Spatio-Temporal Dynamics and Contributing Factors of Irrigation Water Use in the Loess Plateau

Abstract

The “Grain for Green” policy has led to a reduction in cultivated land area in the Loess Plateau, intensifying the conflict between ecological conservation and food security. As a key strategy to mitigate this tension, irrigated farmland has undergone significant changes in both its spatial extent and water consumption, which may further exacerbate the water crisis. Hence, the spatio-temporal dynamics and driving forces behind these changes require greater attention and have not yet been comprehensively explored. This study integrates multi-source datasets and employs piecewise linear regression and the Logarithmic Mean Divisia Index (LMDI) model to analyze the spatio-temporal evolution of cultivated land and irrigation water use. Furthermore, it quantifies the contributions of key factors such as cultivated land area, irrigation intensity, and crop planting structure to irrigation water dynamics. The results show that (1) The total cultivated land area in the Loess Plateau decreased by 12.4% from 1985 to 2020, with increases primarily concentrated along the Yellow River between Hekou and Longmen, while decreases were predominantly observed around major cities such as Xi’an, Taiyuan, and Yuncheng. Conversely, the irrigated area exhibited an overall upward trend, with minor declines occurring between 1977 and 1985. (2) While the total irrigation water use increased overall, piecewise linear regression analysis identified four distinct phases, with the first three phases showing growth, followed by a decline after 2001. (3) The expansion of agricultural irrigation areas emerged as the primary driver of increased irrigation water use, whereas advancements in irrigation efficiency effectively reduced water consumption. This study provides novel insights into the spatio-temporal dynamics of irrigation water use in the Loess Plateau and offers valuable guidance for optimizing water resource management and advancing sustainable development in the region.

1. Introduction

Water resources are fundamental to agricultural production [1,2], especially in arid and semi-arid regions, where they underpin food security and ecosystem stability [3,4]. The spatio-temporal dynamics of irrigation water use directly affect regional water resource allocation, agricultural productivity, and ecological protection [5]. Understanding these dynamics is essential for addressing water scarcity and ensuring sustainable development in these vulnerable regions. However, the complexities of water use in arid zones, driven by factors such as climate variability, agricultural expansion, and technological advancements, pose significant challenges to effective management [6,7]. Therefore, systematically analyzing the spatio-temporal evolution of irrigation water use in arid and semi-arid regions is crucial for improving water use efficiency and optimizing water management strategies [3].

Current studies on water use and its driving factors have predominantly focused on urban areas or specific river basins, with limited emphasis on agricultural irrigation in rural and ecologically fragile regions. For instance, Deng et al. analyzed the temporal and spatial evolution of water supply and consumption structures in Zhangjiakou, identifying population growth, urban expansion, and ecological changes as primary drivers [8]. At the national scale, Chen et al. used principal component analysis and regression models to investigate irrigation water use, finding that crop planting scale, surface water use ratio, and planting structure significantly contributed to changes in irrigation water use, while technological advancements reduced water consumption [9]. Similarly, Zhao et al. utilized an extended Stochastic Impacts by Regression on Population, Affluence, and Technology (STIRPAT) model to analyze agricultural water use in China, highlighting the roles of population growth, urbanization, and dietary changes as key drivers [10]. However, existing models, such as regression-based and STIRPAT models, often fail to fully decompose the complex interactions among key drivers of irrigation water use. The Logarithmic Mean Divisia Index (LMDI) model, initially developed for energy studies, has proven effective for analyzing irrigation water use, as it eliminates unexplained residuals during factor decomposition [11,12]. Zhao et al. identified the driving forces of domestic water consumption changes in Shandong province through LMDI model, and found that economic effect and technological effect respectively promoted and hindered the change of domestic water consumption in Shandong Province [13]. Despite these advancements, most research remains centered on urbanized areas and river basins, with insufficient attention given to rural and agriculturally dominated regions, especially those in ecological transition zones. Irrigation water use is shaped by a combination of economic, technological, and environmental factors, but their interactions and dynamic changes remain underexplored.

The Loess Plateau, located in the Yellow River Basin in China, is a typical ecologically fragile region and a key area for China’s national ecological security [14,15,16]. Historically, the region has suffered from severe soil erosion and ecological crises due to intensive agricultural activities [17,18]. These activities have led to widespread soil erosion, land desertification, and other ecological crises [19,20,21]. Since the 1990s, China has implemented major ecological restoration projects in the Loess Plateau, such as small watershed soil conservation and the “Grain for Green” policy, converting sloping farmlands into forests and grasslands. These measures have significantly improved vegetation cover, reduced soil erosion, and generated notable economic, ecological, and social benefits [22,23,24,25]. However, the unbalanced layout of ecological restoration efforts has also affected local water resource sustainability and food security. The combined pressures of ecological restoration and social-economic development have further exacerbated the conflicts between water resources, food production, and ecological protection in the region [26]. Irrigation water use in the Loess Plateau plays a pivotal role in ensuring food security, managing water resources, conserving ecosystems, and supporting regional development. It is a critical sector for achieving regional sustainable development. Notably, the “Grain for Green” policy has led to a continuous reduction in cultivated land area in the Loess Plateau, while the region’s total grain production has paradoxically increased. This counterintuitive trend is closely related to the spatio-temporal changes in both cultivated land and irrigation water use, which have not received sufficient attention in previous studies. Furthermore, the implementation of the Yellow River Basin Ecological Protection and High-Quality Development Strategy has heightened the urgency for sustainable agricultural development in the Loess Plateau. Therefore, elucidating the spatial-temporal dynamics of cultivated land and irrigation water use, and identifying the primary drivers of irrigation water changes in different periods, is of paramount importance for optimizing reforestation strategies, improving water resource management, and alleviating the food–water–ecosystem conflicts in the Loess Plateau.

To bridge these research gaps, this study systematically examines the spatio-temporal evolution of irrigation water use in the Loess Plateau from 1965 to 2013 and identifies its key driving factors. By integrating multi-source datasets, including remote sensing imagery and statistical records, and employing piecewise linear regression and the Logarithmic Mean Divisia Index (LMDI) model, this study aims to (1) map the spatial and temporal distribution of cultivated land; (2) analyze the changes in irrigation water use across different regions and time periods; and (3) quantify the contributions of key driving factors, such as cultivated land area, irrigation intensity, and crop planting structure, to irrigation water use changes. This research could provide crucial insights into the dynamics of irrigation water use, offering practical guidance for optimizing water resource management and fostering sustainable development in the Loess Plateau.

2. Materials and Methods

2.1. Overview of the Study Area

The Loess Plateau is located in the middle reaches of the Yellow River and spans 35 prefecture-level cities across the provinces of Shanxi, Shaanxi, Henan, Gansu, Inner Mongolia, and Ningxia. As shown in Figure 1, the region has a semi-arid to semi-humid climate, with annual precipitation ranging from 50 to 900 mm [27], which exhibits significant temporal and spatial variability [27,28]. The terrain is complex, characterized by a mix of ravines, mountains, valleys, and basins, and severe soil erosion. The ecological environment in the region is extremely fragile. The population density in the Loess Plateau reaches 167 people per square kilometer [29], and human activities are diverse and intense, exerting substantial pressure on the ecosystem [14]. Currently, as the region continues to develop economically and socially, water use for socio-economic activities is increasing, leading to ecological degradation, imbalances between water supply and demand, and water pollution [30,31,32]. Furthermore, the water use efficiency in the Loess Plateau is only about 0.4 [32], and water wastage is severe, further exacerbating the human–water conflict in the region.

2.2. Data Sources

The data used in this study include water resource data and land use data. Water resource data from 1965 to 2013, including city-level irrigation water use (IRR, 1965–2013)—which quantifies annual agricultural withdrawals, including losses during conveyance and field application (excluding pasture/aquaculture uses)—were compiled from national water resource surveys, provincial annual water bulletins covering the Loess Plateau, and FENG’s research data [33]. Land use data, spanning 1985 to 2020, were derived from the China Land Cover Dataset (CLCD) [34], which classifies land cover into nine categories: farmland, forest, shrubs, grassland, water, ice/snow, bare land, impervious surfaces, and wetlands. The dataset has a spatial resolution of 30 m and an overall accuracy of 80%. Irrigated area data combined historical records from Feng’s research (1965–2013) with governmental agricultural statistics (2000–2019) [33], ensuring temporal continuity.

2.3. Research Methods

2.3.1. Research Framework

Based on the collected datasets, we compiled long-term time series for the entire study area and individual prefecture-level cities and analyzed the spatio-temporal dynamics of both total cultivated land and irrigated farmland (Figure 2). A piecewise linear regression method was applied to identify and delineate distinct historical stages of irrigation water use evolution. Furthermore, the Logarithmic Mean Divisia Index (LMDI) model was employed to quantify the contribution of key driving factors—namely irrigated area, crop structure, and irrigation water use intensity—to changes in total irrigation water use. Based on these results, region-specific land and water management strategies were proposed for the Loess Plateau.

2.3.2. Piecewise Linear Regression

Piecewise linear regression is effective for analyzing long-term data trends, addressing the limitation of the inability of simple linear regression to capture changes in trends over long periods. This method identifies significant turning points in a time series by fitting different linear models to consecutive segments of the data. Each turning point (breakpoint) represents a change in the slope of irrigation water use, corresponding to possible policy shifts, technological changes, or land-use transitions.

The method allows the dependent variable IRR (irrigation water use) to be modeled as a set of linear functions over time, segmented by identified turning years ${\mathrm{Y}}_{\mathrm{i}}$. The model can be expressed as follows:

$$IRR=\left\{\begin{array}{ll}{\alpha }_0+{\alpha }_{1}t+{\epsilon }_t,& t\le Y_1\\ {\alpha }_0+{\displaystyle \sum _{i=1}^k}{\alpha }_{i+1}(t-Y_i)+{\epsilon }_t,& Y_i<t\le Y_{i+1}\\ \dots & \\ {\alpha }_0+{\displaystyle \sum _{i=1}^{k+1}}{\alpha }_{i+1}(t-Y_i)+{\epsilon }_t,& t>Y_k\end{array}\right.$$

(1)

where ${\mathsf{\alpha }}_0$ is the intercept, ${\mathsf{\alpha }}_{\mathrm{i}}$ represents the slope for each segment, ${\mathrm{Y}}_{\mathrm{i}}$ denotes the turning point (year), and ${\mathsf{\epsilon }}_{\mathrm{t}}$ is the residual error.

The least-squares error method was used to determine the best-fitting model by minimizing the residual sum of squares. Turning points were chosen based on the following: 1. minimizing the sum of squared residuals across segments; 2. statistical significance of slope changes between segments, tested using a t-test under the null hypothesis that slope differences are not significant.

2.3.3. Index Selection and LMDI Model

This paper incorporates the expansion of irrigated areas, changes in crop planting structure, and irrigation intensity as driving factors in the model to analyze the driving forces behind agricultural irrigation water use during different periods.

In this study, we selected irrigated area, irrigation intensity (i.e., water use per unit area), and crop planting structure as the driving factors for the LMDI decomposition model. These three indicators directly reflect actual irrigation demand and changes in resource allocation during agricultural production. Specifically, irrigated area captures farmland expansion, irrigation intensity indicates the efficiency of water use, and crop structure reflects the impact of cropping system adjustments on water demand. Moreover, these indicators offer consistent and accessible data over time and have been widely adopted in previous LMDI-based studies on irrigation or energy consumption [9,12]. Although climatic factors such as precipitation may influence irrigation demand, studies have shown that in the arid and semi-arid regions of the Loess Plateau, precipitation during the crop growing season accounts for only a small fraction of total crop water requirements and is generally insufficient to significantly reduce irrigation water use [27,35]. Therefore, excluding precipitation from the LMDI decomposition framework in this study is considered both practical and justified.

The LMDI model decomposes irrigation water use changes into three key factors: irrigated area expansion, changes in crop planting structure, and irrigation intensity. The calculation formulas are as follows:

$$△W{U}_{IRR}=\hspace{0.17em}△W{U}_a +△W{U}_c +△W{U}_u$$

(2)

$$△W{U}_a=\sum _j{W}_j\mathit{ln}{\displaystyle \frac{a^t}{a^{t-1}}}$$

(3)

$$△W{U}_c=\sum _j{W}_j\mathit{ln}{\displaystyle \frac{c^t}{c^{t-1}}}$$

(4)

$$△W{U}_u=\sum _j{W}_j\mathit{ln}{\displaystyle \frac{u^t}{u^{t-1}}}$$

(5)

$$W_j={\displaystyle \frac{W{U}_j^t-W{U}_j^{t-1}}{\mathit{ln}(W{U}_j^t)-\mathit{ln}(W{U}_j^{t-1})}}$$

(6)

3. Results and Analysis

3.1. Spatial-Temporal Changes in Cultivated Land Distribution

From a quantitative perspective, the total cultivated land area in the study region decreased by 32,604.33 km² from 1985 to 2020, accounting for 12.37% of the cultivated land area in 1985. However, this overall reduction resulted from the combined effects of various types of land conversion (Figure 3). Spatially, cultivated land declined by 78,925.64 km² (29.96%) between 1985 and 2020, with 15.95% attributed to urban expansion and 84.05% to the Grain for Green program. The cultivated land lost to urban expansion was primarily concentrated around the built-up areas of major cities within the study region, whereas the land abandoned due to the Grain for Green program was more scattered, mainly distributed in the Mu Us Desert and mountainous areas with steep slopes. Additionally, during this period, newly added cultivated land amounted to 46,321.31 km², accounting for 20.06% of the total cultivated land in 2020. These newly expanded farmlands were mainly distributed around pre-existing cultivated areas, particularly within large-scale irrigation districts, where farmland expansion occurred outward from the original cultivated land.

Given that most of the newly expanded farmland was concentrated in major irrigation districts, we further analyzed the spatiotemporal distribution of irrigated farmland (Figure 4). From 1965 to 2019, the irrigated farmland area in the study region increased from 22,219.03 km² in 1965 to 50,505.50 km² in 2019, with an average annual growth rate of 523.82 km² (Figure 4a). The temporal evolution of irrigated farmland can be divided into four distinct phases, with 1970, 1977, and 1985 identified as critical transition points: (1) from 1965 to 1970, the irrigated area slightly increased; (2) from 1970 to 1977, the irrigated area showed a significant increase, reaching a peak of 30,263 km² in 1977; (3) from 1977 to 1985, the irrigated area slightly decreased, reaching a low of 28,515 km² in 1985; (4) after 1985, the irrigated area fluctuated upward, reaching a peak of 42,443 km² in 2018. The spatial distribution of newly established irrigated farmland is presented in Figure 4b. A comparison with the overall farmland expansion pattern (Figure 3) indicates that a substantial proportion of newly cultivated land in recent decades has been allocated to irrigated agriculture. This suggests that irrigation development has played a pivotal role in farmland expansion within the study region.

3.2. Spatial-Temporal Changes in Irrigation Water Use

Temporally, from 1965 to 2013, irrigation water use in the middle reaches of the Yellow River exhibited an overall trend of initial increase followed by a subsequent decline (Figure 5). Piecewise linear regression identified four distinct phases in the evolution of irrigation water use: 1965–1977, 1977–1987, 1987–2001, and 2001–2013. During 1965–1977, agricultural irrigation water use increased rapidly, rising from 14.18 km³ yr⁻¹ to 19.05 km³ yr⁻¹. During 1977–1987, a decline in irrigation water use was followed by a stabilization period, with irrigation water use reaching a minimum of 18.40 km³ yr⁻¹. Between 1987 and 2001, irrigation water use exhibited a fluctuating upward trend, reaching a peak of 24.96 km³ yr⁻¹ in 2000. However, after 2001, irrigation water use showed a declining trend with fluctuations.

We also quantified the variations in agricultural irrigation water use across 35 prefecture-level cities for each phase, as illustrated in Figure 6. The results suggest that agricultural irrigation water use in the Loess Plateau exhibited a general increasing trend from 1965 to 2013, though with considerable regional variations (Figure 6). From 1965 to 1977, irrigation water use increased to varying extents across all cities, with Weinan, Xi’an, and Xianyang in the southern part of the plateau exhibiting the most rapid growth. From 1977 to 1987, irrigation water use declined in most regions, with notable variations among cities. Agricultural water use increased substantially in Ningxia, particularly in Wuzhong, whereas in northern Inner Mongolia, cities such as Linhe exhibited pronounced declines. From 1987 to 2001, agricultural water use in Ningxia either slowed or declined, whereas cities such as Linhe and Dongsheng in Inner Mongolia experienced rapid increases. From 2001 to 2013, agricultural water use exhibited a general declining trend, with Linhe, Wuzhong, and Dongsheng undergoing substantial reductions, whereas Weinan and Yuncheng in the south recorded slight increases. Overall, irrigation water use in small and medium-sized cities remained relatively stable, whereas larger cities such as Linhe and Wuzhong demonstrated more pronounced fluctuations.

3.3. Analysis of the Driving Factors Behind Changes in Irrigation Water Use

The LMDI model was applied to analyze the driving forces behind changes in irrigation water use, focusing on factors such as irrigated area expansion, changes in crop mix, and irrigation intensity, as shown in Figure 7 and Figure 8 and

Table 1. The driving factors of agricultural irrigation water use cause changes in the amount of water used for agricultural irrigation.

Year/Unit (km3 yr−1)	Irrigation Water	Irrigation Area Expansion	Crop Mix	Water Use Intensity per Unit Area
1965~1977	4.62	6.15	0.14	0.82
1977~1987	0.45	−0.54	0.42	2.54
1987~2001	5.02	3.42	0.27	−5.02
2001~2013	−2.86	2.06	−0.68	−6.63

. The results indicate that the change in agricultural irrigation water use in the middle reaches of the Yellow River ranged from −4 to 2 ${\mathrm{k}\mathrm{m}}^3{\mathrm{y}\mathrm{r}}^{-1}$ during the entire study period. From 1965 to 1977, irrigation water use increased year by year, primarily driven by the expansion of irrigated areas, with a smaller contribution from changes in crop mix (0.14 ${\mathrm{k}\mathrm{m}}^3{\mathrm{y}\mathrm{r}}^{-1}$) and a more significant contribution from irrigation intensity (0.82 ${\mathrm{k}\mathrm{m}}^3{\mathrm{y}\mathrm{r}}^{-1}$). From 1977 to 1987, irrigation water use remained relatively stable, with irrigation intensity playing a positive role, while the effects of irrigated area expansion and changes in crop mix were weaker. Between 1987 and 2001, irrigation water use increased by 5.02 ${\mathrm{k}\mathrm{m}}^3{\mathrm{y}\mathrm{r}}^{-1}$, with irrigated area expansion being the main driving force. After 2001, irrigation water use decreased year by year, with irrigated area expansion continuing to promote increased water use, while changes in crop mix and irrigation intensity exerted inhibitory effects, reducing water use by 6.63 ${\mathrm{k}\mathrm{m}}^3{\mathrm{y}\mathrm{r}}^{-1}$ due to improvements in irrigation efficiency.

Figure 9 reveals clear spatio-temporal heterogeneity in the dominant driving factors of irrigation water use across four stages. From 1965 to 1977, irrigated area expansion dominated most regions, while Ningxia was primarily influenced by increased irrigation intensity, and parts of the Hetao Plain were shaped by changes in crop structure. Between 1977 and 1987, reductions in irrigated areas dominated the Hetao region, and declining irrigation intensity was prevalent in several cities in Shanxi and Shaanxi. From 1987 to 2001, irrigated area expansion again became the main driver in the northwest, while irrigation intensity was the dominant factor in the Weihe River Basin and surrounding areas. During 2001–2013, declining irrigation intensity emerged as the most widespread driver, particularly across the Hetao Plain and the northern Mu Us Sandy Land, while other factors showed a more scattered spatial distribution.

4. Discussion

4.1. Characteristics of Irrigation Water Use Evolution in the Loess Plateau

This study analyzed the overall trends in cultivated land in the Loess Plateau and revealed a clear spatial heterogeneity in land use changes. In the area from the Yellow River’s estuary to Longmen (along the Shanxi–Shaanxi border), cultivated land gradually transformed into natural land, while in urban areas such as Xi’an, Taiyuan, Yinchuan, and Hohhot, cultivated land was converted into impervious surfaces (Figure 3). In the Hetao Plain, Ningxia Plain, and Longzhong Plateau, cultivated land increased. These changes are closely related to regional development strategies. For example, in areas with significant slope farmland, such as the region from the estuary to Longmen, ecological restoration measures prompted large-scale reforestation [36]. Meanwhile, urban expansion driven by the Western Development Strategy inevitably occupied some cultivated land [37]. The expansion of irrigated farmland was concentrated around traditional irrigation zones, suggesting that economic development and advancements in irrigation technology may have played positive driving roles (Figure 4).

Regarding the spatial-temporal evolution of irrigation water use, this study found varying rates of change across different periods (Figure 5), with significant spatial differentiation (Figure 6) and diverse driving factors (Figure 7, Figure 8 and Figure 9). The trends highlight the dynamic nature of irrigation water use over time, reflecting the underlying combined influence of agricultural expansion, water management policies, and hydrological variability in the region. From 1965 to 1977, irrigation water use changed uniformly, mainly driven by the expansion of irrigated areas. From 1977 to 1987, overall water use remained relatively stable, with agricultural water use in cities like Ningxia increasing rapidly, while northern cities experienced a reduction. This change was closely related to soil and water conservation measures and afforestation projects [36]. Irrigation intensity increased significantly in agricultural areas due to enhanced water reuse [36]. From 1987 to 2001, irrigation water use increased as irrigated areas expanded in major agricultural zones, promoting further water use. However, despite an increase in irrigated land, irrigation intensity also rose, driving higher yields and income for farmers without improving water efficiency [38,39]. Since 2001, irrigation water use has declined, despite the continued expansion of irrigated areas in major agricultural zones like the Hetao and Ningxia Plains. The Hetao and Ningxia Plains continued to see expansion in irrigated areas, yet improvements in irrigation technology effectively reduced water use. This corroborates KANG’s findings on water-saving agriculture [40]. Similar studies in other arid regions, such as the North China Plain and countries in the Middle East and North Africa, have reported comparable trends, where irrigation efficiency improvements led to reduced water use despite increasing crop yields [7,41]. The reduction in water use can be attributed to the “Grain for Green” project and changes in agricultural production methods [25]. As ecological conditions improved and production factors shifted, agricultural production gradually transitioned from extensive to intensive management [42]. Additionally, supporting measures, such as farmland water conservation, rural energy development, and ecological migration, improved agricultural production capacity and increased farmers’ incomes [43,44]. Research has also shown that adjustments in agricultural structure have contributed to farmers’ income growth [45], though changes in crop planting structure had a limited impact on irrigation water use, as traditional practices of extensive farming still dominated due to low economic returns. Consequently, structural adjustments were relatively slow.

In addition to the main structural drivers—irrigated area, irrigation intensity, and crop structure—irrigation water use is also influenced by a range of complex factors, including climatic variability, market forces, and institutional mechanisms. Previous studies have suggested that precipitation variability and temperature anomalies may affect irrigation demand and crop growth in arid regions [46]. However, in the Loess Plateau, the observed warming trend over recent decades has been relatively mild, and the interannual variation trend of precipitation is not significant, occurring mostly outside the critical crop growth stages [35]. Despite slight increases in rainfall [47], irrigation water use has continued to rise, indicating that precipitation plays a limited role in alleviating irrigation demand in this region.

Market-driven cropping shifts, such as from grain to cash crops, can intensify water use when high-value crops are prioritized [48]. Moreover, institutional factors—such as water allocation policies and inter-basin transfer schemes in the Yellow River Basin—substantially affect local water availability [49,50], with marked disparities in irrigation infrastructure and governance across sub-regions [51]. These complex and region-specific drivers warrant further integration into future research and tailored policy recommendations.

4.2. Land Management Measures

In the context of China’s shift from rapid economic growth to high-quality development and the consolidation phase of ecological governance, it is essential to devise diversified land management measures based on land use structure [16,49]. These measures should aim to adjust agricultural structure while maintaining ecological balance, modernize facilities and technologies, and achieve coordinated ecological and economic development [49,52]. Specifically, agricultural activities should focus on controlling the scale of cultivated land while optimizing water supply and maintaining food production capacity. This includes modernizing agricultural infrastructure, updating irrigation technologies, promoting water-saving agriculture, and improving water use efficiency while enhancing resilience to climate change. In areas with concentrated urban and agricultural land, protecting ecological source regions is crucial. Furthermore, there is a need to establish and improve water-saving policies, regulations, and standards, raise public awareness of water conservation, appropriately adjust agricultural water pricing, and, if necessary, consider water transfers from outside the region [42,46]. Additionally, optimizing the configuration of forests and grasslands, delineating red lines for protecting ecologically fragile areas, and establishing water source conservation zones can help reduce human disturbances and balance ecological protection with socio-economic development [25,36]. By adopting an integrated approach, managing the demand for water production, and ensuring that water use aligns with regional water supply capacity, scientifically sound policy measures can be developed and implemented to support the sustainable development of the Loess Plateau and improve the well-being of its residents [49].

In addition, the strong regional differentiation in dominant driving factors highlights the need for spatially targeted land and water management strategies. In regions where declining irrigation intensity was the primary driver—such as Ningxia and the northern Mu Us Sandy Land—further promotion of high-efficiency irrigation technologies and crop-water structure optimization is recommended to consolidate existing water-saving outcomes. In contrast, areas where irrigated area expansion persists, such as Ordos and the Hetao irrigation zone, should prioritize stricter irrigation access controls and proactive water use regulations to prevent the accumulation of pressure on regional water resources due to excessive expansion.

4.3. Research Limitations and Outlook

Due to the limited time span and spatial resolution of available data, this study primarily focuses on city-level analyses, which may obscure finer spatial differences. To further reveal the dynamic changes in irrigation water use, future research plans include extending the study period, improving data accuracy, and combining the results of water use structure and driving factor analyses to propose more scientifically sound water resource management policies and measures, promoting the sustainable use of water resources.

Different plant species exhibit varying water use characteristics, which may significantly influence irrigation demand. For example, Wei et al. found that native species such as Quercus liaotungensis demonstrated more stable radial growth and lower water use sensitivity to drought compared to introduced species like Robinia pseudoacacia in the Yangjuangou catchment of the Loess Plateau [53]. This suggests that species replacement, often driven by ecological restoration policies, may alter regional water balances. Future research should consider the hydrological impacts of vegetation structure and species composition when evaluating land use transitions and ecosystem recovery in arid and semi-arid regions.

5. Conclusions

Using the LMDI model, this study analyzed the spatial-temporal trends and driving factors behind agricultural irrigation water use in 35 prefecture-level cities in the Loess Plateau from 1965 to 2013. The key findings are as follows:

Cultivated land area in the Loess Plateau decreased, while irrigated area increased. A significant portion of the cultivated land was converted to natural land or impervious surfaces, with natural land conversion concentrated in areas along the Yellow River, and impervious surface conversion mainly occurring around urban areas. Some land was also converted to irrigated farmland, particularly in the Hetao Plain, Ningxia Plain, and Longzhong Plateau.

Temporally, irrigation water use exhibited significant fluctuations from 1965 to 2013, with a general trend of first increasing and then decreasing, particularly after 2001. Spatially, irrigation water use in the Loess Plateau showed an overall upward trend, though with substantial regional differences. Larger cities like Linhe and Wuzhong experienced more pronounced fluctuations, while smaller cities exhibited relatively stable water use.

The study identified irrigated area expansion as the primary driver of increased agricultural water use in the Yellow River’s middle reaches, while changes in crop planting structure played a secondary role. Over time, irrigation intensity contributed increasingly to the reduction of water use, mainly due to improvements in irrigation technology.