Large-Scale Linkages of Socioeconomic Drought with Climate Variability and Its Evolution Characteristics in Northwest China

Socioeconomic drought is one of the most frequent natural disasters in the world and is closely related to human life. /e main cause of socioeconomic drought is the contradiction between water supply and demand; hence, as local reservoirs play a major role in improving water supply and coping with extreme climate, it is reasonable to estimate socioeconomic drought based on reservoir operations. /e multivariate standardized reliability and resilience index (MSRRI) is utilized to evaluate socioeconomic drought, considering the characteristics of reservoir management and storage water resources. /erefore, with the MSRRI, this study takes the Heihe River Basin in northwestern China, which is controlled by two reservoirs dominating the upstream and downstream regions, as a case study to reveal the evolution characteristics of socioeconomic drought in the basin and the external impacts of climate variability./e results showed that (1) the drought intensity in the up-midstream region is stronger than that in the downstream region; in view of the hysteresis in the downstream region, the occurrence of drought in the up-midstream region could be regarded as an early warning to implement preventive measures in the downstream region; (2) an increasing trend in socioeconomic drought throughout the basin exists on both monthly and annual scales, which indicates that the increasing possibility of drought should be effectively addressed; (3) cross wavelet analysis indicated that the large-scale climate indices contribute to the variations in the socioeconomic droughts throughout the basin, indicating that climate variability may provide a reference for managers to deal with socioeconomic drought in the HRB.


Introduction
As one of the natural disasters most closely related to human life, droughts occur in almost all areas of the world and cause an average of $6∼8 billion global damage annually [1,2]. Drought places large demands on water resources in urban and rural areas and an immense burden on social development. erefore, the identification of drought events and the timely determination of drought intensity could provide assistance to procedures for reducing the impacts of drought [3]. e world-recognized classification of drought has four categories: meteorological, hydrological, agricultural, and socioeconomic [4,5]. Among these categories, the first three types are regarded to be physical phenomena. Nevertheless, drought caused by social and economic development always refers to the insufficient supply of local water [3]. Due to the increasing water demands for economic and social development, when socioeconomic drought occurs, water shortages usually affect production and consumption activities, resulting in economic, social, and environmental losses [6,7]. It can be concluded that socioeconomic drought is most closely related to human life, but it has received the least attention [8][9][10]. With the expansion of industry and urbanization, in recent years, the rapid increase in water consumption has been difficult to fully satisfy with the limited water resources; thus, socioeconomic drought has gradually proven to be a serious issue requiring attention in many areas of the world [11].
Generally, the water supply of an area usually entails withdrawals from rivers. From one point of view, it can be understood that the correlation between the water intake and water demand determines the possibility of socioeconomic drought [3]. Notably, reservoirs are important projects to manage the supply and distribution of regional water resources. e storage and discharge of a reservoir may dominate the distribution of downstream water resources, signifying that a reservoir is a system with resilience that has the ability to cope with extreme events such as floods and droughts. erefore, the reservoir system is considered a possibility for evaluating socioeconomic drought [12].
For measuring socioeconomic drought, various kinds of indices have been developed, such as reliability, resilience, vulnerability, and integrated indices [13][14][15][16][17]; however, few of these indices take into account local reservoir resilience based on reservoir storage and demand [8,10]. In this study, inspired by Mehran et al. [12], a nonparametric multivariate statistical framework to assess socioeconomic droughts was introduced for characterizing socioeconomic drought. is methodology addresses the problem that socioeconomic drought is difficult to quantify [10]. e approach is integrated by two indices, an inflow versus water demand reliability index (IDR) and a water storage resilience indicator (WSR), where the two indices are also the main characteristics of the reservoir system to cope with climate change.
en, the two indices are integrated into the proposed framework to form a new comprehensive index (i.e., multivariate standardized reliability and resilience index, MSRRI) for characterizing socioeconomic droughts.
is method provides an innovative way of evaluating socioeconomic droughts, and many scholars have utilized it to conduct related research and even made improvements to it. In Mehran et al.'s [12] paper, the indicator was used to assess socioeconomic drought during the Australian Millennium drought (1998-2010) and the 2011-2014 California drought. e results show that MSRRI is superior to univariate indices because it captures both early onset and persistence of water stress over time. Shi et al. [9] set two boundary conditions with water shortage and drought duration and proposed the socioeconomic drought index (SEDI) based on the original MSRRI. Considering the different modes of reservoir operation and management, Guo et al. [10] improved the MSRRI under the effect of reservoir operation rules. Zhao et al. [18] applied this index to Datong River Basin and compared it with historical drought events. ese studies indicate that the MSRRI has a good performance in the evaluation of socioeconomic drought. However, most of these studies focused on a single reservoir, and few of them involved two or more reservoirs in watersheds despite the strong correlations. In this study, it is considered that there are spatial differences in drought conditions in the same watershed. According to the different control basin of the reservoir, each reservoir could only represent the drought situation in its own control area.
China possesses abundant rivers, and inland river basins occupy one-third of the area [8]. It is worth mentioning that the Heihe River Basin (HRB), which is the second largest inland river basin in China, has suffered from frequent water stresses and widespread desertification due to the joint impacts of climatic variations and human activities [19][20][21]. Two reservoirs (Huangzangsi Reservoir and Zhengyixia Reservoir) and seven hydropower stations from cascade hydropower station systems in the HRB oversee a large amount of the water and energy resources. erefore, the HRB has long been a critical area in the arid area due to its significant role in Northwest China [8,22,23]. e whole basin can be divided into two regions by the two reservoirs, the up-midstream and the downstream areas. Most of the population and industries are concentrated in the up-midstream, and the downstream areas are mainly grassland and woodland. Because the water inflow downstream mainly comes from the upper and middle regions, the occurrence of drought in the lower reaches is closely related to the up-midstream. Previous studies also focused on the drought situations in the HRB, but there little attention has been paid to the correlations between the two regions.
Globally, climate variability has proven to be one of the major elements with intense teleconnections with droughts [24][25][26]. Advanced understanding of the spatial and temporal conjunctions between the large-scale climate indices and the variations of drought indices can enhance the science of management in water resources systems [27].
e El Niño Southern Oscillation (ENSO), which is a periodical ocean and atmospheric phenomenon, has been shown to have a strong influence on the global climate [28][29][30][31][32]. Many studies have verified the notable linkage between drought/flood events and extreme ENSO events existing in many parts of the world [32,33]. In addition to ENSO, various of other studies applied the correlation analysis to evaluate the large-scale climate indices phenomena on different drought indices, such as the Pacific/ North American (PNA), the North Atlantic Oscillation (NAO), Arctic Oscillation (AO), Pacific Decadal Oscillation (PDO), and East Asian summer monsoon (EASM) et al. [27,[34][35][36][37][38][39]. Besides, the utilization of spectral-based methods such as cross wavelet transform technique could improve the latent associations between a pair of time series [40]. e cross wavelet transform technique breaks down data into time and frequency space and detects the dominant variability and associated variation, which is robust in nonstationary signals [27,41]. Considering the studies related to teleconnection indices with droughts in China, accordingly, the impacts of four large-scale climate signals (ENSO/PNA/NAO/EASM) on socioeconomic drought are examined in the present study with cross wavelet transform method.
Hence, this study takes the Heihe River Basin (HRB) in northwestern China, which is controlled by two large reservoirs, as a case study for research. e purposes of this study are as follows: (1) utilizing the MSRRI multivariate drought index to characterize the socioeconomic drought in two main reservoirs in the up-midstream and downstream regions of the HRB, respectively, and explore the relationship and differentiation of drought events between the two reaches; (2) analyzing the characteristics of the variations in socioeconomic droughts in the HRB in terms of the trends and periodicity; and (3) revealing the teleconnections between anomalous atmospheric circulation patterns (ENSO/PNA/NAO/EASM) and socioeconomic drought in the HRB.

Study Area and Data
Located in the central part of Eurasia, the Heihe River Basin (HRB) is the second largest inland river watershed in northwestern China. e scope of the HRB ranges from 96°E∼102°E to 37.5°N∼42.4°N, with a total length of 928 km and a watershed area of 128,700 km 2 . Affected by human activities and climatic variations, the distribution of water resources in the HRB is extremely uneven, and population and human activities are mainly concentrated in the upmidstream area. With typical drought characteristics, the entire region of the HRB has an average annual precipitation of approximately 400 mm and an average annual potential evaporation of approximately 1,600 mm [42]. Drought has become the main cause of social development restriction and ecological environment deterioration in this area.
To solve the long-standing drought and ecological issues in the basin, two reservoirs, the Huangzangsi Reservoir and Zhengyixia Reservoir, were constructed in the HRB (Figure 1). Located in the upstream area, the Huangzangsi Reservoir is a within-year reservoir that dominates with4.06 × 10 8 m 3 total storage and 3.34 × 10 8 m 3 regulating storage, as well as 6,020 kW guaranteed output and 49,000 kW installed capacity.
e Zhengyixia Reservoir is the main reservoir downstream, with a total reservoir capacity of 4.6 × 10 8 m 3 , which has the main task of allocating the water demands for the ecosystems downstream. As these reservoirs regulate extensive water and energy resources, they will play a leading role in water resource scheduling and management in the HRB when the reservoirs are completed in 2020. Since the reservoirs have not yet been completed, only measured data of the inflow process can be accessed, and simulations of the outflow processes from July 1958 to June 2014 [43] are adopted in this study. e simulated outflow process results are verified to meet the design requirements of the reservoir. e details could be referred to [43]. Furthermore, the time series of the monthly Niño 3.4 index covering 1958∼2014 was adopted to characterize ENSO events in this study, and the data were obtained from the NOAA Earth System Research Laboratory (http://www.esrl.noaa.gov/psd/data/correlation/ nina34.data). e monthly PNA/NAO/EASM time series were available on the NCEP Climate Prediction Centre web page (http:// www.ncdc.noaa.gov/teleconnections/ao.php). e historical data of drought records could be accessed by the China Meteorological Data Service Centre (http://data.cma.cn/).

e Multivariate Standardized Reliability and Resilience Index of Socioeconomic Drought.
e multivariate drought index of MSRRI, as previously mentioned, consists of two individual indices, the inflow-demand reliability (IDR) and the water storage resilience (WSR) [15]. Usually, according to the regulation performance of reservoirs, reservoirs can be divided into two types: within-year and over-year reservoirs [44]. Within-year reservoirs are more sensitive to seasonality, while over-year reservoirs are more sensitive to longterm water deficiency, i.e., drought. is classification indicates that the time period has an effect on reservoir operation. With this idea, a framework was developed with the definitions for the reservoir systems being 6 months or 12 months for within-year or over-year reservoirs, respectively. en, two indices are defined under the frame: the IDR index and WSR index. On behalf of the available inflow, the IDR indicates if there are sufficient available water resources to satisfy the water demands for human life, regardless of the reservoir storage. Consequently, the IDR is derived by calculating the sum of the percent variation in inflow with respect to the water requirements in the predetermined time frame:

Advances in Meteorology
where Q in i and Q est t represent the monthly inflow of the reservoir and the total estimated water demand, respectively, and i � 1, . . . , N; t � 13, . . . , N.
en, the WSR is calculated based on inflow, storage, and water demand at monthly timescales. Correspondingly, it is determined by whether the reservoir storages could fulfil the water demand within the time frame (m): where St t is the reservoir storage at month t and Q in i and Q out t represent the monthly inflow of the monthly water demand. en, Q min denotes the smallest reservoir storage.
For the convenience of comparison, the standardization of the two indices with the standard normal distribution is carried out [12]. e marginal probabilities p(x t ) are calculated by where i represents the nonzero index ranking from small to large and n is the sample size. en, p(x t ) is converted to a standardization index as follows. e same method is used to obtain the IDR index SI(α t ) and WSR index SI(β t ).
Hence, the integration index is generated within a multivariate framework based on the Gringorten plotting position [45][46][47]: where P jt is the joint empirical probability at month t, I represents the number of occurrences of the pair (SI(α t ), SI(β t )) for SI(α) ≤ SI(α t ) and SI(β) ≤ SI(β t ), and N is the size of the sample. e MSRRI is formulated by the standardized joint distribution function of the IDR index SI(α t ) and WSR index SI(β t ) [46]: e MSRRI is an integration index relying on fundamental indices that could be utilized to estimate socioeconomic drought by measuring the supply and water storage amount related to demand. Similar to the other drought indices, this index is based on the positive and negative values to determine whether drought occurs. e occurrence of drought events and the severity of drought need to be judged by establishing criteria. In addition to the positive and negative index values, however, because of the resiliency of the reservoir's water storage capacity, some negative values have little effect on the running process of the system. In this study, − 0.8 indicates that the cumulative probability of the joint distribution of the IDR and WSR series is 0.2 and was adopted as a threshold to be the condition for drought occurrence in multivariate drought analysis [10,46,48,49].
Simultaneously, the duration of drought is also an important feature for judging drought. According to the duration of drought, four grades are also established (Table 1).
Additionally, drought intensity is also an important drought characteristic. e definition of drought intensity (DI) is based on the percent of the total drought index value ( VI) to the months of drought duration (M d ) [8,10], as shown by the following formula: Consequently, for each detected socioeconomic drought event, the drought level (SDL) is defined with the indicator value V g and duration grades D g :

e Modified Mann-Kendall (MMK) Trend Test.
To analyze the features of tendency, the modified Mann-Kendall (MMK) test is adopted to examine the trends of the socioeconomic drought index series. Commonly, the Mann-Kendall (MK) method is the most widely used nonparametric method for time series trend analysis [50]. However, MK results are easily influenced by the consistency of the time series. Afterwards, the MMK trend test method, which is more robust for detecting the tendency of hydrometeorological series, was proposed [51]. Hence, this study adopted the MMK trend test method to detect the tendency of the socioeconomic drought index series.

Moving-Window Correlation Analysis (MWCA) for
Periodic Component Recognition. Periodic component analysis is of great significance for understanding various hydrologic processes and predicting the future hydrological regime of a watershed or region. e moving-window correlation analysis (MWCA) proposed by Xie et al. [52] adopts a new way to test the significance of period for hydrologic series. MWCA constructs the periodic processes through correlation analysis of periodic processes and original series. To analyze the time-frequency characteristics of hydrologic series, the time-frequency centre (TFC) is also proposed to investigate the local time and frequency domain of the time series. It has been proven that MWCA exhibits good performance in identifying true periods, extracting reliable periodic components, and detecting the active time ranges of various periodic components. erefore, MWCA is chosen to conduct the periodic analysis of socioeconomic drought series in HRB.
e specific procedures could be referred to Xie et al. [52] for details.

3.4.
e Cross Wavelet Analysis for Impacts of Climate Variations. Wavelet analysis has been widely used in hydrology, meteorology, and other disciplines in recent years because of its good time-frequency localization characteristics and multiresolution analysis performance. However, wavelet analysis can only explore the time-frequency characteristics of a single time series, and it is difficult to analyze the interaction and time-frequency correlation between multiple time elements. Cross wavelet transform is a new multisignal multiscale analysis technique developed on the basis of traditional wavelet analysis. is technique can not only effectively analyze the correlation degree between two associated time series but also reflect the phase structure and detailed characteristics of both time and frequency domains [53,54].
Assuming that the background power spectra of two time series X and Y are Fourier red noise spectra P X k and P Y k , the theoretical power spectrum distribution of the cross wavelet can be expressed as [54] where σ X and σ Y are the standard deviations of time series X and Y, respectively; Z v (p) is the confidence level associated with the probability p; and v is degree of freedom. For two hydrological time series X and Y, the upper 95% confidence limit of the power spectrum of red noise is obtained first. When equation (9) exceeds the confidence limit, it is considered that the results have passed the test of the standard spectrum of red noise under the condition of significance level α � 0.05, and a significant correlation exists. e relevant codes can be freely downloaded from the website http://www.pol.ac.uk/home/research/waveletcoherence/.
Drought events are irregular in time and space in terms of distribution and severity. Some meteorological factors, such as ENSO, PNA, NAO, and EASM, are associated with drought variability. Consequently, this study explored the relevance between the MSRRI and ENSO/PNA/NAO/ EASM to reveal the influences of climate indices on the socioeconomic drought in the HRB basin, which is expected to be helpful for the mitigation of local natural hazards.

e Evolution Characteristics of Socioeconomic Droughts in the HRB.
e cascade reservoir hydropower station system in the HRB consists of two reservoirs (Huangzangsi and Zhengyixia Reservoirs) and seven hydropower stations. erefore, the characteristics of the indicators of the two reservoirs could indicate the socioeconomic drought state of the up-middle and downstream regions of the HRB. e two reservoir systems both belong to the within-year system because they take one year to fill. en, the time frame for them is set as 6 months. e IDR and WSR values represent the hydrological characteristics and reservoir conditions, respectively, both relative to the demand [12]. For specifics, IDR<0 signifies a low-inflow occurrence (i.e., hydrological drought) relative to water demand, whilst WSR>0 represents the reservoir storage is sufficient for demand. erefore, there is a phenomenon that hydrological indicators symbolize the occurrence of a drought, but the demand is satisfied by available storage, indicating that a hydrological drought may not cause a socioeconomic drought [18]. In contrary, when IDR>0 and WSR<0, it corresponds to a situation in which there is no hydrological drought based on input to reservoirs, while the system is still suffering from a socioeconomic drought as the available storage cannot meet the demand.
As integration of IDR and WSR, the MSRRI implies the synthetic information of the overall system. From Figure 2, the three lines of IDR, WSR, and MSRRI series are roughly consistent.
e Pearson correlation coefficients of the monthly MSRRI series in 1958∼2014 with the corresponding IDR and WSR series are 0.73 and 0.79 for Huangzangsi reservoir, respectively, as well as 0.82 and 0.88 for Zhengyixia reservoir, which indicates the reliability and effectiveness of the MSRRI in characterizing socioeconomic droughts. As integration of IDR and WSR, the smaller the MSRRI, the more severe the drought and the more serious the water shortage. e performance of the MSRRI (including IDR and WSR) applied to reservoirs is shown in Figures 2 and 3.
Overall, the MSRRI of the Huangzangsi Reservoir is lower than that of the Zhengyixia Reservoir; that is to say, the socioeconomic drought in the up-middle reaches is more serious than that in the lower reaches.
ere are four combinations of drought events in the upper and lower reaches: "drought occurs in both reaches," "drought occurs in only the up-midstream," "drought occurs in only the downstream," and "no drought occurs in both reaches." On a monthly scale, the probability of socioeconomic drought events occurring in the upper-middle reaches is the highest, approximately 59.8%. When drought occurs in both reaches, the drought severity in the up-midstream is greater than that in downstream (rate of 65.6%). is result confirms that the drought severity in the up-midstream is more serious than that in the downstream. e time period of drought events (1973∼1983) is selected for detailed analysis. According to the historical drought records of China Meteorological Data Service Centre, the HRB basin experienced serious drought or extreme drought in 1973, 1979∼1980, 1982, and 1984, respectively. It could be seen from Figure 3 that the drought records during this period have been detected, which shows that MSRRI is applicable to the evaluation of socioeconomic droughts in HRB. e occurrence of socioeconomic drought time in the lower reaches often lags behind to a certain extent. In 1973, for example, the occurrence of drought events in the lower reaches was delayed, while the severity was higher than that in the up-middle reaches. Similarly, in approximately 1975Similarly, in approximately , 1978Similarly, in approximately -1980Similarly, in approximately , and 1983, the downstream also experienced a lag. It can be inferred that, since the downstream mainly relies on the inflow of water from the Advances in Meteorology up-midstream, when the up-midstream suffers from socioeconomic drought, the downstream of the HRB may also suffer from the resulting socioeconomic drought. e duration and intensity of drought are usually the main characteristics of drought events. erefore, the intensity and duration of drought are used as the basis for the comprehensive classification of socioeconomic drought in this study. e specific classification criteria are shown in Table 1 above, and the statistical results are shown in Figure 4.
It can be inferred from Figure 4 that there are distinct differences between the socioeconomic drought events in the   1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1973 Year  1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1973 Year is result is mainly reflected in the relatively high intensity of socioeconomic drought in the up-middle reaches. e statistical analysis of intensity and duration showed that more than 80% of the drought events in the up-middle reaches were classified as grades III or IV (49.3% and 31.5%, respectively) and that value in the downstream reaches was 60.7% (Figure 4). e number of drought events lasting for more than 12 months in the two areas was 15 and 11, respectively. Specifically, for both streams, only one drought event lasted longer than 36 months, and two drought events lasted longer than 24 months. From this point of view, the distribution of long-duration drought events in the upmiddle and lower reaches is consistent. In conclusion, consistent with the above monthly MSRRI sequence analysis, the severity of socioeconomic drought in the upmidstream is higher than that in the downstream. It is necessary to be vigilant against socioeconomic drought in both the up-midstream and downstream regions due to the high risk; in particular, in the case of drought in the upmidstream, measures shall be taken in advance in the downstream.

e Trends of Socioeconomic Drought of MSRRI in the HRB.
e tendency analysis in this study was carried out by the MMK trend test method of the annual and monthly MSRRI in 1958∼2014 in the HRB. e results are shown in Figure 5 and Table 2.
Obviously, the MSRRI series of the two reservoirs show a growing trend on both annual and monthly scales. Among them, only the annual MSRRI of the Huangzangsi Reservoir shows a significant increasing tendency. Consistently, according to the annual trend line drawn, both reservoirs also exhibit an obvious growing trend. Consequently, the up-midstream and downstream regions in the HRB are consistent in the occurrence trend of socioeconomic drought. Moreover, this is a warning that the possibility of drought in the whole basin has been increasing.

e Periodic Component Analysis of Socioeconomic Drought in the HRB.
is study utilized the MWCA method to carry out the periodicity analysis of the annul MSRRI series in the HRB at a 5% confidence level. e advantage of this method is that not only the true periodic components according to the period spectrum of time series could be found but also the estimated active time span of significant periods could be revealed with the TFC point distribution [52]. e MWCA method is used for multiple recognition rounds of the periodic components of the annual MSRRI in the HRB. e first round only identifies the most significant period of the original sequence, and then the most significant period of the remaining components of the last round is identified until no significant period occurs. Figures 6 and 7 display the results of the periodicity analysis of the two series, in which the former is the timedomain coverage graph and the latter is the time-frequency centre distribution graph. In Figure 6, MWCA manifests one apparent period (T � 2 years) for the Huangzangsi Reservoir. e coverage ratio is 0.441. e Zhengyixia Reservoir has two periods (T � 9 and 11 years), with coverage ratios of 0.6176 and 0.4998. In addition, the estimated active time ranges of the significant period for the Huangzangsi Reservoir (T � 3 years) were concentrated in 1980-1996, while those for the Zhengyixia Reservoir (T � 9 and 11 years) were most concentrated in 1971-1988 and 1982-2013. Figure 7 shows the interruption of the 2-year period for the Huangzangsi Reservoir, which indicates that the waveform of the periodic component undergoes some changes over time. Relatively speaking, the periodic component of the upmidstream socioeconomic drought is relatively less significant than that of the downstream drought.       Advances in Meteorology HRB are exhibited in Figures 8-11, respectively. e cross wavelet power spectrum emphasizes the correlations between drought sequence and its subsequent factors. e solid black line in the graph is the influence cone of the wavelet boundary effect, and the thick black line indicates that the domain passes the red noise test with 0.05 significance level. e relative phase relationship is represented as arrows (with anti-phase pointing left, in-phase pointing right).
Apparently, it can be seen from Figures 8-11 that these four climatic indices are more or less significantly related to the annual MSRRI series in the HRB, which indicates that the climate indices have a strong influence on the performance of socioeconomic droughts. ENSO events and the MSRRI show an inverse phase relationship; i.e., they show a strong negative phase correlation in both the up-middle and lower reaches of the region (Figure 8). In particular, the ENSO events demonstrated statistically notable negative correlations with the MSRRI series at a scale of 2∼4 years in 1985∼1999 in the up-midstream. For the downstream region, a significant negative linkage also exists for 2-6 years during 1969∼1978. In particular, the ENSO events have a stronger influence on the downstream than on the upmidstream. Similarly, PNA also demonstrated apparent impacts on the features of socioeconomic droughts in the HRB (Figure 9). Specifically, for the upper and middle streams, PNA shows statistically notable positive linkages   with the annual MSRRI series with a 2∼3-year signal in 1994∼1996 and a 4∼6-year signal in 1988∼1994. However, remarkable negative linkages exist with the annual MSRRI series with 3∼5-year periods in 1967∼1976. e degrees of the impact of the PNA on the socioeconomic drought events in the up-middle and lower reaches are similar. e effect of NAO on the drought of HRB basin is relatively insignificant, and the circumstances of upstream and downstream are basically consistent (Figure 10). In the whole performance period, the positive and negative positions alternate each other. For specifics, there were 2-year periodic signals with in-phase and anti-phase orientations alternately in 1968∼1973 and 1993∼1998, respectively. For ESAM, its impact is relatively significant and negative ( Figure 11). For the up-midstream, there was a signal of about 1-4 years in 1993∼1998 and a signal of about 3 years in 1978∼1983. For the downstream, the signal strength is relatively obvious in 1993∼1998 as the upstream with 1∼2 years. In addition, there is a relatively less significant signal of 6-8 years within 1983∼1993.
Overall, the comprehensive influence of the large-scale climate indices contributes to the variations in the socioeconomic droughts throughout the basin. Both positive correlations and negative correlations exist, and the periodic signals are localized. Relatively, ENSO has prominently broader and stronger impacts on socioeconomic droughts than the other teleconnection indexes. In particular, the influence of ENSO and PNA on the downstream is slightly higher than that in the up-midstream of the HRB, while the reverse is true for the NAO and EASM.

Conclusions
As the key facility for managing regional surface water resources, reservoirs play an important role in dealing with drought and climate change. e multivariate standardized reliability and resilience index (MSRRI) is a comprehensive index for evaluating socioeconomic drought considering the water demand and water storage resilience of reservoirs. In this study, the Heihe River Basin in Northwest China, which is controlled by two large reservoirs (Huangzangsi Reservoir and Zhengyixia Reservoir) is taken as a case study, and the MSRRI is utilized to characterize the evolution features of socioeconomic drought and the linkages of large-scale climate indices. e main results of this study showed the following: (1) It could be inferred that the degree of socioeconomic drought in the up-midstream is higher than that in the downstream, as the rate of the annual MSRRI value in the up-mid stream, lower than that in downstream, is 65.6%. In view of the intensity and duration of drought, the statistical analysis shows that more than 80% of the drought events in the middle and upper streams are severe and extreme droughts, and only 60.7% are in the downstream, which also confirms our result. As the drought hysteresis exists in the downstream region, the occurrence of drought in the up-midstream region could be regarded as an early warning to implement preventive measures in the downstream region. e Zhengyixia Reservoir has two periods (T � 9 and 11 years), with coverage ratios of 0.6176 and 0.4998. (4) Cross wavelet analysis indicated that the large-scale climate indices contribute to the variations in the socioeconomic droughts throughout the basin, indicating that climate variability may provide a reference for managers to deal with socioeconomic drought in the HRB. Relatively, ENSO has prominently broader and stronger impacts on socioeconomic droughts than the other teleconnection indexes.
In conclusion, it is necessary for the administrative department of the Heihe River Basin to pay attention to the socioeconomic drought. is study revealed the evolution characteristics of socioeconomic drought and its relationship with climate change in the HRB, and it is expected to provide help for local socioeconomic drought resistance and water resource management.
Data Availability e data of Nino 3.4 index used to support the findings of this study are available at the NOAA Earth System Research Laboratory (http://www.esrl.noaa.gov/psd/data/correlation/ nina34.data). e data of PNA/NAO/EASM index time series used to support the findings of this study are available at the NCEP Climate Prediction Centre webpage (http://www. cpc.ncep.noaa.gov). e simulation data of outflow and storage capacity processes of reservoirs used to support the findings of this study are confidential and not accessible to the public.

Conflicts of Interest
e authors declare that there are no conflicts of interest.  1958 1963 1968 1973 1978 1983 1988 1993 1998