Material and methods

Study area

The study area (Xitiaoxi Watershed) was located within Lake Taihu Basin in eastern China (Fig. 1). The watershed included 24 mountain sub-watersheds (30°23’N~30°45’N and 119°14’E~119°48’E), and 143 lowland artificial sub-watersheds (30°45’N~31°5’N and 119°48’E~120°9’E). The elevations range from 0 m (lowland) to 1576 m (mountain) above sea level.
The mountain watersheds are located at southwest of the study area, with areas of 1,367 km2. River Xitiaoxi contributed 27.7% of the water resources in Lake Taihu (Chen et al., 2019). Annually average precipitation is 1465.8 mm (Chen et al., 2019). Forestlands account for 74% of the mountain watersheds, following 15% paddy lands, 4% residential areas and small areas of surface water and grasslands.
The lowland watersheds (polders) are located at the northeast with a total area of 743 km2. To develop agriculture and protect villages from floods, these polders are enclosed by dikes. Therefore, water exchange between rivers and polders are manually controlled by pumps. The main cultivate land (69%) is paddy field with rice–wheat rotation, other land use types of polder systems contain dry lands (14%), ponds (6%) and residential areas (11%). During rice seasons (from May to Nov.), supplementary water source from surrounding rivers need to pour into paddy lands for irrigation. Meantime, artificial drainage is closed, and excess runoffs are delivered into ponds for keeping the soil moisture saturation. In case that the water level of ponds is too high to harm the crop growth during heavy rainfall events, pond water will be exported into surrounding rivers by pumps. During wheat season, runoff water would not kept in ponds due to its useless for wheat.
Fig. 1 Location of the study area (Xitiaoxi Watershed) with the distribution of hydrological and weathers stations.

Data

The required dataset for these two models included meteorological and hydrological data, land use, and Digital Elevation Model (DEM). 1) Meteorological data were collected based on the national weather station (Huzhou: No.58450). The precipitation data were obtained from 13 automatic rain gauges (Fig. 1). The pan evaporation was substituted by the reference evapotranspiration (ET0) using Penman Equation. 2) Among hydrological data, daily water level was measured using a water level logger. The discharge for mountain watersheds was verified using the daily runoff data from 2009 to 2012 at the national station of Hengtangcun. 3) Land use and digital elevation model (DEM) with a spatial resolution of 30 m were obtained from the satellite image interpretation of 2010 and International Scientific and Technical Data Mirror Site, Computer Network Information Center, Chinese Academy of Sciences (http://www.gscloud.cn) respectively. Above-mentioned data were rescaled to an identical resolution of 100 m for hydrological modelling. We used DEM to delineate the boundaries of the mountain watersheds and to classify the slope.
Table 1 Data collected for the Xin’anjiang model and Nitrogen Dynamic Polder (NDP) model.

Model description

Two hydrological models (Xin’anjiang and NDP) were used to simulate the hydrological processes in the mountain watersheds and lowland artificial watersheds, respectively, and to compare their different responses to climate and land use change. Xin’anjiang model, a widely used hydrological model for simulating streamflow in the humid and semi-humid regions in China, was used to simulate discharge for mountain watersheds at a daily time scale. NDP model was specially developed by Huang et al. (2018a) to describe the unique processes of hydrological and nitrogen dynamics in lowland polders. In this study, it was used to distinguish the water balance components in lowland watersheds.

Raster-based Xin’anjiang model

The Raster-based Xin’anjiang model was developed based on the original Xin’anjiang model originally developed by Zhao (1984), incorporating the merits of the original conceptual rainfall-runoff model (Huang et al., 2018b). Its core concept was runoff formation on the repletion of storage capacity, which implied that the excess rainfall became the runoff until the soil water content of the aeration zone reached its field capacity (Yao et al., 2014). For each cell, the outflow was simulated based on four modules: evapotranspiration module, runoff generation module, runoff separation module and runoff routing module. Total runoff was separated into three components including surface, interflow, and groundwater runoff.
To develop the raster-based Xin’anjiang model for the mountain watersheds, the following inputs were required: 1) initial data : it included initial discharge of surface water, interflow and groundwater, tension water storage and free water storage. A spatial resolution of 100×100 m was used for the model. 2) Forcing data : weather conditions included the variables of sunshine hours, wind direction, wind speed, average air temperature, precipitation, maximum and minimum air temperature in a day. 3) Boundary data : the daily runoff and the locations of streamflows should be input. 4) Model parameters : there were 14 parameters in this model. Value ranges of these parameters as well as their calibrated values were given in Supporting Information.

NDP model

The model was developed based on water balance equations in four land-use types (residential area, surface water area, paddy and dry land). It included the hydrological processes of precipitation, irrigation, evaporation, evapotranspiration, surface runoff, infiltration, water exchange between groundwater and surface water. Notably, water management modules describing the artificial drainage of polder systems were also included. Flood drainage, culvert drainage and seepage acted as outflow pathways to surrounding rivers.
The initial conditions of NDP included the areas of four land use types in each polder. In case that hydrological component was selected, the initial water level and land use types were required. Input data included time series meteorological data and parameters. NDP included 28 parameters in the water balance and water management modules, all the parameter values were obtained from previous studies in a typical polder located not far (about 38 km) from the study area (See Supporting Information).

Forecasting hydrological response in the context of climate and land use change

The Scenario Model Intercomparison Project (ScenarioMIP) is the primary activity within Phase 6 of the Coupled Model Intercomparison Project (CMIP6) that will provide multi-model climate projections based on alternative scenarios of future emissions and land use changes produced with integrated assessment models. BCC-CSM2-MR configured for CMIP6 are used for generating precipitation and temperature for three developed alternative future societal development pathways (the SSPs) and emissions and land use scenarios based on RCPs (O’Neill et al., 2016). Historical simulations from 2000 to 2014 were as the reference period, and four distinct periods including 2029~2032, 2049~2052, 2069~2072 and 2089~2092 were chosen for future scenario runs representing the 2030s, 2050s, 2070s and 2090s, respectively. The predicted climate scenarios indicated a 3.45% increase in annual precipitation for the 2030s, a 7.54% increase for the 2050s, a 16.18% increase for the 2070s, and a 14.43% increase for the 2090s. Temperature outputs indicated a 0.98 °C increase for the 2030s, a 1.63 °C increase for the 2050s, a 2.39 °C increase for the 2070s and a 3.03 °C increase for the 2090s. Compared with BCC-CSM1.1m from CMIP5, BCC-CSM2-MR shows significant improvements in many aspects including the tropospheric air temperature and precipitation at global and regional scales in East Asia (Wu et al., 2019), which fit for our research.
Developed land use scenarios for the 2050s are all under the ‘current rate’ according to Chen et al. (2009). In mountain watersheds, the most frequent land use changes will occur in expanding 272.4% residential area at the expense of cultivate land and forest land by 86.5 km2 and 50.8 km2 respectively, following by the conversion between cultivate land and forest land. In lowland artificial watershed, converting 36.1% (221.7 km2) cultivate land to residential land is overwhelming than other land-use conversion patterns (See Supporting Information).