengkou) showed a downward trend; among them, the core areas of Linhe and Wuyuan presented high-value zones (dark red), indicating the most significant reduction in FVC. In the eastern region (a local part of Urad Front Banner), scattered green patches indicated an upward trend in FVC. The Dengkou area was mainly in yellow tones, corresponding to a relatively small change in FVC, which was related to the natural background of this area being in the transition zone between desert and oasis and having a relatively low background vegetation coverage. By the end of the engineering construction (Figure 4b), the main areas of the central and western regions (Linhe, Wuyuan, Hangjinhouqi, and Dengkou) were mainly in orange-red tones, indicating a continuous decline in FVC during this period; among them, the core areas of Linhe and Wuyuan remained high-value zones (dark red), reflecting a significant reduction in FVC. In the eastern region (a local part of Urad Front Banner), the scattered green patches persisted, indicating an increasing trend in FVC. The Dengkou area maintained yellow tones, with a relatively small change in FVC, consistent with the background of a low base coverage in the transition zone. At the beginning of the engineering operation, compared to the construction period, the spatial pattern presented a “central inversion, eastern decline” feature. The central core areas (Linhe and Wuyuan) changed from the “orange-yellow alternating (slowing decline)” during the construction period to continuous green patches, with FVC changing from “slowing decline” to “significant increase”; the eastern Urad Front Banner area changed from the “green expansion (increase)” during the later construction period to large areas of orange-red, with FVC changing from “increase” to “significant decline”; and the western Dengkou and Hangjinhouqi areas were mainly in yellow tones, with a small change in FVC and a fluctuating state. By the mid-term of the completion, the central Linhe and Wuyuan areas changed from “continuous green (increase)” to orange-red tones again, with FVC showing a downward trend again; the western Hangjinhouqi and Dengkou areas in the local area appeared scattered green patches, with FVC changing from “fluctuating changes” to “local increase”; the eastern Urad Front Banner area still mainly had orange-red tones, but with a small amount of green patches, and the decline in FVC slowed down. By the end of the completion, the spatial variation presented a “west degradation, east returning to increase” trend, continuing the “regional rotation” feature. Compared to the mid-term, the western Hangjinhouqi and Linhe changed from “local green (increase)” to orange-red, with FVC showing a downward trend; the eastern Urad Front Banner area changed from “mainly orange-red (decrease)” to the expansion of green patches, with FVC returning to an upward state; and the Dengkou area maintained yellow tones, with a small change in FVC.

The initial “central area reversal, eastern area decline” pattern during the construction period is the core evidence of the positive effect of the irrigation and drainage projects: the central core area changed from orange-yellow to a continuous green color, and the FVC (Farmland Water Conservation Capacity) changed from a gradual decrease to a significant increase, indicating that after the irrigation system was connected, the guarantee rate of farmland irrigation improved and the water replenishment guarantee for urban green spaces was activated. The vegetation recovered rapidly and even surpassed the pre-construction level, confirming the synergistic enhancement effect of the project on agricultural productivity and ecological service functions. The red color change in the eastern Ula’t Prefecture reflects that in the early stage of the project’s full operation, the focus of water resource allocation shifted westward, resulting in a temporary decrease in local water replenishment intensity and a slowdown in vegetation recovery, presenting a phased characteristic of “regional rotation”. The fluctuations of “western area reduction and eastern area recovery” in the middle and later stages of completion further reveal the coupling influence of operation optimization and planting structure adjustment: the local areas of Hangjin Hetao Prefecture and Dengkou in the west turned green again, possibly benefiting from the stable rise in groundwater levels after water-saving renovations and the implementation of the policy of converting low-efficiency farmland to grassland and returning it to wetland buffer zones [47]; the FVC in the eastern Ula’t Prefecture changed from reduction to increase, which is closely related to the regular ecological water replenishment of Wulansuhai Lake and the initiation of the wetland buffer zone restoration project. This “this decreasing and that increasing” regional rotation pattern is essentially a landscape reflection of the dynamic reallocation of water resources between agriculture and ecology in the project, ensuring stable production in the irrigation core area while promoting the restoration of ecological vulnerable areas through elastic scheduling, highlighting the pivotal regulatory value of the irrigation system in the integrated governance of mountains, rivers, forests, farmlands, lakes, grasslands, and sands in arid regions [48]. Overall, although there was an instantaneous impact during the construction period, after the project was put into use, the FVC in the study area generally showed a favorable trend of “central area stable increase, eastern area fluctuation recovery, and western area local improvement”, verifying the sustainability of major water conservancy facilities in ecologically fragile areas [49].

Figure 5 presents the spatiotemporal variation characteristics of surface temperature (LST) in the study area during different periods. During the initial construction period (1985–1991), a cooling trend was observed in the central part of the region, while the eastern part showed a temperature increase. Specifically, the central areas of Linhe and Wuyuan were characterized by a blue tone (lower LST). In contrast, the eastern Ula’t Prefecture was dominated by a red tone (with an increase in LST). The western areas of Dengkou and Hangjin Houcheng presented a “red-blue alternating” distribution (with significant fluctuations in LST). During the same period, the vegetation coverage (FVC) showed a “decrease in the central area and a slight increase in the eastern area”, but LST and FVC did not show a completely inverse relationship (such as a decrease in FVC in the central area leading to a decrease in LST), which might be influenced by irrigation activities (the thermal inertia of water bodies causing a slow temperature increase). In the later construction period, the west warmed, while the east cooled. The western areas of Dengkou, Hangjin Houcheng, and Linhe were dominated by a red tone (with a significant increase in LST). In contrast, the eastern Ula’t Prefecture was dominated by a blue tone (with a significant decrease in LST). During the same period, FVC showed a “local increase in the west and a decrease in the east”, and LST and FVC presented a clear inverse relationship: in the western areas with increased FVC, LST increased (possibly due to the weakened evaporative cooling effect after the improvement of vegetation coverage), while in the eastern areas with decreased FVC, LST decreased (or affected by precipitation fluctuations). In the initial construction period, the central area warmed, while the east and west showed fluctuations. Linhe in the central area was dominated by a red tone (with a significant increase in LST). In contrast, the areas of Wuyuan and some local areas in the east were dominated by a blue tone (with a decrease in LST). Hangjin Houcheng in the west presented a “red-orange alternating” distribution (with mild fluctuations in LST). During the same period, FVC showed a “increase in the central area and a decrease in the east”, and LST and FVC presented a local inverse relationship: in the western areas with increased FVC, LST increased (possibly due to the interference of the urban expansion-induced heat island effect), while in the eastern areas with decreased FVC, LST decreased (or affected by the fluctuation of precipitation). In the later construction period, the west cooled while the central and eastern areas warmed. Hangjin Houcheng and Dengkou in the west were dominated by a blue tone (with a decrease in LST), while Linhe in the central area and some areas in the east were dominated by a red tone (with a significant increase in LST). During the same period, FVC showed a “decrease in the west and a slight increase in the east”, and LST and FVC presented a local inverse relationship: in the eastern areas with decreased FVC, LST increased (possibly due to the interference of the urban expansion-induced heat island effect), while in the western areas with increased FVC, LST decreased (or affected by the fluctuation of the cooling in the desert area). The relationship between LST and VFC presents spatiotemporal misalignment, which is driven by the combined effect of irrigation water thermal inertia and crop phenology changes: During the construction phase (1985–1997): The central region showed a 0.8–1.2 °C decrease in LST despite a 15% reduction in VFC, mainly due to the thermal inertia of irrigation water (canal excavation and water storage increased surface water coverage by 8%, and the specific heat capacity of water is much higher than that of soil, leading to anomalous cooling) [13]. After completion (1997–2015): The linear negative correlation between LST and VFC was disrupted, with the central region showing a 1.5–2.0 °C increase in LST while VFC increased by 25%. This is mainly caused by changes in crop phenology: the popularization of plastic film mulching (coverage rate 78% in 2015) increased surface albedo by 12%, reducing evaporative cooling, and the shift from grain crops to cash crops (e.g., tomatoes and peppers) shortened the growing period, leading to mismatched peaks of VFC and LST [30,50].

The contradiction between the significant increase in LST in the central part and the increase in FVC at the initial stage of operation further indicates that modern agricultural intensification (with an increase in the proportion of facility agriculture and the use of plastic film mulching) has disrupted the traditional negative correlation between LST and FVC by altering the surface albedo and turbulent fluxes. In the later stages of completion, the spatial pattern reversed: simultaneous cooling in the central part and warming in the eastern part was observed, revealing the long-term effect of water resource redistribution under the operation of the project: in the central irrigation area, due to the improvement of field engineering facilities and the promotion of water-saving technologies, the actual irrigation volume decreased and the evapotranspiration cooling weakened; in the eastern Urad Front Banner, as the end beneficiary area of the project, the water level rise promoted the recovery of wetland vegetation, while the increase in VFC was accompanied by an increase in LST, reflecting the unique thermal response mechanism of the wetland ecosystem (dominated by the water body heat storage effect). It is worth noting that the synchronous phenomenon of LST decrease and FVC reduction in the western Dengkou area in the later stage of completion may be related to the reduction of water diversion and supplementary irrigation from the Yellow River in the edge area of the Ulanbuhesh Desert, resulting in a decrease in surface aridity and an increase in nocturnal radiative cooling, leading to an increase in VFC. Overall, the phased reversal and spatial heterogeneity of the LST-FVC relationship not only indicate that irrigation and drainage projects differentially regulate the surface energy balance through three paths—irrigation thermal effect, transformation of planting structure, and redistribution of hydrological conditions—but also highlight that when evaluating the ecological effects of projects in arid and semi-arid regions, it is necessary to separate the vegetation greenness signal and focus on the deep mechanism of the transformation of land cover functions [50,51,52].

4.3. Discussion on Water Resources Carrying Capacity and WEF Nexus Optimization

The spatiotemporal evolution of the agricultural–ecological structure observed in this study reflects the shifting pressures on the region’s environmental limits. To rigorously interpret these shifts, it is necessary to consider the Water Resources Carrying Capacity (WRCC). Recent frameworks, such as the dynamic multi-criteria analysis by Keyvanfar et al. [53], highlight how hydro-anthropic pressures can push water systems into “critical” or “overloaded” states. Aligning our spatial findings with quantitative capacity assessments and considering the driving mechanisms of the ecological–geological capacity identified by Dai et al. [54] would provide a deeper understanding of the system’s sustainability. Furthermore, future planning should leverage machine learning-based prediction models of ecological carrying capacity, as demonstrated by Zhang et al. [55], to anticipate the long-term impacts of land use transitions.

Effective governance of irrigation and drainage projects requires moving beyond simple land use planning to an integrated Water–Energy–Food (WEF) Nexus management strategy. The “hub regulation model” proposed here should be operationalized using multi-objective optimization tools. For instance, Ashkevari et al. [56] developed a simulation–optimization framework for irrigation networks that successfully balances water scarcity reduction with agricultural profit and greenhouse gas mitigation. Adopting such scenarios, alongside energy performance metrics (EPMI), as suggested by Salman et al. [57], and regional hydro-economic optimization models [58], enables a more comprehensive assessment of project sustainability throughout its life cycle. In the Hetao Plain, the model can be optimized by (1) allocating 60–70% of water diversion to grain production (ensuring food security), 15–20% to ecological water supplement (maintaining VFC threshold), and 10–15% to urban use; (2) reducing energy consumption by 8–10% through water-saving technologies (e.g., drip irrigation) to balance energy input and agricultural output; and (3) establishing a dynamic adjustment mechanism based on annual precipitation and WRCC assessment results.

4.4. Research Gap and Limitation Regarding Temporal Scope

This study focuses on the period 1985–2015, covering the entire life cycle of the core irrigation and drainage projects in the Hetao Plain (construction phase: 1985–1997; operational phase: 1997–2009; post-operation optimization phase: 2009–2015). A notable research gap is the lack of data from 2015 to the present (nearly 10 years). However, this temporal gap has limited impact on the core conclusions of the study for the following reasons: First, the research targets the ecological and spatial effects driven by irrigation and drainage projects, and the main modernization and renovation phases of the projects (1961–2015) have been fully covered, including the key stages of surface disturbance, ecosystem adaptation, and functional optimization. After 2015, the project entered a stable operation period, with no large-scale reconstruction or functional adjustment, and the spatial structure and ecological thermal environment have tended to be stable, so the missing data are unlikely to alter the identified evolution path of “short-term intense disturbance–long-term stable optimization”. Second, the core mechanisms revealed in this study (e.g., the spatiotemporal misalignment of LST-VFC; the coordinated evolution pattern of urban–agricultural–ecological spaces) are derived from the long-term response of the ecosystem to engineering-driven water resource redistribution, which has universal characteristics in arid regions and is not significantly affected by the short-term (nearly 10 years) stable operation stage.

Nevertheless, the lack of post-2015 data still limits the analysis of long-term stability and potential new changes (e.g., the impact of climate change intensification or new water-saving policies on the project’s ecological effects). Future research should supplement multi-source data (e.g., high-resolution remote sensing images and hydrological monitoring data) from 2015 to the present, further verify the sustainability of the “hub regulation model” integrated with WEF Nexus, and explore the adaptive adjustment mechanism of the urban–agricultural–ecological spatial structure under new environmental constraints.

5. Conclusions

Based on multitemporal remote sensing data from 1985 to 2015, this study systematically revealed the spatiotemporal differentiation patterns and regulatory mechanisms underlying the evolution of the urban–agricultural–ecological spatial structure and the ecological thermal environment response driven by irrigation and drainage projects in the Inner Mongolia Hetao Plain. The key conclusions are as follows:

(1)

The spatial structure evolution follows the path of “short-term intense disturbance–long-term stable optimization”. Urban space expands with a shift from external encroachment to internal filling; agricultural space stability increases by 4.8%, and the ecological core area retention rate exceeds 90%, realizing “stable grain yield with unchanged cultivated land area and improved ecological quality with controlled green space loss”.

(2)

Urban, agricultural, and ecological spaces exhibit a “northwest–southeast” consistent pattern of coordinated evolution, with ecological space shifting from independent western distribution to filling the gaps between urban and agricultural spaces, forming a complementary integration pattern.

(3)

Vegetation coverage shows spatial differentiation and stage fluctuations: central area stable increase (annual growth rate 0.8%), eastern area fluctuating recovery (cyclic amplitude ±12%), and western area local improvement (key patches increased by 18%), confirming the ecological effect trajectory of “interference–reconstruction–optimization”.

(4)

The LST-VFC relationship presents spatiotemporal misalignment: irrigation water thermal inertia causes anomalous cooling during construction, while crop phenology changes (plastic film mulching and planting structure adjustment) disrupt the linear correlation after completion. The “hub regulation model” integrated with WEF Nexus provides a practical solution for balanced development.

Based on the research findings, the following policy recommendations are proposed: (1) implement zoning management of water resources: allocate 60–70% of water diversion to the central agricultural core area to ensure grain production, and reserve 15–20% as ecological water for the eastern wetland and western desert, maintaining the VFC at 60% and 30%, respectively; (2) promote ecological compensation for irrigation areas: establish a “water-saving benefit transfer” mechanism, where water-saving gains from agricultural areas are used to fund ecological restoration in the western desert (e.g., artificial grass planting) and eastern wetland (e.g., water level regulation); and (3) optimize agricultural practices: reduce the use of plastic film in high-temperature regions (central Linhe District) and promote biodegradable film, reducing the impact of surface albedo on the LST-VFC relationship.

Future research can strengthen the coupling analysis of “water quantity–land use–ecological effect” by integrating hydrological monitoring data and ecological model simulation and expand the research to other large irrigation areas in arid regions to verify the replicability of the research paradigm.