Analysis of the Impact of Land Use Change on Grain Production in Jiangsu Province, China

Abstract

Located in the Yangtze River Delta region, Jiangsu Province has become the major grain production area of China and plays an important role in ensuring national food security. With rapid economic development and urbanization, the amount of cultivated land has decreased, which greatly affects food security. Based on the statistical data of grain production in Jiangsu Province since 2000 and the remote sensing data of 2000, 2010, and 2020, this paper used the stochastic frontier production function to calculate the output elasticity of various factors and the technical efficiency of grain production. The agglomeration effect of food production was investigated by using spatial correlation analysis. Finally, regression analysis was applied to examine the impact of land use change on grain yield and the technical efficiency of production. The results show that the grain-sown area is the decisive factor for the increase in grain output in Jiangsu Province. The technical efficiency of grain production in the province has been maintained at a relatively high level since 2000, showing a fluctuating upward trend, and the efficiency value in southern Jiangsu Province is greater than that in central and northern Jiangsu. The analysis of the spatial distribution characteristics of grain production technical efficiency shows that grain production has an agglomeration effect. The regression results showed that the complexity of land use and the density of the cultivated land patch were negatively correlated with grain yield and grain production technical efficiency, while the location of cultivated land was positively correlated with grain yield and grain production technical efficiency. The conclusion of this paper has important policy significance for promoting food production and ensuring food security.

1. Introduction

The Yangtze River Delta region has been a major grain production area in China since ancient times. With the development of industrialization and urbanization, this region has become an important manufacturing base of China. Since 2000, there has been an increase in population and the urban agglomeration scale in the Yangtze River Delta. The traditional grain production space has been greatly squeezed, and the per capita cultivated land area has been decreasing. With the rising living standard and migration of rural population to cities, the overall demand for food has significantly increased. The imbalance between supply and demand in food production has threatened China’s food security [1].

Jiangsu province is an important economically developed area of the Yangtze River Delta region. In 2022, the GDP of the province reached CNY 12.29 trillion, accounting for 10.2% of national GDP. Meanwhile, it is an important grain production area of China. In 2022, Jiangsu province’s grain output reached 37.69 million tons, accounting for 5.49% of China’s total food production. However, with the rapid economic development of Jiangsu Province, a lot of cultivated land has been transformed into construction land, and the cultivated land area of the province has decreased from 5,016,300 hectares in 2000 to 4,075,900 hectares in 2020, a reduction of 18.7%. In addition to the decrease in the total amount of cultivated land, the regional distribution of cultivated land has also changed, among which the total grain output and cultivated land area have shown a significant decline trend in Southern Jiangsu Province due to its higher urbanization rate. Therefore, food security is top of the agenda of all stakeholders of the food chain.

Increasing factor input to increase grain yield per unit area is an important measure to alleviate the contradiction between grain supply and demand caused by the decrease in cultivated land [2]. However, the input of factors cannot increase without limit. With the continuous input of factors, the marginal output of factors will continue to decline until zero, according to the law of diminishing marginal utility. Therefore, improving the technical efficiency of grain production has become an important way to ensure the increase in grain output [3].

The calculation of the technical efficiency of grain production is generally divided into non-parametric methods and parametric methods. Data envelopment analysis (DEA) is a more commonly used non-parametric method. Zhang et al. (2018) used the DEA method and Tobit regression model to examine grain production efficiency and its influencing factors in 13 major grain production areas of China from 2006 to 2016 [4]. Yadava et al. (2021) used the DEA method to calculate the technical efficiency of fertilizer use in India [5]. Salam et al. (2022) used the DEA method to calculate the technical efficiency of grain production in Punjab [6], and Berk et al. (2022) used the DEA method to calculate the resource utilization efficiency in corn production in Turkey [7]. The DEA method can only assess relative efficiency and ignores the effect of random errors. The advantage of SFA over DEA in terms of parametric methods is that it considers the influence of random error on the results. Determining the production function form in advance and then studying the production process can improve the accuracy of calculating technical efficiency and also allows for the analysis of the correlation between efficiency and influencing factors. Ghosh and Mazumdar (2018) calculated rice production efficiency in India using stochastic frontier production function [8]. Eguyen et al. (2003) analyzed the technical efficiency of rice production in the Mekong Delta of Vietnam based on the stochastic frontier function [9]. The stochastic frontier production function was applied to study the relationship between global warming and grain production efficiency [10] and it is also used to calculate the mechanical efficiency of rice production in China [11] and the technical efficiency of rice fertilizer use in China [12]. Alem (2021) analyzed the production performance of Norwegian grain production areas using a modified stochastic frontier production function [13]. Nathan et al. (2021) used the stochastic frontier production function to calculate the relationship between precision agriculture technology adoption and technical efficiency in the United States [14]. Chandel et al. (2022) used the stochastic frontier production function to calculate the technical efficiency of rice production in the Ganges Plain [15].

Regarding the influencing factors of grain production efficiency, Wang et al. (2011) measured and decomposed the grain production productivity of 138 counties in Hebei Province [16]. Their findings indicated that the improvement in agricultural technology, farmers’ income, land consolidation, and other factors would improve grain production efficiency. With the rapid advancement of urbanization, the decrease in cultivated land resources and the increase in food demand have gradually become important constraints in the process of development. Therefore, scholars at home and abroad have conducted a wealth of studies on the impact of land use change on food production. First, land use change affects the grain-sown area, which in turn make an impact on grain production. A few studies have been conducted on changes in cultivated land and its impact on grain production in some areas of China, e.g., Chongqing, Inner Mongolia, Jiangsu, and Huang-Huai-hai Plain [17,18,19,20]. The results show that there is a causal relationship between land sown area and total grain production. Adjei et al. (2020) conducted a study on land use change and how this has an impact on food production in Ghana [21]. Liu et al. (2021) studied the overall characteristics of grain production changes in China and found that the change in sown area was the direct cause of grain crop yield changes at the national and regional scales. By exploring the impact of cultivated land use area on food security in Baltic countries [22], Ambros and Granvik (2020) proposed that the blind construction of large-scale farms was of great harm to food security [23]. Bhermana et al. (2011) suggested the proposal of the intensification of cultivated land planning. Second, land use changes have an impact on grain production efficiency [24]. Based on the data of different land use types and statistical yearbook data of Henan Province from 2000 to 2018, Guo (2021) studied the spatial–temporal change in land use in Henan Province and its impact on grain production efficiency [3]. Rahman and Rahman (2009) explored the impact of land fragmentation on grain production efficiency in Bangladesh [25]. Manjunatha et al. (2013) studied the production efficiency of groundwater-irrigated farms in southern India and found that small farms had a relatively high resource utilization efficiency, while large farms faced resource waste due to land fragmentation and had a low production efficiency [26].

The current literature on the relationship between land use and grain production mainly focuses on the micro-scale, especially on the input of labor, agricultural machinery, and pesticides and fertilizers. The research data have mainly been obtained by sampling surveys. There are relatively few studies on the impact of land resources on grain production efficiency through remote sensing data. The mechanism of the effects of land use change on grain production efficiency still needs further investigation and understanding. Therefore, it is necessary to construct a systematic study on the impact of land use change on grain yield and efficiency. Based on the panel data of 53 regions in Jiangsu Province from 2000 to 2020, this paper estimated the grain production efficiency of this Province. Afterward, the key factors of the impact of land use change on grain production efficiency and yield were analyzed by using remote sensing images from 2000, 2010, and 2020 to interpret land use data. Finally, suggestions on land management are put forward so as to improve the technical efficiency of grain production.

2. Theoretical Research Framework

According to the production function theory, grain output is related to capital, labor force, and land input into grain production. For example, in Formula (1), Y represents total grain output, L represents capital, K represents labor force, and N represents land.

$$Y=f\left(L,K,N\right)$$

(1)

The capital input of general food production is replaced by actual material input, such as agricultural machinery and fertilizer, and the land input of food production is measured by the grain-sown area. Increasing agricultural machinery, fertilizer, labor force, and grain-sown area can all increase grain yield Y, but the limited cultivated land space in Jiangsu Province limits the maximum grain-sown area. Meanwhile, the marginal theory tells us that in the case of a certain sown area, increasing the input of agricultural machinery, fertilizer, and labor will have a diminishing marginal output effect, and the effect on the increase in grain output will become smaller. Therefore, the feasible way is to achieve the purpose of increasing grain production by improving technical efficiency.

On the one hand, the question is how to measure the existing technical efficiency, and how much technical inefficiency input is present in production? In this paper, the concept of total factor productivity is introduced, and the technical efficiency of grain production is measured by Solow residuals. By increasing the residuals in Formula (1) and selecting the appropriate grain production function, the technical efficiency of grain production can be obtained by estimating the residuals. On the other hand, what is the impact of land use change on technical efficiency? Industrialization and urbanization have a strong spatial competition for grain planting, which is unlikely to increase the existing cultivated land area and grain production area. However, through the adjustment and optimization of the spatial layout of cultivated land and grain production, it is possible to increase grain yield and technical efficiency.

According to the theory of spatial economics, location has a very important impact on resource allocation and production efficiency. Then, the change in the spatial layout of cultivated land will also have an important impact on the technical efficiency of grain production. With the continuous expansion of cities, cultivated land has become further away from suburban areas. The rising population and improvement in transportation networks has increased the fragmentation of cultivated land. The development of the manufacturing industry has also increased the intensity of land development and the complexity of land use. Continuous changes in the location, morphology, and surrounding environment of cultivated land have had a significant impact on food production.

This paper aims to establish a grain production function model of Jiangsu Province to verify that the sown area is the determining factor of grain output under the current agricultural development stage of Jiangsu Province. After calculating the technical efficiency of grain production, we analyze the effects of the spatial morphology of cultivated land on grain yield and technical efficiency. Finally, the spatial optimization scheme is proposed to streamline the spatial layout of grain production and improve grain productivity.

3. Methods and Data

3.1. Research Method

3.1.1. The Calculation Method of Grain Production Technical Efficiency

Stochastic frontier analysis is one of the most commonly used methods for measuring technical efficiency. The stochastic frontier analysis method was first proposed by Aigner [27] in 1977. After improvement and refinement by Battese and Coelli et al. [28], this model was widely adopted in subsequent research and has become the mainstream method for studying technical efficiency. In order to improve the accuracy of the model and better reflect the combined impact of different input factors on grain production efficiency in the production function, this study applied the translog stochastic frontier production function model for the analysis [29], as follows:

$$\begin{array}{l}\ln \left(Y_{it}\right)={\beta }_0+{\displaystyle \sum _j{\beta }_j}\ln X_{jit}+\frac{1}{2}{\displaystyle \sum _j{\displaystyle \sum _l{\beta }_{jl}\ln }}{X}_{jit}\ln X_{lit}+{\beta }_{t}t\\ \hspace{1em}\hspace{1em}\hspace{1em}\hspace{1em}+{\beta }_{tt}{t}^2+{\displaystyle \sum _j{\beta }_{jt}}t\ln X_{jit}+v_{it}-u_{it}\end{array}$$

(2)

$$T{E}_{it}=\exp (-u_{it})$$

(3)

$Y$ represents the grain output of region i in year t. ${\beta }_0$ is the intercept term. ${\beta }_j$,${\beta }_t$, ${\beta }_{jl}$, ${\beta }_{tt}$, ${\beta }_{jt}$ are the parameters to be estimated. $t$ represents the year. $X$ is the input factor. v and u are the error terms of the model, where v represents the random error term subject to normal distribution, u represents the technical inefficiency term that follows the truncated normal distribution, and v and u are independent of each other. $T{E}_{it}$ represents the technical efficiency level of grain production in region i in year t. Input factors $X_{jit}$(j = 1,2,3,4) are sown area, labor force, agricultural machinery power, and fertilizer application amount.

In order to overcome the shortcomings of a single model, Hansen and Racine [30] proposed the cutter model averaging method, which determines the weight by minimizing the cross-validation criterion, and then assigns different cutter weights to different models so as to obtain more comprehensive and comprehensive estimation results.

Assuming there are m candidate models, the knife cut the fit value of each candidate model’s explanatory variable ${\widehat{y}}^m=\left({\widehat{y}}_1^m,\cdots ,{\widehat{y}}_n^m\right)$, for which ${\widehat{y}}_i^m$ is the fair value of the explanatory variable after excluding the $i$-th sample. The weight $w_m$ of the knife cut model averaging method is a set of nonnegative vectors and sum to 1. The calculation formula is as follows:

$$\begin{array}{l}{w}^{*}=\underset{w=\left(w_1,\cdots ,w_M\right)\in {\Omega }_M}{argmin}C{V}_n\left(w\right)=\frac{1}{n}\widehat{e}{\left(w\right)}^{\prime }\widehat{e}\left(w\right)\\ \hspace{1em}\hspace{0.17em}\hspace{0.17em}=\frac{1}{n}{\left(y-{\displaystyle \sum _{m=1}^M{w}_m{\widehat{y}}^m}\right)}^{\prime }\left(y-{\displaystyle \sum _{m=1}^M{w}_m{\widehat{y}}^m}\right)\end{array}$$

(4)

Among them, $w^{*}$ is the cutting weight, $\widehat{e}\left(w\right)$ is the weighted average residual, and ${\displaystyle \sum _{m=1}^M{w}_m{\widehat{y}}^m}$ is the weighted average of the knife cut fit values. After obtaining the cutting weight $w^{*}$, the regression model of this study is as follows:

$$\begin{array}{l}{y}_{it}={\displaystyle \sum _{m=1}^M{w}_m^{\ast }}{\widehat{f}}_m(x_{it})={\displaystyle \sum _{m=1}^M{w}_m^{\ast }}({\widehat{\beta }}_{0it}^m+{\displaystyle \sum _{j=1}^4{\widehat{\beta }}_{x_j}^m}{x}_{jit}+\frac{1}{2}{\displaystyle \sum _{j=1}^4{\displaystyle \sum _{l=1}^4{\widehat{\beta }}_{x_{jl}}^m}}{x}_{jit}{x}_{lit}\\ \hspace{1em}\hspace{1em}\hspace{0.17em}+{\widehat{\beta }}_t^{m}t+{\widehat{\beta }}_{tt}^m{t}^2+{\displaystyle \sum _{j=1}^4{\widehat{\beta }}_{x_{jt}}^m}t{x}_{jit}+{\widehat{v}}_{it}^m-{\widehat{u}}_{it}^m\end{array}$$

(5)

Among them, $w_m^{*}$ is the cutting weight of the m-th model, ${\widehat{f}}_m(x_{it})$ is the regression result of the m-th model, and $x_{jit}$(j = 1,2,3,4) represents the sown area, labor force, agricultural machinery power, and fertilizer application amount. After deduction, the elasticity coefficient of the grain-sown area, labor force, agricultural machinery power, and fertilizer application amount of the final model is obtained as follows:

$$B_{x_j}={\displaystyle \sum _{m=1}^M{w}_m^{\ast }}({\widehat{\beta }}_{x_j}^m+{\displaystyle \sum _{l=1}^4{\widehat{\beta }}_{x_{jl}}^m}{x}_{lit}+{\widehat{\beta }}_{x_{jt}}^{m}t)(j=1,2,3,4)$$

(6)

3.1.2. Spatial Correlation Analysis of Technical Efficiency in Grain Production

Spatial econometric analysis is an important component of spatial statistics and an effective means of understanding spatial patterns [31]. For spatial econometric regression models, when the results are calculated, it is necessary to analyze the significance of their spatial correlation to see whether it is significantly different from zero. In the spatial error model, the common methods for the spatial autocorrelation test of random disturbance terms include Moran’s I test, likelihood ratio (LR) test, Lagrange coefficient (LM) test, Wald test, etc. The difference is that Moran’s I test is based on least squares estimates, while the likelihood ratio (LR) test, Lagrange coefficient (LM) test, and Wald test are based on maximum likelihood estimates.

Moran’s I statistic for spatial autocorrelation is:

$$I=\frac{{\displaystyle \sum _{i=1}^n{\displaystyle \sum _{j=1}^n{\omega }_{ij}}(x_i-\overline{x})(x_j-\overline{x})}}{\frac{{\displaystyle \sum _{i=1}^n{(x_i-\overline{x})}^2}}{n}{\displaystyle \sum _{i=1}^n{\displaystyle \sum _{j=1}^n{\omega }_{ij}}}}$$

(7)

In the equation, $x_j$ represents the grain production efficiency of regional units i and j, and ${\omega }_{ij}$ is the spatial weight value between i and j. This paper uses a Queen’s second-order adjacency weight matrix, where Moran’s I is progressive normal distribution. $\overline{x}$ is the average grain production efficiency in Jiangsu Province. The Moran’s I index has a range of values between −1 and 1. At the given significance level, Moran’s I index is greater than 0, indicating that there is a positive spatial autocorrelation of grain production efficiency, showing a clustering trend. Moran’s I is less than 0, indicating that there is a negative spatial autocorrelation of grain production efficiency, showing a discrete trend. When Moran’s I = 0, there is no spatial autocorrelation.

Moran’s I spatial autocorrelation coefficient only reflects the spatial distribution pattern of the research phenomenon in the entire study area and cannot obtain the location of the aggregation area of the research phenomenon from it. To explore this phenomenon, Getis–Ord $G_i^{*}$ statistics were used for local spatial autocorrelation analysis. In spatial clustering analysis, local clusters with higher values are called hotspots, while local clusters formed by lower values are called cold spots.

$$G_i^{*}(d)=\frac{{\displaystyle \sum _{j=1}^n{\omega }_{ij}}(d)x_j}{{\displaystyle \sum _{j=1}^n{x}_j}}$$

(8)

In the formula, ${\omega }_{ij}\left(d\right)$ represents the spatial weight matrix, and $x_j$ represents the sample values of region j. The larger the absolute value of $G_i^{*}$ is, the more concentrated the grain production efficiency is, that is, the hot spot area is formed. When $G_i^{*}$ is positive, this region is a positive hotspot region, and when $G_i^{*}$ is negative, this region is a negative hotspot region. When the value of $G_i^{*}$ is close to 0, it indicates that there is no aggregation of grain production efficiency.

3.1.3. Analysis of Influencing Factors of Technical Efficiency of Grain Production

The Tobit model is a restricted dependent variable regression model. Compared to the traditional OLS regression and discrete models, it can effectively avoid bias in the regression estimation process, and thus it has been widely used [32]. The explained variable in this paper is the technical efficiency of grain production. As the technical efficiency of grain production calculated by using the stochastic frontier production function ranges from 0 to 1, it is a typical constrained dependent variable. If OLS is used to regression the model, the results may be biased, so the Tobit regression model is chosen for analysis. The specific model is as follows:

$$T{E}_i=\left\{\begin{array}{l}{\sigma }_0+{\displaystyle \sum _k^{}{\sigma }_k{Z}_{ki}+{\xi }_i,{\delta }_0+{\displaystyle \sum _k^{}{\delta }_k{Z}_{ki}+{\xi }_i>0}}\\ 0,{\delta }_0+{\displaystyle \sum _k^{}{\delta }_k{Z}_{ki}+{\xi }_i\le 0}\end{array}\right.$$

(9)

In the formula, TE represents the technical efficiency level of grain production. $Z_{ki}$ is the independent variable that affects the efficiency of grain production. ${\delta }_0$ and ${\delta }_k$ are the estimated parameters of the model. ${\xi }_i$ is the model’s error term, which belongs to a normal distribution.

Z₁ Land use diversity index [33,34] represents the impact of land use diversity, calculated using the Shannon index. The specific formula is as follows:

$$Z_1=1-{\displaystyle \sum d_i^2}/{\left({\displaystyle \sum d_i}\right)}^2$$

(10)

In the formula, d_i represents the area of the i-th land use type in the study area, reflecting the complexity and utilization degree of land use types. The value ranges from 0 to 1. The larger the value, the higher the degree of land use diversity, and the lower the degree of land use diversity.

Z₂ Cultivated land patch density represents the degree of fragmentation of cultivated land, calculated as the ratio of the number of cultivated land patches to the area of cultivated land, i.e., the number of patches per unit area of cultivated land. The number of cultivated land patches is extracted from remote sensing image interpretation data using ArcGIS10.2 software.

Z₃ Cultivated land area ratio refers to the ratio of the cultivated land area to the total area in each region, reflecting the size of cultivated land in each region.

Z₄ Cultivated land distance refers to the linear distance from the geometric center of cultivated land to the administrative center in each region, reflecting the spatial positioning of cultivated land in the city.

Z₅ Land development intensity refers to the proportion of construction land area in the total area of a region. The high development intensity means that the proportion of construction land is high.

3.2. Data Source

Variables for calculating the technical efficiency of grain production include grain output, grain-sown area, labor force, agricultural machinery power, and fertilizer application amount (see

Table 1. Statistical description of the main input variables for grain production efficiency.

Variable	Mean Value	Standard Deviation
Grain-sown area (1000 hectares)	404.707	240.468
Labor force (10,000 people)	259.970	114.927
Power of agricultural machinery (10,000 kilowatts)	306.853	175.893
Fertilizer application amount (10,000 tons)	25.080	17.880

). The data mainly come from the Jiangsu Provincial Statistical Yearbook (2000–2020). For data statistics and continuity convenience, statistical analysis was conducted based on the administrative regions divided by the 2020 Jiangsu Provincial Statistical Yearbook. Data from 53 sample districts were obtained. Because the statistical yearbook only includes statistics on the total power of agricultural machinery and the net amount of fertilizer application, it is not subdivided into various indicator values for grain production and use, except for the grain-sown area. Therefore, the data need to be processed accordingly. Drawing on relevant research and using the weight coefficient method [35] to separate the input factors of grain production, assuming that the material inputs per unit area of grain planting and other agricultural production are equal and that the labor force input and output are also equal, then:

Labor force = (Total agricultural output value/Total output value of agriculture, forestry, animal husbandry, and fishery) × (Grain-sown area/Total area of crop sown) × (Number of labor force in agriculture, forestry, animal husbandry, and fishery).

Power of agricultural machinery = (Grain-sown area/Total area of crop sown) × (Total power of agricultural machinery).

Fertilizer application amount = (Grain-sown area/Total area of crop sown) × (the net amount of chemical fertilizer applied).

According to the land use data interpreted from remote sensing images, there were 280,086, 216,224, and 218,805 land use patches in 2000, 2010, and 2020, respectively. Among them, the number of cultivated land patches was 39,398, 24,525, and 42,234. Based on this, the land use data of various districts and counties in Jiangsu Province were compiled and combined with partial statistical yearbook data, and the land use diversity index and cultivated land patch density were calculated.

4. Results

4.1. Output Elasticity of Grain Production Factors

The Stata 15 software was used to estimate three types of panel random frontier production functions, and the instruction jackknife was used to calculate the knife cut fit value ${\widehat{y}}^m$ of the explanatory variables for the three models. Then, the respective weights were obtained according to the cross-validation criterion, as shown in

Table 2. Estimation results of the production function of different types of stochastic frontiers.

Variable	BC92 Model	BC95 Model	Kumb90 Model
Grain-sown area	1.002	0.961	0.989
Labor force	0.035	0.029	−0.015
Power of agricultural machinery	0.052	0.0013	0.056
Fertilizer application amount	−0.036	−0.0176	−0.005
Technical efficiency	0.869	0.922	0.851
Knife cutting weight	0.335	0.326	0.339

Note: BC92 model is a stochastic frontier production function model proposed by Battese and Coelli in 1992. The expression of its inefficiency term is as follows: $u_{it}={\eta }_t{u}_i=\left\{\mathrm{e}\mathrm{x}\mathrm{p}[-\eta \left(t-T\right)]\right\}{u}_i$. BC95 model is a stochastic frontier production function model proposed by Battese and Coelli in 1995. The expression of its inefficiency term is as follows: $u_{it}=z_{it}\delta +w_{it}$. Kumb90 model is a stochastic frontier production function model proposed by Kumbhakar in 1990. The expression of its inefficiency term is as follows: $u_{it}=\gamma \left(t\right)u_i={\left[1+\mathrm{e}\mathrm{x}\mathrm{p}(bt+c{t}^2)\right]}^{-1}+u_i$.

.

Based on the estimated coefficients and cutting weights of different types of random frontier production functions, the elasticity coefficient is calculated. The calculation results are shown in

Table 3. Calculation of elastic coefficient values using the average method of cutting model.

Variable	Grain-Sown Area	Labor Force	Power of Agricultural Machinery	Fertilizer Application Amount
Elastic coefficient	0.984	0.016	0.037	−0.019

.

From the above table, it can be seen that the output elasticity of the grain-sown area, labor force, and power of agricultural machinery is positive. The output elasticity of the fertilizer application amount is negative. The results showed that increasing the input of the grain-sown area, labor force, and agricultural machinery power could increase grain yield, but increasing the input of fertilizer application would decrease grain yield. The yield elasticity of the grain-sown area was the largest, reaching 0.984, indicating that under the given technical conditions, keeping other input factors unchanged and increasing the sown area by an additional 1%, the grain output of the province would increase by 0.984%. This reflects the critical importance of land factors for food security in Jiangsu Province.

4.2. Technical Efficiency of Food Production

According to the calculation results of the Stata 15.0 software, the grain technical efficiency curve of each city in Jiangsu Province was drawn, as shown in Figure 1.

It is calculated that the average technical efficiency of grain production is 0.880 during 2000–2020 in Jiangsu Province. On the whole, the production efficiency showed a rising trend in the fluctuation. The efficiency of grain production in various cities of Jiangsu Province varies depending on the region. The efficiency of grain production in southern Jiangsu is higher, followed by central Jiangsu, and it is the lowest in northern Jiangsu.

Specifically, the average level of grain production efficiency in the southern region of Jiangsu is 0.921. Among the five cities, Changzhou has the highest average level of grain production efficiency at 0.950, while Zhenjiang has the lowest average level of grain production efficiency at 0.892.

The average level of grain production efficiency in the central Jiangsu region is 0.876. Among the three cities, Taizhou City has the highest average level of grain production efficiency at 0.943. Nantong City has a relatively low average level of grain production efficiency at only 0.789. It can be seen that there is a significant difference in the efficiency of grain production among different cities in the central Jiangsu region.

The average level of grain production efficiency in the northern Jiangsu region is 0.859. Among the five cities, Yancheng City has the highest average level of grain production efficiency at 0.891.

The three time nodes of 2000, 2010, and 2020 are selected. As can be seen from Figure 2 and Figure 3, the spatial distribution of the grain production efficiency in Jiangsu Province has obvious changes.

In order to further explore grain production efficiency at the county level, this paper analyzes the spatial characteristics of grain production efficiency in 2000, 2010, and 2020 from the perspective of the county level, as shown in Figure 3. From 2000 to 2010, the spatial distribution of the grain production efficiency mainly showed a decrease in grain production efficiency. From 2010 to 2020, except for some areas such as Xuzhou urban area, Suqian urban area, Funing County, and Suining County, the rest of the region basically maintained a stable recovery of grain production efficiency.

4.3. Spatial Correlation of Technical Efficiency

The ArcGIS10.2 software was used to calculate the global Moran’s I index of the spatial distribution of the grain production efficiency in Jiangsu Province in 2000, 2010, and 2020 from the city level and county level, respectively (

Table 4. Global Moran index of grain production in Jiangsu Province in 2000, 2010, and 2020.

	Time	Moran’s I	z	p
At the city level	2000	0.379	2.372	0.017
	2010	0.155	1.278	0.201
	2020	0.231	1.681	0.092
At the county level	2000	0.211	2.710	0.006
	2010	0.184	2.581	0.009
	2020	0.221	3.047	0.002

), and all of the Moran’s I indexes were positive. The results showed that the grain production efficiency in Jiangsu Province showed a positive spatial autocorrelation. The spatial agglomeration of grain production in Jiangsu Province showed that cities with higher (lower) grain production tended to cluster near each other.

The Getis–Ord G*i index was used to calculate the grain production efficiency in Jiangsu Province in 2000, 2010, and 2020, and the spatial agglomeration distribution pattern of grain production in Jiangsu Province is shown in Figure 4 and Figure 5.

From Figure 4, the cold points of the grain production efficiency in 2000 are in Xuzhou, Suqian, and Lianyungang cities in northern Jiangsu. In 2020, the cold points of the grain production efficiency are only in Xuzhou and Suqian, which belong to the regions with low and significant cold points. The hot spots of the grain production efficiency are mainly concentrated in southern Jiangsu, including Zhenjiang, Changzhou, and Wuxi. From 2000 to 2020, Zhenjiang has always been a medium significant hotspot area, Changzhou has changed from a high significant hotspot area to a medium significant hotspot area and then to a high significant hotspot area again until 2020, while Wuxi has gradually changed from a medium significant hotspot area to a low significant hotspot area. There is a large difference in grain production efficiency between northern Jiangsu and southern Jiangsu. In northern Jiangsu, there is a significant spatial agglomeration feature of low grain production efficiency, while in southern Jiangsu, there is a significant agglomeration feature of high grain production efficiency.

From Figure 5, the distribution of hot and cold spots of grain production efficiency from 2000 to 2010 at the county level changed mainly from a random distribution area to a low significant cold point area in Siyang, from a low significant cold point area to a medium significant cold point area in Nantong, and from a medium significant hot point area to a low significant hot point area in Changzhou and Wuxi. From 2010 to 2020, no other changes were noticed except Suzhou urban area and Wuxi urban area. Both areas changed from hot low significant to hot high significant areas. From the distribution point of view, the cold point agglomeration of grain production efficiency is concentrated in Nantong city, Qidong city, Suqian city, and Siyang city, indicating that there is a spatial agglomeration of low grain production efficiency in these areas. The hotspots are concentrated in parts of Changzhou, Wuxi, and Suzhou. From the county level, there have not been any hotspots and significant areas in the province since 2000. There is no high value spatial agglomeration of food production efficiency.

4.4. Effects of Land Use Change on Food Production

Taking grain yield (Y) and technical efficiency (TE) as dependent variables, OLS and Tobit methods were used for the regression calculation, and the coefficients of the land use diversity index, cultivated land patch density, cultivated land area ratio, cultivated land distance, and land development intensity were obtained, as shown in

Table 5. Panel data estimation of the impact of land use change on grain yield in Jiangsu Province.

	Coefficient	Standard Deviation	t-Statistic
Land use diversity index	−4.597 ***	0.256	−17.97
Cultivated land patch density	−0.289 ***	0.024	−12.04
Cultivated land area ratio	0.230 ***	0.065	3.54
Cultivated land distance	0.102 ***	0.031	3.35
Land development intensity	−1.423 ***	0.099	−14.39
Constant	−4.597 ***	0.256	−17.97

Note: *** indicate significance levels at 1%.

and

Table 6. Panel data estimation of the impact of land use change on grain technology efficiency in Jiangsu Province.

	Coefficient	Standard Deviation	t-Statistic
Land use diversity index	−0.062	0.104	−0.60
Cultivated land patch density	−0.001 *	0.000	−1.74
Cultivated land area ratio	−0.195 ***	0.052	−3.77
Cultivated land distance	0.002	0.002	0.86
Land development intensity	−0.309 ***	0.064	−4.86
Constant	−0.062	0.104	−0.60

Note: *** and * indicate significance levels at 1% and 10%.

.

The regression results showed that the land use diversity index and cultivated land patch density had negative effects on grain yield and technical efficiency, cultivated land distance was positively correlated with grain yield and technical efficiency, and the improvement in land development intensity was negatively correlated with grain yield and technical efficiency.

5. Discussion

When food security is top of the agenda of different stakeholders, it is of great significance to examine the optimization of cultivated land use. Our study reveals the strong correlation between cultivated land area and grain production in Jiangsu Province through the analysis of the stochastic frontier production function of grain output and input factors. The technical efficiency of grain production is calculated at the city and county level of the province. Using the spatial data of land distribution creatively, the impacts of land use change on grain yield and efficiency in the province are analyzed. According to the above research results, the output elasticity of grain production input factors, the spatial distribution of the technical efficiency of grain production, the agglomeration effect, and the influencing factors of land use change on grain production will be discussed.

According to the calculation results of the output elasticity of the production factors, it can be seen that land factors play a decisive role in the current stage of grain production in Jiangsu Province. This result is basically consistent with the finding of Li (2017) [36]. In his research, the output elasticity of the grain-sown area is the largest, reaching 0.75, while labor force and agricultural machinery account for 0.07 and 0.02, respectively. The same value was found in this paper. The output elasticity of labor force is 0.016, indicating that an additional 1% increase in labor force will increase food production by 0.016%, keeping other input factors and technical conditions unchanged. On the one hand, this conclusion is similar to the research conclusion of Zeng et al. (2012) [32], that is, farmers are the main laborers in agricultural production; on the other hand, the coefficient of 0.016 indicates that the marginal income of rural labor is already low under the current conditions, and the input of a large number of labor forces does not contribute much to the growth of food production. The output elasticity of farm machinery power is 0.037, indicating that an additional 1% increase in total farm machinery power will increase grain output by 0.037%. This value is small, and the contribution of increasing this input factor to output growth is small. The main reason for this may be that in order to avoid being affected by natural disasters, in busy farming seasons such as spring ploughing and autumn harvest, farmers may invest multiple machines to work at the same time to complete the seeding and harvesting process as soon as possible, which may lead to the excessive input of machinery, as found in the study of Guan (2006) [37]. Compared with the national average, the output elasticity of the sown area in Jiangsu Province accounts for a more prominent proportion, reaching 0.984, which is about 0.23 higher than the national level. As a developed region in the eastern coastal region, this province has a high level of agricultural mechanization, a large amount of machinery input, a dense population, and abundant labor resources. Therefore, the marginal diminishing effect of agricultural machinery and labor input is more obvious. In this way, the factor of sown area becomes more important. What is more different is that, compared with the national fertilizer output elasticity of 0.18, the fertilizer output elasticity of Jiangsu Province is −0.19, indicating that an additional 1% increase in fertilizer application will reduce grain output by 0.019%. This is consistent with the conclusion put forward by Wang (2018) [38], that is, the application rate of chemical fertilizer has shown a decreasing marginal utility. An excessive amount of chemical fertilizer application will cause soil quality degradation and pollution [39], which will endanger the further improvement in grain yield.

According to the efficiency calculation and spatial distribution in the past 20 years, the technical efficiency of grain production in southern Jiangsu is generally higher than that in central and northern Jiangsu, which is consistent with Zhang Qian’s research findings [40]. In 2020, the average grain production efficiency of cities in southern Jiangsu, central Jiangsu, and northern Jiangsu was 0.927, 0.880, and 0.866, respectively. There are historical reasons, natural reasons, and economic and social reasons for the formation of the different grain production efficiency. Historically, south Jiangsu has always been the main grain-producing area of our country, known as the “granary”, accumulating rich experience of intensive farming. In terms of natural conditions, the southern region of Jiangsu has a dense river network, convenient irrigation, abundant rainfall, and fertile soil, while the conditions in the central and northern regions of Jiangsu are relatively poor; in particular, the soil salinization in the northern region of Jiangsu has a certain impact. From the perspective of economic and social development, southern Jiangsu has a developed economy, advanced technology, and convenient transportation, while central and northern Jiangsu are relatively backward. The sown area of grain crops in central and northern Jiangsu is more than that in southern Jiangsu, and the labor force engaged in grain production is also more than that in southern Jiangsu, but the grain production efficiency is relatively low. Therefore, to improve the efficiency of grain production in southern Jiangsu, the most important thing is to ensure that there is enough cultivated land. Specifically, it is necessary to strictly observe the red line of cultivated land, increase the development of cultivated land, and innovate the mode of urban development. For northern Jiangsu, it is necessary to reduce the investment of redundant factors in grain production, increase the investment of capital and scientific research, accelerate the construction of agricultural infrastructure, and attract talents to accelerate the development of modern agriculture.

The cold spots of the grain production efficiency are mainly concentrated in the north of Jiangsu, and the hot spots of the grain production efficiency are mainly concentrated in the south of Jiangsu. Through the analysis of spatial aggregation characteristics, it can be seen that the efficiency of the counties with strong traditional grain production has been consolidated and improved, and the high-value agglomeration areas have also been transferred to these counties. We should focus on breaking through the technical bottleneck of improving the agricultural production technical efficiency in northern Jiangsu. Since 2000, the growth rate of the grain production efficiency in northern Jiangsu has been slow and at a low value. The government should provide relevant policy guidance for grain production in northern Jiangsu and help these areas break through the bottleneck of improving the grain production technical efficiency by further optimizing the allocation of agricultural production resources and improving the level of production management.

The regression results of the impact of land use change on grain production show that the improvement in the land use diversity index in Jiangsu province has a negative impact on grain production. With the growth of population and the rapid development of social economy, the change in land use in Jiangsu Province has intensified. Land has gradually been enriched from a relatively single use to multiple uses, such as industry, transportation, entertainment, and residential, and the proportion of various uses has increased, except as cultivated land. Land use structure has become more complex, and the diversity index of land use has also been increasing. The effect on grain yield and efficiency is also significant. The model estimation results show that grain yield and efficiency decrease as the land use diversity index increases. The cultivated land patch density has a negative impact on grain yield and efficiency, which is similar to the conclusion of Ntihinyurwa et al. (2021) [41]. With the Mosaic occupation of cultivated land by other land types, especially the continuous expansion of rural settlements and urban construction land, and the rapid development of the transportation network, the fragmentation of cultivated land continues to increase, hindering the promotion of mechanical equipment and large-scale agricultural technology, which is not conducive to the intensive production of grain and is bound to affect grain production. Berk (2022) also concluded that agricultural land fragmentation may lead to inefficiency [7]. The distance of cultivated land has a positive correlation with grain output, indicating that in recent years, with the shift of the center of gravity of cultivated land to the urban periphery, the distance between it and the administrative center has increased, and the probability of land being used for urbanization construction has become smaller and smaller. In addition to the definition of “three districts and three lines”, most of these areas are outside the urban development boundary, vigorously developing agriculture has become the leading direction, and the grain-producing areas are basically stable, which is conducive to the increase in grain output. The improvement in the land development intensity is negatively correlated with grain output, indicating that the intensity of land development is increasing with the improvement in the urbanization level. The crowding out of a large amount of cultivated land by construction land will inevitably lead to a prominent contradiction between human and land, resulting in the decline in grain output and efficiency.

In this paper, the effects of grain production efficiency and land use change on grain production efficiency in Jiangsu Province over the last 20 years were discussed, and some useful conclusions were obtained. The deficiency of this paper is that the overall situation of food production has been studied, and the output structure of food production has not been studied. Subdividing the production of major food crops, such as the impact of land use change on rice, wheat, etc., is a direction that can be studied in the future.

6. Conclusions

This study uses the stochastic frontier production function to verify the important role of grain-sown area in grain production, calculates the technical efficiency of grain production by matching the statistical data of grain production in Jiangsu province with the remote sensing data of cultivated land layout, and analyzes the impact of land use change on grain production yield and technical efficiency from a spatial perspective. The main conclusions of this study are as follows:

Firstly, through the calculation and analysis of the stochastic frontier production function, the grain-sown area is the decisive factor for the increase in grain output in Jiangsu province. Under the current technology level, grain output will increase by 0.984% for every 1% increase in sown area. The increase in agricultural machinery power and labor force has very little impact on the increase in grain yield, and the impact of the fertilizer application rate on the increase in grain yield is even negative. This shows that under the current level, the most effective way to increase grain output and ensure food security in Jiangsu Province is to increase the sown area. The size of the grain-sown area is directly related to the cultivated land area, so it is very important to ensure that the cultivated land area does not decrease and meet the demand for grain sowing.

Secondly, since 2000, the technical efficiency of grain production in Jiangsu Province has been maintained at a relatively high level, showing a fluctuating upward trend. In terms of the whole province, the grain production efficiency in southern Jiangsu is generally higher than that in northern and central Jiangsu. The southern region of Jiangsu relies on advanced agricultural technology to drive grain production, while the economic development level of the northern region of Jiangsu is relatively low, and the promotion of modern agricultural technology and agricultural infrastructure construction is not as good as that of the southern region of Jiangsu. In the future, it is necessary to continuously improve the level of modern agricultural technology in northern Jiangsu and make good use of the cultivated land space in northern Jiangsu.

Thirdly, the analysis of the spatial distribution characteristics of the technical efficiency of grain production in Jiangsu Province shows that there is a positive spatial correlation of grain production in Jiangsu province, and grain production presents agglomeration benefits. In the past 20 years, the distribution of cold hot spots of grain production efficiency in Jiangsu province has not changed much, the cold spots of grain production technical efficiency are mainly concentrated in the north of Jiangsu, and the hot spots of grain production technical efficiency are mainly concentrated in the south of Jiangsu. Through the analysis of spatial aggregation characteristics, it can be seen that the efficiency of the counties with strong traditional grain production has been consolidated and improved, and the high-value agglomeration areas have also been transferred to these counties.

Fourthly, through the analysis of the impact of land use change on the total grain output and the technical efficiency of grain production in the province, it is found that the increase in the complexity of land use, especially the increase in the proportion of construction land, has a negative impact on both grain output and the technical efficiency of grain production. From the perspective of the location of cultivated land, the farther cultivated land is away from the city, the more conducive it is to the improvement in grain output and grain production efficiency. From the perspective of cultivated land shape, the larger the density of the cultivated land patch and the more fragmented the cultivated land are, the more unfavorable the increase in grain yield and technical efficiency of grain production is.

In short, in order to ensure food security, Jiangsu province should strengthen the protection of cultivated land to ensure that the due grain-sown area is not reduced. Second, we should reduce the fragmentation of cultivated land in the province and improve the technical efficiency of grain production. Third, we should break the unbalanced development of the region and focus on improving the technical efficiency of northern Jiangsu.