Farmland is a fundamentally important natural resource [1
]. Farmland can provide humans with basic necessities [2
], contribute to biodiversity [3
] and impact our climate, e.g., [4
]. However, demographic shifts are accompanied by large-scale urbanization and have affected the magnitude, pattern and process of farmland use [5
]. Reductions in farmland have caused the degradation of agricultural environments, natural environments [6
] and threatened food security and ecosystem services [7
]. To mitigate such impacts, in 2001, the European Union (EU) took political measures to combat soil erosion, maintain soil structure and organic matter [8
], but major modifications to EU agriculture policies were introduced by the Common Agricultural Policy (CAP) in 2003; the innovation of this reform was that farmers were no longer required to maintain production for receiving CAP payments, but had to keep land in good environmental conditions [9
]. Then, in 2004 and 2007, the EU incorporated twelve eastern and southern European countries to increase farmland by 500,000 hectares, which approaches the total area of France, Germany and the U.K., profoundly affecting the overall ecological statues of European ecosystems [10
]. Under the guidance of the EU, agri-environment schemes (AES) and other funding mechanisms have been implemented under the CAP to achieve nature-friendly management and ecological farming [11
]. Following these important measures, many others have been implemented, laying a foundation for sustainable development in other countries. As China has experienced rapid urbanization, industrialization and economic development, China’s farmland has changed significantly. According to the 2008 Bulletin of National Land and Resources data, since 1996, China’s farmland has decreased by 8.3 million hm2
, or 595.24 thousand hm2
per year [12
]. Nonetheless, agriculture is very important in China, and a large population requires food produced by farmland to maintain life; thus, the limited farmland resources are experiencing production pressure [13
]. Furthermore, due to unplanned land use, a series of ecological environmental problems, such as water pollution, air pollution and soil erosion, took place in the process of urbanization [14
]. Tension between food security, ecological protection and economic development has become intense. China’s central government has implemented policies to address the contradiction, and the dynamic change of farmland and its driving forces have attracted the attention of scholars. However, most of these study areas focused on the development of nations, provinces and urban agglomeration. Only a limited number of studies have focused on less-developed regions or ecologically-fragile, mountainous areas [16
]. In China, mountainous areas account for 70% of the total land area with almost 40% of the population [17
]. Mountainous areas are therefore important for land use and population studies. As a result of geographical and soil conditions, farmland is not prevalent in China’s mountainous areas. In fact, farmland accounts for less than 15% of the land in China’s mountainous areas. The soil quality is not very high [18
], and food from such limited farmland cannot meet the demands of the population. Importantly, the ecology in mountainous areas is often fragile: once destroyed, it is difficult to recover. Since the acceleration of urbanization, a large amount of farmland in China has been converted into other land use types, creating ecological threats. Therefore, there is a pressing need to manage farmland change and its effects on natural environments, particularly in ecologically-fragile mountainous regions.
Recently, a wide variety of researchers has modeled spatiotemporal patterns of farmland conversion, investigated the causes of farmland conversion and analyzed the impacts of land use change [19
]. Until now, there have been two main methods used to describe the characteristics of farmland change: the landscape index [23
] and spatial statistics [24
]. These two methods are based on remote sensing data. Remote sensing (RS) has been recognized as a powerful and effective tool for detecting the spatiotemporal dynamics of land use, and spatially-explicit time series of land use change can be developed based on RS [25
]. Forecasting trends in farmland change has mainly included the development of a simple Markov chain model and the Conversion of Land Use and its Effects at small regional extent (CLUE-s) model [26
]. These two models have been of great value to policymakers in predicting farmland changes and making informed decisions about such changes. Of course, various forces influence farmland change, such as national policy, population change, economy, science and technology, per capita income, agricultural production, land use policy, topography and natural disasters. These forces can be broadly divided into physical and anthropogenic categories. Physical drivers include climate [28
], terrain [29
], soil [30
], hydrology and locust hazard [31
]. These physical forces can result in reduced agricultural harvests and food shortages, even large numbers of deaths. Anthropogenic drivers include the economy, population, agricultural modernization [32
], technology [33
], culture [34
], agricultural policies [35
] and food regimes [36
]. Farmland changes, while restricted by physical conditions, are mainly driven by anthropogenic factors and characterized by changes in built-up land and forest, both of which are closely related to human production activities and national policy [37
In China’s mountainous areas, altitude, slope, aspect and other factors result in three-dimensional differences in precipitation, temperature and other elements, all of which affect the distribution of vegetation and settlements [38
]. As a result, we see vertical differentiation within the natural and cultural landscape in China. Scholars agree that altitude and slope are significant factors affecting the vertical differentiation of farmland [39
]. However, previous studies have mainly focused on the value, structure and spatial distribution of altitude and farmland or slope and farmland. There is less research on conversion characteristics between farmland and its closely related land types (e.g., built-up land, grassland and forest) according to variations in altitude and slope [40
]. To address the lack of research, our study looked into the driving mechanisms involved in the change of farmland to other land use types in Panxi. Our results may be practically applied to current land management issues.
As the ecological security barrier for the Yangtze River, the Panxi region serves as an agricultural engineering base and essential reserve of agricultural land. Panxi is also an important mining area with a high degree of industrialization. In recent years, Panxi’s farmland has experienced structural changes, as well as increased occupation due to two factors. First, industrialization and urbanization have caused the farmland near cities to be converted into built-up land. If continued, this urbanization could cause Panxi to cross its farmland “red line,” which is the minimum amount of farmland required for the region, and impact regional food security. Second, national construction projects intended to protect the natural environment have had tremendous impacts on regional land utilization. For example, the Returning Farmland to Forest and Natural Forest Protection Project has required that land must be converted to forest if the land is of great significance to ecological protection. As a result of these factors, it is particularly important to understand farmland dynamics from a macroscopic point of view. To this end, our study takes the Panxi region as a case study. Based on land use data from 1990, 2000 and 2010, we adopt remote sensing, GIS technology and statistical methods to investigate the characteristics of farmland dynamics and analyze the drivers of farmland change. We aim to provide insight into mountainous landscape change and its drivers, insight that may help optimize regional land use, implement a farmland requisition-compensation balance, which means the occupied farmland and supplied farmland must be equal in quantity and quality, and ensure food security.
Based on the land use data of Panxi region in 1990, 2000 and 2010, this study describes farmland dynamics and transfer characteristics through spatial econometric methods. We also discuss spatial heterogeneity due to altitude and slope and explore factors driving farmland change with principal component analysis.
Our results show that most of Panxi’s farmland was turned into forest and grassland from 1990 to 2010. The changes in Panxi were very different from those found in the plains of China, where farmland was mainly converted to built-up land. This was determined according to two aspects. The first was the orientation of Panxi region. A primary function of many counties in Panxi is to provide an ecological security barrier, which is the area for regional ecological protection, and the Chinese government has taken measures to protect the environment, while many counties have become the key objects of national ecological protection, which requires the restriction of large-scale urban development, such as Yanbian county. The second was the natural conditions in Panxi’s mountainous areas. To encourage development of the region, several economic measures were taken by the central government, but it was difficult to gather a large population, and many projects were unsuccessful because of geographical limitations. According to the statistics (Figure 10
), Panxi’s urbanization has been increasing, but at a relatively slow rate of change. From 2000 to 2014, the urbanization rate in Panxi developed from 18.75% to 19.88%, only increasing1.13%. Its GDP in 2014 accounted for only 0.35% of the entire country, while its proportion of land area in the whole country is 0.7%, which demonstrated that the economic output per km2
of farmland was low.
This study presented some important issues about the relationship between land and water in Panxi. The standard deviational ellipse for forest and grassland showed a distinct landform-oriented tendency, which meant the forest and grassland were mainly distributed in the high-altitude area. This finding agrees with the fact that the net primary productivity (NPP) was higher in the same high-altitude area [51
]. However, the standard deviational ellipse for farmland and built-up land followed the direction of the Anning River. This suggests that the presence of a river encourages land development. Therefore, settlements and infrastructure should be planned as close as possible to rivers in order to utilize land and water resources most efficiently. Meanwhile, the ratio of y
axes in the standard deviational ellipse was the largest for built-up land, which meant the stability coefficient of built-up land was low, and there would be more freedom in selecting the location for built-up land when the abundance and security for water and land resources are relatively high [52
Our study also identified two other interesting phenomena in Panxi. First, regional farmland has largely been turned into forest on slopes steeper than 15°. This inflection point differs from the standard one in China (≥25°), as identified in the Returning Farmland to Forest and Grassland policy. Due to the fragile ecological environment and the combination of intensified human activities, many areas with slopes less than 25° had already suffered from serious erosion. Hence, it is of practical significance for these farmland regions to return to forest land [53
]. Second, at altitudes below 2500 m, there was a trend of forest conversion into farmland; at altitudes above 2500 m, the trend reversed. Thus, 2500 m tends to serve as a boundary between farmland and forest, which was close to the upper limit for population centers of Sichuan province (2600 m) [54
]. Both of these findings indicate the presence of anthropogenic drivers for land use change.
To capture physical factors affecting farmland changes, this study focused first on temperature and precipitation. Then, in consideration of altitudinal differences, elevation and slope were included in the scope of influencing factors [55
]. Finally, this study selected the distance to rivers (Dis-river) and roads (Dis-road) to represent the effect of location on farmland change. This decision was informed by previous studies that point out the relationship between a farm’s proximity to roads or rivers and that farm’s transformation possibility [56
]. The indicators of physical factors are complete, and the results are reasonable. However, as for the anthropogenic factors, this paper only selected population density and GDP due to data restrictions. We did not consider the quantitative expression of agricultural production efficiency, labor migration or national policy in influencing farmland change, but such factors should be included in future research. It should also be noted that the method of principal component analysis used in this study to explore the driving forces is only based on present data and understanding. Further research is needed to monitor farmland change over longer periods of time to gain a better understanding of the relationships between human activities, natural factors and farmland change.
The spatial pattern of mountainous areas is broken, and the shape of most land use types is haphazard, which brings many challenges in land use interpretation. Therefore, our study adopted the Automated Registration and Orthorectification Package (AROP) to apply ortho-photo rectification and geometric registration to the images. Additionally, a classification scheme, which includes acquirement and expression of knowledge, as well as the application of knowledge in land use classification [57
] and the fusion technology of multi-source remote sensing image [58
], has been applied to obtain land classification. This approach was successful, and the classification accuracy is high (Table 4
). When overlaying different spatial data for factor analysis, information was not lost, even though the resolution of the remote-sensing data was 30 m, and the topographic data were computed based on Shuttle Radar Topography Mission (SRTM) at a 90-mresolution with economic, population and climate data at a 1-km resolution. In conclusion, the results of this study are plausible.