Introduction

Wetlands are closely related to human survival and development and are subject to human activities and climatic changes (Lee et al. 2015). As critical parts of the global ecological system and carbon pools, wetlands provide crucial ecological functions and effects that cannot be replaced: they offer a large number of ecosystem services, including freshwater storage, flood control, water purification, habitats for wildlife, regulation of climate and preservation of aquatic life (Wang et al. 2012; Sulman et al. 2013; Kim and Suh 2016). For example, China’s natural wetlands are home to 3200 animal species and 700 fish species, and provide habitats for 60% of crane species and 26% of goose and duck species in the world (State Forestry Administration 2005). However, over the past century, ~50% of global wetlands have been lost due to anthropogenic factors (Khaznadar et al. 2009; Wang et al. 2011). Similarly, China has lost 23.0% of its freshwater marshes, 16.1% of its lakes, 15.3% of its rivers, and 51.2% of its coastal wetlands due to agricultural development (Gong et al. 2010).

Wetland landscapes contain wetland units of spatial heterogeneity or metapopulations from the micro- to macroscale (Walker et al. 2012). The wetland landscape is a result of the combined action of various ecological processes, which have an important influence on the structure and function of the landscape (Weber et al. 2001). Changes in the wetland landscape cause significant regional ecological effects (Mitsch and Hernandez 2013). These are a focus not only of scientific research on wetland landscape dynamics but also in the field of ecology (Kang and Kim 2015). Analyses of landscape dynamics are useful for exploring the relationship between landscape changes and ecological processes (Sueltenfuss et al. 2013). Quantification of the spatial distribution of landscape components is the basis for further study of the function and dynamic changes of the landscape (Fu et al. 2005; Maltchik et al. 2014). Accordingly, research on the processes underlying the dynamics and driving forces of wetland landscapes should improve our understanding of typical wetland evolution and help advance techniques to protect wetlands. The underlying processes and driving forces of wetland landscapes are core research topics in current wetland science (Liu et al. 2009). Wetlands distributed along the Tumen River, which is one of the main international rivers in China, are a major focus of eco-environmental protection in Northeast Asia (Zhu et al. 2014). However, in recent years, the area has also become the focus of multinational cooperation with the implementation of a general opening-up strategy. The environment has been interfered with and damaged to different degrees. Such degeneration has induced the fragmentation of the wetland landscape and decreased biodiversity, thus endangering the survival and sustainable development of the Tumen River Basin (Zhu et al. 2014).

Previous studies of the Tumen River Basin have focused on various aspects of the environment, including wetland ecosystem health assessment, wetland-change detection and the effects of land use changes on ecosystem services (Nan et al. 2013; Zhu et al. 2014). However, many features of these wetlands remain unknown (Zhao et al. 2010). In particular, research on the long-term temporal and spatial variations in wetland landscape dynamics remains insufficient (Zhao et al. 2010). Therefore, we used multi-source remote sensing images [Landsat Multispectral Scanner System (MSS), Thematic Mapper (TM), and Enhanced Thematic Mapper (ETM)] and high-resolution data (Corona, Alos) to explore the wetland landscape changes and driving forces of the Tumen River Basin. The major objectives of this research are: (1) to describe the processes of wetland changes in the Tumen River Basin between 1960 and 2009 at the landscape and pixel levels; and (2) to analyze the forces that drive wetland landscape changes, and distinguish between the effects of human activities and natural factors in the region.

Materials and methods

Study area

The Tumen River Basin is located in the northeastern part of Jilin Province, China. The Tumen is an international river located within the national boundaries of China, North Korea, and Russia. The Tumen River Basin has various types of wetland ecosystems and ranges from 41°59′47″ to 44°30′42″N and from 127°27′43″ to 131°18′33″E (Zhu et al. 2014) (Fig. 1), and borders China, North Korea and Russia, near the sea of Japan in the east. It is characterized by a typical temperate-zone monsoon climate with a mean annual precipitation of 400–650 mm. The average temperature is 2–6 °C, and the maximum and minimum temperatures are 34–38 and −34 to −23 °C, respectively (Zhu et al. 2014). The mean annual rainfall is 1000−1500 mm in the upstream region of the Tumen River Basin, which is the highest in the entire basin. Furthermore, this area is unique because of the types of wetland, which are a transient habitat for endangered migratory birds, such as red-crowned cranes, and various rare animals and plants. It occupies a central position in the Northeast Asia ecological network. In this latter area, the typical wetlands we selected upstream of the Tumen River were located in the Yuan pool of An’tu County and the town of Chongshan, Helong City, east of Changbai Mountain. The first tributary, the Hongqi River, flows through the area. The typical wetlands in the midstream Tumen River Basin were located in Yanji City, Longjing City and Tumen City. The Ga’ya, Bu’er hatong, Hailan, Yanji and Chaoyang rivers flow through the area and then into the Tumen River. The typical wetlands downstream of the Tumen River are located in the town of Jinxin, Hunchun City.

Fig. 1
figure 1

Scope of the study

Full size image

Data

Remote-sensing data used in this study include one paid Corona image, ten cloud-free Landsat images and two paid Alos images to cover the entire area. Landsat images include Landsat MSS images of the 1970s (track nos. 124/030, 125/031), Landsat TM images of the 1980s and 1990s (track nos. 115/030, 116/031), and Landsat ETM images (track nos. 116/031) of 2009. The Alos images include four multispectral bands (with a spatial resolution of 10 m) and one panchromatic band (with a spatial resolution of 2.5 m) of 2009 (Table 1). Ten cloud-free Landsat images were acquired for free from the United States Geological Survey (USGS), and the Corona and Alos images were obtained from early American spy satellites and the Advanced Earth Observing Satellite, respectively. All images were recorded from June to October, during which plant growth in the Jilin Province of China is highest (Dong et al. 2014). All data were used in conjunction with field investigation sites. The images improved the identification of wetlands. Because these data products alone cannot accurately reflect the distribution of wetland types, this study also used 1:50,000 topographic maps produced by the National Geomatics Center of China, soil maps, drainage system diagrams, digital elevation model (DEM) figures, digital photographs and global positioning system data obtained by field investigation of a quality that is sufficient for the study’s purpose (Li et al. 2007) to verify the accuracy of the wetland-type classification (Zhu et al. 2014).

Table 1 Data sources of landscape information for the study area
Full size table

Data processing

ENVI 4.7 image processing software and ArcGIS 9.3 software were used for data processing. Different spatial data were integrated into the Beijing 1954 Coordinate System. The corresponding 1:50,000 topographic maps were selected as the source of basic research data for the study area. The main steps include image geometric correction, image fusion, and color composite using the band combination of Landsat (MSS) 3, 2, 4 (red, green, blue; RGB), Landsat (TM) of 5, 4, 3 (RGB) and Alos 3, 2, 1 (RGB). For remote sensing images with color synthesis, the image will be closest to the actual object color after color synthesis, and is thus better able to identify the wetland landscape patches, vegetation characteristics, and characteristics of water bodies in the Tumen River Basin (Dong et al. 2014).

Wetland area mapping

The principle for wetland mapping in this study was based on the fact that there is a period of flooding when the wetlands are inundated by flooding streams and runoff (Dong et al. 2014). The wetland indices such as hydrology, soil, vegetation, topography etc. were used. From late April to early June, local wetland fields are mainly flooded by shallow water in preparation for the successive planting of seedlings and their transplantation (Dong et al. 2014). We used the high-definition Corona image, Alos image and the cloud-free TM image from late June to early August to mark the wetland.

Definition of wetlands

The USGS defines “wetland” as a general term applied to land areas that are seasonally or permanently waterlogged, including lakes, rivers, estuaries, and freshwater marshes, and as an area of low-lying land submerged or inundated periodically by fresh or saline water (Walker et al. 2012). In agreement with the Ramsar Classification System for Wetland Types (Davis and Brock 2008), the China Wetland Classification System (Lü and Jiang 2004) and literature research (Dong et al. 2014; Wang et al. 2011 and 2012), and taking into consideration the USGS and Ramsar definitions and wetland site conditions in the study area, “wetland” in this study is divided into “natural wetlands” and “artificial wetlands.” The natural wetlands include marshes, lakes and rivers, and the artificial wetlands include paddy fields and other artificial wetlands (Zhu and Koo 2006; Zhu et al. 2007). Finally, the wetlands in this study are classified as “marshes,” “lakes,” “rivers,” “paddy fields,” and “other artificial wetlands,” such as reservoirs, fish ponds and hatching pools.

Extraction of wetland information

We selected all data relating to wetlands in the region for the period between 1960 and 2009 from remote sensing images (Table 1) (Jiang et al. 2014). We combined the extracted data with land use/cover data, DEM data, the map of the river system and soil types for the study area from the Chinese Academy of Science (http://westdc.westgis.ac.cn/) and a 1:50,000 topographic map (obtained by a field study in 1960) for supervised classification (Jiang et al. 2014). Using reference classification results obtained via visual interpretation, the wetland resources maps for the 1960s, 1970s, 1980s, 1990s and 2000s were obtained. Using ENVI 4.7, a stratified random sampling method for assessing accuracy, 50 points were selected evenly in ENVI 4.7 from the study region, with field notes used for verification. In particular, high resolution Alos images can be directly assessed by visual interpretation of surface features; thus, the manual interpretation of corrected images in ArcGIS 9.3 is needed to determine the characteristics of images of the wetlands. Wetland information was obtained from topographic maps and field survey data, and more accurate wetland information was obtained after several iterations (Fig. 2: 2000s).

Fig. 2a–e
figure 2

Typical wetland types of the Tumen River Basin in 1960–2009

Full size image

Dynamic analysis method

The magnitude and rate of wetland area changes showed significant temporal and spatial variations due to the influence of both natural and human factors. We can quantify the rate of wetland landscape changes with a dynamic model, which can reveal not only the temporal changes of a single landscape but also the entire study area and the spatial heterogeneity of different types of landscape patterns. Therefore, the wetland dynamics were further explored in this study (Fukushima et al. 2007).

Dynamic trends

To further explore the range of variations in wetland areas within a certain time period in the study area, the dynamic concept of wetlands is introduced as follows:

$$ {\text{LC}} = ({\text{Ub}} - {\text{Ua}}) \times {\text{Ua}}^{ - 1} \times T^{ - 1} \times 100 \,\% . $$
(1)

Ua and Ub denote the wetland areas at the initial and late stages, respectively, and LC denotes the wetland dynamics for T years. The dynamic trends can describe the degree of change in a wetland and highlight the temporal characteristics of these changes, which are important when studying wetland dynamics (Li et al. 2016).

Transition matrix

A cross-tabulation detection method was used to determine changes in the wetland landscape, and a transition matrix was produced (Li et al. 2007). The average transformation condition of every wetland landscape type can be calculated by an established transition matrix of wetland landscape type from 1960 to 2009. The conversion process of wetland landscape elements, especially the key wetland type, can be intuitively understood, and the change matrix provides information on the main types of changes in the research area (Li et al. 2016).

Landscape index

The landscape indices are quantitative indicators that reflect the composition and spatial configuration characteristics of the landscape (Zhao et al. 2009). To reflect the direction, speed and space differences of wetland landscape evolution and analyze the driving forces caused by this evolution, we selected the patch number (NP), the area percentage of landscape patches (LAND), the mean patch size (MPS), the patch density (PD), Shannon’s diversity index (SHDI), and Shannon’s evenness index (SHEI) for use in the quantitative analysis of the processes involved in dynamic changes in the wetland landscape pattern. Further details of these indices can be found in previous studies (Zhao et al. 2009; Wang et al. 2011). The NP, LAND, MPS and PD were calculated using FRAGSTATS 3.3 and ArcGIS 9.3. SHDI and SHEI were calculated according to Eqs. 2 and 3, respectively:

$$ {\text{SHDI}} = \sum\limits_{k = 1}^{m} {(P_{\text{k}} )\log_{2} (P_{\text{k}} )} , $$
(2)

where P k denotes the ratio of the area of class k to the total landscape area and m is the number of classes present. If the landscape consists of one class, it is homogeneous with a diversity index of 0; if the landscape consists of more than two classes with equal percentages, its landscape diversity is the highest, but the landscape diversity will decrease if the difference between the percent of different classes increases (Bai et al. 2008).

$$ {\text{SHEI}} = \frac{{\sum\nolimits_{k = 1}^{n} {P_{k} \log_{{_{2} }} P_{k} } }}{{\log_{2} n}} , $$
(3)

where SHEI indicates that an uneven distribution of landscape patches had been created in the area. P k is the kth plaque type proportion in the wetland landscape, and n is the total number of landscape types in the study area.

Results

Accuracy assessment

The producer, user, overall accuracy, and κ coefficient were used to evaluate the results (Lu et al. 2003). The accuracy of the classification is presented in Table 2.

Table 2 Assessment of accuracy of classification results for upstream, midstream, and downstream of the Tumen River Basin
Full size table

The overall accuracy of the wetland mapping was 83.72% for upstream, 86.14 for midstream and 85.76% for downstream, respectively. κ coefficients represent how well a classification performs (Dong et al. 2014). The results (upstream, 0.833; midstream, 0.857; downstream, 0.853) showed that the wetland mapping can achieve higher accuracy.

Changes in wetland area

Figure 2 shows the distribution of wetlands in the Tumen River Basin in the 1960s, 1970s, 1980s, 1990s and 2000s. Table 3 lists the details of the wetland areas.

Table 3 Mapped wetland area from 1960 to 2009 upstream, midstream, downstream of the Tumen River Basin
Full size table

Over the past 50 years, the total area of wetlands upstream of the Tumen River decreased by 55.08% (Table 3). The area of marshes and paddy fields decreased, with the largest change in marshes, which decreased by 68.17%. The area of paddy fields decreased by 36.74%, whereas the area of lakes and rivers increased. The lake area was more than 12 times larger by the end of the study period than in the 1960s, and the river area increased by 68.27%. The dynamic trend of the wetland area in different periods shows that the marsh area decreased year by year, particularly from 1985 to 1999, when it decreased fastest with a dynamic degree of 3.87% (Table 4). The lake area decreased early on in the study period but then increased year by year after the 1970s, especially from 1999 to 2009 when it increased with a dynamic rate of up to 82.60% (Table 4). The paddy fields area decreased year by year in a more uniform manner. The river area decreased evenly over the study period but increased from 1977 to 1985. The river area has increased considerably since 1977–1985 and was 68.27% in 2009 compared to 1960 (Table 3).

Table 4 The dynamic trend (LC) of the wetland area in different periods upstream, midstream, downstream of the Tumen River Basin
Full size table

Table 3 shows that the total area of wetlands midstream of the Tumen River decreased by 50%. During the period examined, the largest change occurred in the lakes, which corresponds to the natural lakes disappearing entirely. The smallest change occurred in rivers, with a decrease of 12.05%. The area of natural wetlands decreased, with the marshes decreasing by 97.76% (Table 3) during the study period with a dynamic degree of −2.00% (Table 4). However, the proportion of artificial wetlands increased. Paddy fields decreased by 51.56%, and the area of other types of artificial wetlands increased threefold with a dynamic degree of 6.08% (Table 4).

Downstream, the maximum amplitude was observed in the paddy fields at different times; the area decreased by 55.90% between 1960 and 1976 (Table 3). The dynamic degree was −3.49% (Table 4). The area of paddy fields increased from 1986 to 1996 with a dynamic degree of 9.05% (Table 4). The area of other artificial wetlands was 2.13 times larger in 2009 compared with that in 1986 (Table 3), with a dynamic degree of 5.31% (Table 4). In addition, the proportion of natural wetlands in the lower, marsh and river landscapes decreased, with the largest decrease in marsh wetlands (68.25%, with a dynamic degree of−1.39%). The smallest change was the increase in lake area of 23.5% (Table 3).

Transition characteristics

The transfer matrix of the wetland landscape types midstream of the Tumen River Basin based on the 1960–2009 data is shown in Table 5. There was a transformation in landscape type among the natural wetlands, artificial wetlands and other land use types in the upstream and midstream area of the Tumen River Basin from 1960 to 2009: a large area of natural wetlands changed into other land use types, and the conversion of marshes was the largest change. There was a massive transformation between paddy fields and other land use types.

Table 5 The transfer matrix of wetlands from 1960 to 2009 midstream and downstream of the Tumen River Basin
Full size table

Landscape type conversion occurred between natural wetlands and artificial wetlands, wetlands and non-wetlands downstream of the Tumen River from 1960 to 2009. A large area of natural wetlands turned into artificial wetland and non-wetland; marshes had the maximum conversion rate. From 1960 to 2009, a total of 433.69 hm 2 of marshes and 320.87 hm 2 of lakes turned into artificial wetlands, and 326.47 hm 2 of marshes and 821.46 hm 2 of rivers tuned into non-wetlands. Approximately 29.86% of marshes and 58.93% of lakes were transformed into artificial wetlands, such as fish ponds and hatching pools, by human activities. Approximately 22.48% of marshes and 52.48% of lakes and other natural wetlands disappeared due to lack of water, resulting in their transformation into non-wetlands (Table 5).

Landscape dynamics

Table 6 shows that the number of wetland patches in the Tumen River upstream area decreased from 382 in 1960 to 105 in 2009. The MPS increased from 67.5676 to 109.8901 hm 2. The largest changes were in the marshes and the patch numbers of rivers. The number of marsh patches decreased by 75.98% in different types of wetlands. The average patch area of marshes increased by 32.35% (23.97 hm 2).

Table 6 Landscape indices of wetlands from 1960 to 2009 upstream, midstream, and downstream of the Tumen River Basin
Full size table

This analysis demonstrates that the marshes (89.40%) and rivers (9.13%) accounted for a larger proportion of the total wetlands in 1960. The percentage of marshes decreased by 63.35% from 1977 to 2009 even though it accounted for a large percentage of the total wetlands. The proportion of rivers increased by 34.19%. The wetland PD index decreased from 0.0148 to 0.0091 from 1960 to 2009, indicating increased landscape fragmentation; the maximum degree of fragmentation occurred in the lakes. The PD index of the rivers decreased significantly from 0.0293 to 0.0061 due to human activities. In addition, the landscape diversity index (SHDI) and landscape evenness index (SHEI) decreased gradually from 5.5561 to 1.1871 and from 11.1123 to 2.3742, respectively, from 1960 to 2009 (Table 6). This shows that the artificial wetlands in the Tumen River Basin have experienced an increasing degree of fragmentation during the past 50 years due to human activities.

The number of wetland patches decreased from 452 in 1960 to 305 in 2009 in the Tumen River midstream area, and the MPS decreased from 61.6205 hm 2 to 40.3624 hm 2 (Table 6). Table 6 shows that in the Tumen River midstream area, marshes (20.77%) and paddy fields (66.07%) accounted for a larger proportion of total wetlands in 1960, and rivers (25.67%) and artificial wetlands, especially paddy fields (72.33%), accounted for the largest proportion of total wetlands from 1976 to 2009. The wetland PD index decreased from 1.41 to 0.42 from 1960 to 2009, indicating increased landscape fragmentation. The maximum degree of fragmentation was of marsh. The PD index for marshes increased more than threefold, whereas the PD index of the rivers decreased significantly from 0.0423 to 0.0155. The PD index of the paddy fields increased from 0.0103 to 0.0237 due to human activities. The SHDI decreased gradually from 1.2406 to 1.1813, whereas the SHEI increased from 2.8806 to 3.3626 from 1960 to 2009. These trends show that the degree of fragmentation of the wetland landscape in the Tumen River Basin has been increasing during the past 50 years. The patch type diversity decreased due to human activities. There was no clear advantage for the spot type in the current wetlands, except for the paddy fields, and the block type has a uniform distribution in the landscape. This uniform distribution is mainly because roads have been built in the region, and ponds and artificial canals were constructed during this period. Therefore, the integrity of the wetlands was undermined, and wetland fragmentation accelerated.

Table 6 shows that in the Tumen River downstream area, the number of wetland patches increased from 110 to 256 from 1960 to 2009. The MPS decreased from 43.3814 to 18.7717 hm 2 during this period. In the wetland types, the variation in the number of river patches was the largest, whereas the variation in the number of lake patches was smallest. First, the number of marsh patches doubled, and the MPS of marshes decreased by 79% (34.78 hm 2). Second, the number of river patches decreased rapidly from 39 to three, whereas the MPS of rivers increased eightfold, which indicated that in recent years, the water level dropped, the area of rivers decreased, and some small reaches disappeared. Third, the number of paddy patches increased from 31 to 126 from 1960 to 1996, but decreased to 64 by 2009. The MPS of paddy fields changed from 39.0404 to 13.4762 hm2 from 1960 to 1996, but had increased to 24.2760 hm2 by 2009. In addition, the marshes and rivers accounted for a larger proportion of the total wetlands from 1960 to 1976. The rivers accounted for a larger proportion in 1986, and the artificial wetlands accounted for the largest proportion in 2009; in particular, paddy fields accounted for the largest proportion (32.33%). In addition, the wetland PD index increased from 0.0231 to 0.0533 from 1960 to 2009, indicating increased landscape fragmentation, with the maximum degree of fragmentation occurring in marshes. The PD index for marshes increased by 79.05%, whereas the PD indices of rivers and paddy fields decreased significantly from 0.0319 to 0.0031 and from 0.0742 to 0.0412, respectively, due to human activities. Moreover, the SHDI decreased gradually, even though the SHEI increased. From 1960 to 2009, the SHDI decreased from 0.5458 to 0.4795 and the SHEI increased from 1.0916 to 1.1134. These trends show that, during the past 50 years, the degree of fragmentation of artificial wetlands has been increasing due to human activities. The patch type diversity has decreased. Except for paddy fields, there is no clear advantage of the spot type in the current wetlands, and the block type has a uniform distribution in the landscape. This uniform distribution is mainly a consequence of the roads, ponds and artificial canals constructed during the study period which, combined with population growth, undermined the integrity of the wetlands and accelerated wetland fragmentation.

Discussion

Driving factors

Natural factors

Climate is the primary factor controlling the dynamics of wetland formation and material exchange (Sutula et al. 2003), and the influence of climate change on wetlands is expressed primarily through changes in precipitation and temperature, which determine wetland plant physiology (Dawson et al. 2003). Accordingly, as previous studies have shown, temperature is the most fundamental factor controlling the dynamics of wetland landscape patterning (Song et al. 2012). The temperature increased throughout our study period from 1960 in the Tumen River area, and the annual average temperature increased by 2.27 °C over the past 50 years (Fig. 3).

The change in temperature plays a role in the water evaporation process. For example, natural wetlands were reclaimed over a large area, which indirectly led to an increase in air temperature in the region. Some environmental factors changed because of changes in climate, such as the land surface temperature, amount of precipitation, average temperature, sunshine duration, and water content in the soil, which had a major impact on the wetland plant community distribution in the Tumen River Basin.

Fig. 3
figure 3

Change in annual average temperature for the Tumen River Basin

Full size image

Water is a critical factor controlling the formation and development of wetland vegetation and soil (Kong et al. 2009). Accordingly, in most cases, changes in hydrology have direct effects on the energy cycle and material flow in wetland systems; such systems are particularly sensitive to hydrological changes such as variations in the quantity and quality of their water sources (Wang et al. 2011). In the last 50 years, the annual average rainfall decreased by 127.4 mm in the Tumen River Basin (Fig. 4), which had a profound effect on the processes involved in the maintenance of the wetland landscape pattern. Wetlands are particularly sensitive to changes in the water supply. For example, the wetland area decreased considerably in the Tumen River Basin because the wetlands were reclaimed, resulting in soil desertification. The wetland soil fertility also declined. As the period of reclamation increased, the content of wetland soil organic matter and other factors decreased each year. With the further development of agricultural activities, soil water content and fertility were reduced, preventing soil reversal to natural wetlands. Furthermore, the plant community structure and its distribution pattern would also have changed, which will also hinder efforts to protect biodiversity and enable recovery of the ecology of the Tumen River Basin.

Fig. 4
figure 4

Change in precipitation for the Tumen River Basin

Full size image

Human factors

The relationship between changes in wetland landscape patterns and human activities is currently the topic of considerable discussion in the field of landscape ecology. Many previous studies have posited that high-density human activity is the primary factor controlling the evolution of the wetland landscape pattern (He et al. 2011; Song et al. 2012).

Our results demonstrate that, first, changes in and the development of the regional ecological environment are often affected by national policy, and that changes in policy have a significant influence on wetlands in the Tumen River Basin. Second, statistical data have shown that a large number of people have migrated to the Yanbian region over the past 50 years due to the relatively stable environment and rich land resources in the Tumen River Basin, which has led to a rapid increase in the population in this region. Third, the construction of hydrological infrastructure has affected the area; villages depicting folk customs and golf courses were established on the Hai Lan River on the outskirts of Yanji midstream of the Tumen River in 2003, and downstream of the Hai Lan River was intercepted by artificial barriers, which reduced the flow of the river. Finally, economic factors are the significant driving forces leading to changes in the wetland landscape since the mid-1980s in the Tumen River region.

Research and the collection of statistical data have been limited due to administrative divisions, leading to databases that were not unified and thus contained errors. RS and GIS each have advantages, such as high accuracy, objectivity, and efficiency compared to the previous methods (Bai et al. 2008). Thus, operating efficiency and data accuracy can be improved using these techniques (Liu et al. 2009). However, this study was also limited by the data and technological means; the accuracy of image interpretation could be improved by overcoming these limitations, such as data phase, clarity, resolution, and other details (Sulman et al. 2013). For example, upstream of the Tumen River is located in the Changbai Mountain, thus research is restricted by topography. We could not verify the data through a long-term field survey. A mixture of multi-type, multi-temporal and multi-band RS data can be useful for solving these problems (Alados et al. 2004). Furthermore, in our study, the selected landscape indices were all calculated from data obtained from remote sensing images; this technique often results in considerable errors relating to incorrect classification during the calculation of landscape pattern indices (Langford et al. 2006).

We also attempted to explain the causes of the wetland landscape dynamics, and we will investigate the relationship between different types of driving forces and landscape dynamics in a future study. For example, we will determine why the plant community changes with the community structure and distribution pattern of adaptation (Alados et al. 2004). The original habitat declined, and the structure of the wetland plant community changed, which led to the extinction of organisms with weak adaptability, and to a decrease in biodiversity.

Furthermore, the wetland landscape dynamics were affected by both natural and human factors and thus exhibit a high degree of homogeneity and regionalism. Multivariate statistical methods enable the quantification of different types of driving forces (Wang et al. 2013). Despite this, human activities, which lead to significant changes in land use/cover and affect the wetland considerably, are still difficult to quantify (Maltchik et al. 2014); however, the driving force analysis that we propose in this study provides a reference for related studies. We also intend to solve this problem in the future. In addition, an integrated multi-scale framework must be applied to wetlands to reveal the driving forces of landscape dynamics (Jiang et al. 2014). For this, analysis is mainly at the river basin scale, but a better result would likely be obtained by reducing the scale to the township or even household level (Jiang et al. 2014).

Finally, the eco-environmental effects of wetlands are quite complex; temporal and spatial variations were investigated here via RS. A further study on the effects of the regional wetland eco-environment may provide a better foundation for wetland restoration efforts (Merot et al. 2003).

Conclusion

Digital analyses based on topographic maps and RS images at a decadal scale for the past 50 years showed that the wetland area of the Tumen River Basin has decreased by 50% over the last 50 years. Aside from the Yuanchi upstream and five natural marshes downstream, all other natural lakes nearly disappeared during the study period. The areas of marshes, rivers, and paddy fields decreased, whereas the area of other types of artificial wetland showed a significant increase; regards landscape conversion, a large amount of natural and artificial wetlands were transformed into non-wetlands, and a large number of paddy fields were transformed into dry farmland and building sites. Regards landscape dynamics, the number of landscape patches, the mean patch area, and the landscape diversity decreased, whereas the landscape fragmentation index and evenness index increased.

Wetland changes in the Tumen River Basin were driven by the interaction between natural factors and human intervention. Climate warming created a potential environment for the conversion of wetlands into non-wetlands. The increase in cropland and built-up land in the region was related to the rapid population growth and the improved quality of life since the implementation of the opening-up policy in China. As a result, the Tumen River Basin suffered environmental consequences, which cannot be ignored.

Finally, the results of our study may help others to better understand wetland variations in other important regions of China and even in other countries. The results and conclusions presented here are useful for the formulation of governmental policies that encourage the ecologically and environmentally sensitive utilization of land resources, sustainability, and proper ecosystem management of areas under increased pressure from population increases and climate change.