Modeling the hydrological effects of climate and land use/cover changes in Chinese lowland polder using an improved WALRUS model

Hydrological processes in lowland polders, especially those for paddy rice planting, are affected by complicated factors. The improved Wageningen Lowland Runoff Simulator (WALRUS) model incorporates an irrigation and drainage scheme, and a new stage–discharge relationship to account for hydrological processes in multi-land-use polder with paddy fields and pumping stations. Here, this model was applied to assess how climate and land use changes affected the runoff of a Chinese polder in Poyang Lake basin in the past two decades. Simulated results showed that the runoff in the autumn–winter transition and midsummer months increased significantly, whereas those in the other months decreased slightly during the period of 1996–2005, primarily affected by climate change. For the period of 2006–2014, the runoff in the autumn–winter transition and midsummer increased, while that in the other months declined, affected by both climate and land use/cover changes. The land use/cover change resulting from the conversion of rice–wheat rotation to dominantly double-rice cropping and the expansion of residential area, increased the runoff during this period by demanding more irrigation water from the outside basin. This is an Open Access article distributed under the terms of the Creative Commons Attribution Licence (CC BY 4.0), which permits copying, adaptation and redistribution, provided the original work is properly cited (http://creativecommons.org/licenses/by/4.0/). doi: 10.2166/nh.2016.204 om http://iwaponline.com/hr/article-pdf/47/S1/84/367291/nh047s10084.pdf er 2020 Renhua Yan Junfeng Gao (corresponding author) Lingling Li Key Laboratory of Watershed Geographic Sciences, Nanjing Institute of Geography and Limnology, Chinese Academy of Sciences, 73 East Beijing Road, Nanjing 210008, China E-mail: gaojunf@niglas.ac.cn Renhua Yan School of Geographical Sciences, Southwest University, Chongqing 400715, China Renhua Yan Lingling Li University of Chinese Academy of Sciences, Beijing 100049, China


INTRODUCTION
, temperature (Syvitski et al. ), and seepage (Immerzeel et al. ) are closely related to climate and land use/cover changes.
These changes exacerbate catchment damage from extreme weather events (e.g. droughts and floods). The low elevation and flat terrain of lowland polders, as well as intensified human activities, have made lowland polders more vulnerable to the effects of droughts and flooding linked with both changes (Brauer et al. a). Additionally, the hydrological response of polders to the climate and land use/cover changes will also have a direct effect on the transformation and delivery of pollutants into the downstream rivers, consequently affecting lake water quality because water flow is a carrier of agricultural pollutants. Therefore, better study of the influences of land use/cover and climate changes on polder hydrology is crucial to mitigating natural disasters, reducing eutrophication, and advancing water resources  (Yan et al. a). The simulated results from the calibrated models can compensate for the scarcity and discontinuity of hydrological datasets. However, these existing models have been designed mainly for freely draining catchments with sloping surfaces rather than for flat polders with shallow groundwater and complicated water management operations, because some polder-specific hydrological characteristics (e.g. capillary rise and effect of surface water on groundwater table) are not explicitly considered. Faced with these problems, Brauer et al. (b) recently proposed a lumped rainfall-runoff model called the Wageningen Lowland Runoff Simulator (WALRUS) to account for three essential processes in lowland polders: groundwater À surface water feedback, saturated and unsaturated zone coupling, and wetness-dependent flow routes. Based on this model, an improved WALRUS model was developed for application in multi-land-use polder with paddy fields and pumping stations in Asian polders. The improved WALRUS model describes the runoff from different sources (e.g. paddy fields, residential areas, dry farmlands, and water areas) and introduces an irrigation and drainage scheme to control water management in paddy fields.
In this study, the proposed model is applied to quantify and discriminate the hydrological consequences of climate and land use/cover changes in the large polder of Poyang Lake basin in East China in the past few decades. The objectives of this study were to (1) simulate the response of the seasonal and annual runoff to climate change alone and land use/cover change alone, respectively, and (2)

Model description
The improved WALRUS model was developed in R and conceptually based on the WALRUS model. The user can choose at which times output should be produced (e.g. hourly, daily or non-equidistant). Internally, the computation time step is reduced if necessary (e.g. too much rain or water level variation). This variable time step approach increases stability. The model consists of four compartments: dry farmland, paddy field, water area (surface water), and residential area.
The dry farmland and paddy field include two reservoirs: a quickflow reservoir and a soil reservoir (coupled vadose-groundwater reservoir). Precipitation is the most important water source. At the field surface, precipitation is divided by the soil wetness index (W ) into a portion that percolates to the soil matrix (P V ) and another portion that is directly led to the surface water through a quick flow path (P Q ). Water is depleted by evapotranspiration from the vadose zone (ET V ). The dryness of the vadose zone is expressed by a state variable of storage deficit (d V ), that denotes the amount of water required to saturate the soil profile and determines the evapotranspiration reduction (β) and wetness index (W ). Taking dry farmland as an example, the relation between evapotranspiration reduction (β 2 ) and storage deficit (d V2 ) is specified as follows: Thus, the evapotranspiration of the vadose zone (ET V2 ) is calculated as follows: where ET pot2 denotes the potential evapotranspiration of the dry farmland, a G2 the dry farmland area fraction, and ζ 1 and ζ 2 the two parameters. A detailed description of these variables and parameters is listed in Table A1 (available with the online version of this paper) and can also be referred to Brauer et al. (b).
The groundwater table (d G ) has a dynamic response to change in storage deficit and controls in combination with the surface water level (hs), groundwater drainage or infiltra- can be either positive or negative depending on the differences in water level between the surface water (hs) and groundwater (d G ), that considers the groundwater-surface water interaction and infiltration of surface water into the soil reservoir with higher surface water level. All water that does not pass through the soil matrix is delivered to the surface water through a quickflow route ( f QS ), which represents overland flow, local ponding, and macropore flow. Seepage ( f XG ) is also taken into account in the model and represents the water being added to or subtracted from the soil reservoir. The aforementioned processes are shared by the dry farmland and paddy field. In addition, a water management scheme with three critical water depths (lower limit of appropriate depth (h Q,min1 ), upper limit of appropriate depth (h Q,max1 ), and maximum submergence-tolerant water level (h Q,flood1 ) for rice growth) (Guo is incorporated into the model to control the irrigation and drainage in the paddy field. The water depth in the field surface (i.e. level of quickflow reservoir of paddy field h Q1 ) should be kept within the range between the h Q,min1 and h Q,max1 to provide preferable moisture condition for rice growth. In case the water depth in the field surface (h Q1 ) drops below h Q,min1 , denoting that moisture might threaten rice growth, the irrigation operation is required to execute until the depth reaches h Q,max1 . In contrast, with extreme rainfall event occurring, in case the water depth exceeds h Q,flood1 , field drainage operation needs to be implemented. In other words, the quickflow in the paddy field ( f QS ) will be activated only if the depth of the quickflow reservoir exceeds h Q,flood1 . We should note that the three critical depths are changeable over the cropping regime and growth stages of paddy rice (Tables 2 and 3).
For the residential area, which is considered as an impervious area, precipitation is assumed not to percolate into the soil matrix but directly routes into the adjoining Date 6/21-6/30 7/1-7/20 7/21-8/3 8/4-8/10 8/11-9/3 9/4-9/10 9/11-9/25 9/26-10/10  For the water area, water inputs include precipitation (P s ), the drainages of paddy field ( f QS1 þ f GS1 ) and dry farm- , and residential area ( f rS ). Water consumption terms consist of evapotranspiration (ET s ) and catchment runoff (Q). Catchment runoff (Q) is calculated as a function of surface water level (h S ) and threshold water level to start pumping (hS start pump ): where Q pump is obtained from converting the pump capacity (m 3 /s) to mm/h by dividing the polder area.
The modeling approach is shown schematically in Figure 2. Table A1 displays the model equations, variables, and parameters.

Model calibration and validation
There are four types of model parameters including the wetness index parameter (c w1 and c w2 ), vadose zone relaxation time (c v1 and c v2 ), groundwater reservoir constant (c G1 , c G2 , and c G3 ), and quickflow reservoir constant (c Q1 and c Q2 ), that need to be optimized to represent the catchment-specific characteristics. Based on the field investigation, the threshold water level to start pumping (hS start pump ) and ditch depth (c D ) in the study area were estimated to be 1,000 and 1,500 mm, respectively. The three critical water depths for the paddy field irrigation and drainage scheme are listed in Tables 2 and 3 The main input data for the calibration and validation are shown in Figure 3.

Model scenarios and simulation
According to the study of Tian (), the annual hydrometeorology variables of the Poyang Lake basin (Jiangxiang polder is subordinate to this basin), especially temperature and precipitation, tended to increase since the 1990s and their abrupt change points were identified around 1996 by using the non-parametric Mann-Kendall trend test (Mann ; Kendall ) and Mann-Kendall-Sneyers method  The improved WALRUS simulation of runoff was implemented for the three periods. In order to assess the hydrological impacts of historical land use/cover and climate changes, five scenarios (Table 4) were established and were modeled with the calibrated model as follows: Scenario 1, for which the climate data during the period 1986-1995 and the land use data from 1990 were utilized, was regarded as the conditions for the baseline period (1986)(1987)(1988)(1989)(1990)(1991)(1992)(1993)(1994)(1995). In Scenario 2 and Scenario 3, the land use were quantified using the following equations:

RESULTS AND DISCUSSION Calibration and validation
Most parameters values were preset by referring to the values reported for the Jianwei polder, Taihu basin, improved WALRUS application reported by Yan et al.
(b), and then were calibrated with a trial-and-error method ( Table 5). The calibration and validation results are graphically presented in Figures 4 and 5. Moreover, the daily simulated values are shown in Figure 6.       [1986][1987][1988][1989][1990][1991][1992][1993][1994][1995]. A large variation in the monthly runoff was found between the latter two periods and the baseline period, indicating that climate change had a considerable effect on the monthly runoff for the two latter periods. The runoff significantly increased for the autumn-winter transition (November-January) and midsummer (August) months. For example, the runoff for both periods increased by approximately 200% compared to that for the baseline period in December. Conversely, the runoff in the other months, including early spring and autumn, slightly decreased at a rate lower than 50%.

Climate change impact
This seasonal variation in the climate effect is the combined result of the changes in precipitation and temperature ( Figure 8). In the autumn-winter transition and midsummer months, the precipitation in the latter two periods was much larger than those in the baseline period. Furthermore, the lower temperature contributed to the decrease in evapotranspiration relative to the baseline period. Hence, the expanded net rainfall caused an increase in the monthly runoff for the later two periods.
In contrast, a decrease in monthly precipitation during the early spring and autumn was accompanied by an increased temperature, which in turn resulted in an increase in evapotranspiration from the baseline period.
Thus, the runoff during early spring and autumn was reduced by the reduction in precipitation and increase in evapotranspiration. However, the effect of precipitation change on streamflow was larger than that of temperature in terms of the degree of similarity between the trends in  climate factors and runoff. That might also be proven by the correlation of runoff with climate factor changes, which implied that although the two climate factors were statistically significant as related to runoff, the runoff was more correlated with precipitation (R ¼ 0.706, p ¼ 0.000) than with the temperature (R ¼ À0.466, p ¼ 0.022) for both the periods. Moreover, it is remarkable that the precipitation increased during the early summer (June and July), whereas the runoff decreased. It might be attributable to the buffer capacity of the soil. Soil moisture was depleted more rapidly by evapotranspiration in the drier spring than in the baseline period (as evidenced by the increased storage deficit), which provided a buffer capacity for the increased early summer precipitation (Figure 8(c)).
As a result, although the early summer period experienced high precipitation, little runoff occurred because the increased rainfall was stored in the soil to compensate for the excessive depletion of water in the previous

Combination of both impacts
In order to evaluate the combined impact of climate and land use/cover changes, the comparison of the runoff during the latter two periods (Scenarios 4 and 5) and in the baseline period (Scenario 1) are depicted in Figure 11.

Model performance and uncertainties
The model is relatively simple, flexible, and computationally efficient, hence it shortens computation time and saves computation resources when modeling complicated scenarios.
More importantly, the model is capable of running with lim- This result would ultimately degrade the water quality of the entire floodplain basin, and even probably induce the eutrophication of Poyang Lake. Therefore, the results of this study have implications for the management of water quality and aquatic ecosystem health in the floodplain basin.

CONCLUSIONS
In this study, an improved WALRUS model was applied to evaluate and discriminate the effects from observed histori- Despite the uncertainties of this study, the model results and approaches proposed might have some implications for flood control and water quality management, and may provide guidance in the design of adaptation strategies for dealing with environmental issues arising from climate and land use/cover changes. Further efforts will be oriented towards validating the model further by using long-term daily datasets and appropriate soil moisture parameter values for a specific study area from location-specific soil samples.