Introduction

Land use and land cover change, proposed by the International Geosphere-Biosphere Program (IGBP) and International Human Dimensions Program on Global Environmental Change (IHDP), is not only the core of land changes science but also the intuitive performance of global changes (Turner et al. 1995). Arable land plays a critical role in rural livelihood sustainability and development, especially in developing countries where agriculture is the mainstay economy. However, there is always a conflict of land use interests between arable land and economic developments as reflected by the loss of arable land to economic developments and urbanization (Long et al. 2009). Therefore, arable land use change reflects the response of a society or the human-land relationship to agricultural production and socio-economic development (Feng et al. 2011).

Land use and land cover change analysis approaches have considerably developed in alignment with the development of remote sensing technology (Liu et al. 2005a, 2014a; Turner et al. 2003), a fact that has allowed a better understanding of the change processes and hence better management. Land use/cover change methods based on remote sensing are mainly divided into two categories: bi-temporal change detection and temporal trajectory analysis (Coppin et al. 2004). The former usually uses remote sensing data of more than two epochs to analyze the land use spatial and temporal dynamics (Zhou et al. 2008a). A limitation of bi-temporal change detection method is that it is not possible to do a detailed land use change of the processes and trends. Temporal trajectory land cover change defined as successional process of land cover types during more than two epochs (Lambin 1997) was proposed to overcome the shortcomings of the bi-temporal change detection method. This approach has been successfully applied to highlight land cover changes and their driving forces (Lu et al. 2004; Kuemmerle et al. 2009; Zhou et al. 2008b; Wang et al. 2012).

Nanjing has experienced rapid and high economic developments and exacerbated urban expansion. Thus, concerns on the arable land spatial temporal dynamics are obvious. However, previous studies on land use of Nanjing metropolitan region are rare. So, what are the detailed change processes and trend in arable land during 1985–2010? This paper applies the temporal trajectory analysis approach to evaluate the spatial–temporal dynamics of arable lands in Nanjing metropolitan region. Thus, this study aimed specifically to: (1) extract arable land and model change trajectories based on Landsat data; (2) analyze spatial pattern of arable land change trajectories coupling with landscape metrics; and (3) illustrate the relationship between arable land change and human-induced factors.

Data and methodology

Study area

Nanjing is the capital of Jiangu and is located in the southwest of Jiangsu Province (Fig. 1). Nanjing metropolitan region includes Jiangning, Pukou, Liuhe District and Jiangnan Eight Districts (including Jianye, Gulou, Baixia, Xuanwu, Xiaguan, Qinhuai, Yuhuatai and Qixia). It holds 7.4 million permanent residents and covers administrative areas of 6,515 km2 (Ouyang et al. 2012). Nanjing has a humid subtropical climate and is under the influence of the East Asia Monsoon (Shi et al. 2012). It experiences a mean annual temperature of 15.7 °C and an average precipitation is 1,106.5 mm (Xu et al. 2010).

Fig. 1
figure 1

The location of Nanjing metropolitan region and the distribution of its directions

Full size image
Fig. 2
figure 2

Changes in arable land area in Nanjing metropolitan region, 1985–2010

Full size image
Fig. 3
figure 3

Arable land change trajectory map

Full size image
Fig. 4
figure 4

The area of arable land change trajectories in various counties

Full size image

Urban expansion of this region is rapid, especially within the metropolitan due to urban developments. This in turn has resulted in loss in arable land. Changes in arable land are mainly located in the urban fringe zone of Nanjing. Therefore, this paper selected Nanjing metropolitan region as study area due to the extended urbanization in the region.

Data source

Landsat 5 Multispectral Scanner (MSS) and Thematic Mapper (TM) remotely sensed imageries covering Nanjing metropolitan region were acquired from US Geological Survey from 1985 to 2010. Optimum imageries were selected based on crop planting calendar, and presence of no cloud cover (Table 1). Land cover samples were obtained using handheld GPS while the socio-economic data were acquired from Nanjing Statistical Yearbook.

Table 1 Path/row and acquiring date of remotely sensed imageries
Full size table

Remote sensing images processing and classification

The ground control points from field work were used to rectify Landsat remotely sensed imageries from 1985 to 2010. Radiometric calibration was used to transform digital values into reflectance values (Markham and Barker 1987; Price 1987; Moran et al. 1992; Liu et al. 2014b). The Landsat MSS image was upscale by resampling it into 30 m.

Five land cover categories, namely, arable land, forest, water body, built-up land and unused land were developed based on the classification criteria of State Bureau of Land Administration and Chinese Academy of Sciences (Liu et al. 2005b, 2014c). Maximum likelihood classification method was employed to classify the images. Finally, the land use classification results were reclassified into arable and non-arable land maps.

Modelling of arable land change trajectories

Land cover change temporal trajectory was defined as chronological arrangement of land cover types. Arable and non-arable land was coded A and O, respectively. Arable land trajectories were calculated using formula 1 as follows:

$$N_{\text{trac}} = \sum\limits_{i = 1}^{n} {\left( {2^{i} \times A_{i} } \right)}$$
(1)

where \(N_{\text{trac}}\) is the number of arable land change trajectories, \(n\) is epoch, \(A_{i}\) is the arable land map in epoch \(i\).

In this article, change trajectories were categorized as: stable arable land, stable non-arable land, change into arable land, change into non-arable land and ephemeral arable land (Table 2).

Table 2 Change trajectories’ categories and their definitions
Full size table

Spatial analysis of arable land change trajectories

Arable land change trajectory analysis in various directions

To reveal spatial variation of arable land changes in different directions, the study area was divided into eight equal-angle sectors around city center (Fig. 1). The directions were numbered counterclockwise as D1–D8 representing the eight directions as follows: D1 “East–Northeast (0°–45º)”, D2 “Northeast–North (45°–90º)”, D3 “North–Northwest (90°–135º)”, D4 “Northwest–West (135°–180º)”, D5 “West–Southwest (180°–225º)”, D6 “Southwest–South (225°–270º)”, D7 “South–Southeast (270°–315º)”, D8 “Southeast–East (315°–360º)”.

Arable land change trajectory analysis in various counties

To analyze the variations of arable land changes in Nanjing metropolitan, the analysis was performed as per the administrative counties.

Arable land change trajectory analysis in various buffers

To reveal spatial variation along the urban–rural gradient, the buffers of 5, 10, 15, 20, 25 and 30 km from the city center were developed using buffer function in the ESRI ArcGIS 9.3 software (Fig. 5).

Fig. 5
figure 5

Distribution of arable land change trajectories in the various buffers

Full size image

Landscape metrics selection

In this paper, number of patches, Interspersion and Juxtaposition and percentage of landscape (Table 3) from class level were selected to analyze fragmentation, composition, coherence and dispersion of arable land change trend in FRAGSTATS 3.3 software (McGarigal et al. 2002).

Table 3 Description of landscape pattern metrics used in this paper
Full size table

Result and analysis

Spatial and temporal dynamics of arable land in Nanjing metropolitan region

Arable and non-arable land maps in six epochs from 1985 to 2010 were obtained with overall accuracy and Kappa coefficient higher than 85 % and 0.84, respectively. Arable land increased from 1985 to 1990, then decreased dramatically until 2010 (Fig. 2). Arable land decreased by 57.24 % during 1985–2010. Arable land increased by 7.74 % between 1985 and 1990 while it decreased by 60.31 % from 1990 to 2010. The maximum change magnitude of arable land was 38.93 %, which occurred in the 2007–2010 period.

Spatial distribution of arable land change trajectory

Arable land changes trajectory in various directions

Arable land change trajectory map was extracted based on the arable/non-arable land maps from 1985 to 2010 (Fig. 3). The percentage of five arable land change trajectory types was calculated in eight directions (Table 4). Stable arable land category had the largest proportion (56.89 %) in direction three (North–Northwest, 90°–135º) and the smallest proportion in direction one. Stable non-arable land is mainly dominant in the directions one and three. Change into non-arable land was mainly dominant in the directions one, six, and eight, which revealed the main direction of urban expansion.

Table 4 Percentage of arable land change trajectories in eight directions (%)
Full size table

Arable land changes trajectory in various counties

Liuhe District had the highest stable arable land (895.92 km2) (Fig. 4). Jiangning District was the second largest with stable arable (696.37 km2) while Jiangnan Districts had smallest stable arable land. Change into non-arable land was the maximum in Jiangning District where numerous arable lands were transformed into construction land. In addition to urbanization, another factor that contributed to the observed arable land loss was construction of aquaculture ponds.

Arable land changes trajectory analysis in various buffers

Stable arable land was mainly concentrated in the area far away from city center of Nanjing with 55.16 % being in the buffer between 25 and 30 km (Fig. 5; Table 5). The percentage of the stable arable land increased with increasing distance from the city center. A complete reverse trend was, however, observed for the non-arable dynamics. The percentage of stable non-arable land decreased from 70.3 % in buffer 0–5 km to 8.745 in buffer 25–30 km. Ephemeral arable land decreased from 1.97 to 0.77 % with increasing distance from the city center. The change trend of ephemeral arable land illustrated that areas nearby the city center changed dramatically and arable land was the most non-stable. Non-arable land experienced the highest (59.2 %) and the lowest (23.88 %) change within 10–15 km and 0–5 km buffer, respectively. Change into arable land was the lowest (<3 % in six buffers) among the change trend types.

Table 5 Statistics of arable land change in different buffers
Full size table

Arable land change trajectory landscape pattern analysis

The percentage of stable arable land was 46.72 %, implying that it constitutes a major land use component in the study area (Table 6). The second largest land cover was change into non-arable land (36.93 %). The percentage of change into non-arable land revealed sharp reduction of arable land. Stable non-arable land accounted for 13.72 %, including built-up land, water body and forest. Change into arable land and ephemeral arable land accounted for 1.83 and 0.80 %, respectively. The former was caused by land reclaimed due to land policy while the latter was mainly caused by agricultural restructuring.

Table 6 Landscape metrics of arable land change trajectories
Full size table

Interspersion and Juxtaposition (IJI) of stable arable land and stable non-arable land were both less than 35, implying the two types had high spatial association and high stability. Ephemeral arable land had the highest IJI value (80.25) reflecting the extreme unstable nature of these land cover types. The patch number of change into non-arable land was the largest, indicating high landscape fragmentation associated with conversion of arable land into built-up land as a result of urbanization. Among all the change trend types, stable non-arable land had the lowest patch numbers implying the solidarity nature of the stable non-arable lands.

Driving factors analysis

The changes in arable land were mainly human induced. Rapid economic developments, urban expansion and agricultural restructuring were the main driving forces associated with the observed decreased trends in arable land during 1985–2010. Thus, urbanization, agricultural restructuring, policies and rural labor migration were selected to illustrate the mechanism of arable land changes.

Urbanization

Loss of arable land was associated with increase in build-up areas as a result of urbanization (Zhang et al. 2011). The major urbanization was located in the south of Yangtze River. The southwest and east of Nanjing have been both major urban expansion directions since 1988, especially in Jiangning District (Xu et al. 2007). Urban expansion encroaches the nearby arable lands, which explains the high instability of arable lands within the vicinity of the center. Land use intensity in the industrial parks and economic development zones was low, which caused loss in high quality arable land. In addition, large numbers of national infrastructures (such as expressway) were developed on arable lands.

Rural labor transformation

Due to economic development, a huge new floating population of rural–urban migrants is transforming the urban labor force in China (Seeborg et al. 2000; Bhattacharyya and Parker 1999). According to the statistical data from National Bureau of Statistics of China (NBSC), the number of rural labors in Nanjing Metropolitan Region has gradually decreased by 49.2 % from 1985 to 2010 (NBSC 2011). In addition, the age structure of rural labor also changed, with increasing trends of large numbers of young labors opting for non-agricultural city jobs. This has had indirect implications on the arable land change dynamics in that arable lands were abandoned due to lack of labor or high cost thereof, giving room for conversion of the land into other land cover uses. A report from the National Bureau of Statistics of China showed that youth laborers between the ages of 26–45 occupied 64.41 % of the total rural labor migration in 2005 (NBSC 2005).

Agricultural restructuring

Driven by variations in the market efficiency of agricultural products (Zhao et al. 2012; Veeck and Pannell 1989), farmers were enticed to transform arable land to aquaculture and orchard. However, this process is unstable and can be sometimes inversed caused by changes in market efficiency, land use and water resource protection policies. This agricultural restructuring has resulted in instability of arable land changes.

Policy-driven factors

Land policy and rural tax reforms and agricultural subsidies have acted to slow down arable land reduction. Chinese Land Management Laws were promulgated and revised twice in 1998 and 2004 to establish land use controlling system. The revised land policies proposed arable land protection (Nanjing Municipal Government 2007). Reforms on rural taxes changed agricultural production distribution income and this increased farmers’ income. Agricultural subsidies, such as grain, seed and fertilizer subsidies, improved the comparative advantage of agriculture to increase grain production (Long and Zou 2010).

Conclusions

Detailed arable land change dynamics using the temporal change trajectory approach based on Landsat data was analyzed. Conversion of arable land into non-arable land was mainly dominant within the urban fringes due to rapid urbanization and industrialization. Change into non-arable land was dramatic and characterized by high fragmentation compared to other change trends. Such change trends were attributed to high urbanization. Rural labor migration, economic development, policies and urban expansion are the major driving forces to the observed arable land dynamics.