Application of a distributed catchment model to investigate hydrological impacts of climate change within Poyang Lake catchment (China)

The increasing atmospheric concentrations of greenhouse gases, especially carbon dioxide, that have led to warming of the earth's surface (Feng et al. 2014) and associated climate change, are predicted to modify the world's terrestrial hydrological systems, with a strong impact on freshwater resources (Intergovernmental Panel on Climate Change (IPCC) 2013). Climate change is expected to vary spatiotemporal distributions of rainfall and snowfall in many regions, including increased frequency and severity of extreme climate events (Liuzzo et al. 2010; Jung et al. 2012; IPCC 2013; Emam et al. 2015). Actually, there is evidence that hydrological systems have already begun to respond to global warming in recent decades (e.g., Christensen et al. 2007; Bates et al. 2008; Murphy & Ellis 2014).

The availability of catchment water resources is closely linked to variations in climate. Climate change affects catchments through the direct influence of atmosphere drivers (e.g., temperature, precipitation and wind speed) and associated changes in catchment hydrological processes (Kienzle et al. 2012). While several climate variables are influenced by climate change, precipitation and temperature are the two dominant factors affecting the variations of river discharge on the catchment scale (Liu et al. 2011). Catchment discharge behavior may be considered as a sentinel for regional water cycles, by exhibiting a particularly robust signal that integrates climate and landscape stresses and that reflects significant changes in these (e.g., Cornelissen et al. 2013; Xu et al. 2013; Fan & Shibata 2015). However, quantifying the catchment hydrological response to climate variations is challenging, given the inherent complexity of water flow pathways within watersheds (e.g., Teutschbein & Seibert 2012; Musau et al. 2015; Vezzoli et al. 2015).

The catchment hydrological response to climate change has been studied through the application of hydrological models using atmospheric inputs from various climate models (e.g., Xu et al. 2005; Chen et al. 2012; Kienzle et al. 2012; Awan et al. 2016). From a theoretical viewpoint, a physically based catchment model represents the underlying hydrological and land surface processes in greater detail than conceptual and statistical models (Beven 2001). Spatially distributed catchment models are ideally suited to assessing the impacts of climate change on both surface and subsurface hydrological responses, given the critical interdependencies between surface water and groundwater systems (Goderniaux et al. 2009). For example, Graham et al. (2011) and Kienzle et al. (2012) used the distributed ACRU (Agricultural Catchments Research Unit) agro-hydrological model to simulate the impacts of climate change in the Thukela River basin (South Africa) and the Cline River catchment (Canada), respectively. Cornelissen et al. (2013) and Fan & Shibata (2015) employed the semi-distributed SWAT (Soil and Water Assessment Tool) model to assess the impact of climate and land-use changes on river discharge in a tropical catchment (West Africa) and the Teshio River catchment (Japan), respectively. Hydrological projections and their associated applications to river discharge remain an important topic given increased confidence in changing climatic patterns around the globe.

During the past two decades, numerous large-scale climate models have been developed for global and/or regional hydrological assessment and modeling (Li et al. 2015a). Many previous studies have attempted to develop methods to downscale global climate model predictions to finer resolutions that better account for important spatial variability within catchments. Such methods can be classified into four types: (1) simple interpolation of the climate model data to a finer spatial scale, (2) statistical downscaling using empirical relationship between coarse-scale model output and local climate, (3) regional climate models (RCMs), and (4) global atmospheric general circulation models with high or variable resolution (Mearns et al. 2001; Arnell et al. 2003). Among these, RCMs have become one of the most popular methods to impart higher-resolution catchment characteristics, e.g., geographical features, vegetation variability, topography, etc., in simulations of both current and future climate patterns (Li et al. 2015a). Thus, RCMs are widely used to generate atmospheric input to hydrology models of a wide range of catchment scales (e.g., Bergstrom et al. 2001; Graham et al. 2011; Chen et al. 2012).

Poyang Lake, the largest freshwater lake in China, is located in the middle reach of the Yangtze River, and has a catchment area of 1.62 × 10⁵ km² (Shankman et al. 2006). The lake and its surrounding catchment are in a wet, subtropical climate zone, which has experienced more frequent and intensive floods and droughts in recent decades (e.g., Shankman et al. 2006; Zhang et al. 2011). Poyang Lake's water balance is controlled primarily by catchment inflows and discharge to/from the Yangtze River. Inflows from Poyang Lake's catchment are the primary drivers of seasonal variations in the lake's water level (Shankman et al. 2006; Hu et al. 2007).

Changes to the hydrology of the region, and the subsequent societal, economic and environmental impacts, have attracted considerable research attention (e.g., Yin & Li 2001; Zhang et al. 2015). Previous studies of Poyang Lake's catchment have drawn a number of consistent conclusions regarding its historical and projected future response to climate change. For example, Ye et al. (2013) considered coupled water and energy budgets in differentiating the relative impacts of climate change and human activities on historical variations in river discharge from the Poyang Lake catchment during 1960–2007. Their empirical analysis found the annual river discharge has been modified primarily by climate change, while human activities within the catchment play a secondary role. Guo et al. (2008) used the SWAT model to examine climate (1953–2002) and land-use (2000) effects on river discharge in the Xinjiang River basin of Poyang Lake, and found that climate variability is the main driver of annual river discharge trends, while land-use change may have a moderate impact. Zhao et al. (2010) used statistical analysis (for 1951–2003) to arrive at similar conclusions to those reported by Guo et al. (2008). These previous studies indicate that the variations of river discharge in Poyang Lake catchment are strongly related to regional climate variability and change, while human activities such as land-use change also have significant impact.

Limited attempts to predict the hydrological impact of future climate change on Poyang Lake's river discharge have been made. Ye et al. (2011) developed a regression model based on the relationship between Poyang Lake water levels and the catchment inflows and temperature to predict lake water level changes under the future time period of 2011–2050 by adopting the global climate model ECHAM5. For the lake's catchment, their investigation mainly focused on the changes of the total inflows by using the catchment model WATLAC (Zhang & Werner 2009; Ye et al. 2011), which may not be adequate to provide critical insights into the discharge behavior at sub-catchment scale regarding the spatial flow variability (Guo et al. 2007). Sun et al. (2012) employed a water balance model to evaluate the future change (2061–2100) of annual river discharge in the Ganjiang River and Xinjiang River basins of Poyang Lake. Their investigation mainly focused on the upstream mountainous regions and concluded that the annual river discharge shows an uptrend due to projected increases in both precipitation and evapotranspiration. More recently, Yan et al. (2015) combined a raster-based Xin'anjiang model and the Coupled Model Inter-comparison Project Phase 5 (CMIP 5) multi-model ensemble to simulate future river discharge under three climate change scenarios and two land-use change conditions in the Xinjiang River basin of Poyang Lake. They found that future climate and land-use changes affect not only the seasonal distributions of river discharge, but also the annual amounts of discharge. There is an urgent need to provide water resource managers and policy makers with detailed information on the potential impacts of future climate change on river discharge across the Poyang Lake catchment.

The purpose of this study is to quantify potential impacts of future climate change, estimated from an RCM, on the hydrology of the extensive Poyang Lake catchment. Our specific objectives are to: (1) calibrate and evaluate the reliability of a coupled catchment model for the baseline period (1986–2005); (2) analyze spatiotemporal changes in projected precipitation and temperature for the near future (2016–2035) under three IPCC scenarios; and (3) investigate the hydrological responses of river discharge in the different sub-catchments of Poyang Lake under future climate scenarios. We extend on previous studies of climate change impacts on Poyang Lake catchment by using a distributed hydrological model with automatic calibration scheme and a high-resolution RCM.

The Poyang Lake catchment has a mean annual precipitation and potential evapotranspiration of 1,666 mm/yr and 1,034 mm/yr, respectively (Li et al. 2014). Precipitation records show distinct wet and dry seasons, with 45% of annual precipitation concentrated in the wet season from April to June, and only 16% of rainfall occurring during September to December, on average. Temperatures are also highly seasonal, with a June-August average of 27.3 °C, a December-February average of 7.1 °C, and an annual average of 17.6 °C (Zhang et al. 2015). A digital elevation model of the catchment, derived from the National Geomatics Center of China, shows that the topography varies from upstream mountainous areas at elevations of 2,100 m (above sea level) to downstream plain areas around the lake at elevations of about 32 m (Figure 1). The land-use data (for 2005), obtained from the Department of Soil Survey of Jiangxi Province, can be categorized into forest (61%), farmland (28%), water bodies (5%), pasture (4%) and urbanization (2%) (Li et al. 2014). The soils in the whole catchment are classified according to the Genetic Soil Classification of China (Shi et al. 2004), and have the following catchment-aggregated proportions: red soil (43%), latosol (23%), paddy soil (21%), yellow soil (7%), alluvial soil (3%) and others (3%) (Ye et al. 2011).

Measured daily precipitation and temperature from 14 weather stations inside Poyang Lake catchment (Figure 1), all of which are standard national weather stations, were obtained from the National Climate Center of China Meteorological Administration. Observations of daily river discharge at the most downstream gauging stations of the Ganjiang, Fuhe, Xinjiang, Raohe and Xiushui Rivers (Figure 1) were used to represent runoff from the five sub-catchments of Poyang Lake. These data were obtained from the Hydrological Bureau of Jiangxi Province and the Hydrological Bureau of the Yangtze River Water Resources Commission of the Ministry of Water Resources of China. Hydro-climatic data are available at daily time steps for the period 1986–2005. The basic features of hydro-climatic records are listed in Table 1.

The commonly used scenarios, called Representative Concentration Pathways (RCP), are designed to accommodate a wide range of possibilities in social and economic development consistent with specific radiative forcing paths (Kharin et al. 2013). We consider future climate conditions predicted for the scenarios RCP2.6 (low emission), RCP4.5 (intermediate emission) and RCP8.5 (moderately high emission), in terms of precipitation and temperature outputs from the RCM COSMO-CLM (Steppeler et al. 2003). COSMO-CLM, developed by the Potsdam Institute for Climate Impact Research, has been successfully validated for the East Asian monsoon region (Fischer et al. 2013; Wang et al. 2013). COSMO-CLM uses a grid with a resolution of 0.5 ° over the East Asian domain. Details of COSMO-CLM are described in the above references and are omitted here for brevity.

When using the precipitation and temperature outputs of RCMs, it is necessary to implement a bias correction. Bias in RCM simulation results can arise from shortcomings of the RCMs themselves, but also from erroneous forcing data (Schoetter et al. 2012; Teutschbein & Seibert 2012). The simulations and projections of COSMO-CLM are decomposed for the baseline period 1986–2005, which is used to evaluate and bias-correct the model. The corrections are then applied to the emission scenarios (RCP2.6, 4.5 and 8.5) for the projection period 2016–2035 (Rockel et al. 2008; Huang et al. 2015).

In this study, the systematic biases in the climate variables are corrected for the COSMO-CLM projections by using the equidistant cumulative distribution function (EDCDF) matching method proposed by Li et al. (2010). The EDCDF method is based on transfer functions that are generated to map the distribution of the simulated historical data to that of the observations. For a given percentile, this method assumes that the differences between the modeled and observed values of the baseline period also apply to the projected future period, which means that the adjustment function remains the same. The method can be used to produce auxiliary ensemble scenarios for various climate impact-oriented applications. Details of the mathematics and the associated parameters are described by Li et al. (2010) and Huang et al. (2015).

The catchment model WATLAC (Zhang & Li 2009; Zhang & Werner 2009) is a spatially distributed hydrological model that was developed for large-scale, physically based, continuous rainfall-runoff simulation. The model includes such processes as canopy interception, overland flow, stream flow routing, soil lateral flow, evapotranspiration, groundwater recharge and groundwater flow. Precipitation and potential evapotranspiration are the main driving forces for WATLAC. The model represents landscape and stream flow hydrology at a daily time step. WATLAC has already been successfully applied to a number of case studies, including the Fuxianhu Lake catchment (Zhang & Werner 2009), the Xitiaoxi catchment of Taihu Lake (Zhang & Li 2009), and the Poyang Lake catchment (Ye et al. 2011; Li et al. 2014). WATLAC is described in detail in the above references, and therefore only a brief description is given here.

The WATLAC model used in the present study is the same as the model used in a previous Poyang Lake catchment investigation by Li et al. (2014). Li et al. (2014) improved the Poyang Lake catchment model of Ye et al. (2011) by adopting a finer 1 km × 1 km grid and an automatic calibration software, PEST (Parameter ESTimation) (Doherty 2005), to better capture the heterogeneity of the land surface and reduce the calibration error. They used PEST to modify key model parameters: overland runoff (α), soil infiltration coefficient (β₁), groundwater recharge coefficient (β₂), flood travel time (k) in the Muskingum flood routing method, and specified yield (S_y) of the shallow unconfined aquifer. Li et al. (2014) calibrated and validated the Poyang Lake catchment model for the time sequence of 2000–2008, and produced a satisfactory accuracy for the river discharges at the Waizhou, Xiajiang, Lijiadu, Meigang, Shizhenjie and Wanjiabu gauging stations. These results demonstrated that the model was able to simulate the catchment's rainfall-runoff behavior.

A temperature-based approach (i.e., Hamon 1961) was employed to convert RCM output to potential evapotranspiration for input to WATLAC, rather than pan evaporation (as used by Li et al. (2014)). This allowed for analyses of the influence of temperature changes on catchment hydrology, pertaining to past and future time periods. Other WATLAC model aspects are essentially the same as the approach of Li et al. (2014), and are described in detail therein. The effects of climate change on river discharge are often assessed by calibrating and validating hydrological models for current climate conditions, and hence the differences between the past and future climate change scenarios of model simulations are quantified. In the current study, the previous WATLAC model was re-calibrated using time series of precipitation and temperature derived from 14 national meteorological stations (Figure 1). These data cover the baseline period for both the model calibration (1986–1995) and validation (1996–2005). The river discharge at Waizhou, Lijiadu, Meigang, Shizhenjie and Wanjiabu gauging stations (Figure 1), representing the Ganjiang, Fuhe, Xinjiang, Raohe and Xiushui Rivers, respectively, provided the observation dataset for WATLAC model calibration and validation. Calibration involved modifying the five parameters (i.e., α, β₁, β₂, k and S_y) within expected ranges using PEST until a reasonable match to river discharge was obtained.

The initial parameter values and parameter bounds were specified according to those adopted by Li et al. (2014), who used literature values. The optimized WATLAC parameters are listed in Table 3. These optimized values are close to previous calibration results obtained by Li et al. (2014), indicating consistency in the approach and reliability of catchment runoff predictions.

The annual temperature is expected to be higher than the mean temperature of 1985–2005 in all stations and climate scenarios, except the six most northerly stations in scenario RCP4.5 (Figure 4(b)). This is consistent with the findings of Sun et al. (2012), who show that, in general, precipitation and evapotranspiration (as a surrogate for temperature) are projected to increase in Poyang Lake catchment under future climate conditions. For the case of the mean annual temperature, it does not manifest a systematic increasing trend in the future time 2016–2035 under RCP2.6, 4.5 and 8.5 scenarios. That is, the magnitude of projected increases in mean annual temperature under RCP2.6 (>0.5 °C) is higher than that of the RCP4.5 and 8.5 (<0.5 °C) at some stations (Figure 4(b)), which is consistent with previous studies that concluded that the RCP2.6 may have a warmer global mean surface temperature than RCP4.5 and/or RCP8.5 despite a lower greenhouse gas forcing during 2018–2037 (Chalmers et al. 2012; Miao et al. 2014).

Relative to the precipitation changes, the monthly mean temperature tends to show a fairly consistent increasing trend throughout the year, except a slightly decreasing trend for the Xinjiang, Raohe and Xiushui Rivers with the RCP4.5 scenario (Figure 5). The RCP2.6 scenario appears to contribute most to the late winter (January-February), spring (March-May) and autumn (September-November) warming across the Poyang Lake catchment, while the RCP4.5 and 8.5 scenarios contribute the most to the summer (June-August) and early winter (December) warming. It is projected that the monthly mean temperature will vary from approximately 6.1 °C to 28.7 °C under future climate scenarios. The range of monthly temperature increase, averaged across the Poyang Lake catchment, is 0.3–1.4 °C for the RCP2.6 scenario, 0.2–0.7 °C for RCP4.5, and 0.2–1.2 °C for RCP8.5, indicating the strong likelihood of warmer temperatures in the near future. In all three emission scenarios, the increase in temperature is distinctly larger from November to April (up to 1.5 °C) compared to other months of the year (<1 °C), suggesting a warmer winter and spring in the future.

Table 5 summarizes the changes in mean annual river discharge in the five sub-catchments and the whole Poyang Lake catchment. The annual discharge change with respect to the baseline period has the range between −1.0% and 5.0% under the three climate scenarios. Taking the average of the three climate scenarios, the mean annual river discharge is projected to increase in all of the rivers, with the exception of the Xinjiang River. For the whole catchment, the mean river discharge is projected to increase by 1.2%, averaged across the three scenarios.

The potential implications of climate change on the hydrological system over the Poyang Lake catchment (China) have been assessed using the WATLAC model and the RCM COSMO-CLM. Previous studies have concluded that although the land-use change exerts some impacts on the river discharge, climate change is the primary factor causing the river discharge change of the Poyang Lake catchment (Guo et al. 2008; Zhao et al. 2010; Ye et al. 2013), and therefore, the effect of land-use/cover change on river discharge was not considered in this study. The major sources of uncertainty are the path of future emissions used in this study. However, there is no universally accepted methodology for evaluating the relative quality of RCMs with regard to their ability to project the future climate (Bader et al. 2008; Kienzle et al. 2012). In the current study, a suite of climate change scenarios was applied to cover a wider range of climate change uncertainties, which predict principally progressive future changes in the precipitation and temperature regime.

It is also acknowledged that obvious differences between climate observations and COSMO-CLM simulations were observed in this study. According to the bias-corrected scenarios, the projected climate variables and associated river discharges for different RCPs do not show a distinctly overall increase or decrease across the Poyang Lake catchment. However, the mean monthly temperature shows a fairly consistent increasing trend for different sub-catchments regardless of climate scenarios. The future climate in the catchment is projected to have an increased mean monthly temperature ranging from 0.3–1.4 °C, 0.2–0.7 °C, and 0.2–1.2 °C for RCP2.6, 4.5 and 8.5 scenarios, respectively. Although the variation pattern of monthly precipitation has slight differences between the RCPs, the results tend to show a general increase in precipitation in all seasons, with the exception of winter. Changes in the annual cycle of precipitation are mostly projected to increase in March-October, with the largest increase of up to 62% in September, while precipitation is predicted to decrease in November-February, with the largest decrease of up to 39% in January.

Hydrological simulations indicate that the projected increases in mean annual river discharge are expected to vary between 0.2% and 4% in the future time period, except for the decreasing trend for the Xinjiang River sub-catchment. The projected changes in precipitation and temperature are likely contributing to the decreased winter river discharges and increased summer river discharges. The projections also show a great likelihood of increases in high flows and decreases in low flows in terms of both frequency and magnitude, suggesting an increasing frequency and severity of flooding and droughts in the Poyang Lake catchment in the future climate. The broad range of projected river discharges is primarily the result of uncertainty in projections of future precipitation, and the variability of precipitation across the projections can be attributed to the effect of emissions pathways. Based on a wide range of RCMs, Bucchignani et al. (2014) concluded that all previous RCMs performed over China have non-negligible precipitation and temperature biases, mainly due to shortcomings of the RCMs in reproducing atmospheric dynamics and partly due to low quality of the gridded datasets used for validation.

The severe water shortages of Poyang Lake have aroused wide concern recently and raised questions such as whether the lake water levels will increase in the dry period under future climate change, which is of high importance for policy makers as it may affect various aspects of the lake (Dai et al. 2008; Zhang et al. 2015). Model simulations in Ye et al. (2011) indicate that the total inflows from the catchment will increase in the wet period and decrease in the dry period under future climate conditions, resulting in a possible change in water levels for the flood (0.1 to 1.34 m) and dry (−0.32 to −1.31 m) seasons in the lake. Considering the outcomes of this study, the significant decreases in winter river discharge for different rivers, resulting from changes in catchment climate, are more likely to further worsen the seasonal dryness of the lake. Thus, water resource managers and wildlife managers in Poyang Lake and its catchment may need to adapt to less water availability during the dry months, and as peak flows become more extreme, more works should be directed at quantifying their impact on existing infrastructure.

This work is jointly supported by the National Basic Research Program of China (973 Program) (2012CB417003), the Collaborative Innovation Center for Major Ecological Security Issues of Jiangxi Province and Monitoring Implementation (JXS-EW-00), the Natural Science Foundation of Jiangsu Province (BK20141061), and the National Natural Science Foundation of China (41401031). The authors are grateful to Adrian D. Werner from School of the Environment, Flinders University, for providing valuable suggestions on this paper.