Evolution Characteristics and Driving Forces of Wetland Changes in the Poyang Lake Eco-economic Zone of China

Wetland ecosystem is known as the “kidney” of earth and the “gene pool” of species. It has the functions of regulating climate, flood storage and degradation of pollution. In this paper, based on GIS technology and landscape ecology, wetland changes and its driving forces in the Poyang Lake Eco-economic Zone of China are analyzed. The analysis of landscape pattern demonstrates that there is an increase in the degree of fragmentation of wetlands in the study area. At the same time, the overall aggregation degree of the lake is in the rise. The increased Perimeter-area fractal dimension indicates that the shape of wetland becomes more and more rules. The main driving forces of wetland changes in the Poyang Lake Eco-economic Zone include natural factors and human activities. This study also indicates the main natural factor is the changes of precipitation meanwhile the increase of average temperature. Meanwhile, the rapid population growth, regional economic development and other human activities are also the key driving forces of wetland landscape changes in the Poyang Lake Eco-economic Zone.


Introduction
Land use/land cover change (LUCC) in the field of global environmental change research has been gained increasing degree of attention because of its role in the social and ecological environment (Vitousek et al., 1997;Li, 1996). As one of the important land types, wetlands are also increasingly widespread concern by many scholars Nagabhatla et al., 2012;Scott et al., 2012). Wetland ecosystem, which is known as the "kidney" of earth and the "gene pool" of species, has become one of the most productive ecosystems on the planet (Whigham, 1999;Cserhalmi et al.,2011). Wetland not only is a valuable natural resource for human survival but also is one of the most important environments (Kingsford, 2011). It not only directly provides the raw material for the production and human life and also some functions of regulating climate, flood storage and control the pollution and degradation of pollution and other environmental function (Traill et al., 2010). The analysis of the driving force of the various wetland changes is one of priorities and focuses of wetlands LUCC research ( Urbanization is an irresistible trend of the development of human society in the 21 st century. Most of arable lands have been converted to construction land by urban expansion in China. This made a large number of ecological land including wetlands, which plays a pivotal role in ecological service function, exploited to arable land for meeting the object of arable land protection. The newly formed wetlands in western China were caused primarily by climate warming over that region whereas the newly created artificial wetlands were caused by economic developments (Gong et al., 2010). Therefore, it is meaningful to explore the mechanisms of wetlands evolution.
Poyang Lake Region of China is recognized as one of the fundamental ecological function districts by the World Wide Fund for Nature (WWF). It plays a vital role in the provision of fresh water resources, the maintenance of the regional water balance, the homogenization of the flood, the regulation of regional climate and the conservation of biological resources (Deng et al., 2011;Yan et al., 2013). In recent years, some activities including reclamation of land from the lake and agricultural development have made wetland of Poyang Lake region change dramatically, which brought increasing obvious ecological problems Feng et al., 2012;Shankman and Liang, 2003). In view of the high service value of wetland and the vast eco-environment effect of LUCC, it is necessary to study process and mechanism of wetland in the Poyang Lake Eco-economic Zone . The Mountain-River-Lake Program (MRL) was implemented since 25 years ago in the Poyang Lake basin, southern China. It consists of series of forest restoration projects that aim to address severe soil and water losses, and improve farmer's livelihoods . Therefore, the study of changes in wetlands in the Poyang Lake Region of China becomes increasingly urgent. As we all known, the study of wetland landscape pattern can better understand the ecological processes. Exploring the change of natural wetland landscape pattern over time and revealing its driving forces are an urgent need to study the issue for the Poyang Lake Yan et al., 2013).
The main purposes of this study are: 1) how to study the characteristics of wetland changes based on the theories and methods of landscape ecology; 2) to explore the evolution pattern of different kinds of wetlands; 3) and to find the driving forces of wetland changes in the Poyang Lake Eco-economic Zone for Sustainable Watershed Management.

Study Area
The study area (28°30′N -30°06′N, 114°29′E -117°25′E) is located in Jiangxi Province, a southern region of China, with a surface of approximately 51,200 km 2 ( Figure 1). The area belongs to the subtropical humid climate zone, with an annual average temperature of 16～18°C and an annual average rainfall of 1,600 mm. Annual average sunshine is about 1,473.3~2,077.5 hours. Annual sunshine total radiation is about 97~114.5 Kcal/cm 2 . Soils are predominantly red soil, yellow soil and paddy soil. Poyang Lake is the largest freshwater lake in China and is one of the six wetlands with rich biodiversity in the World. Taking Poyang Lake as the core and relying on the Poyang Lake city circle, the Poyang Lake Eco-economic Region is the significant economic zone for protecting the ecology and developing economics. The study area includes 38 counties and has a population of 20.06 million and GDP of 3,948. 17 billion Yuan (RMB) in 2008. One of goals of the study area is to build an international demonstration zone for the harmonious ecological and economic development.

Data
Land use data of 1990, 2000 and 2005 employed in this study came from the 1:100,000 national land use database of the Data Center for Resources and Environmental Sciences, Chinese Academy of Sciences (RESDC). Wetland types in this study divided into three classes and 11 subclasses (see Table 1). Based on the ArcGIS9.3 software, land use data resampled at the spatial resolution of 100 m × 100 m. Climatic data came from the China Meteorological Data Sharing Service System. Social-economical data at the county level in this study derived from the Jiangxi province statistics yearbook from 1985 to 2006.

Land dynamic degree
Land dynamic degree is to measure the number of changes in the situation for some time within a certain land use types. The formula is given as: Where K is the dynamic degree of a certain land types within the study period; Ua and Ub respectively represent the area of land use types in the beginning and at the end of the study; T is the length of the study period. When T is set for the year, the value of K is the average annual rate of the area change of a certain land type.

Transfer matrix of land use change
On the basis of the transfer matrix of land use types, transition probability matrix of land use types is established to describe the changes in the intensity of land use types. The formula is given as: (2) Where Dij is the transition probability of land use type i converted into land use type j in the study period; Si is the total area land use type i of the beginning of the study; dSi-j is the sum of the areas of land use type i converted into the land use type j in the study period; n is the number of land use types changed in the study area.

Conversion contribution ratios
The transfer matrix method describes the evolution of different land use types. In order to fully reflect the status and role of information of different land types in the land use pattern, the method of conversion contribution ratios of transfer to/from the land use types is conducted in this study. It can comparatively analyze of spatial pattern and quantity characteristics of the transfer-in and transfer-out of the various types of land use. The formula of conversion contribution ratios of transfer-in of land use types is given as: Where Lji means the proportion of the area of other kinds of land use types except i converted into land use type i accounting for the total area transferred; Sji refers to the transfer area of the land use type j converted to land use type i; St is the total area of land use type transferred; n is the number of land use types. Lji can be used to compare the area differences of increment allocated for the various kinds of land in the transfer-in process of land dynamic change.
The formula of conversion contribution ratios of transfer-out of land use types is given as: Where L0j means the proportion of the area of land use type i converted into other kinds of land use types accounting for the total area transferred; Sij refers to the transfer area of the land use type i converted to land use type j; St is the total area of land use type transferred; n is the number of land use types. L0j can be used to compare the area differences of decrement allocated for the various kinds of land in the transfer-out process of land dynamic change.

Landscape pattern analysis
Landscape ecology can provide new theories and methods for a comprehensive solution to the resource and environmental problems and to carry out a detailed ecological environment construction. In this study, we selected seven landscape indices to reflect the characteristics of the wetlands changes at landscape and class level.
Number of patches of a particular patch type is a basic measure of the extent of subdivision or fragmentation of the patch type. The formula of Number of Patches (NP) is given as: i n NP  (5) Where NP represents the Number of Patches; ni is the number of patches in the landscape of patch type (class) i.
Patch density has the same basic utility as the number of patches as an index, except that it expresses the number of patches on a per unit area basis that facilitates comparisons among landscapes of varying size. The formula of Patch Density (PD) is given as: Where PD represents the patch density; ni is the number of patches in the landscape of patch type (class) i; A is the total landscape area.
Largest Patch Index at the class level quantifies the percentage of total landscape area comprised by the largest patch. As such, it is a basic measure of dominance. The formula of Largest Patch Index (LPI) is given as: Where LPI represents the Largest Patch Index; ai is the area of patch (class); A is the total landscape area.
Perimeter-area fractal dimension is appealing because it reflects shape complexity across a range of spatial scales (patch sizes). A fractal dimension greater than 1 for a 2-dimensional landscape mosaic indicates a departure from a Euclidean geometry (i.e., An increase in patch shape complexity). The formula of Perimeter-area fractal dimension (PAFRAC) is given as: (8) Where PAFRAC represents the Perimeter-area fractal dimension; A is the area of patch (class); P is the perimeter of patch (class).
Landscape Division is based on the cumulative patch area distribution and is interpreted as the probability that two randomly chosen pixels in the landscape are not situated in the same patch. The formula of Landscape Division Index is given as: Where DIVISION represents the Landscape Division Index x; ai is the area of patch (class); A is the total landscape area.
Shannon's Diversity Index is a popular measure of diversity in community ecology, applied here to landscapes. The formula of Shannon's Diversity Index (SHDI) is given as: Where SHDI represents the Shannon's Diversity Index; pi is the proportion of the landscape occupied by patch type (class) i; Aggregation index is calculated from an adjacency matrix, which shows the frequency with which different pairs of patch types (including like adjacencies between the same patch type appear side-by-side on the map. The formula of Aggregation Index (AI) is given as: ) 100 ]( max [ ii ii g g AI   (11) Where AI represents the Aggregation Index; gii is the number of like adjacencies (joins) between pixels of patch type (class) i based on the single-count method.

Overall analysis of wetland dynamic changes
The areas and its changes of different kinds of wetland during the period 1990-2005 in the Poyang Lake Eco-economic Zone are listed in the Table 2. From Table 2, we can see that there was an increasing trend of natural wetland and the artificial wetland in the study area showed a decreasing trend from 1990 to 2005. For natural wetland, lake increased from 200, 523 hm 2 in 1990 to 284, 300 hm 2 in 2005, an increase of 41. 78%. As can be seen from the Figure 2, there is a rapid growth of the construction land. Figure 2 also shows that the increase in the lake is more obvious from 1990 to 2005 due to the conversion of a large area of the lakeshore.  26.24 Other non-wetland 612 632 632 20 3.27    The transfer matrix of different types of wetlands from 1990 to 2005 is listed in the Table 3. Many lakes, rivers, lakeshores, reservoir and ponds transformed into paddy fields, the area respectively 2955hm 2 , 189 hm 2 and 3072 hm 2 (see Table 3). This means that many natural wetlands were replaced by the artificial wetlands. As can be seen from the Table 3, 34% of the paddy fields were pushed back the natural wetlands, and 25% of the paddy fields were occupied by the constructed land during the period 1990-2005 in the Poyang Lake Eco-economic Zone. At the same time, there is a greater conversion between the different types of wetlands. The main types mainly transferred to the lakes are the beaches, marshes and rivers. 8 The conversion contribution ratio of different kinds of wetlands from 1990 to 2005 is listed in the Table 4. Compared with the contribution ratios of other transfer-in landscape components, the contribution ratios of transferring into the lake are the largest, 62.18% during the period 2000-2005 (see Table 4). While the contribution ratio of transfer-in is less than transfer-out during the period 1990-2000, but in general, the contribution ratio of transferring into the lake is greater than transferring out the lake. This is because there are a large number of lakes have been reclaimed into the paddy fields during the period 1990-2000.  4.91 The concern is that the contribution rate of transferring into or out the landscape components directly or indirectly controlled by the change of the lake in the Poyang Lake Eco-economic Zone. During the period 1990-2000, the conversion contribution ratio of transferring into the lakeshore is the maximum, 42.11% because large areas of ponds and lake converted into lakeshore. During the period 1990-2005, the conversion contribution ratio of transferring into the lake is the maximum, 51.22% due to the large number of the lakeshore and swamp converted into lake. During the period 1990-2000, the conversion contribution ratio of transferring out the paddy field is the maximum, 25.14% due to the large areas of paddy field converted into constructed land. Because of the large areas of lakeshore converted into the lake, the conversion contribution ratio of transferring out the lakeshore is the maximum, 37.13% during the period 2000-2500and 30.8% during the period 1990-2000. This is mainly because of the increase in rainfall during this period.

Pattern change of wetland landscape
The change of landscape indices of whole wetland in the Poyang Lake Eco-economic Zone from 1990 to 2005 is listed in Table 5. As can be seen from Table 5, the patch number of wetland landscape increased from 15974 in 1990 to 15988 in 2005, which showed an increasing trend in the Poyang Lake Eco-economic Zone. At the same time, Landscape Division Index (DIVISION) showed a downward trend from 1990 to 2005. It means that the separation degree of wetland increased. According to the changes of Number of Patches (NP) and Landscape Division Index (DIVISION), we can infer that there is an increase in the degree of fragmentation of wetlands in the study area. Average fractal dimension index means the self-similar degree of the patch, to some extent, and it can reflect the impact degree of human activities on the patch. The Perimeter-area fractal dimension increases from 1.5132 in 1990 to 1.5216 in 2005 (see Table 5). The increase of average fractal dimension means that the shape similarity of wetland landscape patch increases and the shape become more and more rules. This is mainly because of the large number of marsh wetlands reclamation become more regular paddy fields.
From Table 5, we can see that Shannon's Diversity Index (SHDI) decreased from 1.0361 in 1990 to 0.9942 in 2005. Shannon's Diversity Index (SHDI) can reflect the landscape heterogeneity and is extremely sensitive to the non-equilibrium distribution of each patch type in the landscape. It emphasizes the contribution of rare patch types of information. The decrease of Shannon's Diversity Index shows that wetland type in the regional landscape is more monotonous, and the contribution of information of rare patch types reduced.
The result of landscape indices of different kinds of wetlands in 1990 and 2005 is listed in Table 6. In respect of the lake wetland, as can be seen from Table 6, the number patches of the lake are in decline, and the largest patch index shows an upward trend from 1990 to 2005. The largest patch index (LPI) can reflect the effect degree of the maximum patch on the entire landscape. The increase in the largest patch index of lake indicated that the degree of Poyang Lake landscape controlling the whole wetland showed an enhanced trend. Meanwhile, Aggregation Index (AI) of lake increased from 92.7145 in 1990 to 94.6958 in 2005. It shows that the overall aggregation degree of wetland is in the rise.
As for the swap wetland, the largest patch index shows a downward trend from 1990 to 2005. Meanwhile, the Perimeter-area fractal dimension decreased from 1.4568 in 1990 to 1.4296 in 2005 (see Table 6). It indicates that the large patch of wetlands is fragmented and becomes increasingly irregular. The increase in the number of patches and reduces patch density of swamp also supports this conclusion.
As for the reservoir and pond, the patch number and largest patch index shows a downward trend from 1990 to 2005. Meanwhile, the Aggregation Index (AI) decreased from 72.37 in 1990 to 71.84 in 2005. The Perimeter-area fractal dimension increased from 1.4297 in 1990 to 1.4415 in 2005 (see Table 6). This means that pond wetland is fragmented and became increasingly regular by the human disturbance.
As can be seen from Table 6, for all wetlands, the patch number and patch density of paddy field is largest. This means that the paddy field is the largest wetland type interfered by human. Aggregation degree of the lake is highest, which means that the connectivity of the lake is the best. Overall, the patch number of artificial wetlands is greater than natural wetlands. Simultaneously, the patch density and separation degree of artificial wetlands is greater than natural wetlands. This is mainly a result of manual interference. Table 6 shows that the connectivity of natural wetlands is higher than artificial wetlands.

Driving forces of wetland change
The main driving factors of wetland changes in Poyang Lake Eco-economic Zone include natural factors and human activities. Natural factors usually include climate, geology, geomorphology, hydrology, vegetation, soil, and so on. Human activities are mainly reflected in the demographic, economic, and policy aspects. Natural factors often have a role in the landscape at the larger spatial and temporal scales. In other words, environmental backgrounds control the main changes of wetland. Meanwhile, factors of human activities are the main driving force of the dynamic changes of the wetland at a shorter time scale. Wetland is a special type in the watershed landscape. Water is the fulcrum to maintain its ecological structure and function and spatial characteristics of landscape and is the main carrier of material flow, energy flow and information flow within the wetlands and other land type. Water environment is the motivating factors directly promoting the formation and evolution of wetland.
Atmospheric rainfall becomes main replenishment water for the wetland. How much is precipitation directly impacts on the regional wetland area. Figure 4 shows the change of annual mean precipitation in Poyang Lake Region from 1981 to 2010. According to Table 2, lake decreased from 200, 523 hm 2 in 1990 to 194,575 hm 2 in 2000, then increased 284, 300 hm 2 in 2005. At the same time, annual precipitation decreased from 1543mm in 1990 to 1455mm in 2000, then1472mm increased in 2005 (see Figure 4). There is a strong correlation between the average annual rainfall and the area of Poyang Lake. This is mainly because a lot of lakeshore land turned into lakes when annual precipitation is abundant. In addition to the effects of precipitation, temperature changes are also critical factors that affect the wetland landscape changes. Temperature not only affects the vegetation growth status and biomass, but also it affects the process of evaporation, intensity of surface and surface evaporation. Figure 5 shows the change of annual mean temperature in the Poyang Lake Region from 1981 to 2010. As can be seen from figure 5, in general, the annual average temperature in the Poyang Lake Eco-economic Zone shows a rising trend. From figure 3, we can see that the average annual rainfall of Poyang Lake Ecological Economic Zone shows an overall decreasing trend. Reduced rainfall will reduce the water supply of upstream river runoff on wetlands, reduce soil moisture, exacerbates the drought level, eventually leading to the degradation of the swamp. The rise in temperature will increase the surface evaporation, affecting the wetland area. Meanwhile, some rivers shrinking dry and the many bubble marsh shrink or disappear due to the reduction of water. Therefore, the changes of annual rainfall and temperature become the main driving forces of natural wetlands change.
With the rapid economic development, over-exploitation of wetland resources is the main reason for the loss of a large area of wetlands in the Poyang Lake Eco-economic Zone. This is mainly because the economic development and population growth will inevitably lead to the rapid increase of the construction land and other non-wetlands, which occupied more natural wetlands, especially the beaches.
According to the analysis related statistics, during the period 1990-2005, the non-agricultural population of the Poyang Lake Eco-economic Zone is growing rapidly from 3,591,700 in 1990 to 5, 678, 500 in 2005, an average annual growth of 139, 100. Accordingly, the proportion of the non-agricultural population increased from 21.3% in 1990 to 28.3% in 2005. The proportion of the non-agricultural population growth reflects a trend of urbanization and industrialization. Therefore, industrialization and urbanization are also extremely crucial driving forces of wetland changes in the Poyang Lake Eco-economic Zone. Industrialization and urbanization in the study area has made the land use of non-farm through the population concentration, industry concentration and the geographical spread of occupied land. As the weak link in the land use structure, wetland has become the most obvious type of land use change in the non-agricultural conversion process. In the past 15 years, a large number of ponds and paddy fields occupied for construction land in the study area, which can be seen from Table 3. Therefore, a large number of natural wetlands have gradually reclaimed to the artificial wetlands or artificial landscape in the interference of human activities for economic benefits. It is mainly in the agricultural development activities such as paddy field development and urban construction such as transportation and urban settlements.
Some policies have a profound impact on the wetlands change of Poyang Lake Eco-economic Zone, especially in playing a key role in the process of protecting their reasonable evolution. Since the late 1980s, the study area is in the implementation of the Mountain River Lake Development Program, which is a watershed integrated management program for the sustainable development of Poyang Lake watershed. The object of the project is to achieve the harmonious development of the economy, society and environment. It has formed a governance guideline "Governing lakes must first regulating the rivers, regulating the rivers must first managing mountains, managing the mountains must reduce poverty ". The implementation of the project has protected the water quality of Poyang Lake and reduced the degradation of wetlands.
Other policies like "cultivated land balance" has negative on the wetland changes of Poyang Lake Eco-economic Zone. In the context of increasing demand for construction land, swamp and pond are facing the threat of agricultural development because of meeting the requirements of "cultivated land balance". Implementation of forest restoration projects and improve the livelihoods farmers are important measures to protect wetlands in the Poyang Lake Eco-economic Zone. How to use quantified methods, such as correlation analysis, factor analysis, and so on, to explore the driving forces of wetland changes is important for future research.

Conclusions
There is an increasing trend of natural wetland and artificial wetland in the study area shows a decreasing trend from 1990 to 2005. The value of dynamic degree of the lake wetland is largest, 9.22 during the period 2000-2005.
The main types mainly transferred to the lakes are the beaches, marshes and rivers during the period1990-2005.
The contribution ratio of transferring into the lake is the largest, 62.18% during the period 2000-2005. Overall, the contribution ratio of transferring into the lake is greater than transferring out the lake from 1990 to 2005. During the period 1990-2005, the conversion contribution ratio of transferring into the lake is the maximum, 51.22% due to the large number of the lake and swamp converted into lake. 25% of the paddy fields were occupied by the constructed land during the period 1990-2005 in the Poyang Lake Eco-economic Zone.
The analysis of landscape pattern indicates that there is an increase in the degree of fragmentation of wetlands in the study area. The Perimeter-area fractal dimension increased from 1.5132 in 1990 to 1.5216 in 2005, which indicates the shape of wetland becomes more and more rules. The Aggregation Index (AI) of lake increases from 92.7145 in 1990 to 94.6958 in 2005, which means that the overall aggregation degree of the lake is in the rise. The aggregation degree of the lake is highest, which means that the connectivity of the lake is the best.
The change of temperature and precipitation of the study area has a significant impact on wetland changes. The rapid population growth, regional economic development and other human activities are also the key driving forces of wetland landscape changes in the Poyang Lake Eco-economic Zone.
The above conclusions in this study provide the basis for the sustainable management and decision-making of the Poyang Lake Watershed. Cyanobacterial blooms especially Microcysits blooms are rampant in eutrophicated freshwater lakes, which bring the crisis for the human and ecological health. However, understanding the mechanism of Microcystis colonies formation and controlling blooms development still need a lot to explore and research. A lot of studies have showed that predation by zooplankton [1,2] , Microcystin [3] , and nutrients [4] may induce unicellular Microcystis 15 cells to form colony to survive. During the process of form tranformation, the surface characterization of cells changed firstly and could give us a direct information. However, the surface characterization of algal cells have received little attention.

鄱阳湖生态经济区湿地景观演变特征及其影响因素分析
As we know, algal cells excret extracellular polymer substances to promote the growth or to protect themselves. These substances contained active functional groups such as hydroxyl, amine, and carboxylic groups, which play a major role in surface binding capacity, biomineralisation, and adhesion [5,6] . The adhesion mechianism was generally dominated by the electrostatic interactions or by the hyrophobicity interactions on the surface, which could be expressed by the zeta potential and the hydrophobicity [7] . Hydrophobicity was higher, which was better for the aggregation or the colony formation [8] . Additionally, zeta potential was consider to be the useful tool to estimate the efficiency of removement during the water treatment [9] . So considering the Microcystis blooms are serious and harmful, understanding the surface characterization of Microcystis cells was essential in explaining the mechanism of colony formation and controlling the blooms. In present study, through combining the field investigation and culture experiment, we investigated the cell surface charaterisation (zeta potential and hydrophobicity) of Microcystis spp. in eutrophic Lake Taihu.
2 Materials and methods：

Field samples collection
The field cyanobacteria were collected in eutrophic Taihu lake ecosystem research station (TLLER) in Meiliang Bay during the period from April to October in 2012. During this period, the cyanobacteria samples were collected using plankton net at a fixed site in the station. Then the collected samples were immediately sent to Nanjing Laboratory for the zeta potential measurement. For the hydrophobicity determination, the collected filed cyanobacteria in August were seperated into five groups according to the colony sizes, and they are < 20  m, 20  100  m, 100  200  m, 200  400  m and > 400  m, respectively. The hydrophobicity of cell surface corresponding to these different sizes were measured.

Experimental culture
Microcystis spp. including Microcystis aeruginosa (FACHB-1214) and Microcystis flos-aquae (FACHB-1028) in the pure culture were obtained from the institute of hydrobiology, Chinese Academy of Sciences, China. The cells were cultured in BG11 medium, but with 10 times of carbon in BG11, 1/50 times of nitrogen and phosphorus in BG11. Cultrues were grown at 25 o C, under 39 mol photons/m 2 /s irradiance and 12:12 h (light/dark) cycle in the constant temperature incubator. The whole incubation time lasted for 30 days.

Zeta potential analysis
The charge surface of cells were assessed by a computer with software based on the Helmholtz-Smoluchowski equation [10] . The zeta potential was detemied using a JS94G+ microelectrophoresis apparatus made in China [11] . Cultured algal suspension was harvested directly in triplicate at low speed centrifugalization and resuspended in 10 ml of 0.1 mol/L NaNO3. For field algae, disintegration of Microcystis colonies was done by ultrasonic treatment before centrifugalization. The pH of the washed cell suspensions was adjusted from 3 to 10 by HNO3 or NaOH (0.1 M) addition. The zeta potential was evaluated at room temperature in the electrophoresis cell. For each sample, triplicate cultures were taken for measurement and for each data, approximately, 10 readings were done. The average values were reported in this paper. 16

Influence of pretreatment of cells on hydrophobicity
In order to estimate the effect of common pretreatment on hydrophobicity, three ways were included as follows: (1) filtration only; (2) centrifugation only; (3) centrifugation after formalin fixation. Considering the collected field algal colony were floating on the water surface, the low frequency ultrosound was used to make the cells suspend [12,13] .

Hydrophobicity analysis
The hydrophobicity of the cell surface was evaluated by xylene-water two-phase system [14] . Field cyanobacterial cells were sonicated firstly and harvested by centrifugation, and then resuspended in 0.2 M sodium hydrogen phosphate buffer (pH=7.0). The concentrations of algal cells counted by Microscope were N0. A volume of 0.5 mL of xylene was added to 4 mL of cell suspension. The two-phase system was vortexed for 30 s and settled for 15 min. The alagal cells in aqueous phase were counted as Nt. The percentage of hydrophobicity (H) was calculated by H=(N0-Nt)/N0)  100%, where N0 is the cells numbers after settlement and Nt is the cells before settlement.

statistical analysis
One-way ANOVA analysis was used between treatment groups, and the difference between them was analized by Tukey HSD comparison.
3 Results and discussion:   The zeta potential is the potential at the slid surface of a colloid electrical double layer, and it is closely related to the state of charge on the surface of particles. The surface zeta potential of algal cells could affect the physiological status and nutrient absorption. Recently, the cyanobacterial blooms was ranpant and serious, the algal zeta potential was also an important affecting factor in engineering treatment [15,16] . When the surface charge of algal cells was positive, the zeta potential was positive. When the surface charge of algal cells was negative, the zeta potential was negative. The results showed that the zeta potential values on the surface of the cultured and field cyanobacterial cells were both negative, and decreased by increasing pH values. As Fig. 1 and Fig. 4 shown, the zeta potential of cyanobacterial cells changed by time. The zeta potential of field cyanobacteria at pH 7.0 were in the range of -8.3  -20.9 mV. Among them, the zeta potential of cells in June was similar to the investigated value (-18 mV) of field cyanobacteria during the blooming period [17] . The zeta potential values of cultured pure Microcystis cells at pH 7.0 ranged from -13.1 to -21.0 mV. The values on the cultured cells surface during the exponential phase ( Fig. 4 and Fig. 5) were similar to the reported values by Hadjoudja et al [18] . By comparison to the early stage of incubation, the negative charge of cells surface at day 30 was decreased, which meant that the zeta potential was increased. This may be due to the changes of the composition and contents of bound extracellular polymeric substances (bEPS) [19] . 19 The zeta potentials of algal cells changed by pH values (3  10) were shown in Fig. 2 and Fig. 5. The results showed that, the zeta potentials of field and cultured Microcystis cells were both negative, suggesting that the surface charge of Microcystis cells was negative. The zeta potential value at pH 3.0 was still negative, indicating that the isoelectric point was less than 3.0. This result was in aggrement with the reported value of Microcystis [16][17][18] . According to the report by Henderson et al [16] , the isoelectric point of EPS from Microcystis aeruginosa during the exponential phase and stationary phase was about 2. This result showed that the surface characterization of cells could be reflected by the EPS. During the whole culture process, the zeta potential increased by decreasing pH values. When pH value increased from 3 to 5, the zeta potential values decreased rapidly, which may be caused by deprotonation of carboxyl functional groups in the EPS [18,20] .
As shown in Fig.1, the absolute values of zeta potentials on the filed cyanobacterial cells surface showed a trend of changes from high to low during the seasons from Spring to Autumn. Generally, the absolute values of zeta potential on the cells surface at different pH values in Spring were higher than those in Summer (Fig. 2). The decrease of the absolute zeta potential values means that the reduce of electrostatic repulsive force. When the Van der Wals force was above the repulsive force, the cells in waters were inclined to condense. The cyanobacterial cells in Summer and Autumn were easier to aggregate into colony and then to form blooms. The reduce of the absolute values of zeta potential in Summer, which could help explain that why cyanobacterial colony sizes in Summer and Autumn were larger than those in Spring [21] .
As shown in Fig, 5, for both the species of Microcystis aeruginosa and Microcystis flos-aquae, the absolute zeta potential values of cells surface in the stationary phase were higher than those in the exponential phase, which not only suggested that the growth phase could affect the zeta potential, but also meant that the stability of cells in the exponential phase was worse than that in the stationary phase. So the cells in the exponnetial phase were easier to form conlonies or aggregate with other organic matters. There was still difference of zeta potential between the species. Whatever in the exponential phase or in the stationary phase, the amplitude of variation of zeta potential of Microcystis aeruginosa cells was a little larger than that of Microcystis flos-aquae. This may be due to the surface variable charge of Microcystis flos-aquae cells less than that of Microcystis aeruginosa cells. Additionally, we noted that when the pH value was more than 7, the absolute zeta potentials values of Microcystis flos-aquae were lower than those of Microcystis aeruginosa. However, when the the pH value was less than 7, the absolute zeta potentials values of Microcystis flos-aquae were higher than those of Microcystis aeruginosa especially during the stationary phase. In natural waters, the pH value was not less than 7 due to the photosynthesis of phytoplankton during the cyanobacterial blooms periods [22] . This suggested that Microcysits cells in waters could absorb higher negative charge. Considering Microcystis aeruginosa could absorb more negative charge than Microcystis flos-aquae during the cyanobacterial blooming periods, the more stability of cells would be for Microcystis aeruginosa. In the alkaline aquatic environment, some metals such as manganese ion were relatively insoluble. So the increase of negative charge on the cells surface would be critical for the aborption and settlement [23] , which could enhance the capacity of binding cations, and then increase the competition for trace elements such as iron ion. Previous studies have shown that iron limitatioin was not only important for phytoplankton growth in ocean, but also critical for field cyanobacteria in Lake Taihu, iron supply in Lake Taihu could be beneficial for the growth of Microcystis aeruginosa [24] . Microcystis cells in natural waters could absorb relatively high negative charge, which absorb more nutrients (such as iron ion) or other cations, and would be beneficial for the Microcystis cells to dominate. The results reported by Li et al [25] also showed that the EPS of field Microcystis have the strong capacity of binding metals. 20   Previous studies have reported that the surface of Microcystis colony was hydrophobic, whereas single cell in culture was hydrophilic [8] . So in this study, the hydrophobicity of single Microcystis cell in culture was so low that it was not decteced by xylene-water two-phase system. By comparison of the effect of pretreatment on field algal hydrophobicity, there was little difference betweeen filtration and centrifugation. But the hydrophobicity after formalin fixation was significantly higher than those without fixation (p < 0.05), suggesting that formalin fixation changed the characterization of algal surface. In order to avoid formalin to contaminate the samples, choosing filtration or centrifugation for collection would be appropriate.

Hydrophobicity
By determining the hydrophobicity of different colony sizes, there were significant differences among them (p < 0.05). The strongest hydrophobicity was seen in the colony size less than 20  m, and the weakest one was belonged to the colony size larger than 400 m. The hydrophobicity of colony cells with size less than 20 m was significantly stronger than that of colony with size between 20 and 100  m and size larger than 400  m. There were no significant differences between the other sizes. The stronger hydrophobicity suggested that the adhesion capacity of cells surface was stronger, which was more beneficial to the cells aggregate to form colony [8,26] . As the results shown, the hydrophobicity of colony in small size fraction was stronger than those in large size fraction, 21 suggesting that a large amount of colony in small size fraction during the early stage of the blooms would be more easier to aggregate to form large colony. According to the result reported by Yang et al [8] , the hydrophobicity of Microcystis cells was mainly due to the surface polysaccharides. The higher content of monosachrides (rhamnose, fucose, and galactose) were in polysaccharides, the stronger hydrophobicity would be [27] . Through the investigation of the compositions of polysaccharides in bEPS of field Microsystis colony by Li et al. [25] , the ratio of the sum of rhamnose, fucose, and galactose to total polysaccharides was up to 63. 3%. So hydrophobicity was the direct good index for the evaluating the cells surface characterization and would be helpful to recoganize the colony formation mechanism.

Conclusions:
This paper has presented the surface zeta potential and hydrophobicity of blooming Microcystis in the field and in the experimental culture. pH values, growth phase, and the species could affect the surface zeta potential of Microcystis. The zeta potential values were negative for cyanobacteria in the filed and in the experiment, and changed by seasons and by culture time. The absolute zeta potential values in Spring were lower than those in Autumn. And the absolute zeta potential values in the stationary phase were higher than those in the exponential phase. Aditonally, the zeta potential decreased by increasing pH values (310). The hydrophobicity of colony in small size fraction was stronger than that in large size fraction. These informatin would help us to explain the phenomenon of small colony in spring and then form large colony in Summer and Autumn. So this paper has provided a comprehensive knowledge of surface characterization of Microcystis for understanding the mechanisms of colony formation during the blooms. Such a recognization will assist in controlling the outbreaks of blooms. Abstract：Naproxen is a widely used non-steroidal anti-inflammatory drug (NSAID), using for the reduction of pain, fever and inflammation, and thus also released into the aquatic environment, such as lake. Naproxen is chronically toxic to aquatic organisms and bioaccumulates in fish tissue. Here, we propose a new miniaturized naproxen extraction method for aqueous and fish samples by extraction with 1:1 hexane: ethyl acetate and gas chromatography mass spectrometry (GC/MS). Response surface methodology (RSM) was adopted to optimize extraction PH, solvent volume and extraction time. The optimum extraction conditions predicted with the experimental ranges were as follows: pH 2, solvent volume 1.5 mL, and extraction time 5 min. The method was validated using samples fortified with naproxen at levels of 10, 30 and 50 ng/g, the mean recovery exceeds 86%.
Key words: miniaturization, naproxen, response surface modelling, Box-Behnken design 1 Introduction In recent years, residues of pharmaceuticals have been detected in measurable concentrations in various compartments of the aquatic environment (1,2). Of all detected pharmaceuticals, the naproxen is consistently detected at higher concentration in the effluents of more WWTPs than other drug (3,4).
Naproxen [(S)-6-methoxy-α-methyl-2-naphthalene acetic acid] is a widely used non-steroidal 24 anti-inflammatory drug (NSAID), known for the reduction of pain, fever and inflammation. In the United States, the Food and Drug Administration (FDA) approved its use as an over-the-counter (OTC) dug in 1994. As a consequence of its high amount application, naproxen and its metabolites formed in the human body are excreted unchanged and continuously discharged into municipal wastewaters. In Waste Water Treatment Plants (WWTPs), incomplete elimination of naproxen via biotransformation (5,6) result in its presence in effluents at concentrations of up to 5.22 μg/L (7) and finally in surface waters at about 0.4 μg/L (8). Therefore, the safety of naproxen has been questioned with regards to environmental and human health. The phototransformation appears to be the main mechanisms of elimination in the environment for naproxen and the photoproducts are more toxic for the aquatic organisms both for acute and chronic values than the parent compound itself (9,10). This drug is recognized to be highly effective and clinically safe, and the concentration elicit the toxicological effects to fish and other organisms are several magnitudes higher than its found in the environment (11). However, it has been stated that naproxen has some side-effects such as gastrointestinal toxicity, nephrotoxicity, jaundice, hepatotoxicity and induce oxidative stress in the isolated perfused rat liver (12). According to previous report, naproxen and its metabolites have been detected in the bile of rainbow trout (Oncorhynchus mykiss) with bioaccumulation factors of 500 to 2,300 (11).
Zebrafish, a small tropical fish native to the rivers of India and South Asia, is an animal of great scientific interest due to the advantageous features over other vertebrate model systems (13). Low test cost of using adult zebrafish facilitates bioaccumulation kinetics of research, meanwhile zebrafish have over 80% genomic homology to humans, which enables a significant correlation of the data obtained between the two species. In addition, zebrafish is one of the model species recommended by OECD for bioconcentration Test 305, which evaluates the accumulation a dissolved chemical in adult fish by measuring its final concentration in both, the fish and the surrounding media after an equilibration time (14). However, little attention has been devoted to the extraction, purification and characterization in fish sample of naproxen. Herein, we report in detail the optimization of extracting parameters, the purification and the preliminary characterization of naproxen and its metabolites.
The response surface methodology (RSM) has been extensively utilized to optimize culture conditions and medium composition of fermentation process, conditions of enzyme reaction, and processing parameters in the production of food and drug (15). Box-Behnken design (BBD), one of RSM, only have three levels (low, medium and high, coded as -1, 0 and +1), and need fewer experiments. It is more efficient and easier to arrange and interpret experiments in comparison with others (16)(17)(18). Therefore, BBD of RSM was used to optimize the extracting parameters of naproxen in the present work. Firstly, single-factor experimental designs (extracting temperature, extracting time, extracting pH, and ratio of extracting solvent to raw material) were carried out before RSM experiments. Secondly, three factors (extracting temperature, extracting time and ratio of water to raw material) were chosen based on single-factor designs for further optimization by employing a three-level, three-variable BBD from RSM. Furthermore, naproxen was purified through liquid-liquid extraction (LLE) and solid-phase extraction (SPE) and the isolated fractions were characterized using GC-MC. 25 2 Materials and methods

Reagents and chemicals
Naproxen (purity 98%) and internal standards rac-ibuprofen-d3(purity 98%) were purchased from Sigma (St. Louis, MO, USA) and J&K Chemical Co., respectively. For the pH-value adjustment, the samples were acidified with 1 N HCl and pH value was measured with the pH meter (Thermo Orion pH meter 520A, Cambridge, UK). Analytical-grade hydrochloric acid (37% Hangzhou, China) used as the reagent, and other solvents used in experiments, including n-hexane, ethyl acetate, acetonitrile, methylene chloride, acetone, methanol) were of HPLC grade and purchased from Tedia Company Inc. (Fairfield, OH, USA). Water was purified in Millipore, Milford, MA, USA). Solid-Phase Extraction (SPE) cartridges (Florisil, 500 mg of adsorbent/3 mL of reservoir) were obtained from Anpel Scientific Instrument Co., Ltd (Shanghai, China).

Reference standard solutions
Standard solution of naproxen was prepared in methanol at the concentration of 1,000 mg/L as stock solution and preserved at 4 o C in the darkness. Working solutions were prepared by a series of 10-fold dilutions of the stock solution and also stored in the dark at 4 o C.

Sample preparation
Adult zebrafish were blended in a laboratory blender and frozen at -20 o C, before analysis. It showed that no detectable residue of naproxen was contained and it can be used as a negative controls. Screening experiments were carried out on zebrafish spiked with naproxen. The standard naproxen ibuprofen-d3 internal standard solution (100 mg/L in methanol) were spiked into homogenised fish samples and then kept at 4 o C for 24 hours, resulting in a final concentration of 5 μg/g naproxen of sample for use in the evaluation of the subcritical extraction.
The optimisation procedure was applied to the analysis of real samples of adult zebrafish. Adult zebrafish (Danio rerio ) obtained from a local commercial fish farm and were acclimated in our laboratory at least for one month before using for experiment. In the experiment, 10 adult zebrafish were randomly distributed in 5L glass tanks containing different concentrations of naproxen (0.1, 1, 10, 100 μg/L) for two months. Three replicates were run for each concentration, and the exposure solutions were renewed completely on a daily basis. During the experiments the water temperature was maintained at 27±1 o C on a light/dark cycle of 12 h: 12 h. The zebrafish were fed twice daily with live brine shrimp. At the end of exposure, zebrafish were taken from the pool, euthanised and stored at -20 o C for later determination.

Extraction procedure
Fish sample of 200 mg were transferred to 10 mL FEP tube containing 5.0 mL acidified Milli-Q water and 1:1 hexane: ethyl acetate (0.5-1.5 mL) was added followed by vigorous mechanical shaking for 30 sec, ultrasound bath for 5-15 min, and centrifugation at 10000×g for 2 min, and then the upper organic layer was removed to a clean 5.0 mL glass vial. Lower layer was re-extracted one more time and both organic phases were combined and evaporated to dryness 26 using nitrogen gas at 40 o C. Extracts thus obtained were dissolved in 1 mL 7.5:1.5:1 methylene chloride: acetone: methanol.
Extracts were cleaned up by Florisil-SPE column. The cartridges were conditioned with 5 mL methylene chloride and 3 mL 7.5:1.5:1 methylene chloride: acetone: methanol. Above solution was loaded into the cartridge and then 7.5:1.5:1 methylene chloride: acetone: methanol (2 mL) was used to clean the glass vial for three times, in total of 6 mL, followed by loading into the cartridge together and 3 mL n-hexane was used to wash some impurities and fat, followed by air drying the cartridges; naproxen was eluted using 1.5 methanol and then the solvent was evaporated to dryness using nitrogen gas, followed by dissolved in 100 μL 1:1 n-hexane : ethyl acetate. Then, 200 μL of 1N HCl in methanol was added to each sample and heated for 30 min at 80 o C. After the reaction, the solution was extracted two times with 1.0 mL 1:1 n-hexane: ethyl acetate. The organic layer was evaporated to dryness under a stream of nitrogen gas and dissolved in n-hexane (1 mL) and kept until GC/MS analysis. All experiments were carried out in triplicate and the means of results were used for optimisation.

Optimization of process variables and experimental design
Multiple variables could affect the efficiency of naproxen extration, the response surface methodology (RSM) is an effective technique for optimising the process (19). Three process variables (sample pH, extraction solvent volume and extraction time) were identified to investigate their influence on the extraction of naproxen form adult zebrafish. In the present work, a three-variable and three-level Box-Behnken design (BBD), a method of response surface methodology (RSM), was applied to optimize the process variables. The coded and uncoded independent variables used in the BBD were listed in Table 1. The levels of the independent parameters were based on preliminary experimental results and research reports (20,21). The experimental design was based on the Box Behnken design with three central points as shown in Table 2. The experimental data obtained were analyzed by the response surface regression procedure using the following second-order polynomial equation: Where Y is the estimate response (extraction rate%), β0 is defined as a constant, β1, β2 and β3 are the linear coefficients, β11, β22 and β33 are quadratic coefficients, β12 ,β13 and β23 are interaction coefficients between the three factors. Three-dimensional surface response plots were generated using in fitted model by varying two variables within the experimental range and holding the other constant at the central point. The coefficients of the response surface equation were estimated by using the software of Design Expert 8.0.7.1 (State-Ease, Inc., Minneapolis). The test of statistical significance was based on the total error criteria with a confidence level of 95.0%.

Chromatographic conditions
An Agilent series gas chromatography mass spectrometry (GC/MS) system was used in this study. The ion source was operated in an electron ionization mode (EI; 70 eV, 230 o C). Select-scan mess spectra (m/z 185) was recorded for analyte identification. The instrument parameters of the GC/MS detector was: splitless mode; t helium gas flow rate, 1.0 mL/min; injector temperature, 27 310 o C; transfer line temperature, 300℃; oven temperature, programmed from 140 o C at 12 o C/min to 320 o C (hold for 10 min).
3 Results and discussion

Experimental design and regression model
The design matrix and the corresponding extraction rates of RSM experiments to determine the effects of the three independent variables including pH (X1), extraction solvent volume (X2) and extraction time (X3) were presented in Table 2. According to these experimental results, the model for the predicted response Y could be expressed by the following empirical second-order polynomial equation (in the form of coded values): The analysis of variance (ANOVA) is shown in Table 3. The significance of the model was determined using F-test. The value of the coefficient of determination (R 2 ) and adjusted coefficient of determination were 0.9739 and 0.9268, respectively, indicating that a high degree of correlation between the observed and predicted values. For any of the terms in the model, a large F-values and a small P-value would be more significant (22). The F-value of 20.71 and p<0.05 implied this model was significant. Conversely, the F-value and P-value of the lack of fit were 9.53 and 0.0964, respectively, which implied that it was not significant (p>0.05). The significance of each coefficient was determined using F-value and P-value. The results were given in Table 3. It could be seen that the linear parameters (X1, X2) and one quadratic terms (X1 2 ) were significant (p<0.05). Results also showed that the independent variable X1 was the most significant factor on the extraction rate of naproxen.

Three dimensional (3D) response surface
It has been reported that 3D response surface plots was a function of two factors, maintaining all other factors at fixed levels helpful in understanding both the main and their interaction effects of these two factors (23). Thus, in order to gain a better understanding of the effect of independent variables on a dependent variable, 3D response surface plots were shown in Fig. 1

Fig. 1(A)
illustrates the effect of extraction pH and solvent volume on the extraction rate. When extraction time was fixed at 0 level, the extraction rate increased with increasing extraction solvent volume (X2) and declined with the pH (X1). Extraction rate increased sharply when the pH decreased from 4 to 2, indicating that the high recovery for extraction of naproxen was at lower pH value. As shown in Fig. 1(B), when extraction solvent volume was fixed at 0 level, the extraction rate was increased with decreases in pH. However, the influence of the extraction time on the extraction rate is negligible. The results of Fig. 1(C) showed that when pH was fixed at 0 level, the extraction rate was increased with increase in extraction solvent volume and extraction time displayed negligible effect on the extraction rate. 3.3 Optimization of extraction condition 28 The naproxen extraction conditions would be considered optimum if the extraction rate reached maximum value. From the solutions predicted by the model, the experimental conditions set at the pH of 2, extraction solvent volume 1.5 mL and extraction time 5 min, could give an extraction rate of 90. 78%. Under the optimum conditions, the experimental extraction rate was 89.7% (N=3), which was close to the predicted value.

Results of method validation
Under the optimum conditions described, the calibration curves were linear (R 2 =0.9923) over the range of 25-500 ng/ml. The accuracy and precision of the method were determined using fish samples fortified with naproxen at levels 10, 30 and 50 ng/g. All experiments were repeated four times. The mean recovery of naproxen was above 86%. Table 1 Uncoded and coded independent variables used in Box Benken Design

Short-term TN and TP prediction by back propagation artificial network based on genetic algorithm in Poyang Lake Basin, China
Yujie  L. et al., 2005)), can simulate nitrogen and phosphorus concentration of water body. However, it is very expensive and time-consuming to obtain huge amounts of data (e.g., long term of daily rainfall, runoff, sediment and water quality data) for modeling. Moreover, high requirements of the model users make these models difficult or unavailable to be widely applied in solving practical problems (Suen and Eheart, 2003;

Wen et al., 2013).
Concentration of nitrogen and phosphorus in water body can be affected by many factors (e.g., temperature, pH and dissolved oxygen) (Mcqueen and Lean, 1987). Because of the complex and nonlinear relationship among these factors, there are high uncertainties in the traditional methods for the short-term nitrogen and phosphorus prediction (Omer Faruk, 2010). Thus, a new effective and accurate method for predicting the short-term nitrogen and phosphorus in water body is in great need to be developed. In the last two decades, with the leapfrog development of computer hardware and software, high speed computing performance of computer provides a solid platform for the developments of artificial intelligence technology. Methods involving artificial neural network (ANN) is believed to be an effective tool for dealing with non-linear problems, and have been gradually applied in various water resources problems, such as hydrological processes ( . Therefore, the ANN method can also be considered as an alternative for the short-term nitrogen and phosphorus prediction in water body. For the ANN method, the back propagation (BP) algorithm has been applied widely. However, the ANN trained by BP algorithm (BP ANN) often suffers from converging too slowly and being easily trapped into a local optimum (Bakhbakhi, 2012;Li et al., 2013). In recent, some scholars stated that the training process of ANN can be considered as a classical optimization problem (Alexandridis et al., 2014), and the determination of ANN structure, parameters (e.g., weights, threshold and bias) play a crucial role in training accuracy and generalization ability for ANN (Heuvelmans et al., 2006).  (Hou et al., 2014) proposed improved Pareto ACO algorithm (PACA) for optimal spatial allocation of water resources, results of the improved PACA are superior to those of other intelligent optimization algorithms including ACO, multi-objective GA and BP ANN. Compared with the traditional ANNs, these hybrid algorithms not only accelerate the convergence rate, but also are easier to obtain the global optimal solution.
Compared to PSO and ACO, GA has a relative mature convergence analysis method with estimable convergence rate, and has the ability to deal with discrete problems ( In addition, GA has also been used to train deterministic models. Sahoo (Sahoo, D. et al., 2010) used GA for optimizing model parameters for the Hydrologic Simulation Program-Fortran (HSPF), and the loosely coupled GA-HSPF model shows a more objective and less time-consuming ability than traditional ones. Similarly, GA can be also adopted for SWAT model parameter optimization (Zhang, X. Zhang, X.S. et al., 2010). Thus, these studies have revealed that GA is a powerful, potential approach to train ANNs and other models.
Despite a high level of water quality prediction accuracy in some researches, improving the performance of the prediction model is still the first-line goal of this present study, which focuses on developing an effective and accurate hybrid model for the short-term nitrogen and phosphorus prediction. Combining with a variety of known understandings, including field monitoring, laboratory analysis and artificial intelligence modeling, we proposed a BP ANN coupled with GA (the hybrid GA-BP ANN) for the short-term nitrogen and phosphorus prediction in Poyang Lake Basin. For this model, the genetic algorithm is introduced to adjust and optimize the connection weights of BP ANN, to overcome its premature convergence problem and to accelerate the convergence speed, and finally to achieve the global optimum. To evaluate the performance of this proposed model, a comparison between Total Nitrogen (TN) and Total Phosphorus (TP) predictions from the hybrid GA-BP ANN and the traditional BP ANN was carried out.

Study area description
Poyang Lake Basin is the largest freshwater lake in China, located in the southern bank of the Yangtze River, north of Jiangxi province. The water surface area of the lake changes greatly with seasons, reaching its maximum of more than 4125 km 2 in high water level (20m) and minimum of less than 500 km 2 in low water level (12m). This lake is primarily fed by five big rivers including Gan River, Rao River, Xin River, Fu River and Xiu River. Moreover, it belongs to subtropics moist monsoon climatic region, with moderate climate, adequate lighting, and abundant rainfall (annual average precipitation is 1636 mm).
Four typical sampling sites including Kangshan (S1), Lianhu (S2), Duchang (S3) and Hamashi (S4; Fig.1) within the Poyang Lake were employed to investigate the short-term TN and TP predictions based on the locations of the sites, which are almost in a line from south to north in Poyang Lake.  Davis, 1991.). In 1975, the idea of this method, which was inspired by the theory of natural evolution, was firstly proposed by Professor John Holland from Michigan University. Compared with traditional optimization algorithms, the superiority of the GA are, firstly, it hardly fall into local optima in the search process, even if the definition of the fitness function is not continuous, irregular or noise, it can also find the global optimal solution with great probability (Song, 2013). Secondly, due to the inherent parallelism, GA is very suitable for large-scale parallel computer (Bahlous et al., 2013;Song, 2013).
To solve optimization problems, GA works with a population of individual strings (chromosomes), each representing a possible solution (Goldberg and Holland, 1988). Abiding by the law of "survival of the fittest", each chromosome is assigned a fitness value according to the result of the fitness function. Highly fit chromosomes are given more opportunities to reproduce, named reproduction; and then the chromosomes are produced by two genetic operations: crossover and mutation. From generation to generation, the offspring sharing features taken from their parents finally converge to the best chromosome at the end of the evolution. Therefore, GA can be summarized as two operation processes: genetic operators (including crossover and mutation) and evolutionary operators (selection).
In this present study, a hybrid GA-BP ANN is proposed based on GA's global search ability and BP's nonlinear capability. Thus, the hybrid GA-BP ANN can not only overcome the blindness of optimizing process, but also avoid the occurrence of local convergence (Koehler, 1997). The structure and the algorithm process of the hybrid GA-BP model are detailed in Fig.2. Fig.2 The BP ANN process optimized by Genetic Algorithm

Evaluation criteria
In this study, the predictability of the proposed models can be evaluated by average relative deviation (ARD), coefficient of determination (R²) and the Nash-Suttcliffe coefficient of efficiency (NSE) (Nash and Sutcliffe, 1970). The ARD, R² and NSE are expressed by the following equations, respectively: Where N is the number of sample data. p Q and o Q is the simulated and the observed data, separately.
p Q is the average of the simulated, while o Q is the average of the observed.

Dataset preprocessing (1) Determination and selection of water quality indices
In the present study, the dataset from S1, S2, S3 and S4 were monitored with equal time intervals in rainy, average flow and dry seasons during 2009 to 2010. Combining with field monitoring and laboratory analyses, eight water quality indices (Table 1) Table 1 shows that the annual average of NH4 + -N, BOD5, TN and TP at S2 are significantly higher than the others, probably because of the extrinsic factors, such as anthropogenic sources (S2 located below the Lianhu Bridge). Consequently, the dataset from S1, S3, and S4 are selected for the present work.
(2) Interpolation Given the dataset collected in this work, to carry out monthly short-term TN and TP prediction, three values should be interpolated between each two adjacent time points so that the frequency of the dataset was improved to per month.
The interpolation method was realized by interpolation function programming based on Matlab R2012b platform. Through comprehensively comparing of four kinds of interpolation function [i.e., Linear (linear interpolation), Nearest (near point interpolation), Cubic (cubic polynomial interpolation) and Spline (cubic spline interpolation)], the Spline function was finally adopted in this study. The time range was defined as t (t = 1.00, 1.25… 6.00), where the integers represented the real monitoring time and the decimals represented the interpolation time from 2009 to 2010. Thus, each sampling site was interpolated to 21 samples. Fig.3 shows the observed and the interpolated series, and the interpolation function curves in S1 site. Fig.3 The fluctuations of each water quality index based on the Spline function in S1 site As a data-driven ANN model, data preprocess can significantly influence the model performance. For equally being treated for all the variables in training and test stages, the input and output variables have been normalized to range from 0.1 to 0.9, which is preferable when a logistic activation function whose outputs are between 0-1 was used in the network (Nash and Sutcliffe, 1970), the normalization equation is given as: Where max x and min x are the maximum and minimum of the input and output data, respectively.

Architecture
In this present study, two water quality indices, (t+1)-month concentration of Total Nitrogen and Total Phosphorus, hereafter denoted by TNt+1 and TPt+1, respectively, were determined as the output variables of the hybrid GA-BP ANN. Moreover, eight water quality indices (Table 1) can be chosen as the network input variables. However, some studies have shown that increased input variables may increase the complexity of the neural network, resulting in low efficiency of the algorithm and poor prediction effect (Jeong et al., 2006). By analyzing the correlation of the eight water quality indices (Table 1), CODmn has poor correlation with both TN and TP (-0.23 and -0.17, respectively), consequently, seven variables of t-moth (i.e., Tt, pHt, DOt, NH4 + -Nt, BOD5t, TNt, TPt) were finally chosen as the input variables ( Fig.4).

Fig.4 Input and output variables of the hybrid GA-BP model for short-term TN and TP prediction
After dataset preprocessing, the available dataset were generally divided into training and test subsets for developing the hybrid GA-BP model. The training set was used to estimate the unknown parameters (e.g., weights, threshold and bias), and the test set was used to assess the generalization ability of the trained model (Wen et al., 2013). In this study, 42 samples of data from S1 and S3 were selected as training set and the other 21 samples of data from S4 were selected as test set, respectively. The main parameters of the hybrid GA-BP model are listed in Table 2, and the other parameters are set as default values. Note that traditional BP model uses the same parameters as the hybrid GA-BP model except for population size and the maximum generation. The optimum number of the hidden-layer neurons was determined by the minimum ARD value of TN and TP prediction based on the trail and error procedure. An experiment was performed with a variation of 3-11 neurons.
To avoid contingency, each architecture configuration was implemented for 50 runs with different initializations. And then, the average value of ARD was calculated to evaluate the performance of the hybrid GA-BP model with different neurons (Table 3).  Table 3 shows that when the number of hidden-layer neurons is less than 8, the performance (ARD) for TN and TP prediction gradually decreases for both training and test. Nevertheless, when the neurons is more than 8, the ARD for TN and TP prediction changes very slightly for training set. Meanwhile, the ARD presents a more or less rising trend for test set, revealing that excessive number of hidden-layer neurons may result in "overfitting" for training, and reduce the generalization ability of the network (Tetko et al., 1995). Hence, the number of the hidden-layer neurons is finally determined to 8, and the hybrid GA-BP ANN architecture is 7-8-2.

Results and discussion
Using the proposed hybrid GA-BP ANN model with architecture of 7-8-2 and the traditional BP ANN, we predicted the short-term TN and TP in Poyang Lake Basin ( Fig.5 and 6). It is important to note that, the traditional BP ANN was trained under the same condition (i.e. input and output variables, the number of hidden-layer neurons) as the hybrid GA-BP's for comparison, other scenarios were not considered, as only in this way can the improvement brought by GA be contrastively analyzed.
As depicted in Fig.5 and 6, both the simulated TN and TP from the different models generally fluctuate around the observed. Related to the traditional BP ANN, the simulated TN and TP from GA-BP ANN model are much closer to the observations, basically indicating this proposed model can capture the characteristics of the TN and TP fluctuations better during both the training and test stages.   R. et al., 2007), and the reason for this shortcoming is the premature convergence problem of the BP ANN, in other words, the BP ANN is easily trapped into a local optimum with bad generalization ability, resulting in lower accuracy of test set than that of training set.     Fig.11, the hybrid GA-BP ANN is superior in terms of convergence rate and accuracy, as indicated by fewer iterations and lower MSE, revealing that the hybrid GA-BP ANN has been overcome the premature convergence problem of BP ANN by optimizing the connection weights of BP ANN, and finally achieve the global optimum.
It should be noted that, as new exploration, the application of the hybrid GA-BP ANN for short-term TN and TP prediction is still at the primary stage. In the present study, there are still some uncertain aspects including the number of the training and test subsets, the influence of the interpolation method and the optimization of the network structure, which need to be further explored. Nevertheless, it can be concluded that the hybrid GA-BP ANN proposed outperforms the traditional BP ANN, and can be an alternative model for short-term TN and TP prediction in Poyang Lake Basin.

Conclusion
In this study, a new hybrid GA-BP ANN model is developed and is applied to predict the short-term TN and TP in Poyang Lake Basin. Comparing the performance of the hybrid GA-BP ANN against the traditional BP ANN, the evaluation results are very encouraging even with a relatively small number of dataset employed for training and test. Therefore, conclusion can be drawn as, for short-term TN and TP prediction, the hybrid GA-BP ANN performed better than the traditional BP ANN, with a lower ARD, and higher R² and NSE during both the training and test stages. In addition, the simulated TN and TP of the proposed model fluctuates better around the observed than the traditional BP ANN, revealing that the proposed model has efficiently overcome the shortage premature convergence of the traditional BP ANN, and can be considered as an alternative model in predicting the short-term TN and TP in Poyang Lake. Moreover, new hybrid ANNs based on the other swarm intelligence algorithms (i.e., PSO and ACO) will be a focus. In the future, we will follow up on this subject and focus on developing new hybrid models to solve more realistic problems. 47 Simulation investigation of relationship between water level and surface area of Poyang Lake in typical hydrologic years 1 Introduction Poyang Lake is a river-connecting lake, located in the south of the middle and lower reaches of The Yangtze river, and its latitude is between 28°24'N and 29°46'N, longitude 115°49'E and 116°46'E. The topography of the lake region is like a "Bowel", the slope of lake basin is from southeast to northwest). Due to the double impacts of five rivers flows (Ganjiang river, Fuhe river, Xinjiang river, Raohe river, Xiushui river) importing into the lake and 48 the withstanding of water level in the Yangtze River. The annual and interannual variabilities of the water level are large. The annual maximum amplitude is 9.59m to 15.36m, interannual maximum variability is 16.69m [3,5] .
Generally, water level get start to grow up to the flood period owing to an increasing amount of rainfall since late March, it reach the peak from June to August and it does not decrease until October, the following year in March is the dry season, the water level dropped, bottomland bare. In the area, the highest water level is 22.59m (Hukou station, July 31, 1998), the corresponding water surface area is about 4, 070 km 2 , the lowest water level 5.9m, (Hukou station, February 6, 1963), and river-connecting lake area is less than 200 km 2 [1,4] , the area change significantly.
Lake area is an important ecological variable and it can be extracted by remote sensing information and actual measurement or calculated by the statistical relationship between the water level and area. Such as Ding (2010) measured the area curve of reservoir water level by combined with DEM and remote sensing [2] . Zhang (2012) etc. used the remote sensing data to estimate relationship between the level and the area of Poyang Lake [11] . Liu (2011) etc. studied the relationship model of water area of Poyang and water level of Hukou [6] .
Due to the complex terrain of Poyang Lake, the traditional assessment methods are not only time-consuming, but also difficult and costly. The method of remote sensing exist the shortcomings of timeliness and low temporal resolution. Therefore, how to calculate the surface area of Poyang effectively and quickly is a problem that deserves to be discussed.
In this paper, in order to build the statistical relationship of Lake area and water level for deducing the surface areas by actual water levels, four sites in different parts of upstream and downstream in Poyang Lake were selected, daily surface area data of high temporal resolution were simulated on the basis of numerical simulation of hydrodynamic model, statistical model of that was built up by correlation analysis of actual water level and lake area in four sites. Finally, the statistical model was verified by surface area data using remote sensing methods.

Hydrodynamic model and main control equation
In this paper, the lake area data were gotten by the hydrodynamic model, and used to build the statistic model of water level and lake area. The basic hydrodynamic model is EFDC (Environmental Fluid Dynamics Code) model, which is developed by the Virginia Institute of Marine and funded by the United States Environmental Protection Administration. It combined with hydrodynamic, sediment, contaminants transport and water quality etc.. The EFDC model can be used to simulate the 1-dimensional, 2-dimensional, 3-dimensional numerical variables on different temporal and spatial scales (such as Lake, reservoir, estuarine, bay and wetland etc.) [9,10] .
EFDC model adopts orthogonal curvilinear coordinate in the horizontal direction but sigma coordinate in the vertical direction. The momentum equation is: Continuity equation is: State equation is: Where H is the total depth, H=h+  , h is the average sea level (Yellow sea elevation) distance to the seabed,  means the instantaneous water level; u, v, w velocity components were fitted borderline orthogonal curvilinear coordinates, x, y, z direction; where m = m x m y, m x and m y are the diagonal elements of the metric tensor The square root; m is the square root of the determinant of the metric tensor; f is the Coriolis number; g is the acceleration of gravity; p is the pressure; Av is the vertical turbulent viscosity coefficient; Qu, Qv respectively x, y direction momentum sink term; ρ is the mixed density; ρ0 is the reference density; S is the salinity; T is the temperature [6,8] . Salinity of the equation S = 0, and assuming that the water is an incompressible fluid, the density ρ and the temperature T is constant.

Poyang Lake hydrodynamic modeling methods and data sources
By using remote sensing images during flood period of Poyang Lake in 1998 as a reference and combining with GIS data of levee in Poyang Lake, the maximum extent boundary of Poyang lake was determined, and on this basis, Poyang Lake was meshed by using orthogonal curvilinear grid, the total number of grids is 96004, the resolution of grids is between 178m and 205m, and the orthogonal grid parameter is less than 0.2. The DEM of Poyang Lake was provided by Hydrology Bureau of Jiangxi Province, which was measured in 2010 with a map scale of 1:10,000. The upper boundary of the model is daily measured flow data in Qiujin, Wanjiabu, Waizhou, Lijiadu, Meigang, Hushan and Dufengkeng station (m 3 /s), the low boundary is daily measured water level data in Hukou station (m, yellow sea height)( fig.1). The judgment parameters of dry-wet in the model are: water depth of dry grid is 0.16m, time step of dry grid is 16s.  [7] , therefore, in this paper, they are regarded as typical years for normal, dry and wet hydrologic situation.
Daily dry-wet status of every computational grid in typical years can be simulated by established Poyang Lake hydrodynamic model, in which dry-status means grids without water while wet-status presents grids with water. Therefore, the total water surface area of Poyang Lake, which will be used in subsequent anlysis of this article, can be gotten by accumulation of each wet-grid area.

Analysis of the statistical relationship between water level and surface area in typical years
The daily lake surface areas in typical years (2005,2006 and 2010) can be obtained by using hydrodynamic model of Poyang Lake. In order to compare mutual relationships between water levels and lake surface area in upstream and downstream of Poyang Lake, and build representative regressive relation, which uses water level as independent variable, lake area as the dependent variable, Kangshan, Tangyin, Duchang and Xingzi were selected as four actual water level sites in different parts of upstream and downstream in Poyang Lake. Finally, the mutual relationships of water level and lake area in wet, normal and dry hydrologic year were analyzed with daily simulated surface area and actual water level in above-mentioned years.

Analysis of the relationship in dry hydrologic year
Through curve regression analysis between the actual water and simulated area of Poyang Lake in 2006 (dry 51 hydrologic year), the relationships between water level and lake area of Kangshan, Tangyin and Xingzi present a cubic equation, Duchang is quadratic equation (see Figure 2). The Curve regression relationships can be expressed as following:  The figure 2 shows, the correlation of water level and lake area of four sites have some differences. From the space perspective, northern sites correlation is better than that in the south of lake. The correlation coefficient of Kangshan, Tangyin, Duchang and Xingzi is 0.746, 0.907, 0.952 and 0.953 respectively, the correlation of the Xingzi station is best, and Kangshan station is relatively poor, but all of them show strong correlation in 0.01 level (bilateral).
The North-South differences of the correlation of water level and surface water area mainly relates to the terrain of Poyang Lake and the hydrological characteristics of "High water level like a lake and low water level like a river". Inflowing rivers of Poyang Lake are mainly distributed in the south of lake. The terrain of lake in South is higher than that in North leads to the annual variation of southern sites (such as Kangshan) is less, and northern sites (such as Xingzi, Duchang and so on) is larger. When the water level variation of downstream (north of lake) or the heartland of lake does not affect the water level of upstream, the water level of sites in south can not reflect the water level and area of lake. In this situation, the error is greater.
As can be seen from the figure 2, when the area of Poyang Lake is 1400km 2 to 1600km 2 , 2800km 2 to 3000km 2 , there is a obvious phenomenon that there are several water area at the same level of each site, this phenomenon also affect the correlation of water level and water area in to some extent, which is relate to the process of 'swelling' or 'recession'. That is the lake area corresponding to the process of 'lifting' is different at the same level. Because the paper length, the statistical relationship of water level and lake area in the process of 'lifting' will be discussed in subsequent papers. 3   Compared with dry hydrologic year, there is more affluent inflow into the lake in normal hydrologic year, which results in the variation of water level of Poyang Lake being larger than that of dry hydrologic year. Therefore, the variation of surface area of Poyang Lake is also larger. Similarly, owing to the higher ground of southern stations (e.g., Kangshan station), its amplitude of water level in wet, normal and dry hydrologic year is smaller than that of lake district, thus the correlation of actual water level and simulated area of southern stations is poorer than dry hydrologic year. The process of 'lifting' in normal hydrologic year is stronger than that in dry hydrologic year, resulting in the variation of corresponding surface area of Poyang Lake being also larger in the same condition. It can be clearly informed from figure 3 that the plots is more scattered than those in dry hydrologic year when the area of Poyang Lake is in 2, 000km 2 to 2, 500km 2 or 3, 000km 2 , and it indicates the corresponding relationship of water level and surface area is more complex. 54

Analysis of the relationship in wet hydrologic year
The statistical relationships for water level and lake area of Kangshan, Tangyin, Duchang and Xingzi present Cubic equation, and coefficient of determination (R 2 ) is close to 1,according to the curvilinear regression analysis for 2010 year (wet hydrologic year). The statistical relationship is as follows:  As can be seen in Figure 4, the curve overall trend of the actual water level and simulated area is similar to that in dry, normal hydrologic year. The correlation coefficient of Kangshan, Tangyin, Duchang and Xingzi is 0.745, 0.924, 0.960 and 0.962 respectively and passed the significance test at 0.01 level (bilateral). The correlation of Xingzi station is the most obvious. 55 Due to inflow into lake is most plentiful in wet hydrologic year, the lake water level amplitude is largest. The southern sites of higher ground that are greatly affected by water level of Poyang lake present the same trend as the change of surface area. Consequently, the relationship between water level and lake area is approximately the same as that in dry hydrologic year.

Result verification
To verify the relationship of actual water level and simulated area of Poyang Lake, 6 scenes of remote sensing images of Landsat 7 ETM are downloaded respectively in some days of 2005, 2006 and 2010, in which days water level is in the medium and lower level. The actual surface areas of Poyang Lake are calculated by ArcGIS software using remote sensing data.  Table 1 lists the imaging time of six scenes ETM and the corresponding actual surface areas calculated by remote sensing data. Statistical relationship between water level and surface area in different years are established, using water level in the imaging time of remote sensing data as the independent variable, and plugging it into statistical equations of water level and surface area that has been established, lake surface area can be deduced. Table 1 also lists estimated surface areas. By contrasting the two areas, errors between simulated areas and actual areas are assessed. Table 1 presents absolute errors and relative errors of actual areas and simulated areas respectively, while figure 5 presents the contrasted schematic diagram of two areas. It is noteworthy that the correlation in Duchang station is relatively stronger, therefore the simulated areas in table 1 is estimated by statistical equation of water level and surface area in Duchang station. From the table 1 and figure 5 it can been seen that the error between estimated area and actual area of Poyang lake in dry, normal and wet hydrologic year, and the absolute error is in 4.85km 2 to 270.47km 2 , the relative error in 0.24% to 15.17%.
Since cloudy and rainy during high water level, the remote-sensing images without cloud in the Poyang Lake are difficult to obtain. In the error analysis, there are no any remote-sensing images to be used to verify the estimated area of Poyang Lake during high water level.

Conclusion and discussion
Based on hydrodynamic model of Poyang Lake, the lake area of day by day in 2005, 2006 and 2010 year (typical normal, dry and wet hydrologic year) were simulated in this paper. By analyzing the statistical relationships between lake area simulated and measured water level from the different positions of the four lakes (Kangshan, Tangyin, Duchang and Xingzi), The Statistical equation , which is obtained by using the water level to estimate lake area of the four sites in Poyang lake, was built.
Results indicate that there were positive correlations between the water level and area, The average correlation between the water level and lake area of Duchang site in dry, normal and wet hydrologic year is stronger. At the time of estimated the lake area by the measured water level, the statistical equations between water level and lake area in Duchang can be used. Because of the error being relatively large, in the period of simulating low water level by Poyang Lake hydrodynamic model, the error of simulated area is relatively large. In this case, there must be some errors when the established equation is used to estimate the relationship between the actual water level and the lake area. Therefore, the surface area calculated from the actual water level exist uncertainty especially during the low water level. However, it is noteworthy that the established equation is in accord with statistical law relatively, due to high temporal resolution of simulated surface area data. In conclusion, there is a scientific significance.
The built equations for estimating lake surface area only represent overall relations between water level and lake area of Poyang Lake in one year, in fact, through above mentioned analysis, the overall relations of water level and surface area of Poyang Lake also contain many secondary relationships in different processes of 'lifting, and these relations lead to the obvious phenomenon that different surface areas of Poyang Lake corresponding to the same water level. To build more elaborate and accurate relationship of water level and surface area, it is necessary to explore interrelation of that in different process of 'lifting'. The details will be dealt with in the further study because of the paper length.     Abstract: Urban lakes are an important part of cities,with a variety functions such as water storage,flood control,ecology,landscapes and so on,while the negative effect on urban lakes caused by urban construction and expansion can not be ignored.This paper,by taking Jiujiang City as an example,makes a comprehensive exposition that water environment protection of lakes should be taken seriously during the process of urban development,and puts forward the corresponding countermeasures.   Abstract: Poyang Lake Ecological Economic Zone construction is an important strategic initiative for Jiangxi Province to transformits economic development mode and explore the new path for coordinate development of ecology and economy. With a superior natural environment, Poyang Lake Ecological Economic Zone has abundant eco-tourism resources and rich history and cultural heritage. Based on the present situation of Jiangxi Province, this article first analyzes the scientific development and Green Rise conditions and environmentithas, then, focusing on the theme of harmonious development of green tourism and ecological construction of Poyang LakeEcological Economic Zone, this paperstudies and discusses how to give full play to the ecological advantages and implement the strategy of increasing competitiveness by developing tourism.Next, the paper elaboratesthe great importance of the ecological resources for the economic growth of Jiangxi Province, and puts forward ideas and suggestions on green tourism development and ecological protection.  Abstract ： The characteristics of water and sediment and the change rule of the Gan River into Poyang Lake are analysed by making use of the monitoring data of the runoff and suspended sediment discharge from1957 to 2010 in hydrological control station of Gan river downstream (Waizhou hydrological station), the results showed that:(1) The annual runoff from Gan River into the lake has experienced the change process of increased first and then decreased in the last 50 years, Since 1998 the last ten years is significantly reduced, but the overall trend is still increasing; In the last ten years (after 1998), the annual runoff from Gan River into the lake decreases obviously, the annual runoff of Downstream decreases fastest, the middle reaches is second, upstream is slower; Changes of annual runoff in the Gan River into the lake is mainly controled by annual precipitation change, the influence of human activities on annual runoff changes in many years is very small, it have clearly reflect In the hydrology statistical results. (2) The annual sediment discharge of Gan River into the lake in the last 50 years has also experienced 79 a changing process increases at first and then decreases, but the increasing speed is very small, the duration is very short also, the overall trend is increasing. Since 1991,because of a number of large and medium-sized reservoir in the middle reaches of Gan River are impounding, The suspended sediment in Gan River is intercepted of large number, the annual sediment discharge is quickly greatly reducing. (3) The related point groups of Gan River into the lake annual sediment yield and annual runoff showed a double distribution, the hierarchical time (year) nodes is in 1991, the running time is consistent with the main reservoir was put into. Each layer was rather close correlation, can set up a ideal years experience formula; Correlation of product between annual sediment yield and annual runoff and peak flow were significantly correlated, but appeared before, after two year period of stratification, the relationship between stratification time nodes and annual sediment yield and annual runoff is the same. The annual sediment yield empirical formula of established based on the universal soil loss equation is higher improved of the accuracy Stability and reliability than did not join the experience formula, and is more practical.   Abstract: The lake plays an important role in the sustainable development of the whole economic society as a kind of important natural resources, with functions of storage, water supply, aquaculture, shipping ,tourism 。 In view of the existing problems of lake protection and management seriously affected the work of flood control and disaster management system disorder ， such as the area of atrophy; function degradation; serious water pollution; blind fence breeding. Put forward to some suggestions such as law enforcement, lake participation mechanism, the restoration of water environment quality and the lake management mode.      Abstract: Nowadays ecotourism has become the main tendency of tourism development, and it also becomes the effective way to protect wetland ecosystem and explore wetland resources. The paper took the national natural reserve of Lake Poyang-Nanji wetland as a case study. Based on the analyses of characteristics of tourism resources and environment for Lake Poyang-Nanji wetland and the tourism exploitation actuality, the development model of wetland ecotourism was discussed from functional division, ecotourism product development, community participation in tourism development, the protection of wetland resources and environment, and the construction of tourism service facilities, so to provide the reference and instruction for the ecotourism development in wetland natural reserves.  [12] 。 而当前对畜牧业中抗生素的研究主要集中于畜禽生物体残留上 [13,14] ， 对于伴随粪便尿液进入环境中的抗生素研究则相对较少，因而其危害性也不是很清楚。 2 Key Laboratory of Poyang Lake, Jiangxi Academy of Sciences, Nanchang 330096 *Author for correspondence E-mail: 7086006@qq.com Abstract: In Jiangxi province, the status of water pollution scale pig farms were analyzed that Swine wastewater in high concentrations of organic matter, heavy metals, antibiotics and other pollutants will bring potential hazards to air, water and soil. Summarizing and analyzing the status quo of scale pig farm wastewater treatment, then proposed for small, medium, large-scale pig farm wastewater corresponding technical feasibility recommended.  [4] 。日本从 1973 年开始进行"农村村镇排 水工程"建设，主要是集中处理农村生活排水，后来开发的净化槽技术在分散式生活污水处 理中得到了广泛应用 [5,6] 。丹麦 1987 年对农村生活污水排放标准进行立法，颁布了分散式生 活污水处理指导守则 [7] 。德国从 2003 年起实施"分散市镇基础设施系统"项目研究，利用膜 生物反应器净化偏远村镇生活污水 [5] 。澳大利亚提出了菲儿脱污水处理土地利用系统 [8] 。目 前，欧洲和美国有 20%~30%的人口使用分散污水处理设施，日本有 66%的人口使用净化槽 技术，分散式处理技术在这些国家已经应用成熟 [4] 。 国内从 20 世纪 80 年代开始，对分散式污水处理技术同样进行了有益的探索和实践， 人 工湿地、稳定塘、生物滤池、厌氧好氧组合工艺在中国农村地区得到了研究和应用。太湖流 域农村生活污水处理开展了以"厌氧水解+跌水充氧接触氧化+折板潜流式人工湿地组合技 100 术"，"塔式蚯蚓生态滤池组合技术"及"厌氧发酵+生态土壤+蔬菜种植组合技术"为核心的示 范工程建设，取得了良好的有机物脱氮除磷效果 [9] 。中国各地关于农村污水治理的规定和标 准陆续出台，如：上海制定了《村镇排水工程技术规程》和《上海市农村生活污水处理技术 指南》 ；江苏出台的《江苏省太湖水污染治理工作方案》规定太湖一级保护区农村生活污水 处理率达到 70%，其他地区农村生活污水处理率达到 40%；浙江编制了《浙江省农村生活 污水适用技术与实例》等 [10] 。 2 分散式生活污水处理主要技术 2.1 人工湿地 人工湿地是通过模拟自然湿地，人为设计与建造的由基质、植物、微生物和水体组成的 复合系统，利用"基质-微生物-植物"复合生态系统的物理、化学和生物的三重协同作用， 通过过滤、吸附、沉淀、离子交换、植物吸收和微生物分解来实现对污水的净化，具有高效、 低耗、投资省、适用范围广等诸多优点 [11,12] 。常见的人工湿地主要有垂直流、表面流和潜 流 3 种形式人工湿地， 其中用于农村生活污水处理的主要是表面流和潜流 2 种形式人工湿地 [13] 。 自上世纪 70 年代德国学者 Kichuth 等提出根区法理论以后，国内外学者从污染物去除 机理、影响因素、基质选择、植物配置、工艺设计等各方面对人工湿地污水技术进行了大量 的研究 [11][12][13][14][15] 。帖靖玺等认为降低容积负荷有利于保持系统的除污效率，二级湿地采用粒径较 小的填料有助于维持系统对 NH4 + -N、TN 和 TP 去除效果的稳定性 [16] 。Hench 等通过美国一 个小型人工湿地连续 2 年的观测，发现进水负荷在 19Ld -1 时，TSS、BOD5 和 TN 的平均去 除率分别为 83%、42%和 55%，且第一年污染物去除效果较好 [12] 。孙亚兵等认为自动增氧型 潜 流 人 工 湿 地 具 有 较 强 的 抗 冲 击 负 荷 能 力， 当 COD 、 NH4 + -N 、 TP 进 水 浓 度 分 别 在  [35] 。厌氧消 化技术在分散生活污水处理中得到了广泛的研究与应用， 发展了越来越多的高速处理设备和 技术，如厌氧滤池、升流式污泥床反应器、厌氧膨胀颗粒污泥床等。荷兰、巴西、哥伦比亚、 印度等国家已建成生产性升流式污泥床反应器来处理生活污水。 日本开发的净化槽主要处于 一种厌氧-兼氧的环境条件，根据结构不同分别采用"厌氧过滤+接触曝气"、"反硝化型厌氧 过滤+接触氧化"、"新型膜分离净化槽"工艺，既提高了出水质量，又起到了脱氮除磷的效果 [35] 。 中国在农村地区推广小型装置包括无动力地埋式污水处理装置、 厌氧沼气池处理技术等。 102 冯华军等针对厌氧工艺脱氮性能差的问题，在 ABR 厌氧处理工艺的基础上对填料式厌 氧折流板反应器进行了优化， 开发了一种适合用于处理分散式低浓度生活污水的无回流脱氮 反应器，适用于分散式生活污水的达标处理 [36] 。研究进一步发现，分散式生活污水的有机负 荷冲击不会对填料式厌氧折流板反应器的出水造成显著影响，COD 去除率可达 80% 以上， 但在中温(18℃)和低温(10℃)条件下，反应器抗有机负荷冲击的稳定性随温度的下降而下降 [37] 。刘志强等采用缺氧/好氧膜生物反应器((A/O-MBR)试验装置进行中试试验，研究表明 [5,8,38] [31,[39][40][41][42][43] 。陈鹏等通过研究厌氧接触工艺和改型潜流湿地的污染物去 除性能，进行模块化单元优化组合出一种厌氧接触与生态组合处理工艺，COD 和 SS 平均去 除率分别为 82. 5%、96.5%，其中 NH4 + -N 和 TP 的去除主要发生在湿地单元 [35] 。赵迎迎等利 用"厌氧池+射流充氧滴滤塔+人工湿地"组合工艺，进行了厌氧、好氧和人工湿地三级处理， COD、NH4 + -N、TN、TP 的去除率分别可达 85%、74%、76%、80% [44] 。冉全和吕锡武研究 表明，"厌氧池+接触氧化+潜流人工湿地"组合工艺 COD、NH4 + -N、TN、TP 的平均去除率 分别为 73%、97％、87％、93％ [45] 。吴磊等研究表明脉冲滴滤池在水力负荷为 7.0 m 3 m -2 d -1 ， 布水周期为 20 min， 回流比为 2.0 的最佳运行条件下， 与水力停留时间为 12 h 的前置水解池， 以及水力负荷 50cmd -1 的后置人工湿地组合运行时， 水解池+脉冲滴滤池+人工湿地组合工艺 对 COD、TN 和 TP 的平均去除率分别达 91％、95％和 95％ [46] 。以上结果表明，这些组合 工艺能以较低的处理成本，取得良好的污染物去除效果，获得稳定的出水水质，具有更强的 适用性和应用性。 5 结语 分散式污水处理工艺已经成为低密度、 人口分散居住的广大农村地区生活污水处理优先 选择的工艺，但污水工艺技术的选择是否得当直接影响到污水处理运行效果和稳定程度， 也 对工程基建投资、运行费用、管理操作等产生根本影响。中国农村幅员辽阔，农村地区的生 态环境、经济发展水平各异，在实际应用中应根据自然环境、地形条件、人口规模、投资情 况等因素，因地制宜选择污水处理工艺，充分实现成本的最小化和效用的最大化，促进可持 Abstract: Decentralized treatment of domestic sewage has become a new idea. Due to many advantages such as small investment, convenience of management ，high purification efficiency, and low maintenance costs, etc, it is suited for rural domestic sewage treatment. However ， the choice of techniques of sewage treatment is directly related to the effects of sewage treatment and the environmental health in an area. In this paper, the present status of treatment of rural domestic sewage both in China and abroad were reviewed, and several technologies of decentralized treatment domestic sewage were introduced, such as constractd wetlands, stabilization pond, artificial floating island, biological filte and anaerobic biological treatment technology. Based on the comparison of advantages and disadvantages of many decentralized treatment technologies being used in China and abroad，their treatment effect，environmental and economic conditions were discussed. Finally, we suggested that, hybrid of different processes has stronger applicability for decentralized treatment domestic sewage, which promoting decentralized treatment facilities in the future.   16   Abstract: In order to understand characteristics of microbial contamination and water use function in Poyang Lake Natural Reserve ( Poyang Lake Migratory Bird Sanctuary ) ， Bang Lake, Da huchi,Sha Lake, Gan River and Xiu River were selected as research object. Samples were set in Bang Lake, Sha Lake and so on in different level seasons. Three sampling sections were set about Bang Lake by the direction of flow in wet season, level season and dry season in 2012-2013. Four kinds of indictor bacteria including the total number of colonies, total coliforms, fecal coliforms and enterococcus, three kinds of pathogenic bacteria including salmonella, golden staphylococcus and hayes bacteria were tested and analyzed. The results showed that:(1)The pollution levels of five waters : Bang Lake＞Sha Lake＞Da huchi Lake＞Xiu River＞Gan River. Faecal pollution mainly came from the human being. The detecton rate of pathogenic bacteria: Bang Lake ＞Sha lake =Da huchi＞Gan River=Xiu River. (2) The total number of colonies in Bang Lake: wet season ＜ level season ＜ dry season. Various seasons were polluted by faecal, faecal pollution mainly came from the human being in wet season and level season, however faecal pollution mainly came from the animal in dry season. The water quality of three seasons were not conform to requirements of drinking water and recreational water, different degree of pathogenic bacteria were detected in three seasons. (3)    Abstract: Soil is a key component of wetland ecosystem, and remediation is the basis for restoration of damaged wetland ecosystems .This study studied the chemical and physical aspects of damaged wetland soil in Poyang Lake, and revealed their spatial distribution features of five different damaged wetland soil types. The damaged soil types included a boggy soil，two meadow soils，a paddy soil and a meadow-boggy soil. The result showed that the damaged wetland soils has low soil pH and cation exchange capacity (CEC), and soil nutrients were in the impoverished state. There was a similar spatial distribution character between soil organic matter (SOM) and total nitrogen (TN) in 5 types of soils. The accumulation of SOM and TN showed the same tendency: the concentrations of SOM were the highest in the surface layer, but declined in the deeper levels. Soil texture and soil moisture significantly affected the spatial distribution of N. Due to higher vertical traveling speed of NO3 --N in soil, sandy soil was more easily leaching than sludge soil. On the other hand, the concentrations of NH4 + -N in different soil layers were significantly higher than the content of soil NO3 --N in the counterparts, indicating soil has strong adsorption ability for NH4 + -N. However, soil NH4 + -N content of different soil types in the same layer had no significantly differences. The result also showed that the heavy metals gathered in the middle layer (20～40 cm) and resulted from the same source. The heavy metal content of boggy soil near Poyang lake were the highest in all five types of soils. In addition, the heavy metals level in sampling site, including Zn, Cu, Pb, Hg, As, were lower than those in soil environmental quality standard. The results could provide the reference for damaged wetland ecological restoration and reconstruction choosing the appropriate technology model. Abstract: The water quality dynamic monitoring and water ecosystem for Poyang lake during the nearly 10 years (2004~2013) were studied in this paper. The results showed that the water quality in general presented a decreasing trend, the current eutrophication status was moderate, and the main pollutants were the total phosphorus (TP) and ammonia nitrogen (NH3-N). By using the seasonal Kendall test for the main pollutants of TP and NH3-N, it was found that TP and NH3-N appeared a significantly rising trend for most of monitoring cross-sections in the lake and the controlling-sections in the lake outlet. From the results of water bloom evaluation, it was still safe for Poyang lake and the algal density has no obvious variation trend. Moreover, some protection countermeasures were proposed according to current situation of water quality and water ecosystem in Poyang lake. 生物浮床(Ecology Floating-beds System)是一项重要的人工湿地技术，自从上世纪80 年代在德国被发明并投入应用以来，生物浮床技术因其在富营养化水体治理中独特的优势， 脱颖而出，已经被越来越多的科学家及政府部门认可。生物浮床，也称"人工生物浮床" [1] ， "人工浮岛" [2,3] ， "浮床无土栽培" [4 ]，它运用无土栽培原理，把水生植物或改良的陆生植 物以人工构建的浮床为载体，栽种到富营养化水体的表面，通过植物本身在生长过程中从水 体中吸收氮、磷等营养物质的特点，除去水体中的氮、磷等营养元素；通过发达的根系， 发 挥对污染物的吸附降解作用； 同时为微生物的生存及降解水体中的有机污染物提供必要的场 所；同时还通过与藻类竞争养分、氧气起到抑制藻类生长的作用(图1-1) ，已经广泛应用于 包括农业废水处理，城市雨水处理，养殖废水，富营养湖泊治理等各个方面 [5][6][7][8][9]