Analysis of Land Use Change and the Role of Policy Dimensions in Ecologically Complex Areas: A Case Study in Chongqing

Abstract

China has adopted policies, such as the Grain for Green program (GFGP) and China’s Western Development Strategy, to maintain ecosystem sustainability and the rational use of land resources based on economic development. Existing studies have revealed the impact of these policies on land use and land cover change (LUCC). However, more research is needed to identify what would happen if the original trajectory of land use change were to continue unaffected by policy. In this research, we employed the future land use (FLUS) model to simulate land use changes in Chongqing under the natural scenario in 2020, assuming the existence of policy and natural contexts. The relative contribution conceptual model (RCCM) estimated the contribution of policies to LUCC, assessed the characteristics of LUCC in both situations using a complex network model, and analyzed the policies affecting LUCC. The findings revealed that cropland was the key land use type in both contexts, and the stability of the land use system in the natural context was greater than in the policy context. This research contributes to new research ideas for analyzing land use change and comprehending the role of policy execution in land use change.

1. Introduction

Land use and land cover change (LUCC) is linked to numerous changes in terrestrial ecosystems over time [1,2,3,4]. LUCC is the complex result of natural, policy, economic, and other factors [5]. Global warming is one of the great challenges facing humanity today; xeric biomes are becoming more widely distributed [6], and the risk of coastal marshes being submerged is increasing [7]. At the same time, human activities, such as afforestation and urbanization, are driving changes in land use [8]. The total amount of available land is limited, but human factors, such as the acceleration of urbanization, have led to the continuous replacement of other land types with construction land, and the efficient and rational use of land is imminent [9,10]. Therefore, policy makers and related scholars need to have a good grasp of the current land use situation, be able to predict LUCC under natural scenarios, and assess the relative contribution of policies to LUCC.

Regionally relevant policies can essentially determine trends in LUCC. For example, the prohibition of deforestation directly promotes an increase in forest coverage [11]. The destruction of the ecological environment will result in societal disasters, such as floods, and, thus, many countries have successively introduced policies to restore the environment [12]. China has also invested large-scale funds for ecological restoration, and these promulgated policies have lasted for a long time [13]. These policies and measures show China’s determination to maintain ecological security, and, thus far, good results have been achieved. The Grain for Green program (GFGP), the Western China Development Strategy, and other strict land conservation and intensive use policies were all implemented after 2000, and the Three Gorges Dam project (TGDP) was completed in 2009. The GFGP is one of the most influential ecological restoration policies and restores farmland that is unsuitable for cultivation into woodland or grassland [13]. The GFGP has not only brought about beneficial effects, such as the improvement of the greening rate [14], but it has also caused adverse effects on the ecological environment system in certain areas [15,16]. In addition, the Western China Development Strategy, the TGDP, and other strict land conservation and intensive use policies have also had important impacts on LUCC in Chongqing, with these policies being implemented to promote the development of the western region, to prevent flooding, and to promote the intensive and economical use of land, respectively. A quantitative assessment of the contribution of these policies to various land use types can help to augment our understanding of the role of current policies in this context and improve future policies in a timely manner.

In this study, natural scenarios refer to scenarios that are less affected by policies, and policy scenarios refer to scenarios that are affected by both natural scenarios and policies. Previous studies have typically identified policy impacts only within the policy context, neglecting the role of natural scenarios, and quantitative analyses of policy impacts on LUCC are scarce. Tian et al. studied the effects of human factors and climate factors on vegetation cover protection [17]. Hoek pointed out that there are regional differences in the impact of policies on forest coverage and how in some places, forest coverage decreases under policy scenarios [18]. Fu et al. evaluated ecosystem services based on land use changes under the business-as-usual, economic development, and ecological conservation scenarios [19]. However, there are few quantitative analyses of policy impacts and no distinction between natural and policy scenarios. Therefore, a quantitative analysis of the impact of both scenarios on LUCC is necessary.

The future land use simulation Includes two parts: the prediction of land use demand and a spatial distribution simulation. For the prediction of land use demand, Markov chain models have been used to forecast land use demand at various study scales [20]. The model creates a transfer probability matrix based on the land use status in one period and predicts the land use demand in the next period based on this matrix. It does not require independent data, such as logistic regression, and it is very convenient to use the Markov chain model when encountering dependent data [21]. Based on the Markov model, the amount of land use in the Beijing-Tianjin-Hebei region in 2030 under the business-as-usual, cropland protection, and ecological security scenarios and the land use demand in Shanghai in 2030 under the unconstrained development and development with planning intervention scenarios were obtained [22,23]. Therefore, the Markov model can effectively simulate land use demand under different scenarios.

For spatial distribution simulation, the cellular automata (CA) model is often used to study urban LUCC [24], and the Conversion of Land Use and its Effect at Small regional extent (CLUE-S) model has been used to study future land use trends in the county [25]. Previous models focused too much attention on the impact of human factors on LUCC and could not effectively simulate LUCC in the context of climate change. However, climate change is one of the key factors affecting simulation results. The future land use simulation (FLUS) model considers the effects of climate change [26]. The FLUS model is designed with adaptive inertia and competition mechanisms to estimate the likelihood of transfer between different land use types and is a comprehensive model that can simulate future land use conditions based on social and natural factors [26,27]. Based on the FLUS model, the urban expansion of China under the baseline development, the rapid development, the harmonious development, and the slow development scenarios from 2010 to 2050 and the urban growth boundary of the Pearl River Delta region from 2020 to 2050 were simulated [28,29]. Compared with the common CLUE-S and CA models, previous studies have confirmed that the FLUS model has higher simulation accuracy and can simulate LUCC under a variety of development strategies [26]. Therefore, the FLUS model has more advantages in the simulation of land use change.

The FLUS model has previously been used to simulate urban growth boundaries [30], to simulate LUCC under multiple scenarios [27,31], and to study ecosystem services in combination with models such as the InVEST model [32]. However, fewer studies have analyzed and compared land use system characteristics such as key land use types under different scenarios based on FLUS model simulation results. The complex network model reveals the relationship between nodes in the network from the perspective of the network as a whole, and can be used to analyze key land use types by calculating the betweenness centrality and the overall stability of the land use system by calculating the average shortest path length [33]. It has been used to study the efficiency of information dissemination and language translation [34,35], but has rarely been used to analyze and compare the characteristics of land use systems in different scenarios. Therefore, the complex network model was chosen to analyze overall land use characteristics in this paper.

In this study, we simulated the land use status under the natural scenario by using the Markov and FLUS models and analyzed the LUCC under the two scenarios accordingly. To quantitatively assess the impact of policies on land use change in Chongqing, we used the relative contribution conceptual model (RCCM) to analyze the relative contribution of each land use type under two scenarios [27]. The key land use types and the stability of the land use system were analyzed based on a complex network model. In summary, the contribution of this paper is to distinguish between natural and policy scenarios, to quantitatively assess the relative contribution of policies to LUCC using RCCM, and to innovatively compare the overall characteristics of land use systems under natural and policy scenarios using a complex network model. Therefore, our research objectives were: (i) to simulate land use change based on a Markov-FLUS model; (ii) to quantitatively assess the relative contribution of policies to each land use type; (iii) to comparatively analyze the overall characteristics of the land use system; and (iv) to analyze the role of policy dimensions affecting LUCC. The research results can provide a theoretical basis for the next stage of relevant policy decisions in Chongqing.

2. Data and Methods

2.1. Study Area

Chongqing was one of the earliest approved central cities and is an important financial center located in the upper reaches of the Yangtze River. Located in southwestern China, Chongqing is a municipality directly under the jurisdiction of the central government, with an area of 82,400 km² and over nearly 40 districts and counties (Figure 1). The mountainous areas in Chongqing account for more than 70% of the total area. The geographical environment is complex, the overall forest coverage rate exceeds 50%, the water and air quality are good, and the area is suitable for living. With a permanent population of more than 32 million, the gross domestic product (GDP) of the region exceeds RMB 2.5 trillion, and the mileage of expressways open to traffic is more than 3500 km (https://www.cq.gov.cn/zjzq/ (accessed on 10 January 2022)).

2.2. Data Sources

Land use data was acquired from the Resource and Environment Data Cloud Platform (http://www.resdc.cn (accessed on 10 January 2022)). The selected drivers of the FLUS model in this paper are shown in

Table 1. A description of the data used in the model.

Category	Data	Spatial Resolution	Data Resource
Terrain	Aspect	1000 m	Calculated according to DEM
	Slope	1000 m	Calculated according to DEM
Soil	Silt		CAS (http://www.resdc.cn (accessed on 10 January 2022))
Temperature	Annual average temperature	1000 m	CAS (http://www.resdc.cn (accessed on 10 January 2022))
Precipitation	Annual precipitation	1000 m	CAS (http://www.resdc.cn (accessed on 10 January 2022))
Human influence	Population density	1000 m	CAS (http://www.resdc.cn (accessed on 10 January 2022))
	GDP	1000 m	CAS (http://www.resdc.cn (accessed on 10 January 2022))
	Distance to railway	Vector	GDEMDEM (http://www.gscloud.cn/ (accessed on 10 January 2022))
	Distance to road	Vector	GDEMDEM (http://www.gscloud.cn/ (accessed on 10 January 2022))
	Distance to residential area	Vector	GDEMDEM (http://www.gscloud.cn/ (accessed on 10 January 2022))

and Figure 2. Our process is described as follows: preprocess raster data and unify raster information, including resolution, number of columns and rows, and projection. In this paper, we simulated the land use status in 2000 based on the land use status in 1995 to verify the accuracy of the simulation results of the FLUS model in the absence of a policy context.

2.3. Method

2.3.1. Research Framework

First, we found the FLUS model parameter settings that achieved the appropriate kappa coefficient values in the natural context. In order to simulate land use conditions under natural scenarios, we needed to predict land use demand and then predict the spatial distribution of land use. Based on the relevant data from 1990 to 1995, Markov models were used to predict the amount of land use in 2000, and then the prediction results were input into the FLUS model to set appropriate parameter values to make the spatial distribution simulation results more accurate. Second, the LUCCs of previous years as well as the quantitative assessment of the policy effect on LUCC were analyzed. Based on the land use data from 1990–2000, a Markov model was used to predict the land use demand in 2020 under the natural scenario. The Markov model predictions were then input into the FLUS model to simulate the spatial distribution of land use. The RCCM model was used to compare and analyze the contributions of various land use types in natural and policy contexts in 2020 [27]. Third, a complex network model was used to analyze the overall characteristics of land use in different time periods during 1990–2020. Fourth, we analyzed the role of policy dimensions affecting LUCC. The research flowchart is shown in Figure 3.

2.3.2. Scenario Definitions and Driving Factors

Natural scenarios refer to scenarios that are less affected by policies, and policy scenarios refer to scenarios that are affected by both natural scenarios and policies. The GFGP, Western China Development Strategy, and other strict land conservation and intensive use policies were all implemented after 2000, and the TGDP was completed in 2009. Since 2000, these policies have had a significant impact on land use change and have been in place for a long time, so the policy scenario in this paper mainly considers the impact of these policies. Therefore, this paper takes the year 2000 as the boundary. The policy scenarios refer to the scenarios influenced by the above four policies. The natural scenario refers to the scenarios that are not influenced by the GFGP or the Western China Development Strategy, and those that are less influenced by land use conservation and intensification policies and TGDP.

LUCC before 2000 was mainly influenced by natural background changes. In this paper, the real land use status of 1990–2000 is in the natural context, and the real land use status uses land use data downloaded from the Resource and Environment Data Cloud Platform (http://www.resdc.cn (accessed on 10 January 2022)). The land use status in 2020 in the natural context was simulated based on the real land use status in 1990–2000. The land use status in 2020 in the policy context is the actual situation, i.e., the land use data in 2020 in the policy scenario uses the real values in 2020 downloaded from the Resource and Environment Data Cloud Platform (http://www.resdc.cn (accessed on 10 January 2022)).

This research used the receiver operating characteristic (ROC) curve in SPSS 25 software to discern the correlation between each land use category and the drivers. Considering that both human and natural factors have an impact on LUCC as well as referring to the drivers chosen by other scholars [26,27,29,36], we selected 10 relevant influencing factors commonly used in the FLUS model as the drivers of the model, and the relevant data are shown in Table 1. If the receiver operating curve (AUC) values under the ROC curve is greater than 0.7, the selected factors are considered to have a relatively strong explanatory power [37]. The results show that the AUC values are all greater than 0.7 (Figure A1). Thus, the 10 factors we selected can better explain the LUCC [37].

2.3.3. Markov Chain Model

The Markov chain model can establish a transition probability matrix based on the land use data of different years [38]. The greater the transition probability, the more likely a transition will occur, and based on this matrix, it can predict the land use demand in the next period [39,40,41,42]. It does not ignore the description of the change process when simulating land use changes [43]. Both the transfer matrix and the transfer probability matrix can be calculated by the Markov module in the FLUS software package [26].

The expressions describing the probability of land use change in different periods are given as follows:

$$P_{ij}=\frac{n_{ij}}{n_i}$$

(1)

$$\begin{array}{c}{P}_{ij}=\left[\begin{array}{cc}\begin{array}{cc}{P}_{11}& P_{12}\\ P_{21}& P_{22}\end{array}& \begin{array}{cc}\cdots & P_{1k}\\ \cdots & P_{2k}\end{array}\\ \begin{array}{cc}⋮& ⋮\\ P_{k1}& P_{k2}\end{array}& \begin{array}{cc}\ddots & ⋮\\ \cdots & P_{kk}\end{array}\end{array}\right]\\ 0\le P_{ij} \le 1,{\displaystyle \sum }_{j=1}^k{P}_{ij}=1(\mathrm{I},j=1,2,3....,k)\end{array}$$

(2)

$n_i$ is the number of pixels of category $i$ that have changed. $n_{ij}$ is the number of pixels transformed from category $i$ to $j$. $k$ represents the type of land use.

From the land use situation at time t ($S_t)$, the situation at time t+1 ($S{ }_{t+1})$ can be predicted [20,44,45]:

$$S{}_{t+1}=S_t\times P_{ij}$$

(3)

2.3.4. FLUS Model

The FLUS model combines the SD and CA models to calculate the occurrence probability of each land type on each plot through factors, such as climate and human activities; the model also solves the competition problems of different land types through a certain mechanism and finally obtains the future land use status [26]. The FLUS model performs well when solving the problem of competition among various categories and has been proven to have a high level of simulation accuracy in previous studies [19,26,28,29,36,46]. The model requires the input of the number of land use projections, and in this paper, the results of Markov model projections were input [26]. All operations of the spatial land use distribution simulation in this paper were done through the FLUS module in the FLUS software package [26], the user’s guide of which explains in detail the operational details, including the input of Markov values, and is very useful (http://www.geosimulation.cn/FLUS.html (accessed on 10 January 2022)) [26]. Therefore, the detailed operation procedures will not be discussed in this paper.

2.3.5. Relative Contribution Conceptual Model (RCCM)

The RCCM can quantitatively evaluate the contribution of various land use types in different contexts [27]. If government policy guidance is added to the natural context, the RCCM can be used to evaluate the effect of policy implementation on LUCC [27]. Previous studies mainly compared the overall changes in land use, did not compare specific land types, and did not distinguish between the two backgrounds of no policy and policy [47]. In contrast, this paper distinguishes between natural and policy contexts to obtain the relative contribution of specific land categories to LUCC in a policy context. The absolute value of the change in a land type under a certain background is divided by the sum of the absolute value of the change in the land type under two backgrounds to obtain the relative contribution value [27].

In Figure 4, the solid blue line in the $T_1–T_3$ time period represents the change in the number of one land use type under the natural scenario. In other words, the trend in the number of land use types in the $T_2–T_3$ time period is the same as in the $T_1–T_2$. The solid red line in the $T_2–T_3$ time period indicates the change in the number of one land use type under the policy scenario. Land use change begin to be subject to policy intervention at time $T_2$, resulting in the deviation of the LUCC from the natural scenario in the policy scenario. The upward arrow on the right side of Figure 4 indicates the positive contribution of this land use type to LUCC, while the opposite indicates its negative contribution. The green and yellow arrows indicate the change in the quantity of a land use type caused by a natural scenario and a policy scenario, respectively, and the absolute value of this change in quantity is the contribution of a scenario to a land use type. The percentage of the relative contribution of the natural scenario or the policy scenario to LUCC can be obtained by dividing the absolute value of the change in the quantity of the land use type caused by the natural scenario or the policy scenario by the sum of the absolute values of the change in quantity under the two scenarios [27]. Figure 4a indicates that both policy and natural contexts contribute positively to changes in land use types. Figure 4b indicates that the positive contribution of the natural context is not enough to offset the negative contribution of the policy context, which eventually makes the amount of this land use type in the natural context higher than the amount in the policy context. The natural context in Figure 4c,d contributes negatively to the change in land use type [27].

2.3.6. Complex Network Model

The complex network model is composed of nodes representing different land use types and edges with conversion weights. The direction of the edge indicates the direction of conversion, and the weight of the edge indicates the amount of conversion. The weighted in-degree and the weighted out-degree represent the land use amount at the end point and the initial moment, respectively. Using Gephi 0.9.2 to build a complex network model, we could comprehensively study land use changes during this period, including stability and key land types.

(1)

Node degree

The number of related directed edges on a node is defined as a degree [48,49]. The more related edges there are, the more closely connected the node is with other nodes [50,51]. The degree can be used as an auxiliary index to judge the importance of nodes.

$$k_i={\displaystyle \sum }_{j\in N}{A}_{ij}$$

(4)

$A_{ij}$ is a matrix used to express the correlation of nodes. When node I is related to node j, $A_{ij}$=1; otherwise, it is 0.

(2)

Betweenness centrality

Nodes can play a role in transmitting information to other nodes in the network, and the betweenness reflects this influence of nodes [52]. When measuring the importance of a node, it is not feasible to rely solely on the degree of the node. This also needs to be based on the betweenness centrality of the node [52].

$$Cb\left(i\right)={\displaystyle \sum }_{j\ne k}{g}_{jk}\left(i\right)/g_{jk}$$

(5)

where Cb(i) is the value of betweenness centrality, $g_{jk}$ is the number of shortest paths between nodes j and k, and $g_{jk}\left(i\right)$ needs to pass through node $i$ on the basis of $g_{jk}$.

(3)

Average shortest path length

There are many ways to reach any two nodes, and the path with the smallest average distance is the average shortest path [53,54]. This indicator reflects the connection efficiency between nodes and can measure the stability of the entire system.

$$L=\frac{1}{N\left(N-1\right)}{\displaystyle \sum }_{i\ne j}{d}_{ij}$$

(6)

where $d_{ij}$ is the edge between nodes$i$ and $j$, and $N$ is the number of land use types [55].

2.3.7. Urban Expansion Elasticity Factor

The elasticity of the urban expansion coefficient was used to describe the harmonious relationship between the growth rate of urban built-up areas and the growth rate of the urban population.

$$R_{\left(i\right)}=\frac{A_{\left(i\right)}}{Po{p}_{\left(i\right)}}$$

(7)

where $R_{\left(i\right)}$ represents the elasticity coefficient of urban expansion, $A_{\left(i\right)}$ is the annual average urban built-up area growth rate of city $i$, and $Po{p}_{\left(i\right)}$ is the annual average urban population growth rate of city $i$. $R_{\left(i\right)}$ = 1.12 indicates that the urban built-up area expansion and population growth are in a coordinated state, $R_{\left(i\right)}$ > 1.12 indicates that the expansion of built-up areas is too fast, and $R_{\left(i\right)}$ < 1.12 indicates that the urban population growth is too fast [56,57].

3. Results

3.1. Simulation Accuracy Evaluation

This paper used relevant data from 1990 to 1995 to simulate the LUCC in 2000 to verify the simulation accuracy of the FLUS model in the context of no policies. Based on the land use data from 1990 to 1995, the Markov model was used to predict the 2000 data, the results were input into the FLUS model, and the relevant parameters were set to obtain the land use status map in 2000 (Figure 5b). The accuracy of the land use simulation in 2000 was verified by calculating the kappa coefficient (

Table 2. Simulation accuracy of the FLUS model in a no policy context.

	Producer Accuracy	User Accuracy	Kappa Coefficient	Overall Accuracy
Cultivated land	0.990254	0.993056	0.978426	0.986650
Forest	0.991205	0.987987
Grassland	0.980322	0.974222
Water area	0.988095	1
Construction land	0.726471	0.730159
Unused land	1	1

). The kappa coefficient and overall accuracy were both higher than 0.95, and all producer accuracies and user accuracies were higher than 0.7. It is generally considered that the simulation results are better if the kappa coefficient is greater than 0.8 [25]. Therefore, the setting of the FLUS model parameters in the natural context met the accuracy requirements, while the rationality of the driver selection was again verified.

3.2. The Relative Contributions of Natural and Policy Contexts to LUCC

The FLUS model was used to simulate the LUCC in the 2020 natural scenario based on land use data from 1990–2000, and the real land use status map for 2020 was used as the land use data for the policy scenario, with the use of real data from the Resource and Environment Data Cloud Platform (http://www.resdc.cn (accessed on 10 January 2022)). As shown in

Table 3. Quantitative status of land use types in 2020 under two scenarios (unit: km²).

	Cultivated Land	Forest	Grassland	Water Area	Construction Land	Unused Land
Policy background	37,504	33,575	7608	1312	2399	12
Natural background	38,381	30,407	11,747	929	925	15

, compared with the natural background, arable land was reduced by 877 km², forestland increased by 3168 km², grassland decreased by 4139 km², and construction land increased by 1474 km² under the policy background. Due to the different urbanization processes in different time periods, there was a large difference between the predicted value of the amount of construction land and the real value. The main changes in both contexts for the 2000–2020 period were a decrease in arable land and grassland and an expansion of forest and building land.

The RCCM was used to analyze the contribution of various land use types in natural and policy contexts from 2000–2020 [27]. The absolute value of changes in the quantity of land use types due to the natural scenario or the policy scenario divided by the sum of the absolute values of the quantity change under the two scenarios gives the relative percentage contribution of the natural scenario or the policy scenario to LUCC [27]. The policy scenario was unfavorable to the expansion of cultivated land. In 2020, arable land was reduced by 877 km² in the policy scenario compared to the natural scenario, generating a negative contribution of 73.82% from 2000–2020 (Figure 6a). The natural scenario produced a negative contribution of 26.18% to the arable land. The positive impact of these policies on woodlands was 98.08% (Figure 6b), with an expansion of 3168 km² compared to the no policy context. The negative impact of these policies on grasslands was 98.08% (Figure 6c), the positive contribution to watershed areas was 99.48% (Figure 6d), and the positive contribution to built-up land was 81.8% (Figure 6e). In summary, these policies have had a significant impact on the changes in various land use types.

The theoretical contributions of the land use types on each other from 2000–2020 under both scenarios are shown in Figure 7. Both the natural scenario and the policy scenario showed a decreasing trend for arable land and grassland in Chongqing and an increasing trend for forest, watershed, and construction land. There was little change in land use under the natural scenario, and only cropland and building land contributed more to each other. The natural scenario showed a decreasing trend in arable land, which was mainly converted to construction land. The reduction in cropland was more obvious in the policy scenario, and this reduction was mainly due to the conversion of cropland to construction land, while the amount of grassland replenished to cropland was higher but not enough to offset the reduction in cropland. In both scenarios, cropland and grassland contributed positively to woodland, and both cropland and woodland contributed negatively to grassland, while the growth of watershed and building land depended on the positive contribution of cropland.

3.3. LUCC between 1990–2000 and 2000–2020

A number of important policies implemented after 2000 had a significant impact on the quantity and distribution of various types of land use. In this paper, we argued that LUCC was not affected by policy contexts such as the GFGP prior to 2000, but it was affected by these policy contexts after 2000.

The transition matrix showed the changes in LUCC in Chongqing for different time periods from 1990 to 2020 (

Table 4. 1990–2020 Land Use Transfer Matrix (unit: km²).

Scenario		Cultivated Land	Forest	Grassland	Water Area	Construction Land	Unused Land	Losses *
1990–2000 (reality)	Cultivated land	38,615	74	13	2	146	0	235
	Forest	42	30,101	152	1	17	0	212
	Grassland	33	170	11,663	2	2	0	207
	Water area	2	0	0	922	2	0	4
	Construction land	0	0	0	0	430	0	0
	Unused land	0	0	0	0	0	15	0
	Gains *	77	244	165	5	167	0
2000–2020 (under policy scenarios)	Cultivated land	25,172	8941	2363	538	1635	10	13,487
	Forest	7942	20,207	1682	224	248	1	10,097
	Grassland	3893	4238	3501	97	85	1	8314
	Water area	313	100	37	400	76	0	526
	Construction land	143	46	9	46	353	0	244
	Unused land	8	2	0	4	1	0	15
	Gains *	12,299	13,327	4091	909	2045	12
2000–2020 (under natural scenarios)	Cultivated land	37,880	54	26	275	457	0	812
	Forest	31	30,304	10	0	0	0	41
	Grassland	68	49	11,711	0	0	0	117
	Water area	270	0	0	647	10	0	280
	Construction land	132	0	0	7	458	0	139
	Unused land	0	0	0	0	0	15	0
	Gains *	501	103	36	282	467	0

* Gains means the amount of increase in area of one land use type due to the transfer in of other land use types. Losses means the amount of decrease in area due to the transfer out of one land use type to other land use types.

), including the increases and decreases in each type. Areas where no land type change occurred were represented by the values on the diagonal of the matrix. The spatial distribution of land use under the natural and policy scenarios from 1990 to 2020 is shown in Figure 8. During the 1990–2000 period, except for a decrease in the total amount of cultivated land by 158 km² and an increase in the total amount of built-up land by 167 km², the area of other land types did not change significantly, and the total change area was 658 km².

The more obvious and rapid changes in the policy context from 2000 to 2020 are described as follows: arable land and grassland decreased, and forestland and construction land increased, with a total change by 32,683 km². Arable land decreased by 13,487 km², which was more significant compared to the last decade. The reduced arable land was replaced by forestland (8941 km²), grassland (2363 km²), water (538 km²), and construction land (1635 km²), which was related to the implementation of the GFGP, the TGDP, and others. During this period, with accelerated urbanization and the development of industry and commerce, there was a large influx of rural population into cities, and construction land increased greatly. The degradation of grasslands was more significant than their expansion, so the grasslands eventually saw a decrease in quantity, and cultivated land and forestland replaced most grasslands. In contrast, less transformation occurred in various land use types in the natural context from 2000–2020, and the watershed was basically unchanged. The expansion of woodland occurred mainly in the southwestern part of Chongqing, where the land previously surrounding the city was replaced by a large amount of built-up land. Comparing the LUCC for 1990–2000 and the policy scenario for 2000–2020, the results showed that the main changes were a more pronounced trend of decreasing cropland and grassland and accelerating growth in terms of building and forest land.

Based on the amount of various land use types from 2005 to 2020, a Markov model was used to project the amount of land use under the policy scenario in 2035. The land use data were updated to 2020, but the data of relevant influencing factors, such as GDP and population in 2015, were the latest data; thus, 2020 was not appropriate as a model year, and we only predicted the land use quantity in 2035 without modeling the land use distribution status. From

Table 5. Quantitative status of land use types in 2035 under the policy scenario.

	Cultivated Land	Forest	Grassland	Water Area	Construction Land	Unused Land
Policy background	36,861	34,000	6548	1557	3427	12

, we can see that if the current policy continues to be implemented, arable land and grassland will be further reduced in the future, while forestland, the water area, and construction land will continue to increase; therefore, the avoidance of the inefficient use of construction land and ensuring food security in the future are still key issues.

3.4. Analysis of Land Use Changes Based on Complex Network Models

The complex network model can be used to represent the circulation of different land use types. Six nodes represent six land use types, and the greater the weight of directed edges between nodes, the greater the conversion area. From 1990 to 2000, all land use types except unused land were involved in land use area conversion, and the conversions with higher weights were the interconversion between forest land and grassland and the conversion of cropland to construction land. However, on the whole, all land use conversions from 1990 to 2000 had little weight. Under the policy scenario, all land use types were involved in land use area conversion from 2000 to 2020, with high weight conversion occurring between the three land use types of cropland, forest land, and grassland. Under the natural scenario, a high weight conversion occurred between arable land and construction land and between arable land and watershed in 2000–2020. From 2000 to 2020, compared with the natural background, more arable land and grassland were converted to forest under the policy background, and the conversion of other land use types to water was accelerated (Figure 9). If the ratio of the weighted output degree to the weighted input degree is greater than 1, it indicates that the amount of this kind of land is reduced, which is represented as output land; otherwise, it is input land (

Table 6. Weighted in-degree and weighted out-degree of complex network nodes in 1990–2020 (unit: km²).

Scenario	Land Use Types	Output Degree	Input Degree	Output Degree/Input Degree
1990–2000 (reality)	Cultivated land	38,850	38,692	1.004
	Forest	30,313	30,345	0.999
	Grassland	11,870	11,828	1.004
	Water area	926	927	0.999
	Construction land	430	597	0.720
	Unused land	15	15	1.000
2000–2020 (under policy scenarios)	Cultivated land	38,659	37,464	1.032
	Forest	30,304	33,534	0.904
	Grassland	11,815	7592	1.556
	Water area	926	1309	0.707
	Construction land	597	2398	0.249
	Unused land	8	12	0.667
2000–2020 (under natural scenarios)	Cultivated land	38,692	38,381	1.008
	Forest	30,345	30,407	0.998
	Grassland	11,828	11,747	1.007
	Water area	927	929	0.998
	Construction land	597	925	0.645
	Unused land	15	15	1.000

) [58]. From 1990 to 2000, except for the fact that construction land was obviously input land, the ratio of weighted output to the weighted input of other land was close to 1 and showed little change. From 2000 to 2020, forest land, water area, and construction land were all input land under the policy background and the natural background, but the ratio of weighted output degree to weighted input degree under the policy background was smaller, indicating a more significant increase in quantity.

The degree reflects the closeness of the connection between the node and other nodes [48]. Both the degree and betweenness centrality can identify important land use types, but relying solely on degree is not rigorous [58]. The key land use categories were cultivated land in 1990–2000, cultivated land and forest land under the policy background, and cultivated land under the natural background for 2000–2020. Therefore, arable land has a certain importance in all periods (

Table 7. Betweenness centrality of land use changes in different periods.

Land Use Type	1990–2000	2000–2020 (Policy Background)	2000–2020 (Natural Background)
Cultivated land	2.00	0.92	8.0
Forest	0.00	0.92	0.0
Grassland	0.00	0.67	0.0
Water area	0.00	0.25	0.0
Construction land	0.00	0.25	0.0
Unused land	0.00	0.00	0.0

and

Table 8. Degree of land use type.

Land Use Type	1990–2000	2000–2020 (Policy Background)	2000–2020 (Natural Background)
Cultivated land	9	12	10
Forest	8	12	6
Grassland	8	11	6
Water area	7	11	6
Construction land	6	11	6
Unused land	2	7	2
Average node degree	3.333	5.33	3

).

Realistic land use data from 1990 to 2020 showed that the average shortest path value decreased from 1.125 to 1.1, indicating a slight decrease in the stability of the land use system. From 2000 to 2020, the average shortest path value of the natural context was larger than that of the policy context, and the land use system stability was higher (

Table 9. Average shortest path and average node degree in different periods.

Period	Average Shortest Path
1990–2000	1.125
2000–2020 (Policy background)	1.1
2000–2020 (Natural background)	1.4

).

3.5. Coordination Analysis of Urban Expansion and Population Development in Chongqing

Due to the availability of data, we started to analyze the changes in the elasticity coefficient of urban expansion in 2006. The research of the China Institute of Urban Planning and Design concluded that the ratio of the average annual growth rate of urban built-up areas to the average annual growth rate of urban population is 1.12, which is a reasonable value [56,59]. The overall urban expansion elasticity coefficient of Chongqing is 2.37 for the 2006–2020 period, which is higher than 1.12, indicating that the city is expanding too quickly (

Table 10. Built-up Area and Urban Population in Chongqing, 2006–2020.

	2006	2010	2015	2020
Built-up area	631.35	870.23	1329.45	1565.61
City population	747.02	831.69	1032.63	1213.56

and

Table 11. Growth in urban built-up areas and urban population growth.

	Average Annual Growth Rate of Urban Built-Up Area (%)	Average Annual Urban Population Growth Rate (%)	Elasticity Factor
2006–2010	9.46	2.83	3.34
2010–2015	10.55	4.83	2.18
2015–2020	3.55	3.50	1.01
2006–2020	10.57	4.46	2.37

). This result indicates that the rate of urban land expansion during 2006–2020 did not match the rate of urbanization, and more land was converted to non-agriculturalized land, leading to the idle and inefficient use of construction land [59]. Three time periods, 2006–2010, 2010–2015, and 2015–2020, show a decreasing trend in the elasticity coefficient of urban expansion and gradually tend to become reasonable.

4. Discussion

4.1. The Three Gorges Dam Project (TGDP)

As shown in Figure 6d, the positive contribution of the policy scenario to the water area is 99.48%. The Three Gorges Dam project has led to significant land use changes in Chongqing, which are mainly manifested in this paper as a significant increase in water area (Table 4). The Three Gorges Dam, located on the main stream of the Yangtze River between Chongqing and Yichang, China, is the largest hydrological hub project in the world today. The Three Gorges Reservoir Area (Chongqing section) includes 22 districts and counties in Chongqing, accounting for 85.6% of the total area of the reservoir area, and most of the Three Gorges Reservoir Area is located in the Chongqing section. The TGDP required 17 years to complete, from 1993 to 2009, including three engineering construction periods. This dam can withstand mega-flood disasters, provides power resource support for the region, and significantly improves the navigability of the Yangtze River while reducing shipping costs. However, the Three Gorges project has resulted in the migration of millions of people to the reservoir area and has had an impact on the ecological environment [60,61]. The first phase of the project was dominated by the interception of the Yangtze River. The second phase of construction lasted from 1998 to 2003, resulting in the construction of a number of engineering facilities. The third phase entailed the completion of the construction of remaining works [61]. During 1990–2000, when the Three Gorges Dam was not yet completed, there was no measurable change in water area. The water area changed significantly from 2000 to 2020, and the water area in 2020 in the policy context was 383 km² more than that in the natural context (Table 3). The increase in water area was mainly from arable land and forest land, with no significant change in water area in the natural context (Table 4). As shown in Table 8, the nodal degree value of water area in the 2020 policy scenario is higher than in the natural scenario, and the importance of water area in the process of land use change increases. If the current development trend is maintained, the watershed area will continue to increase between 2020–2035 (Table 5). Therefore, the government should guard against the potential adverse effects of the TGDP on land use change.

4.2. China’s Western Development Strategy

As shown in Figure 6e, the positive contribution of policies to construction land is 81.8%, which shows that policies have a significant impact on changes in the amount of construction land. Construction land in Chongqing has grown rapidly over the past 20 years, with some of the growth in construction land caused by increased infrastructure (Table 4). With accelerated urbanization, the increase in construction land has become an irreversible trend. Since 2000, relevant national policies, such as the Western China Development Strategy, have given more resources and opportunities to the west, and a large number of rural people have flocked to cities, promoting the urbanization process [62]. China implemented the Western Development Strategy in early 2000 to promote the development of the western region (including the Chongqing Municipality) and narrow the gap between the western region and other regions through the implementation of several policies. Infrastructure development is one of the key tasks of the Western Development Strategy, and the central government has invested a great deal and achieved positive results in this regard [63]. The construction of major railroads and highways to build a transportation network system has made the western region more accessible, and the construction land in Chongqing increased by 1474 km² in the policy scenario in 2020 compared to the natural scenario (Table 3), part of which was due to the increase in infrastructure for the Western Development Strategy [63]. The built-up area of Chongqing increased during the 2006–2020 period, and construction land has played an important role in land use changes (Table 8 and Table 10). Therefore, when implementing the Western China Development Strategy, the government should consider how well the urbanization process matches the growth rate of land for construction.

4.3. Economical and Intensive Land Use

China first proposed the policy of the economical use of land in 1953 and has been deepening and improving this policy since that time. In 2008, the State Council issued the Circular on Promoting the Economical and Intensive Use of Land, which proposed the intensive and economical use of land in the future. The 18th Party Congress emphasized the “economical and intensive use of resources”. In this context, the Regulations on the Economical and Intensive Use of Land were introduced in 2014, and the content of the system of economical and intensive use of land began to form. In 2021, the newly revised Regulations for the Implementation of the Land Management Law called for improving the efficiency of construction land use and promoting the economical and intensive use of land in terms of various aspects. As shown in Table 11, the overall expansion elasticity coefficient of Chongqing city during 2006–2020 is 2.37, which is much higher than the reasonable value of 1.12, indicating that urban expansion is too rapid and that the growth rate of construction land does not match the urbanization process. The amount of construction land under the policy scenario in 2020 is 1474 km² more than that under the natural scenario (Table 3), but some of the increased construction land is idle and being used inefficiently. The elasticity coefficients of urban expansion in the three time periods of 2006–2010, 2010–2015, and 2015–2020 show a decreasing trend and gradually tend to be reasonable, indicating that the excessive expansion of construction land has been alleviated and that the policy of economical and intensive utilization of construction land has achieved certain results (Table 11). The government should continue to implement and improve this policy in the future.

4.4. The Grain for Green Program (GFGP)

The Grain for Green program is one of the most important ecological restoration projects in China, with a wide range of impacts and a large scale of investment. The first round of the Grain for Green program was implemented at the end of 1999, and the second round was launched in 2014. Before 2000, due to urbanization, some arable land was taken away, and the demand for food as a result of population growth and other land types were blindly developed for agricultural production, which led to increased soil erosion and flooding. The project aims to improve the ecological environment by preventing the use of agricultural land that is unsuitable for cultivation and restoring vegetation in a planned manner. As shown in Figure 6, the positive impact of the policy scenario on forest land is 98.08%, and the negative impact on cultivated land is 73.82%. In other words, the increase in forest land and the decrease in cultivated land have a strong correlation with government policies. There are many policies that will reduce cultivated land, and the GFGP has had a more pronounced impact [62]. From 2000 to 2020, in the policy context, 8941 km² of arable land was converted to forestland and 2363 km² of arable land was converted to grassland in Chongqing, while the conversion of arable land to forestland and grassland in the natural context was 54 km² and 26 km², respectively (Table 4). From 2000 to 2020, arable land and forestland were the key land use types in the policy context, while the key land use type in the natural context was arable land (Table 7). The implementation of the GFGP highlighted the role of forestland in the process of land use change and greatly changed the land use status (Table 7 and Table 8).

A total of 31.49% of the reduced arable land from 1990 to 2000 was converted to forest land, and, from 2000 to 2020, this percentage reached 66.29%. Prior to the implementation of the GFGP, the irrational use of land destabilized the ecosystem. For example, massive deforestation increased dust storms. With the implementation of the GFGP and other policies, the area of forested land increased significantly after 2000, which was beneficial in the context of wind and sand control. From 2000 to 2020, the increase in forestland in the policy context was mainly from cropland and grassland (Figure 7). The GFGP has a positive effect on the ecological environment by contributing to soil and water conservation and increasing the area of forest land. The stability of the land use system in the 2000–2020 policy context was lower than that in the natural context (Table 9). Therefore, the implementation of policies should be tailored to local conditions, such as selecting suitable crop types and providing complete post-management, otherwise the policy will have a negative effect [64]. The arable land occupied by the GFGP is not highly productive and has little impact on food production [65]. However, the food problem is still a key issue for China, which has a population of 1.4 billion people. In the future, it is necessary to improve the level of food production technology, coordinate the relationship between cultivated land and forest land, and ensure a sufficient amount of cultivated land.

5. Policy Implications

5.1. Promotion of Natural Forest Restoration

Although the GFGP is worth continued implementation, we should still pay attention to native trees. We must not only avoid negatively affecting native trees in the process of planting and afforestation but also restore natural forests as much as possible [66]. Forest restoration can not only effectively increase forest coverage but also help to restore ecosystem functions at low cost [67]. Different restoration measures should be taken in consideration of specific factors, such as the previous use of the land, type of vegetation, and time when it was abandoned. The restoration of vegetation and the verification of results is a long process, and long-term effective solutions should be sought instead of only considering short-term results [68].

5.2. Local Conditions Should Be Considered When Afforestation Occurs

If deep-rooted trees are planted in an area with limited precipitation, the water table of the plot will be lowered and the local grassland will be degraded [69]. We should avoid changes in the structures of forests caused by tree invasion to prevent adverse effects on the local ecosystem, which are difficult to observe over a short-term period [70]. Although trees play a key role in improving the ecological environment, the factor of water resources should also be considered in the process of afforestation [71]. China has invested considerable labor and material resources in afforestation, but it is likely to exacerbate water shortages in arid areas and make the ecological environment more fragile. Therefore, we should not blindly pursue an increase in the number of trees and should choose appropriate afforestation areas and methods based on actual local conditions [72].

5.3. Efficient Use of Construction Land and Protection of Agricultural Land

In recent decades, land use has been greatly affected by human activities, and human factors in different local areas have caused different land use patterns [73]. One of the three major challenges in China’s urbanization process put forward in previous studies is the reduction in high-quality arable land [74]. Its large population puts China under pressure with regard to food shortages [75]. As construction land plays a greater role in economic growth than cultivated land, LUCC is moving toward nonagricultural land use. Economic development has made urbanization an irreversible trend, and future construction land will increase with further increases in the urbanization process. However, the current growth rate of China’s construction land does not match the corresponding urbanization process, resulting in the unreasonable expansion and inefficient use of construction land. The government should plan and use construction land rationally, avoid the abandonment of construction land and use abandoned land appropriately. The expansion of construction land has resulted in the occupation of much high-quality farmland, especially farmland located near cities. The productivity of arable land has increased with the improvement of related technologies, but the arable land area in Chongqing has been showing a decreasing trend in recent decades and cannot guarantee future food security in the long run; thus, we still need to protect arable land. In addition, the expansion of construction land does not necessarily require the sacrifice of arable land. We should use construction land effectively in order to match the growth of construction land with the urbanization process, avoid losses due to excessive increases in construction land, and bring the value of the elasticity coefficient of urban expansion back to a reasonable value.

6. Conclusions

This article examined the overall characteristics of land use change in Chongqing using a complex network model, quantified the relative contributions of four policies to LUCC using the RCCM, and investigated the role of policy dimensions influencing LUCC by discriminating between natural and policy scenarios. The simulation results of the Markov and FLUS models revealed a considerable shift in land use in Chongqing from 2000 to 2020, with forest land, water area, and building land expanding greatly under the policy scenario, while arable land and grassland decreased. The policies had negative contributions of 73.82% and 98.08% to cropland and grassland, respectively, and positive contributions of 98.08% and 99.48% to forest land and watershed, respectively. Cropland was the most important form of land use at various times, and the stability of the land use system under the policy scenario was lower than that under the natural scenario, resulting in the state of land use change being unstable. All four main policies had a significant impact on LUCC. The TGDP mainly led to an increase in water area, China’s Western Development Strategy increased infrastructure construction, which led to a partial increase in construction land, and the GFGP had a significant impact on changes in cultivated land, forest land, and grassland. The policy of the economical and intensive utilization of construction land has been effective in mitigating the excessive expansion of construction land. The implementation of relevant policies will have important implications for LUCC and will need to be further refined to achieve better results in the future. While additional land for construction is needed in the future to support the growing urban population, it is also necessary to maintain sufficient arable land and match urban expansion with population growth.