Correlation between hydrological drought, climatic factors, reservoir operation, and vegetation cover in the Xijiang Basin, South China

Accepted Manuscript Research papers Correlation between Hydrological Drought, Climatic Factors, Reservoir Operation, and Vegetation Cover in the Xijiang Basin, South China Qingxia Lin, Zhiyong Wu, Vijay P. Singh, S.H.R. Sadeghi, Hai He, Guihua Lu PII: DOI: Reference:

S0022-1694(17)30240-8 http://dx.doi.org/10.1016/j.jhydrol.2017.04.020 HYDROL 21951

To appear in:

Journal of Hydrology

Received Date: Revised Date: Accepted Date:

2 December 2016 20 March 2017 10 April 2017

Please cite this article as: Lin, Q., Wu, Z., Singh, V.P., Sadeghi, S.H.R., He, H., Lu, G., Correlation between Hydrological Drought, Climatic Factors, Reservoir Operation, and Vegetation Cover in the Xijiang Basin, South China, Journal of Hydrology (2017), doi: http://dx.doi.org/10.1016/j.jhydrol.2017.04.020

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Correlation between Hydrological Drought, Climatic Factors, Reservoir Operation, and Vegetation Cover in the Xijiang Basin, South China Qingxia Lin a,b,c, Zhiyong Wu a,∗, Vijay P. Singh b,c, S.H.R. Sadeghi b,c,d, Hai Hea and Guihua Lu a a

Institute of Water Problems, College of Hydrology and Water Resources, Hohai University, Nanjing

210098, China b

Department of Biological and Agricultural Engineering, Texas A&M University, College Station, TX

77843-2117, USA c

Zachry Department of Civil Engineering, Texas A&M University, College Station, TX 77843-2117,

USA d

Department of Watershed Management Engineering, Faculty of Natural Resources, Tarbiat Modares

University, P.O. Box 46417-76489, Noor, Iran

Abstract The Xijiang River is known as the Golden Watercourse because of its role in the development of the Pearl River Delta Regional Economic System in China, which was made possible by its abundant water resources. At present, the hydrological regime of the Xijiang River has now become complicated, the water shortages and successive droughts pose a threat to regional economic development. However, the complexity of hydroclimatological processes with emphasizes on drought has not been comprehended. In order to effectively predict and develop the adaptation strategies to cope with the water scarcity damage caused by hydrological droughts, it is essential to thoroughly analyze the relationship between hydrological droughts and pre/post-dependent hydroclimatological factors. To

∗

Corresponding author. E-mail address: [email protected] (Z.Y. Wu) 1

accomplish this, the extreme-point symmetric mode decomposition method (ESMD) was utilized to reveal the periodic variation in hydrological droughts that is characterized by the Standardized Drought Index (SDI). In addition, the cross-wavelet transform method was applied to investigate the correlation between large-scale climate indices and drought. The results showed that hydrological drought had the most significant response to spring ENSO (El Niño-Southern Oscillation), and the response lags in sub-basins were mostly 8–9 months except that in Yujiang River were mainly 5 or 8 months. Signal reservoir operation in the Yujiang River reduced drought severity by 52–95.8% from January to April over the 2003–2014 time period. Similarly, the cascade reservoir alleviated winter and spring droughts in the Hongshuihe River Basin. However, autumn drought was aggravated with severity increased by 41.9% in September and by 160.9% in October, so that the land surface models without considering human intervention must be used with caution in the hydrological simulation. The response lags of the VCI (Vegetation Condition Index) to hydrological drought were different in the sub-basins. The response lag for the Hongshuihe, Yujiang, and Liujiang River Basins were mostly 0–4 months, 0–1 months, and 2–3 months, respectively, but there was no obvious regular change pattern in the Guijiang River Basin. Keywords: Hydrological drought; Drought variability; Climate change; Vegetation dynamic; Disaster management; Water management

Introduction Drought is one of the most frequently occurring disasters in the Xijiang Basin, because of its uneven spatial-temporal distribution of precipitation (Fischer et al., 2013; Qiang et al., 2014; Niu et al., 2015). The basin experienced a serious drought under an extreme shortage of precipitation in 1962–1963,

2

and encountered a continuous drought that lasted nine years from 1984 to 1992. Severe or long lasting droughts have shown more predictable trends in recent years; in other words, a local or regional drought would occur each year from 2003 to 2015 in southern China (Yu et al., 2014; Zhang et al., 2014; Chen and Sun, 2017), and the Xijiang Basin had repeatedly suffered from severe droughts, which had a significant impact on the regional economy in 2004–2005, 2009–2010, and 2012–2013 (Niu et al., 2013; Zhang et al., 2013; Chen et al., 2015). The drought has not only attracting the attention of policymakers and the public, but it also has become a hot issue in international scientific research (Trenberth et al., 2014; Van Huijgevoort et al., 2014; Hao and Singh, 2015; Carlton et al., 2016). A natural precipitation shortage will eventually translate into a runoff deficit, which is a major indicator of hydrological drought (Van Loon, 2015; Wang et al., 2016). Like other types of drought, hydrological drought is characterized by spatio-temporal evolution, frequency of occurrence or return period, severity, durtion, and areal extent (Hisdal et al., 2004). In the Xijiang Basin, the spatio-temporal characteristics of these drought characteritics have been investigated using different methods, such as the Standardized Drought Index (SDI) (Huang et al., 2015; Niu et al., 2015) or the copula function (Zhang et al., 2012, 2013; Chen et al., 2013; Wu et al., 2015). However, it is remaining unclear as to why drought exhibit these spatio-temporal characteristics, so as to what is the drought effect on the underlying surface. Hydrological drought regularity is mainly influenced by climate change and human activities (Harding et al., 2014; Gan et al., 2016; Van Loon et al., 2016). The former influences the drought process through spatio-temporal variation of precipitation (Leng et al., 2015; Sam et al., 2016), while human activities, such as water conservancy project operations, lake reclamation, and changes in forest cover, change the status of hydrological droughts (Davudirad et al., 2016; Liu et al., 2016). However, the 3

hydrological model often has been blindly used to model drought without considering human intervention. Drought effects vary according to specific drought characteristics, and are directly reflected in the change of vegetation cover (Kumar et al., 2015; Gouveia et al., 2016). To understand the causes and the effects of drought, the influence or reponse of climate variability, human activity, and vegetation cover on hydrological drought must be investigated. Current studies have quantified the individual connection based on regression analysis, sensitivity analysis, and hydrological modeling (Niu et al., 2015; Byzedi and Khazaei, 2016; Huang et al., 2016), and show that the connection vary according to research method, time scale, and study region (Cheng et al., 2016; Liu et al., 2016). In these studies, the effects of climate change on runoff have been primarily focused on future scenarios. The impacts of human activities on drought have been primarily investigated with hydrological models (with or without water management module) that have inherent uncertainties, which can be subdivided in structural uncertainty, parametric uncertainty and numeric uncertainty (Van Loon, 2015). Meanwhile, the impacts are mostly evaluated on an annual scale rather than on seasonal or monthly scales, which are important for determining the proper allocation of water resources during the year. Futhermore, a comprehensive analysis on the causes and effects of hydrological drought (aside from runoff deficit) is still needed. The Xijiang Basin has a complicated climate and is dramatically affected by human activities. It is influnced by monsoon circulation because of its geographical location (Zhu et al., 2012), and has a number of large-scale hydropower reservoirs for power generation, flood control, and shipping. The basin has widely distributed karst formations and a large ecologically fragile area, so the vegetation cover is sensitive to drought. Therefore, the causes and effects of hydrological drought in the Xijiang Basin are more complex than that of the river basins in southern China. An investigation of the connections between hydrological drought, climatic factor, reservoir operation, and vegetation cover in 4

the Xijiang Basin is very important for drought management. In this research, the effect of climate change on hydrological drought over the past 55 years was investigated by applying the cross-wavelet transform method to drought severity and the large-scale climatic factor indices (AMO, AO, ENSO and PDO). The effect of human activity on hydrological drought in different seasons was investigated using the measured daily runoff of large hydropower reservoirs and the Muskingum method. In addition, the seasonal sensitivity of vegetation cover to hydrological drought was identified by combining the Global Inventory Modeling and Mapping Studies (GIMMS) dataset and the Moderate Resolution Imaging Spectroradiometer (MODIS) remote-sensing Normalized Difference Vegetation Index (NDVI) dataset. The results of this study will improve our understaning of hydrological drought characteristics in a complicated hydrological regime, and provide a basis for drought prediction and water resource allocation in the Xijiang Basin.

1. Study area and data series The Xijiang River Basin spreads across the middle and southern subtropic region of China and has an area of 360,000 km2, which accounts for 79% of the total area of the Pearl Basin. The river originates in Yunnan and flows through the Yunnan, Guizhou, Guangxi, and Guangdong Provinces, ultimately flowing down into the South China Sea. The average annual runoff of the Xijiang River is second to that of the Yangtze River, five times that of the Yellow River, and 4.5 times that of the European Rhine River, but its sediment concentration is only one-third that of the Yangtze River. The water resources of the Xijiang River are abundant but are unevenly distributed; the resources decrease in quantity from southeast to northwest. The main stream is 2,214 km long and consists of five river sections: Nanpanjiang, Hongshuihe, Qianjiang, Xunjiang, and Xijiang. An overview of the basin, its main river, and its tributaries is presented in Fig. 1. 5

The position of Fig. 1 The 1961–2014 daily precipitation dataset at 0.5°×0.5°.resolution was download from the China Meteorological Data Network (http://data.cma.cn/). The datasets for the daily runoff and daily regulation of hydropower reservoirs were obtained from the Ministry of Water Resources Information Center. The daily runoff dataset represents the daily discharge at five representative stations in the 1961–2014 time period: Hongshuihe (Qianjiang station), Yujiang (Guigang station), Liujiang (Liuzhou station), Guijiang (Pingle station), and Xijiang River Basin (Gaoyao station). The corresponding missing value percentages at five hydrological stations are 2.17%, 0.49%, 4.27%, 5.32% and 4.64%, respectively. The measured runoff and the simulated runoff (Lu et al., 2015) from Variable Infiltration Capacity (VIC) model were assumed to have the equal frequency on a same missing day, then the value in the measured series with that frequency was chosen to insert. The regulation dataset is comprised of the daily inflow and outflow datasets under the regulation of the Tianshengqiao, Longtan, and Yantan hydropower reservoirs (2004–2014) on the Hongshuihe River, and the Baise reservoir (2003–2014) on the Yujiang River. The large-scale climate indices investigated in this study include the Atlantic Multidecadal Oscillation (AMO), the El Niño/Southern Oscillation (ENSO), the Arctic Oscillation (AO), and the Pacific Decadal Oscillation (PDO). The AMO and ENSO indices were derived from the National Oceanic and Atmospheric Administration (NOAA) Earth System Research Laboratory (www.esrl.noaa.gov/psd/data/correlation/amon.us.data and www.esrl.noaa.gov/psd/data/correlation/nina34.data), the AO index was derived from the NOAA National Climatic Data Center (www.ncdc.noaa.gov/teleconnections/ao/), and the PDO index was derived from the Tokyo Climate Center (ds.data.jma.go.jp/tcc/tcc/products/elnino/decadal/annpdo.txt). 6

The long-term GIMMS vegetation index for 1983–2003 time period was provided by the Environmental and Ecological Science Data Center for West China, and the National Natural Science Foundation of China (westdc.westgis.ac.cn). The high-precision MODIS vegetation index data was obtained

from

the

Atmosphere

Archive

and

Distribution

System

(LAADS,

ladsweb.nascom.nasa.gov/search/).

2. Research methodology 3.1 Hydrological drought index and drought identification Assuming that the runoff data followed a probability distribution such as gamma or another function, the SDI could be obtained by normalizing the runoff series(Tabari et al., 2013; Hong et al., 2015; Hao et al., 2016). It was used for characterizing the degrees of drought at different spatial scales and was calculated as follows: SDI =

Ri − R

δR

(1)

where SDI denotes the standardized drought index,
 stands for the runoff in line with a distribution function,
 and  represent the multi-scale (yearly, monthly, or daily) mean and square deviation of the
 values, respectively. Based on a previous study on the Xijiang Basin (Wu et al., 2015), the log-normal distribution was used for fitting runoff data, and then the SDI was constructed. 3.2 The extreme-point symmetric mode decomposition method (ESMD) Empirical Mode Decomposition (EMD) is a popular method used for analyzing nonlinear and nonstationary time series, because its decomposition frequency and amplitude are variable (Sang et al., 2014). However, the EMD method has a disadvantage in that the filter number is uncertain and the decomposition residual is too coarse, hence it is not ideal for the Hilbert-Huang transform (HHT) of the 7

frequency of each eigenmode. At the same time, there is no accepted solution to the problem of boundary treatment, which plagues the EMD (Li et al., 2015). The extreme-point symmetric mode decomposition (ESMD) is a new adaptive method used for the analysis of nonlinear and non-stationary time series. It uses the idea of EMD to transform external envelope interpolation into internal extreme-point symmetry interpolation, and borrows the idea of least squares to optimize the final residual mode, in order to generate an adaptive global mean curve. It abandons the traditional method of spectrum analysis, which relies on integral transforms, and instead creatively proposes a direct interpolation (DI) method that can consider the inherent defects of all integral transforms. These transforms include the Hilbert transform, which analyzes time-frequency variations, and intuitively reflects the time-varying properties of amplitude and frequency of each mode. Wang and Li (2013) have described the principles, advantages, and disadvantages of the method. In this study, the ESMD method was chosen to study the cyclic characteristics of hydrological drought in each sub-basin. 3.3 Cross-wavelet transform method Wavelet analysis has a good time-frequency localization feature and strong multi-resolution analysis ability, but can only describe the time-frequency characteristics of a single time series.Therefore, it is difficult to analyze the mutual influence and time-frequency correlation of multiple time elements. The cross-wavelet transform method is a new multi-signal and multi-scale analysis technique based on traditional wavelet analysis (Miao et al., 2014). It is capable of not only analyzing the correlation between time series, but it can also reflect the phase structure and detailed features in the time-frequency domain. Suppose that  (S) and  (S) are the continuous wavelet transforms of two time series: X = , , … ,   and Y =  ,  , … ,  . The cross wavelet transform between them would 8

be  (S) =  (S) ∗ (S), where  ∗ (S) represents the complex conjugate of  (S), and S is the delay (also referred to as time shift). The power spectrum of the cross wavelet transform can be defined as | (S)|, which contains the time-frequency-amplitude information. When the value is greater, the degree of correlationbetween the time series is higher. For two stationary random processes, the standardized form of the cross-wavelet transform can be written as the wavelet cross-correlation coefficient:

∑

r( X ,Y ) =

n i =1

(Wi X ( s ) − Wi X ( s ))(WiY ( s ) − WiY ( s ))

n

∑ i=1 (Wi X ( s) − Wi X ( s))2

n

(2)

∑ i =1 (WiY ( s) − Wi X ( s))2

In this study, the cross-wavelet transform method was used to study the response of hydrological drought to different climatic factors.

3. Results and discussion The above scientific datasets and research methods were employed to investigate the spatial-temporal characteristics of hydrological drought and the correlation between drought, climatic factor, reservoir operation, and vegetation cover, in order to achieve a better understanding of the causes and effects of drought. 4.1 Statistical characteristics of hydrological drought 4.1.1 Typical hydrological drought events since the 2000s Since the 2000s, hydrological drought has been a serious concern for decision makers because of its frequent occurrence in and widespread influence on the Xijiang Basin. Hydrological drought has regional characteristics, so evaluating its characteristics based only on the discharge of a basin would most likely overlook local drought features. Therefore, the smallest sub-basin area, Guijiang Basin, was used as a reference to establish whether a drought did or did not occur. Thus, typical hydrological 9

drought events in the Xijang Basin were obtained based on area threshold, and are listed in Table 1. The position of table 1 Table 1 indicates that severe hydrological droughts occurred on a nearly annual basis in the district. A total of seven hydrological droughts, which started in summer or autumn and lasted until the following spring or summer, were identified in the 2003–2013 time period. Since 2003, the most serious hydrological droughts mostly appeared in the Hongshuihe river. In the 2003–2004 drought period, the driest areas were the Liujiang and Guijiang River Basins in July through September. These were consistent with the actual drought disaster that occurred in the 2003–2004 period (Huang et al., 2004). From 2004 to 2008, the Xijiang Basin suffered droughts so severe that the basin management institution was forced to transfer salt water into the area to supplement the freshwater supply. This action was taken in order to deal with the serious droughts that occurred in the in 2004–2005, 2005–2006, and 2007–2008 time periods. In the 2009–2010 period, the Hongshuihe experienced its worst runoff deficit in September through November of 2009. Subsequently, the drought center transferred to the Liujiang in January through May of 2010, and the worst hydrological drought periods were in March and April of 2010. Actually, southwest China had suffered a serious blow from an extreme drought that had drawn nationwide attention in the 2009–2010 period (Barriopedro et al., 2012). The previously mentioned studies demonstrated that the occurrence periods and regional locations of identified droughts correlated with actual occurences, therefore the SDI-based research method for hydrological droughts was validated. 4.1.2 Annual change in drought severity As previously mentioned, the log-normal distribution was chosen to fit natural runoff and to obtain the SDI in four typical sub-basins. Drought severity based on the daily scale was calculated and 10

exptrapolated to the annual scale; its variation process is shown in Fig.2. The position of Fig. 2 Fig. 2 shows different variations in hydrological drought severity among the four sub-basins, but all of the sub-basins shared the same three severe periods: 1963–1966, 1986–1993 and the whole 2000s. Hongshuihe shows four extremely unusual drought years. Sorting by severity in descending order, those years were 2011, 1963, 1989, and 2009. The extremely drought years in the Yujiang River Basin were 1963 and 2006. In the Liujiang River Basin, only one year, 1963, showed an extremely serious situation. The years 1965, 1998, and 1999 were the three abnormal years for the Guijiang River Basin. The drought severities showed increasing trends of 6.6 SDI/10a, 3.8 SDI/10a, and 5.3 SDI/10a in the Hongshuihe, Yujiang, and Luijiang River Basins, respectively, while exhibiting a declining trend, -4.5 SDI/10a, in the Liujiang River Basin. The desecnding ratios of precipitation in Hongshuihe, Yujiang, Liujiang, and Guijiang sub-basins were 28.2 mm/10a, 12.1 mm/10a, 21.7 mm/10a, and 15.6 mm/10a, respectively. Although precipitation decreased in Guijiang, the annual runoff coefficient showed an increasing trend of 0.19/100a, indicating that the hydrological cycle of the underlying surface also affected the characteristics of hydrological drought. 4.1.3 Periodic variation in Hydrological Drought Periodic variation in hydrological drought characteristics is an important factor to consider when examining regional water resource evolution. Hence, intra-annual, inter-annual, and inter-decadal cycles of four sub-basins from 1961 to 2014 were analyzed using the ESMD method. The decomposition results of the Hongshui River Basin are plotted in Fig. 3. The position of Fig. 3 In Fig. 3, the previous seven decomposition intrinsic mode functions (IMFs) of SDI series showed 11

that the drought variation in temporal scales was gradually increasing from IMF1 to IMF7. Specifically, the first and second IMFs mainly reflected the intra-annual shocks with quasi-periodic periods of 3 and 6 months. The third to sixth IMFs reflected inter-annual shocks of 1, 2, 3.4, and 9.6 years, and the seventh IMF indicated a 17 year inter-decadal shock. The research on periodic variations of other tributaries showed that in the Yujiang River Basin, there existed intra-annual shocks of 3 and 6 months, inter-annual shocks of 1, 2.1, 4.2, and 9.7 years, and inter-decadal shocks of 15.8 years. The Liujiang River Basin had intra-annual shocks of 3 and 7 months, inter-annual shocks of 1, 1.9, 3.8, and 7.6 years, and an inter-decadal shock of 14.4 years. The quasi-periodic shocks of the Guijiang River Basin were 3 and 6 months, 1,1.9, 3.4, and 6.8 years, and 11.2 and 18.7 years. In general, the periodic variations of hydrological droughts in the four tributary basins were relatively consistent, with intra-annual shocks of approximately 3 and 6 months, inter-annual fluctuations of 1, 2, 3.4–4.2, and 6.8–9.7 years, and inter-decadal shocks of 14.4–18.7 years. In fact, the precipitation had 3 and 7–9 months of intra-annual shocks, 1.2, 2, 3–3.6, and 5.7–7.4 years of inter-annual shocks, and 10.7–17.4 years of inter-decadal shocks in the four tributary basins. Therefore, the periodic variation in hydrological drought in the Xijiang River Basin was mainly affected by precipitation cycle. 4.2 Response in hydrological drought to large-scale climatic factors In the field of drought formation research, it has been recognized that changes in global ocean temperature or large-scale climatic factors such as ENSO, North Atlantic Oscillation (NAO), and PDO may be crucial for drought formation on the inter-decadal scale (Hosseinzadeh Talaee et al., 2014; Zhang et al., 2015). Accordingly, correlations between hydrological drought severity and the four climatic factors (AMO, AO, ENSO and PDO) are studied through cross-wavelet transform analysis (Fig. 4). 12

The position of Fig. 4 Fig. 4 shows the various temporal-scale correlations between hydrological drought and AMO, AO, ENSO, and PDO. The arrows pointing from left to right indicate that the hydrological drought is in phase with the change in climate index, and the arrows pointing from right to left indicate the negative phase. The arrows pointing vertically downward or upward represent the drought ahead or behind the climate index of the 1/4 cycle, and they had a non-linear correlation. The correlation between hydrological drought and ENSO was higher than that of other indices. Fig. 4(c) shows that the effect of ENSO on hydrological drought was mainly on the scale of 8–45 months. The positive correlation between ENSO and hydrological drought was mainly reflected on the scale of 20–40 months in 1970–1975 and on the scale of 10–32 months in 2006–2014. On the contrary, the negative correlation between them was mainly manifested on the scale of 30–45 months in 1986-2000. In order to reveal the response process of hydrological drought to ENSO events more clearly, the temporal variation of drought severity and ENSO anomaly percentage in 1961–2014 are plotted in Fig. 5. The position of Fig. 5 Fig. 5 shows that the hydrological drought severity and the ENSO anomaly percentage yielded slowly increasing trends over the 1961–2014 time period. A peak value of ENSO was usually accompanied by a peak value of drought severity (①~⑩). When the ENSO peaks increased, most of the hydrological drought peaks increased as well, meaning that the ENSO and hydrological drought shared a good correlation with each other. For this reason, the correlation analysis of the ENSO index and the monthly SDI during the ENSO events was carried out, and the results are listed in Table 2. The position of table 2 Table 2 shows that the hydrological droughts were most sensitive to spring ENSO events in the four 13

sub-basins. After spring, winter ENSO events had the next most significant impact on the droughts, and autumn events showed no impact. The efficacy of the spring ENSO events on droughts has to do with the warm phase of ENSO in spring, which causes the South China Sea Monsoon to be weaker and occur later, thus strengthening the East Asian Subtropical Monsoon. For example, the South China Sea (SCS) summer monsoon onset index showed that the weakest summer monsoon years before the year 2000 were 1970, 1983, 1987, 1988, 1993, 1995, 1996, and 1988, and the late outbreak years were 1963, 1968, 1970, 1973, 1975, 1982, 1987, 1991, and 1993. In fact, hydrological droughts occurred in the Xijiang Basin in many of these years (1963, 1970, 1974–1975, 1987, 1988, 1991, 1993, 1998). At the same time, the summer typhoon intensity of the coastal areas in China had a significantly negative correlation with the subtropical monsoon (Wang et al., 2006). The subtropical monsoon over East Asia was not conducive to the formation and development of tropical cyclones in Guangxi, so the mitigation effect of hydrological drought was limited. In addition, the winter ENSO warm phase would cause the southwest monsoon onset time to occur later. For example, sea surface temperature anomalies in the winter of 1986 and 1990 were relatively high, and the onset of the summer monsoon in the following year (1987, 1991) was relatively late. Correspondingly, hydrological droughts occurred in the Xijiang Basin during the 12/1986–8/1987, 12/1990–5/1991, and 7/1991–12/1991 time periods. The degree of response to the spring ENSO in the four tributary basins was varied. The responses of the Liujiang and Guijiang River Basins were higher, as both correlation coefficient had reached more than 0.6. This was because the runoff coefficients of the Liujiang and Guijiang Rivers were higher than those of the Hongshuihe and Yujiang Rivers. The reduction in the equivalent precipitation caused by ENSO would cause the runoff of the former basins to decrease more significantly, and the presence of the droughts would be more evident. Meanwhile, the dispatching of large hydropower reservoirs on the 14

former basins would inevitably reduce the sensitivity of the regional hydrological drought response to the ENSO events. However, droughts in different tributary basins had a relatively consistent response in the lag times for spring ENSO events. The lag times were 8–9 months in most cases in the Hongshuihe, Liujiang, and Guijiang River Basins, and the response time was 5 or 8 months in the Yujiang River Basin. It is possible to predict the hydrological drought in the sub-basins in autumn and winter based on the strengthened ENSO in spring. However, there is an obvious seasonal dependence on the prediction error of the ENSO event, and the maximum error usually rises quickly in spring. At the same time, the cold and warm phases of the ENSO event are significantly asymmetric, and so the prediction error of positive phase ENSO is stronger than that of the negative phase (Duan and Wei, 2013; Yu et al., 2014). Therefore, predicting regional hydrological drought based on the spring ENSO index still needs to overcome the spring predictability barrier (SPB). 4.3 Effect of hydropower reservoir operation on hydrological drought A number of regulatory cascade hydropower reservoirs, such as the Tianshengqiao, Longtan, Yantan, and Baise reservoirs have been built to generate electricity or to control heavy flooding in the Xijiang River Basin. Their operation mode would directly affect the temporal characteristics of hydrological drought. Here, the operation effects were investigated based on the actual runoff datasets of large-scale hydropower reservoirs by using the Muskingum method to deal with the runoff reduction calculation. 4.3.1 Effect of single reservoir operation on hydrological drought The Baise reservoir is a large-scale hydropower project equipped with comprehensive utilization of power generation, irrigation, shipping, and water supply for the Xijiang tributaries (Yujiang River). The

15

normal water level of the reservoir is 228 m and its corresponding storage capacity is 4.8 × 1010 m3. The dead water level is 203 m and the corresponding storage capacity is 2.18 × 1010 m3. The reservoir has incomplete multi-year regulation and the regulation storage capacity is 2.62 × 1010 m3. The inter-annual variability and intra-annual distribution of hydrological drought severity were investigated based on the SDI series of the Guigang station under natural condition and reservoir scheduling during the 2003–2014 period (Fig. 6). The position of Fig. 6 Fig. 6 describes the time-dependent variation of the hydrological drought severity in 2002–2014. It illustrates a good reflection of the development phases of the Baise reservoir, including the beginning of the construction phase (October 2001), the river closure phase (October 2002), the water storage phase (October 2006), and the beginning of the operation phase (2007). This indicates that the reservoir operation data is reasonable. The inter-decadal variation of drought severity showed that the Baise reservoir partly alleviated hydrological droughts in 2007, 2009, and 2013. The intra-annual distribution of drought severity showed that the mitigation effect of the Baise reservoir on drought mainly occurred from January to April, during which time the drought severity decreased by 95.8%, 96.5%, 73%, and 52%, respectively. This means that the relieving effects took place primarily in winter and spring. Two typical hydrological droughts that occurred during the 2006–2007 and 2009–2010 periods in the Yujiang River Basin were selected in order to investigate the impact of hydropower reservoirs on the occurrence and development of hydrological drought (Fig. 7). The position of Fig. 7 In Fig. 7, reservoir adjustment volume refers to the difference between the outflow and inflow of the reservoir. A positive value means the outflow is greater than the inflow, and a negative value means 16

the opposite. The reservoir mainly reduced the 2006–2007 drought from December to February, during which the drought severity decreased by 86.1%, 98.5%, and 100%, respectively. However, the engineering operation had little effect on the drought from April to August 2007. The natural hydrological drought at Guigang station in 2009–2010 was characterized by two components: the short-term drought (from February to March) in the first half and the long-term drought (from August to December) in the second half of the year. The reservoir had a strong mitigating effect on the drought in the first half of the year, achieving a reduction in drought severity of 98% in February and March. However, in the second half of the year, there was only a small adjustment to the hydrological drought. 4.3.2 Effect of cascade reservoir operation on hydrological drought The Hongshuihe is the main river for the cascade hydropower reservoirs in the Xijiang Basin, and it contains three large-scale reservoirs: the Tianshengqiao, Longtan, and Yantan projects. Tianshengqiao is a multi-year regulating reservoir built for the main purpose of generating electricity. Its total capacity is 1.03 × 1011 m3, of which the adjustment capacity is 5.80 × 10 10 m3. Longtan hydropower is the second largest reservoir in China (the Three Gorges station is the largest). Its normal water level is 400 m and its total capacity is 2.73 × 10 11 m3. Yantan hydropower is the second largest reservoir in the Guangxi province. It is located in the middle of the Hongshui River and belongs to the annual regulation reservoir. Its normal water level is 223 m and the corresponding capacity is 2.60 × 10 10 m3, of which the adjustment capacity is 1.56 × 10 10 m3. The inter-annual variability and intra-annual distribution characteristics of hydrological drought severity were analyzed based on the SDI series under the influence of natural status and reservoir scheduling in 2004–2014 (Fig. 8). The position of Fig. 8 Fig. 8 shows that the drought under the influence of cascade reservoir joint scheduling has a 17

complex relationship with that under natural condition. The inter-annual variability of the drought severity indicated that the reservoir operation alleviated the hydrological drought to a certain extent in 2007 and 2011, but the drought intensified in 2009 and 2013. The intra-annual distribution of drought severity showed that hydropower reservoirs provided drought relief primarily in winter and spring, because drought severities decreased by 60.0% and 92.9% in February and March, respectively, and by 73.2% and 41.5% in May and June, respectively. However, the adjustment of the reservoir intensified severe autumn drought. The drought severity in September increased by 41.9%, while that in October was as high as 160.9%. Here, a typical hydrological drought in 2009–2010 in the Hongshuihe River Basin was selected to explore the effect of cascade reservoir joint operation on hydrological drought (Fig. 9). The position of Fig. 9 Fig. 9 shows that the peak of hydrological dryness during the 2009–2010 drought event, which was identified based on the runoff data with a reverse calculation of reservoir operational log, was observed in December 2009 and March 2010. These two periods correspond to the peak of ENSO and the southwest meteorological drought, respectively, indicating that the reverse calculation was reasonable. In the figure, drought relief was more evident in March 2010, but the drought under the reservoir regulation anomaly intensified from August to November 2009. Based on the scheduling process of three major hydropower reservoirs on the Hongshuihe River, the water storage operation in the Longtan and Yantan reservoirs from July 2009 to January 2010 were frequent, which increased the drought severity by 994.3%, 582.0%, and 420.7% from August to November in 2009, respectively. The inter/intra-annual variability of the hydrological drought severity and the typical drought processes in the natural state or under hydropower reservoir scheduling showed that the increased 18

discharge of a hydropower reservoir for the purpose of shipping, irrigation, etc. in the dry season relieved winter and spring droughts. However, when there was less water in flood season, hydropower reservoirs had still retained the water for the purpose of power generation, hence the summer and autumn droughts could not be effectively alleviated, and were, in fact, further aggravated. The hydrological modeling methods without considering human intervention (land cover change, reservoir operation, agricultural irrigation and water withdrawal from river channels, etc.) have been widely used, especially in Asian countries, to reveal the regional drought characteristics in many cases (Liu et al., 2016; Van Loon et al., 2016). However, the results indicate the blindly use of the hydrological model worthy of serious treatment. Hydrological drought caused by human activities has become a new challenging issue that needs to be noticed and discussed. The storage design capacities of the hydropower reservoirs on Hongshuihe and Yujiang have not been fully explored for opportunities to increase discharge and control the drought during the impoundment period when the precipitation is too low. The reason why hydrological droughts in summer and autumn are exacerbated under the operation of hydropower reservoirs is that regulations give priority to social production and the livelihood of the people. For example, the Hongshui River experienced the most severe hydrological drought since its hydrological record in the summer of 2011, and the electricity shortage in Guangxi reached 30%, which was the most significant electricity shortage in the past 20 years. At this time, in order to maintain stable operation of the power network and reduce the cost of drought resistance and drought loss, the major reservoirs in the Hongshui and Yujiang River Basins were rarely used for drought regulation. Instead, they were primarily used to conserve water to increase power generation. Therefore, the operation of hydropower reservoirs only alleviated hydrological and agricultural droughts in spring, and economic droughts during the peak period of 19

electricity consumption in summer and autumn. However, the hydrological and agricultural droughts in summer and autumn were exacerbated. In order to alleviate the impact of hydropower reservoirs on hydrological drought, the storage time of the reservoir can be prolonged appropriately, so that discharge can be increased to reduce the impact of general drought during the impoundment period. At the same time, coal storage in the region can be expanded so that when the water supply is insufficient, the power gap can be supplemented by thermal power. In addition, the encounter probability analysis of regional hydrological droughts should be strengthened. When a region is in a drought with insufficient electrical energy, the near drought-free basin can provide the electricity so that more water resources in the basin can be allocated to alleviate the drought. 4.4 Response of vegetation to hydrological drought The evolution of meteorological–hydrological factors directly affects the growth of vegetation (Sawada et al., 2014). The vegetation ecosystem had a certain tolerance to drought disturbance, and a persistent dry spell would have a significant impact on the composition, structure, and function of a vegetation ecosystem (Van Loon, 2015). The VCI is a normalization of NDVI for filtering out the contribution of local geographic resources to the spatial variability of NDVI (Quiring and Ganesh, 2010; Belal et al., 2014). It has a better drought reflection ability in many parts of the word and has been applied in daily drought monitoring by National Atmospheric and Ocean Bureau (NOAA) of United States and National Satellite Meteorological Center (NSMC) of China. The delayed response of sub-basin vegetation to hydrological droughts over different months was studied based on VCI (Fig. 10). The SDI and VCI in a typical drought (Fig. 11), and the spatial variation of VCI are plotted in Fig. 12. The position of Fig. 10 20

The positive correlation coefficient, denoted by 0 in Fig.10, indicates that there was no correlation between hydrological drought and VCI. The average correlation coefficients of sub-basins in 1984–1993, 1994–2003, and 2004–2014 were -0.51, -0.42, and -0.66, respectively, indicating that the negative correlation in 1984–2014 first experienced a decline before an increase. This change was particularly evident from September to December, which yielded average correlation coefficients of -0.43, -0.24, and -0.77 in the three periods, respectively. The correlation coefficients in the Liujiang and Guijiang River Basins were 13.7% and 9.8% higher than the mean in 1984–1993. The coefficients ranged from -0.4 to -0.43 for 1994–2003 and that in the sub-basins were consistent. In 2004–2014, the correlation in the Yujiang River Basin was 9.1% higher than the average level. The Xijiang Basin has a diverse vegetation ecosystem that encompasses forests, grasslands, rocky hills, and wetlands. The average growth rate of the VCI was 2.2/10a in 2004–2011 and mainly improved as the forest changed. The annual net growth rate of forests in the Guangxi province was 1.25% from 2005 to 2010. However, the forest increases mainly manifested in the growth of man-made forests, while natural forests degenerated into shrubs, grasslands, and secondary forests. Man-made forests had the characteristics of fast growth, thus the shortage of water for several months had a significant influence on the plantation. In order to avoid drought disaster, the Chinese fir and the Pinus massoniana were mainly chosen from the coniferous plantations to grow in the basin, and so did the Vernicia fordii and Eucalyptus. All of the plantations had strong drought resistance ability. For example, the growth of Vernicia fordii would not be too slow or wither until the drought persisted for more than two months, and the genes of the Eucalyptus population can naturally adapt to a drought environment. However, man-made forests are vulnerable to drought disaster because of their low diversity and weak regulation ability. For instance, the planting area of man-made Eucalyptus plantations in southern and southeastern 21

Guangxi occupied 44.3% and 35.1% of total forest area, respectively. The negative correlation between drought severity and VCI in Yujiang was more significant than that in other sub-basins during the 2004–2014 period. Therefore, although the drought severities were similar in the 1983–1993 and 2004–2014 time periods, the latter had higher vegetation coverage, therefore its vegetation change was more susceptible to hydrological drought. On balance, as the Xijiang Basin develops through the ecological transformation period, where the natural forest is restored and man-made forests become more diverse, the negative correlation of VCI and hydrological drought would gradually return to the level of 1983–1994, or lower. The negative correlation between drought severity and VCI decreased in the 1984–1993 and 1994–2003 periods. The 1994–2003 period in particular showed strong rates of decrease: 0.056/m, 0.057/m, 0.038/m, and 0.042/m in the Hongshuihe, Yujiang, Liujiang, and Guijiang Basins, respectively. The decreasing trend over the course of one year indicated that the hydrological drought in spring had a higher impact on vegetation coverage than other seasons in 1984–2003. In contrast, the negative correlation of each sub-basin increased in the 2004–2014 period. Although the increasing trend was weaker than the decreasing trend, this reality still indicated that the effects of autumn and winter drought on vegetation coverage were beyond the spring drought. In the figure, the response lag of VCI to hydrological drought in different sub-basins varied over the 1984–2014 period. The lags in the Hongshuihe and Yujiang Basins were generally 0–4 months and 0–1 month, respectively. In the Liujiang Basin, the lag times lastly mostly for 2–3 months, or 7–8 months less frequently. There was no general rule for the Guijiang Basin. The lag time in the Hongshuihe, Yujiang, Liujiang and Guijiang River Basins were mostly 4, 0, 3 and 1 (or 3) months during 1984–1993, respectively, generally 2, 1, 2–3, or 8–9 months during 1994–2003, and 0, 1, 2 (or 7), 22

and 4 months during 2004–2014. The typical drought in the 2004–2005 period was selected to plot the VCI variation along SDI (Fig. 11), and to explore the vegetation response to a typical hydrological drought event in different sub-basins. The position of Fig. 11 Fig. 11 shows that the 2004–2005 continuous drought event was concentrated in October 2004 and was mainly located in the Yujiang Basin. The lag time of the VCI to hydrological drought in the Hongshuihe, Yujiang, Liujiang, Guijiang River Basins was 2, 3, 5, and 5 months, respectively. In the basin where the vegetation coverage was higher (the Liujiang and Guijiang Basins), the delay of vegetation variation to drought was longer. The moving VCI sequences (moving with lag time) showed good consistency with SDI, and the negative correlation coefficients above the sub-basins were -0.11, -0.41, -0.53 and -0.70, respectively. This indicates that the vegetation change had a good response to hydrological drought. The spatial variability of vegetation after hydrological drought in October, January, March, and May during a typical 2004–2005 drought was investigated. The VCI anomaly percentage for the above four months was calculated using the monthly average VCI during 2004–2014, and was used to describe the degree of vegetation cover. Anomaly percentages less than -35% stand for extremely low vegetation cover, -35% to -25% for low vegetation coverage, -25% to -15% for relatively low vegetation coverage, -15% to 15% for normal coverage, 15% to 25% for relatively high vegetation coverage, 25% to 35% for high vegetation coverage, and more than 35% for extremely high vegetation coverage. The spatial distribution of VCI is plotted in Fig. 12. The position of Fig. 12 23

The 2004–2005 hydrological drought began in August 2004 and peaked in October 2004. In August, the vegetation was still in the normal growth state and the VCI anomalies were mostly between -15% and 15%, indicating that there is a lag effect of vegetation growth to hydrological drought (Fig. 12a). In January 2005, the vegetation coverage in the middle of the basin was significantly lower than the average value. The area ratios of relatively low, low, and extremely low covers were 12.2%, 6.5% and 14.9%, respectively. The lower coverage was concentrated in the downstream of the Liujiang River Basin, and the low and very low values were mainly located in the downstream of the Qianjiang and Yujiang River Basins. The hydrological droughts in January 2005, December 2004, October 2004, and October 2004 in above sub-basins had the greatest effect on the vegetation reduction in March 2005, respectively. The dryness of the Hongshuihe River Basin had short-term relief in March, and that of the Yujiang basin was reduced to a light drought value. However, light and medium droughts occurred in the Liujiang and Guijiang basins, respectively, and the extremely low vegetation cover in the central part of the basin transferred from the south to the north. In January and March 2005, the low vegetation cover area was mostly located in low-lying hills and pain areas, and forest land and crops mainly distributed on it. The hydrological drought in the Xijiang Basin was relieved in May, but the vegetation coverage in the east and west were still lower than the mean value, except that the low vegetation coverage was dispersed and the widely-spread vegetation damage disappeared (Fig. 12d). Fig. 12 shows that the vegetation damage has a continuous distribution and it changes with the hydrological drought dynamics in the spatial. Therefore, the response lag of vegetation to drought in sub-basins revealed here is beneficial for formulating distinct plans for crop protection and afforestation in different sub-basins.

4. Conclusion The hydrological drought severities of the four tributary sub-basins in the Xijiang Basin reflect 24

different variations from 1961 to 2014. The droughts in the Hongshuihe and Yujiang River Basins are more serious than those in the Liujiang and Guijiang Basins. Drought severities in the Hongshui, Yujiang, and Liujiang River Basins show upward trends (6.6 SDI/10a, 3.8 SDI/10a, 5.3 SDI/10a), while the drought severity in Guijiang shows the opposite. However, all tributary basins indicate three severe drought periods: 1963–1966, 1986–1993, and 2000–2009. The periodic variations of hydrological drought in the four tributary basins are mainly influenced by the variation in precipitation cycle, and their variations were consistent from 1961 to 2014, including 3 and 6 months of intra-annual shocks, 1a, 2a, 3.4a–4.2a, 6.8a–9.7a of inter-annual shocks, and 14.4a–18.7a of inter-decadal shocks. In four tributary basins, hydrological droughts have the most significant response to the spring ENSO index and then to that in winter, but has little response to autumn index. The response levels in individual tributary basins are different. It is higher in Liujiang and Guijiang sub-basins for the correlation coefficients between the spring ENSO index and drought severity can reach more than 0.6. However, the response lags in the tributary basins are only slightly different; they are 8–9 months for the Hongshuihe, Liujiang, and Guijiang River Basins, and 5 months or 8 months for the Guijiang River Basin. Hydrological drought under the regulation of the single reservoir in the Yujiang Basin is mainly alleviated from January to April. The cascade hydropower reservoir regulations have eased the winter and spring drought in Hongshuihe River Basin. However, the regional drought in summer and autumn cannot be effectively alleviated and even intensified, with drought severity increasing by 41.9% in September and increasing even further to 160.9% in October. Therefore, the land use models should consider human intervention when using for hydrological modeling in the research region. The correlations between drought severity and VCI had experienced a decline first and then an 25

increase from 1984 to 2014. These variations are particularly evident from September to December, when the average correlation coefficients in the three historical periods, 1984–1993, 1994–2003 and 2004–2014, were -0.43, -0.24 and -0.77 respectively. The intra-annual negative correlations in 1984–1993 and 1994–2003 showed a decreasing trend, but showed an increasing trend in 2004–2014. The response lags of the VCI to hydrological drought in Hongshuihe, Yujiang, Liujiang River Basins were mostly 0-4 months, 0-1 months and 2-3 months (followed by 7-8 months), respectively. This research has deepened our understanding of the causes and effects of regional hydrological drought in the Xijiang Basin. However, there are still some issues worthy of in-depth discussion, for instance, the development of a method to tailor the hydropower reservoir scheduling scheme to meet the dual requirements of power generation and drought resistance when severe drought occurs. Another worthy endeavor would be involving the development of a method for determining the response characteristics of vegetation cover to hydrological drought under the changing background of land use and land cover. To clarify the above problems, further research is necessary.

Acknowledgment This study was supported by the National Natural Science Foundation of China (grant No. 51579065), the Special Public Sector Research Program of Ministry of Water Resources (Grants No. 201301040 and 201401008), and the Program for New Century Excellent Talents in University (Grant no. NCET-12-0842). The authors gratefully acknowledge Texas A&M University for providing laboratory and electronics resources for use in this study.

References Barriopedro, D., Gouveia, C. M., Trigo, R. M., Wang, L., 2012. The 2009/10 drought in China: possible

26

causes and impacts on vegetation. J. Hydrometeorol., 13(4), 1251–1267. https://doi.org/10.1175/jhm-d-11-074.1 Belal, A. A., El-Ramady, H. R., Mohamed, E. S., Saleh, A. M., 2014. Drought risk assessment using remote sensing and GIS techniques. Arab. J. of Geosci., 7(1), 35-53. https://doi.org/10.1007/s12517-012-0707-2 Byzedi, M., Khazaei, M. R., 2016. The evaluation of climate change effects on hydrological drought in the Sirwan basin. J. Curr. Res. Sci., S (2), 60–67. Carlton, J. S., Mase, A. S., Knutson, C. L., Lemos, M. C., Haigh, T., Todey, D. P., Prokopy, L. S., 2016. The effects of extreme drought on climate change beliefs, risk perceptions, and adaptation attitudes. Clim. Change,135(2), 211–226. https://doi.org/10.1007/s10584-015-1561-5 Chen, H., Sun, J., 2017. Anthropogenic warming has caused hot droughts more frequently in China. J. Hydrol., 544, 306-318. http://dx.doi.org/10.1016/j.jhydrol.2016.11.044 Chen, Y. D., Zhang, Q., Xiao, M., Singh, V. P., 2015. Transition probability behaviors of drought events in the Pearl River basin, China. Stoch. Environ. Res. Risk Assess., 1–12. https://doi.org/10.1007/s00477-015-1178-2 Chen, Z., Liu, Z., Huang, Q., 2013. Probability of hydrologic drought duration and intensity of Xijiang River. J. Lake Sci., 4, 017. https://doi.org/10.18307/2013.0416 Cheng, L., Hoerling, M., AghaKouchak, A., Livneh, B., Quan, X. W., Eischeid, J., 2016. How Has Human-Induced Climate Change Affected California Drought Risk? J. Clim., 29(1), 111–120. https://doi.org/10.1175/jcli-d-15-0260.1 Davudirad, A. A., Sadeghi, S. H., Sadoddin, A., 2016. The Impact of Development Plans on Hydrological Changes in the Shazand Watershed, Iran. Land. Degrad. Dev., 27(4), 1236–1244. 27

https://doi.org/10.1002/ldr.2523 Duan, W., Wei, C., 2013. The ‘spring predictability barrier’ for ENSO predictions and its possible mechanism: results from a fully coupled model. Int. J. Climatol., 33(5), 1280–1292. https://doi.org/10.1002/joc.3513 Fischer, T., Gemmer, M., Su, B., Scholten, T., 2013. Hydrological long-term dry and wet periods in the Xijiang River basin, South China. Hydrol. Earth. Syst. Sc., 17(1), 135–148. https://doi.org/10.5194/hess-17-135-2013 Gan, T. Y., Ito, M., Hülsmann, S., Qin, X., Lu, X. X., Liong, S. Y., Rutschman, P., Disse M., Koivusalo, H., 2016. Possible climate change/variability and human impacts, vulnerability of drought-prone regions, water resources and capacity building for Africa. Hydrol. Sci. J., 1–18. https://doi.org/10.1080/02626667.2015.1057143 Gouveia, C. M., Trigo, R. M., Beguería, S., Vicente-Serrano, S. M., 2016. Drought impacts on vegetation activity in the Mediterranean region: An assessment using remote sensing data and multi-scale drought indicators. Glob. Planet. Change. https://doi.org/10.1016/j.gloplacha.2016.06.011 Hao, Z., Singh, V. P., 2015. Drought characterization from a multivariate perspective: A review. J. Hydrol., 527, 668–678. https://doi.org/10.1016/j.jhydrol.2015.05.031 Hao, Z., Hao, F., Singh, V. P., Sun, A. Y., Xia, Y., 2016. Probabilistic prediction of hydrologic drought using a conditional probability approach based on the meta-Gaussian model. J. Hydrol., 542, 772–780. https://doi.org/10.1016/j.jhydrol.2016.09.048 Harding, R., Reynard, N., Kay, A., 2014. Current understanding of climate change impacts on extreme events. Hydrometeorological Hazards: Interfacing Science and Policy, 27–47. 28

https://doi.org/10.1002/9781118629567.ch1b Hisdal, H., Tallaksen, L. M., Clausen, B., Peters, E., Gustard, A., 2004. Hydrological drought characteristics. Hydrological drought–Processes and estimation methods for streamflow and groundwater, 48, 139–198. Hong, X., Guo, S., Zhou, Y., Xiong, L., 2015. Uncertainties in assessing hydrological drought using streamflow drought index for the upper Yangtze River basin. Stoch. Env. Res. Risk. A., 29(4), 1235–1247. https://doi.org/10.1007/s00477-014-0949-5 Hosseinzadeh Talaee, P., Tabari, H., Sobhan Ardakani, S., 2014. Hydrological drought in the west of Iran and possible association with large‐scale atmospheric circulation patterns. Hydrol. Processes, 28(3), 764–773. https://doi.org/10.1002/hyp.9586 Huang, Q., Chen, Z.S., Chen, Q.X., Wu, L., 2014. A multi-scale perspective on drought monitoring and assessment for the Pearl River basin. Meteor. Environ. Res., 5(10), 25. Huang X.S, Qin Z.N., Tang, B. L., 2004. The impact of drought from summer of 2003 to winter of 2004 in Guangxi and the preparedness policies. J. Catastrophology, 19(S0): 29–33. https://doi.org/10.3969/j.issn.1000-811X.2004.z1.006 Huang, Y., Xu, C., Yang, H., Wang, J., Jiang, D., Zhao, C., 2015. Temporal and Spatial Variability of Droughts in Southwest China from 1961 to 2012. Sustainability, 7(10), 13,597–13,609. https://doi.org/10.3390/su71013597 Kumar, S., Purohit, H., Mathur, M., Bastin, G., Friedel, M., 2015. Assessment of Impact of Drought on Desert Vegetation through Remote Sensing. Ann. Arid Zone., 54(3&4), 63–70. Leng, G., Tang, Q., Rayburg, S., 2015. Climate change impacts on meteorological, agricultural and hydrological droughts in China. Global. Planet. Change, 126, 23–34. 29

https://doi.org/10.1016/j.gloplacha.2015.01.003 Li, Y., Xu, M., Wei, Y., Huang, W., 2015. An improvement EMD method based on the optimized rational Hermite interpolation approach and its application to gear fault diagnosis. Measurement, 63, 330–345. https://doi.org/10.1016/j.measurement.2014.12.021 Liu, Y., Ren, L., Zhu, Y., Yang, X., Yuan, F., Jiang, S., Ma, M., 2016. Evolution of Hydrological Drought in Human Disturbed Areas: A Case Study in the Laohahe Catchment, Northern China. Adv. Meteorol., 2016. https://doi.org/10.1155/2016/5102568 Lu, G., Liu, J., Wu, Z., He, H., Xu, H., Lin, Q., 2015. Development of a large-scale routing model with scale independent by considering the damping effect of sub-basins. Water Resour. Manage., 29(14), 5237-5253. https://doi.org/10.1007/s11269-015-1115-7 Miao, J., Liu, G., Cao, B., Hao, Y., Chen, J., 2014. Identification of Strong Karst Groundwater Runoff Belt by Cross Wavelet Transform. Water Resour. Manage, 28(10), 2,903–2,916. https://doi.org/10.1007/s11269-014-0645-8 Niu J, Chen J, Sivakumar B., 2013. Hydrological drought characterizations with SPI and SRI in the Pearl River basin in South China. 10th Annual Meeting and Geosciences World Community Exhibition of the Asia Oceania Geosciences Society, AOGS 2013. https://doi.org/10.5353/th_b4558746 Niu, J., Chen, J., Sun, L., 2015. Exploration of drought evolution using numerical simulations over the Xijiang (West River) basin in South China. J. Hydrol., 526, 68–77. https://doi.org/10.1016/j.jhydrol.2014.11.029 Sam, A. S., Abbas, A., Arshad, M., Kächele, H., 2016. Analysing Vulnerability to Climate Change in India with Special Reference to Drought Risk: Results from a Field Survey. In Climate Change Challenge (3C) and Social-Economic-Ecological Interface-Building (pp. 409–427). Springer 30

International Publishing. https://doi.org/10.1007/978-3-319-31014-5_25 Sang, Y. F., Wang, Z., Liu, C., 2014. Comparison of the MK test and EMD method for trend identification in hydrological time series. J. Hydrol., 510, 293–298. https://doi.org/10.1016/j.jhydrol.2013.12.039 Sawada, Y., Koike, T., Jaranilla-Sanchez, P. A., 2014. Modeling hydrologic and ecologic responses using a new eco‐hydrological model for identification of droughts. Water Resour. Res., 50(7), 6,214–6,235. https://doi.org/10.1002/2013wr014847 Tabari, H., Nikbakht, J., Talaee, P. H., 2013. Hydrological drought assessment in Northwestern Iran based on streamflow drought index (SDI). Water Resour. Manage, 27(1), 137–151. https://doi.org/10.1007/s11269-012-0173-3 Trenberth, K. E., Dai, A., van der Schrier, G., Jones, P. D., Barichivich, J., Briffa, K. R., Sheffield, J., 2014. Global warming and changes in drought. Nat. Clim. Change, 4(1), 17–22. https://doi.org/10.1029/2015eo034713 Van Huijgevoort, M. H. J., Van Lanen, H. A. J., Teuling, A. J., Uijlenhoet, R., 2014. Identification of changes in hydrological drought characteristics from a multi-GCM driven ensemble constrained by observed discharge. J. Hydrol., 512, 421–434. https://doi.org/10.1016/j.jhydrol.2014.02.060 Van Loon, A. F., 2015. Hydrological drought explained. Wiley Interdisciplinary Reviews: Water, 2(4), 359–392. https://doi.org/10.1002/wat2.1085 Van Loon, A. F., Gleeson, T., Clark, J., Van Dijk, A. I., Stahl, K., Hannaford, J., Baldassarre, G. D., Teuling, A. J., Tallaksen, L. M., Uijlenhoet, R., Hannah, D. M., Sheffield, J., Svoboda, M., Verbeiren, B., Wagener, T., Rangecroft, S., Wanders, N., Van Lanen, H. A. J., Hannah, D. M., 2016. Drought in the Anthropocene. Nat. Geosci., 9(2), 89–91. https://doi.org/10.1038/ngeo2646 31

Wang, H., Ding Y. H., He, J. H., 2006. Influence of Western North Pacific Summer Monsoon Changes on Typhoon Genesis. Acta Meteorol. Sin., 64(3), 345–355. https://doi.org/doi:10.11676/qxxb2006.033 033 (In Chinese) Wang, J. L., Li, Z. J., 2013. Extreme-Point Symmetric Mode Decomposition Method for Data Analysis. Adv. Adapt. Data Anal., 5. https://doi.org/10.1142/s1793536913500155 Wang, W., Ertsen, M. W., Svoboda, M. D., Hafeez, M., 2016. Propagation of Drought: From Meteorological Drought to Agricultural and Hydrological Drought. Adv. Meteorol., 2016. https://doi.org/10.1155/2016/6547209 Wu, Z., Lin, Q., Lu, G., He, H., Qu, J. J., 2015. Analysis of hydrological drought frequency for the Xijiang River Basin in South China using observed streamflow data. Nat. Hazards, 77(3), 1,655–1,677. https://doi.org/10.1007/s11069-015-1668-z Yu, M., Li, Q., Hayes, M. J., Svoboda, M. D., Heim, R. R., 2014. Are droughts becoming more frequent or severe in China based on the standardized precipitation evapotranspiration index: 1951–2010?. Int. J. Climatol., 34(3), 545-558. https://doi.org/10.1002/joc.3701 Yu, L., Mu, M., Yu, Y., 2014. Role of parameter errors in the spring predictability barrier for ENSO events in the Zebiak-Cane model. Adv. Atmos. Sci., 31(3), 647–656. https://doi.org/10.1007/s00376-013-3058-3 Zhang, Q., Xiao, M., Singh, V. P., Chen, X., 2013. Copula-based risk evaluation of droughts across the Pearl River basin, China. Theor. Appl. Climatol., 111(1-2), 119–131. https://doi.org/10.1007/s00704-012-0656-4 Zhang, Q., Xiao, M., Singh, V. P., Li, J., 2012. Regionalization and spatial changing properties of droughts across the Pearl River basin, China. J. Hydrol., 472, 355–366. 32

https://doi.org/10.1016/j.jhydrol.2012.09.054 Zhang, W., Jin, F. F., Turner, A., 2014. Increasing autumn drought over southern China associated with ENSO regime shift. Geophysical Research Letters, 41(11), 4020-4026. https://doi.org/10.1002/2014GL060130 Zhang, Z., Chao, B. F., Chen, J., Wilson, C. R., 2015. Terrestrial water storage anomalies of Yangtze River Basin droughts observed by GRACE and connections with ENSO. Global. Planet. Change, 126, 35–45. https://doi.org/10.1016/j.gloplacha.2015.01.002 Zhu, Y. J., Guo, C. Q., Huang, X. K., 2012. Trend analysis of precipitation extreme values of Xijiang River Basin. J. China Hydrol., 32(2), 72–77. https://doi.org/10.3969/j.issn.1000-0852.2012.02

Figure captions Fig. 1 Distribution of the main rivers, typical hydrological stations, and hydropower reservoirs in the Xijiang Basin; R., S., and H. are abbreviations for river, hydrological station, and hydropower reservoir project, respectively. Fig. 2 Inter-annual variations and trends of hydrological drought severity from 1961 to 2014 in typical sub-basins of the Xijiang Basin Fig. 3 The ESMD decomposition results of the standardized drought index (SDI) in the Hongshuihe River from 1961 to 2014 Fig. 4 Crossing wavelet analysis of large-scale climate factors and hydrological drought severities at Gaoyao station from January, 1961 to December, 2014. (a)AMO, (b)AO, (c)ENSO, (d)PDO Fig. 5 The variation and tendency of hydrological drought severity and ENSO anomaly percentage from January 1, 1961 to December 31, 2014 in the Xijiang Basin 33

Fig. 6 Inter/intra annual variation of hydrological drought under natural condition and reservoir operation from 2002 to 2014 in the Yujiang River Fig. 7 The reservoir operation schemes and hydrological drought severities under natural condition and reservoir operation in two typical drought periods of the Yujiang River Basin. (a) 2006 –2007 drought, (b) 2009–2010 drought Fig. 8 Inter/intra annual variation of hydrological drought under natural condition and reservoir operation from 2004 to 2014 in the Hongshuihe River Fig. 9 Hydrological drought severities under the natural condition and reservoir operation (left) and the reservoir operation schemes during 2009–2010 drought (right) in the Hongshuihe River Basin Fig. 10 Correlation of drought severity and VCI in four sub-basins in different time periods. (Upper: 1984–1993, Middle: 1994–2003, Lower: 2004–2014) Fig. 11 Variation of SDI and shifted VCI anomaly percentage in 2004–2005 drought from Autumn to Spring in the Xijiang Basin. There are left shifts for (a) 2 months in Hongshuihe, (b) 3 months in Yujiang, (c) 5 months in Liujiang, and (d) 5 months in Guijiang Fig. 12 Spatial response of VCI to hydrological drought in 2004–2005 from autumn to spring in the Xijiang Basin. (a) August 2004, (b) January 2005, (c) March 2005, and (d) May 2005

Table captions Table 1 Hydrological drought events identified based on SDI and the run theory in the Xijiang Basin since the 2000s Table 2 The response difference of hydrological drought to seasonal ENSO variation in 4 tributaries in the Xijiang Basin 34

Fig. 1 Distribution of the main rivers, typical hydrological stations, and hydropower reservoirs in the Xijiang Basin; R., S., and H. are abbreviations for river, hydrological station, and hydropower reservoir project, respectively.

35

Fig. 2 Inter-annual variations and trends of hydrological drought severity form 1961 to 2014 in typical sub-basins of the Xijiang basin

36

Fig. 3 The ESMD decomposition results of the standardized drought index (SDI) in the Hongshuihe River from 1961 to 2014

37

Fig. 4 Crossing wavelet analysis of large-scale climate factors and hydrological drought severities at Gaoyao station from January, 1961 to December, 2014. (a)AMO, (b)AO, (c)ENSO, (d)PDO

Fig. 5 The variation and tendency of hydrological drought severity and ENSO anomaly percentage from January 1, 1961 to December 31, 2014 in the Xijiang basin 38

Fig. 6 Inter/intra annual variation of hydrological drought under the natural condition and reservoir operation from 2002 to 2014 in the Yujiang River

Fig. 7 The reservoir operation schemes and hydrological drought severities under natural condition and reservoir operation in two typical drought periods of the Yujiang River basin. (a) 2006–2007 drought, (b)2009 –2010 drought 39

Fig. 8 Inter/intra annual variation of hydrological drought under natural condition and reservoir operation from 2004 to 2014 in Hongshuihe River

Fig. 9 Hydrological drought severities under the natural condition and reservoir operation (left) and the reservoir operation schemes during 2009–2010 drought (right) in the Hongshuihe River Basin 40

Fig. 10 Correlation of drought severity and VCI in four sub-basins in different time periods. (Upper: 1984–1993, Middle: 1994–2003, Lower: 2004–2014)

41

Fig. 11 Variation of SDI and shifted VCI anomaly percentage in 2004–2005 drought from Autumn to Spring in the Xijiang basin. There are left shifts for (a) 2 months in Hongshuihe, (b) 3 months in Yujiang, (c) 5 months in Liujiang, and (d) 5 months in Guijiang

42

Fig. 12 Spatial response of VCI to hydrological drought in 2004–2005 from autumn to spring in the Xijiang Basin. (a) August 2004, (b) January 2005, (c) March 2005 and (d) May 2005

43

Table 1 Hydrological drought events identified based on SDI and the run theory in the Xijiang Basin since the 2000s Peak value of Month of Location of Duration Rank Period drought severity (month) peak value peak value (SDI) 1 2003.07-2004.06 12 -1.96 6 Yujiang 2 2004.08-2005.05 10 -1.76 10 Yujiang 3 2005.07-2006.06 12 -2.13 2 Hongshuihe 4 2007.10-2008.05 8 -1.56 4 Guijiang 5 2009.08-2010.05 10 -2.82 10 Hongshuihe 6 2011.06-2012.02 9 -3.37 8 Hongshuihe 7 2013.06-2013.11 6 -2.99 7 Hongshuihe

44

Table 2 The response difference of hydrological drought to seasonal ENSO variation in 4 tributaries in the Xijiang Basin Qianjiang Guigang Liuzhou Pingle Lag time (month) spr. sum. aut. win. spr. sum. aut. win. spr. sum. aut. win. spr. sum. aut. win. 0 0.25 0.24 0.37 0.15 0.09 0.24 0.50 0.33 0.38 0.26 0.40 0.24 0.27 0.41 0.32 0.48 1 -0.11 0.05 0.39 0.31 -0.39 0.30 0.61 0.46 0.29 0.00 0.45 0.43 0.15 0.20 0.44 0.57 2 0.10 0.22 0.21 0.41 0.02 0.20 0.49 0.51 0.38 0.17 0.38 0.49 0.53 0.25 0.41 0.58 3 0.10 0.20 0.00 0.52 0.02 0.19 0.40 0.38 0.24 0.28 0.20 0.51 0.26 0.37 0.31 0.52 4 -0.08 0.06 0.04 0.42 -0.24 0.07 0.36 0.17 -0.02 0.18 0.16 0.43 0.08 0.26 0.18 0.45 5 -0.30 -0.06 0.16 0.28 -0.38 0.09 0.44 0.21 -0.35 0.20 0.14 0.43 -0.32 0.25 0.18 0.44 6 -0.35 -0.10 0.29 0.20 -0.25 0.26 0.29 0.26 -0.51 0.25 0.27 0.13 -0.43 0.29 0.28 0.21 7 -0.28 0.07 0.34 -0.25 -0.23 0.25 0.40 -0.03 -0.48 0.28 0.36 -0.16 -0.47 0.26 0.33 -0.15 8 -0.42 0.08 0.06 -0.21 -0.35 0.23 0.29 -0.10 -0.62 0.10 0.24 -0.20 -0.66 0.10 0.19 -0.25 9 -0.43 -0.09 0.03 -0.07 -0.16 0.01 0.16 0.08 -0.64 -0.02 0.10 -0.08 -0.58 0.03 -0.10 -0.17 10 -0.27 -0.18 0.00 -0.01 -0.08 -0.01 0.09 0.17 -0.44 -0.04 0.03 -0.08 -0.38 -0.04 -0.06 -0.23 11 -0.33 -0.03 -0.05 0.03 -0.24 0.01 0.06 0.26 -0.43 0.18 0.10 -0.10 -0.36 0.12 -0.02 -0.20 12 -0.32 0.09 -0.10 0.07 -0.17 0.00 0.01 0.29 -0.35 0.30 0.03 -0.10 -0.26 0.07 -0.25 -0.15

45

Highlights 1. Response of hydrological drought to large-scale climate index was revealed. 2. The drought character was greatly influenced by the reservoir operation. 3. The delayed effect of hydrological drought on vegetation cover was investigated.

46