Introduction

The research on regional landscape change has been one of the hottest fields, which draws general interest of the geographers and ecologists all over the world (Forman and Godron 1986; O’Neill and Gardner 1988). Landscape change and eco-process are interrelated and interact on each other (Forman 1995; Xiao and Bu 1997). Studying on eco-process takes quantitative analysis of landscape change as preexisting. The methods of quantitative analysis of landscape pattern structure and its change based on landscape indexes have been the core of landscape ecology researches (Roy 1999; George and Perry 2002).

With the development of geographical information system (GIS) and remote sensing (RS) technique, quantitative analysis of landscape change has been getting in greater depth (Forman and Godron 1986; Zonneveld and Forman 1990; Turner and Gardner 1991; Forman 1995; Naveh and Liberman 1994; Xiao 1996; Frohn 1998). There have been some academic reports (Li and LI 2004; Zhan and Deng 2006) on landscape change analysis of the Chinese Loess Plateau, but very few reports (Liu and Dong 2007) on landscape change analysis at river basin scale based on GIS and high-resolution RS images can be found. This paper has a double purpose: first, to present the methods to quantify landscape change at river basin scale using GIS and RS images; and second, to apply the methods to the Weigou River basin on the Chinese Loess Plateau to analyze the landscape change during the period from1998 to 2004.

Materials and methods

Study area

The Weigou River basin, located in Jingning county, Gansu province, China, between 105°29′–105°48′E and 35°25′–35°38′N (Fig. 1), has a population of 43,690 inhabitants and covers an area of 29,444 ha. It is one of the sub river basins of the Huluhe River, which is a tributary of the Weihe River water system. Lying at the west piedmont of Liupanshan Mountain in the third hill region of the loess plateau, it has a temperate semi-arid climate. The average annual precipitation is 450 mm, with a maximum of 690 mm and a minimum of 228 mm. The precipitation from June to September accounts for 69% of the total annual precipitation. The average annual temperature is 7.37°C. More than 10°C accumulated temperature is 2551.2°C. There are 160 frost free days a year. The altitude is between 1,700 and 2,100 m above the sea level. The density of ravines is 3–5 km/km2. The main soil types are loessial soil (about 78%), heilu soil (about 18.9%), red clay, and entisol. The natural vegetation is sparse due to the dense ravines and fragmentized landform. Gully erosion, gravity and engineering-induced erosion are prevalent, with an erosion modulus of 9348.3 t km−2 a−1 in 2004. And these have long been the main ecological problem of the region (Committee of Jingning’s Chorography 2005).

Fig. 1
figure 1

Study area

Full size image

Data

In this paper, 1:50,000 scale topographic maps, SPOT HRV (high resolution visible) images and Landsat TM (thematic mapper) composed of band 4, 3, and 2 images from 1998 to 2004 have been adopted to establish the 10-m resolution SPOT + TM images. All the Weigou River basin has been completely covered (Fig. 1). These space information and spectral information of SPOT images and TM images can entirely meet the demand of this study (Wang and Feng 1991; Wang and Qi 2003). Other data used in the study include the climatic information recorded at Jingning weather station from 1960 to 2003 (see Table 1; Fig. 1).

Table 1 Data
Full size table

Methods

The three key links of landscape changes analysis using RS images and GIS techniques are satellite remote sensing monitoring, data processing, indoor man–machine interactive interpretation, integration issues and comprehensive analysis (see Fig. 2). Firstly, a collection of related data or program records, including RS data, land form data, climate data, etc., should be executed. Secondly, data processing is requested, including data standardization, data classification, accuracy estimation, checking and so on. Thirdly, the changes of landscape are quantified and a comprehensive analysis is conducted.

Fig. 2
figure 2

Workflow figure

Full size image

In this study, the land use/cover in the Weigou River basin has been investigated. The RS images have been studied completely, so as to get an overall understanding of the spectrum information (see Table 2), radiation information and geometric features, to establish a unified classification system in the light of local condition, and to establish the interpretation key (Carper et al. 1990; Chavez et al. 1991; Bocco et al. 2001).

Table 2 Data spectrum information
Full size table

Data processing

Data processing is very important for obtaining accurate data (Baret et al. 1991; Brivio et al. 2001; Bricaud et al. 2002; Qin et al. 2006; Masoud 2006). ERDAS IMAGINE has been used for the process of image radiometric enhancement, geometric rectification and merging images (Dang and Wang 2003).

To identify ground control points, all images have been enhanced by linear contrast stretching and histogram equalization methods, so that the original brightness values of pixels could be kept unchanged (Wang and Feng 1991; Wang and Qi 2003; Li and Li 2004). To keep the precision within one or two pixels, the accurate corresponding points between images system and topographic maps (65 for each SPOT image and 42 for TM) have been carefully identified and geometric rectification has been conducted. The quadratic polynomial model and nearest neighbor resampling model have been used in the geometric rectification of these images. The former, which perform coordinate transformation to the images directly, is widely used for geometric rectification of various images of satellite sensors. However, the complicated high order exponent may lead to the unexpected changes in length, hence, the quadratic polynomial model was selected. The latter, keeping the DN (digital number) of pixels unchanged without smoothing, was considered effective in the landscape classification (Dang and Wang 2003). Each image has been rectified by Albers Equal Area Conic projection based on 1:50,000 scale topographic maps.

The principal component transformation model has been adopted when merging the SPOT images and TM images (Pohl 1998; Li and James 2002; Dang and Wang 2003). Firstly, a spatial registration should be carried into execution to input multispectral image (TM image, with 30-m resolution), and then a principal component transformation. Secondly, the histogram of the panchromatic image (SPOT HRV images, with 10-m resolution) should be matched to the main principal component (PC 1) of the input multispectral image. Thirdly, the stretched panchromatic image takes the place of PC 1 and takes a reverse principal component transformation in the processing of images merged (see Fig. 3).

Fig. 3
figure 3

RS data merging based on principal component transformation model

Full size image

The resultant root-mean-squared errors (RMS) are 0.46 pixels (4.6 m) for the 1998 SPOT + TM images, and 0.51 pixels (5.1 m) for the 2004 images.

Data extraction

Satellite remote sensing and GIS techniques are efficient tools to monitor and quantify the landscape changes. By the use GIS software, dynamic data reflecting the landscape changes in the Weigou River basin, such as land use, land use changes, change area and change direction, etc., could be extracted from the satellite remote sensing images.

For extracting, processing and displaying vector- and grid-data, both the Arctools model and the Grid model of ARC/INFO have been used (ESRI 1996, 1997).

By the first-hand investigation of land use/cover in the Weigou River basin and in-door analysis, all the images have been classified into six primary landscape types (farmland, forest, grassland, water, construction area and desert). In more detail each image was classified into 12 sub-types (plain farmland, sloping farmland, terraced farmland, arbor forest, shrubbery forest, economic forest, natural grassland, artificial grassland, water, town construction area, rural residential area, and desert).

The vector and attribute data have been obtained under man–machine interactive interpretation method, with which the information of remote sensing images is extracted by veterans of landscape changes analysis and remote sensing application under GIS software. By this method, it has complementary advantages for data extraction, because the veterans and the software are so well coordinated. So it plays a unexchangeable role in remote sensing information extraction (Pu 1992). The precision of the vector has been kept within one or two pixels.

In July 2004, the fieldwork was carried out in order to check data accuracy of various landscape types. The GVG system (car global position system (GPS) combined with digital video system and GIS system) was adopted to obtain 1947 ground truth pictures along the intersected linear routes. Then superposition analysis was conducted on the interpreted results and thereby the error matrix was established. The results show that the total error rate is 2.47%, and therefore the total interpretation accuracy approaches 97.53% (see Table 3).

Table 3 Error matrix (m2)
Full size table

Landscape transition matrix and landscape indexes

Area change, change ratio and change direction of various landscape types can be shown up by the landscape transition matrix (Li and Lu 2001; Li and Li 2004). In this paper, the landscape transition matrix has been quantified by Grid data of all the landscape types in the Weigou River basin established by the Grid model of ARC/INFO.

To study the interrelation between landscape pattern and eco-process, it is necessary to monitor and analyze the landscape change at the accurately same scale of the landscape characteristics (Wu 2000). Landscape index has demonstrated its efficacy in researches on scale effects and scale characteristics of landscape pattern (Turner and O’Neill 1989; Jelinski and Wu 1996; O’Neill and Hunsaker 1996; Wu 2000; Turner and Gardner 2001). The structural characteristics of landscape can be pointed out by comparing landscape indexes at class level and at landscape level (Lu and Li 2001; Xiao and Li 2005). Nine landscape indexes at class level and eleven indexes at landscape level (Table 4) have been quantified by the use of the spatial pattern analysis program for categorical maps (FRAGSTATS) for monitoring the landscape change at river basin scale. The expressions of the indexes provided by FRAGSTATS have been adopted (McGarigalv and Marks 1994; McGarigalv et al. 2002). The landscape vector data have been converted to 10-m resolution grid under Grid model of ARC/INFO. The edge of the river basin and landscape has been treated as the edge of patches. The minimum search radius has been assigned to 2,000 m, respectively.

Table 4 FRAGSTATS metrics
Full size table

Results

Analysis of landscape transition matrix

As seen in Table 5 and Fig. 4, farmland dominates the study area. There has been a considerable change (1,390.30 ha, 4.72% of the total area) in the Weigou River basin during the 6-year-period. Forest, grassland, construction area, and water have increased (by 63.93, 0.65, 4.04 and 2.75%, respectively).

Table 5 Landscape change in six primary types, 1998–2004 (ha)
Full size table
Fig. 4
figure 4

Area change of primary landscape types

Full size image

A further analysis shows that all the sub-types of landscape, the area of terraced farmland have decreased most significantly (by 715.11 ha) and economic forest has had the largest increment in area (by 511.19 ha) (Fig. 5). Plain farmland, slope farmland and terraced farmland have decreased by 206.53, 117.72, and 715.11 ha, respectively. Arbor forest, shrubbery forest and economic forest have increased by 163.04, 271.23 and 511.19 ha, respectively. Natural grassland has decreased by 256.92 ha, while the artificial one has increased by 288.84 ha.

Fig. 5
figure 5

Area change of 12 sub-types

Full size image

Table 6 presents the transition matrix of landscape change. The results indicate that, the two most prominent conversions are the conversion from terraced farmland to artificial grassland and the one from terraced farmland to economic forest (284.84 and 273.64 ha, respectively). Conversions from plain farmland to economic forest, from natural grassland to shrubbery forest, and from terraced farmland to shrubbery forest are also notable (137.79, 137.61, and 106.23 ha, respectively). Arbor forest has been transformed mainly from terraced farmland (40.95%) and natural grassland (39.34%). Of the area converted to town construction area, 88.93% were originally plain farmland. Of the area converted to rural residential, 65.89% were originally plain farmland, 34.11% were originally terraced farmland.

Table 6 Landscape change matrix, 1998–2004 (ha)
Full size table

As shown in Tables 5, and 6, from 1998 to 2004, the primary type of grassland changed a little. However, to illuminate the remarkable changes among farmland, grassland and forest, the transformations related to sub-types of grassland cannot be ignored.

Analysis of landscape indexes at class level

The changes in landscape indexes at class level are shown in Table 7. The interspersion and juxtaposition index (IJI) of all landscape types increased from 1998 to 2004, indicating more uniform landscape configuration and diversity landscape at class level in the Weigou River basin.

Table 7 Landscape matrix change at class level, 1998–2004
Full size table

First of all, among landscape types, terraced farmland is the highest variable patch type: its percentage of landscape index (PLAND) decreased from 57.06% in 1998 to 54.63% in 2004, whereas its number of patches index (NP) increased from 90 to 111 and the mean patch size (MPS) declined. Both its area-weighted mean patch fractal dimension (AWMPFD) and area-weighted mean shape index (AWMSI) decreased, indicating that the 2004 terraced farmland patches were more isolated and less uniform than the 1998 patches. That is also supported by the mean proximity index (MPI), which decreased from 6242.61 to 4093.02 and mean nearest-neighbors distance (MNN), which decreased from 54.48 to 51.03. Therefore, it shows that the landscape pattern, which was dominated by large terraced farmland patch, has changed to a certain degree. The area of slope farmland and NP index decreased, while the MPS index increased. The AWMPFD index decreased while the MNN index increased. This indicates that the slope farmland was regulated to a certain degree.

Secondly, PLAND index and NP index of forest increased. Economic forest changed most significantly among the sub-types of forest, from 2.78 to 4.48% and 761 to 834, respectively. All of its MPS index, AWMSI index and AWMPFD index increased, while MNN index decreased, indicating that distribution of forest changed from fragment to clumped, and the inner-connection were further strengthened.

Besides farmland and forest, grassland is another flexible patch type. PLAND index of artificial grassland increased, while natural grassland decreased. Therefore the total area of grassland changed a little. All MPS index, AWMSI index, AWMPFD index, and MNN index of the two sub-types of grassland decreased, indicating that artificial grassland protected the grassland ecosystem in the situation of natural grassland degradation.

At last, the changes in landscape indexes of construction area indicate that: the construction area has been expanded, more clumped and more regulated, and the inner-connection has been strengthened. Additionally, indexes of water and desert changes point out that water area increased and clumped. Desert decreased little, but has been regulated and changed to uniform.

Analysis of landscape indexes at landscape level

From 1998 to 2004, in the Weigou River basin, the NP index increased and MPS index decreased, showing a trend towards increasing fragmentation and fine-grained landscape (Table 8). Both Shannon’s Diversity Index (SHDI) and Shannon’s Evenness Index (SHEI) became larger. Thus, the landscape heterogeneity and evenness increased. Furthermore, landscape dominated by large patches of farmland became subdued. The spatial context of landscape had significant changes. For instance, the MNN index and the MPI index declined, displaying that the spatial distribution of patched became more continuous. In addition, the IJI index increased and the contagion index (CONTAG) became smaller, illustrating that the spatial distribution of main patches (terraced farmland, natural grassland, plain farmland and slope farmland) in the landscape became less clumped, and the dominance of one or several patch types decreased and had an even greater connectivity. That also supported that the landscape became more uniform and more continuous.

Table 8 Landscape matrix change at landscape level, 1998–2004
Full size table

Drive factors of landscape pattern

Climate factor

The Weigou River basin has a temperate semi-arid climate. The climate data recorded at Jingning weather stations from 1960 on (Figs. 1, 6; Table 9) shows that the average annual precipitation is 450 mm with a disparate time distribution (from June to September, accounting for 69% of annual total precipitation) and mainly in the form of rainstorm, which is the effect of general circulation of atmosphere (west wind). The average annual temperature is 7.37°C and more than 10°C accumulated temperature is 2,551.2°C. The climate condition caused an asynchronous incidence of solar radiation, solar heat and water condition. The quality and condition of vegetation is not so good. Soil erosion is severe and ecological environment is broken up. These are the climatic reasons for the current situation of landscape pattern and structure.

Fig. 6
figure 6

Difference between evaporation and precipitation

Full size image
Table 9 Climate condition in Weigou river basin
Full size table

Terrain factor

The Weigou River basin is an offshoot of the Huajialing Mountain range, located in the third hill region of the loess plateau and at the west piedmont of the Liupanshan Mountain. As shown in Table 10, only 9.34% of the area has slopes below 5°; 32.14%, 5°–15°; 39.72%, 15°–25°; 18.8%, above 25°. Due to dense ravines and fragmentized landform, gully erosion, gravity and engineering—induced erosion are prevalent, with a mean erosion modulus of 9,360 × 104 t km−2 and a ravines density of 3–5 km/km2. These are the topographic reasons for the current situation of landscape pattern and structure.

Table 10 Soil erosion intensity in Weigou river basin
Full size table

Soil condition factor

As shown in Table 11, the main soil types in the Weigou River basin are loessial soil (accounting for about 78%), heilu soil (18.9%), red clay (2.1%), and entisol (1%). Loessial soil, with visible vertical joint, lower anti-erosion capacity, shallow ripe layer, strong collapsibility, and the lower fertility characteristics, often occurs on slopes of residual loess plain, mid and top of weir and mound slopes, and in gully areas. The soil distribution pattern causes the severe soil and water losses.

Table 11 Soil condition in Weigou river basin
Full size table

Human activities factor

The Weigou River basin is located in the agro-pasture zigzag zone. The population density is 148 inhabitants per km2. The intensified land reclamation for the development of agriculture and animal husbandry, severe deforestation and overgrazing has led to the desertification of the impoverished land. These have caused the formation of the landscape pattern dominated by large farmland patches. Since 1980s, people have been aware of the fact that ecological environmental protection is important to social and economic development in the western regions. In recent years, with the implementation of ecological rehabilitation projects, including returning farmland to forest or grass, afforestation and building terrace fields in the Weigou river basin, the eco-environmental conditions have exhibited some favorable changes. For instance, the cooperation proportion values of farmland, grassland and forested rise from 27.28:3.77:1 in 1981 (Committee of Jingning’s Chorography 2005) to 14.33:2.91:1 in 1998 and to 8.32:1.79:1 in 2004 (see Table 12), calculated results by revised universal soil loss equation (RUSLE) shows that the mean erosion modulus in the Weigou River basin during the 6-year-period has decreased from 9559.4 t km−2 a−1 to 9348.3 t km−2 a−1 and the erosion area has decreased by 389.8 ha. The soil erosion severity has changed from weak and moderate degrees to slight degree.

Table 12 Soil erosion change
Full size table

Drive factors of landscape change

As seen in Fig. 7, the climate data recorded at Jingning Weather Station shows that, since 1960, the precipitation have been descending on the whole, while during the period from 1998 to 2004, the precipitation has been ascending on a partial scale. Because of the relatively favorable water condition, survival rate of vegetation moves upward. The implemented ecological rehabilitation projects expound the last paragraph. The land of total area of 1039.36 ha has retired from crop cultivation, with trees planted, tree-shielding systems established and small watersheds treated, covering an area of 697.42 ha. Artificially sown grass field and improved meadows have reached 296.11 ha (see Tables 5, 6). Therefore, the soil and water losses have been checked to a certain extent. The landscape structure has been ameliorated. Landscape heterogeneity and evenness have increased.

Fig. 7
figure 7

Precipitation in Weigou river basin

Full size image

Conclusions and discussion

This paper describes how the GIS analysis technology and satellite remote sensing images are combined to address the landscape change in the Weigou River basin, during the period of 1998–2004. The forest has been found increased notably, while farmland has decreased significantly due to the conversion to forest and grassland. The landscape heterogeneity and evenness have increased, with a greater connectivity. The dominance of farmland patch has decreased. The project of returning land from farming to forest and grassland and the development of supporting policies could be expected to benefit from the use of landscape information.

Located at a special geographic zone, the Weigou River basin has a rather fragile ecosystem and semi-arid climate on Chinese Loess Plateau. Ecological landscape changes in this region are mainly affected by natural and human factors. Through complex interaction, the natural factors and human activities alter the landscape structure and decide the service function of regional ecosystem (Wang and Feng 1991; Zhang and Wang 2006).

The natural factors include climate, landform, soil and water condition, etc., the climate has become warmer and drier mainly by natural driving force. As a result, the environmental aridification and vegetation degradation have become exacerbated, and the already broken terrain has become even more fragmentized due to the water erosion-induced dissection. The fluctuated arid-climate boundary has moved southward, and the land desertification (such as water erosion, wind erosion, gravitational erosion, soil salination and fertility reduction) become aggravated. Because of the relatively favorable water condition (expounded in “Drive factors of landscape change”), the survival rate of vegetation moves upward. The landscape structure has been ameliorated. And the landscape heterogeneity and evenness have increased.

The human activities, i.e. the rapid population growth and predatory economic activities, are also the important factors leading to the eco-environmental deterioration of the region. In the past, the erosion of loose loessial soil in the region was accelerated and the service function of regional ecosystem declined, which even caused serious conflicts. Such environmental degradations were caused by following factors: the lack of environmental awareness, the seeking of instant economic benefits at the expense of the environment, uncontrolled removal of vegetation, over-cutting and over-reclamation. Since 1998, with the implementation of ecological restoration projects, some favorable changes to the eco-environment have occurred. For example, some farmlands have been transformed back into forestland and grassland, some grassland has been transformed into forest land, and some sloping croplands have been transformed into terrace fields. As a result, the surface vegetation has been restored to a certain degree, the erosion intensity of heavy rain has been reduced, the soil loss has been decreased, and hence the ecological environment has started to develop in a reversing direction.

However, the restoration of ecosystem structure and service function and the elimination of poverty in the region of a fragile ecosystem are a long-term process. Some schemes to solve the conflicts may slow down the region’s economic development in a short period of time or even cause rebound of poverty, but in the long run, solving the conflicts concerning the service function of regional ecosystem and maintaining its service function are the foundation for the region’s sustainable development, modern civilization and human existence. Therefore, it is important to weigh the pros and cons and solve the conflicts, overcome the contradiction between resource use and the harmonious development of society and environment, and enhance the service function of regional ecosystem.

In Weigou River basin, farmland, forest, grassland, water, construction area and desert are not only major self-contained parts of the ecosystem, but also connected with each other. The natural and human activities made a weak service function of regional ecosystem. In the environmental and ecological systems engineering points of view, reestablishing vegetation and building terrace fields are just the beginning of the improvement of the local eco-environmental conditions. The vegetation is just restored, water erosion and gravitational erosion are just brought under control to a certain degree, and hence the ecological environment started to develop in a reversing direction.

The image interpretation coupled with GIS has demonstrated its ability to provide comprehensive information on the direction, rate, and location of landscape changes as a result of land management. Landscape metrics have proved useful in monitoring the dramatic landscape changes that have taken place in the Weigou River basin. Meanwhile, the transition matrix and the transition map of landscape patch types are helpful tools to the analysis of the changes in landscape structure and composition.