Characteristics of Land Use Change in China before and after 2000

Abstract

China, with notable population blooming and economic development in the last decades, has experienced profound land-use changes, which, in turn, dramatically impacted the regional, even global and environment system. However, characteristics of land-use changes in China have not yet been well addressed, especially around the year 2000 when a series of land policies were put forward, such as the project of “returning farmland to forest”. To fill this gap, this paper investigated the temporal and spatial patterns of land use changes in China for the period from 1987 to 2010, by taking advantage of the continually updated China Land Use Database developed from remote sensing images. The land-use dynamic matrix, zonal model, and transition matrix were employed to characterize land-use change patterns for four time intervals (1987–1995, 1995–2000, 2000–2005, and 2005–2010) on the dimensions of conversion and modification. Results showed that land-use change affected 4 × 10⁵ km² (4.5%) of the total landscape in China for more than the past twenty years. Of the six land-use types, built-up land experienced the largest net increase by almost 30% (52,434 km²), with the rate of expansion accelerating after 1995. The area of cropland increased before 2000 and declined afterwards, ending with a net increase in 14,280 km², approximately 1% of its original area. The loss in the eastern coastal region is attributed mainly to built-up land expansion, while the gain in northern China, with the price of grassland and woodland shrinking, reshaped the cropland distribution in China. The area of woodland decreased slightly by 7880 km² without a clear pattern over time. The modification of woodland indicated an intensive forest management in terms of planting fast-growing trees in the south of China. Grassland continues to shrink at a decreasing rate, and the modification of grassland shows a tendency of transformation from sparse grassland into a dense one in the 21st century. Trade-offs among demands on food security, economic development, and environment protection forced and shaped the contemporary land-use change in China. These results contribute to understanding the trends and causes of land use change in China, which could provide underpinning knowledge for assessing environmental change, and provide insights on future land planning.

1. Introduction

The monitoring of land-use change has a fundamental role in understanding the impacts of anthropogenic activities on the global environment [1,2,3] and its feedbacks [4,5]. This also has a significant impact on the harmonious and sustainable development of the environment and human beings. As one of the most active economies in the world, China has been experiencing profound land-use change, especially following economic reform and opening up in the 1980s. These changes were forced by the huge and increasing demand on natural resources caused by population growth, as well as rising income level. The enormous disturbance of environmental systems, such as climate change [6,7], land degradation [8,9], water depletion [10,11], loss of biodiversity [12,13], and so on, spurred by substantial land-use change has attracted ongoing domestic and international concern.

Around 2000, a series of land policies were put forward in order to cope with ecological environment change, a large population, and the shortage of land resources caused by urbanization development, and to ensure the quality and quantity of cultivated land [14]. For example, the pilot project of “returning farmland to forest” was carried out in 1999 [15], and the policy of “returning farmland to forest” was implemented in 2002. The year 2000 was also an important point of land-use and land-cover change [16,17]. Jiyuan Liu pointed out that compared with the last decade of the 20th century and the first decade of the 21st century, the spatial pattern of land-use change in China showed some new characteristics [18]. By revealing the in-depth characteristics of land-use change in China from 1987 to 2010, we can focus on the details of this transition.

Research into the driving forces behind land-use change in this period has also attracted much more interest with a view to providing insights into sustainable policy formulation [19,20]. A spatially and temporally explicit investigation of land-use change across the country, as the foundation of this research, has been urgently needed. Large-scale surveys and long time series have been difficult until the advent of remote sensing. Since the emergence of these techniques, they have been used to conduct extensive research into land-use change in China [21,22,23], providing much of our knowledge of the main processes and hot spots of land-use transformation [24,25,26]. However, exiting research cannot fully meet the needs. In the 1980s, a group of national agencies in China, including the Agricultural Regional Planning, National Administration of Surveying, etc., constructed a national land-use map of China using Landsat MSS (Multispectral Scanner). However, no review of subsequent land-use change has been conducted on the basis of this database. Later, the Ministry of Land and Resources conducted another land-use mapping of China, completed in 1996, on the basis of aerial photographs, field surveys, and other available maps [19]. Land-use change has been monitored on the basis of this database every year since its establishment. However, a survey approach was used, and biases were anticipated because of the subjectivity of this type of approach. Although the latest update was conducted in 2007 using remote-sensing techniques, both the data and methods are different from those of the initial mapping. Spatially and temporally consistent analysis is not possible based on this database.

In the late 1990s, the Chinese Academy of Sciences, together with the Ministry of Agriculture, National Administration of Forest, and other agencies, started to build the China Land Use Database (CLUD) using remote-sensing data. Land-use maps of China in 1995 were created. Following that, land-use-change maps in the periods of 1995–2000, 1987–2000, 2000–2005, 2005–2008, and 2008–2010 were generated sequentially. On the basis of these, land use maps in 1987, 2000, 2005, 2008, and 2010 were obtained. Jiyuan Liu [27] introduced the initiation of the construction of this database, and then analyzed the spatial pattern of land-use change in China over this period [28,29]. Spatial patterns and driving forces of land-use and cover change (LUCC) in the first five years of the 21st century were later investigated [30]. However, the spatial and temporal patterns of land-use change around 2000, an important transition period, are still unexplored.

Moreover, the previous research on land-use change consistently focused on those land-use changes in which one land-use type was completely replaced by another. More subtle changes in operations on the land, which do not change its overall use classification (i.e., changes between subtypes of certain land-use types), have not been well addressed. These changes also have a remarkable influence on global environmental change. For example, an increase in paddy land for rice production, a subtype of cropland, will most likely be accompanied by an increase in methane emissions, which is recognized as one of the most important greenhouse gases [31].

This paper explores the temporal–spatial pattern of land-use changes of China for from the late 1980s to the first decade of the new century, considering both land-conversion and land-modification dimensions. Here, we define land use conversion as the transformation of land use types. We refer to land use modification to mean the land-use type is not changed, but the subtype has been changed. First, a land-use-change dataset with approximate 5-year intervals from 1987 to 2010 was established by reorganizing and integrating the land-use-change maps of the CLUD. On the basis of this, overall changes of land-use system in China for the past two decades were characterized by employing a land-use dynamic matrix (LUDM). Thematic temporal–spatial change patterns were also analyzed for the land-use types experiencing major changes, i.e., cropland, woodland, grassland, and built-up land, to highlight the main processes of land-use change. Then, transition patterns of these land-use types were analyzed to disclose the proximate causes of land use changes. Finally, the underlying causes of land-use change in China were discussed.

2. Data

The CLUD, scaled at 1:100,000, was employed in this paper for temporal—spatial analysis. To facilitate the analysis of temporal patterns of land-use change, the land-use maps were integrated into a new land-use-change dataset with evenly distributed time intervals. A land-use dynamic matrix was developed for quantitatively and comprehensively measuring the amount and extent of both land-use conversion and land-use modification. A zonal model and transition matrix were employed for characterizing the thematic changes of largely changed land-use types and the proximate cause of these changes.

2.1. China Land Use Database (CLUD)

The CLUD was developed using remote-sensing images obtained mainly from Landsat TM (Thematic Mapper)/ETM+ (Enhanced Thematic Mapper Plus). Photographic film and thousands of images from CBERS (China–Brazil Earth Resource Satellite) and HJ-1 (Small Satellite Constellation for Environment and Disaster Monitoring Forecasting) were adopted when TM/ETM+ images were not available. The CLUD includes land-use maps for 1987, 1995, 2000, 2005, 2008, and 2010, and land-use-change maps for the periods 1987–2000, 1995–2000, 2000–2005, 2005–2008, and 2008–2010. The classifications used in this database are given in

Table 1. Classification of China Land Use Database.

Land Use Types	Subtypes
Cropland	Paddy land/Dry land
Woodland	Forest/Shrubs/Woods/Others
Grassland	Dense grassland/Moderate grassland/Sparse grassland
Water area	River and drainage/Lake/Reservoir and pond/Ice and permanent snow/Tideland/Bottomland
Built-up land	Urban area/Rural area/Other construction land (such as factory, mine, industrial zone, and so on)
Unused land	Sandy land/Gobi/Saline land/Swampland/Bared land/Rocky land/Others (such as tundra)

.

Visual interpretation was used to detect land-use change by integrating analyses of the spectral reflectance, location, and shape of objects [32]. A series of auxiliary datasets, such as maps of soil type and vegetation, topographic maps, information on regional planning, and so on, were also combined to improve the precision of interpretation. Research groups from eight institutes of the Chinese Academy of Sciences were involved in the construction of the CLUD. Most of them were located in different provinces and conducted a visual interpretation of land use in the region near their location. Field surveys were conducted to assess the accuracy of land-use databases, as well as to consolidate the experts’ experience in land-use identification [33]. Details of the assessed database accuracy are presented in

Table 2. The accuracy of China Land Use Database according to field surveys in different periods.

Period		Cropland	Woodland	Grassland	Built-Up Land	Water	Unused Land	Total
1995	Num	5058	4104	1512	1714	912		13,300
	A (%)	94.94	90.13	88.16	96.32	95.72		92.92
1980s–2000	Num	20,153	7553	4903	1147	1857	676	36,289
	A (%)	97.02	97.30	96.80	99.04	94.67	95.56	96.67
1995–2000	Num	99,867	4506	6631	8055	2561	1526	33,929
	A (%)	99.09	98.92	97.96	98.92	96.86	97.88	98.04
2000–2005	Num	11,701	6266	5199	7382	4056	1392	35,877
	A (%)	99.30	97.75	98.62	97.01	97.61	98.28	98.56
2005–2008	Num	4235	9584	2073	7589	2025	540	26,046
	A (%)	98.23	99.08	96.53	95.41	95.46	96.67	97.34
2008–2010	Num	3218	9975	1560	10,872	2194	296	27,565
	A (%)	98.94	97.36	98.59	96.44	95.92	97.64	97.15

Note: Num denotes number of verified patches; A denotes accuracy (=correct patches within total verified patches/total verified patches). In 1995, water and unused land were verified together, with a total of 912 patches, and the accuracy was 95.72%.

[33].

2.2. Integrity of Land-Use-Change Database (CLUD)

As the temporal intervals between the original land-use-change maps in the CLUD are not all the same, comparison of the characteristics of land-use change among these periods would be unreasonable. Although annual change-rate maps could solve this problem, this is not the best option, as the detection of land-use change at a finer temporal scale captures more detailed information. To facilitate the analysis of temporal patterns in the land-use changes, we developed a set of land-use-change maps, at approximately five-year intervals, on the basis of the original database. To facilitate the integration, the land-use-change maps in vector format in the CLUD were converted into maps in a 100 m grid format.

First, land-use-change maps for the period 1987–1995 were generated by integrating two maps based on changes from 1987 to 2000 and from 1995 to 2000. We divided the overall land-use-change patches in the two periods into three types of situations: (1) change patches existing only in the period 1987–2000; (2) change patches existing only in the period 1995–2000; and (3) change patches existing in both periods. For Situation 1, the change type in the target period was deemed to be the same as that in the period 1987–2000, while for Situation 2, the change type was the opposite of that in the period 1995–2000. For Situation 3, the land-use type that was converted from ‘lost’ in the period 1987–2000 was assigned as the land-use type ‘lost’ in the period 1987–1995, while the land-use type that was converted from ‘lost’ in the period 1995–2000 was assigned as the land-use type ‘gained’ in the target period 1987–1995.

Second, land-use-change maps for the period 2005–2010 were generated by integrating two maps covering the periods 2005–2008 and 2008–2010. The overall land-use-change patches in both periods were also divided into three types: (1) change patches existing only in the period 2005–2008; (2) change patches existing only in the period 2008–2010; and (3) change patches existing in both periods. For the land-use change in the target period 2005–2010, we assigned the same land-use-change type as in the periods 2005–2008 and 2008–2010 in Situations 1 and 2, respectively. In the third situation, we used the land-use type that was lost in the period 2005–2008 as the land-use type ‘lost’ in the period 2005–2010, while the land-use type gained in the period 2008–2010 was assigned as the land-use type ‘gained’ in the period 2005–2010.

Finally, by using the same method as in the second step, we can produce land-use-change maps for the entire period from 1987 to 2010. In this way, the land-use-change dataset now includes land-use-change maps for the periods 1987–1995, 1995–2000, 2000–2005, 2005–2010, and an aggregated map for 1987–2010.

3. Methods

3.1. Land-Use Dynamic Matrix for Understanding Overall Land-Use Change

In a land-cover system, the two types of changes are referred to as land-cover conversions and land-cover modifications [4,34]. Here, we use land-use conversions and land-use modifications to define the corresponding phenomena in a land-use-change system. For each land-use type, we developed an LUDM including three dimensions: net conversion (NC), gross conversion (GC), and modification (MD). NC means the areal change in each land-use type, indicating the overall result of land-use change. GC is the summation of the areas where the land-use type is either gained or lost, indicating the total area affected by any changes in that land-use type. MD refers to the area undergoing intra-type changes. Furthermore, three variables are used to quantitatively describe the changes in these dimensions: change amount (CA), change rate (CR), and change percentage (CP). CA is the total area of the change, CR is the annual change in area over a certain period of time, and CP is the percentage of CA relative to the original area of the corresponding land-use type. Finally, the land-use dynamic matrix can be described as:

LUDM = [NC_ij, GC_ij, MD_ij],

(1)

where i refers to one of the six land-use types, i.e., cropland, woodland, grassland, water areas, built-up land, or unused land, and j refers to one of the three variables of land-use change, i.e., CA, CR, or CP. We use the LUDM to describe the overall change in land use in China, and some selected indexes of this matrix are adopted for temporal and spatial pattern analysis.

3.2. Zonal Model for Spatial Pattern Detection

A zonal model, providing a visible spatial distribution map of land-use change, was used extensively in the analysis of land-use-change characteristics [35]. In this paper, two types of 10 km² gridded zonal products were obtained. One shows spatial pattern of land-use conversion, including the amount of gross conversion for the whole land-use system and net conversion for four largely changed land-use types. The other shows the spatial pattern of land-use modification, including the amount of modification of the whole land-use system, as well as four largely changed land-use types.

Specifically, we first extracted dynamic patches from the land-use-change maps. Then, a grid frame for the whole of China was generated in vector format using the FISHNET module in ArcGIS software. Each cell of the grid was 10 km by 10 km. Finally, we used the 10 km² grid to intersect with the dynamic patches, and thereby obtained the change amount in each 10 km² grid square.

3.3. Transition Matrix for Proximate Causes Exploration

A transition matrix, which identifies the direction and magnitude of land-use transformations [36,37], is a useful technique for analyzing observed land-use changes. This type of information is important in understanding the causes of land-use change, and also in analyzing the effects of these changes on the eco-environmental and climatic systems [37,38,39,40]. Here, we used three attributes of the land-use-change map to calculate transition matrixes for each period—the land-use-type codes of the transformation patches at the beginning and end of each period and the areas of the patches. Cross-tabulation tables were then calculated to obtain transition matrices.

The area–percentage method was adopted to explore the structure of land-use types as the source of certain newly emerging land-use types and the structure of the target land-use types into which certain land-use types were being converted. The contribution of one land-use type to the increase of another land-use type, CR, and the contribution of one land-use type to the loss of another land-use type, OP, are defined as

$${\mathrm{CR}}_i={\mathrm{A}}_{ji}/{\mathrm{A}}_{ii}$$

(2)

$${\mathrm{OP}}_i={\mathrm{A}}_{ij}/{\mathrm{A}}_{il}$$

(3)

where A_ji denotes the area converted from land-use type j to land-use type i; A_ii denotes the total gained area of land-use type i; A_ij denotes the area converted from land-use type i to land-use type j; and A_il denotes the total lost area of land-use type i.

We chose the three largest CR or OP values to best describe the transition patterns of land-use conversion of each type. To characterize land-use modification, the three types of land-use modification with the largest area were listed for each land-use type. However, there were two exceptions. For cropland, there are only two subtypes and, therefore, only two types of modification, which were listed. For grassland, the first four types were listed, as these were required to determine whether the modifications lead to degradation or recovery of grassland.

4. Results

For the entire period from 1987 to 2010, an area of 4 × 10⁵ km², or approximately 4.5% of the total landscape, has been either converted (3.0%) or modified (1.6%), while approximately 0.1% of the landscape has experienced both types of change. Within the region experiencing either land-use conversion or land-use modification, 19.6% was observed to have changed more than once across the four periods (i.e., 1987–1995, 1995–2000, 2000–2005, and 2005–2010): 17.7% changed twice, 1.8% changed three times, and 0.13% changed four times.

4.1. Overall Changes during the Entire Period

The land-use system in China has experienced enormous changes in both dimensions of land-use conversion and land-use modification. Land-use conversion always results in an area change of land-use types, while land-use modifications affect the structure of land-use types, leaving its area unchanged. During the period in question, land-use conversion was more widely distributed compared with land-use modification (

Table 3. Land use changes in China described by Land-Use Dynamic Matrix for the entire study period.

Land-Use Type	Net Change			Gross Change			Modification
	CA (km2)	CR (km2/y)	CP (%)	CA (km2)	CR (km2/y)	CP (%)	CA (km2)	CR (km2/y)	CP (%)
Cropland	14,280	621	1.0	159,930	6954	11.3	25,979	1130	1.8
Woodland	−7880	−343	−0.4	68,877	2995	3.0	51,748	2250	2.3
Grassland	−50,963	−2216	−1.8	127,709	5553	4.4	29,510	1283	1.0
Water area	4731	206	1.8	28,791	1252	11.1	9926	432	3.8
Built-up land	52,434	2280	29.9	53,062	2307	30.3	3353	146	1.9
Unused land	−10,972	−477	−0.5	51,637	2245	2.4	2532	110	0.1

Note: CA denotes change amount; CR denotes change rate; CP denotes change percentage.

).

Three out of six land-use types increased in area: cropland, water areas, and built-up land. Built-up land changed the most, increasing by 52,433 km², thereby adding 29.9% to its entire area in 1987. Largely converted from other types of land use while rarely converted into other land-use types, the gross conversion of built-up land, which is 53,062 km², was approximately the same as its net conversion area. Conversions related to cropland affected the landscape most widely, with a gross area of change of 159,930 km², or approximately 11.3% of the 1987 cropland area. The cropland area showed a net increase of 14,280 km². Woodland and grassland, two land-use types related to natural vegetation, decreased in area by 7880 km² and 50,963 km², respectively, during the study period. Grassland was also the second-most-disturbed land-use type, following cropland, with a gross conversion area of 127,709 km², and the largest net decline of any types in area of 50,963 km².

The gross conversion of each land-use type is usually larger than the net conversion. The ratio of gross conversion area to net conversion area ranged, in our study, from 1.0 (built-up land) to 11.2 (cropland), indicating that the total area affected by land-use conversion is substantially more widely distributed than is suggested by the net changed area.

In the entire study period, woodland had the largest area of modification, 51,748 km² or 2.3% of its 1987 area. The modification of water areas reached 3.8% of the total area in 1987, although the modified area was only 9926 km². Grassland and cropland also had relatively high modification areas: 29,510 km² and 25,979 km², respectively; 1.0% and 1.8%, in terms of percentage.

4.2. Temporal Patterns for Land-Use Change

Yearly gross land-use conversion in China declined continuously during the study period from 16,195 km²/y for the period 1987–1995 to 10,161 km²/y for the period 2005–2010 (Figure 1b). This indicates a mitigation of anthropogenic landscape disturbance. Yearly land-use modification, of a much smaller magnitude than the gross conversion, fluctuated within a narrow range between 7578 km²/y (2000–2005) and 8236 km²/y (2005–2010) (Figure 1c).

Figure 1 and Figure 2 show the variation characters in different periods. The net area of cropland increased prior to 2000, at a rate of 1547 km²/y for the period 1987–1995 and 3422 km²/y for the period 1995–2000 (Figure 1a). However, after 2000, cropland area declined at an increasing rate, specifically at 1377 km²/y for the first five years and 1664 km²/y for the second five years. The gross conversion rate of cropland showed a declining tendency, dropping dramatically from 10,814 km²/y in the first period to 6339 km²/y in the last period (Figure 1b). A similar trend was also found in the land-modification rates (Figure 1c). Through the reduction of cropland, we can see that the scope of the impact of human activity was shrinking.

Built-up land area continued to expand during the study period, with the rate soaring to 4029 km²/y in the period 2005–2010, which is 4.2 times the rate in the period 1995–2000, the period of slowest growth. The modification of built-up land was the lowest among the six land-use types.

Counter to this, grassland was continuously lost, but at a declining rate, from 2523 km²/y at the beginning of the overall study period to 1295 km²/y at the end. The gross area of grassland conversion decreased from 9121 km²/y in the first period to less than half this amount at the end. The rate of grassland modification dropped to 494 km²/y for the period 2005–2010 after peaking at 3067 km²/y between 1995 and 2000. A mitigating influence on grassland disturbance clearly emerged during the study period.

Woodland, another type of natural vegetation, slightly increased and then shrank at a rate of 2498 km²/y between 1995 and 2000, yet increased again by a slight expansion. Although the rate of gross conversion of woodland decreased from 6135 km²/y at the beginning to almost one-third of this rate at the end, the modification rate within woodland rose rapidly to 4097 km²/y after reaching the bottom in the period 1995–2000. Intra-type modification had become a more important role in the change of woodland, compared with the conversion.

4.3. Spatial Patterns for Land-Use Change

Hot spots of land-use conversion since 1987 have been concentrated mainly in three parts of China: the northeast, the northwest, and the eastern coastal area (Figure 3a). For the first two regions, conversions among cropland, grassland, and woodland were the dominant land-use change types, while built-up land expansion was clearly evident in the third region. The areas of greatest land-use modification were in northeast and southeast China, mainly in vegetation-related land-use-change types (Figure 3b).

Regions with a high intensity of cropland conversion coincided with the areas of major land-use conversion mentioned above. Northeast and northwest China, including Heilongjiang, Jilin, Liaoning, west of Inner Mongolia, and Xinjiang provinces, were the main areas where cropland expanded dramatically (Figure 4a). Grassland and woodland were the dominant sources of land for cropland expansion in these regions (Figure 4b,c). The eastern coastal region was the area with greatest cropland loss, especially in the Jing-Jin-Tang region, the Yangtze River Delta, and the Pearl River Delta. Some areas in Hubei, Sichuan, and Chongqing also surfaced as hot spots of cropland loss. All of these regions experienced rapid economic development, and built-up land expansion was the main reason for cropland loss in these regions. Cropland modification occurred mostly in northeast China, including Heilongjiang, Jilin, and Liaoning provinces (Figure 4b). Conversion from dry land to paddy land accounted for more area than the opposite conversion. Jiyuan Liu also pointed out that the center of gravity of new croplands gradually moved from the northeast to the northwest [18].

Woodland conversions were concentrated in northeast China, the Loess Plateau, southeast China, and southwest China (Figure 4c). Among these regions, northeast China suffered the greatest woodland loss, which can be attributed to cropland expansion, while in other regions, the area of woodland increased in the study period. Grassland was the main source contributing to the woodland expansion in southeast China, while mixed grassland and cropland were the dominant sources for the Loess Plateau and southwest China. Great Khingan Mountain and Lesser Khingan Mountain in northeast China were among the hot spots for woodland modification (Figure 4d). Transformation from forest to shrubs, woods, and others was the main woodland modification type. Another hot spot of woodland modification was in southern China, including Fujian, Guangdong, and Guangxi provinces, where transformations resulted mainly from logging or fire slashing, followed by replacement with fast-growing forest.

Most of the grassland in China is in the northwest area, where a large amount of grassland conversion was also observed (Figure 4e). Within this region, most of Xinjiang province showed a dramatic loss of grassland due to cropland expansion, while grassland loss and gain coexisted in Inner Mongolia province. Cropland and unused land were the main source for grassland expansion, whereas grassland loss was attributed mainly to conversion to cropland. The modification of grassland was most evident in Inner Mongolia province (Figure 4f). The degradation of grassland resulting from grassland modification outweighed the areas of recovery.

Built-up land expansion is considered an irreversible process, so built-up land conversion presented mainly as gains in built-up land area. The expansion of built-up land occurred mainly in the east of China. Urban agglomeration regions in the middle reaches of the Yangtze River, Sichuan, and Chongqing also showed a rapid expansion of built-up land (Figure 4g). Cropland was undoubtedly the main source. Modification of built-up land occurred mainly in the North China Plain, the Yangtze River Delta, and the Pearl River Delta (Figure 4h). Transformations from rural land, and traffic and construction land to urban land were the main types of built-up land modification. Jiyuan Liu also summarized that the basic characteristic of urban and rural construction land change is the accelerated expansion from the eastern to the central and western regions [18].

4.4. Proximate Causes for Land-Use Change

4.4.1. Proximate Causes for Land-Use Conversion

Cropland was the most widely disturbed land-use type in China over the study period. Basically, built-up land, grassland, and woodland were three major land-use types that claimed area from cropland (

Table 4. Transition patterns of land use conversion for different periods.

		1987–1995		1995–2000		2000–2005		2005–2010
		LUT	(%)	LUT	(%)	LUT	(%)	LUT	(%)
OP for Woodland	1st	Cropland	56.4	Cropland	57.7	Grassland	37	Built-up land	43.4
	2nd	Grassland	37.1	Grassland	38	Cropland	34.3	Grassland	26.6
	3rd	Built-up land	2.9	Built-up land	1.6	Built-up land	22.3	Cropland	20.5
OP for Grassland	1st	Cropland	55.9	Cropland	53.3	Cropland	47.1	Cropland	54.1
	2nd	Woodland	31.5	Unused land	26.3	Unused land	24.5	Woodland	27.3
	3rd	Unused land	10.2	Woodland	13.5	Woodland	21.5	Built-up land	9.5
CR for Woodland	1st	Grassland	57.4	Grassland	62.1	Grassland	53.4	Grassland	64.2
	2nd	Cropland	38.8	Cropland	33.2	Cropland	42.3	Cropland	33.3
	3rd	Unused land	2.2	Unused land	2.8	Unused land	2.9	Water area	1.2
CR for Grassland	1st	Cropland	39.7	Woodland	45.9	Cropland	47.5	Unused land	42.1
	2nd	Woodland	33	Cropland	27.4	Unused land	25.7	Cropland	29.2
	3rd	Unused land	22.3	Unused land	23.6	Woodland	21.4	Woodland	20.9
CR for Built-up land	1st	Cropland	85.7	Cropland	78.9	Cropland	75.1	Cropland	69.9
	2nd	Woodland	5.3	Grassland	8.2	Woodland	9.9	Woodland	12.1
	3rd	Grassland	3.3	Woodland	6.2	Water area	5.5	Grassland	6.4

Note: first three transition types contributing most to the loss and gain of each land-use type are given in the table. OP denotes the contribution of one land-use type (LUT) to the loss of target land-use type; CR denotes the contribution of one land-use type to the gain of target land-use type.

). Among these land-use types, built-up land was the largest consumer of cropland. The percentage of cropland loss due to built-up land expansion saw a slight drop in the period 1995–2000 followed by a rapid increase that peaked at 70.7% at the end of the study period. The sources for cropland expansion were primarily grassland, woodland, and unused land. Grassland made the major contribution throughout the study period, with its contribution to cropland gain rising from 52.7% in the period 1987–1995 to 63.2% in the period 2005–2010. The contribution of unused land also became increasingly important, while the woodland contribution had the opposite tendency.

The expansion of cropland, grassland, and built-up land was the main reason for woodland loss. The percentage of woodland loss due to cropland expansion increased slightly in the second period and then declined to 20.5% in the period 2005–2010. Expansion of built-up land, on the other hand, had an increasing influence, causing 43.4% of the total woodland loss in the last period. Throughout the study period, grassland was the major source of areas of woodland gain, followed by cropland. Unused land and water areas also contributed a little to woodland gain.

The conversion of grassland was associated mainly with changes in cropland, woodland, and unused land. Conversion to cropland was the most important reason for grassland loss. The percentage of grassland loss due to new cropland varied from 55.9% in the first period to a low of 47.1% between 2000 and 2005 before rebounding to 54.1% at the end of the period. Conversions from grassland to woodland and unused land were also very common in the study period. The sources for grassland expansion were basically cropland, woodland, and unused land. The contribution of unused land grew steadily during the study period, while the contribution of woodland showed the opposite tendency. Although the percentage of cropland lost to grassland fluctuated, it still played a significant role in grassland gain.

Built-up land expansion was a dominant land-use-change process over the study period, and cropland was always its major source. However, the area of built-up land sourced from cropland showed a continuous decline in percentage terms, from 85.7% at the beginning to 69.9% at the end. Apart from cropland, woodland, grassland, and water areas were also sources of built-up land expansion.

4.4.2. Proximate Causes for Land-Use Modification

Land-use modification is a type of land-use change that always leads to the alternations to the structure of land-use types. For more than the past twenty years, cropland modification did not show a clear pattern in terms of temporal development, while built-up land modification was dominated in all periods by conversions from rural land and others to urban land, accounting for more than 90% of the modifications (

Table 5. Transition patterns of land use modification for different periods.

		1987–1995		1995–2000		2000–2005		2005–2010
		Type	CP (%)	Type	CP (%)	Type	CP (%)	Type	CP (%)
Crop-land	1st	Paddy→Dry	55.1	Dry→Paddy	80.5	Paddy→Dry	68.4	Dry→Paddy	61.9
	2nd	Dry→Paddy	44.9	Paddy→Dry	19.5	Dry→Paddy	31.6	Paddy→Dry	38.1
Wood land	1st	Others→Forest	29.6	Forest→Shrubs	53.4	Forest→Others	33.4	Forest→Others	47.4
	2nd	Forest→Shrubs	18.1	Forest→Woods	13.3	Others→Forest	17.5	Others→Forest	21.0
	3rd	Forest→Woods	9.8	Woods→Forest	7.8	Woods→Forest	15.5	Woods→Forest	8.5
Grass land	1st	Sparse→Dense	29.2	Dense→Moderate	31.8	Moderate→Dense	23.0	Sparse→Moderate	32.9
	2nd	Dense→Moderate	20.2	Moderate→Sparse	25.2	Moderate→Sparse	21.7	Moderate→Dense	31.1
	3rd	Moderate→Dense	19.0	Dense→Sparse	18.8	Sparse→Moderate	20.7	Sparse→Dense	15.3
	4th	Sparse→Moderate	14.5	Sparse→Moderate	11.1	Sparse→Dense	12.6	Moderate→Sparse	10.0
Built-up land	1st	Rural→Urban	84.7	Rural→Urban	89.8	Rural→Urban	83.1	Rural→Urban	70.2
	2nd	Others→Urban	11.7	Others→Urban	5.0	Others→Urban	15.5	Others→Urban	25.1
	3rd	Rural→Others	1.3	Rural→Others	3.3	Others→Rural	0.8	Rural→Others	1.5

Note: CP denotes percentage of the area of certain transition type in total modification area. Generally, first three transition types with largest area were listed, except cropland (only first two) and grassland (first four).

).

For woodland, significant deforestation occurred in the period 1995–2000, with the transformation from forest to shrubs or woods, accounting for 66.8% of woodland modification. However, on entering the 21st century, large areas of other types of woodland were transformed into forest, accounting for 33.0% and 29.5% of woodland modification in the first and second periods, respectively. Although the largest amount of woodland modification after 2000 was the transformation from forest to others, it did not mean a loss of forest area, as the major component of others, according to our data, was logging or fire slashing replaced with fast-growing species that would become forest again.

A dominant modification of grassland was transformation from sparser one to denser one in the period 1987–1995. The situation reversed in the next period, with at least 75.8% of the modification attributed to the transformation from denser grassland to sparser grassland. However, after 2000, a recovery of the grassland ecosystem seemed to occur as the transformations of grassland from sparer back to denser became dominant: 56.2% and 79.4% of grassland modification in the periods 2000–2005 and 2005–2010, respectively.

5. Discussion

Land-use change is a nonlinear and complex process, and is coupled with other societal and biophysical system changes [41]. Population growth directly affects the land-use mosaic and leads to changes. However, there are additional underlying forces that induce land-use change, as anthropogenic activities largely respond to economic opportunities, which are mediated by institutional factors [42,43]. Moreover, with rising awareness of the importance of environment protection, policies aimed at reducing negative feedback arising from depletion of key resources or a decline in the provision of important ecosystem goods and services were also found to be crucial forces in modern land-use change [41,44]. In China, the trade-offs among demands for sustainable food production, economic growth, and environmental protection make up the driving forces behind the rapid alterations in contemporary land use.

First, trade-offs between demands on food security and economic development influenced the changing pattern of cropland. Cropland is considered to be a crucial land use, as it has enormous consequences for food security [45,46], especially in China, the world’s most populous country. Cropland expanded to meet the food demand of the increasing population before 2000 (Figure 1a). However, the rapid rate of economic growth, accompanied by development of large areas of industrial and residential land, led to a large loss of cropland [47,48,49], as found in this research (Section 4.2 and Section 4.3). Although some policies, such as Regulations on the Protection of Basic Farmland, sought to protect cropland from being consumed by built-up land, the effects seem to have been limited. Both the amount and the percentage of cropland loss due to built-up land expansion increased during the study period. To ensure domestic food security, the central government of China launched a policy named the Dynamic Balance (no net loss) of Cropland at the end of the 1990s. Northeast and northwest China, with vast cropland potential, therefore, experienced significant cropland expansion (Figure 4a).

Second, expectations of food security, economic development, and environment protection substantially influenced the woodland dynamics. Demands for food supply and logging resulted in the decline of woodland areas (Table 4). However, the increasing environmental problem, along with the shrinkage of woodland, evoked the government’s awareness of environment protection. A series of policies, including Grain for Green and the Three-North Forest Program, were launched to mitigate the pace of woodland loss. The pace of the decline in woodland area slowed and even reversed to an area increase between 2000 and 2005, as shown in Figure 1. However, pursuing economic profits has led to the development of fast-growing forest areas. This was the largest part of woodland modification in the 21st century (Table 5), and has caused serious ecological problems, even without any decrease in forest area.

Third, the trade-off between food security and environment protection shaped the general pattern of grassland change. Grassland has played an important role in food security, as it was generally considered to be a backup for cropland. During the study period, most new cropland was a conversion from grassland (Table 4). The arid and semiarid region of northwest China, containing vast areas of grassland and previously unused land, was designated as a place with enormous reserves of land for farming [50], following a national survey to identify more cultivable land [51]. However, the transition from grassland to cropland in this region also brought about severe ecological problems, such as soil erosion and sandstorms. The Grain for Green policy aimed to mitigate the environmental degradation and led to a phenomenon of increasing amount of cropland converted back into grassland in the period of 2000–2005, directly after the implementation of the policy (Table 4), as well as improvement of grassland coverage (Table 5).

Finally, economic development, along with urbanization and industrialization, was the major factor driving built-up land expansion [52], which resulted in a large amount of cropland loss (Table 4). Policies on economic development, such as “Western Expansion” and “Rise of Central China”, also impacted the spatial pattern of built-up land change. Some regions in central and western China became hot spots of built-up land expansion (Figure 4g).

6. Conclusions

China has been experiencing tremendous land-use changes in last decades, especially after the reform and opening up of the 1980s, putting intensive pressures on regional, even global, environmental system. Taking advantage of the CLUD, this study characterized the temporal and spatial pattern of land-use change in China over the period from 1987 to 2010 considering the dimensions of both inter-type and intra-type change. An area of 4 × 10⁵ km², 4.5% of the total landscape in China, experienced land-use change during the period in question. Of the changed area, 19.6% was found to have been altered more than once during the four sub-periods of analysis. Cropland, woodland, grassland, and built-up land were four thematic land-use types involved in significant conversions and modifications.

Trade-offs between demands on food security, economic development, and environment protection largely forced and shaped land-use changes in China. Economic development, along with urbanization and industrialization, led to the tremendous expansion of built-up land, resulting in an increase of 30.6% from its 1987 area, the greatest proportional change of the six land-use types. Built-up land expansion resulted in a massive loss of cropland with a basically increasing rate, in eastern coastal region. However, a large amount of cropland gain, on the demand of domestic food security, was evidenced in north of China with the price of area loss in grassland and woodland. The conversion of cropland affected an area of 159,930 km², or 11.7% of the original 1987 area, and ended with a net area increase of 14,280 km² during the period in question.

The area of woodland declined slightly by an area of 7880 km², or 0.4% of its area in 1987, during the past over twenty years. However, following the implementation of several policies on environmental protection, a consistent and slight increase in the sub-periods after 2000 was evidenced. At the same period, modification of woodland highlighted the intensive forest management in terms of planting fast-growing trees. This was induced by economic profits pursuing. Grassland, the major source of cropland gain, continued to shrink for more than the past twenty years, although, at a decreasing rate. The area of grassland decreased by 50,963 km² during the study period. Policies on environment protection, such as Grain for Green, saw its effects as a larger amount of cropland was converted back to grassland in the period of 2000–2005, directly after the policy’s launch. Modification of grassland, dominated by the transformations from lower density grassland to higher one after 2000, gave more evidence of the positive effects of environment policy.

Temporally and spatially explicit analysis of land-use-change patterns provides a comprehensive understanding of land-use patterns in China. Dominant features and hot spots of land-use change can be explored on the basis of this analysis. In the future, specific research on changes in certain land-use types should be conducted to address impacts on the environment quantitatively and to help formulate future policies to achieve sustainable food, environmental, and economic development.