How can we reduce phosphorus export from lowland polders? Implications from a sensitivity analysis of a coupled model

How can we reduce phosphorus export from lowland polders? Implications from a sensitivity analysis of a coupled model

Science of the Total Environment 562 (2016) 946–952 Contents lists available at ScienceDirect Science of the Total Environment journal homepage: www...

1MB Sizes 0 Downloads 18 Views

Science of the Total Environment 562 (2016) 946–952

Contents lists available at ScienceDirect

Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv

How can we reduce phosphorus export from lowland polders? Implications from a sensitivity analysis of a coupled model Jiacong Huang, Junfeng Gao ⁎, Renhua Yan Key Laboratory of Watershed Geographic Sciences, Nanjing Institute of Geography and Limnology, Chinese Academy of Sciences, 73 East Beijing Road, Nanjing 210008, China

H I G H L I G H T S

G R A P H I C A L

A B S T R A C T

• Three models were coupled to describe phosphorus (P) dynamics in Polder Jian, China. • Sensitive factors affecting annual P export from Polder Jian were identified. • Proper fertilization should be a strategical priority for reducing polder P export.

a r t i c l e

i n f o

Article history: Received 30 December 2015 Received in revised form 31 March 2016 Accepted 10 April 2016 Available online 3 May 2016 Editor: Simon Pollard Keywords: Polder Phosphorus Sensitivity analysis Soil erosion Pond

a b s t r a c t Phosphorus (P) export from lowland polders has caused severe water pollution. Numerical models are an important resource that help water managers control P export. This study coupled three models, i.e., Phosphorus Dynamic model for Polders (PDP), Integrated Catchments model of Phosphorus dynamics (INCA-P) and Universal Soil Loss Equation (USLE), to describe the P dynamics in polders. Based on the coupled models and a dataset collected from Polder Jian in China, sensitivity analysis were carried out to analyze the cause-effect relationships between environmental factors and P export from Polder Jian. The sensitivity analysis results showed that P export from Polder Jian were strongly affected by air temperature, precipitation and fertilization. Proper fertilization management should be a strategic priority for reducing P export from Polder Jian. This study demonstrated the success of model coupling, and its application in investigating potential strategies to support pollution control in polder systems. © 2016 Elsevier B.V. All rights reserved.

1. Introduction

⁎ Corresponding author. E-mail address: [email protected] (J. Gao).

http://dx.doi.org/10.1016/j.scitotenv.2016.04.068 0048-9697/© 2016 Elsevier B.V. All rights reserved.

Polder systems are widely located in the lowland areas around rivers, lakes and seas. Phosphorus (P) export from these polders was severe due to the intensive farming of agricultural lands, and caused

J. Huang et al. / Science of the Total Environment 562 (2016) 946–952

severe eutrophication problems in the surrounding aquatic systems (Hellmann and Vermaat, 2012; Puijenbroek et al., 2004). To reduce P export, many strategic options, such as wetland and ecological ditch construction, have been extensively tested (Fisher and Acreman, 2004; Olli et al., 2009). In general, these strategies have proven to be timeconsuming and costly. It is therefore worthwhile to evaluate the potential for reducing P export before implementing these strategies and to identify the most cost-efficient one. Such an evaluation is challenging due to the complexity of polder systems and our inadequate understanding of P dynamics. Numerical models are a means of predicting P export under future conditions, and for evaluating alternative strategies to reduce P export (Shen et al., 2012). Many watershed models, such as SWAT, HSPF, AGNPS, USLE and INCA-P, have been developed and describe the detailed processes related to P dynamics in freely draining catchments (Arnold and Fohrer, 2005; Borah and Bera, 2004; Daniel et al., 2011; Ongley et al., 2010; Renard et al., 1991; Wade et al., 2002). These models effectively simulated P dynamics in various catchments, investigated potential P export responses to scenarios of changing land use and climate, and helped water managers set appropriate goals for P export reduction (Jackson-Blake et al., 2015). However, the application of these models to lowland polder systems has been limited to date. In contrast to freely draining catchments, polder systems exchange water with their surrounding rivers by manually controlled processes, such as irrigation, culvert and flood drainage (Lindenschmidt et al., 2009). These manually controlled processes were recently described in a simple Phosphorus Dynamic model for lowland Polder systems (PDP) developed by Huang et al. (2016). However, this model did not describe the detailed processes related to P dynamics in agricultural lands (e.g., dry and paddy lands). Therefore, model coupling may overcome the weaknesses of existing watershed models and PDP.

947

This study used Polder Jian in China as an example and aimed to examine how P export from this polder responded to environmental factors (e.g., land use and meteorological conditions) and to investigate potential strategies for reducing P export from this polder. The detailed processes related to P dynamics in the polder were described based on the tight coupling of three models, i.e., PDP, INCA-P and USLE. Sensitivity analysis was then carried out to quantify the response of P export to changes in land use, vegetation and manual water-management practice. Potential strategies to control P export were then discussed based on the results of the sensitivity analysis. 2. Material and methods 2.1. Model description To elucidate the cause-effect relationships between environmental factors and P export, three models were coupled: the Phosphorus Dynamic model for Polders (PDP), Integrated Catchments model of Phosphorus dynamics (INCA-P) and Universal Soil Loss Equation (USLE) (Fig. 1). PDP described the polder water balance and P dynamics in the water areas. INCA-P simulated the dissolved inorganic and organic phosphorus in the runoff water from dry and paddy lands. USLE estimated the particulate phosphorus in the surface water. The P forms in these models included dissolved organic phosphorus (DOP), dissolved inorganic phosphorus (DIP), dissolved phosphorus (DP), particulate phosphorus (PP) and total phosphorus (TP). DP is the sum of DIP and DOP, and TP is the sum of DP and PP. 2.1.1. Phosphorus Dynamic model for lowland Polder systems (PDP) The Phosphorus Dynamic model for Polders (PDP) was originally developed by Huang et al. (2016) and was based on mass conservation.

Fig. 1. Conceptual diagram for the tight coupling of the Phosphorus Dynamic model for Polders (PDP), Integrated Catchments model of Phosphorus dynamics (INCA-P) and Universal Soil Loss Equation (USLE). DOP, dissolved organic phosphorus; DIP, dissolved inorganic phosphorus; DP, dissolved phosphorus; PP, particulate phosphorus.

948

J. Huang et al. / Science of the Total Environment 562 (2016) 946–952

The model included six modules: water-area water balance module, residential-area water balance module, paddy-land water balance module, dry-land water balance module, water management module and the phosphorus balance module (Fig. 1). These modules were based on a daily time step and described the physical and biological processes related to P dynamics as well as the manually controlled processes, such as water export by pumping or through a culvert. The PDP model was calibrated and validated based on a dataset collected from Polder Jian, China. The validation results showed an acceptable model fit with an R2 (coefficient of determination) value of 0.594. However, the weakness of PDP is that it did not include the P-related processes in the dry and paddy lands. The DP and PP in the runoff water from dry and paddy lands were obtained from measured data. This weakness meant that PDP required intensive data input to implement a simulation, and could not be used to investigate the response of P export to the changes in dry and paddy lands. 2.1.2. Integrated Catchments model of Phosphorus dynamics (INCA-P) The Integrated Catchments model of Phosphorus dynamics (INCA-P) is a dynamic and process-based model that was originally developed by Wade et al. (2002). Over the past decade, it has been greatly improved and widely applied to estimate catchment P export and evaluate alternative strategies for controlling P pollution (Jackson-Blake et al., 2015; Whitehead et al., 2014). The model described the P dynamics in soils and groundwater based on different land use and incorporated different P sources, such as fertilization and manure, with a daily time step. In this study, INCA-P was coupled to estimate DP in the runoff water, as determined by DOP and DIP in the soil water. DP Runoff ¼ DOP Soil þ DIP Soil

ð1Þ

DOP and DIP dynamics in the soil zone were mainly affected by precipitation, fertilization, crop uptake, mineralization and immobilization. dDOP Soil dDOP FB ¼ DOP Sour þ DIP Immo −DOP Mine −DOP Uptake − dt dt

ð2Þ

dDIP Soil dDIP FB ¼ DIP Sour þ DOP Mine −DIP Immo −DIP Uptake − dt dt

ð3Þ

DOP

DIP

DIPUptake ¼ kUptake ST SSeason SSMD

DOP Soil V Soil

DIP Soil V Soil

ð4Þ ð5Þ

where kUptakeDOP and kUptakeDIP are the plant uptake rates for DOP and DIP, respectively. SSeason , ST and SSMD are the seasonal plant growth index, temperature and soil moisture factor, respectively. Vsoil is the volume of soil water. The flux of DOP mineralization and DIP immobilization (DOPMine and DIPImmo) are calculated as (Wade et al., 2002) DOP

DOP Mine ¼ kMine ST SSMD DIP

DIP Immo ¼ kMine ST SSMD

dDOP FB DOP Soil DOP FB DOP DOP ¼ kSoil2 FB −k FB2Soil dt V Soil V Soil

ð8Þ

dDIP FB DIP Soil DIP FB DIP DIP ¼ kSoil2 FB −k FB2Soil dt V Soil V Soil

ð9Þ

where kSoil2FBDOP and kSoil2FB DIP are the conversion rate of soil DOP and DIP to firmly bound DOP and DIP, respectively. kSoil2FB DOP and kFB2Soil DIP are the conversion rate of firmly bound DOP and DIP to soil DOP and DIP, respectively. Further details about the P-related processes and their corresponding equations can be found in Wade et al. (2002). 2.1.3. Universal Soil Loss Equation (USLE) The Universal Soil Loss Equation (USLE) was used in this study to estimate the annual sediment yield (Sed, ton/a). Sed ¼ R  K  L  S  C  P

ð10Þ

where R is the rainfall erosion factor, K is the soil erodibility factor, C is the crop management factor, P is the erosion control practice factor, L is the topographic factor, and S is the coarse fragment factor. Equations to calculate these factors can be found in Neitsch et al. (2005). Based on the calculated annual sediment yield (Sed), PP concentration in the runoff water could be estimated as follows. PP Runoff ¼ P Soil Sed

ð11Þ

where PPWater and PSoil are PP concentrations in the runoff water and soil water, respectively. 2.2. Study area and data

where DOPSoil and DIPSoil are the DOP and DIP in the soil water. DOPSour and DIPSour are the DOP and DIP load through precipitation. DOPMine is the mass change of DOP due to mineralization. DIPImmo is the mass change of DIP due to immobilization. DOPUptake and DIPUptake are the mass change of DOP and DIP due to plant uptake. DOPFB and DIPFB are the firmly-bound DOP and DIP. DOP and DIP uptake by crop (DOPUptake and DIPUptake) depended on the season, soil temperature and moisture, and could be described by the following equations (Wade et al., 2002). DOP Uptake ¼ kUptake SSeason ST SSMD

where kMineDOP and kMineDIP are the DOP mineralization and DIP immobilization rate, respectively. The P exchange between soil P and firmly-bound P are calculated as (Wade et al., 2002)

DOP Soil V Soil

ð6Þ

DIP Soil V Soil

ð7Þ

Polder Jian is located in Lake Taihu Basin, China, and has an area of 10.6 hm2 (Fig. 2). Its land use is a mixture of paddy land (50.1%), dry land (21.7%), residential area (19.2%) and water surface (9%). The water surface includes a ditch network and ponds for water transport and for storage purposes. During the dry period, the ditches transport irrigation water from the surrounding rivers to agricultural lands to meet their crop water requirement. During the heavy rainfall period, the ditch network receive runoff water from different land uses to ponds. Polder Jian is characterized by a semitropical climate with an annual precipitation of approximately 1116 mm. Most of the rainfall occurs between March and September. Heavy rainfall events (daily precipitation N25 mm) frequently occur during the summer period and result in severe soil erosion and the release of pollution. In contrast to freely draining catchments, Polder Jian is surrounded by large rivers and exchanges water with them through manually controlled systems, such as pumps and culverts. During heavy rainfall events, the ponds are water storage areas. Its water level thus increases significantly, and may harm crops. When the pond's water level rises higher than that of its surrounding river, the pond water can be exported through culverts. However, when the pond's water level is lower than that of its surrounding river, the pond water should be exported by pumps. A dataset collected from Polder Jian by Huang et al. (2016) was used to calibrate and validate the coupled model. This dataset included land use, vegetation coverage, and meteorological and water quality data from Polder Jian collected between Jan. 1 and Dec. 31, 2014. The land use data were obtained from interpretations of satellite images. The vegetation coverage data were obtained by monthly surveys at five sites (V1–5 in Fig. 2). The meteorological data were collected based on an automatic rain gauge at Polder Jian and a national weather station

J. Huang et al. / Science of the Total Environment 562 (2016) 946–952

949

Fig. 2. The locations of Polder Jian in China, descriptions of its land use and sampling sites (Huang et al., 2016).

near the polder. The water quality data were collected by water and sediment sampling at W1–7 and S1–3 (Fig. 2), respectively. Further details about this dataset can be found in Huang et al. (2016). 2.3. Calibration and validation A method for global sensitivity analysis (variance-based) was used to screen the sensitive parameters affecting the target variable (TP in the water area). This method computed a sensitivity value (STi) to compare parameter sensitivity. STi ¼

EX i ðV X i ðYjX i jÞ V ðY Þ

ð12Þ

where Xi is the i-th parameter. X~ i denotes the matrix of all parameters but Xi. V(Y) is the total unconditional variance. EX~ i(VXi(Y|X~ i) is the expected variance that would be left if all parameters but Xi could be fixed. A higher value of STi implies that the i-th parameter was more sensitive to the target state variable (TP in the water area). n simulations are needed to compute V(Y), while k ∗ n simulations are needed to compute EX~ i(VXi(Y|X~ i). n is the sample size (n = 200). k is the number of input parameters (k = 34). This sensitivity value has been widely used and greatly encouraged for its potential in screening the most sensitive parameter (Dai and Ye, 2015; Saltelli and Annoni, 2010; Shahsavani and Grimvall, 2011). Further information about the variance-based sensitivity analysis approach can be found in Saltelli and Annoni (2010). Thirty-four parameters were adjusted by trial-and-error method based on their values from the literature. Four sensitive parameters

(Table 1) were further optimized using the Genetic Algorithm method proposed by Goldberg (1989). A k-fold cross-validation was then used for model validation. The one-year dataset collected from Polder Jian was divided into four sub-sets (Jan. to Mar., Apr. to Jun., Jul. to Sep. and Oct. to Dec.). These four sub-sets were used in turn for validation. Implementation details on the cross-validation approach can be found in Huang et al. (2016). The model fits were assessed using five statistical measures, i.e., mean absolute error (MAE), mean absolute percent error (MAPE), root mean square error (RMSE), coefficient of determination (R2) and index of agreement (d). 2.4. Sensitivity analysis P export from lowland polders by infiltration, culvert and flood drainage was the major P source from the polder to surrounding rivers (Huang et al., 2016). Therefore, sensitivity analysis was carried out to screen for environmental factors influencing P export via infiltration, culvert and flood drainage. The variance-based sensitivity analysis method (Eq. (12)) that was used to screen sensitive parameters in Section 2.3 was also used here. For the sensitivity analysis of environmental factors, the symbols in Eq. (12) were defined as follows, • • • • •

STi: The sensitivity value of the i-th variable. Xi: The i-th variable. X~ i: The matrix of all variables but Xi. V(Y): The total unconditional variance. EX~ i(VXi(Y| X~ i): The expected variance that would be left if all variables but Xi could be fixed.

Table 1 Sensitive parameters (sensitivity value higher than 0.05) in the coupled model. Sensitive parameter

Sensitivity value

Value range (reference)

Calibrated value

Maximum water storage of the paddy land Maximum resuspension rate of particulate phosphorus from sediment Setting rate of particulate phosphorus to sediment Phosphorus uptake rate of plant

0.32 0.30 0.07 0.07

160–180 0–0.01 0–0.0625 (Mao et al., 2008; Pei and Ma, 2002) 0–0.3 (Hu et al., 2006)

162 mm 0.0024 0.0059 0.0085

950

J. Huang et al. / Science of the Total Environment 562 (2016) 946–952

A sample size of 500 was used to compute V(Y) and EX~ i(VXi(Y|X~ i). Eleven environmental factors was tested, including meteorology (air temperature and precipitation), land use (water area and residential area), water level (culvert and flood controlling), aquatic plant (aquatic plant coverage and harvest) and P sources (fertilization, P deposition and soil erosion). The sensitivity value of a testing factor for P export and water export was then calculated. A higher value of STi implies that the i-th variable was more sensitive to the target state variable (P export or water export). 3. Results 3.1. Model calibration The TP in the water area was well simulated, with a R2 value as high as 0.72 and a RMSE value of 0.023 mg/L. Its trend of increase from Jan. to May and decrease from Sep. to Nov. was well simulated. This model fit was slightly better than that of PDP (Table 2). However, several TP peaks (e.g., the high TP on Nov. 26, 2014) were not well reproduced. The PP dynamics in the water area were better described than the DP dynamics. The PP variation in 2014 was significantly larger than the DP variation (Fig. 3). 3.2. Sensitivity analysis The sensitivity analysis results showed that P export from Polder Jian was sensitive to air temperature, precipitation and fertilization with a sensitivity value larger than 0.05, and it was most sensitive to precipitation with a sensitivity value of 0.65. Changes to the aquatic plant coverage, residential and water areas moderately affected P export with a sensitivity value larger than 0.01. Other environmental factors showed a slight effect on P export with a sensitivity value b 0.01 (Fig. 4). P export from Polder Jian was sensitive to fertilization (Fig. 4), implying that intensive fertilization would result in large P export via its infiltration to agricultural lands. P export from Polder Jian was not sensitive to P deposition through precipitation because P import from precipitation (wet deposition) was far less than that from fertilization based on the P balance calculation by Huang et al. (2016). The culvert and flood controlling operations had negligible effects on P export because their magnitudes were small (200 mm) in Polder Jian. A large change in these two parameters was not acceptable for water management. For example, an extremely high threshold of water storage in the water area for starting the flood pump would result in a flooding disaster in the polder. 4. Discussion 4.1. Model performance From the perspective of water management, TP was a greater concern than DP and PP. Compared with previous case studies for aquatic modeling reported by Arhonditsis and Brett (2004), the R2 value of TP

Table 2 Model fits of the coupled model and Phosphorus Dynamic model for Polders (PDP) for measurement and simulation of the phosphorus in Polder Jian water area. TP, total phosphorus; DP, dissolved phosphorus; PP, particulate phosphorus. Model

Items

MAE (mg/L)

MAPE (%)

RMSE (mg/L)

R2

d

PDP

PP DP TP PP DP TP

0.019 0.012 0.020 0.021 0.015 0.015

29.4 42.6 20.5 33.6 58.8 15.0

0.026 0.015 0.027 0.027 0.017 0.023

0.53 0.18 0.59 0.57 0.11 0.72

0.83 0.62 0.86 0.83 0.53 0.91

Coupled model

Note: The model fits of the Phosphorus Dynamic model for Polders (PDP) were obtained from Huang et al. (2016).

(Table 2) implied that the coupled model performed better than 70% of the reported case studies. Such performance was acceptable for modeling the P dynamics of Polder Jian, which was strongly influenced by human activity, considering that simulating phosphorus dynamics in aquatic systems has been widely recognized as a challenging task (Robson, 2014). The DP dynamics were not well simulated by the coupled model (Table 2), implying that the processes (e.g., P uptake by phytoplankton and transformation between inorganic and organic P) related to DP dynamics were not adequately described. Such detailed description of these processes was not difficult, because many complex models (e.g., Environmental Fluid Dynamics Code, AQUATOX and PCLake) for aquatic ecosystems have already been developed with detailed description of P dynamics (Janse et al., 2010; Park et al., 2008; Tetra Tech, Inc., 2007). However, to control model complexity and data requirement, these processes were so far simplified or not described in the coupled model. 4.2. Potential strategies to reduce phosphorus export Three environmental factors (air temperature, precipitation and fertilization) showed a high sensitivity for affecting P export. However, air temperature and precipitation were difficult to control. From a water management perspective, controlling fertilization is important for reducing P export from Polder Jian because excessive P fertilization in agricultural lands would result in P infiltration into the groundwater, and P loss in the runoff water (Li et al., 2014). An ideal fertilization situation is when the amount of P precisely meets the crop requirements, but without excessive P accumulation in the soil of agricultural lands. Such fertilization would minimize the P loss from agricultural lands. Several strategies, such as evaluating P requirements of the crop, predicting P dynamics in soils and avoiding fertilization before heavy rainfall events, were encouraged to achieve better fertilization practices. P export from Polder Jian was moderately sensitive to three environmental factors (residential area change, water area change and aquatic plant coverage in the water area) (Fig. 4), implying that additional strategies could be used to reduce P export in combination with controlling fertilization. Among these three factors, aquatic plants have been widely demonstrated to play an important role in P removal in aquatic ecosystems (Fisher and Acreman, 2004; Silvan et al., 2004; Vymazal, 2007). Aquatic plants increase P retention by uptaking P directly and by constructing a stable environment for P settling (Vymazal, 2007). Corresponding strategies to increase aquatic plant coverage, such as wetland construction in pond areas, are helpful for improving P retention (Land et al., 2013). However, it is important to note that aquatic plant harvest is important for removing P from water area and for avoiding P pollution due to plant decay (Vymazal, 2007). Changes to the residential area moderately affected P export by the P load intensity from human sewage and livestock excretion. To date, sewage water plants have not been built in Polder Jian, resulting in a relatively large P export from the residential area. A change in water area moderately affected P export from Polder Jian. This is because water area (ditches and ponds) was a means of transporting water from the polder to its surrounding rivers. Several P-related processes (e.g., P uptake, settling and resuspension) occurred during the transport period. P export was not sensitive to soil erosion in agricultural lands, implying that erosion controlling practices in Polder Jian were not as important as other factors (e.g., proper fertilization management) in reducing P export. This is perhaps because the sediment transport capacity was low in this lowland polder with a low hydraulic gradient of ditches. Considerable sediment from soil erosion would settle into the bottom area, rather than being transported elsewhere. P deposition had a slight effect on P export due to its limited proportion (10.5% for Polder Jian in 2014) of total P import. This is different from nitrogen, where atmospheric deposition may constitute a larger proportion of its total import (Zhao et al., 2011b).

J. Huang et al. / Science of the Total Environment 562 (2016) 946–952

951

Fig. 3. Phosphorus from simulation results and measurement data at W1 (Fig. 2) in 2014. DP, dissolved phosphorus; PP, particulate phosphorus; TP, total phosphorus.

4.3. Advantages and disadvantages of model coupling 4.3.1. More effective investigation of cause-effect relationships Model coupling in this study had the advantage of implementing a more comprehensive investigation of the response of P export to environmental factors. Such an investigation would assist water managers in identifying the key factors affecting P export, so they can implement cost-effective measures for reducing P export. For example, previous studies found that both proper fertilization management and increasing aquatic plants were considered important strategies for reducing P export (Taguchi and Nakata, 2009; Xue et al., 2014). However, the sensitivity analysis results from this study revealed that P export from Polder Jian was more sensitive to fertilization than to aquatic plants, implying that proper fertilization management should be a priority for reducing P export from Polder Jian.

4.3.2. Better understanding of the mechanisms of P-dynamics in polder systems Based on the simulation results of the coupled model, further investigations of P dynamics in Polder Jian were implemented. For example, P dynamics in dry and paddy lands were not described in PDP but were described in INCA-P. The coupling of PDP and INCA-P facilitated us to investigate the effects of P dynamics in agricultural lands on P export from Polder Jian. Similarly, the coupling of PDP and USLE facilitated us to evaluate the contribution of soil erosion from agricultural lands to P export from Polder Jian. Three sensitive factors (air temperature, precipitation and fertilization) were identified as sensitive factors affecting P export from Polder Jian. However, the sensitivity value values in Fig. 4 implied that they had different mechanisms for changing P export. Precipitation and air temperature primarily affected P export by changing the exchange of water between Polder Jian and its surrounding river. Fertilization

affected P export by affecting P cycles in the polder, and did not impact water export (Fig. 4). 4.3.3. Increase in model complexity and data requirements Although the benefits of model coupling is clear, it is important to note that coupling more modules does not necessarily result in a better models. The major disadvantage of model coupling is the increase in the model complexity and data requirements. In this study, the parameter number increased from 24 in PDP to 34 in this coupled model. This change would increase modeling challenge in determine these parameter values. From this perspective, it is important to evaluate that whether coupling is needed considering the coupled model's intended use, and if possible considering data availability. The ‘build for today’ philosophy suggested by Huang et al. (2012) should be encouraged in the practice of model coupling. Moreover, it is better to couple widely used models (such as USLE and INCA-P in this study) rather than the newly developed models in a regional area because the widely used models have undergone increased validation and because previous case studies could provide value ranges for parameter determination. 5. Conclusions Based on a dataset collected from Polder Jian in China, sensitivity analysis was carried out by coupling three models (PDP, INCA-P and USLE) to investigate the response of P export to environmental factors. The results of the sensitivity analysis showed that P export from Polder Jian was the most sensitive to air temperature, precipitation and fertilization. Potential strategies for achieving better fertilization practices would be helpful for reducing P export from Polder Jian. The success of the model coupling in this study provides useful insights into polder P dynamics, and can help managers make better decisions for controlling P export from polders.

Fig. 4. Sensitivity value of environmental factors to water and phosphorus export from Polder Jian in 2014. T, air temperature; Pr, precipitation; LuWater, water area; LuTown, residential area; WLCul, culvert controlling; WLFlood, flood controlling; PlantC, aquatic plant coverage; PlantH, aquatic plant harvest; FerP, fertilization; PrP, Phosphorus deposition; EroP, soil erosion.

952

J. Huang et al. / Science of the Total Environment 562 (2016) 946–952

Acknowledgments The project was financially supported by National Natural Science Foundation of China (41301574). The authors would like to thank China Meteorological Data Sharing Service System for providing the measured data for the model development. Special thanks to Dr. Hongbin Yin, Dr. Wei Huang, Dr. Qi Huang and Mr. Desheng Zhu for their helps on the samplings in Polder Jian.

References Arhonditsis, G.B., Brett, M.T., 2004. Evaluation of the current state of mechanistic aquatic biogeochemical modeling. Mar. Ecol. Prog. Ser. 271, 13–26. Arnold, J., Fohrer, N., 2005. SWAT2000: current capabilities and research opportunities in applied watershed modelling. Hydrol. Process. 19 (3), 563–572. Borah, D.K., Bera, M., 2004. Watershed-scale hydrologic and nonpoint-source pollution models: review of applications. Trans. Am. Soc. Agric. Eng. 47 (3), 789–803. Dai, H., Ye, M., 2015. Variance-based global sensitivity analysis for multiple scenarios and models with implementation using sparse grid collocation. J. Hydrol. 528, 286–300. Daniel, E.B., Camp, J.V., LeBoeuf, E.J., Penrod, J.R., Dobbins, J.P., Abkowitz, M.D., 2011. Watershed modeling and its applications: a state-of-the-art review. Open Hydrol. J. 5, 26–50. Fisher, J., Acreman, M., 2004. Wetland nutrient removal: a review of the evidence. Hydrol. Earth Syst. Sci. 8 (4), 673–685. Goldberg, D., 1989. Genetic algorithms in search, optimization and machine learning. Addison-Wesley, Boston. Hellmann, F., Vermaat, J.E., 2012. Impact of climate change on water management in Dutch peat polders. Ecol. Model. 240, 74–83. Hu, W., Jørgensen, S.E., Zhang, F., 2006. A vertical-compressed three-dimensional ecological model in Lake Taihu, China. Ecol. Model. 190 (3-4), 367–398. Huang, J., Gao, J., Hörmann, G., Mooij, W.M., 2012. Integrating three lake models into a phytoplankton prediction system for Lake Taihu (Taihu PPS) with python. J. Hydroinf. 14 (2), 523–534. Huang, J., Gao, J., Yan, R., 2016. A Phosphorus Dynamic model for lowland Polder systems (PDP). Ecol. Eng. 88, 242–255. Jackson-Blake, L.A., Dunn, S.M., Helliwell, R.C., Skeffington, R.A., Stutter, M.I., Wade, A.J., 2015. How well can we model stream phosphorus concentrations in agricultural catchments? Environ. Model. Softw. 64, 31–46. Janse, J.H., Scheffer, M., Lijklema, L., Van Liere, L., Sloot, J.S., Mooij, W.M., 2010. Estimating the critical phosphorus loading of shallow lakes with the ecosystem model PCLake: sensitivity, calibration and uncertainty. Ecol. Model. 221 (4), 654–665. Land, M., Granéli, W., Grimvall, A., Hoffmann, C.C., Mitsch, W.J., Tonderski, K.S., Verhoeven, J.T., 2013. How effective are created or restored freshwater wetlands for nitrogen and phosphorus removal? A systematic review protocol. Environ. Evid. 2 (1), 16. Li, Y., Šimůnek, J., Jing, L., Zhang, Z., Ni, L., 2014. Evaluation of water movement and water losses in a direct-seeded-rice field experiment using Hydrus-1D. Agric. Water Manag. 142, 38–46.

Lindenschmidt, K.-E., Pech, I., Baborowski, M., 2009. Environmental risk of dissolved oxygen depletion of diverted flood waters in river polder systems—a quasi-2D flood modelling approach. Sci. Total Environ. 407 (5), 1598–1612. Mao, J., Chen, Q., Chen, Y., 2008. Three-dimensional eutrophication model and application to Taihu Lake, China. J. Environ. Sci. 20 (3), 278–284. Neitsch, S., Arnold, J., Kiniry, J., Williams, J., King, K., 2005. SWAT theoretical documentation. Soil Water Res. Lab. Grassl. 494, 234–235. Olli, G., Darracq, A., Destouni, G., 2009. Field study of phosphorous transport and retention in drainage reaches. J. Hydrol. 365 (1), 46–55. Ongley, E.D., Zhang, X., Yu, T., 2010. Current status of agricultural and rural non-point source pollution assessment in China. Environ. Pollut. 158 (5), 1159–1168. Park, R.A., Clough, J.S., Wellman, M.C., 2008. AQUATOX: modeling environmental fate and ecological effects in aquatic ecosystems. Ecol. Model. 213 (1), 1–15. Pei, H., Ma, J., 2002. Study on the algal dynamic model for West Lake, Hangzhou. Ecol. Model. 148 (1), 67–77. Puijenbroek, P., Janse, J.H., Knoop, J.M., 2004. Integrated modelling for nutrient loading and ecology of lakes in the Netherlands. Ecol. Model. 174 (1-2), 127–141. Renard, K.G., Foster, G.R., Weesies, G.A., Porter, J.P., 1991. RUSLE: revised universal soil loss equation. J. Soil Water Conserv. 46 (1), 30–33. Robson, B.J., 2014. State of the art in modelling of phosphorus in aquatic systems: review, criticisms and commentary. Environ. Model. Softw. 61, 339–359. Saltelli, A., Annoni, P., 2010. How to avoid a perfunctory sensitivity analysis. Environ. Model. Softw. 25 (12), 1508–1517. Shahsavani, D., Grimvall, A., 2011. Variance-based sensitivity analysis of model outputs using surrogate models. Environ. Model. Softw. 26 (6), 723–730. Shen, Z., Liao, Q., Hong, Q., Gong, Y., 2012. An overview of research on agricultural nonpoint source pollution modelling in China. Sep. Purif. Technol. 84, 104–111. Silvan, N., Vasander, H., Laine, J., 2004. Vegetation is the main factor in nutrient retention in a constructed wetland buffer. Plant Soil 258 (1), 179–187. Taguchi, K., Nakata, K., 2009. Evaluation of biological water purification functions of inland lakes using an aquatic ecosystem model. Ecol. Model. 220 (18), 2255–2271. Tetra Tech, Inc., 2007. The environmental fluid dynamics code: theory and computation. US EPA, Fairfax, VA. Vymazal, J., 2007. Removal of nutrients in various types of constructed wetlands. Sci. Total Environ. 380 (1-3), 48–65. Wade, A., Whitehead, P., Butterfield, D., 2002. The integrated catchments model of phosphorus dynamics (INCA-P), a new approach for multiple source assessment in heterogeneous river systems: model structure and equations. Hydrol. Earth Syst. Sci. 6 (3), 583–606. Whitehead, P.G., Jin, L., Crossman, J., Comber, S., Johnes, P.J., Daldorph, P., Flynn, N., Collins, A.L., Butterfield, D., Mistry, R., Bardon, R., Pope, L., Willows, R., 2014. Distributed and dynamic modelling of hydrology, phosphorus and ecology in the Hampshire Avon and Blashford Lakes: evaluating alternative strategies to meet WFD standards. Sci. Total Environ. 481, 157–166. Xue, L., Yu, Y., Yang, L., 2014. Maintaining yields and reducing nitrogen loss in rice–wheat rotation system in Taihu Lake region with proper fertilizer management. Environ. Res. Lett. 9 (11), 115010. Zhao, G., Hörmann, G., Fohrer, N., Li, H., Gao, J., Tian, K., 2011b. Development and application of a nitrogen simulation model in a data scarce catchment in South China. Agric. Water Manag. 98 (4), 619–631.