EN中文

Journal of Resources and Ecology >

2022 , Vol. 13 >Issue 3: 394 - 406

DOI: https://doi.org/10.5814/j.issn.1674-764x.2022.03.005

Land Use and Sustainable Development

Patterns and Driving Forces of Cropland Abandonment in Mountainous Areas

  • CHEN Shuanglong , 1, 2 ,
  • SONG Wei , 1, * ,
  • LI Han 1, 3 ,
  • LI Han 1, 4
Expand
  • 1. Key Laboratory of Land Surface Pattern and Simulation, Institute of Geographic Sciences and Natural Resources Research, Chinese Academy of Sciences, Beijing 100101, China
  • 2. College of Electrical and Information, Northeast Agricultural University, Harbin 150030, China
  • 3. College of geography and Environment, Shandong Normal University, Jinan 250358, China
  • 4. College of Resources and Environment, University of Chinese Academy of Sciences, Beijing 100049, China
*

CHEN Shuanglong, E-mail:

Received date: 2020-11-14

  Accepted date: 2021-08-20

  Online published: 2022-04-18

Supported by

The National Natural Science Foundation of China(42071233)

The Project of the Second Tibetan Plateau Scientific Expedition and Research(2019QZKK0603)

The Project of the Strategic Priority Research Program of Chinese Academy of Sciences(XDA20040201)

Fold

Abstract

Cropland abandonment is spreading from developed countries to developing countries such as China. Cropland abandonment in China commonly occurs in mountainous areas due to their specific natural and geographical conditions. However, due to the lack of dependable monitoring methods via medium-high-resolution remote sensing images, the scale of abandoned cropland in many mountainous areas of China is unclear, and the mechanisms driving cropland abandonment have not been clearly identified. To overcome these limitations, we took Zhong County of Chongqing in China as an example, and used Landsat 8 OLI_TIRS remote sensing image data to develop a method for mapping abandoned cropland in mountainous areas based on annual land use change monitoring. At the same time, the ridge regression method was adopted to analyze the factors influencing cropland abandonment. These analyses showed that the cropland abandonment rate in Zhong County of Chongqing was as high as 7.86%, while the overall accuracy of identifying abandoned cropland was as high as 90.82%. Among the social and economic factors that affect cropland abandonment, the rural population, economic development, and livestock husbandry development were the most important ones. At the land parcel scale, large-scale cropland abandonment occurred in areas at elevations above 650 m or with slopes of more than 15°.

Key words： cropland abandonment; remote sensing; driving factors; Zhong County; China

Cite this article

CHEN Shuanglong , SONG Wei , LI Han , LI Han . Patterns and Driving Forces of Cropland Abandonment in Mountainous Areas[J]. Journal of Resources and Ecology, 2022 , 13(3) : 394 -406 . DOI: 10.5814/j.issn.1674-764x.2022.03.005

1 Introduction

Globally, cropland abandonment is widespread in developed and developing countries (Díaz et al., 2011; Vacquie et al., 2015; Smaliychuk et al., 2016; Kolecka et al., 2017; Yu and Lu, 2018). In Europe, 33%, 18%, 33%, 22%, and 32% of the cropland areas were abandoned in Russia, Belarus, the Ukraine, Kazakhstan, and the Czech Republic, respectively, from 2000 to 2010 (Lieskovský et al., 2015; Lesiv et al., 2018). In South America, 45.05% of the cropland in Chile was abandoned from 1985 to 2007 (Díaz et al., 2011), while in Asia, cropland abandonment in China (Shi et al., 2018), Japan (Kamada and Nakagoshi, 1997), and Thailand (Lambin and Meyfroidt, 2010) has frequently been reported in recent years.
Currently, data on cropland abandonment in China are largely lacking, and the acquisition of such data mainly relies on rural household surveys (Yan et al., 2016). Although this approach is accessible, it is limited by the high survey costs and fails to map the overall cropland abandonment of a given study area. In addition, considering the influences of the survey techniques used by different investigators and the reliability of the sample selection, there is some uncertainty in the reliability of the survey results. In recent years, remote sensing technology has shown advantages for monitoring the spatial-temporal evolution of large-scale land use and land cover changes, and this approach has been widely used to acquire data on abandoned cropland. For example, Estel et al. (2015) assessed the cropland abandonment status of Europe using MODIS data, while Alcantara et al. (2012) used MODIS data to assess the cropland abandonment status of the Baltic States, Russia, and Belarus, and found that the area of abandoned cropland accounted for 15% of the total study area. Nevertheless, the authors observed that the low resolution of MODIS data (220-230 m) led to larger errors when identifying the abandoned cropland in regions with high parcel fragmentation.
To address the above problem, some authors have used remote sensing images with a resolution of 30 m to identify the abandoned cropland. For example, Prishchepov et al. (2012) and Dara et al. (2018) extracted the abandoned cropland plaques in some regions of Belarus, Lithuania, Poland, and Kazakhstan, and found that the abandoned cropland plaques in the regions with fragmented terrain could be well identified. However, in some regions with extremely fragmented terrain, such as the ravine zones with serious soil erosion, remote sensing images with a resolution of 30 m were inadequate for extracting abandoned cropland plaques. The increasing availability of aerial images provides more accurate data for the extraction of abandoned cropland in these regions. For example, Kolecka et al. (2017) employed the images taken by unmanned aerial vehicles to establish three-dimensional point cloud data and successfully extracted the abandoned cropland data, achieving a good recognition effect. However, this method has complicated image data processing requirements and numerous parameters, so it is difficult to apply this method in large areas.
Cropland abandonment is jointly influenced by socioeconomic factors and changes in natural conditions. On the macro level, changes in socioeconomic conditions are generally considered to be the main factors of cropland abandonment (MacDonald et al., 2000; Romero-Calcerrada and Perry, 2004). In particular, the transfer of agricultural labor to non-agricultural employment (Khanal and Watanabe, 2006) and the aging of the rural labor force that is left-behind make the marginal cropland with poor production conditions and low production efficiency face high abandonment risks (Nikodemus et al., 2005). In addition, the rising cost of agricultural production and the price fluctuations lead to a decrease in cropland yield, which also has a significant influence on cropland abandonment (Gellrich et al., 2007; Van Doorn and Bakker, 2007; Sklenicka et al., 2014). On the micro level, quantitative analyses based on rural household survey data have shown that labor force quantity, labor force age, education level of the labor force, gender structure of the labor force, fertilizer cost, and cropland consolidation degree (Van Doorn and Bakker, 2007; Meyfroidt et al., 2016; Zhang et al., 2016) may also impact cropland abandonment. In terms of natural factors, changes in temperature and precipitation can sometimes affect the degree of cropland abandonment (Osawa et al., 2016; Lasanta et al., 2017).
In recent years, with the rapid development of industrialization and urbanization in China, cropland abandonment has presented a trend of expansion, especially in mountainous areas (Shi et al., 2016, 2018; Levers et al., 2018; Guo and Song, 2019). However, due to the lack of reliable abandoned cropland extraction methods based on medium-resolution remote sensing images, the scale of abandoned cropland in China's mountainous areas and the corresponding driving mechanisms are still unclear. Specifically, in previous abandoned cropland extraction efforts by using remote sensing data, the differences between fallow fields and abandoned cropland have generally been ignored, leading to a reduced recognition accuracy of abandoned cropland. In view of this problem, this paper developed an abandoned cropland recognition method based on annual land use change monitoring. Through tracking the trajectory of annual cropland use, fallow fields and abandoned cropland in mountainous areas can be effectively differentiated. Against this background, the aims of this paper are three-fold: 1) To develop an abandoned cropland identifying method based on annual land use change monitoring; 2) To identify the distribution of abandoned cropland in Zhong County of Chongqing in China; and 3) To analyze the factors driving cropland abandonment in Zhong County of Chongqing.

2 Study area and data sources

2.1 Study area

Zhong County is located in the central part of Chongqing and represents the vital part of the Three Gorges Reservoir area. It spans across 107°32'E to 108°14'E and 30°03'N to 30°35'N (Fig. 1), with a mountain climate of the subtropical southeast monsoon region. Within its territory, low mountains spread up and down, with elevations between 117 m and 1680 m, and the terrain is mostly hilly. The study region covers an area of 2187 km2 and administers four subdistricts and 25 towns. The registered household number is 347300, with a registered population of 1.0094 million, including an agricultural population of 0.7748 million. The main crops are wheat, rice, corn, vegetables, and oil crops; the economic forest is given priority for citrus, and the citrus planting area is increasing progressively.
Fig. 1 Geographic location of Zhong County in China

2.2 Data sources

We adopted Landsat 8 OLI satellite remote sensing images with a resolution of 30 m from 2013 to 2018, downloaded from NASA (The National Aeronautics and Space Administration), as the basis for identifying abandoned cropland. The specific image parameters are listed in Table 1. Remote sensing images from June to October in each year were chosen for abandoned cropland extraction, since vegetation was abundant during this period and the vegetation cover types such as cropland, forest land, and grassland can easily be distinguished. The cloud cover of the remote sensing images was set to be less than 10%, with the exception of the images from 2015 which had a high cloud cover level (so the threshold value was set at 20%).
Table 1 Specific parameters of the acquired Landsat 8 images
Acquisition date Cloud cover (%) Solar azimuth (°) Solar height (°)
2013-08-12 10.22 120.83 64.07
2014-07-30 6.06 113.21 65.49
2015-01-22 18.70 151.16 34.62
2016-06-17 0.14 103.77 68.91
2017-07-22 6.20 109.64 66.38
2018-06-07 4.42 105.74 68.90
The digital elevation model (DEM) data used were obtained from the geospatial data cloud platform of the Computer Network Information Center, Chinese Academy of Sciences, with a spatial resolution of 30 m. High-resolution satellite image data for different years from Google Earth Pro were used as the basic data for selecting and screening the training samples and for verifying the classification results of the land use types. The social and economic data used in this study were obtained from the statistical yearbook of Zhong County.

3 Research methods

3.1 Research process

The research process includes four major steps, i.e., preprocessing of remote sensing images, land use type classification using the Classification and Regression Trees (CART) algorithm, extraction of abandoned cropland, and driving factor analysis of the abandoned cropland. Specifically, the major steps for abandoned cropland extraction and driving force analysis in Zhong County were as follows (Fig. 2):
Fig. 2 Technical process
(1) The remote sensing images were preprocessed, including cloud removal, geometric correction, stripe removal, radiometric calibration, and image cropping.
(2) Landsat 8 OLI remote sensing images for 2013 were processed by artificial visual interpretation. The land use distribution map of Zhong County in 2013, which has a higher reliability, was plotted, and the classification accuracy was verified. This data for land use status was used to evaluate the 2013 results obtained by CART to assess the accuracy of the CART model.
(3) The cropland boundaries in 2013 were extracted and used as the boundaries for identifying the abandoned cropland in 2014 to 2018.
(4) Land use change monitoring from 2014 to 2018, within the cropland boundary of 2013, was carried out via the CART decision tree, thus generating the annual land use variation maps for 2014 to 2018.
(5) On the basis of the annual land use change trajectories, the abandoned cropland and fallow cropland were defined and subsequently identified.
(6) The driving forces for cropland abandonment were analyzed.

3.2 Classification and Regression Trees (CART) algorithm

The CART algorithm was proposed by Breiman et al. (1984).The method adopts the top-down, divide-and-conquer recursive thinking and can deduce the algorithm rule in the form of decision trees from a set of disordered, random data. The algorithm rule can divide the data in the search space into several unrelated subsets. Subsequently, the CART algorithm can determine the classification thresholds for different subsets according to the selected training samples. The CART algorithm can automatically build a decision tree using the spectral feature information, effectively avoiding the influences of various factors, such as the spatial resolution of remote sensing images, different bodies with the same spectrum, and identical bodies with different spectra, on the classification results.
Regarding the decision tree, a non-cotyledon represents a certain test on the value of the land use type, a branch represents the test result, and a leaf node on the tree represents the land use type after image interpretation. The purpose of this division is to increase data purity. The purer the node classification, the more accurate the classification result of the decision tree; and the more obvious the spectral information characteristics of the remote sensing images, the clearer the land use types and the better the classification effect. The CART algorithm adopts the Gini Index to assess the node purity, which is calculated as follows (Breiman et al. 1984):
$Gini\left( p \right)=1-\underset{i\in I}{\mathop \sum }\,p_{i}^{2}$
where $Gini\left( p \right)$ represents the Gini coefficient, i represents land use type, I represents the total number of land use types and ${{p}_{i}}$ represents the probability of the i-th type among the samples. The formula of the Gini Index for the sample set S is as follows:
$Gini\left( S \right)=1-\underset{i\in I}{\mathop \sum }\,{{\left| \frac{{{d}_{i}}}{S} \right|}^{2}}$
where Gini(S) represents the Gini coefficient of sample dataset S, and ${{d}_{i}}$ is the sample subset belonging to the i-th land use type in the sample dataset S.

3.3 Drawing the maps of land use types

Since the crops in Zhong County follow a double cropping system, the selection of remote sensing images should also consider the phenological information of the crops. When selecting training samples and verification samples, the synthesized multivariate data images were first superposed to Google Earth Pro images with a high spatial resolution of 0.26 m, and the subsequent training and verification samples for cropland, forest land, grassland, water area, construction land, and bare land were extracted year by year according to the historical image datasets at different time periods in Google Earth Pro. Then, CART-extended tools in the ENVI 5.3 software were employed for data classification and the classification results were verified. The above process was executed until high-accuracy maps of land use types are obtained.
The visual interpretation classification results for 2013 were verified by the following steps. According to the status of land use types in the study area, a total of 326 verification samples were randomly selected, including 63 cropland samples, 64 forest land samples, 65 grassland samples, 59 water area samples, 59 construction land samples, and 16 bare land samples. The verification samples were superposed on the visual interpretation classification results, and subsequently, the visual interpretation results were individually contrasted with the actual land use status. Among the randomly selected 326 verification pixels belonging to six kinds of land use types, a total of 311 pixels were found to have land use types consistent with the classification results; namely, the verification accuracy of the overall classification of land use types in 2013 was as high as 95.40%.
The accuracy of the classification results of the Landsat satellite image data from 2014 to 2018 by the CART decision tree was verified by the built-in confusion matrix tool in ENVI 5.3 software. The overall classification accuracy of the confusion matrix was equal to the total number of correctly classified pixels divided by the total number of pixels. The Kappa coefficient can measure the accuracy based on the confusion matrix, and its formula is as follows (Congalton, 1991):
${{p}_{e}}=\frac{\sum\limits_{i=1}^{m}{{{a}_{i}}{{b}_{i}}}}{{{n}^{2}}}$
$k=\frac{{{p}_{o}}-{{p}_{e}}}{1-{{p}_{e}}}$
where pe is the intermediate variable, po is the overall classification accuracy, ai is the number of true samples for each land use type, bi is the number of predicted samples for each land use type, m is the total number of land use types, n is the total number of samples, and k is the Kappa coefficient.
Through verification by the confusion matrix, the overall classification accuracy of the classification results by the CART decision tree was 94.48%, and the Kappa coefficient was 0.928. Among the classification results, there were large classification errors for the two land use types of grassland and cropland.

3.4 Identification of fallow fields, abandoned cropland, and rehabilitated land

Currently, the definition of abandoned cropland used by scholars is still controversial, and the main debate focuses on the number of years for which farming had been stopped. Generally, cropland can be considered abandoned if it has not been cultivated for 1 or 2 years (Shi et al., 2016; Lesiv et al., 2018). However, some researchers define cropland abandonment as the cropland remaining unmanaged for at least 5 years during a 6-year period (Smaliychuk et al., 2016). In this study, according to the basic terms in the standard system of China's land resources in 2016, abandoned cropland is defined as land that has been fallow or idle for 3 or more years, while the fallow land includes grass-crop rotation fields mainly for crop cultivation as well as fallow fields that have been abandoned for less than 3 years.
The rule for identifying abandoned cropland is as follows: for the cropland in the t-th year, if in the following 3 years (until the (t+3)-th year), the land use type of this cropland is always forest land, grassland, or bare land, and this cropland is not cultivated within these 3 years, then this cropland is considered to be abandoned. The formula for calculating the cropland abandonment rate is as follows:
${{P}_{l}}=\frac{{{a}_{t+3}}}{A}\times 100\text{ }\!\!%\!\!\text{ }$
where ${{P}_{l}}$represents the cropland abandonment rate; A represents the total cropland area; and ${{a}_{t+3}}$represents the total area of abandoned cropland in the (t+3)-th year.
The rule for identifying fallow land is as follows: for the cropland in the t-th year, if in the following 3 years (until the (t+3)-th year), this cropland is converted into forest land, grassland, or bare land, and the land is cultivated and harvested at least once, then it is considered to be fallow land. The formula for calculating the cropland fallow rate is as follows:
${{P}_{x}}=\frac{X}{A}$
where ${{P}_{x}}$represents the fallow rate; A represents the total cropland area; and $X$represents the total area of fallow cropland during the 3 years.
Large-area cropland abandonment directly influences the cultivation production and the grain acreage, thus affecting national food security and hindering the economic development and social stability. Under certain conditions, abandoned cropland can be restored and rehabilitated; namely, the existing abandoned land that can be used for agricultural production can be converted into cropland. In this study, the rehabilitation of abandoned cropland is defined as follows: for the abandoned cropland in the t-th year, if the land use type is converted into cropland in the (t+1)-th year, the land is considered to be rehabilitated land. The specific formula for abandoned cropland rehabilitation is as follows:
${{P}_{f}}=\frac{{{F}_{t+1}}}{{{a}_{t}}}\times 100%$
where ${{P}_{f}}$represents the land rehabilitation rate; ${{a}_{t}}$ represents the total area of abandoned cropland in the t-th year; and ${{F}_{t+1}}$ represents the total area of the rehabilitated land in the (t+1)-th year.
The tool “Create Random Points” in ArcGIS 10.4 software was used to create 120 random sample points in the spatial distribution map of the extracted abandoned cropland, and 98 abandoned cropland parcels nearest to the random points were selected by the Near tool as the verification samples. Subsequently, on the Google Earth images with a resolution of 0.27 m, the plaques corresponding to the verification samples were identified. The abandonment status of each plaque was determined one by one, and the results were checked via field investigations. Based on the verification, among the 98 identified abandoned cropland parcels, only nine parcels had been cultivated during three successive years. Therefore, the accuracy rate of the abandoned cropland identification was as high as 90.82%.

3.5 Analysis of socioeconomic driving mechanisms of cropland abandonment

In this study, the ridge regression model was employed to analyze the driving mechanisms of cropland abandonment. The ridge regression model is a type of biased estimation regression method for colinear data analysis which can be used to correct the computational formula for the regression coefficient of the least square estimation. The model is mainly used to solve the problems of inadequate data quantity and multicollinearity between explanatory variables. The ridge regression model makes the regression coefficient more stable at the expense of information loss and a decrease in accuracy; in particular, the plus or minus symbols before explanatory variables coincide better with the actual problems.
The ridge trace method was used to draw the ridge trace map of ${{\theta }_{\lambda }}$. Within the range where the regression coefficients of all explanatory variables generally tend to be stable, the minimum $\lambda $ value was selected to make the mean square error (MSE) be the lowest (Hoerl and Kennard, 1970).
The hypothesis function of the ridge regression model is as follows:
${{Y}_{\theta }}\left( x \right)={{\theta }_{0}}+{{\theta }_{1}}{{x}_{1}}+{{\theta }_{2}}{{x}_{2}}+\cdots +{{\theta }_{n}}{{x}_{n}}$
where ${{Y}_{\theta }}\left( x \right)$ is the abandonment rates of all towns in Zhong County; ${{\theta }_{n}}$ is the regression coefficient; ${{x}_{n}}$is the independent variable.
The regression coefficient can be expressed as:
${{\theta }_{\lambda }}={{\left( {x}'x+\lambda {{t}_{n}} \right)}^{-1}}{x}'Y$
where ${{\theta }_{\lambda }}$is the parameter, Y represents the dependent variable; x is the independent variable; ${x}'$is the transposed matrix of x; λ is a non-negative factor, which is called ridge parameter or biased parameter; ${{t}_{n}}$is the n-order unit matrix.
By introducing regularization into the loss function, the ridge regression model can deal with the overfitting problem of linear regression and the irreversible problem of $\text{ }\!\!~\!\!\text{ }{x}'x$in the process of solving $\theta $ through the normal equation. The loss function of the ridge regression is as follows:
${{J}_{\theta }}=\frac{1}{2m}\underset{i=1}{\overset{m}{\mathop \sum }}\,{{({{h}_{\theta }}({{x}^{i}})-{{y}^{i}})}^{2}}+\lambda \underset{j=1}{\overset{n}{\mathop \sum }}\,\theta _{j}^{2}$
where ${{J}_{\theta }}$is the loss function of ridge regression; ${{x}^{i}}$is the i-th independent variable; ${{y}^{i}}$is the abandonment rate of the i-th township in Zhongxian County from 2013 to 2018; m is the number of rows of the matrix, i=1, 2, 3, ···, m; n is the number of columns of the matrix, j=1, 2, 3, ···, n; ${{h}_{\theta }}({{x}^{i}})$is the abandonment rate of the i-th township in Zhongxian County from 2013 to 2018 predicted by ridge regression model; ${{\theta }_{j}}$is the weight coefficient.
In the ridge regression analysis of cropland abandonment in Zhong County, the dependent variables were the abandonment rates of all towns in Zhong County, while the independent variables could be classified into three groups. The first group was the population characteristics of the corresponding study area (including birth rate, out-migrants, natural growth rate, rural population and student enrollment). The second group was the economic development characteristics of the corresponding study area (including capital introduction from outside the county, average disposable personal income of rural residents, added value of agriculture, industrial added value, industrial investment, total retail sales of consumer goods, individual business, and meat production). The third group involved the influences of other industries on cropland abandonment, including citrus cultivation area (Table 2).
Table 2 Social and economic development variables, summary statistics, and the expected effects of each variable
Variables Minimum Maximum Mean Standard
deviation		 Expected effect
Dependent variable Cropland abandonment rate (%) 0.87 24.04 8.66 1.110
Independent
variable		 Explanatory variables birth rate (%) 7.51 14.56 11.38 0.392 -
Out-migrants (person) 32 2761 326.36 92.954 +
Natural population growth rate (%) -3.26 6.34 2.11 0.586 -
Rural population (person) 6932 49629 26787.61 2279.835 -
Student enrollment (person) 385 37717 4000.71 1321.580 +
Capital introduction from outside the county (104 yuan) 1000 10500 5633.21 585.161 +/-
Average disposable personal income of rural residents (yuan) 8350 12236 9959.25 203.303 +
Added value of agriculture (104 yuan) 1425 21500 11163.79 999.254 -
Industrial added value (104 yuan) 978 499812 29073.50 17848.175 +
Industrial investment (104 yuan) 350 193312 16850.25 8843.985 +
Total retail sales of consumer goods (104 yuan) 1099 376457 20573.68 13228.794 +/-
Individual business (household) 100 10131 1045.07 346.572 +
Meat production (t) 408 4715 2641.89 233.094 -
Citrus cultivation area (km2) 0.08 23.12 8.00 1.35 +/-

Note: Out-migration refers to residents perpetually or persistently migrating from an emigration area to a relocation area; Capital introduction from outside the county refers to the amount of money introduced from outside the study area through investment promotion; Individual business refers to the families engaging in industrial and commercial business.

4 Results

4.1 Land use changes in Zhong County from 2013 to 2018

The areas of cropland, forest land, grassland, water bodies, construction land, and bare land were 723.03, 1083.52, 211.36, 81.71, 75.75, and 22.11 km2 in 2013, respectively (Fig. 3), accounting for 32.90%, 49.31%, 9.62%, 3.72%, 3.45%, and 1.01% of the total area of Zhong County, respectively. Construction land was mainly distributed in urban and residential areas, and the central cities were distributed along both sides of the Yangtze River. In these regions, the social economy was well-developed, and settlements were widely distributed. Cropland and grassland were scattered across these areas, with a high fragmentation degree; cropland was not used intensively and mainly distributed in the southwestern and northeastern regions with a smaller topographic relief, while grassland was distributed along both sides of the Yangtze River. Forests covered a large area, with a high forest coverage rate. They were distributed in the central and northern mountainous areas with high elevations and a larger topographic relief.
Fig. 3 Distribution of land use types in Zhong County in 2013
Within the cropland boundaries in 2013, some cropland was quickly converted into other land use types (Fig. 4), mainly grassland and bare land. Grassland, bare land, and forest land which had been converted from cropland were mainly distributed in the regions with larger slopes. Over time, only small areas of cropland were converted into water bodies, with slightly varying water areas. The areas of forest land and construction land increased continuously; construction land expanded mainly into rural residential and urban areas, while the newly added forests were mainly concentrated in unprofitable farming regions with larger slopes.
Fig. 4 Land-use changes within the cropland boundary of Zhong County in 2013, 2014, 2015, 2016, 2017, 2018.

4.2 Fallow land, abandoned land, and rehabilitated land

According to the identification methods for fallow land,abandoned land, and rehabilitated land, the spatial distribution data were extracted for these land use types in Zhong County (Fig. 5). In the study area, the cumulative total area of fallow land during the six years was 171.87 km2, and the fallow rate was 34.36%, so nearly a third of the total cropland area was fallow. The total area of abandoned land during the six years was 39.34 km2, and the total abandonment rate in the study area was 7.86%. The total area of rehabilitated land was 9.48 km2, accounting for 24.10% of the abandoned land area. The rehabilitation degree of abandoned cropland was low, and most of the abandoned cropland had been abandoned for a long time. There was an obvious spatial distribution of fallow land, abandoned land, and rehabilitated land. The numbers of land parcels of fallow land and abandoned land were 15147 and 7369, respectively, and they were scattered throughout the regions with a greater topographic relief and higher elevations, showing a high degree of fragmentation. Fallow land was distributed in western Chongqing, while abandoned land was mainly distributed in regions with steeper slopes, mountainous margins, and in regions far away from residential areas. Rehabilitated areas were distributed in the eastern region of Zhong County, mainly on flat terrain and near water sources and residential areas.
Fig. 5 Spatial distribution of fallow land, abandoned land, and rehabilitated land in Zhong County from 2013 to 2018.

4.3 Influences of social and economic development on cropland abandonment

The ridge regression results show that eight of the 14 explanatory variables were statistically correlated (Table 3). For the explanatory variables of population and labor force distribution, rural population and student enrollment were correlated with cropland abandonment at a significance level above 0.05, and their influences on cropland abandonment were consistent with their expected effects. In other words, an increase in rural population resulted in a decrease in abandoned land, while an increase in student enrollment enhanced the risk of cropland abandonment. For the explanatory variables of economic development, in creases in either the average disposable personal income of rural residents, the capital introduction from outside the county, or the total retail sales of consumer goods increased the possibility of cropland abandonment, and these three explanatory variables were statistically significant at the 0.05 level. The added value of agriculture was correlated with cropland abandonment at the 0.01 level, and an increase in the added value of agriculture reduced the risk of cropland abandonment. Among other aspects of industrial development, an increasing meat production reduced the possibility of cropland abandonment, and meat production was correlated with cropland abandonment at the 0.01 level. An increase in individual industrial and commercial households aggravated cropland abandonment. In the regression analysis, the R2 value was equal to 0.75, indicating that the ridge regression model can explain the influences of social and economic variables on cropland abandonment pretty well.
Table 3 Regression results of the social and economic explanatory variables
Explanatory variable Ridge regression model
Estimated parameter Normalized parameter Standard error t value
Constant 9.03
Birth rate -0.19 -1.010 0.803 0.803
Out-migrants 0.18 0.942 1.345 1.345
Natural population growth rate 0.19 0.978 0.757 0.757
Rural population -0.41** -2.117 2.332 2.332
Student enrollment 0.31** 1.609 2.453 2.453
Capital introduction from outside the county 0.63** 3.297 2.520 2.520
Average disposable personal income of rural residents 0.74*** 3.868 2.810 2.810
Added value of agriculture -0.67*** -3.464 3.528 3.528
Industrial added value 0.17 0.875 1.289 1.289
Industrial investment -0.17 -0.863 0.656 0.656
Total retail sales of consumer goods 0.27** 1.392 2.138 2.138
Individual business 0.21* 1.104 1.817 1.817
Meat production -0.47*** -2.418 2.596 2.596
Citrus cultivation area -0.42 -2.166 1.55 1.55

Note: ***, **, and * represent significance at P < 0.01, P < 0.05, and P < 0.1, respectively.

4.4 Physical geographical characteristics of abandoned cropland parcels

Spatial analysis of the abandoned land shows that the area of abandoned cropland generally presented an increasing trend with an increase in farming radius (i.e., the distance from residential settlements). Within a farming radius of 720 m, the abandonment rate increased obviously with increasing farming radius, but in the farming radius range from 720 to 1200 m, the abandonment rate was only slightly influenced by farming radius. The maximum cropland abandonment rate was observed in the farming radius range from 1080 to 1200 m, with a value of 7.81% (Fig. 6).
Fig. 6 Cropland area, abandoned land area, and abandonment rate at different farming radii.
The area of abandoned cropland decreased with an increasing distance to the forest edge, along with the cropland abandonment rate. As Fig. 7 shows, the average abandonment rate decreased significantly within the distance ranges of 0-30 m and 30-60 m. When the distance from abandoned cropland to the forest edge was less than 60 m, the cropland abandonment rate was greater than the average cropland abandonment rate in the study area. At various distances of more than 60 m, the cropland abandonment rate variation was low, and the influences of the distance from abandoned cropland to the forest edge on this rate were weak (Fig. 7).
Fig. 7 Cropland area, abandoned land area, and abandonment rate at different distances from the abandoned cropland to the forest edge.
Cropland was mainly distributed in the regions with smooth slopes of 2-15°, while cropland in regions with slopes of more than 25° only accounted for 1.40% of the total cropland area (Table 4). The abandonment rate increased with increasing slopes. Within the scope of smaller slopes (i.e., up to 15°), the abandonment rate was low; however, at slopes of more than 15°, the rate increased significantly, and at slopes of more than 35°, a maximum abandonment rate of 65% was reached.
Table 4 Influences of different slope grades on cropland abandonment in Zhong County
Slope (°) 0-2 2-6 6-15 15-25 25-35 ≥35
Abandoned cropland (km2) 0.61 3.61 14.71 14.33 5.27 0.80
Cropland (km2) 37.13 169.44 232.44 56.81 4.14 0.20
Abandonment rate (%) 1.40 1.81 5.00 15.95 40.81 65.04

Note: Abandonment rate refers to the area of abandoned cropland with different slope divided by the area of cropland with corresponding slope in 13 years.

In the study area, cropland was concentrated in the hills and low mountains within an elevation range of 200-650 m (Table 5), while the cropland at elevations of more than 650 m accounted for only 4.59% of the total cropland area. The abandonment rate increased with the increasing elevation. At 0-650 m, the rate was lower than the overall rate in the study area, while at elevations above 650 m, the average cropland abandonment rate was above 15.22%.
Table 5 Influences of different elevation grades on cropland abandonment in Zhong County
Elevation (m) 0-200 200-350 350-500 500-650 650-800 ≥800
Abandoned cropland (km2) 1.26 7.04 15.81 11.64 3.12 0.46
Cropland (km2) 18.41 97.01 213.04 148.72 20.53 2.44
The abandonment rate (%) 6.82 7.26 7.42 7.83 15.22 18.79

Note: Abandonment rate refers to the area of abandoned cropland with different elevations divided by the area of cropland with corresponding elevation in 13 years.

5 Discussion

The cropland abandonment rate of Zhong County was 7.86%, which is lower than the average abandonment rate of China's mountainous counties (14.32%) from a household survey (Li et al., 2017). Notably, the terrain of this study area is hilly, and the terrain conditions and the development level of agricultural mechanization are superior to those of the mountainous regions. Hence, it is reasonable that the abandonment rate in the study area is slightly lower. In addition, the area of fallow land in the study area was 171.87 km2, and the cropland was characterized by fallow crop rotation. According to the statistical yearbook of Zhong County, the total cropland area at the end of 2013 was 722.08 km2, of which the area of temporary cropland was 173.34 km2, accounting for 24.01% of the total cropland area. Temporary cropland is relatively barren, and the production infrastructure is generally weak. As a result, such cropland is more likely to be fallow and abandoned (Lesiv et al., 2018). However, in our study, the abandoned cropland was extracted on the basis of the cropland boundaries in 2013, while some temporarily fallow cropland beyond these boundaries was not considered. Hence, the measured cropland abandonment rate may be slightly lower than the actual abandonment rate of the study area.
We selected 14 variables covering demographic factors, the economic development situation, and other industry development situations of all 28 towns in Zhong County, with the aim of comprehensively assessing the influences of social and economic development on cropland abandonment. The ridge regression results show that the rural population and student enrollment are significantly correlated with cropland abandonment. The rural population has a great influence on cropland abandonment, which accords with the results of previous studies (Deng et al., 2018; Xu et al., 2019). With the transformation of traditional concepts, rural households pay more attention to their children's education. These children have either directly or indirectly participated in agricultural production activities and are an important part of the agricultural labor force. Hence, the increase in student enrollment increases the risk of cropland abandonment.
Among the explanatory variables representing the economic development distribution, the added value of agriculture reduces cropland abandonment and can be used to measure the income of farmers. An increase in farming income can effectively reduce the possibility of cropland abandonment. By contrast, the average disposable personal income of rural residents, the capital introduction from outside the county, and the total retail sales of consumer goods can increase the possibility of cropland abandonment, and these three explanatory variables showed statistical significance at the 0.05 level. Towns with higher capital introduction from outside have more money to develop their local industry, the tourism sector, and their service industry, thus providing more employment opportunities for the local rural labor force. According to China's rural land law, land is assigned to farmers based on the average rural population. In the regions where the average annual disposable income of rural residents is higher, the proportion of non-agricultural employment in the rural population is greater. With improved living standards, rural residents are reluctant to engage in hard work and agricultural activities with low production efficiency, but willing to take up other non-agricultural employment activities with higher economic benefits, which directly or indirectly leads to the transfer of more rural laborers to non-agricultural employment, thus aggravating cropland abandonment (Sikor et al., 2009; Deng et al., 2018). In terms of other industrial development, animal husbandry has a great influence on cropland abandonment and can significantly reduce the cropland abandonment phenomenon at the 0.01 significance level. The development of animal husbandry increases the demand for feed crops, and agricultural products can further flow to livestock production, thus reducing cropland abandonment.
Previous studies have found that the land parcel status has a stronger explanatory power for the spatial distribution of cropland abandonment than households or villages (Prishchepov et al., 2013; Zhang et al., 2014). In this study, we divided different slopes, different elevations, different farm radii, and different distances from the land parcel to the forest edge into multiple grades on the land parcel scale and conducted a quantitative analysis on the abandonment rates at each grade. The results showed that in regions with similar economic and social development levels and a consistent policy system, from the perspective of the macro landform, the larger the slope and the higher the elevation of the land parcel, the higher the possibility of cropland abandonment. At larger slopes and higher elevations, mechanized agricultural production is not feasible (Shi et al., 2016). From the perspective of the farming radius, a larger farming radius means a farther commuting distance and higher management costs, and such land parcels are therefore more likely to be abandoned. The smaller the distance from the land parcel to the forest edge, the greater the influences of tree shade and seed rain on that land parcel, potentially reducing cropland production; therefore, such land parcels are also highly likely to be abandoned.
In this study, there are some deficiencies in the analysis of the driving forces of cropland abandonment. First, the selected socio-economic factors cannot fully reflect the social and economic situation. The socio-economic indicators considered in this research are just the conceptualization of the structures and characteristics of the socio-economic system from different perspectives. They can only reflect part of the characteristics of the socio-economic system because the actual social economy is a complex system (Jin et al., 2019). Therefore, the analysis of the driving forces of cropland abandonment is inevitably affected by many random factors. Secondly, the change of the land system has a profound impact on the social and economic development. The change of the land system is an adaptation to the changes of social and economic development. The continuous development of the social economy requires the continuous change and development of the land system. As for the implementation of the national “Grain for Green Program” (GGP), this scheme was initiated in 2001 and aims to convert marginal farmland into forested land by compensating households for planting trees on retired farmland (Zhang et al., 2017). Many studies have shown that it will have an impact on cropland abandonment. For example, the changing trend of the cropland abandonment rate in Guangxi karst mountainous area from 2001 to 2019 is consistent with the implementation of “Returning Farmland to Forest Plan” (GGP) to a certain extent (Han and Song, 2020). Finally, compared with other research results, some representative socio-economic indicators are not included in the indicator system used here. As affected by geographic location, each region has its own unique characteristics, so it is difficult to develop a unified standard for fully evaluating the influencing factors of cropland abandonment. Therefore, determining how to scientifically and accurately select indicators to reflect the driving forces of cropland abandonment remains to be further explored.

6 Conclusions

In this study, we developed a method for mapping abandoned cropland in mountainous areas based on annual land use change monitoring. The results show that the cropland abandonment rate in Zhong County of Chongqing was 7.86%, while the overall extraction accuracy of abandoned cropland was as high as 90.82%. Abandoned cropland was scattered throughout the region, with a high degree of fragmentation. Cropland abandonment presented a trend of spreading from the mountainous regions with higher elevations and larger slopes to hilly regions with more favorable terrain conditions.
The population, the degree of economic development, and the animal husbandry sector in the study area all significantly influenced cropland abandonment. Among the social and economic factors that affect cropland abandonment, the rural population, economic development, and livestock husbandry development were the most important ones. At the land parcel scale, large-scale cropland abandonment occurred in areas with elevations above 650 m or slopes of more than 15°.

References

Publishing order | Descend order by publishing year | Descend order by cited within
[1]
Alcantara C, Kuemmerle T, Prishchepov A V, et al. 2012. Mapping abandoned agriculture with multi-temporal MODIS satellite data. Remote Sensing of Environment, 124: 334-347.

DOI

[2]
Breiman L, Friedman J H, Olshen R A, et al. 1984. Classification and regression trees. New York, USA: Routledge.

[3]
Dara A, Baumann M, Kuemmerle T, et al. 2018. Mapping the timing of cropland abandonment and recultivation in northern Kazakhstan using annual Landsat time series. Remote Sensing of Environment, 213: 49-60.

DOI

[4]
Deng X, Xu D D, Qi Y B, et al. 2018. Labor off-farm employment and cropland abandonment in rural China: Spatial distribution and empirical analysis. International Journal of Environmental Research and Public Health, 15(9): 1808. DOI: 10.3390/ijerph15091808.

DOI

[5]
Díaz G I, Nahuelhual L, Echeverría C, et al. 2011. Drivers of land abandonment in Southern Chile and implications for landscape planning. Landscape and Urban Planning, 99(3-4): 207-217.

DOI

[6]
Estel S, Kuemmerle T, Alcántara C, et al. 2015. Mapping farmland abandonment and recultivation across Europe using MODIS NDVI time series. Remote Sensing of Environment, 163: 312-325.

DOI

[7]
Gellrich M, Baur P, Koch B, et al. 2007. Agricultural land abandonment and natural forest re-growth in the Swiss mountains: A spatially explicit economic analysis. Agriculture, Ecosystems & Environment, 118(1-4): 93-108.

[8]
Guo Y C, Song W. 2019. Spatial distribution and simulation of cropland abandonment in Wushan County, Chongqing, China. Sustainability, 11(5): 1367. DOI: 10.3390/su11051367.

DOI

[9]
Han Z, Song W. 2020. Abandoned cropland: Patterns and determinants within the Guangxi Karst mountainous area, China. Applied Geography, 122: 102245. DOI: 10.1016/j.apgeog.2020.102245.

DOI

[10]
Hoerl A E, Kennard R W. 1970. Ridge regression: Applications to nonorthogonal problems. Technometrics, 12(1): 69-82.

DOI

[11]
Jin G, Chen K, Wang P, et al. 2019. Trade-offs in land-use competition and sustainable land development in the North China Plain. Technological Forecasting and Social Change, 141: 36-46.

DOI

[12]
Kamada M, Nakagoshi N. 1997. Influence of cultural factors on landscapes of mountainous farm villages in western Japan. Landscape and Urban Planning, 37(1-2): 85-90.

DOI

[13]
Khanal R N, Watanabe T. 2006. Abandonment of agricultural land and its consequences. Mountain Research and Development, 26(1): 32-40.

DOI

[14]
Kolecka N, Kozak J, Kaim D, et al. 2017. Understanding farmland abandonment in the Polish Carpathians. Applied Geography, 88: 62-72.

DOI

[15]
Lambin E F, Meyfroidt P. 2010. Land use transitions: Socio-ecological feedback versus socio-economic change. Land Use Policy, 27(2): 108-118.

DOI

[16]
Lasanta T, Arnáez J, Pascual N, et al. 2017. Space-time process and drivers of land abandonment in Europe. CATENA, 149: 810-823.

DOI

[17]
Lesiv M, Schepaschenko D, Moltchanova E, et al. 2018. Spatial distribution of arable and abandoned land across former Soviet Union countries. Scientific Data, 5: 180056. DOI: 10.1038/sdata.2018.56.

DOI

[18]
Levers C, Schneider M, Prishchepov A V, et al. 2018. Spatial variation in determinants of agricultural land abandonment in Europe. Science of the Total Environment, 644: 95-111.

DOI

[19]
Li S F, Li X B, Xin L J, et al. 2017. Extent and distribution of cropland abandonment in Chinese mountainous areas. Resources Science, 39(10): 1801-1811. (in Chinese)

[20]
Lieskovský J, Bezák P, Špulerová J, et al. 2015. The abandonment of traditional agricultural landscape in Slovakia-Analysis of extent and driving forces. Journal of Rural Studies, 37: 75-84.

DOI

[21]
MacDonald D, Crabtree J R, Wiesinger G, et al. 2000. Agricultural abandonment in mountain areas of Europe: Environmental consequences and policy response. Journal of Environmental Management, 59(1): 47-69.

DOI

[22]
Meyfroidt P, Schierhorn F, Prishchepov A V, et al. 2016. Drivers, constraints and trade-offs associated with recultivating abandoned cropland in Russia, Ukraine and Kazakhstan. Global Environmental Change, 37: 1-15.

DOI

[23]
Nikodemus O, Bell S, Grı̄ne I, et al. 2005. The impact of economic, social and political factors on the landscape structure of the Vidzeme Uplands in Latvia. Landscape and Urban Planning, 70(1-2): 57-67.

DOI

[24]
Osawa T, Kohyama K, Mitsuhashi H. 2016. Multiple factors drive regional agricultural abandonment. Science of the Total Environment, 542: 478-483.

DOI

[25]
Prishchepov A V, Müller D, Dubinin M, et al. 2013. Determinants of agricultural land abandonment in post-Soviet European Russia. Land Use Policy, 30(1): 873-884.

DOI

[26]
Prishchepov A V, Radeloff V C, Dubinin M, et al. 2012. The effect of Landsat ETM/ETM + image acquisition dates on the detection of agricultural land abandonment in Eastern Europe. Remote Sensing of Environment, 126: 195-209.

DOI

[27]
Romero-Calcerrada R, Perry G L W. 2004. The role of land abandonment in landscape dynamics in the SPA ‘Encinares del río Alberche y Cofio, Central Spain, 1984-1999. Landscape and Urban Planning, 66(4): 217-232.

DOI

[28]
Shi T C, Li X B, Xin L J, et al. 2016. Analysis of farmland abandonment at parcel level: A case study in the mountainous area of China. Sustainability, 8(10): 988. DOI: 10.3390/su8100988.

DOI

[29]
Shi T C, Li X B, Xin L J, et al. 2018. The spatial distribution of farmland abandonment and its influential factors at the township level: A case study in the mountainous area of China. Land Use Policy, 70: 510-520.

DOI

[30]
Sikor T, Müller D, Stahl J. 2009. Land fragmentation and cropland abandonment in Albania: Implications for the roles of state and community in post-socialist land consolidation. World Development, 37(8): 1411-1423.

DOI

[31]
Sklenicka P, Janovska V, Salek M, et al. 2014. The farmland rental paradox: Extreme land ownership fragmentation as a new form of land degradation. Land Use Policy, 38: 587-593.

DOI

[32]
Smaliychuk A, Müller D, Prishchepov A V, et al. 2016. Recultivation of abandoned agricultural lands in Ukraine: Patterns and drivers. Global Environmental Change, 38: 70-81.

DOI

[33]
Vacquie L A, Houet T, Sohl T L, et al. 2015. Modelling regional land change scenarios to assess land abandonment and reforestation dynamics in the Pyrenees (France). Journal of Mountain Science, 12(4): 905-920.

DOI

[34]
Van Doorn A M, Bakker M M. 2007. The destination of arable land in a marginal agricultural landscape in South Portugal: An exploration of land use change determinants. Landscape Ecology, 22(7): 1073-1087.

DOI

[35]
Xu D D, Deng X, Guo S L, et al. 2019. Labor migration and farmland abandonment in rural China: Empirical results and policy implications. Journal of Environmental Management, 232: 738-750.

DOI

[36]
Yan J Z, Yang Z Y, Li Z H, et al. 2016. Drivers of cropland abandonment in mountainous areas: A household decision model on farming scale in Southwest China. Land Use Policy, 57: 459-469.

DOI

[37]
Yu Z, Lu C Q. 2018. Historical cropland expansion and abandonment in the continental US during 1850 to 2016. Global Ecology and Biogeography, 27(3): 322-333.

[38]
Zhang Y, Li X B, Song W. 2014. Determinants of cropland abandonment at the parcel, household and village levels in mountain areas of China: A multi-level analysis. Land Use Policy, 41: 186-192.

DOI

[39]
Zhang Y, Li X B, Song W, et al. 2016. Land abandonment under rural restructuring in China explained from a cost-benefit perspective. Journal of Rural Studies, 47: 524-532.

DOI

[40]
Zhang Z M, Zinda J A, Li W Q. 2017. Forest transitions in Chinese villages: Explaining community-level variation under the returning forest to farmland program. Land Use Policy, 64: 245-257.

DOI

Options
Outlines

/

〈 〉