1. Introduction
Global warming not only changes the mean state of climatic conditions, but it also alters climate extremes (Dankers and Hiederer 2008; IPCC 2013; Zhang et al. 2013a). Such changes have significant influences on the water cycle and water hazards such as water availability and floods (Vörösmarty et al. 2000; Rosenzweig et al. 2001). Floods are one of the most devastating natural disasters and have killed over 175 000 people and affected more than 2.2 billion people during 1971–2001 around the world (Jonkman 2005). Variations in water availability may increase conflicts between water supply and water demands, competition among water users, and probability of droughts. All of these have significant socioeconomic and environmental impacts (Oki and Kanae 2006; Wang et al. 2012).
As more and more evidence has proven that global warming can exacerbate climatic extremes, changes in floods and water availability across different regions of the world have been widely studied (e.g., Milly et al. 2005; Wilby et al. 2008; Raff et al. 2009; Ranger et al. 2011). However, it is difficult to generalize results from different studies because of the differences in the models and scenarios (Dankers et al. 2014). On the other hand, the Intergovernmental Panel on Climate Change (IPCC) special report on extremes states that the confidence in projections of changes in fluvial floods is low because of limited evidence and because the causes of regional changes are complex (IPCC 2012).
The newly released Inter-Sectoral Impact Model Intercomparison Project (ISI-MIP) attempts to estimate uncertainties based on the quantification of intermodel variations for both global climate models (GCMs) and impact models (IMs; Warszawski et al. 2014). ISI-MIP also endeavors to assess global climate change impacts in a consistent setting across multiple sectors and to initiate an ongoing coordinated impact-modeling improvement and intercomparison program. In ISI-MIP, IMs are driven by bias-corrected simulations of state-of-the-art GCMs from phase 5 of the Coupled Model Intercomparison Project (CMIP5) under a new generation of scenarios called representative concentration pathways (RCPs; Moss et al. 2010; Taylor et al. 2012; Warszawski et al. 2014). The first series of global assessments of floods, water scarcity, and water availability on the basis of ISI-MIP have been carried out very recently (e.g., Dankers et al. 2014; Elliott et al. 2014; Schewe et al. 2014). Although these studies have offered many insights for understanding worldwide changes, a comprehensive study at the country level is of great importance and demand for developing national strategies to manage hydrologic hazards and water resources.
As the most populous country with a vast territory, China highly depends on agriculture for food supply and has always been very sensitive to water availability and very vulnerable to floods and droughts. Nevertheless, few studies have been conducted so far to investigate the potential future changes in floods and water availability in China. Among a few previous studies, most of them used only one hydrologic model forced by one GCM or one regional climate model (RCM) with different settings, making it difficult to generalize results and quantify uncertainties from hydrologic modeling and climate modeling in different studies. Wang et al. (2012) used an RCM called the Providing Regional Climates for Impacts Studies (PRECIS) to force a hydrologic model called the Variable Infiltration Capacity model (VIC) to assess potential future changes in water resources in China under scenarios from phase 3 of the Coupled Model Intercomparison Project (CMIP3). They concluded that the prevailing pattern of “north dry and south wet” in China is likely to be exacerbated in the future. On the other hand, precipitation mean and extremes in China are expected to increase in the future (Xu and Xu 2012; Li et al. 2013a). A number of studies have projected future changes in floods and droughts purely based on the spatiotemporal variations in precipitation mean and extremes (e.g., Chen et al. 2012; Li et al. 2013a,b), but such studies only considered the change in climate and did not consider watershed hydrologic processes. Therefore, studies of changes in floods and water availability based on various simulations of discharge with the consideration of uncertainties are very much needed.
In this study, therefore, we analyze ISI-MIP simulations to 1) assess ISI-MIP simulations of historical floods and water availability against observed streamflow data, 2) evaluate the climate change impacts on floods and water availability during 2070–99 relative to 1971–2000 across China under RCP2.6 (low emission) and RCP8.5 (high emission) with the consideration of multimodel uncertainties, and 3) explore the linkage between changes in floods and water availability and changing climate.
2. Data
We analyze simulated daily discharge and monthly actual evapotranspiration from eight hydrologic models (HMs), one type of IM in ISI-MIP, forced by bias-corrected outputs of five CMIP5 GCMs from ISI-MIP (Tables 1, 2; Warszawski et al. 2014). Spatial resolutions of GCMs are usually too coarse for hydrologic studies, so ISI-MIP provides GCM outputs at a 0.5° × 0.5° spatial resolution, which were interpolated from their original spatial resolutions (Table 1) by a statistical bias correction method developed in ISI-MIP (Hempel et al. 2013). This method ensures the long-term statistics of GCM outputs are in agreement with the Water and Global Change (WATCH) data of 1960–99 (Weedon et al. 2011; Warszawski et al. 2014). Simulations under the historical, RCP2.6, and RCP8.5 scenarios are collected. RCP2.6 is a low-emission scenario that achieves the 2°C global average warming target, in which the radiative forcing increases to about 3 W m−2 around the middle of the twenty-first century and then decreases to 2.6 W m−2 by 2100 (van Vuuren et al. 2011). RCP8.5 is a high-emission scenario in which the radiative forcing increases to 8.5 W m−2 by 2100 (Riahi et al. 2011). Details about most of the HMs can be found in Haddeland et al. (2011) and Davie et al. (2013). All simulations are obtained through runs without human impact, that is, no model inputs of socioeconomic factors such as water withdrawal for irrigation, population, and gross domestic product (GDP). Therefore, the outputs of the five GCMs listed in Table 1 are used to force each of the eight HMs listed in Table 2, and eventually we have 40 discharge simulations from the combinations of HMs and GCMs. These 40 discharge simulations are used to calculate the multimodel ensemble. Simulated daily precipitation from bias-corrected outputs of the five GCMs (Table 1) under the historical, RCP2.6, and RCP8.5 from ISI-MIP is also collected to analyze the changes in precipitation mean and extremes at the 0.5° × 0.5° spatial resolution.
Information about the GCMs from CMIP5.
Information about the HMs in ISI-MIP.
All of China is divided into 10 major river basins as in the previous studies (Fig. 1) (Wang et al. 2012; Li et al. 2013a). We collect observed daily streamflow data from four hydrologic stations, including Wuzhou and Boluo on the main stem of the Pearl River and Tangnaihai and Longmen on the main stem of the Yellow River. The Pearl River is a major river basin in the humid areas of southern China and the Yellow River is a major river basin in the semiarid and arid areas of northern China. The averages of naturalized streamflow of Datong on the Yangtze River, Luanxian on the Haihe River, Bengbu on the Huaihe River, and Tieling on the Lioahe River are also collected to assess simulations of water availability. Streamflow data are obtained from the Yellow River Conservancy Commission, 1958–80, Hydrologic Yearbook (in Chinese) and the data quality has been confirmed (Zhang et al. 2013b, 2014). The details of the hydrologic stations are listed in Table 3. Simulations of ISI-MIP used in this study do not consider societal and human influences, so the time period with the least human activities is chosen for validation. Since the late 1970s, major rivers in northern China have been intensively regulated by human activities (Li et al. 2011; Wang et al. 2012). Therefore, with the consideration of the length of ISI-MIP simulations (starting from 1971), the period of 1971–80 (streamflow data available at all the eight selected stations) is selected to evaluate the performance of ISI-MIP in modeling floods and water availability.
Major river basins in China. Red triangles denote hydrologic stations for validation in this study. Numbers denote river basins: 1) Songhuajiang River, 2) Liaohe River, 3) Haihe River, 4) Yellow River, 5) Huaihe River, 6) Yangtze River, 7) southeast rivers, 8) Pearl River, 9) southwest rivers, and 10) northwest rivers.
Citation: Journal of Hydrometeorology 17, 4; 10.1175/JHM-D-15-0074.1
Information about the hydrologic stations.
3. Methodology
a. Floods and water availability
For each combination of HM and GCM, annual maximum daily discharge is estimated from simulated daily discharge in each grid cell. Annual maximum daily discharge series are then fitted into the generalized extreme value (GEV) distribution using the L-moment method (Hosking 1990; Embrechts 1997; Katz et al. 2002). A flood event is indicated by the annual maximum daily discharge with specific n-yr return period (n-yr flood hereafter), which is determined by the probability of exceedance. The n-yr flood is calculated based on a fitted GEV distribution and is the annual maximum daily discharge with the exceedance probability of 1/n. In this study, 5-, 20-, 30-, and 50-yr floods are estimated. For instance, 50-yr flood means the annual maximum daily discharge with 50-yr return period and its probability of exceedance in any given year is 1/50. Floods are estimated at each grid cell for the periods of 1971–2000 and 2070–99 separately. Water availability is estimated as the mean discharge of a considered period (Manabe et al. 2004).
b. Quantile–quantile plot
Quantile–quantile (QQ) plot is a graphical technique for comparing two probability distributions (Wilk and Gnanadesikan 1968). If two distributions are identical, the plot of their quantiles against each other should be on a straight line with slope of unity pointing toward the origin (1:1 line hereafter). If the relationship of two variables is linear, the QQ plot will be linear but with changed location and slope. In this study, the QQ plot is used to compare simulated discharge against observations.
c. Relationships of changes in runoff, precipitation, and actual evapotranspiration
d. Model agreement of change rates
The changes in the future are represented by change rate (%) and model agreement (%). Change rate is the ratio of the difference between the future period (2070–99) and the historical period (1971–2000) to the value of the historical period. Change rate of each combination of HM and GCM is calculated individually. For discharge simulations, model agreement is denoted by the proportion of all 40 model combinations (%) that project the change rate within the corresponding category (Davie et al. 2013). For example, for a cell, if 32 out of the 40 combinations project the change rates of 30-yr flood falling in the range of 10%–50%, then the model agreement of the changes of 10%–50% in that cell is 80% (e.g., 32/40). Model agreement is presented by four categories, <50%, 50%–60%, 60%–70%, and 70%–100%, in different colors. For precipitation simulations, model agreement is denoted by number of models, and model agreement is presented by four categories, ≤2, 3, 4, and 5, in different colors. In the following results, only cells with changes of ≥50% model agreement for discharge simulations or ≥3 models for precipitation are shown. Cells with lower model agreement are not presented and not considered.
4. Results
We investigate the future changes in floods and water availability across China through the following five steps. First, floods and water availability derived from ISI-MIP simulations are compared with the past observations to evaluate the reliability of ISI-MIP outputs. Second, the future changes in floods and water availability across the country are presented. Third, we analyze the causes of the future changes in floods and water availability. Fourth, the implications of changing climate (i.e., changes in precipitation and temperature) on floods and water availability are examined. Fifth, the major source of the uncertainties associated with the projected floods and water availability is identified and discussed. The technical procedures and study results are described in detail as follows.
a. Evaluation of floods and water availability projected by ISI-MIP
Even though ISI-MIP simulations have been used to study global changes in the water cycle (e.g., Dankers et al. 2014; Elliott et al. 2014; Schewe et al. 2014), a regional assessment of China is still important to apply the simulations at the regional scale. Therefore, the floods and water availability derived from ISI-MIP simulations are first compared to the past observations to evaluate the model performances. First, the capability of ISI-MIP in modeling annual maximum daily discharge is evaluated by QQ plots of observed and simulated annual maximum daily discharge at Wuzhou, Boluo, Tangnaihai, and Longmen stations (Fig. 2) (Khan et al. 2006). We can see that the uncertainties among different individual combinations are large and that there are systematic biases for most combinations. For the multimodel ensemble median of the 40 combinations of HMs and GCMs, the simulated values are shown to be close to observations, and therefore the probability distributions of observed and simulated discharges can be considered similar. But the ensemble median tends to underestimate in the upper tail for annual maximum daily discharge.
QQ plots of observed and simulated annual maximum daily discharge of (a) Wuzhou, (b) Boluo, (c) Tangnaihai, and (d) Longmen during 1971–80. The black line denotes the 1:1 line. In the legend, G, H, I, M, and N indicate GFDL-ESM2M, HadGEM2-ES, IPSL-CM5A-LR, MIROC-ESM-CHEM, and NorESM1-M, respectively.
Citation: Journal of Hydrometeorology 17, 4; 10.1175/JHM-D-15-0074.1
Second, the simulated 5-, 20-, 30-, and 50-yr floods are compared against observations (Table 4). Generally, the observed floods fall in the ranges of simulations, and the multimodel ensemble medians in most cases are close to observations, indicating that the ability of ISI-MIP in modeling floods is acceptable. The multimodel median tends to underestimate floods just as the situation of annual maximum daily discharge shown in Fig. 2. All observed floods are within the range of the 25th and 75th percentiles of simulations. This comparison indicates ISI-MIP simulations have an acceptable ability in modeling 5-, 20-, 30-, and 50-yr floods for the Pearl River and the Yellow River, but tend to underestimate them. The uncertainties, however, are generally large among individual combinations. For instance, the 25th to 75th values of the 50-yr flood can range from 4549 to 13 732 m3 s−1.
Floods (m3 s−1) derived from observations and simulations of median and 25th and 75th percentiles.
The above procedures are also applied to evaluate water availability. QQ plots are made to assess the performance of ISI-MIP in modeling annual mean daily discharge first and then the water availability simulated by ISI-MIP is compared to observations. Figure 3 shows an acceptable performance in general, but overestimation of annual mean daily discharge at Longmen station is identified (Fig. 3d). The uncertainties among individual models are also large in annual mean daily discharge. Correspondingly, the simulated water availability fits observations quite nicely at most stations, but overestimates the water availability at Longmen and Tieling as shown in Table 5.
As in Fig. 2, but for annual mean daily discharge.
Citation: Journal of Hydrometeorology 17, 4; 10.1175/JHM-D-15-0074.1
Water availability (m3 s−1) derived from observations and simulations of median and 25th and 75th percentiles. For Datong, Luanxian, Bengbu, and Tieling, values in the observation column are the average of naturalized streamflow.
Generally, the performance of ISI-MIP in simulating the historical floods and water availability is acceptable. However, the ensemble median tends to underestimate floods but overestimate water availability. It is important to keep in mind that ISI-MIP simulations differ considerably even though their multimodel medians fit observations. Therefore, model agreement is an important indicator to quantify the uncertainties among different combinations. Furthermore, in section 4e, we analyze and identify the major source, GCM or HM, of the uncertainties associated with discharge simulations.
b. Future changes in floods and water availability across China
Figure 4 presents the projected changes and model agreement of 5-, 30-, and 50-yr floods as well as water availability across China during 2070–99 under RCP8.5 and RCP2.6 relative to 1971–2000 under the historical scenario. Under RCP8.5, the multimodel ensemble reaches agreement on the increase of 5-yr floods in the Pearl River, southeast rivers, southwest rivers, the Yangtze River, the Yellow River, the Songhuajiang River, the western part of the Haihe River, and the southern part of northwest rivers (Fig. 4a). However, it does not reach agreement on the changes in the Huaihe River, the Liaohe River, and most parts of northwest rivers, indicating the uncertainties in the changes of 5-yr floods are too large to provide an acceptable projection on the potential changes. For most rivers where models reach agreement, the increase rates of 5-yr floods are from 10% to 50%. The 5-yr floods in the southern part of northwest rivers and the western part of southwest rivers increase by more than 50%. In the Pearl River basin, 50%–60% of the combinations indicate that 5-yr floods in the north increase by 10%–50%. In the southeastern part of the Pearl River basin where the lower main stem is located, the increases of 5-yr floods are identified with higher model agreement. In the middle of the Pearl River main stem, more than 70% of the combinations indicate that the changes of 5-yr floods are from −10% to 10% during 2070–99 under RCP8.5. Given the large uncertainties in simulated floods from ISI-MIP (refer to section 4a), such low change rates can be considered negligible. In the Yangtze River, 5-yr floods in the western (upstream) part of the basin increase with higher model agreement of more than 60%. In the middle Yangtze River, 50%–60% of simulations project that 5-yr floods increase by 10%–50%. Nevertheless, at the outlet of the Yangtze River, no model agreement is reached. In the upper and middle Yellow River, the increase of 5-yr floods is projected consistently by model simulations, especially in the middle reach where model agreement is larger than 70%. In the lower Yellow River, however, model agreement does not reach 50%. These results clearly indicate the substantial spatial variability in the changes of 5-yr floods within the same river basin. For the 30-yr floods, increase is also projected on the main stems of the Pearl River, the Yangtze River, the Yellow River, southeast rivers, southwest rivers, the northern part of the Songhuajiang River, and the southern part of northwest rivers, but the model agreements are smaller than those of the 5-yr floods (Fig. 4c). The changes in 20-yr floods are similar with the 30-yr floods (figures not shown). Simulations reach agreement on the increase of 50-yr floods on the main stems of the Pearl River, the upper and the southeastern parts of the Yangtze River, and the upper Yellow River (Fig. 4e). On the other hand, 50-yr floods in the northeastern Pearl River increase by 50% with high model agreement. The 50-yr floods in the southern part of southwest rivers also increase more compared to the 5- and 30-yr floods, but with lower model agreements. In other words, larger floods increase to greater extents than those of shorter return periods, although the uncertainties are also larger. In summary, floods across China generally increase by 10%–50% with acceptable model agreement, especially in the Pearl River, the Yangtze River, the Yellow River, southeast rivers, southwest rivers, and the Songhuajiang River.
Multimodel ensemble median of changes (%) and model agreement (%) of floods and water availability across China over the period of 2070–99 relative to the period of 1971–2000. Shown are change rates of (a),(b) 5-yr flood; (c),(d) 30-yr flood; (e),(f) 50-yr flood; and (g),(h) water availability under (left) RCP8.5 and (right) RCP2.6. Each color denotes the corresponding change rate. The lighter (darker) color denotes the lower (higher) model agreement of that category of change rates.
Citation: Journal of Hydrometeorology 17, 4; 10.1175/JHM-D-15-0074.1
Different from the consistent increases in floods, water availability exhibits more complicated changing characteristics under RCP8.5 (Fig. 4g). Unlike increases in floods on the whole main stem of the Pearl River with high model agreement, decreases in water availability are projected with model agreement on the lower main stem of the Pearl River. ISI-MIP simulations project 10%–50% increases in floods, but negligible changes in water availability with sufficient model agreements of 50%–60% along the main stem of the Yangtze River toward the end of the twenty-first century. However, in the tributaries and the outlet of the Yangtze River, a decrease in water availability is indicated by more than 70% of the combinations. Increasing floods but decreasing water availability are also revealed in southeast rivers and the northern part of the Songhuajiang River. In southwest rivers, 10%–50% increases in water availability, which are lower than those in 30- and 50-yr floods, are projected with over 50% model agreement.
Under the low-emission scenario (i.e., RCP2.6), as expected, the changes in floods are generally smaller in comparison with the high-emission scenario (i.e., RCP8.5; Fig. 4). For 5-yr floods, the projections indicate negligible changes on the whole main stem of the Yangtze River, but increases in the southeastern and northwestern parts of this river basin. In the Yellow River, the 5-yr floods change negligibly on the upper main stem but increase in the middle main stem. In the lower Yellow River, however, ISI-MIP simulations do not produce results with sufficient consistency. As shown in Figs. 4d and 4f, the floods of longer return periods (30 and 50 years) generally increase more than the 5-yr floods. The 30- and 50-yr floods increase similarly in the Pearl River, the upper and middle Yangtze River, the middle Yellow River, the Songhuajiang River, and southwest rivers, but the model agreements are relatively lower than the simulations of 30- and 50-yr floods, implying larger uncertainties associated with the simulations of greater floods. To summarize, 5-yr floods experience negligible changes, but 30- and 50-yr floods increase substantially across China under RCP2.6. With regard to water availability as shown in Fig. 4h, however, the changes are generally negligible in large parts of the country except in the western part of the Songhuajiang River and the northern part of northwest rivers. Overall, these results indicate increases in floods, especially those of longer return periods and negligible changes in water availability across China under RCP2.6.
In summary, the ISI-MIP simulation results indicate that floods increase across China under both RCP8.5 and RCP2.6, but to greater extents under RCP8.5. The 5-yr floods in the Pearl River, the Yangtze River, and the Yellow River change negligibly under RCP2.6, but increase significantly under RCP8.5. The changes in water availability are negligible under RCP2.6, but negative under RCP8.5 in the Pearl River and the Yellow River. Overall, two general change patterns are identified, that is, increasing floods and decreasing water availability in southern China under RCP8.5, and increasing floods and negligible changes in water availability in China as a whole under RCP2.6. The projected changes in the Pearl River, the Yangtze River, the Yellow River, southeast rivers, southwest rivers, and the Songhuajiang River carry more certainties in the ISI-MIP simulations. Generally, the model agreements are higher in flood simulations compared to water availability, implying it is more certain to project the changes in floods than in water availability.
c. Potential causes for reduced water availability and intensified floods in southern China
Toward the end of the twenty-first century, the projected spatial patterns of the changes in floods and water availability and their differences are obvious as described above. Under RCP8.5, it is understandable that water availability increases and floods become more intensive as the climate gets wetter in northern China, but the simulations also project reduced water availability and intensified floods, which seems to be contradictory in southern China.
As described in the methodology section, floods are calculated from the annual maximum daily discharge series, and water availability is determined by annual mean daily discharge series. Here, we examine the relative changes in annual maximum daily discharge and annual mean daily discharge and use their ratios to explain this contradiction. As shown in Fig. 5a, this ratio increases by 10%–50% with high model agreement in southern China including the Pearl River, the Yangtze River, southeast rivers, and southwest rivers, which is caused by the increase in annual maximum daily discharge (Fig. 5c) and the decrease in mean discharge (Fig. 4g; water availability quantified by the long-term mean of the discharge). As a result, floods increase but water availability decreases in southern China under RCP8.5 (Fig. 4). In most parts of northern China, especially on the main stem of the Yellow River, there is negligible change in the ratio, indicating that annual maximum daily discharge and annual mean daily discharge increase at similar rates. This matches the increases in floods and water availability in northern China (Figs. 4, 5). Under RCP2.6, the multimodel ensemble produces negligible changes in the ratio across China because of negligible changes in maximum and mean discharge (Fig. 5). Therefore, floods with short return periods and water availability change negligibly under RCP2.6.
As in Fig. 4, but for (a),(b) change rates of the ratio of annual maximum daily discharge to annual mean daily discharge and (c),(d) annual maximum daily discharge.
Citation: Journal of Hydrometeorology 17, 4; 10.1175/JHM-D-15-0074.1
Precipitation is normally the most dominant factor determining river discharge (Pavelsky and Smith 2006; Q. Zhang et al. 2011). In the future, not only the mean state but also the extremes of precipitation will change (IPCC 2013; Li et al. 2013b), and such changes in precipitation will bring about substantial changes in discharge (Vörösmarty et al. 2000; Wang et al. 2012; Schewe et al. 2014). Precipitation extreme is the key factor influencing the extremes of discharge, and mean precipitation usually plays a more important role in determining water availability. The spatial patterns of the ratio of annual maximum daily precipitation to annual mean daily precipitation match those of the ratio of annual maximum daily discharge to annual mean daily discharge (Figs. 5, 6). Under RCP8.5, these ratios in southern China increase, while those in northern China change negligibly (Fig. 6a). In southern China, the increases of the ratios are caused by a negligible change in mean precipitation (Fig. 6e) but a significant increase in annual maximum daily precipitation (Fig. 6c), which means that although mean precipitation does not change considerably, the extremes increase substantially in southern China. Such a change pattern indicates an intensification of precipitation regime, which is an important aspect of changing precipitation and has been well observed and extensively discussed in previous studies (e.g., Q. Zhang et al. 2011, 2012). The precipitation tends to become more intensive because of the increase in lower-tropospheric water vapor as a consequence of rising temperature (Held and Soden 2006). Therefore, in summary, increasing floods but decreasing water availability is caused under the circumstances that precipitation gets more intensified, but with a negligible change in the mean in southern China under RCP8.5.
As in Fig. 4, but for (a),(b) change rates of the ratio of annual maximum daily precipitation to annual mean daily precipitation; (c),(d) annual maximum daily precipitation; and (e),(f) annual mean daily precipitation. Instead of being represented by the proportion of models as in Fig. 4, the model agreement is represented by the number of models that agree with that category of change rates.
Citation: Journal of Hydrometeorology 17, 4; 10.1175/JHM-D-15-0074.1
d. Implications of changing climate on floods and water availability
The changes in floods and water availability are the combined effects of the changes of different climatic factors. The above results show that the wet tendency in terms of precipitation causes an increase in floods and water availability in northern China, while the combination of negligible changes in mean precipitation and intensified extremes bring about the decrease in water availability but increase in floods in southern China. Furthermore, the changing temperature also plays an important role in the changes of floods and water availability by affecting actual and potential evapotranspiration. Gao et al. (2007) confirmed that change in actual evapotranspiration is basically controlled by potential evapotranspiration in southeastern China where sufficient water is normally available for evapotranspiration under the humid climate. Potential evapotranspiration is primarily controlled by air temperature, wind speed, sunshine duration, relative humidity, etc. (Yang and Yang 2012; Yong et al. 2013). A diagnostic study by Laine et al. (2014) found that the increase in evaporation over tropical regions projected by CMIP5 is mainly due to the contribution from the increase in air temperature, although this contribution is partially offset by weakening surface winds and increasing near-surface relative humidity. Furthermore, an increase in temperature can bring about an increase in actual evapotranspiration, but in turn an increase in actual evapotranspiration can result in a decrease in temperature (Davie 2008). In this study, HMs are forced by GCMs without the consideration of this feedback to temperature.
Here, we focus on the effects of changing precipitation and temperature on floods and water availability. At the basin scale, the simulated discharge is routed from runoff at each grid cell within the basin. Runoff generated at each grid is determined by the climatic factors of that grid cell. Therefore, the discharge change of each grid is controlled by the changes in the climatic factors of this grid and all other upstream grids. Figure 7 compares change rates of floods and water availability at the outlets of the Pearl River, the Yangtze River, and the Yellow River, and change rates of mean and extreme precipitation in these basins. The change rates of mean and extreme precipitation are estimated from the areal averages of the whole basin. At the outlet of the Pearl River, floods of longer return periods increase more than those of shorter return periods while the water availability decreases (Fig. 7a). Similarly, extreme precipitation (i.e., the n-yr values of annual maximum daily precipitation) of longer return periods increases more, and all extreme values of different magnitudes increase more than the mean values (Fig. 7b). The similarities between changes in discharge and precipitation indicate the dominant role of changes in precipitation in the whole basin for changes in floods and water availability. Similar findings are obtained in the Yangtze River and the Yellow River (Fig. 7).
Box-and-whisker plots of changes (%) of 5-, 20-, 30-, and 50-yr floods and water availability (WA) at the outlets of (a) the Pearl River, (c) the Yangtze River, and (e) the Yellow River, as well as change rates of 5-, 20-, 30-, and 50-yr values of annual maximum daily precipitation and mean precipitation at the outlets of (b) the Pearl River, (d) the Yangtze River, and (f) the Yellow River during 2070–99 relative to 1971–2000. In the box, the central line is the multimodel ensemble median, and the edges are the 25th and 75th percentiles, respectively. The whiskers indicate the most extreme changes in the ensemble.
Citation: Journal of Hydrometeorology 17, 4; 10.1175/JHM-D-15-0074.1
However, the increase rates of precipitation are generally larger than those of floods and water availability, which is especially obvious under RCP8.5, implying that other climatic factors also considerably influence the changes in floods and water availability. Furthermore, water availability decreases while mean precipitation changes negligibly in southern China toward the end of the twenty-first century. According to the long-term water balance, runoff equals precipitation minus actual evapotranspiration, and discharge is the routed runoff. Therefore, the increase in actual evapotranspiration mainly caused by temperature increase should contribute to such differences. Since temperature under RCP8.5 increases much more significantly than under RCP2.6 (IPCC 2013), the actual evapotranspiration increases more significantly under RCP8.5 (Fig. 8a) but changes negligibly under RCP2.6 (Fig. 8b); thus, the difference between water availability and mean precipitation under RCP8.5 is larger than that under RCP2.6. If the long-term mean actual evapotranspiration increases more than mean precipitation, which is the case in southern China under RCP8.5 (Fig. 8c), the mean runoff will decrease consequently, and hence water availability decreases (Fig. 4g). On the other hand, if mean actual evapotranspiration increases less than the mean precipitation, as the situation in northern China under RCP8.5, the runoff will increase consequently, and hence water availability will increase.
The changes (%) and model agreement of annual mean daily actual evapotranspiration under (a) RCP8.5 and (b) RCP2.6, and the difference between precipitation and actual evapotranspiration under (c) RCP8.5 and (d) RCP2.6.
Citation: Journal of Hydrometeorology 17, 4; 10.1175/JHM-D-15-0074.1
In summary, toward the end of the twenty-first century, intensified precipitation extremes may cause larger floods in China. Under RCP8.5, although actual evapotranspiration increases in northern China, the wet tendency in precipitation is larger and raises water availability. However, in southern China, negligible changes in precipitation with significant increases in actual evapotranspiration reduce water availability. Under RCP2.6, negligible changes in precipitation and actual evapotranspiration lead to negligible changes in water availability.
e. The major source of the uncertainties: GCM or HM?
Uncertainties associated with discharge simulations arise from uncertainties of both GCMs and HMs (Teng et al. 2012). In this section, the sources of uncertainties associated with floods and water availability projections are explored by visually assessing the consistency of outputs generated by the 40 model combinations and numerically comparing the variances of outputs from each GCM in combination with eight HMs to those from each HM in combination with five GCMs at each cell.
Figure 9 shows the changes in 30-yr floods under RCP8.5 of each model combination of HM and GCM. For different HMs forced by the same GCM (by the same column in Fig. 9), the spatial patterns of changes in 30-yr floods are highly consistent, but the exact values are quite different and the variations are generally larger among model combinations of different GCMs with the same HM (defined as GCM variance) than different HMs with the same GCM (defined as HM variance), indicating the greater influence of climatic forcing from GCMs on the hydrologic simulations, which has also been found by Teng et al. (2012) in Australia. This is also true for the changes in water availability under RCP8.5 (figures not shown). For each grid cell, eight GCM variances (each of the eight HMs with five GCMs) and five HM variances (each of the five GCMs with eight HMs) are averaged and then the average GCM variance is divided by the corresponding HM variance. This ratio is presented in Fig. 10 for the 30-yr floods and water availability, respectively, to demonstrate the relative contribution of GCMs and HMs to model uncertainty. The red (blue) grid cells indicate that the ratio of average GCM variance to average HM variance is larger (less) than 1.1 (0.9), which means GCM (IM) variance dominates. It is clear that GCM variance dominates in the eastern and southern parts of the country, coinciding with the humid and semihumid regions. Interestingly, the boundary of the red-colored grid cells (the ratio larger than one) on the east and southeast generally follows the annual isohyet of 400 mm as the divide between humid and arid regions in China. To the west of this isohyet, blue-colored grid cells (the ratio smaller than one) do exist in some locations with very dry climate and indicate greater model uncertainties caused by HMs.
Changes (%) in 30-yr flood simulated by eight HMs forced by five bias-corrected GCMs under RCP8.5. Rows represent HMs (D, H, Ma, MA, MP, P, V, and W denote DBH, H08, Mac-PDM, MATSIRO, MPI-HM, PCR-GLOBWB, VIC, and WBM, respectively), and columns represent GCMs (G, H, I, M, and N denote GFDL-ESM2M, HadGEM2-ES, IPSL-CM5A-LR, MIROC-ESM-CHEM, and NorESM1-M, respectively).
Citation: Journal of Hydrometeorology 17, 4; 10.1175/JHM-D-15-0074.1
Ratio of average GCM variance to average HM variance of (a) 30-yr flood and (b) water availability under RCP8.5. The variance of changes in the 30-yr flood/water availability of all five GCMs of each individual HM is calculated, and then the average GCM variance is the average of variances of all eight HMs. The definition of average HM variance is the same as average GCM variance, but for HMs. A variance ratio larger (less) than 1 indicates uncertainties from GCMs (IMs) predominate.
Citation: Journal of Hydrometeorology 17, 4; 10.1175/JHM-D-15-0074.1
5. Discussion
In this study, we project changes in floods and water availability over China toward the end of the twenty-first century by employing outputs of 40 combinations of GCMs and HMs. Different combinations do generate varying results for the same region. Thus, it is necessary to analyze and interpret these results with careful consideration of model agreements. Nevertheless, for different regions or different variables, the model agreements vary substantially. Compared with other regions, model agreements are generally higher in the Pearl River, the Yangtze River, the Yellow River, southeast rivers, southwest rivers, and the Songhuajiang River, especially along the main stems. First, hydrologic regimes are better monitored in these major river basins than the others. Second, the river discharges are large enough relative to the systematical biases and errors in HMs and GCMs, which helps to generate more robust results. On the other hand, in northwestern China, the model agreements seldom reach 50%, probably because of the lack of sufficient discharge measurements and the relatively small discharge.
The model agreements of the changes in the ratio of annual maximum daily discharge to annual mean daily discharge are higher in comparison with their individual changes (Figs. 4, 5). Hamlet and Lettenmaier (2007) used the VIC to evaluate changes in flood risk in the western United States and found out that it performed better to simulate the ratio of the 100-yr flood to the mean annual flood than the absolute flood magnitude. They indicated that the bias in mean annual flood simulation was offset in the ratio calculation although the absolute errors are on the order of 50%.
Generally, model agreements of changes in annual maximum daily discharge are higher than those of annual mean daily discharge, and correspondingly those of floods are higher than those of water availability. The fundamental reason behind this is the discrepancy of future precipitation projected by different GCMs as explained in the following. Figure 11 shows remarkable variability in mean precipitation among five GCMs. Especially in southeastern China, GFDL-ESM2M and MIROC-ESM-CHEM project negligible changes in mean precipitation under RCP8.5, HadGEM2-ES and NorESM1-M project an increase, and IPSL-CM5A-LR projects a decrease. In contrast, all GCMs project significant increases in extreme precipitation as quantified by the 30-yr annual maximum daily precipitation. Since outputs of GCMs contribute substantially to the uncertainties associated with floods and water availability, especially in humid regions, model agreements of the increases in floods are higher than those of water availability.
Changes (%) of 30-yr annual maximum daily precipitation (Extreme) and annual mean daily precipitation (Mean) from five bias-corrected GCMs under RCP8.5 (G, H, I, M, and N denote GFDL-ESM2M, HadGEM2-ES, IPSL-CM5A-LR, MIROC-ESM-CHEM, and NorESM1-M, respectively).
Citation: Journal of Hydrometeorology 17, 4; 10.1175/JHM-D-15-0074.1
A major limitation of this study is that the changes in vegetation due to changing climate and hydrologic conditions are not taken into consideration in the HMs. As shown by L. Zhang et al. (2011), variations in vegetation distribution have considerable impacts on streamflow. Hence, the changes in vegetation alter the hydrologic cycle and water budget, which then influence the distribution of vegetation subsequently. Another type of model, called a biome model, does consider CO2 impacts and vegetation dynamics; however, such consideration brings about other uncertainties (Davie et al. 2013). Another limitation is that the impacts of human activities are also not considered. In practice, water withdrawal, reservoir regulation, and other human activities can heavily affect the instream flow and groundwater storage. These complex human activities are very difficult to be quantified and projected in the models, but since this study mainly focuses on the impacts of climate change on floods and water availability, it is reasonable to ignore human activities. However, to have deeper insights of future changes in floods and water availability, the consideration of human activities will be very desirable in a future study. Finally, although ISI-MIP has been applied in other studies of global hydrology (e.g., Dankers et al. 2014; Elliott et al. 2014) and our assessment indicates acceptable performances of ISI-MIP in simulating floods and water availability in several representative basins, a more thorough validation in more basins would be desirable, but this requires more daily/monthly observations in other basins that are not available in this study.
6. Conclusions
To study the potential changes in floods and water availability across China and their linkage with climate change during 2070–99 relative to 1971–2000, hydrologic outputs of eight HMs forced by five GCMs (total of 40 combinations of HMs and GCMs) from ISI-MIP are analyzed and evaluated. Before ISI-MIP is applied to study floods and water availability, its abilities in modeling floods and water availability are evaluated first in the representative basins where reliable observations are available (i.e., the Pearl River, the Yellow River, the Yangtze River, the Haihe River, the Huaihe River, and the Liaohe River). Then the spatiotemporal changes in 5-, 20-, 30-, and 50-yr floods and water availability with consideration of model agreement are investigated. Afterward, the linkages of changes in floods and water availability with changing climate are analyzed. Finally, we examine the contributions of GCMs and HMs to the uncertainties associated with model results. The following conclusions can be drawn from the research findings:
ISI-MIP simulations have certain abilities in modeling floods and water availability. Multimodel ensemble medians can represent floods and water availability reasonably well, but generally tend to underestimate floods and overestimate water availability. Compared to HMs, GCMs play a more significant role in determining the spatial patterns of changes in floods and water availability. GCM variances dominate the uncertainties of floods and water availability in the humid and semihumid regions in eastern and southern China, while HM variances play a more significant role in the arid and semiarid regions in northwestern China.
During 2070–99, increasing floods but decreasing water availability is detected in southern China under RCP8.5, implying that this region may face greater threats from flood hazards but also endure less available water resources, which may worsen the conflict between water demand and availability, and also increase drought risks. In northern China, our study indicates that floods and water availability both increase. Under RCP2.6, floods increase while water availability changes negligibly across China. Generally, floods in China in the future increase more significantly than water availability. The increase rates of larger floods are higher than those of smaller ones.
Highly intensifying precipitation extremes are the primary reason for a significant increase in floods over China in the future. Under RCP8.5 in southern China, although precipitation extremes intensify, mean precipitation changes negligibly. The precipitation changes combined with increased actual evapotranspiration as a result of rising temperature reduce water availability. In northern China, although actual evapotranspiration increases, the wet tendency of precipitation is found to be even more significant, leading to an increase of water availability. Generally, because of the increase of actual evapotranspiration caused by higher temperature, the changes in floods and water availability are less significant than those in precipitation.
Model agreements for the changes in floods are generally higher than water availability, mainly because of the greater uncertainties associated with the simulated mean precipitation from GCMs. GCMs tend to consistently project significant increases in precipitation extremes over China, but varying changes in mean precipitation, especially in southern China under RCP8.5.
Acknowledgments
The work described in this paper was substantially supported by a grant from the Research Grants Council of the Hong Kong Special Administrative Region, China (Project CUHK441313). J. Li would like to acknowledge the Australian Government Department of Education and Training for the award of the Endeavour Research Fellowship. We acknowledge the World Climate Research Programme’s Working Group on Coupled Modelling, which is responsible for CMIP. We thank the modeling groups that are listed in Tables 1 and 2 of this paper and the ISI-MIP coordination team for their efforts in producing, coordinating, and making the model outputs publically available. Our cordial gratitude should also be extended to the editor, Dr. Faisal Hossain, and four reviewers for their pertinent and professional comments and suggestions.
REFERENCES
Bierkens, M. F. P., and van Beek L. P. H. , 2009: Seasonal predictability of European discharge: NAO and hydrological response time. J. Hydrometeor., 10, 953–968, doi:10.1175/2009JHM1034.1.
Chen, H., Sun J. , Chen X. , and Zhou W. , 2012: CGCM projections of heavy rainfall events in China. Int. J. Climatol., 32, 441–450, doi:10.1002/joc.2278.
Dankers, R., and Hiederer R. , 2008: Extreme temperatures and precipitation in Europe: Analysis of a high-resolution climate change scenario. JRC Scientific and Tech. Rep. EUR 23291, European Union, 66 pp. [Available online at http://esdac.jrc.ec.europa.eu/ESDB_Archive/eusoils_docs/other/EUR23291EN.pdf.]
Dankers, R., and Coauthors, 2014: First look at changes in flood hazard in the Inter-Sectoral Impact Model Intercomparison Project ensemble. Proc. Natl. Acad. Sci. USA, 111, 3257–3261, doi:10.1073/pnas.1302078110.
Davie, J. C. S., and Coauthors, 2013: Comparing projections of future changes in runoff from hydrological and biome models in ISI-MIP. Earth Syst. Dyn., 4, 359–374, doi:10.5194/esd-4-359-2013.
Davie, T., 2008: Fundamentals of Hydrology. Routledge, 220 pp.
Elliott, J., Deryng D. , and Müller C. , 2014: Constraints and potentials of future irrigation water availability on agricultural production under climate change. Proc. Natl. Acad. Sci. USA, 111, 3239–3244, doi:10.1073/pnas.1222474110.
Embrechts, P., 1997: Modelling Extremal Events for Insurance and Finance. 645 pp.
Gao, G., Chen D. , Xu C.-Y. , and Simelton E. , 2007: Trend of estimated actual evapotranspiration over China during 1960–2002. J. Geophys. Res., 112, D11120, doi:10.1029/2006JD008010.
Gosling, S. N., and Arnell N. W. , 2011: Simulating current global river runoff with a global hydrological model: Model revisions, validation, and sensitivity analysis. Hydrol. Processes, 25, 1129–1145, doi:10.1002/hyp.7727.
Gosling, S. N., Bretherton D. , Haines K. , and Arnell N. W. , 2010: Global hydrology modelling and uncertainty: Running multiple ensembles with a campus grid. Philos. Trans. Roy. Soc. London, A368, 4005–4021, doi:10.1098/rsta.2010.0164.
Haddeland, I., and Coauthors, 2011: Multimodel estimate of the global terrestrial water balance: Setup and first results. J. Hydrometeor., 12, 869–884, doi:10.1175/2011JHM1324.1.
Hagemann, S., and Gates L. D. , 2003: Improving a subgrid runoff parameterization scheme for climate models by the use of high resolution data derived from satellite observations. Climate Dyn., 21, 349–359, doi:10.1007/s00382-003-0349-x.
Hamlet, A. F., and Lettenmaier D. P. , 2007: Effects of 20th century warming and climate variability on flood risk in the western U.S. Water Resour. Res., 43, W06427, doi:10.1029/2006WR005099.
Hanasaki, N., Kanae S. , Oki T. , Masuda K. , Motoya K. , Shirakawa N. , Shen Y. , and Tanaka K. , 2008a: An integrated model for the assessment of global water resources—Part 1: Model description and input meteorological forcing. Hydrol. Earth Syst. Sci., 12, 1007–1025, doi:10.5194/hess-12-1007-2008.
Hanasaki, N., Kanae S. , and Oki T. , 2008b: An integrated model for the assessment of global water resources—Part 2: Applications and assessments. Hydrol. Earth Syst. Sci., 12, 1027–1037, doi:10.5194/hess-12-1027-2008.
Held, I. M., and Soden B. J. , 2006: Robust responses of the hydrological cycle to global warming. J. Climate, 19, 5686–5699, doi:10.1175/JCLI3990.1.
Hempel, S., Frieler K. , Warszawski L. , Schewe J. , and Piontek F. , 2013: A trend-preserving bias correction—The ISI-MIP approach. Earth Syst. Dyn., 4, 219–236, doi:10.5194/esd-4-219-2013.
Hosking, J. R. M., 1990: L-moments: Analysis and estimation of distributions using linear combinations of order statistics. J. Roy. Stat. Soc., 52, 105–124.
IPCC, 2012: Managing the Risks of Extreme Events and Disasters to Advance Climate Change Adaptation. Cambridge University Press, 582 pp.
IPCC, 2013: Climate Change 2013: The Physical Science Basis. Cambridge University Press, 1535 pp.
Jonkman, S. N., 2005: Global perspectives on loss of human life caused by floods. Nat. Hazards, 34, 151–175, doi:10.1007/s11069-004-8891-3.
Katz, R. W., Pariange M. B. , and Naveau P. , 2002: Statistics of extremes in hydrology. Adv. Water Resour., 25, 1287–1304, doi:10.1016/S0309-1708(02)00056-8.
Khan, M. S., Coulibaly P. , and Dibike Y. , 2006: Uncertainty analysis of statistical downscaling methods. J. Hydrol., 319, 357–382, doi:10.1016/j.jhydrol.2005.06.035.
Laine, A., Nakamura H. , Nishii K. , and Miyasaka T. , 2014: A diagnostic study of future evaporation changes projected in CMIP5 climate models. Climate Dyn., 42, 2745–2761, doi:10.1007/s00382-014-2087-7.
Li, J., Zhang Q. , Chen X. , and Jiang T. , 2011: Study of ecological instream flow in Yellow River: Considering the hydrological change (in Chinese). Acta Geogr. Sin., 66 (1), 99–110.
Li, J., Zhang Q. , Chen Y. D. , and Singh V. P. , 2013a: GCMs-based spatiotemporal evolution of climate extremes during the 21st century in China. J. Geophys. Res. Atmos., 118, 11 017–11 035, doi:10.1002/jgrd.50851.
Li, J., Zhang Q. , Chen Y. D. , Xu C.-Y. , and Singh V. P. , 2013b: Changing spatiotemporal patterns of precipitation extremes in China during 2071–2100 based on Earth System Models. J. Geophys. Res. Atmos., 118, 12 537–12 555, doi:10.1002/2013JD020300.
Liang, X., Lettenmaier D. P. , Wood E. F. , and Burges S. J. , 1994: A simple hydrologically based model of land surface water and energy fluxes for general circulation models. J. Geophys. Res., 99, 14 415–14 428, doi:10.1029/94JD00483.
Manabe, S., Wetherald R. T. , Milly P. C. D. , Delworth T. L. , and Stouffer R. J. , 2004: Century-scale change in water availability: CO2-quadrupling experiment. Climatic Change, 64, 59–76, doi:10.1023/B:CLIM.0000024674.37725.ca.
Milly, P. C. D., Dunne K. A. , and Vecchia A. V. , 2005: Global pattern of trends in streamflow and water availability in a changing climate. Nature, 438, 347–350, doi:10.1038/nature04312.
Moss, R. H., and Coauthors, 2010: The next generation of scenarios for climate change research and assessment. Nature, 463, 747–756, doi:10.1038/nature08823.
Oki, T., and Kanae S. , 2006: Global hydrological cycles and world water resources. Science, 313, 1068–1072, doi:10.1126/science.1128845.
Pavelsky, T. M., and Smith L. C. , 2006: Intercomparison of four global precipitation data sets and their correlation with increased Eurasian river discharge to the Arctic Ocean. J. Geophys. Res., 111, D21112, doi:10.1029/2006JD007230.
Pokhrel, Y., Hanasaki N. , Koirala S. , Cho J. , Yeh P. J.-F. , Kim H. , Kanae S. , and Oki T. , 2012: Incorporating anthropogenic water regulation modules into a land surface model. J. Hydrometeor., 13, 255–269, doi:10.1175/JHM-D-11-013.1.
Poveda, G., and Coauthors, 2007: Linking long-term water balances and statistical scaling to estimate river flows along the drainage network of Colombia. J. Hydrol. Eng., 12, 4–13, doi:10.1061/(ASCE)1084-0699(2007)12:1(4).
Raff, D. A., Pruitt T. , and Brekke L. D. , 2009: A framework for assessing flood frequency based on climate projection information. Hydrol. Earth Syst. Sci., 13, 2119–2136, doi:10.5194/hess-13-2119-2009.
Ranger, N., and Coauthors, 2011: An assessment of the potential impact of climate change on flood risk in Mumbai. Climatic Change, 104, 139–167, doi:10.1007/s10584-010-9979-2.
Riahi, K., and Coauthors, 2011: RCP8.5—A scenario of comparatively high greenhouse gas emissions. Climatic Change, 109, 33–57, doi:10.1007/s10584-011-0149-y.
Rosenzweig, C., Lglesias A. , Yang X. B. , Epstein P. R. , and Chivian E. , 2001: Climate change and extreme weather events: Implications for food production, plant diseases, and pests. Global Change Hum. Health, 2, 90–104, doi:10.1023/A:1015086831467.
Schewe, J., and Coauthors, 2014: Multimodel assessment of water scarcity under climate change. Proc. Natl. Acad. Sci. USA, 111, 3245–3250, doi:10.1073/pnas.1222460110.
Stacke, T., and Hagemann S. , 2012: Development and evaluation of a global dynamical wetlands extent scheme. Earth Syst. Sci., 16, 2915–2933, doi:10.5194/hess-16-2915-2012.
Tang, Q., Oki T. , and Kanae S. , 2006: A distributed biosphere hydrological model (DBHM) for large river basin. Proc. Hydraul. Eng., 50, 37–42, doi:10.2208/prohe.50.37.
Tang, Q., Oki T. , Kanae S. , and Hu H. , 2007: The influence of precipitation variability and partial irrigation within grid cells on a hydrological simulation. J. Hydrometeor., 8, 499–512, doi:10.1175/JHM589.1.
Taylor, K. E., Stouffer R. J. , and Meehl G. A. , 2012: An overview of CMIP5 and the experiment design. Bull. Amer. Meteor. Soc., 93, 485–498, doi:10.1175/BAMS-D-11-00094.1.
Teng, J., Vaze J. , Chiew F. H. S. , Wang B. , and Perraud J.-M. , 2012: Estimating the relative uncertainties sourced from GCMs and hydrological models in modeling climate change impact on runoff. J. Hydrometeor., 13, 122–139, doi:10.1175/JHM-D-11-058.1.
van Vuuren, D. P., and Coauthors, 2011: RCP2.6: Exploring the possibility to keep global mean temperature increase below 2°C. Climatic Change, 109, 95–116, doi:10.1007/s10584-011-0152-3.
Vörösmarty, C. J., Federer C. A. , and Schloss A. L. , 1998: Potential evaporation functions compared on US watersheds: Possible implications for global-scale water balance and terrestrial ecosystem modeling. J. Hydrol., 207, 147–169, doi:10.1016/S0022-1694(98)00109-7.
Vörösmarty, C. J., Green P. , Salisbury J. , and Lammers R. B. , 2000: Global water resources: Vulnerability from climate change and population growth. Science, 289, 284–288, doi:10.1126/science.289.5477.284.
Wada, Y., van Beek L. P. H. , van Kempen C. M. , Reckman J. W. T. M. , Vasak S. , and Bierkens M. F. P. , 2010: Global depletion of groundwater resources. Geophys. Res. Lett., 37, L20402, doi:10.1029/2010GL044571.
Wang, G. Q., and Coauthors, 2012: Assessing water resources in China using PRECIS projections and VIC model. Hydrol. Earth Syst. Sci., 16, 231–240, doi:10.5194/hess-16-231-2012.
Warszawski, L., Frielet K. , Huber V. , Pointek F. , Serdeczny O. , and Schewe J. , 2014: The Inter-Sectoral Impact Model Intercomparison Project (ISI-MIP): Project framework. Proc. Natl. Acad. Sci. USA, 111, 3228–3232, doi:10.1073/pnas.1312330110.
Weedon, G. P., and Coauthors, 2011: Creation of the WATCH forcing data and its use to assess global and regional reference crop evaporation over land during the twentieth century. J. Hydrometeor., 12, 823–848, doi:10.1175/2011JHM1369.1.
Wilby, R. L., Beven K. J. , and Reynard N. S. , 2008: Climate change and fluvial flood risk in the UK: More of the same? Hydrol. Processes, 22, 2511–2523, doi:10.1002/hyp.6847.
Wilk, M. B., and Gnanadesikan R. , 1968: Probability plotting methods for the analysis of data. Biometrika, 55, 1–17, doi:10.2307/2334448.
Xu, C.-H., and Xu Y. , 2012: The projection of temperature and precipitation over China under RCP scenarios using a CMIP5 Multi-Model Ensemble. Atmos. Oceanic Sci. Lett., 5, 527–533, doi:10.1080/16742834.2012.11447042.
Yang, H., and Yang D. , 2012: Climatic factors influencing changing pan evaporation across China from 1961 to 2001. J. Hydrol., 414–415, 184–193, doi:10.1016/j.jhydrol.2011.10.043.
Yong, B., and Coauthors, 2013: Spatial–temporal changes of water resources in a typical semi-arid basin of north China over the past 50 years and assessment of possible natural and socioeconomic causes. J. Hydrometeor., 14, 1009–1034, doi:10.1175/JHM-D-12-0116.1.
Zhang, L., Zhao F. , Chen Y. , and Dixon R. N. M. , 2011: Estimating effects of plantation expansion and climate variability on streamflow for catchments in Australia. Water Resour. Res., 47, W12539, doi:10.1029/2011WR010711.
Zhang, Q., Singh V. P. , Li J. , and Chen X. , 2011: Analysis of the periods of maximum consecutive wet days in China. J. Geophys. Res., 116, D23106, doi:10.1029/2011JD016088.
Zhang, Q., Li J. , Singh V. P. , Xu C.-Y. , and Bai Y. , 2012: Changing structure of the precipitation process during 1960–2005 in the Xinjiang, China. Theor. Appl. Climatol., 107, 255–264, doi:10.1007/s00704-011-0476-y.
Zhang, Q., Li J. , Singh V. P. , and Xiao M. , 2013a: Spatio-temporal relations between temperature and precipitation regimes: Implications for temperature-induced changes in the hydrological cycle. Global Planet. Change, 111, 57–76, doi:10.1016/j.gloplacha.2013.08.012.
Zhang, Q., Singh V. P. , and Li J. , 2013b: Eco-hydrological requirements in arid and semiarid regions: Case study of the Yellow River in China. J. Hydrol. Eng., 18, 689–697, doi:10.1061/(ASCE)HE.1943-5584.0000653.
Zhang, Q., Singh V. P. , Li K. , and Li J. , 2014: Trend, periodicity and abrupt change in streamflow of the East River, the Pearl River basin. Hydrol. Processes, 28, 305–314, doi:10.1002/hyp.9576.
