Large Scale Wetland Mapping and Evaluation

15570 words (62 pages) Full Dissertation in Full Dissertations

06/06/19 Full Dissertations Reference this

Disclaimer: This work has been submitted by a student. This is not an example of the work produced by our Dissertation Writing Service. You can view samples of our professional work here.

Any opinions, findings, conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of UK Essays.

Abstract: Wetlands are vital for human survival. Considering there is no common wetland definition which is accepted by all domains or sectors, the RAMSAR wetland definition is recommended for large scale wetland mapping. Three kinds of methods for developing wetland dataset at large spatial scale, comprising wetland mapping by remote sensing, compilation of historical datasets and simulation using physical models, are reviewed and assessed.

Large scale wetland mapping is challenging because of the complexity in types and spatial heterogeneity of wetland landscape. Wetland-related datasets suffer from major inconsistencies. Estimates of the areal extent of the remaining global wetlands range from 1.53 to 14.86 million km2. Until recently there are two global wetland simulation datasets which have been developed by Zhu et al. (2014) and Hu et al. (2017) without considering influences of anthropogenic activities. Based on Hu’s simulation of global wetland and Globcover2009 dataset, global wetland loss was at least 33% as of 2009, including 4.58 million km2 of wetlands without open water and 2.64 million km2 of open water. The areal extent of wetland loss has been greatest in Asia, but Europe has the greatest proportion.

The top 100 globally large RAMSAR Sites have been mapped based on MODIS time-series data in 2001 and 2013, and its changing characteristics and landscape integrity were evaluated. Results show that the global wetland area has maintained a stable state with less than 1% of wetland area loss. However, significant fluctuations exist within RAMSAR sites caused by the cumulative effects of natural conditions, i.e. rainfall and temperature, and human activities. The change rate of inland wetland area is higher than that of coastal and artificial wetland. From the perspective of wetland disturbance/degradation, those RAMSAR Sites are still under threat at different degrees. At the same time, the effects of protecting China’s national wetland reserves between 1978 and 2008 have been evaluated. Results show that about 79% of the 91 national wetland reserves are in a poor condition with regards to protection area. These are generally located around the Yangtze River, Eastern Coast, the Three Rivers Source, and Southwest China. Only 15% of national wetland reserves are under sound protection, and these are generally located in the upper reaches of the Songhua River.

Since most experiences of local wetland mapping cannot be directly learned or transferred to regional and global scales. More research is needed to conduct global thematic wetland mapping, to obtain more accurate wetland classes and more detailed ecological and environmental information on wetland ecosystems.

 

Key words: wetland mapping; wetland simulation; wetland evaluation; global/national wetlands

1.     Wetland importance and definitions

Wetlands are vital for human survival. They are among the world’s most productive environments, cradles of biological diversity that provide the water and productivity upon which countless species of plants and animals depend for survival (Mitsch and Gosselink, 1993). They play a critical role in climate change, biodiversity, hydrology, and human well-being (Ramsar, 2001). Wetlands regulate both global and local climate through exchange of water, heat, and energy with the atmosphere, and greenhouse gases sequestration and emission (Parish et al., 1999; Takai, 1970, Munger, 2004). They also provide habitats for a range of fishes and wildlife, including many threatened and endangered species (Ramsar Convention Bureau, 2001). With the function of groundwater replenishment, water movement regulation, and water purification, wetlands are crucial in maintaining the hydrological cycle which, in turn, underpins all ecosystem services and sustainable development (Russi et al., 2013). For human well-being, wetlands not only provide fundamental materials for human healthcare but also have the potential to generate considerable economic value (Millennium Ecosystem Assessment, 2005). It is an accepted fact that a wetland is an indispensable resource for humans and wildlife, and substantial wetland loss could be irreversible (Millennium Ecosystem Assessment, 2005).

Though recognition of the environmental services provided by wetlands is increasingly acceptable across the world, there is no common wetland definition which was accepted by all domains or sectors. There are numerous different definitions of wetlands such as ‘bogs and fens’ (peat-accumulating wetlands), ‘marshes’ (herbaceous, frequently inundated wetlands) or ‘swamps’ (forested wetlands), and no standardization of these terms (Mitra et al., 2005). The Ramsar Convention on Wetlands (1971) produced an international, intergovernmental treaty which defined wetlands as “…areas of marsh, fen, peatland or water, whether natural or artificial, permanent or temporary, with water that is static or flowing, fresh, brackish or salt, including areas of marine water the depth of which at low tide does not exceed six meters” (Ramsar, 2011).

Despite the fact that no consistent, concise definition agreed upon commonly by international parties, the definition of the Ramsar is perhaps the closest substitute, although noted to be too broad to have the precision necessary for further scientific inquiry (Keddy, 2000). The Ramsar definition of wetlands has been accepted by many organizations, such as the International Union for the Conservation of Nature (IUCN, World Conservation Union) on a global scale (Navid, 2014).This definition will serve as the foundation whenever the term “wetland” is used in this chapter in order to best encapsulate wetlands on a global/national/regional scale.

As defined by the Convention, wetlands include a wide variety of habitats such as marshes, peatlands, floodplains, rivers and lakes, and coastal areas such as salt marshes, mangroves, and seagrass beds, but also coral reefs and other marine areas no deeper than six meters at low tide, as well as human-made wetlands such as paddy lands, waste-water treatment ponds and reservoirs (Ramsar Convention Secretariat, 2013). There are three wetland groups in this classification, including inland wetland, coastal wetland and marine wetland. However, considering the availability of data sources, some wetland classes, such as seagrass beds, coral reefs and other marine classes less than 6 meters deep at low tide, will not be considered in discussion here.

2. Methods for global wetland mapping

Remote sensing techniques have commonly been accepted as effective for mapping wetlands, especially at large scale (i.e. global/regional/national). However, due to the complexities of wetland ecosystems, there are many challenges in mapping wetlands when only remote sensing methods were employed. Besides the global wetland/wetland-related datasets based on remote sensing techniques, there are two other kinds of methods – compilation and simulation.

2.1 Remote sensing classification methods

Up to now, nearly all remote sensing data sources including aerial photography, multispectral and hyperspectral sensors, light detection and ranging (LIDAR), synthetic aperture radar (SAR), interferometric SAR (InSAR), passive microwave systems, and gravimeter (e.g., Gravity Recovery and Climate Experiment (GRACE)) are employed to map and characterize wetlands, especially at local scales. However, due to limitations in data acquisitions, complicated images processing, and high costs of computing and human resources, only multispectral and SAR data have been used to classify wetlands at regional and global scales.

Manual interpretation of aerial photography and satellite imagery was the main method used in early studies (Dahl, 1990). Because of the complexity of wetland ecosystems and the small number of spectral bands of images, visual interpretation based on expert knowledge is more effective and reliable than computer classification. The early wetland inventory (NWI) of the US was primarily produced using aerial photographs through photo interpretation, field verification, and use of some collateral data sources (Dahl, 1990). The NWI contains more detailed wetland classes and are generally considered to be the most accurate wetland maps available in the United States. The first wetland map of China which has the accuracy of nearly 90% was also developed primarily from manual interpretation of Landsat imagery (Niu, et al, 2009 & 2012; Gong, et al. 2010). While this method generally has high accuracy, it is time-consuming (Phinnet al., 1999; Wright and Gallant 2007) and subjective (Baker et al., 2006). Therefore, it is expensive and slow to update. Moreover, it is limited to local areas or small regions. At the same time, this method is also generally employed to collect training and validation samples in line with the development of computerized classification (Gong et al., 2013).

Computerized classification methods usually comprise supervised and unsupervised approaches. Supervised classification uses training samples to train classifiers to recognize different classes. The advantage of this method is the ability to specify the desired class types. However, it has some limits that the desired classes may not correspond to spectrally unique classes, and that the acquisition of training data maybe time-consuming and expensive (Ozesmi et al., 2002). The Moderate Resolution Imaging Spectroradiometer (MODIS) land cover classification algorithm (MLCCA), which used a supervised decision tree classification approach to classify five months of MODIS data, was used to develop the MODIS land cover product (BU-MODIS) at 1-km spatial resolution. Both permanent wetlands of large areal extent and water bodies were contained in this dataset (Friedl et al., 2002). A 500-m global land cover dataset, the Global Land Cover by National Mapping Organizations 2008 (GLCNMO version2), was developed by the International Steering Committee for Global Mapping (ISCGM). GLCNMO version2 employed a classification method based on Tasseled cap transformation and supervised decision tree (Tateishi et al., 2014) for six land cover classes, including wetlands, water, and mangrove.

The unsupervised classification (clustering) method groups together pixels with similar spectral values and labels clusters with specific classes based on ancillary information (Piwowar, 2005). This approach eliminates the time-consuming training step and the classes are distinct units. However, the clusters may not correspond to desired class types (Ozesmi et al., 2002). A 1-km global land cover database (DISCover) was developed for the Data and Information Systems (IGBP-DIS) initiative (Loveland et al., 2000), for which unsupervised classification of monthly Advanced Very High Resolution Radiometer (AVHRR) Normalized Difference Vegetation Index (NDVI) of 1992 was used. Another global land cover product for the year 2000 (GLC2000) has been produced using SPOT/VEGETATION data based on the ‘regionally tuned’ approach by an international partnership of 30 research groups (Bartholomé et al., 2005). In this product, the definition of wetland, regularly flooded shrub/herbaceous lands was adopted, instead of permanent wetlands. Building on the success of the GLC2000 project, the European Space Agency (ESA) launched the GLOBCOVER initiative and released higher resolution global land cover (GlobCover2005 and GlobCover2009) products by using 300m resolution ENVISAT/MERIS data for 2005–2006 and 2009 (Arino et al., 2009). However, optical and infrared remote sensing is unable to penetrate clouds and dense vegetation cover, particularly in tropical or sub-tropical regions, which is a major limitation. Prigent et al. (2007) merged the passive and active microwave along with visible and infrared observations through an unsupervised clustering technique to detect global inundation at 0.25° resolution (also called Global Inundation Extent from Multi-Satellites – GIEMS). Based on this time-series products, Fluet-Chouinard et al. (2015) established a new inundated dataset (GIEMS-D15) by using a downscaling method to improve the spatial resolution from 0.25° to 500m (i.e. downscaled 15 arc-seconds).

Other image classification approaches employed in large scale land cover mapping include Decision Tree, Random Forests, Support Vector Machine (SVM), and Classification and Regression Tree (CART). These methods often used with spectral indices, such as the NDVI and TM4/TM2 which may be used to separate water from land (Raabe et al., 1996), and TM5/TM2 which is especially good for discriminating marshes (Dobson, 1995). Ancillary data such as soil maps and topography data may also be used. The first 30m global land cover dataset (Finer resolution observation and monitoring of global land cover (FROM-GLC)) employed four classifiers (Gong, et al. 2013). However, only two wetland classes were included: open water and marshes.

In general, wetlands are defined as ‘lands transitional between terrestrial and aquatic systems where the water table is usually at or near the surface or the land is covered by shallow water’ (Cowardin et al. 1979). Changes in plant phenology and soil moisture, and annual flooding patterns of wetland ecosystems present great challenges to wetland mapping, since most mapping efforts are restricted to ecologically and climatically homogenous regions where the wetland landscapes are not the case. The above traditional statistical image classification techniques for land cover mapping cannot yet be used to map wetlands at large scale, because they mainly focus on vegetation features while wetlands are defined primarily based on soil moistures/water conditions. At the same time, the limited knowledge of wetlands in the existing remotely-sensed wetland-related datasets cannot meet the requirements of conservation and management purposes, since most of them only contain few wetland classes and features, which are snapshots in time and do not contain seasonal wetlands.

Although global inundated dataset are time series and could reflect the dynamic changes of wetland, its coarse resolution and uncertainties such as the underestimation of GIEMS in some forested regions and low accuracy of GIEMS-D15 in mountainous regions (Miller et al., 2007; Fluet-Chouinard et al.,2015) have limited its application.

There is no single source remotely sensed data that is best for mapping wetlands under all circumstances. This is because different types of remotely sensed data are uniquely suited for identifying specific landscape component of wetlands during different time periods. Therefore, the combinations of spatial, temporal, and spectral aspects of remotely-sensed data are needed to map wetlands in the future.

2.2 Compilation methods

Another way to produce global wetland datasets is the compilation of historical data, including independent feature dataset (water, vegetation, and soil) and existing wetland-related datasets (wetland maps and land cover products). In order to estimate area, location and environmental features of global wetlands, Matthews and Fung (1987) combined three independent global digital datasets (vegetation, soil properties, and fractional inundation), and developed a global dataset of natural wetlands at 1° resolution. Through compiling published information and various maps that were basically drawn from regional wetland surveys and monographs, Aselmann and Crutzen (1989) created a dataset on the distribution and seasonality of global freshwater wetlands and rice paddies. Another recent global wetland database is Global Lakes and Wetland Database Level-3 (GLWD-3), which was developed in 2004 by combining the available sources for lakes and wetlands on a global scale with a high spatial resolution of 30 arc-seconds (Lehner and Döll, 2004). For biodiversity and ecosystem modeling, Tuanmu (2014) integrated four global land cover products (DISCover, GLC2000, BU-MODIS, and GLOBCOVER 2005) and developed a global 1-km consensus land cover product, using a generalized classification scheme and an accuracy-based integration approach.

Construction of wetland datasets through compilation could make full use of existing information and represent the best choice when there is no other specific wetland data. However, it is challenging to integrate a variety of datasets, which have different application purposes, wetland definitions, mapping methods and dates, and to reduce the uncertainties within each dataset that could be inherited by the final result. At the same time, any inherent errors in those historical data may be propagated in the new data products. Therefore, these compiled datasets are a bit outdated and static in, and cannot reflect the seasonal hydro-dynamics of wetland. For instance, GLWD-3 was compiled from maps mostly generated prior to 1996 and, in some arid or semiarid regions, this dataset has overestimated some wetlands because of the dynamic inundation (Lehner and Döll, 2004).

2.3 Model simulation methods

Water is a key factor in wetland occurrence. Topography, climate, and soil features have impacts on the spatial-temporal distribution of water simultaneously and their combined effect controls wetland formation and distribution. Some hydrological models that are based on these relationships have been developed to simulate the areal extent and spatial distribution of wetland. For example, Fan et al. (2011, 2013) and Zhu and Gong (2014) simulated global wetland at 1-km spatial resolution based on the relationship between water table depth and wetland distribution, respectively. Although distributed hydrological models could capture the most prominent features of wetland occurrence and extent, the solution of model and the acquisition and precision of the model parameters limit its application, particularly at larger scales (Fan et al., 2011). To avoid this problem, Hu et al. (2017) simulated the potential distribution of global wetlands by employing a new Precipitation Topographic Wetness Index (PTWI) and global remote sensing training samples.

The increasing awareness of the major role wetland plays in global climate system has led an increasing number of Land Surface Models (LSMs) to take account of wetland in their schemes. However, the simulation results of those LSMs commonly have coarse spatial resolutions (the finest one is only 0.5°). Furthermore, there are considerable disagreements on the area of global wetland among their results (Melton et al., 2012). The lack of global validation samples also causes uncertainties in the reliability of LSM simulations.

Wetland modeling represents an efficient way to simulate wetland, especially when physically based models can reflect the formation mechanisms of wetland. One of the most important advantages of wetland models is that these models can not only trace back the historical wetland distribution, but also predict the future changes under different scenarios. But the major hindrance is that all models are simplifications of reality due to the complexity of wetland. There are still considerable uncertainties in the simulation results. In addition, as human interference as a dominant driver of wetland changes, its quantification and integration with the environmental factors in wetland models requires more attention.

3 Global wetland status

3.1 Global wetland loss and status

(1) How much wetland has the world lost as a result of direct human activities

Agriculture and urbanization, two main human activities, directly cause wetland loss (Gong et al., 2013; Niu et al., 2012). Hu et al. (2017) simulated the potential distribution of global wetlands by employing a new Precipitation Topographic Wetness Index (PTWI) and global remote sensing training samples. This simulated global wetland distribution can be regarded as the natural area of global wetlands without human exploration. GlobCover2009 as a global dataset with a 300m resolution provides the spatial extent of direct human activities. Five land-cover categories, including post-flooding and irrigated croplands (11), rain-fed croplands (14), cropland/vegetation mosaics (20), vegetation/cropland mosaics (30), and artificial surfaces and associated areas (190), were considered the results of direct human activities and were extracted from the GlobCover2009.

Areas that are both placed in the wetland category in the simulated global wetland distribution map and directly affected by human activities were extracted from GlobCover2009 and were counted as wetlands lost due to human activity. Overlapping those two data layers shows that, as of 2009, the world had lost 33% of its wetland in area, including 4.58 million km2 of wetland with no open water and 2.64 million km2 of water. Although wetland loss takes place all over the world, the wetland loss situation that was indicated by the ratio of wetland loss (equal to the lost area/potential area) varied greatly among continents (Figure 3-1-1). The largest wetland loss has occurred in Asia with about 2.65 million km2 and the least happened in Oceania with about 0.18 million km2. However, the most serious situation was observed in Europe, which has lost 45% of its wetland. This is followed by South America and Asia with approximately 32% and 27% of wetland lost, respectively. This global wetland loss situation shows a similar trend with population density of 2009 among continents according to the Food and Agriculture Organization Corporate Statistical database (Figure3-1-1), which confirmed that the disappearance of wetlands was principally attributable to human activities. The vast populations, which are always accompanied by high levels of demand for food and housing and then result in the acceleration of the development of agriculture and urbanization, was one of the most important factors that contributed to the severe wetland loss observed worldwide.

Davidson (2014) argued that the long-term global loss of wetland could be 54–57%, which is larger than Hu’s estimation (Hu et al., 2017). The difference could come from any of the following: (ⅰ) some types of wetland, such as intermittently flooded wetland (wet meadows, flooded forested areas in the Amazon and Congo) were not well included in Davidson’s research (Davidson, 2014), which can lead to a smaller base of global wetland areas. (ⅱ) Davidson’s research was based on published papers and reports which are few in number and affected by geographical bias in the numbers of published reports among different regions and at different spatial scales (Davidson, 2014). For example, the shortage of published reports for changes in total wetland area in Africa, the neotropics, and Oceania would inevitably affect the total figure of wetland loss worldwide (Davidson, 2014). (ⅲ) The accuracy of GlobCover2009 products could also have an impact on the assessment of wetland loss. For example, urbanized surfaces and associated areas (urban area < 50%) are not contained in the GlobCover2009 dataset (Arino et al., 2012), which may have caused an underestimation of global wetland losses.

The 8% figure for wetland loss in North America is surprising because it is much lower than the 56% reported by Davidson (Table 3-1-1) and the 50% put forward by Shaw and Fredine (1956) for the United States alone. However, the papers and reports upon which Davidson’s conclusion was based were principally concerned wetlands in the United States and some specific parts of Canada, while omitting those of Central America (Davidson, 2014). Whether these regional scale papers and reports represent the whole continent is debatable. A series of reports on the status and trends of wetlands in the conterminous United States from the 1780s to 2009 show approximately 50% of the wetlands in the conterminous United States have been lost. Among the vanished wetlands, 50–60% were lost to rural and urban development and agriculture (Dahl et al., 1990, 1991, 2000, 2006, 2011). Consequently, it can be concluded that 25–30% of wetlands has been lost because of rural and urban development as well as agriculture in the conterminous United States, which is roughly equal to the current estimate of 24% for wetland loss in this area. However, the small wetland loss ratio in Canada of 2% causes the low overall wetland loss ratio for North America that was observed in Hu et al (2017).

Eurasia is home to 40% of the world’s wetlands, while the area lost represents for 50% of the total loss area of global wetlands. Among all areas globally, Europe shows the most serious situation with respect to wetland loss both Hu’s and Davidson’s research (Table 4). Of all the continents, Asia has the largest overall area of wetland loss. According to the current estimation, nearly 29% of China’s wetlands have been lost due to direct human activity, which is close to the figure cited in a previous study by Niu et al. (2012), which paper stated 33% of China’s wetlands had been lost between 1978 and 2008.

Compared with the ratio of wetland loss caused by agriculture, provided by the OECD (1996), the largest difference was observed among areas such as South America and Africa where the estimate made in the current work is notably larger. However, in other areas, such as North America and Asia, Hu’s estimate is similar to that previous estimate (Table 3-1-1). It is not possible to support or refute the value given by the OECD in Africa or South America using existing information because the limited number of papers, reports, and inventories available for these two areas leave a dearth of a reliable data on wetland areas and this then leads to uncertainty in assessing wetland loss. However, small-scale studies of these regions have shown alarming ratios of wetland loss. For instance, Taylor et al. (1995) found that the Tugela Basin and the Mfolozi catchment in South Africa lost approximately 90% and 58% of their wetlands, respectively. In South America, Colombia’s Cauca River Valley lost 88% of its mapped wetlands between the 1950s and 1980 (Moser et al., 1996). Moreover, according to a 1999 global wetland inventory (Finlayson et al., 1999), wetlands in Africa and South America in tropical and sub-tropical zones have been seeing increasing ratios of loss since the 1950s, alongside their rapid economic development. In this way, the ratio of wetland loss on these two continents may be greater than indicated by the OECD.

(2) Global wetland status

With the rapid population growth and economic development, the decreasing trend of global wetlands continues to occur all over the world (Davidson, 2014; Niu et al. 2012; Zhang et al., 2013). Accurate wetland information is indispensable for understanding the pattern of global wetland changes and for making policies on wise utilization and conservation of wetland. However, the knowledge on global wetlands is limited up to now due to various causes (Hu et al. 2017). Wetlands can be regarded as one special land cover type, special ecosystem and natural resources in different fields, by which the datasets on global wetlands was developed in respective manners with various understanding.

Remote sensing can be a useful tool for large-scale mapping of wetlands. However, extensive discrepancies exist among the various satellite-based global wetland-related land-cover datasets (Jung et al., 2006; Giri et al., 2005). Nakaegawa (2012) found that the inconsistency among six 1-km resolution global wetland-related land-cover datasets (GLCC.S, GLCC.I, GLC2000, BU-MOIDS, GLCNMO, and GLWD-3) was more than 70%. This situation leads to variation in figures for global wetland area, which ranges from 0.29 to 9.78 million km2 (Friedl et al., 2002; Lehner et al., 2004). In this way, none of these global wetland-related datasets alone can reflect the real situation of the world’s remaining wetlands.

To understand the possible coverage of remaining global wetlands, the consistency of estimates of the remaining global wetlands was calculated by synergizing four global wetland-related datasets, GLCC.I, GLC2000, BU-MODIS, and GLWD-3, for which available satellite images can be dated back to circa 2000 (Hu et al. 2017). The consistency was defined as the conformity of the same category among various datasets and was calculated pixel by pixel. If a pixel shows identical categories among the four datasets simultaneously, the consistency is 100%; if three datasets indicate the same category in the pixel, the consistency is 75%, and so on. The results show that the possible areal extent of global wetlands could range from 1.53 to 14.86 million km2 (Table 3-1-2). In addition, the inconsistency of the spatial distribution of wetlands among those various satellite-based global land-cover products is prominent. These four datasets simultaneously identified an area as wetland for less than 8% of the total area.

At present, providing an accurate and acceptable figure related to the areal extent of wetlands on a global scale is very difficult based on currently available information. However, those previous efforts could help outline the sketch of a global wetland map.

3.2 Inconsistency among global wetland-related datasets

Inconsistency of global land cover products has been confirmed by several studies (Herold et al., 2008; Jung et al., 2006; Giri et al., 2005). Furthermore, identification of wetland is the most difficult task during land cover mapping. The inconsistency of wetland among various datasets is prominent which usually manifests in the following aspects:

  1. Different understanding for wetland concept

Wetlands are commonly regarded as the transitional areas between terrestrial ecosystems and aquatic ecosystems, but it is difficult to give a clear definition. For example, although the definition provided by the Ramsar Convention has been widely accepted in the political arena, it is criticized for being too embracing (Sudip et al., 2003). Furthermore, there is no standardization of terms for wetland, such as ‘bogs and fens’ (peat-accumulating wetlands), ‘marshes’ (herbaceous, frequently inundated wetlands) or ‘swamps’ (forested wetlands) (Navid, 1989). Therefore, the aforementioned wetland-related datasets have different wetland definitions and legends depending on the purpose of their application. For instance, wetland in datasets like IGBP and BU-MODIS refer to permanent wetlands while the definition of wetland in GIEMS include not only natural but also man-made wetlands.

  1. Inconsistency of areal extent

The area extent of global wetland varies greatly among different datasets from 0.54 million km2 to 21.26 million km2 (Hu et al. 2017). Even if datasets are produced by the same kind of approach, wetland areas remain largely different. For example, wetland area in land cover products ranged from 0.54 million km2 to 6.04 million km2 (Chen et al., 2015; Bartholomé et al., 2005) and the simulation results compared by the WETCHIMP showed variation in area ranging from 2.7 million km2 to 8.17 million km2 (Melton et al., 2012). The 1999 global wetland inventory report estimated global wetland area to be between 7.48 million km2 and 12.79 million km2 (Finlayson et al., 1999). In contrast, the Millennium Ecosystem Assessment (2005) stated the figure is more than 1.2 million km2. Therefore, according to available findings, it remains difficult to ascertain the global extent of wetlands.

  1. Inconsistency of spatial distribution

The disagreement in spatial distribution among global wetland-related datasets is also prominent. Martin et al. (2009) compared three different datasets, which include Matthews and Fung’s (1987) compilation, the land cover map from the International Satellite Land Surface Climatology Project (ISLSCP map) which is no longer accessible, and DISCover. The result showed that Matthews and Fung’s compilation and the ISLSCP map only matched 57%, and the match among the three datasets was even lower. Nakaegawa (2012) compared three water-related land cover types (snow and ice, wetland, and open water) in six 1-km global land cover datasets (GLCC.S, GLCC.I, GLC2000, BU-MOID, GLCNMO version1 and GLWD-3) by calculating the class-specific consistency. The result indicated that the agreement for open water is about 67%, but the situation of wetlands without surface water reduces to only 30% that a wetland pixel in a 1-km global land cover dataset was classified correctly in the same pixel of the other five datasets.

The chronological inconsistency of datasets in Nakaegawa’s research undoubtedly contributes to the uncertainties in the result. At the same time, the class-specific consistency was calculated just between each two instead of the six datasets. Therefore, in order to avoid these limitations, Hu et al. (2017) compared four wetland-related datasets which were collected before 2000 (Table 3-2) and calculated the class-specific consistency among these datasets simultaneously to check whether, at a particular time, the pixels had the same wetland type. Result shows that the spatial consistency of water among the four datasets is about 23% and the spatial consistency of wetlands without surface water is less than 1% (Figure 3-2).

3.3 Reasons

The widespread inconsistency among global wetland-related datasets mainly comes from two sources.

  1. Different application purposes

The different application purposes of different projects lead to a diverse understanding on wetland definition, wetland classification system, input data sources, and mapping methods, which eventually produce a variety of datasets. For example, remote sensing based datasets mainly come from global land cover products, which have been primarily directed towards identifying land surface characters (e.g. vegetation features) (Gumbricht, 2012). By contrast, compilation datasets were developed under the demands of CH4 emission calculations. Hence, these two kinds of datasets have different classification systems. For example, only permanent wetland and some regularly flooded areas are considered in the classification system of remote sensing datasets, while the compiled datasets have more detailed categories ranging from natural wetland to rice paddies (Table 1). At the same time, the methods adopted in global land cover datasets may not be suitable for wetland detection, since they are not specific to wetland mapping. For instance, the classification method used by IGBP-DISCover was unsuitable to identify wetland, which may have been one reason that led to the underestimation of the wetland area (Loveland et al., 2000). Most of the land cover products use optical images, which cannot penetrate dense vegetation canopies, and thus may underestimate wetlands such as swamp and mangrove in tropical and sub-tropical areas (Hess et al., 2015). The compiled datasets could not only inherit uncertainties from the data sources and the process of aggregating them, but also include seasonal dynamics of wetland (Matthews and Fung,1987; Lehner and Döll, 2004) due to the limitations of the input data sources.

  1. The intrinsic features of wetlands

Wetlands are characterized by seasonal and annual variations in hydrology (Loveland et al., 2002; Herold et al., 2008). The areal extent of wetlands may change considerably across different seasons (Prigent et al. 2007). Therefore, the chronological inconsistencies of the input data sources could result in disagreements among the datasets. Furthermore, it is difficult to distinguish seasonal wetlands if only single-date imagery is used in wetland mapping.

Another wetland feature is its transitional environmental characteristics, which come from its geospatial location between terrestrial and aquatic ecosystems. The spectrum of wetland is always a spectral mixture of water, soil, and vegetation, which poses great challenges to wetland mapping using remote sensing. The sharp spatiotemporal changes of landscape elements often result in confusion between wetland classes.

4. Suitability mapping of global wetland areas

Global Compound Topographic Index (CTI) dataset and the relation between wetland and water table depth were used to model the suitability distribution of global wetland. Considering coastal wetland vulnerable to the influence of tide and its unusual hydrological character, this study was focused on inland wetland. In previous research a sub-grid run-off parameterization scheme had been proposed based on TOPMODEL which used CTI to represent the status of local water storage (expressed as deficit or saturation) (Habets and Saulnier, 2001). This parameterization scheme was then coupled with land surface model ISAB and ORCHIDEE respectively, thus described the proportion of saturation area in each sub-grid more accurately (Decharme and Douville, 2006; Decharme and Douville, 2007; Ringeval et al., 2012). Those methods above offered a reference for Zhu and Gong (2014) to construct the relation between water table depth and CTI, but the scale used in this research was still too coarse for wetland mapping. Zhu and Gong (2014) utilized high resolution topographic dataset and related CTI with a one-layer water balance model. Then the average water table depth at 5min resolution attained with hydrological model was downscaled to 30 second resolution by using CTI. Finally, they characterized the global land surface wetland suitability distribution without accounting for impact of human activities by selecting appropriate threshold for the downscaled water table depth.

4.1. Global wetland suitability areas modeling

The modeling process was divided into two parts: (ⅰ) obtaining soil water content by water balance model, (ⅱ) determining wetland suitability areas with water table depth. The data required include CTI derived from global digital elevation model (DEM), climatic data used to force the water balance model, wetland/no wetland distribution data for parameter calibration.

(1) Datasets

From 60°S to 60°N, Zhu and Gong (2014) used 1km flow accumulation data HydroSHEDS (http://hydrosheds.cr.usgs.gov/) from USGS to calculate CTI based on equation (4-1). The flow accumulation data HydroSHEDS established on Shuttle Radar Topographic Mission (SRTM) had been processed to eliminate dam and false sink and reduced the error caused by DEM interpolation. For area above 60N, the CTI was obtained by applying CTI.aml compiled by USDA Forest Service Rocky Mountain Research Station to the 1km GTOP30 elevation dataset. The global CTI was produced by combining the two datasets and then transformed it to 30 second resolution with WGS-84 coordinate system.

                 (4-1)

where is the flow accumulation and is the slope with radian measure.

The datasets used to force water balance model is the average monthly climate data at 5min resolution during 1950-2000. These datasets was obtained from WorldClim (http://www.worldclim.org/), including average monthly precipitation, average monthly maximum temperature and average monthly minimum temperature.

To acquire global wetland and no wetland samples, Zhu and Gong (2014) collected five global land cover products and seven regional land cover products for 2000 (Table 4-1-1). Because there is a distinct disagreement with wetland definition, wetland area from these land cover products uniformly according to the wetland definition of Ramsar Convention were extracted. Wetland definition in Ramsar Convention contains not only swamps, estuary, peatland and mangrove but lake, river, alpine meadow and tundra. Among the five global land cover datasets, GLWD was specially oriented to wetland and included 12 wetland types, so all of the area was classified as wetland. GLC2000 was accomplished based on cooperation with various institutes so the wetland definition also varied. All land cover types in accordance with wetland definition of Ramsar Convention as wetland (more details shown in Table 4-1-1) were classified. Land cover types like water body, herbaceous wetland, wooded wetland, herbaceous tundra, wooded tundra and mixed tundra in IGBP-DISCover were all classified as wetland. In GLCNMO and BU-MODIS, the land cover type could be regarded as wetland only contained water body and permanent wetland, from which GLCNMO and BU-MODIS products had less wetland type and the other products contained not only permanent wetland but also floodplains and tundra. Apparently, these datasets are comparable, because they can be unified under the framework of wetland definition in Ramsar Convention.

To unify these datasets with different resolutions, all of these datasets were resampled to 1km. In the process of higher resolution datasets conversion, only each pixels with a proportion of above 80% wetland area were determined as wetland to keep purity inside mixed pixels because of fragmented patches of wetland. To acquire a more accurate wetland samples, Zhu and Gong (2014) took the intersection area among different wetland datasets as the wetland sampling area. Meanwhile, water body was eliminated from this intersection result to adjust the proportion of water samples in wetland samples and the global water body dataset MODIS Water Mask (MWM) was obtained from http://glcf.umd.edu/data/watermask/. In order to get more accurate no wetland sampling extent, all of the wetland extent datasets were joined together and expanded outside 5km to get the maximum wetland extent and then eliminated this extent from the global map. Random sampling points were extracted with HawthsTools developed on ArcGIS from the processed wetland and no wetland extent datasets. Because the area of the intersection and water-eliminated wetland extent was already rather small, about 75000 wetland sample points according to 30% of the total area and about 10000 water points from MWM were gathered. 90000 no wetland points were extracted which was 20% more than the number of wetland points. To evaluate the quality of training sample and considering the possible error mainly resulting from wetland points, 4000 points were randomly sampled from wetland sample. By overlapping these points on Google Earth high resolution image, only 67 points could not be confirmed as wetland, indicating the wetland training sample had a high quality.

Because the evolutional mechanism of land surface wetland varied across different area, especially the climatic factors which determined the way of water provision to support land surface wetland under different climatic conditions. In a previous research, according to different climate system and wetting mechanisms, wetland at different area is divided into four types: (1) mostly precipitation and local runoff-fed, such as bogs and marshland which are disconnected from river networks, (2) mostly surface water-fed but groundwater-supported, such as coastal freshwater wetlands, tropical-subtropical forests and inland floodplains, (3) mostly groundwater fed, such as inland freshwater wetlands in semi-arid climate, (4) mostly frozen ground-supported (Fan and Miguez-Macho, 2011). According to this hydrological mechanism partition, it is necessary to model wetland under various climate with different parameters. Freshwater Ecoregions of World (FEOW) (http://www.feow.org/) was used to construct the model units. Although this dataset was aimed at freshwater biodiversity conservation, it was produced mainly based on drainage basins boundaries. The comparison between this dataset and another global basin dataset USGS HydroSHEDS showed a good spatial agreement. FEOW contained 426 regions, six bioclimatic variables were chosen, including annual mean temperature, max temperature of warmest month, min temperature of coldest month, annual precipitation, precipitation of wettest month, precipitation of driest month from WorldClim, and  their the average of these variables for each region as its feature variables were calculated. These feature variables were then standardized by subtracting the average value and divided by the standard deviation. Finally a clustering algorithm k-means was applied to the 426 regions marked by six feature variables to generate 15 climatic-basin clusters (Figure 4-1-1), which were used as our modeling units. The comparison between our climatic-basin clusters and Köppen Climate Classification System indicated our cluster result can reflect the geographical distribution of global climatic region and also distinguish hydrological control of basin on different wetland ecosystems.

(2) Modeling the distribution of soil water content

Our modeling ran at monthly scale. With the deficiency of high resolution climatic forcing datasets the calculation of soil water content was based on a simple water balance model including the processes of precipitation, evapotranspiration, runoff generation and ignoring the processes of vegetation interception, interaction between land surface water and underground water (recharge and discharge). The soil water content was updated monthly:

    (4-2)

where refer to precipitation, evapotranspiration (ET), runoff and soil water content at time t respectively. The calculated soil water content in equation (4-2) should be viewed as the average soil water content in 5 min grid cell rather than the rigid definition of soil water content in pedology. ET was calculated based on the hypothesis that it has a direct proportion relation with the relative soil water content and potential evapotranspiration (PET) (Prentice et al., 1993):

        (4-3)

                              (4-4)

where , and respectively refer to soil permanent wilting point, soil porosity and relative soil water content. The basic soil datasets were obtained from Harmonized World Soil Database v1.2 (http://webarchive.iiasa.ac.at/Research/LUC/External-World-soil-database/HTML/). Soil permanent wilting point and soil porosity were determined with the same parameters used in NOAH (National Centers for Environmental Prediction Oregon State University Air Force Hydrologic Research Lab) to keep consistent with the later NOAH output variables (Ek et al., 2003). The soil parameters were listed in Table 2. PET was calculated following Hargreaves (Hargreaves and Samani, 1985).

To quantify the impact of surface runoff on soil water content, the following parameterization scheme which was expressed as the function of precipitation and relative soil water content ( Bergström and Singh. 1995)  were adopted:

              (4-5)

where was solved using 0-2m depth soil water content data in NOAH LSM offered by Global Land Data Assimilation System (GLDAS). NOAH LSM is an operational LSM coupling physical processes of the soil-vegetation-atmosphere and characterizing the land surface process more accurately than our model but its coarse spatial resolution prevent it being used in high resolution wetland modeling. Therefore each month’s mean soil water content by averaging NOAH output soil water content datasets for 1950-2000 was achieved and then the parameter with equation (2) and (5) was solved. is viewed as a temporal invariable and scale-independent parameter reflecting the nature of basins. Finally with solved our model ran for 30 times to achieve a steady state and then got the soil water content at 5min resolution (30 times was chosen for it could ensure the soil water content keep invariable). Following equation (4), the average value of relative soil water content for 1950-2000 was calculated, which is shown in Figure 2.

Figure 4-1-2 Average value of relative soil water content for 1950-2000

1.3 Determination of land surface wetland distribution based on water table depth

Under the TOPMODEL assumptions that the surface infiltration rate and soil properties are uniform across a basin and that subsurface transmissivity has an exponential profile with water table depth, the local water table depth in point i of the 5 min grid cell can be expressed as a function of the average grid cell water table and grid cell topography as (Beven and Kirkby, 1979):

             (4-6)

where is the 5 min grid cell mean water table depth (WTD), is the local CTI index in point i of 5 min grid, is the 5 min grid cell mean CTI, and M is a scaling parameter. While equation (6) is established over an entire basin, rather than a grid cell, the 5min grid cells are sufficiently large that the relationship is still approximately true, i.e. pixels for which a significant fraction of their upslope contributing area lies outside the cell boundaries make up only a small fraction of the grid cell (Bohn et al., 2007).

Before acquiring WTD at each point, the mean WTD of 5 min grid cell must be achieved. According to a previous research concerned with modeling hydrological process on peatland (Hilbert et al., 2000), it has the following equation:

                 (4-7)

then the relation between and can be denoted as the following:

                                    (4-8)

where c is the parameter to be determined. Finally can be expressed as:

             (4-9)

The equation above shows that as long as the parameters c and M is determined the global WTD at each point can be solved and then get the global distribution of land surface wetland. However, it is hard to get an accurate value of the two parameters globally and the result is wetland distribution rather than global WTD map, so the following method to determine global land surface wetland distribution by using a optimal threshold to distinguish the wetland and non-wetland area was proposed.

The wetland and no wetland sample points were used as binary variable (B) to calibrate the parameters c and M (when it is wetland, B=1; otherwise B=0). Sigmoidal function was introduced, which could map into 0-1 to relate B with . was used as the solving condition where

          (10)

Parameters c and M were individually solved for 15 global climatic-basin regions given that values of these parameters varied across different climate (Table 4-1-3). To quantify the optimal WTD thresholds to distinguish wetland suitability area, wetland and no wetland samples were used as validation data to determine the accuracy with different WTD as thresholds. Finally the corresponding WTDs when the highest accuracy was achieved were chosen as the optimal thresholds to distinguish wetland and no wetland area (Table 4-1-3).

4.2 Results and discussion

Based on water cycle process and relation between WTD and wetland, the global land surface wetland suitability map was eventually produced (Figure 4-2-1). Through numeric solving Darcy’s law and Richards equation, Fan et al (2013) produced global pattern of WTD. 0.25 m was used in Fan et al (2013) as the threshold to distinguish wetland and no wetland area. Since Fan et al (2013) also did not account for human activities, the result should be also viewed as wetland suitability map (Figure 4-2-2). The comparison between the two global wetland suitability map exhibited a good spatial agreement. The modeling result in Zhu and Gong (2013) indicates that the total area of global land surface wetland is 3.316×107km2, and Fan’s result is 3.376×107km2. The two results were aggregated to 5 minresolution to get the proportion of wetland at each grid cell. The difference between the two results indicated that the main wetland area disagreement was located in North Hemisphere especially the high latitude like north of Canada, West Siberia and Northeast of China (Figure 4-2-3). The reason why there are more differences in north area is the different models employed and the different topographical datasets used above North 60

°

.

Wetland and no wetland big samples, which were selected to be consistent with wetland modeling result in spatial scale, were compiled from Gong et al. (2013). Since the amount of wetland points selected from this dataset was much less than no wetland points, international important wetland points from Ramsar Convention were added to form a validation datasets consisting of 8112 wetland points and 12358 no wetland points. This validation datasets were used to assess accuracy of the modeling result, Fan’s result and the five global land cover mapping based on remotely sensed image. The confusion matrix (Table 4) indicates the Zhu and Gong (2014) result has a higher accuracy in both Producer’s Accuracy (PA) and User’s Accuracy (UA) than Fan’s result and the Overall Accuracy (OA) reached 83.7% which is the highest of all datasets. This indicates the result by Zhu and Gong (2014) holds great potential to work as the basis of constructing global wetland database. Latitudinal distribution of seven global wetland (Zhu and Gong, Fan’s result and five remote sensing based wetland mapping) area fraction shows a consistent spatial pattern (Zhu and Gong, 2014)). In areas mostly impacted by human activities (about 40

°

S-55

°

N), model based results have a larger wetland area fraction than the other five remote sensing based observational results and in areas less impacted by human activities (about above 55

°

N), model based wetland area fractions are in the middle of all results. There are two regions with plentiful wetland along global latitudinal distribution: regions around the equator including Amazon, Congo Basin and Victoria Lake and regions around 60

°

N with tundra and peatland.

The area comparison across the seven global wetland results shows the result by Zhu and Gong (2014) has almost the same area with Fan’s modeling result (Table 4-2-1) and both of them are larger than the five remote sensing based results especially far larger than BU-MODIS and GLCNMO wetland extent. The main reasons are: (1) BU-MODIS and GLCNMO wetland extent only include permanent wetland and water body; although the other three wetland extents include floodplain, the maximum extent of short-term inundation is hard to be captured from a single date remotely sensed image. This is especially true for high-seasonal lake like Poyang Lake and the difference between the maximum and minimum inundation area could reach three times (Sun et al., 2013). Thus a single date remote sensing based results lead to less wetland area. However, modeling results based on topography and hydrologic process have the potential to classify these areas into wetland suitability areas. (2) Both of the two modeling results did not account for impact of anthropogenic water regulation like irrigation and reservoirs on wetland which decrease the wetland area. The main reason of no representation of anthropogenic water regulation is the lack of human activities datasets, which is also the current main difficulties in large scale hydrological modeling (Pokhrel et al., 2012; Wood et al., 2011). Moreover, a larger wetland suitability area is also expected, for this area is used to work as the maximum boundary and possibility of wetland beyond this mask is ignored.

With special hydrological process of coastal wetland, its large variation in width along with the topographic structure of sea-land system and being hard to represent the subtle surface relief in a coarse resolution DEM datasets, Zhu and Gong (2014) concentrated on land surface wetland suitability mapping to reduce uncertainty. Their model is relatively simple when compared with models used by Fan and WETCHIMP. However, it shows a better accuracy based on the validation. The first reason is that it adopted a climatic-basin units to distinguish modeling parameters resulting from the different hydrologic control on wetland water source under different climate. Another reason is that it incorporated soil water content datasets from NOAH which have been assessed and adopted by many other researches (Rodell et al., 2007; Syed et al., 2008; Zhang et al., 2008) and showed a good quality. The result also proves that exploring and modeling global wetland at high resolution from a water cycle perspective is feasible. Since the impact of human induced disturbance to natural water cycle was not considered, this result overestimated the area of global wetland. What can be done further is to include more human impact factors like agricultural irrigation and reservoir and add more hydrological processes like interception by vegetation, surface runoff generation and interaction between surface water and groundwater (recharge and discharge), which can be achieved by revising existing LSM or constructing new ones.

5. Wetland mapping of Large Ramsar sites

5.1. Methods and data sources

In 2 February 1971, the Convention on Wetlands of International Importance especially as Waterfowl Habitat was adopted in Ramsar, Iran. It is often written as “Convention on Wetlands (Ramsar, Iran, 1971)” for short. Under the text of the Convention, wetlands are defined as “areas of marsh, fen, peatland or water, whether natural or artificial, permanent or temporary, with water that is static or flowing, fresh, brackish or salt, including areas of marine water the depth of which at low tide does not exceed six meters”. The List of Ramsar Sites (the “Ramsar List”) was put forward accordingly. This list has designated many wetlands for special protection in terms of ecological, botanical, zoological, limnological or hydrological importance, especially being important to waterfowl all the year round.

  1. Top 100 global large Ramsar Sites

2193 sites are designated in the “Ramsar List” as of May 2015, covering 209 million hectares. Based on the principle of covering large area, possessing verifiable spatial distribution and adequate cartographic information, being typical in ecological climate regions and continents in the world, 100 large area (greater than 2000km2) Ramsar sites were monitored and analyzed in this part (Table 5-1). Two of these typical sites were laid special emphasis on.

 (2) Methods and data sources for wetland mapping

Using MODIS 250meter product in 2001 and 2013 (16-day composite, 23 periods a year) as the main data source, global large area Ramsar sites (top 100 sites as shown in table 5-1) were monitored. In addition, high resolution remote sensing images, topographic and other geographic data were utilized as auxiliary data. Filtering, principal component analysis (PCA), supervised classification and unsupervised classification methods were combined to implement large area Ramsar sites remote sensing monitoring and classification. The minimum unit for wetland mapping is about 56.25 hectares (3 rows, 3 columns) (Niu, 2015b).

Training samples were selected using manual interpretation, with aids of high spatial resolution images in Google Earth and photos in Ramsar websites. At the same time, the Enhanced Wetness Index (EWI), Normalized Water Index (NWI) and slope were also referenced to improve the classification results. The validations of classification were conducted for ten Ramsar Sites, which were chosen randomly from table 5-1. 1% of total pixels of each land cover type in the results were chosen to be validated based on high spatial resolution images, such as GF, ZY3 images with 4-5m resolution. The confusion matrix was used to measure the classification accuracies and to calculate the kappa coefficient of the result (Table 5-2 and Table 5-3). As is shown in Tables5-2 and 5-3, the overall accuracy and kappa coefficient were 88% and 0.86 for 2001, 89% and 0.876 for 2013. The high precision of classification show that the method has good performance, and the classification results are reliable and credible for monitoring the wetland changes.

 (3) Evaluation indices of wetland landscape

Wetland change, including the inter-conversion between wetland and non-wetland or among wetlands, can reflect the situation of wetland ecological environment. When the wetlands were transformed to the non-wetland or the natural wetlands were transformed to the artificial wetlands, it generally means the deterioration of the ecological environment of wetlands, because it is not conducive to protect the biological diversity. On the contrary, it was considered that the ecological environment of the Ramsar Sites turns for the better.

Ecological integrity can be defined as the capacity to support and maintain a balanced, integrative, adaptive biological system, having the full range of elements and processes in the natural habitat of a region at ecosystem scale (Li et al., 2010). Landscape integrity can be one practical metrics to measure ecological integrity at landscape scale by using RS and Geographic Information Systems (GIS) techniques. The value of the index of landscape integrity is between 0 and 1. The higher value indicates a higher integrity of wetland landscape and more conducive to maintain and play the functions of the wetland ecosystem. The changes of index in different period also indicate the changes of the integrity of wetlands ecological system. The following formula is used to calculate the wetlands landscape integrity (Niu, et al., 2015b).

Do

=0.5*Rd+Lp*100%

(5-1)

where Rd is the density, Rd = (the number of wetlands plates / the number of total plates) *100%; Lp is the landscape proportion, Lp = (the area of wetlands plates / the area of total plates) * 100%.

It is inevitable to disturb the wetlands ecosystem when the human beings exploit and utilize wetland resource for the production and living in the different degree, such as the agricultural reclamation, fish farming and so on. Besides, the changes of natural environment conditions, including the precipitation and temperature, also have an impact on the wetlands ecosystem. The combined effect of natural and human factors often leads wetland landscape to be transformed into the non-wetland landscape or the artificial wetlands. This transformation processes can be quantitatively monitored by the degree of the wetlands ecosystem Disturbance and degradation. The wetland degradation and Disturbance index is calculated by the following formula:

D=∑inSi*RiS*100%

(5-2)

Where D is wetland degradation and Disturbance index; Si is the area of the i-th land cover; Ri is the Disturbance coefficient the i-th land cover to the wetland ecosystems; S is the area of the Ramsar Sites. The disturbance intensity coefficients for the different land cover types are confirmed by expert estimation method. The value of wetland degradation and Disturbance index (DDI) is between 0 and 1 while the value is 0 if it was not completely disturbed. The higher value means the more serious Disturbance and degradation.

5.2. Spatial distribution and wetland change of globally Ramsar Sites

(1) The spatial distribution of globally Ramsar Sites

Ramsar sites vary enormously in quantity and area in all continents. Europe has the largest number of Ramsar sites, approximately 898 (Figure 5-2-1). The area of African Ramsar sites accounts for 48 percent of the world’s total, and is at the top of the world (Figure 5-2-2). An unequal acreage also exists between each other. The minimum of Ramsar sites is less than 1 hectare, while the largest is up to 6 million hectares.

 (2) Change of acreage and types of globally large Ramsar Sites

Non-wetland area accounts for 56 percent within global large area Ramsar Sites. Between 2001 and 2013, the total wetland area decreased by less than 1%. Thus it indicates that the global wetland area has maintained a stable state. However, significant fluctuations exist within Ramsar sites caused by the cumulative effects of natural conditions, i.e. rainfall and temperature, and human activities.

The change rate of inland wetlands area is higher than that of coastal wetlands and artificial wetlands. For instance, the area of seasonal herbaceous marshes and floodplain wetland increased by approximately 21% (15402km2, accounting for 1.11% of the area of Ramsar Sites, similarly hereinafter) and 6% (694km2, accounting for 0.05%) respectively. While the inland forested/shrub wetlands area reduced nearly 30% (16350km2, accounting for 1.18%); and the area of rivers and lakes reduced by 9% (522km2, accounting for 0.04%) and 12% (682km2, accounting for 0.49%) respectively. On the whole, the inland wetlands area showed a decreasing trend. On one hand, the decreasing trend of global wetlands also exists in these large area Ramsar sites. On the other hand, inter-annual climatic fluctuation is the cause of instability of inland wetlands.

The area of coastal forested/shrub wetlands increased about 13% (611km2, accounting for 0.44%) between 2001 and 2013. In addition, the change rates of other types were less than 3% (150km2, accounting for 0.02%), showing a relatively stable state. This may be concluded that coastal wetlands are less subjected to the impact of changes in hydrological conditions, and they are mainly related to the influence of temperature and other meteorological condition.

Among all continents, wetland area in North American increased slightly, while that value of the other continents showed different degrees of decrease. The types of wetland area loss are mainly forested/shrub wetland, seasonal herbaceous marshes and lakes. Direct influence of human activities on the wetlands is relatively mild, however, wetland changes are more related to climatic fluctuations. Moreover, Africa, South America and Oceania possess large non-wetland proportions (greater than 53%) within Ramsar sites. In all ecological regions, wetland area in the temperate steppe zone shows the largest loss rate. Except for boreal coniferous forest and polar ecological zones, non-wetlands occupy a higher proportion (more than 50%) in the other four zones.

 (3) Landscape ecological evaluation of global large-area Ramsar Sites

On the whole, the Ramsar Sites in North America show the highest wetland landscape integrity. A larger proportion of the low wetland landscape integrity Ramsar sites were distributed in other continents. For example, the proportion reached 86% in Africa. The results also show that water area decreases in 70 of 100 large area Ramsar sites, mainly in Africa, Asia and South America.

From the perspective of wetland disturbance/degradation degree, global Ramsar Sites are still under threat. In 2001 or 2013, the number of Ramsar sites which are under worse interference (disturbance/degradation index >20%) is 19 and 22 respectively.

5.3 Wetland change analysis of some typical Ramsar Sites

(1) Inland wetland: Lake Mar Chiquita in Argentina

Lake Mar Chiquita, covering a total area of 9960km2, was listed as the wetland of international importance in May 2002 (Ramsar No. 1176; the center of its coordinate: 30°23’0″S,62°46’0″W). It is the largest inland floodplain basin of Argentina, and is also one of the largest saline lakes in the world. Animal husbandry, fishery, forestry and agriculture are the main industries of this area, therefore water resources shortage is a huge threat to the wetland. In 2013, the area of lakes with perennial water declined sharply, approximately half of the lake was lost. The disappeared lake was replaced by seasonal herbaceous marshes in a great measure, and small part changed into floodplain wetland. In the next place, non-wetland type changed greatly. Dry land grew twice as large as that in 2001. In general, agricultural reclamation and other human activities made a strong impact on the natural wetlands.

(2) Coastal wetland: Sundarbans Reserved Forest in Bangladesh

Sundarbans Reserved Forest is situated in the southwest coast of Bangladesh, near the Bay of Bengal, with a total area of for 6017km2. In 1992, it was listed as a wetland of international importance (Ramsar No. 560; the central coordinates: 22°2’0″N,89°31’0″E).

Between 2001 and 2013, about 46km2 (accounting for 1% or so) of the dominant wetlands type, coastal forested/shrub wetlands, were converted to dry land at a small scale. Bangladesh has one of the world’s largest mangrove planting project. Since 2000, the World Bank has provided funding to the mangrove ecological environmental protection project, to protect species diversity, to prevent the deterioration of ecological environment. At present, protection of mangrove forests in the region achieved better results. The coastal evergreen forested/shrub wetlands area of 2013 is 7% higher than that in 2001 (about 300km2, accounting for about 5%).

6. Evaluation of China’s national Wetland reserves

Wetlands have the most abundant biodiversity (Costanza, et al. 1997), the highest carbon sequestration capacity (Duarte et al. 2005), and the highest values for ecological services per unit area, of all the world’s ecosystems. Protecting wetlands and their biodiversity has attracted significant attention from the international community (http://www.wetlands.org; 2010). However, the effectiveness of wetland protection has rarely been considered. There is a great need for quantitative assessment of wetland ecosystems to aid understanding of likely global environmental change (Gong et al. 2010). Through over 140 years of practice, the establishment of nature reserves has been shown to be the best method of protecting biodiversity, and an effective measure for maintaining the ecological security of a region (Cui, 2004). Prior to 2008, there were 1.2×105 protected areas around the world, occupying about 2.1×107 km2 in total area (http://www.unepwcmc.org/wdpa/statistics, 2010), and accounting for 14% of the world’s total land area.

Studies on nature reserve classification were carried out in China during the 20th century (Shi, 1987). In 1993, the “Chinese Nature Reserves Classification and Ranking Principle” was published jointly by the former State Environmental Protection Agency and the State Administration of Technological Supervision, and this was accepted as a national standard (GB/T 14529-93) (Xue & Jiang, 1994). By consideration of the main protection roles, this document divided Chinese nature reserves into 3 categories and 9 types: the natural ecosystem class (forest, prairie and meadow, wilderness, interior wetland, and marine/coast), the wildlife class (wild animals and wild plants), and the natural vestige class (geological vestige and extinct organism vestige). Prior to 2011, China had established 614 wetland reserves (Zheng, 2010), and 91 of these are national wetland reserves.

A wetland protection network has been established, which is made up of mostly wetland reserves, along with international important wetlands, wetland parks, special marine reserves, small wetland reserves and wetland multipurpose reserves. However, no report has evaluated the effects of wetland protection, or assessed the spatial distribution of China’s wetland reserves.

6.1 methods and data sources

(1) Data sources and preparation

A. Chinese wetland maps based on satellite data. The Chinese wetland maps were developed primarily by manual interpretation (Niu et al. 2009; Gong, et al. 2010), in which the narrowest river width was 90 m and the area of the minimum map ping unit was about 9×104 m2. Four different maps at 20 m (2008), 30 m(2000, 1990), and 80 m (1978) resolutions were resampled into maps with the same cell size (240 m×240 m) for comparative purposes (Niu, 2015a).

B. Database of wetland reserves. Data for 182 wetland reserves were collected (Wu et al. 2010; Zheng et al. 2009&2010; Gao et al. 2011), the database for protection value evaluation of wetland reserves was established and distribution maps of wetland reserves were created. The websites of the nature reserve authorities were used for checking data and updating it in real time. For details see Zheng (2010). The distribution maps of wetland reserves were scanned, adjusted, and quantified based on the general plan of wetland reserves. Basic information about the reserves from scientific reports was input into an Excel spreadsheet and then loaded into a special database. Data from the “1:4000000 Chinese wetland distribution map” (State Forest Administration, 2003), the “1:4000000 Chinese river system distribution map” (https://219.238.166.215/mcp/index.asp, 2011), “the Chinese map” (Map Publishing House, 2010), “the remote sensing data of Beijing Small satellite”, and “the announcement material on the website of the reserves authorities” were used as supplementary material for checking and analysis.

 (2). Research methods and data processing

Our methods for evaluating the effects of protection were based on the protection value, the measurement of wetland changes, wildlife population proliferation, and functional zone adjustment in wetland reserves. The habitats included interior wetland reserves, marine/coast reserves, and some wild animal reserves, and the reserves of interior wetland were divided into lacustrine wetlands, riverine wetlands, and palustrine wetlands.

A.Evaluation of protection value

Zheng et al. (2010) established an indicator system to evaluate the protection values of wetland reserves; this is a quantitative evaluation of protection value of wetland reserves. Based on the scientific investigation reports of 182 wetland reserves, the protection value of national wetland reserves was evaluated by coupling the analytic hierarchy process (Zheng, 2010; Zheng et al. 1994;Yang et al. 2007), with expert consultation, and the quartile method. The wetland reserves indicator system mainly includes 5 series and 16 sets: lacustrine wetlands (permanent freshwater lakes and permanent saltwater lakes), riverine wetlands (permanent rivers, seasonal and intermittent rivers, and flood plains), palustrine wetlands (peatmire, herbaceous marshes, woody swamps, swamp meadows, pale springs, and oasis wetlands), marine/coast wetlands(mangroves, coral reefs, estuarine waters, shallow waters, seaweed deposits, intertidal wetlands), and wildlife (birds, fish, animals, amphibians, reptiles).

The indicator system of these series has 10 systematic indices: degree of endangeredness, representativeness, rarity, diversity, area suitability, human threats, tourism function, fish function, water function and land function. The system al so used 32 indicators with different weights and grade intervals (for details see Zheng (2010)); permanent fresh water lake reserves are an example.

(i) Degree of endangeredness reflects endangered species, which include the number of essential species, The number of endemic species and The population ratios of endangered species in reserves.

(ii) Representativeness reflects the ability of a wetland to represent national or basin levels; the amount of information about the homogeneity of nature reserves per unit area.

With reference to “the National Wetland Resources Investigation Technical Regulations” (State Forest Administration, 2009), continental territories were divided into 11 first order basins, 83 second order basins, and 225 third order basins. Based on the “Chinese Bay” book (The Editorial Board of Chinese Bay,1998), the coast basin was divided into 4 second order basins and 13 third order basins for research convenience. The number of wetland reserves in each basin was calculated as follows: If a reserve belonged to only 1 basin it was counted as 1; if it crossed 2 basins it was counted as 0.5; if it crossed 3 basins, it was counted as 0.33, and so on.

(iii) Rarity reflects the number of national key protected species in wetland reserves, including the number of national first order and second order species.

(iv) Diversity reflects the biodiversity conditions of the wetland reserves, including 7 indices; the number of vascular plant species, chordates, haloplankton, benthos, birds, habitat types and the population size of birds.

(v) Suitability reflects the scale and the quality of reserves, including 3 indices of the area of reserves, the proportions of core zoning and wetland areas.

(vi) Human threats reflect the intensity at which reserves were affected by human activities. Human threats included 2 indices, the population density inside reserves, and the population size of adjacent communities.

(vii) Tourism function was used as the index of annual environmental carrying capacity.

(viii) The annual catch was used to estimate fish resources in reserves.

(ix) Water quality and quantity was used to estimate water function in reserves.

(x) Cultivated area and area of reed swamp within reserves were used to estimate land function.

The protection value evaluation model used was as follows:

S=    (6-1-1)

where S is the protection value index of China’s national wetland reserves, Wi is the synthesis weight, Xi is the index value, i is the serial number of the index.

B. Evaluation of wetland changes

The speed and the acceleration of wetland changes before and after the establishment of reserves were calculated. The formula used was as follows:

R=A/A     (6-1-2)

dR=(R2R1)/|R1|    (6-1-3)

where R is the wetland change rate in the reserve, A is the total net variation value of the wetland in the reserve, A is the area of the reserve, R1 is the wetland change rate before the establishment of a reserve, R2 is the wetland change rate after the establishment of a reserve, and dR is the variation in the former two.

C. Evaluation of proliferation

Considering the specialty of wildlife reserves, the evaluation index of wetland change was replaced by an evaluation index of expanding propagation and proliferation.

The formula was as follows:

L=   (6-1-4)

where L is the rate of annual increase, N1 is the number of endangered species when the reserve was established, N2is the number after the establishment of the reserve, a1is the year when the reserve was established, a2is the year after its establishment, k is the serial number of endangered species in the reserve, and m is the number of endangered species in the reserve.

D. Evaluation of functional zoning adjustment

With the enclosure of massive areas of wetland by the Chinese government, the functional zoning of wetland reserves was considered unreasonable in some areas. To get rid of the “Nature Reserves Rule”, many local authorities changed the function to that which would give the greatest commercial benefit; this seriously reduced the protection effects. Changes to functional zoning should be minimal and that the total area of reserves should remain the same.

The function zoning adjustment index D of wetland reserves was developed to include this aspect. The formula was as follows:

D=|A2-A1|/A1      (6-1-5)

where D is the functional zoning adjustment index, A1 is the area of wetland reserve before adjustment, and A2 is the area after adjustment.

E. Determination of thresholds

The grade division of evaluation indexes and its basis in the system of protected effect evaluation indexes can be referenced to Zheng et al. (2012)

6.2 Results and discussion

(1) The basic situation of China’s national wetland reserves

As of May 2011, there were 91 national wetland reserves in mainland China, which cover 2.64×105 km2, account for 29% of China’s reserves by area. Provinces with five or more national wetland reserves are Heilongjiang, Jilin, Guangdong and Inner Mongolia (14, 8, 6 and 5 reserves, respectively).Those with 3 or 4 national wetland reserves are Hubei, Hainan, Shandong, Sichuan, Liaoning, Gansu, Anhui, Guangxi, Jiangsu, Henan, Qinghai, Tibet and Yunnan. There are presently no national wetland reserves in Beijing and Shanxi. Qinghai Province has the largest area of national wetland reserves: 1.57 × 105 km2, and 60% of the area of China’s national wetland reserves. Tibet, Heilongjiang and Inner Mongolia all have around 1 × 104– 3 × 104 km2 of national wetland reserves, whereas Tianjin, Ningxia, Hubei, Shanghai, Jiangxi, Hebei, Guizhou, Guangxi, Fujian, Zhejiang and Hainan have 0–1000 km2.

Most of China’s national wetland reserves were established between 1978 and 2000, with some established after 2000, during a promotional period for national reserves. This was probably because China joined the international Convention of Wetlands of International Importance Especially as Waterfowl Habitats” in 1992, and started a project called “Wildlife Protection and Nature Reserves Construction” in 2001. In 2006, this wetland project was included in the Eleventh Five Year Plan.

(2) Wetland changes in protected areas

The area of wetland reserve and the actual wetland area in a reserve are different. The former describes the total area of reserves, generally staying unchanged, except when being merged with another wetland, extended or adjusted. However, the actual wetland area in a reserve is often in a fluctuating state, depending on available water. In general, the area of wetland in a reserve is smaller than the area of wetland reserve.

The area of protected wetlands has shown a downward trend over the last 30 years, with a total net decrease of approximately 8152.47 km2, or 9% of China’s net decrease in wetlands. In particular, Palestrina wetlands, lacustrine wetlands and coastal wetlands are decreasing, while riverine wetlands and artificial wetlands are increasing (Table 6-2-1). Nearly half of the net change in national wetland reserves occurred in 1978–1990. This was probably due to the enclosure of wetlands for cultivation and water conservation, as communities endeavored to solve livelihood problems.

Over the last 30 years, Palestrina wetlands declined by as much as 5686.33 km2, with the maximum reduction in 1990–2000, although this was effectively controlled in 2000–2008.The rate of reduction of lacustrine wetlands was about one third of the decrease of Palestrina wetlands, with the slowest rate of reduction in 1990–2000, although this accelerated in2000–2008. Coastal wetland reduced by approximately 1/5of reduction of Palestrina wetland, with a marked reduction in 1978–1990, although this was relieved in 2000–2008.Riverine wetland increased 2574.60 km2, in 1978–1990 it increased least, in 1990–2000 it increased the most, and in2000–2008 it increased moderately. Artificial wetland increased the least; at approximately half the rate of riverine wetland, although this increased in 1978–1990 and 1990–2000, and diminished in 2000–2008.

Compared with the changes in China’s wetlands, the similar changes in protected wetlands show a declining trend in natural wetlands and an increasing trend in artificial wetlands. The decrease in natural wetlands far exceeds the increase in artificial wetlands; the former has been partly transformed into the latter, and partly into non-wetland. Another similarity is that wetlands have all suffered from changes in wetland types and large-scale degradation. In particular, the important ecological functions of biodiversity protection and ecological security disappear with the loss of natural wetlands.

In China’s coastal wetlands, the net change in national wetland reserves reaches 20% far higher than the national average of 8%. This is because coastal wetland reserves are often located in developed areas with heavy population pressure, where land use often changes as a result of government policy.

The riverine wetland in national wetland reserves has increased by 2574.60 km2, indicating that national riverine wetland reserves are in good condition. Riverine wetlands are essential to wetland reserves and the basin river system, and are the final defense in ensuring the ecological security of wetlands.

(3) Evaluation of protection

A. Protection effects

The area (79%) of poor reserves is far higher than the number (48%) of poor reserves, and both are higher than the areas or numbers of excellent and moderate reserves. Poor reserves thus occupy the leading status (Table 6-2-2).Palestrina reserves are the main type of poor reserves; the numbers of poor lacustrine, marine/coast, and wildlife reserves are only 2%, 3% and 16% respectively. Lacustrine reserves form the majority of the excellent reserves.

B. Analysis of different types of reserves

Of the 20 marine/coast reserves, 11 are poor with a total area of 7.06×103 km2, and an area ratio of 70%. Number and area ratios of excellent and moderate reserves are both 15%. Three reserves in Hainan were not evaluated due to the lack of data. Marine/coast reserves are generally in poor condition. In the three riverine reserves, Henan Dan River is excellent with an area of 640.20 km2, Henan Yellow River is moderate with an area of 680.00km2, and Xinxiang Yellow River is poor with an area of 227.80 km2. Of the 16 lacustrine reserves, 7 are poor and 6 are excellent, with area ratios of 12% and 68% respectively. Of the 28 Palestrina reserves, 13 are poor with a large total area of 1.65×105 km2, and 88% of the area. This shows that the protection for Palestrina reserves is poor as a whole. Of the 24 wildlife reserves, 12 are poor with a total area of 3.33×104 km2, and 91% of the area. This means that wildlife reserves are also generally poor. The palustrine, wildlife, and marine/coast reserves are generally in poor condition, and lacustrine and riverine reserves are, by comparison, in better condition.

C. Analysis of different evaluation indices

Table 6-2-3 shows considerable differences between the protection effect evaluation indices, and thus supports the reliability of evaluation results. Of the 44 poor reserves, 21 are poor in protection value, 24 are poor in wetland changes evaluation, 9 are poor inproliferation, and 5 are poor in functional zoning adjustment.

Of the 16 moderate reserves, 12 have protection value slower than 71. The wetlands changes in Guangxi Shankou, the Inner Mongolia Ordos and Qinghai Lake reserves were slightly lower than the threshold of 10%. In addition, Henan Yellow River, Yangtze River Fish and Yalu River Estuary were places in which functional zoning adjustment were undertaken; the former two staying unchanged in area, while the area of the latter declined by 7%. The 19 excellent reserves had high evaluation results for both protection value and proliferation, without short-sighted functional zoning adjustment causing protection to be replaced by development. The moderate reserves are special and have great potential

for protection. The wetland areas in these reserves often show an increase, and are even higher than that in the excellent reserves, however they often have relatively small areas and low biodiversity, as well as functional zoning adjustment for economic development.

Of the poor reserves, the largest is the Three Rivers Source with an area of 1.52×105 km2, and the smallest is Guangdong Fustian with 10 km2. Of the moderate reserves, the largest is Qinghai Lake with approximately 5.00 ×103 km2, and the smallest is Hubei Shishou with 20 km2. Of the excellent reserves, the largest reserve is Dalai Lake with7.4×103 km2, and the smallest is Hainan Dongzhaigang with 30 km2.

Of the excellent reserves, the Heilongjiang Xingkai Lake has the greatest protection value of 84, and the Heilongjiang Naoli River has the lowest protection value of 71. Of the moderate reserves, the Yangtze River Fish has the greatest protection value of 89, and the Inner Mongolia Erdos has the lowest protection value of 65. Of the poor reserves, the Three Rivers Source has the greatest protection value of 93, and the Guangxi Beilun Estuary has the lowest protection value of 50.

Table 6-2-3 Comparative analysis of different evaluation indices of China’s national wetland reserves

Of the excellent reserves, the Gansu Dunhuang Xihu shows the greatest acceleration in wetland change rate with a two-fold increase, and Inner Mongolia Ke’er’qin shows the smallest acceleration in wetland change rate with a quarter-fold increase. Of the moderate reserves, the Henan Yellow River shows the greatest acceleration in wetland change rate with a nine-fold increase, and Guangxi Shankou shows the smallest acceleration of wetland change rate with a 9% decrease. Of the poor reserves, Dalian Phoca largha also shows a nine-fold increase and Gansu Gahai-zecha shows a seven-fold decrease.

D. Analysis of different regions

(i) At the regional level. The poor reserves are widely distributed throughout each wetland region, but mainly in the wetland regions of the Yangtze River, the Coast, the Three Rivers Source, and Southwest China. In particular, the poor reserves are found in the wetland subregions of the Lower Reaches of the Jinsha River, the Chaochuwan Reaches of the Yangtze River, Poyang Lake, Hubei, and the Tianjin Coast, the North Coast of Shandong Peninsula, the Jiangsu Coast, the South Coast of Fujian, Lazi to the Paixiang Reaches of the Yarlung Zangbo River, and the Lower Reaches of the Lancang River Bijiang Estuary. In addition, they are scattered through the wetland subregions of the Tangshan Mountains North Foothills, the Gansu Corridor, the Lower Reaches of the Nen River, and the Tongnan and Chongming Island Reaches of the Yangtze River (Figure6-2-1).

At present, there are no moderate reserves in the wetland regions of the Three Rivers Source and Southwest China. They occur scattered in various subregions, such as the Yalu River, Sanmenxia to the Huayuankou Reaches of the Yellow River, Coast of Southern Zhejiang, Qinghai Lake, Yibing to Yichang Reaches of the Yangtze River, the Chaochuwan Reaches of the Yangtze River, and the lower reaches of the Yangtze River.

The excellent reserves are mainly distributed in the wetland region of the Songhua River, especially in the subregions of the Nen River, the Wusuli River, the Mudan River and the E’er’guna River. There are also excellent reserves in the wetland subregion of the Main Liao River, for example, the Shuangtai Estuary reserve in Liaoning. The excellent reserves in the wetland region of the Yangtze River exist in the subregions of the Along River and Han River. There are no excellent reserves in the wetlands of the Three Rivers Source or Southwest China. Attention should be focused on those subregions without National wetland reserves, because there are possibly protection gaps here. These subregions mainly include the Suifen River and the Tumen River in the wetland regions of the Songhua River; Mintuo River, Wu River and Tai Lake in the wetland region of the Yangtze River; South and North of the Pan River, Hongliu River, Yu River, West River, North River, East River, Zhu River Delta; Han River and Eastern Guangdong in the wetland region of the Zhu River; Hekou Town to Dragon Gate of the Yellow River, Upstream of Huai River, Mushusi River, North Hai River, Tuhaimajia River, Luan River and the Eastern Hebei Coast, Eastern Liao River, Western Liao River, and Huntai River in the wetland region of Huanghuaihailiao; Chaidamu Basin, Tuhabasin, Altai Mountains South Foothills, the inland river of Central and Western Asia, Gurban Tongut Desert, the source of the Tarim River, Kunlun Mountains North Foothills, the MainTarim River and Tarim Basin desert in the wetland region of Northwest China; the Red River, Nu River and Irawaddy River, Southern Tibet River and Western Tibet River in the wetland region of Southwest China; and the Qiantang River, Eastern Zhejiang River, Sorthern Zhejiang River, Eastern Fujiang River, and Min River in the wetland region of Southeast China.

There is a common characteristic of the wetland subregions of the Western Guangdong Coast and the Wusuli River. Both have excellent, moderate, and poor reserves, such as the Guangdong Zhanjiang mangroves, the Heilongjiang Xingkai Lake, and the Naoli River. There are excellent reserves in these two wetland subregions even though populations are dense, resulting in strong pressure on the ecological conditions. This is because of positive intervention measures taken in these areas, as previously discussed.

(ii) At the provincial level. There are poor reserves in each province except for Guizhou, Hainan and Zhejiang. The excellent reserves are mainly distributed in Heilongjiang, Jilin, Jiangsu and Inner Mongolia; each of these provinces has two excellent reserves. These provinces set an example to those provinces without excellent reserves, such as Anhui, Fujian, Guangxi, Huizhou, Hubei, Hubei, Jiangxi, Ningxia, Qinghai, Shandong, Shanghai, Tianjin, Yunnan and Zhejiang. The reserves in Hunan, Gansu, Heilongjiang, Inner Mongolia, Sichuan, Xinjiang and Henan are in good condition, and the number ratios of the excellent reserves in these provinces are all higher than 50%. The reserves in Fujian, Shanxi, Hubei, Jiangxi, Ningxia, Tianjin, Tibet and Yunnan are all in poor condition. The reserves in Qinghai,Liaoning, Shandong, Anhui, Jiangsu, Hubei and Shanghai are partly in poor condition, and the number ratios of the poor reserves in these provinces are also higher than 50%.

7. Conclusion

Although wetland thematic mapping is not a new topic, it is usually performed at local scale than at global scale. Different from local wetland monitoring, large scale wetland mapping and monitoring can serve the government and policy-make with information on wetland status, changes and trends in long terms. This not only requires knowledge on wetlands, but also data processing techniques of satellite images. However, knowledge on both aspects are limited for large scale wetland mapping since most experiences of local wetland mapping cannot be directly learned or transferred to regional and global scales. Particularly, the existing wetland maps at global scale are highly inconsistent. Therefore, more research is needed to conduct global thematic wetland mapping, to achieve more accurate wetland classes and more detailed ecological and environmental information on wetland ecosystems.

Cite This Work

To export a reference to this article please select a referencing stye below:

Reference Copied to Clipboard.
Reference Copied to Clipboard.
Reference Copied to Clipboard.
Reference Copied to Clipboard.
Reference Copied to Clipboard.
Reference Copied to Clipboard.
Reference Copied to Clipboard.

Related Services

View all

DMCA / Removal Request

If you are the original writer of this essay and no longer wish to have the essay published on the UK Essays website then please:

McAfee SECURE sites help keep you safe from identity theft, credit card fraud, spyware, spam, viruses and online scams Prices from
£29

Undergraduate 2:2 • 250 words • 7 day delivery

Order now

Delivered on-time or your money back

Rated 4.1 out of 5 by
Reviews.co.uk Logo (25 Reviews)

Get help with your dissertation