Analysis of the Spatiotemporal Variation Characteristics and Driving Forces of Crops in the Yellow River Basin from 2000 to 2023

Abstract

In the context of global climate change and growing food security challenges, this study provides a comprehensive analysis of the yields of three staple crops (wheat, corn and rice) in the Yellow River Basin of China, employing multiple quantitative analysis methods including the Mann–Kendall trend test, center of gravity transfer model and hotspot analysis. Our research integrates yield data covering these three crops from 72 prefecture-level cities across the Yellow River Basin, during 2000 to 2023, to systematically examine the temporal variation, spatial variation and spatial agglomeration characteristics of the yields. The study uses GeoDetector to explore the impacts of natural and socioeconomic factors on changes in crop yields from both single-factor and interactive-factor perspectives. While traditional statistical methods often struggle to simultaneously handle complex causal relationships among multiple factors, particularly in effectively distinguishing between direct and indirect influence paths or accounting for the transmission effects of factors through mediating variables, this study adopts Structural Equation Modeling (SEM) to identify which factors directly affect crop yields and which exert indirect effects through other factors. This approach enables us to elucidate the path relationships and underlying mechanisms governing crop yields, thereby revealing the direct and indirect influences among multiple factors. This study conducted an analysis using Structural Equation Modeling (SEM), classifying the intensity of influence based on the absolute value of the impact factor (with >0.3 defined as “strong”, 0.1–0.3 as “moderate” and <0.1 as “weak”), and distinguishing the nature of influence by the positive or negative value (positive values indicate promotion, negative values indicate inhibition). The results show that among natural factors, temperature has a moderate promoting effect on wheat (0.21) and a moderate inhibiting effect on corn (−0.25); precipitation has a moderate inhibiting effect on wheat (−0.28) and a moderate promoting effect on rice (0.17); DEM has a strong inhibiting effect on wheat (−0.33) and corn (−0.58), and a strong promoting effect on rice (0.38); slope has a moderate inhibiting effect on wheat (−0.15) and a moderate promoting effect on corn (0.15). Among socioeconomic factors, GDP has a weak promoting effect on wheat (0.01) and a moderate inhibiting effect on rice (−0.20), while the impact of population is relatively small. In terms of indirect effects, slope indirectly inhibits wheat (−0.051, weak) and promotes corn (0.149, moderate) through its influence on temperature; DEM indirectly promotes rice (0.236, moderate) through its influence on GDP and precipitation. In terms of interaction effects, the synergy between precipitation and temperature has the highest explanatory power for wheat and rice, while the synergy between DEM and precipitation has the strongest explanatory power for corn. The study further analyzes the mechanisms of direct and indirect interactions among various factors and finds that there are significant temporal and spatial differences in crop yields in the Yellow River Basin, with natural factors playing a leading role and socioeconomic factors showing dynamic regulatory effects. These findings provide valuable insights for sustainable agricultural development and food security policy-making in the region.

1. Introduction

Escalating global climate change and anthropogenic activities have exerted a significant impact on agricultural production, making food security a paramount global concern [1]. The increasing frequency and intensity of extreme weather events, particularly droughts and floods, have substantially compromised crop productivity and yield stability [2]. Concurrently, burgeoning population growth has intensified pressure on food supply chains [3], requiring agricultural systems to simultaneously address quantitative sufficiency and qualitative safety requirements [4]. Against this backdrop, in-depth research on the spatiotemporal variation patterns and driving mechanisms of crop yields at the regional scale is of great practical significance for optimizing agricultural production layouts, improving comprehensive grain production capacity and achieving sustainable agricultural development [5].

Research on the spatiotemporal variation in crops can not only reveal dynamic trends in crop yields over time but also provide deeper insights into their spatial distribution characteristics. By integrating long-term series data with spatial analysis techniques, it is possible to effectively quantify the combined influence of natural and socioeconomic factors on crop yields. Consequently, consensus has gradually emerged regarding the importance of food security for human sustainable development. Scholars have extensively studied and explored the combined effects of interactions between natural and socioeconomic factors on crop yields and the relationship between spatiotemporal dimensions and crop yields. However, most studies focus exclusively on a single temporal or spatial dimension, limiting a comprehensive understanding of the patterns governing changes in crop yields. Han Tianfu et al. utilized nationwide farmland monitoring data from the Ministry of Agriculture and Rural Affairs spanning nearly 30 years (1988–2019), focusing primarily on the long-term temporal trends of wheat and corn yields [6]. Similarly, Arata Linda et al. analyzed yield trends over time using 8088 national-level crop yield sequences from the FAO database [7]. On the spatial scale, Li Jianqin et al. emphasized spatial differences in sericulture production among eastern, central and western regions, focusing primarily on static regional comparisons [8]. Similarly, Porwollik Vera et al. also focused on spatial dimensions, revealing spatial differences in crop yields at global, national and regional scales through comparisons of multiple models within the Global Gridded Crop Model Intercomparison Project [9]. Furthermore, current research limitations are evident in studies confined to specific provinces or local areas. For example, Liu Ruixuan et al. used only Qihe County in Shandong Province as their research area, determining the optimal planting periods for winter wheat through temperature threshold methods to assist yield assessment and production planning [10]. Likewise, Shirin Mohammadi et al. concentrated on specific regions in Norway, analyzing the relationship between temperature, precipitation, and crop yields at the county scale, providing reference points for developing local agricultural strategies to adapt to climate change [11]. Importantly, existing research exhibits significant limitations in exploring the driving mechanisms behind crop yield variations. While many studies highlight the significance of analyzing the drivers of crop yield spatiotemporal variation due to its practical importance in formulating precise agricultural policies, ensuring food security and promoting agricultural sustainability [12,13,14,15], their analyses frequently focus exclusively on either natural or socioeconomic factors, overlooking the complex interactions among these factors. Analyzing yield and cultivation area data of major crops in China from 1981 to 2010, Guo Liang et al. found that meteorological factors dominate interannual fluctuations in crop yields but did not deeply explore the interactions between natural and socioeconomic factors or the impacts of regional differences [16]. Although Gerssen-Gondelach Sarah et al. addressed some endogenous factors driving future yields [17], they similarly failed to fully elaborate the interaction mechanisms among various factors, nor did they effectively reveal the differential characteristics of multi-factor interactions across different regions.

Extensive research on crop yield has been conducted globally. In the study of spatiotemporal distribution, spatial analysis technologies have been effectively employed to intuitively present the distribution patterns and evolution trends of crops, such as the Mann–Kendall trend test [18], the Getis–Ord Gi* index [19] and Moran’s I [20]. In the analysis of driving forces, traditional statistical methods such as regression analysis [21] and GeoDetector [22] are widely used, but it is difficult to comprehensively handle the complex interrelationships between multiple variables via these methods. As a multivariate statistical method capable of comprehensively analyzing the causal relationships between variables, Structural Equation Modeling (SEM) [23] has shown significant advantages in fields such as chemistry [24], medicine [25] and education [26]. Nevertheless, its application in the research on the driving forces behind crop yields remains to be further explored and expanded.

The Yellow River Basin (YRB), as an important birthplace of Chinese civilization, is also a major agricultural production area in China [27]. Spanning the Qinghai-–Tibet Plateau, the Loess Plateau, and the North China Plain, the basin covers various climate types such as arid, semi-arid, semi-humid [28], with complex natural conditions [29] and a fragile ecological environment [30]. Under the dual influences of global climate change and national agricultural policy adjustments, the agricultural production patterns in the basin have undergone profound changes [31,32]. These changes include the area of traditional food crops shrinking [33]. In this context, the crops in the YRB show complex spatiotemporal distribution patterns, which demand in-depth research.

This study is based on the shortcomings of current research on the spatiotemporal changes in crops, which mostly focuses on a single temporal or spatial dimension, and on the analysis of the driving factors behind yield changes, which mostly focuses on either natural or social factors, lacking a systematic discussion regarding the characteristics of spatiotemporal coupling evolution and the interactions among various factors. Using spatial analysis and Structural Equation Modeling (SEM), this study analyzes yield data for three main crops (wheat, corn and rice) in the basin [34], examining spatiotemporal variation characteristics in the YRB from 2000 to 2023 and revealing the driving mechanisms of natural and socioeconomic factors, and provides a scientific basis and decision-making reference for the sustainable agricultural development in the YRB.

2. Materials and Methods

2.1. Study Area

The Yellow River originates from the Yueguzonglie Basin at the northern foot of the Bayan Har Mountains on the Qinghai–Tibet Plateau. It flows eastward through nine provinces and autonomous regions, namely Qinghai, Sichuan, Gansu, Ningxia, Inner Mongolia, Shaanxi, Shanxi, Henan and Shandong, before finally emptying into the Bohai Sea, with a total basin area of approximately 795,000 square kilometers [35]. The terrain within the basin is complex and diverse [36], including plateaus, mountains, plains and hills. The upper reaches are mainly the Qinghai–Tibet Plateau and the Inner Mongolia Plateau; the middle reaches flow through the Loess Plateau; and the lower reaches form the North China Plain. As an important agricultural production area in China, the upper reaches are dominated by animal husbandry, with some valley agriculture. For example, the Huangshui River Valley in Qinghai, with its relatively low terrain, higher temperature and sufficient water sources, is suitable for growing crops such as wheat. The middle reaches are a typical dryland agricultural area, mainly planting drought-tolerant crops like wheat and corn. The lower reaches have well-developed irrigated agriculture, primarily cultivating wheat, corn and rice, relying on irrigation water from the Yellow River and its tributaries to ensure crop growth [37]. To maintain crop yield data consistency, this study in accordance with the theories proposed by Qin Chenglin et al., this study defines the YRB study area as 72 prefecture-level administrative regions (corresponding to prefecture-level cities, autonomous prefectures or leagues in administrative division) through which the Yellow River flows, following the three principles of “taking the natural scope of the YRB as the foundation, maintaining the integrity of prefecture-level administrative divisions as much as possible and considering the direct correlation between regional social and economic development and the Yellow River” [38], as shown in Figure 1.

2.2. Data Sources and Preprocessing

The crop yield data are derived from the statistical yearbooks of various provinces. We collected the yield data of wheat, corn and rice in 72 regions within the study area for six time points: 2000, 2005, 2010, 2015, 2020 and 2023. Due to data being missing in some statistical yearbooks and changes in administrative divisions, during data processing, the missing data for certain years were replaced with data from adjacent years. Meanwhile, in accordance with the adjustment of administrative divisions, the data of Laiwu City were merged into those of Jinan City. The DEM data are sourced from the Geospatial Data Cloud with a spatial resolution of 90 m, which were then resampled to 1 km using the ArcGIS 10.8 software platform. The precipitation and temperature data come from the China Meteorological Data Network, with a spatial resolution of 1 km. The data for the six time points from 2000 to 2023 were obtained using the Kriging interpolation method. The GDP and population data are derived from the China Statistical Yearbook. After Kriging interpolation, data with a spatial resolution of 1 km and a temporal resolution of 1 year were obtained. Detailed data sources are shown in

Table 1. Research data and their sources.

Data Type	Data Source	Website and Access Time
Crop yield data	Provincial Statistical Yearbooks	E.g., https://tjj.henan.gov.cn (10 January 2025)
DEM	Geospatial Data Cloud	https://www.gscloud.cn (25 February 2025)
Temperature Precipitation	China Meteorological Data Network	https://data.cma.cn (25 February 2025)
GDP Population	China Statistical Yearbook	https://www.stats.gov.cn (25 February 2025)

. The slope and aspect data were derived from the original DEM data and calculated using the ArcGIS 10.8 software platform. Finally, projections were performed, and the raster data of different influencing factors were extracted by mask according to the scope of the study area.

2.3. Research Methods

2.3.1. Mann–Kendall Trend Test

The Mann–Kendall test method was proposed by H.B. Mann and M.G. Kendall. Initially used to detect the trend of sequence changes, it has since been improved to test trends and mutation times [39]. As a non-parametric test method, it has the advantages of not requiring a specific distribution of samples, being resistant to the interference of outliers and having simple calculations. It can reveal nonlinear change patterns in data and is widely used in the time series analysis of elements with non-normal distribution data in meteorology, hydrology and other fields [40]. In this paper, the Mann–Kendall trend test method is used to conduct a significance test on the change trend results of the three crop yield data at six time points from 2000 to 2023, and its formula is as follows:

$$S={\sum }_{i=1}^{n-1}{\sum }_{j=i+1}^{n}sgn(x_j-x_i)$$

(1)

$$VAR\left(S\right)={\displaystyle \frac{n(n-1)(2n+5)}{18}}$$

(2)

$$Z=\left\{\begin{array}{cc}{\displaystyle \frac{S-1}{\sqrt{VAR\left(S\right)}}}& S>0\\ 0& S=0\\ {\displaystyle \frac{S+1}{\sqrt{VAR\left(S\right)}}}& S<0\end{array}\right.$$

(3)

Among them, in Formulas (1)–(3), for a long time series $X$($x_1$,$x_2$,$x_3$, …,$x_n$), $S$ is the Mann–Kendall trend test statistic, $x_i$ and $x_j$ are the $i$-th and $j$-th data values in the time series, respectively, $n$ is the number of data points, and $sgn\left(x\right)$ is a sign function, such that when $x$ > 0, $sgn\left(x\right)=1$; when $x$ = 0, $sgn\left(x\right)=0$; when $x$ < 0, $sgn\left(x\right)=-$1.

2.3.2. Center of Gravity Transfer Model

The center of gravity refers to a point within the scope of a gravitational field through which the resultant force of gravity acting on the mass points of all components of an object must pass, regardless of the object’s orientation [41]. Based on a spatial coordinate system, this model regards the distribution of the research object as a set of points with a certain mass, where the mass of each point is determined by its corresponding attribute value. The position of the center of gravity of the attribute distribution is obtained by calculating the weighted average coordinates of these points. In this paper, this model is used to analyze the center of gravity of crop yield data in the YRB and to plot the migration trajectory, thereby analyzing the spatial variation characteristics of the yield center of gravity in the YRB [42]. Its formula is as follows:

$$X={\displaystyle \frac{\sum _{i=1}^n{M}_i{x}_i}{\sum _{i=1}^n{M}_i}}$$

(4)

$$Y={\displaystyle \frac{\sum _{i=1}^n{M}_i{y}_i}{\sum _{i=1}^n{M}_i}}$$

(5)

Among them, in Formulas (4) and (5), ($X$,$Y$) represents the coordinates of the crop yield gravity center at a certain time point in the study area, ($x_i,y_i$) denotes the geometric center coordinate value of the $i$-th calculated planar spatial unit, and $M_i$ stands for the attribute value of the $i$-th calculated planar spatial unit, i.e., the crop yield value.

2.3.3. Hotspot Analysis (Getis–Ord G_i*)

Hotspot analysis is a spatial statistical method used to identify regions with clusters of high or low values (i.e., hotspots and coldspots) in spatial data [43]. It determines whether a spatial unit belongs to a hotspot or coldspot by calculating the sum of attribute values of each spatial unit and its surrounding adjacent units and comparing it with the global average [44]. In this paper, this analytical method is employed to conduct cluster analysis on crop yields and its calculation formula is as follows:

$$G_i^{*}(d)={\displaystyle \frac{\sum _{j=1}^n{w}_{ij}\left(d\right)x_j-\overline{X}\sum _{j=1}^n{w}_{ij}\left(d\right)}{\sqrt{\frac{\sum _{j=1}^n{x}_j^2}{n}-{\overline{X}}^2}\sqrt{\frac{\sum _{j=1}^n{w}_{ij}^2\left(d\right)-{\left(\sum _{j=1}^n{w}_{ij}\left(d\right)\right)}^2}{n-1}}}}$$

(6)

Among them, in Formula (6), $x_j$ is the attribute value of region $j$, $\overline{X}$ is the average value of all regional observation values, $w_{ij}\left(d\right)$ is the distance-based (distance d) spatial weight matrix, which represents the spatial relationship between region $i$ and region $j$, and $n$ is the total number of spatial units.

When the $G_i^{*}$ value is positive, it indicates that the observed values in the region and its surrounding adjacent areas are relatively high, belonging to a hotspot area, i.e., a cluster of high values. When the $G_i^{*}$ value is negative, it means that the observed values in the region and its surrounding adjacent areas are relatively low, belonging to a coldspot area, i.e., a cluster of low values. When the $G_i^{*}$ value is close to 0, it suggests that there is no obvious clustering of observed values in the region and its surrounding adjacent areas, showing a random distribution [45].

2.3.4. GeoDetector

GeoDetector is a spatial analysis method used to detect spatial heterogeneity and reveal the driving factors behind it, which is widely applied to explore the influencing factors of various phenomena [46]. The core idea is that if an independent variable has a significant impact on a dependent variable, their spatial distributions should be similar. This method can effectively analyze spatial heterogeneity and has two notable advantages: it can handle both numerical and qualitative data detection and it can accurately detect the interactive effects of two factors on the dependent variable [47]. In this paper, GeoDetector is used to analyze the driving forces affecting the spatiotemporal evolution of crop yields. The formula is as follows:

$$q=1-{\displaystyle \frac{\sum _{h=1}^L{N}_h{\sigma }_h^2}{N{\sigma }^2}}$$

(7)

Among them, in Formula (7), q is the detection factor and its value range is [0, 1]; N is the sample size of the study area and $N_h$ is the sample size of the $h$-th stratum; ${\sigma }^2$ is the variance of Y in the whole region; ${\sigma }_h^2$ is the variance of Y in the $h$-th stratum; and L is the number of strata. The larger the q value, the stronger the explanatory power of the factor on the changes in crop yield.

2.3.5. Structural Equation Modeling (SEM)

Structural Equation Modeling (SEM) originated in the field of social sciences. It constructs causal relationships between variables and modifies the model according to fitting indicators, ultimately achieving the effect of integrating the connections between variables [48]. As a multivariate statistical method and multivariate analysis equation based on covariance matrices, this model can handle multiple variables simultaneously and test the goodness of fit, and has been applied in many fields [49]. In the stage of theoretical model setting for Structural Equation Modeling (SEM), based on the research hypothesis that “natural factors and socioeconomic factors jointly drive changes in crop yields”, the causal relationships between variables are clearly depicted through path diagrams. Among them, natural variables include temperature, precipitation, DEM, slope and aspect; social variables include GDP and population; and the core dependent variables are the yields of three crops: wheat, corn and rice. The measurement model is used to define the corresponding relationships between latent variables and observed variables, while the structural model constructs the influence paths of latent variables on crop yields, including the direct effects of natural factors and socioeconomic factors as well as the interaction effects between them. After the model is set, an adaptability test is conducted, which requires checking that the p-VALUE > 0.05 and the standardized root mean square residual (SRMR) < 0.08 to comprehensively judge the fit between the model and the data. If the fit does not meet the standards, the model is revised (such as deleting observed variables with low factor loadings or adding theoretically reasonable variable associations), and the adaptability test is re-conducted, forming an iterative “revision-test” process until all indicators meet the threshold requirements. This paper applies it in the field of geography to reveal the direct and indirect influencing factors of natural and social variables on crop yields, and its calculation formulas are as follows:

$$x={\Lambda }_x\xi +\delta$$

(8)

$$y={\Lambda }_y\eta +\epsilon$$

(9)

$$\eta =\beta +\Gamma \xi +\zeta$$

(10)

Among them, Formulas (8) and (9) are the measurement model formulas and Formula (10) is the structural model formula. $x$ represents the exogenous observed variables and $y$ denotes the endogenous observed variables; $\xi$ stands for the exogenous latent variables and $\eta$ indicates the endogenous latent variables; $\delta$ is the measurement error of $x$ and $\epsilon$ is the measurement error of $y$; ${\Lambda }_x$ represents the relationship between $x$ and $\xi$; ${\Lambda }_y$ denotes the relationship between $y$ and $\eta$; $\beta$ is the relationship between endogenous latent variables; $\Gamma$ is the impact of $\xi$ on $\eta$; and $\zeta$ is the part of the model that cannot be explained.

3. Results and Analysis

3.1. Spatiotemporal Patterns of Crop Yields in the YRB

3.1.1. Temporal Variation Characteristics

Using the wheat yield data of various regions in the YRB for the years 2000, 2005, 2010, 2015, 2020 and 2023, the Mann–Kendall trend test method was employed to analyze the results from the temporal dimension, as shown in Figure 2.

Significant increases are mainly concentrated in the lower reaches of the YRB, particularly Liaocheng City in Shandong Province and Shangqiu City in Henan Province. This trend results from the synergistic effects of multiple driving factors. The promotion of water-saving irrigation technologies and mechanized operations have jointly driven the continuous increase in yields [50]. Regions with slightly significant decreases are concentrated in the agricultural core areas and industrial and mining concentrated zones in the middle and upper reaches of the YRB, including Zhongwei City, Ningxia Hui Autonomous Region, Yulin City and Tongchuan City in Shaanxi Province and Changzhi City in Shanxi Province. This area suffers from severe soil erosion and intensified soil salinization and the dual pressures of industrial pollution and water resource competition. Reduced agricultural irrigation quotas have contributed to a decrease in wheat acreage, thereby affecting its yield [51]. Regions with extremely significant decreases are mainly distributed in the ecologically fragile zones and alpine arid areas of the YRB, such as Xining City in Qinghai Province, Alxa League in Inner Mongolia Autonomous Region and Gannan Tibetan Autonomous Prefecture in Gansu Province. The cold and arid climate in this area, coupled with the expansion of soil desertification, has resulted in a reduction in wheat yield.

The significant changes in corn yields in the YRB show distinct geographical differentiation characteristics, as shown in Figure 3. From the perspective of spatial distribution, regions with slightly significant increases are concentrated in the hilly areas from the upper to the middle reaches, including Tianshui City in Gansu Province. These areas have a relatively arid climate and traditional agriculture is limited by soil erosion and insufficient irrigation. However, breakthroughs have been achieved through policy support and technological innovation [52]. Regions with extremely significant increases are mainly distributed in the middle and lower reaches and the Hetao Plain, such as Hohhot City in Inner Mongolia Autonomous Region, Kaifeng City in Henan Province and Heze City in Shandong Province. These regions have flat terrain and contiguous cultivated land, coupled with stable irrigation from the main stream and tributaries of the Yellow River, resulting in a significant increase in their yields. The only region with an extremely significant decrease is Shangluo City in Shaanxi Province, which is located in the southeast of Shaanxi Province. The terrain of this city is dominated by mountains and hills, with scattered cultivated land resources and high ecological sensitivity, leading to a substantial reduction in corn yields.

Region changes in the YRB exhibit a zonal distribution, as shown in Figure 4. Regions with slightly significant increases are mainly distributed in the middle reaches of the Yellow River and its tributary areas, including Shizuishan City in Ningxia Hui Autonomous Region, Qingyang City in Gansu Province and Weinan City in Shaanxi Province. These areas are mostly located in the interleaving zone of the Loess Plateau and river valleys, with both potential for water and soil management and local climate stability. Regions with extremely significant increases are concentrated in the middle reaches of the Yellow River, represented by Linfen City in Shanxi Province and Wuhai City in Inner Mongolia Autonomous Region. This is due to the promotion of drought-tolerant varieties and water-saving technologies, which have further improved rice productivity. Regions with slightly significant decreases include Luoyang City, Xinxiang City in Henan Province, and Dezhou City and Zibo City in Shandong Province. These cities are mainly located in the alluvial plain of the middle and lower reaches of the Yellow River, where agricultural intensification is high but they are facing the squeeze of urbanization. Pronounced decreases are more widely distributed, such as Baoji City in Shaanxi Province, Zhengzhou City in Henan Province and Weifang City in Shandong Province. Most of these areas are economic core zones or ecologically fragile zones, affected by both urbanization expansion and resource constraints, which ultimately constrain their yields [53].

3.1.2. Spatial Variation Characteristics

In 2000, the center of wheat production in the YRB was located in Huguan County, Changzhi City. In 2005, it had migrated approximately 25 km southeast to the southern part of Linzhou City, Anyang City. In 2010, it continued to move about 28 km southeast to the northern part of Qi County, Hebi City. In 2015, it further shifted around 16 km southeast to the southern part of Qibin District, Hebi City. In 2020, the center changed direction and moved roughly 15 km northeast to the northwest of Xun County, Hebi City. By 2023, it migrated about 6 km southwest and finally arrived in the western part of Xun County. The overall migration path shows a zigzag trend of “first southeast, then northeast and finally southwest” with a significant total migration range, as shown in Figure 5. This migration is closely related to the promotion of water-saving irrigation technologies and high-yield varieties in the lower reaches [54]. Terrain and climate are the fundamental restrictive factors. The lower plain has a lower altitude, so the annual average temperature is higher than that of the upper plateau, which drives the wheat production center to migrate to the lower reaches with better heat conditions.

In 2000, the center of corn production was located in the northwest of Lucheng District, Changzhi City, Shanxi Province. In 2005, it had migrated approximately 23 km northwest to the northeast of Xiangyuan County, Changzhi City. In 2010, it retreated about 11 km southeast to the west of Licheng County, Changzhi City. In 2015, it turned northwest again, moving roughly 32 km back to the northwest of Xiangyuan County. In 2020, it made a slight adjustment of around 14 km northeast to the border area between Xiangyuan County and Wuxiang County. By 2023, it migrated about 13 km northwest and finally stabilized in the southwest of Wuxiang County. This forms a cyclically fluctuating path of “northwest–southeast–northwest–northeast–northwest”, as shown in Figure 6. The migration trajectory reflects that the Hetao Plain has become a core area of sustained high yields due to the stable irrigation from the Yellow River. The Loess Plateau in the middle reaches, through terrace transformation and the promotion of drought-tolerant varieties [55], has reduced yield volatility, attracting the center to shift westward periodically. However, the arid areas in the upper reaches have long maintained low yields due to water resource constraints and ecological policy restrictions. The improvement of agricultural mechanization and policy support have promoted yield growth in the middle reaches, leading to the dynamic adjustment of the production center in the transition zone between irrigated and dryland agriculture.

In 2000, the center of rice production in the YRB was located at the junction of Tunliu District and Zhangzi County, Changzhi City, Shanxi Province. In 2005, it had migrated approximately 82 km southeast of Huguan County, Changzhi City. In 2010, it moved about 146 km northwest to the border area between Gu County and Anze County, Linfen City. In 2015, the center continued to advance westward by around 18 km to the junction of Gu County and Huozhou City in the northern part of Linfen City. In 2020, it shifted 67 km southeast to the border area between Qinyuan County and Qin County, Changzhi City, Shanxi Province. By 2023, it had migrated 56 km southeast and finally returned to the northeast of Tunliu District. The overall path shows a phased fluctuation of “southeast–northwest–west–west–southeast”, as shown in Figure 7. This phenomenon is closely related to the urbanization expansion in the middle and lower reaches and the improvement of water conservancy facilities in the upper reaches. In the middle and lower reaches, due to the loss of cultivated land and competition for water resources, rice planting areas have shifted to the upper reaches of tributaries, leading to the westward migration of the production center. In the high-altitude areas of the upper reaches, with the promotion of drought-tolerant varieties and water-saving technologies, short-term yield hotspots have been formed. However, due to the limitation of heat conditions, the center eventually returned to the warm and humid irrigation areas in the middle and lower reaches. The water-dependent characteristics of rice make its spatial distribution significantly dependent on water sources and the areas with improved irrigation systems in the tributaries of the Yellow River in the middle and lower reaches have always been the core areas of rice production.

3.1.3. Spatial Clustering Characteristics

The changes in the spatial clustering of wheat yields in the YRB from 2000 to 2023 are shown in Figure 8: From 2000 to 2005, the agglomeration of low-yield areas in some plateau regions of the upper reaches significantly intensified, while the coldspot characteristics in the semi-arid areas of the middle reaches weakened. From 2005 to 2015, no new significant agglomeration areas emerged in the basin, the original pattern remained stable and regional differences tended to balance out. From 2015 to 2020, the high-yield agglomeration in Qingdao City receded into a random distribution. From 2020 to 2023, the overall pattern continued the previous trend with slight local adjustments. In general, the hotspots of wheat yields in the YRB have presented a spatial evolution pattern characterized by “stable high yields in the lower reaches, improved yields in the middle reaches, and intensified low yields in the upper reaches”. The core high-yield areas have remained stable, and the low-yield situation in the marginal semi-arid areas has been alleviated to some extent with policy and technical support. However, due to topographical and water resource constraints, the agglomeration of low yields in some plateau and hilly regions has persisted for a long time.

The changes in the spatial clustering of corn yields in the YRB from 2000 to 2023 are shown in Figure 9: From 2000 to 2010, the agglomeration of yields in some plateau areas increased, while the irrigated plains in the lower reaches maintained their status as core high-yield areas, with a stable pattern. In 2015, the high-yield agglomeration in the lower plains remained generally unchanged, with a small extension to the inland flat areas; during the same period, some plateau regions showed obvious low-yield agglomeration. From 2015 to 2020, some coldspot areas turned into non-significant ones, and the yields in marginal counties improved slightly. From 2020 to 2023, the overall pattern continued, with the upper reaches still maintaining low-yield agglomeration, and the level of local coldspots weakened slightly. In general, the corn yields in the YRB have evolved in a pattern characterized by “stable high yields in the lower reaches, slow improvement in the middle reaches, and persistent low yields in the upper reaches”. The lower plains have maintained their high-yield advantages relying on irrigation and intensive farming. With the support of water conservancy and variety improvement, the coldspot areas in some hilly transition zones have shown a weakening trend. However, due to topographical and climatic constraints, the agglomeration of low yields in plateau regions has persisted for a long time.

As shown in Figure 10, the changes in the spatial clustering of rice yields in the YRB from 2000 to 2023 are as follows: From 2000 to 2005, the high-value agglomeration in some cities shifted to a random distribution, and Zhengzhou began to show high-value agglomeration. From 2005 to 2010, high-value agglomeration formed in some cities in the northwest, while the agglomeration in Shandong Province and Henan Province weakened. From 2010 to 2015, there were no significant changes. From 2015 to 2020, the high-value agglomeration in Weifang and Binzhou strengthened, while the agglomeration in Hebi and Anyang decreased. From 2020 to 2023, the overall pattern tended to be stable; cities such as Weifang maintained high-value agglomeration, and some cities in the northwest returned to a random distribution. In general, the high-value agglomeration areas are mainly located in the middle and lower reaches of the basin, covering parts of Shandong Province, Henan Province and Shanxi Province, where the yields remain high all year round. Other regions are relatively scattered and show a random distribution, reflecting the characteristics of spatial dynamic evolution.

3.2. Analysis of Driving Forces for Spatiotemporal Changes in Crop Yields in the YRB

3.2.1. Single-Factor Analysis

As can be seen from Figure 11, among all the factors affecting wheat yield, the temperature factor shows the highest impact weight, reaching a peak in 2020 with an impact factor of 0.62. From 2000 to 2023, it shows a trend of first rising, then falling, then rising again and then falling, indicating the continuous impact of temperature. The precipitation factor had a relatively high impact weight in 2010, with an impact factor of 0.34, but it dropped sharply to 0.10 in 2005 and then remained at a relatively stable but low level from 2015 to 2023. The impact weight of the GDP factor reached a peak in 2000, with an impact factor of 0.38. It changed significantly from 2010 to 2020, especially in 2020, when it was close to the peak again with an impact factor of 0.37 and then decreased by about 0.9 in 2023, with an impact factor of 0.28, which reflects the dynamic impact of economic development on wheat yield. The impact weight of the population factor reached a peak in 2005, with an impact factor of 0.47, then continued to decline until 2015, dropping to a minimum of 0.38, and slightly rebounded from 2020 to 2023, which is essentially related to the acceleration of urbanization. The impact weight of the DEM factor reached a peak in 2000, with an impact factor of 0.56. There was no significant change from 2000 to 2005 and there was a periodic fluctuation of “decline–rise” from 2010 to 2023. The impact weights of the slope factor and aspect factor are always low, which indicates that the slope and aspect factors have a weak impact on the temporal and spatial changes in wheat yield.

As can be seen from Figure 12, the temperature factor has the highest impact weight on corn yield, reaching a peak in 2015 with an impact factor of 0.57, which confirms the sensitivity of corn, as a thermophilic crop, to heat conditions. The impact weight of the precipitation factor was relatively high at 0.19 in 2000, but it remained low after 2005 and tended to be stable after 2015, with the impact factor fluctuating around 0.13, reflecting that water resource management measures in the YRB have effectively reduced the dependence of corn production on natural precipitation. Although the impact weight of the DEM factor is lower than that of temperature, it showed an upward trend from 2005 to 2010 and a sharp decline occurred from 2010 to 2023, dropping from 0.63 to 0.59 and finally stabilizing at around 0.46. This indicates that the constraints on corn yield in high-altitude areas due to insufficient accumulated temperature and soil degradation have gradually become prominent with climate change. The impact weights of the GDP and population factors reached peaks in 2010 and 2000, respectively, with impact factors of 0.35 and 0.31 and then decreased with the improvement of agricultural mechanization level, indicating that the driving effect of social and economic factors on corn yield is phased. The impact weights of the slope and aspect factors have always been weak, indicating that topographic relief has little impact on large-scale corn cultivation.

As can be seen from Figure 13, the impact weight of the DEM factor on the spatiotemporal changes in rice yield showed a significant increase from 2000 to 2010, indicating that the influence of altitude on the spatial distribution of crops strengthened during this stage. It dropped sharply to 0.23 in 2015 and remained stable from 2015 to 2023, with the impact factor fluctuating around 0.23. The impact weight of the temperature factor did not change significantly from 2000 to 2010, fluctuating around 0.33, while it decreased significantly to 0.29 in 2015. By 2023, the impact weight of the temperature factor showed a continuous increase, reflecting the key role of temperature conditions in crop growth. However, the decrease in the impact weight of the temperature factor also indicates that the optimization of comprehensive irrigation and drainage has reduced the impact of temperature on crops. The impact weight of the precipitation factor was the highest in 2000, with an impact factor of 0.17, but it dropped significantly to the lowest value of 0.08 in 2005 compared with the previous period. By 2010, the weight of the precipitation factor rose to 0.12 and then fluctuated around this value, indicating that the importance of precipitation fluctuations in alleviating drought and ensuring stable crop yields is increasing in recent years. The GDP and population factors showed an upward trend from 2000 to 2010, reflecting that rural economic growth and labor scale had a positive promoting effect on yield improvement in the early stage. However, from 2015 to 2023, they both showed a decline to varying degrees compared with the previous period, indicating that with the improvement of agricultural modernization and mechanization, the explanatory power of regional economic aggregate and population scale gradually weakened. The impact weights of the slope and aspect factors were always less than 0.02 and the overall change was gentle, indicating that the slope and aspect around the YRB had limited impact on the spatiotemporal changes in crops.

3.2.2. Interaction Factor Analysis

In the analysis of interactive factors affecting wheat yield in the YRB from 2000 to 2023, as shown in Figure 14, in 2000, the four groups with high explanatory power were Precipitation ∩ Temperature (0.65027) > DEM ∩ Precipitation (0.64226) > DEM ∩ Temperature (0.61713) > Population ∩ Temperature (0.59512). In 2005, the four groups with high explanatory power were Precipitation ∩ Temperature (0.66114) > DEM ∩ Precipitation (0.64839) > Population ∩ Temperature (0.62496) > DEM ∩ Temperature (0.62315). In 2010, the four groups with high explanatory power were Precipitation ∩ Temperature (0.69445) > DEM ∩ Precipitation (0.64612) > DEM ∩ Temperature (0.61662) > Population ∩ Temperature (0.57531). In 2015, the four groups with high explanatory power were Precipitation ∩ Temperature (0.63075) > DEM ∩ Temperature (0.58725) > DEM ∩ Precipitation (0.58011) > Population ∩ Temperature (0.55671). In 2020, the four groups with high explanatory power were Precipitation ∩ Temperature (0.69253) > DEM ∩ Precipitation (0.6806) > DEM ∩ Temperature (0.66295) > GDP ∩ Temperature (0.6506). In 2023, the four groups with high explanatory power were Precipitation ∩ Temperature (0.61653) > DEM ∩ Temperature (0.60321) > Population ∩ Temperature (0.59137) > DEM ∩ Precipitation (0.58378). From 2000 to 2023, the interaction between precipitation and temperature has always been the primary factor affecting wheat yield, with explanatory power fluctuating between 0.61653 and 0.69445, highlighting the core role of the synergistic cooperation of hydrothermal conditions in wheat growth. The explanatory power of the interactions between DEM and precipitation and between DEM and temperature, has long been among the top, reflecting that topography indirectly affects yield by regulating climate. The explanatory power of the interaction between population and temperature increased in the early stage (2000–2005) and decreased in the later stage (2005–2015), which is related to the fact that agricultural technology alleviated climate fluctuations in the early stage but the effect stabilized in the later stage [56]. In 2020, the interaction between GDP and temperature entered the group with high explanatory power, indicating that economic development had an increasing impact on yield. Overall, natural factors continue to dominate changes in wheat yield, while the influence of human activity factors has gradually formed a complex synergistic relationship with natural factors amid fluctuations, jointly driving the dynamic evolution of wheat yield.

In the analysis of interactive factors affecting corn yield in the YRB from 2000 to 2023, as shown in Figure 15, the mechanism of corn yield affected by the interaction of multiple factors has obvious dynamic evolution characteristics and presents certain phased and structural features. In 2000, the four groups with high explanatory power were DEM ∩ Precipitation (0.69084) > Precipitation ∩ Temperature (0.66727) > DEM ∩ Temperature (0.62782) > DEM ∩ Population (0.5795). In 2005, the four groups with high explanatory power were DEM ∩ Precipitation (0.57399) > DEM ∩ Temperature (0.56904) > DEM ∩ Population (0.54528) > Precipitation ∩ Temperature (0.54381). In 2010, the four groups with high explanatory power were DEM ∩ Precipitation (0.74090) > Precipitation ∩ Temperature (0.67926) > DEM ∩ GDP (0.66034) > DEM ∩ Temperature (0.65468). In 2015, the four groups with high explanatory power were DEM ∩ Precipitation (0.65981) > Precipitation ∩ Temperature (0.65718) > DEM ∩ Temperature (0.63779) > DEM ∩ Population (0.60288). In 2020, the four groups with high explanatory power were DEM ∩ Precipitation (0.63522) > Precipitation ∩ Temperature (0.61057) > DEM ∩ Temperature (0.55380) > DEM ∩ Population (0.48151). In 2023, the four groups with high explanatory power were DEM ∩ Precipitation (0.65112) > Precipitation ∩ Temperature (0.61750) > DEM ∩ Temperature (0.56472) > DEM ∩ Population (0.48547). From 2000 to 2023, the interaction between DEM and precipitation has always been the primary factor affecting corn yield, with its explanatory power fluctuating between 0.57399 and 0.74090, reflecting that the role of topography in regulating the spatial distribution of moisture conditions is significant, thereby affecting corn yield. The interaction between precipitation and temperature has long maintained a high level of explanatory power (0.54381–0.67926), highlighting the importance of the synergistic effect of hydrothermal conditions on the growth and development of corn. The explanatory power of the interaction between DEM and temperature is relatively stable and remains at the forefront for a long time, reflecting that the combined effect of topography and climate continues to be evident. The interaction between population and DEM was more prominent in the early stage but gradually weakened in the later stage, a trend that may be related to the phased changes in population structure and labor input in agricultural production. In addition, the interaction between DEM and GDP jumped to a relatively high level in 2010, indicating that economic factors have gradually become one of the important driving forces for the growth of corn yield. Overall, the interaction of natural factors has always dominated the dynamic changes in corn yield, while economic and social factors have shown a trend of phased enhancement. The interaction mechanism between human activities and natural factors has become more complex and in-depth, jointly promoting the spatial pattern and temporal changes in corn yield.

In the analysis of interactive factors affecting rice yield in the YRB from 2000 to 2023, as shown in Figure 16, in 2000, the four groups with high explanatory power were Precipitation ∩ Temperature (0.53127) > DEM ∩ Precipitation (0.40335) > DEM ∩ Temperature (0.35694) > GDP ∩ Temperature (0.34320). In 2005, the four groups with high explanatory power were Precipitation ∩ Temperature (0.48271) > DEM ∩ Precipitation (0.39097) > GDP ∩ Temperature (0.37499) > Population ∩ Temperature (0.36559). In 2010, the four groups with high explanatory power were Precipitation ∩ Temperature (0.49630) > DEM ∩ Precipitation (0.39071) > GDP ∩ Temperature (0.36797) > DEM ∩ Temperature (0.36448). In 2015, the four groups with high explanatory power were Precipitation ∩ Temperature (0.43823) > DEM ∩ Precipitation (0.36662) > DEM ∩ Temperature (0.31032) > GDP ∩ Temperature (0.30610). In 2020, the four groups with high explanatory power were Precipitation ∩ Temperature (0.51793) > DEM ∩ Precipitation (0.40802) > GDP ∩ Temperature (0.32946) > DEM ∩ Temperature (0.32522). In 2023, the four groups with high explanatory power were Precipitation ∩ Temperature (0.51336) > DEM ∩ Precipitation (0.38627) > GDP ∩ Temperature (0.34567) > Population ∩ Temperature (0.33111). From 2000 to 2023, the interaction between precipitation and temperature remained the dominant factor affecting rice yield, with its explanatory power fluctuating between 0.43823 and 0.53127, highlighting the importance of hydrothermal coordination for rice growth and development. The explanatory power of the interaction between DEM and precipitation has long maintained the second-highest level, indicating that topographic conditions play a significant role in regulating the spatial distribution of water. The explanatory power of the interaction between DEM and temperature generally showed a downward trend, which is related to the fact that technological progress in major rice-producing areas has gradually weakened the constraints of topographic climate [57]. The interaction between GDP and temperature has remained among the top and gradually increased during the study period, reflecting that the positive role of economic development in rice yield has become increasingly evident. The interaction between population and temperature entered the high explanatory power group in some years, playing a certain regulatory role. Overall, the interaction of natural factors continues to dominate changes in rice yield, but the role of human activity factors has gradually increased, forming a more complex and dynamic synergistic effect with natural factors, jointly promoting the spatiotemporal evolution of rice yield.

3.3. Driving Force Analysis Using Structural Equation Modeling (SEM)

3.3.1. Direct Impact

Structural Equation Modeling (SEM) was used to analyze the driving forces among multiple factors, aiming to explore in depth the direct impacts and interaction mechanisms of temperature, slope, aspect, digital elevation model (DEM), gross domestic product (GDP), precipitation and population on the yields of various crops (output). The degree of direct impact of each variable on different crop yields is reflected by impact factors, where positive and negative signs represent the direction of impact (positive for promotion and negative for inhibition) and the absolute value indicates the intensity of impact. Generally, an absolute value of the impact factor greater than 0.3 is considered a relatively large impact, a value between 0.1 and 0.3 is regarded as a moderate impact and a value less than 0.1 is considered a relatively small impact [58].

As can be seen from the analysis results in Figure 17, the impact factor of DEM on wheat yield is −0.33 and its absolute value is greater than 0.3, which indicates that DEM has an inhibitory effect on wheat yield and the intensity of this inhibitory effect is relatively large. In high-altitude areas, the temperature is relatively low, which affects the growth and development of wheat, leading to a decrease in yield. The impact factors of precipitation and slope are −0.28 and −0.15, respectively, indicating that precipitation and slope also have different inhibitory effects on wheat yield and both inhibitory effects are smaller than that of DEM, belonging to moderate inhibitory effects. Excessive precipitation will cause water logging in the fields, resulting in poor soil permeability and affecting the normal growth of plants. When the slope is relatively large, it is not conducive to wheat planting and field management. Machinery is difficult to operate on steep slopes, leading to low efficiency in sowing, fertilization, harvesting and other aspects and increasing labor and time costs. The impact factors of aspect and population are −0.06 and −0.02, respectively, with relatively small inhibitory effects. This is because wheat has strong adaptability to light and well-developed root systems, and the high degree of mechanization in modern wheat planting and policy/technical support, has weakened the impact of human factors. The impact factors of temperature and GDP are 0.21 and 0.01, respectively, indicating that temperature and GDP can promote wheat yield. Temperature has a moderate promoting effect, while the promoting effect of GDP is relatively weak. Suitable temperature can promote physiological processes such as photosynthesis, thereby directly increasing yield. GDP growth means economic development and an increase in social wealth. The government’s increased investment in agriculture has improved agricultural infrastructure, providing better basic conditions for wheat growth.

The results of analyzing the impact of various factors on corn yield are shown in Figure 18. It can be seen that the impact factor of DEM on corn yield is −0.58. The negative sign indicates that it inhibits corn yield and its absolute value is greater than 0.3, meaning the inhibitory effect is relatively strong. In high-altitude areas, heat is insufficient, which slows down the growth and development of corn. Due to weathering in high-altitude areas, the soil is often relatively barren, with low organic matter content and poor soil structure. In such a soil environment, corn roots are difficult to stretch and take root, and their ability to absorb water and nutrients is limited, failing to provide sufficient nutrients for corn growth, which directly restricts the improvement of corn yield. The impact factors of temperature and precipitation are −0.25 and −0.18, respectively, indicating that temperature and precipitation also have different inhibitory effects on corn yield and these inhibitory effects are moderate. Corn is a thermophilic crop and requires a certain amount of accumulated temperature during its growth period to complete each growth stage. During the growth of corn, when the temperature exceeds its suitable range, the stomata will partially close to reduce water loss. After the stomata close, the photosynthesis rate decreases, and the organic substances produced reduce and ultimately directly affect corn yield. When there is excessive precipitation, corn roots are soaked in water for a long time, leading to root poisoning and rot. It can also cause corn plants to lodge, with leaves covering each other, thus affecting photosynthesis and ultimately seriously affecting the normal growth and yield of corn. The impact factor of aspect is −0.01, with an absolute value less than 0.1, indicating that its inhibitory effect is relatively small. This is because during the growth process of corn, its demand for light and heat can be met to a certain extent under different aspect conditions, and the corn plants themselves have a certain ability to regulate and adapt to the environment. Slope and population have a direct positive impact on corn yield. The impact factor of slope is 0.15, which is between 0.1 and 0.3, indicating a moderate promoting effect. The impact factor of population is 0.01, which means its promoting effect on yield is very slight. A moderate slope is conducive to drainage, which can effectively avoid diseases such as root rot caused by waterlogging in corn fields, create a good growth environment for corn roots and also allow corn plants to receive light more evenly, thereby enhancing photosynthesis and increasing yield. A certain scale of population can provide sufficient labor for corn planting. With the increase in population, the consumer demand for corn will also rise, including as food, feed and industrial raw materials. This market demand stimulates farmers to increase yield, ultimately having a direct positive impact on corn yield.

The results of the impact of various factors on rice yield are shown in Figure 19. GDP and slope have direct negative impacts on rice yield. The impact factor of GDP on rice yield is −0.20; the negative sign indicates that it inhibits rice yield and its absolute value is between 0.1 and 0.3, belonging to a moderate inhibitory effect. When GDP grows, more investment tends to be focused on industry and the service sector, resulting in reduced investment in agricultural infrastructure construction. Rice cultivation relies on sound water conservancy facilities to ensure irrigation and drainage. If water conservancy facilities are aging and in disrepair, they cannot provide sufficient water during the critical water demand period of rice, or cannot drain water in time when there is excessive precipitation, which will affect rice growth. At the same time, agricultural science and technology are crucial for increasing rice yield. Insufficient investment in agriculture due to GDP factors will slow down agricultural scientific research progress, making it difficult to cultivate new rice varieties with high yield, good quality and strong stress resistance. The impact factor of slope on rice yield is −0.02, with an absolute value less than 0.1, indicating a small inhibitory effect on rice yield. As an aquatic crop, rice has high requirements for water demand and water layer stability. It is difficult to maintain a stable water layer in sloped plots and water will flow quickly along the slope, resulting in insufficient water in the upper part of the plot and possible water accumulation in the lower part, thereby affecting rice growth and yield. DEM, precipitation and population have direct positive impacts on rice yield. The impact factor of DEM on rice yield is 0.38; the positive sign indicates that it promotes rice yield and its absolute value is greater than 0.3, with a relatively large impact. In high-altitude areas, the air is thin and the atmosphere weakens solar radiation less, resulting in high light intensity and long light duration. Sufficient light provides a strong driving force for rice photosynthesis, enabling it to efficiently synthesize organic substances and significantly increase yield. The impact factor of precipitation on rice yield is 0.17, with an absolute value between 0.1 and 0.3, showing a moderate promoting effect. Rice needs a large amount of water to maintain physiological activities throughout its growth period. Precipitation can supplement soil moisture in paddy fields, maintain appropriate soil humidity and reduce field temperature in hot summer months to avoid high-temperature damage, thereby increasing yield. The impact factor of population on rice yield is 0.06, indicating a small promoting effect on yield. Although the population can provide labor and stimulate demand, the impact is limited.

3.3.2. Indirect Impact

Based on the analysis results of Figure 20 combined with Figure 17, it can be seen that slope, aspect and precipitation have indirect impacts on wheat yield, with indirect impact factors of −0.051, −0.007 and −0.003, respectively, all showing small inhibitory effects. Among them, the indirect impact of slope is the most significant. It indirectly affects wheat yield by influencing intermediate variables such as precipitation and temperature and the path through temperature is particularly critical, with an indirect impact factor of −0.0441. This indicates that when the slope is larger, the heat exchange between the surface and the atmosphere accelerates, leading to a decrease in temperature, which in turn hinders the normal growth and development of wheat and reduces yield. Overall, natural factors still dominate the indirect impacts, while the mediating effect of social and economic factors is weak. This reflects that topographic and climatic conditions impose multi-level constraints on wheat yield through complex paths, while the indirect regulatory role of social and economic factors is limited.

Based on the analysis results of Figure 21 and Figure 18, it can be seen that slope and DEM have indirect impacts on corn yield, with indirect impact factors of 0.149 and −0.058, respectively. Among them, slope has a moderate promoting effect on corn yield, while DEM has a weak inhibitory effect on yield, showing indirect effects of different intensities. Unlike the indirect negative impact of slope on wheat yield mentioned earlier, slope has an indirect positive impact on corn yield. In the indirect impact path of slope, its effect on corn yield through directly influencing temperature is the most significant, with an indirect impact factor of 0.1175, showing a moderate promoting effect. This indicates that when the slope is larger, the heat exchange between the surface and the atmosphere accelerates, resulting in relatively lower temperatures. This low-temperature environment can, on the one hand, help corn avoid or reduce damage from high temperatures, and on the other hand, it can reduce the damage of diseases and pests to corn plants, thereby ensuring the normal growth of corn and indirectly increasing its yield. Overall, in the indirect impacts of slope and DEM on corn yield, the mediating effect of natural factors dominates, reflecting that topographic and climatic conditions form a multi-level impact mechanism on corn yield through specific paths, while the indirect regulatory role of social and economic factors in this process is relatively limited.

As can be seen from the analysis results of Figure 22 and Figure 19, the indirect impact of DEM on rice yield is relatively significant, with an impact factor of 0.236, which is a moderate promoting effect. In the indirect impact path of DEM, its effect on rice yield through influencing GDP and precipitation is the most prominent. Specifically, the indirect impact factor of DEM on the GDP path is 0.120 and that on the precipitation path is 0.1156, both belonging to moderate impacts. This indicates that high-altitude areas, due to topographic constraints, to a certain extent hinder industrial development and increase development costs, resulting in certain restrictions on GDP growth. However, the indirect benefit of restricted GDP growth is that it reduces the occupation of cultivated land by industrial and urban expansion, effectively protecting the rice planting area. At the same time, the relatively low intensity of human activities in high-altitude areas helps reduce environmental pollution, creating a good ecological environment for rice growth. In addition, the impact of moderate high altitude on GDP will also prompt the government to strengthen agricultural policy support, such as increasing subsidies and promoting advanced agricultural technologies, to assist rice production. From the perspective of precipitation, when warm and humid air currents encounter high-altitude terrain, they will cool and condense due to orographic lifting, thereby increasing precipitation and improving the probability of regional precipitation. Sufficient precipitation can accurately meet the water demand of rice at various growth stages, providing good natural conditions for rice growth and development and ultimately promoting an increase in rice yield.

4. Discussion

4.1. Analysis of Spatiotemporal Variation Patterns and Driving Mechanisms of Crops

This study systematically analyzed the spatiotemporal variation patterns of wheat, corn and rice in the YRB from 2000 to 2023, revealing significant geographical differentiation characteristics of these three crops and providing a key basis for optimizing the agricultural production layout in the basin. The research found that wheat yields in the lower reaches have increased significantly due to the synergistic effects of water-saving irrigation technologies, variety improvement and mechanized operations, while yields in the upper plateau areas have continued to decline under the influence of the policy for farmland conversion to forest. This pattern is highly consistent with the research conclusions of Han et al. [6]. The increase in corn yields in the hilly areas of the upper and middle reaches and the Hetao Plain reflects the positive role of policy support and technological innovation in dryland and irrigated agriculture. However, the upper alpine regions have remained in a state of low yields due to insufficient accumulated temperature and soil desertification, which is consistent with the findings of Pang et al. [28] and Mohammadi et al. [11]. Rice yields show a differentiation pattern of “increase in the middle and upper reaches, decrease in the middle and lower plain areas and economic core zones,” which is closely related to urbanization expansion, water resource competition and ecological policies. These findings clearly demonstrate the differential responses of different crops to natural conditions and socioeconomic factors, providing a scientific reference for targeted adjustments to the planting structure [33,37]. Furthermore, this study deeply analyzed the driving mechanisms behind changes in crop yields, enhancing the understanding of the dynamics of the basin’s agricultural system. Temperature and precipitation, as core natural factors, have shown significant regional differences in their impacts. However, variety improvement and the construction of water conservancy facilities can alleviate climate constraints to a certain extent, reflecting the dynamic interaction between natural factors and technical interventions [12,28,37]. Among socioeconomic factors, GDP and population have distinct regional variations in their effects: in the lower plain areas, agricultural mechanization and technological input have promoted yield increases, while in the middle and upper reaches, industrialization and urbanization have led to farmland fragmentation and even non-agriculturalization, negatively affecting yields [15]. This is consistent with the research conclusions of Li et al. [31] on farmland conversion to forests and changes in cultivated land. Notably, the driving effect of population on yields is not directly reflected in the correlation with the number of agricultural practitioners, but through a “demand-driven supply” mechanism. Specifically, the expansion of food demand and the upgrading of dietary structures caused by population growth have incentivized the agricultural production system to make adaptive adjustments in terms of planting scale, structure and input intensity [59]. This analysis of multi-factor interactions breaks through the limitations of single-factor analysis, providing a multi-dimensional perspective for understanding the “development–conservation” balance dilemma in the basin.

4.2. Cross-Regional Comparison and Innovation in Research Methods

From a cross-regional comparison perspective, the differentiated pattern of crop yields in the YRB shares common characteristics with large river agricultural systems worldwide. For instance, the pattern of “yield increase in economic agglomeration zones and yield decrease in ecologically fragile zones” is similar to the trajectory of the Ganges rice belt shifting toward the Nile Delta under urbanization pressure, as well as the dual agricultural structure in the upper and lower reaches of the Nile Basin [60]. This highlights the common challenges faced by large river agricultural systems globally. This finding not only enriches the theoretical connotation of basin agricultural research but also provides experiences from the YRB for international practices in balancing “development and conservation.” Methodologically, this study enhanced the reliability of driving mechanism analysis through the combined application of GeoDetector and Structural Equation Modeling (SEM). GeoDetector first identifies core influencing factors and variable interaction effects, delimiting the scope for subsequent analysis; Structural Equation Modeling (SEM), on this basis, further distinguishes between direct and indirect impacts and clarifies causal paths. The progressive collaboration of these two methods ensures the comprehensiveness of exploration and strengthens the explanatory power of the conclusions, providing a reference methodological framework for similar regional agricultural research [16,17].

4.3. Practical Value, Prospects and Limitations of the Study

Overall, the findings of this study hold significant practical value. To address the issue of declining wheat yields in the upper reaches, measures such as optimizing the balance between farmland conversion policies and agricultural subsidies and promoting drought-resistant varieties can be adopted. In the midstream corn-producing areas with yield-increasing potential, efforts should be made to strengthen terrace transformation and investment in drought-resistant technologies to consolidate production capacity. In the downstream rice-producing regions, it is necessary to coordinate urbanization and farmland protection, prioritizing the guarantee of irrigation water. These specific suggestions provide practical decision support for improving the comprehensive grain production capacity of the basin and achieving ecological protection and agricultural development and are of great significance for ensuring national food security and promoting sustainable agricultural development in the YRB. From the perspective of research prospects, with the continuous advancement of science and technology, future studies can integrate more advanced remote sensing technologies and big data analysis methods to improve the accuracy and timeliness of dynamic monitoring of crop growth in the YRB, thereby providing a more precise basis for crop yield prediction and adjustment of planting structures. In response to the long-term impact of climate change on crop yields, more in-depth simulation and prediction studies can be conducted to provide forward-looking strategies for addressing food security issues under climate change. However, this study also has certain limitations. Due to data gaps in some cities, adjacent year data were used as substitutes, which may have exerted a certain impact on the accuracy of the research results. Additionally, this study is mainly based on existing statistical data and image data, failing to fully consider some micro-level factors that may also play important roles in actual agricultural production and affect crop yields. Although multiple analytical methods and models were comprehensively applied, there is still room for improvement in these methods.

5. Conclusions

The study found that crop production in the YRB shows significant spatiotemporal differentiation. From the perspective of temporal variation characteristics, wheat yields have formed intensive areas in the lower reaches, with significant growth due to technological upgrading and policy support, while yields in the ecologically fragile areas of the middle and upper reaches have continued to decline under the influence of the conversion of farmland to forests and soil degradation; corn yields in the large-scale planting areas of the Hetao Plain have increased significantly due to variety improvement and mechanization promotion, while in the arid areas of the upper reaches, yields have continued to decrease due to water resource constraints and ecological policy restrictions; rice yields in the middle reaches of the Yellow River and tributary areas have increased significantly due to the potential of water and soil management, improved water conservancy facilities and the promotion of drought-resistant varieties and water-saving technologies, while yields in the downstream alluvial plains and economic core areas have decreased due to high agricultural intensification but facing urbanization pressure and resource constraints. In this study, we conducted a systematic analysis of the spatiotemporal evolution and driving mechanisms of wheat, corn and rice yields in the YRB from 2000 to 2023, using the Mann–Kendall trend test, gravity center transfer model and hotspot analysis techniques. The study covered 72 prefecture-level cities across the basin. The main findings of the study are as follows:

(1)

The crop yields in the YRB exhibited significant spatiotemporal heterogeneity. Wheat yields increased significantly in the lower reaches due to technological advances and supportive policies, while yields in the ecologically fragile mid-upper regions continued to decline. Corn yields increased notably in the Hetao Plain due to mechanization and improved cultivars but declined in the arid upper reaches due to resource constraints and policy limitations. Rice yields grew markedly in tributary regions of the middle basin owing to improved irrigation and technology, whereas yields in the lower reaches declined under urbanization pressure.

(2)

Spatially, the center of wheat production followed a trajectory from the southeast to the northeast and then southwest, ultimately stabilizing along the western edge of the North China Plain. The corn production center showed periodic fluctuations before settling in the southwestern part of Wuxiang County, Shanxi. Rice production centers experienced phased shifts and eventually returned near their original locations. In terms of spatial clustering, wheat displayed a pattern of stable high yields in the lower reaches, improving yields in the middle reaches and reinforced low yields in the upper reaches. Corn exhibited a similar trend, with stable high yields downstream, modest improvement midstream and persistently low yields upstream. High rice yields were primarily concentrated in the middle and lower reaches, with evident phase-based fluctuations.

(3)

The driving force analysis revealed that natural factors—particularly climatic conditions such as temperature and precipitation—play a dominant role in influencing crop yield changes. Topographic factors indirectly constrained crop growth by modulating water and heat distribution. Socioeconomic variables (GDP, population) served as dynamic regulators, having a stronger influence in the earlier period, which diminished over time as urbanization progressed.

(4)

Structural Equation Modeling (SEM) showed good model fit: the p-VALUE for wheat, corn and rice models were 0.392, 0.338 and 0.138, respectively, with corresponding SRMR values of 0.03, 0.05 and 0.06, all meeting the thresholds of p-VALUE > 0.05 and SRMR < 0.08. The results confirmed the reliability of the model and further clarified the driving mechanisms. Wheat yield was suppressed by high elevation, low temperatures, excessive precipitation and steep slopes, with temperature exerting a moderate positive effect. Corn yield was significantly hindered by high altitude but moderately promoted by favorable slope conditions. Rice yield benefited from high-altitude light and precipitation but was notably constrained by insufficient GDP investment.

Based on the aforementioned rules, this paper puts forward the paths for the sustainable development of agriculture in the YRB: Optimize the production layout and regional adaptation. For the intensive production areas in the lower reaches, it is necessary to promote the intelligent transformation of high-standard farmland and implement differentiated subsidy policies; for the ecologically fragile areas in the middle and upper reaches, efforts should be made to popularize drought-resistant crops and water-saving technologies and guide farmers to transform through ecological compensation mechanisms. Strengthen technological innovation and resource synergy. Accelerate the research and development of stress-resistant varieties; establish an inter-provincial water diversion cooperation mechanism to ensure irrigation water for the middle and lower reaches and promote rotational irrigation and brackish water utilization technologies to tackle the over-exploitation of groundwater. Improve the institutional guarantee system, coordinate the planting structures of various provinces and regions and avoid homogeneous competition.

This study provides scientific evidence and policy recommendations for optimizing agricultural production layout and promoting sustainable agricultural development in the YRB, offering valuable theoretical and practical contributions.