Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Meteorological Drought, Hydrological Drought, and NDVI in the Heihe River Basin, Northwest China: Evolution and Propagation

Meteorological Drought, Hydrological Drought, and NDVI in the Heihe River Basin, Northwest China:... Hindawi Advances in Meteorology Volume 2020, Article ID 2409068, 26 pages https://doi.org/10.1155/2020/2409068 Research Article Meteorological Drought, Hydrological Drought, and NDVI in the Heihe River Basin, Northwest China: Evolution and Propagation 1 1,2 3 Fanglei Zhong , Qingping Cheng , and Ping Wang Northwest Institute of Eco-Environment and Resources, Chinese Academy of Sciences, Lanzhou 730000, China University of the Chinese Academy of Sciences, Beijing 100039, China School of Tourism and Geographical Sciences, Yunnan Normal University, Kunming 650500, China Correspondence should be addressed to Qingping Cheng; qpchengtyli@foxmail.com Received 24 October 2019; Revised 10 February 2020; Accepted 22 February 2020; Published 9 June 2020 Academic Editor: Budong Qian Copyright © 2020 Fanglei Zhong et al. *is is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Understanding the evolution and propagation of different drought types is crucial to reduce drought hazards in arid and semiarid regions. Here, Standardized Precipitation Index (SPI), Streamflow Drought Index (SDI), and Vegetation Condition Index (VCI) were used to investigate the spatiotemporal variation of different drought types and correlations between Pre (Pre-R)/post (Pos-R)- reservoir. Results showed that the average peak/intensity/duration/severity of meteorological droughts (MD) were greater in the Pre- R than in the Pos-R period in the upstream Heihe River Basin (UHRB), while there was little change between the Pre-R and Pos-R periods in the midstream Heihe River Basin (MHRB). *e average peak/intensity/duration/severity of hydrological drought (HD) decreased in the mainstream for Yingluoxia (Ylx) but increased for Zhengyixia (Zyx) station in the Pos-R period. Propagation time decreased by 3 months (negative effect) in Ylx and increased by 8 months (positive effect) in Zyx compared with the Pre-R period. In the Pos-R period, propagation time increased (1–3 months) for tributaries (positive effect). Propagation times for the mainstream and tributaries varied for different seasons and time periods. Pearson’s correlation coefficient values were lower at short timescales (1–3 months) but higher at long timescales for the Pos-R period in Ylx and Zyx for SDI-1 with different timescales of SPI. *e SDI and SPI had no lag in the UHRB and MHRB. However, VCI with SPI had a significant lag correlation at short timescales in the UHRB (lag 6 months) and MHRB (lag 4 months), and the VCI with SDI had a significant lag correlation for 1 month in the MHRB. *e propagation time from MD to HD has been reduced for Pos-R in the UHRB. *ere was a positive effect (prolonged MD propagation HD time) in Pos-R but still faces serious drought stress in the MHRB. and central portions of northern China since the late 1. Introduction 1990s [11, 12]. Drought is recognized as the world’s most costly and In general, drought can be classified into four types: pressing natural hazard, giving rise to significant losses in meteorological, hydrological, agricultural, and socioeco- the fields of the economy, ecology, and environment (e.g., nomic [13]. Meteorological droughts (MD) are the water crop losses, degradation and desertification, urban water shortages caused by an imbalance in precipitation and supply shortages, and forest fires) [1–4]. *e damage from evaporation; precipitation is commonly used for MD droughts is expected to increase in severity [5]. Compared analysis [14, 15]. Hydrological droughts (HD) are related to a with other natural hazards, the spatial extent of drought is period with inadequate surface and subsurface water re- very large and its time of influence is commonly much sources for established water uses in a given water resource longer. Under global warming, more frequent and severe management system. Streamflow data are widely applied for droughts have been projected in the 21st century [6–8], HD analysis [14, 16]. Because the influence of drought on particularly in the mid-latitudes [9, 10]. Severe and ex- terrestrial ecosystems is becoming increasingly acute, the treme droughts have become more frequent in the eastern vegetation response to drought is a crucial topic in the 2 Advances in Meteorology intensity, severity, and peak, while detecting the spatio- domain of climate research [17–19]. *e Normalized Dif- ference Vegetation Index (NDVI) is a good measure to temporal characteristics of VCI; (2) to investigate VCI and HD propagation times based on MD at different timescales estimate green biomass, leaf area index, and patterns of productivity and has been widely used to assess vegetation and VCI propagation times based on HD at different degradation, ecosystem features, and the physiological timescales; and (3) to explore the potential factors influ- drought conditions of vegetation [20, 21]. encing drought characteristics and propagation times. Drought indices are the best ways to monitor drought and drought events at present. Lloyd-Hughes [22] counted 2. Materials and Methods over 100 drought indicators that have been developed for different types of drought [23]. A commonly used index is 2.1. Study Area. *e Heihe River Basin (HRB) is the second the Standardized Precipitation Index (SPI) [24]. Its calcu- largest inland river basin in the arid zone of northwest lation is simple because it only needs precipitation as an China. *e HRB flows through Qinghai Province, Gansu input, and it can be calculated over different timescales, Province, and the Inner Mongolia Autonomous Region. *e which allows SPI to monitor both short-term droughts, such HRB is located along the land Silk Roads (“One Belt, One ° ° ° ° as agricultural droughts, and long-term droughts, such as Road”) between 97.1 E–102.0 E and 37.7 N–42.7 N with a 5 2 HD [3, 25–27]. For HD index, the Streamflow Drought total area of approximately 1.43 ×10 km [47, 48]. *e HRB Index (SDI) proposed by Nalbantis and Tsakiris [28] includes three sections from south to north: the upstream overcomes the problems of predicting drought onset and Heihe River Basin (UHRB) from the Qilian Mountains to duration using cumulative streamflow volumes and the areal Yingluoxia station (outlet of the mountains) belongs to the extent of drought using a spatially integrated streamflow at a subhumid and semiarid temperate continental monsoon basin outlet; it is also widely used in different regions of the climate zone; the midstream Heihe River Basin (MHRB) world [28–33]. *e Vegetation Condition Index (VCI) is a running from Yingluoxia station to Zhengyixia station be- normalization of the NDVI, which filters out the contri- longs to the arid climate zone; and downstream HRB ter- bution of local geographic resources to the spatial variability minates in the Juyan Lakes (east and west branches, of the NDVI [17, 34]. It reflects agricultural (vegetation) respectively) [49]. From upstream to downstream, the ele- drought accurately in many parts of the world and has been vation decreases, the water availability decreases, and the applied in daily drought monitoring by the National Oceanic landscape changes from glaciers and alpine biomes in the and Atmospheric Administration of the United States UHRB to steppes and agricultural ecosystems in the MHRB (NOAA) and the National Satellite Meteorological Centre of to riparian ecosystems and vast areas of desert in the China [35]. downstream area [47]. In this study, the UHRB and MHRB Many researchers in related fields have focused on were selected as the study area. Figure 1 shows the study area drought evolution, drought prediction, and developing re- and the hydrometeorological stations. Analyses were also liable drought indices using observed and simulated data conducted at the sub-basin level in the UHRB and MHRB. In [4, 36–39]. Recently, studying the propagation time of VCI the MHRB, the annual average precipitation, potential and HD based on MD using Pearson’s or Spearman’s evapotranspiration, and streamflow were 226 mm, 924 mm, correlation coefficients has become a research focus and 88 mm, respectively. Over 90% of the population, grain [8, 10, 18, 19, 21, 23, 40–46]. *ese two kinds of droughts can production, and major industries are concentrated in the reflect the different stages of drought development. Com- MHRB. Approximately 84% of the total available water was monly, MD develops and ends relatively quickly, while HD is consumed for irrigation. On average, a decline in ground- the result of MD [4]. *e corresponding propagation time water of about 1.86 m is attributable to water consumption depends on local landscape conditions [46]. *erefore, in- for irrigation of the farmland area in Heihe River from 1981 vestigating the propagation time from MD to HD [4] and to 2010 [50] with demand constantly increasing [51, 52]. from MD to VCI along with its potential influencing factors *ere are a series of reservoirs (one large reservoir, 9 me- is necessary for establishing an effective monitoring and dium size reservoirs, and 89 small reservoirs) and a relatively warning system for HD and VCI based on MD. complete irrigation system consisting of more than 893 main However, little research has been done into the propa- canals and branch canals [53, 54]. Land use and land cover gation of MD to HD and VCI and HD propagation to VCI, have also significantly changed in the HRB. Urban, culti- in particular in subarid and arid watersheds, which is one of vated, and forest land have increased, while water area and the aims of this paper. *e investigation of VCI and HD grassland have become increasingly degraded. propagation time from MD at different timescales under the influence of global warming and human activities (e.g., construction of reservoirs or dams, land use, and land cover 2.2. Data. Daily precipitation datasets from 1961 to 2012 change) will help to elucidate the impacts of droughts on were collected by five meteorological stations (Table 1) and terrestrial ecosystems and inform planning and optimizing derived from the China Meteorological Data Sharing Service water resource allocations during droughts. It is important System of National Meteorological Information Centre V3.0 to identify the drought propagation process and mecha- (http://www.nmic.gov.cn/). Rigorous quality control was nisms to help establish a drought early warning system. *e conducted by the China National Meteorological Infor- primary objectives of this study are as follows: (1) to inspect mation Centre before the data were released. Monthly ob- the spatiotemporal characteristics of MD and HD duration, served streamflow data, which were also of high quality, for Advances in Meteorology 3 98°E 100°E 102°E 98°E 100°E N N 40°N 40°N 38°N 38°N 38°N Elevation (m) 38°N High : 5297 0 60 120 0 60 120 Kilometers Kilometers Low : 1277 98°E 100°E 102°E 98°E 100°E 102°E Hydrological station LUCC Meteorology station Forest Built-up area River Grassland Unused land Water area Cultivated land (a) (b) 98°E 100°E 38°N 38°N 0 60 120 Kilometers 98°E 100°E Coniferous forest Marshy grassland Alpine vegetation Swamp Cultivated vegetation Glacial snow and salinized land Shrubs Residential land Desert River system Grassland (c) Figure 1: *e upstream and midstream in the Heihe River Basin with meteorological and hydrological stations and digital elevation map (a), land use and land cover (LULC) in 2011 (b), and vegetation types (c). Table 1: Information on the meteorological stations. ° ° Sub-basin Station Longitude ( E) Latitude ( N) Elevation (m) Data period ° ° Qilian (Ql) 100.24 38.19 2787.4 1961–2012 UHRB ° ° Yeniugou (Yng) 99.58 38.41 3286.0 1961–2012 ° ° Zhangye (Zy) 100.46 38.91 1482.7 1961–2012 ° ° MHRB Shandan (Sd) 101.08 38.77 1764.6 1961–2012 ° ° Gaotai (Gt) 99.79 39.36 1332.2 1961–2012 eight hydrological stations (Table 2) were collected from the SPOT/VEGETATION NDVI and MODIS from 1999 to Hydrological Bureau of Gansu Province. Yingluoxia (Ylx) 2012. *e data were obtained from the resource and envi- and Zhengyixia (Zyx) stations were located in the main ronmental science data center of the Chinese Academy of stream of the HRB. *e monthly NDVI (1 km spatial res- Sciences (http://www.resdc.cn), and LULC data in the olution) dataset was the satellite remote sensing data of UHRB (2000 and 2010) and MHRB (1975, 1985, 1995, 2000, 4 Advances in Meteorology Table 2: Information on the hydrological stations. Longitude Latitude Catchment area Elevation Data Abrupt Sub-basin Station River ° ° ( E) ( N) (km ) (m) period year Yingluoxia (Ylx) Heihe River 100.18 38.82 10009 1700 1945–2012 1979 Hongshui Shuangshusi (Sss) 100.83 38.32 578 2490 1957–2012 1974 River UHRB Wafangcheng Dazhuma 100.48 38.43 334 2440 1957–2012 1985 (Wfc) River Liyuanbao (Lyb) Liyuan River 100.00 38.97 2641 1760 1956–2012 1976 Zhengyixia (Zyx) Heihe River 99.47 39.82 35634 1280 1957–2012 1984 Lijiaqiao (Ljq) Maying River 101.13 38.52 1205 2150 1957–2012 1986 MHRB Haichaoba Shunhua (Sh) 100.72 38.59 191.8 1948 1957–2012 1995 River Suyoukou (Syk) Suyou River 100.58 38.79 473.4 1565 1957–2012 1971 Abrupt change year was detected by the Pettitt test. Table 3: Classification of SPI, SDI, and VCI. and 2010) were obtained from the Cold and Arid Regions Science Data Centre (http://westdc.westgis.ac.cn/). *e Value Drought grades historical disaster records were derived from the China SPI/SDI≤ −2.0, VCI< 15 Extreme drought Meteorological Disaster Yearbook (Gansu volume). Zhang −2.0< SPI/SDI≤ −1.5, 15≤ VCI< 30 Severe drought et al. [49] used Pettitt tests to determine abrupt change −1.5< SPI/SDI≤ −1.0, 30≤ VCI< 45 Moderate drought points in the annual streamflow series from 1945 to 2012 and −1.0< SPI/SDI≤ 0, 45≤ VCI< 60 Mild drought from 1957 to 2012 at the Ylx and Zyx stations to ascertain the impact of the reservoir activities in the HRB. *e results SPI. It is worth noting that the procedure should be per- indicated only one significant abrupt change point in 1979 at formed separately for each month and can be calculated for Ylx station and in 1984 at Zyx station [49]. *e impact of any timescale (e.g., 1, 3, 6, 9, and 12 months). *e total human activities is more evident after the Chinese Gov- streamflow X in a given month j and year i depends on the ernment initiated the Ecological Water Diversion Project i,j chosen timescale k and can be calculated as [55] (EWDP) in mainstream Heihe River for 2000 [54]. *ere- fore, in this paper, we focused on the mainstream Ylx and X � 􏽘 V + 􏽘 V , if j< k; Zyx stations, between the Pre-R and Pos-R periods, before i,j i−1,l i,l l�13−k+l i�1 and after EWDP. *e abrupt change points for the (1) tributaries for six stations are shown in Table 2 (Pettitt tests). X � 􏽘 V , if j≥ k, i,j i,l i�j−k+l 2.3. Methods where V and V denote streamflow volumes in years i − 1 i-1,l i,l 2.3.1. Standardized Precipitation Index (SPI). *e SPI index and i, respectively. Due to the three-parameter log-logistic was developed by McKee et al. [24]. Within a certain distribution used to calculate the SPI, we also chose this geographic area, the precipitation usually fluctuates regu- probability distribution to standardize the observed larly. If the precipitation is less than the average annual streamflow data used to get the SDI to facilitate comparison precipitation, a drought may occur. If precipitation exceeds in the HRB. It is emphasized here that positive SDI values the annual average, flooding may occur. To calculate the SPI, reflect wet conditions, while negative values indicate a hy- a frequency distribution function is first constructed from a drological drought. series of long-term precipitation observations. In this paper, a three-parameter log-logistic distribution was used because the frequencies of precipitation accumulated at different 2.3.3. Vegetation Condition Index (VCI). Kogan and Sulli- van [56] proposed the VCI index in 1993. It is here assumed timescales are well modeled using these statistical distri- butions. Because SPI is based on the cumulative probability that the change of the VCI index is only affected by weather factors; in this way, the ecosystem factors and extreme of a given timescale, here the total amounts of precipitation in the current month and previous i months (i � 1, 2, 3, ......) weather factors are separated, the impact of extreme weather on vegetation is evaluated, and agricultural drought is were used to calculate SPI on a timescale of i + 1 month. In our study, we used the SPI (1, 3, 6, 9, ......, 48 months) to monitored. A smaller VCI index correlates to a greater analyze the characteristics of drought change in the HRB. drought. *e VCI index is calculated as follows: For drought classifications, please see Table 3. NDVI − NDVI J min VCI � × 100, (2) NDVI − NDVI max min 2.3.2. Streamflow Drought Index (SDI). *e SDI was de- veloped by Nalbantis and Tsakiris [28] to characterize hy- where NDVI, NDVI , and NDVI are monthly NDVI, max min drological drought based on developing the concepts of the multiyear maximum NDVI, and multiyear minimum NDVI Advances in Meteorology 5 (in this study, 14 years), respectively, for each grid cell. *e D: duration, S: severity, P: peak, D : nondrought duration, T : initiation time, T : termination time, and I: intensity VCI changes from 0 to 100 corresponding to changes in i e 2.5 vegetation condition from extremely unfavorable to optimal. 2.0 In the case of an extremely dry month, the vegetation condition is poor, and the VCI is close to or equal to zero. A 1.5 VCI of 50 reflects fair vegetation conditions. At optimal D = d + d + d 1.0 2 0 1 2 vegetation conditions, the VCI is close to 100 [34]. 0.5 d d d According to t-test, the trend was divided into extremely D 0 2 0.0 significant increase (slope> 0, p< 0.01), significant increase S S 1 T 2 (slope> 0, 0.01≤ p< 0.05), insignificant change (p≥ 0.05), T –0.5 e extremely significant decrease (slope< 0, p< 0.01), and S –1.0 significant decrease (slope< 0, 0.01≤ p< 0.05). –1.5 –2.0 2.4. Drought Variables and Run ;eory. According to McKee S = s + s 2 1 2 –2.5 et al. [24] and Spinoni et al. [57], the derived drought variables based on run theory [58] follow certain definitions –3.0 (Figure 2). When the monthly SPI/SDI values were below Figure 2: Definition of drought characteristics including peak, −1, the corresponding month was potentially identified as a intensity, duration, and magnitude using the SPI/SDI and run drought event (e.g., the four drought events for D , D , d , 0 1 0 theory. and d shown in Figure 2). *resholds of −1 (moderate drought), −1.5 (severe drought), and −2 (extreme drought) were used to identify drought events and drought charac- the cumulative water balance occurring from January to teristics. Drought duration was defined as the number of March [19, 21, 23, 41]. Meanwhile, to determine whether months from the first month in which the indicator goes there is a lag between the SPI (accumulation periods of 1–48 lower than −1 to the last month with a negative value before months) and SDI-1, cross-correlations were calculated for the indicator returns to positive values. If the duration of a SDI-1 series which were lagged by 0 to 6 months after the SPI drought event only contained one month where SPI/ series. In this case, the SPI accumulation period with the SDI< −1, this month was regarded as a single drought event strongest correlation with SDI-1 was denoted as the lagged (e.g., D with a magnitude S ). Drought severity was defined 1 1 SPI-n (1, 3, 6, 9, ....., 48 months) [23]. *e same interpre- as the sum of the monthly absolute values of the index when tation applies to the lag correlation of the VCI with different the index was≤−1.0 (−1.5, −2). A drought event may contain timescales SPI and VCI with different timescales SDI. a few consecutive months with negative SPI/SDI (e.g., du- ration D and its magnitude S ). If the duration of a drought 0 0 3. Results event contained two branches, such as d and d , and the 0 2 interruption period d between d and d was less than 6 1 0 2 3.1. MD and HD for the Mainstream Characteristics between months in which 0< SPI/SDI< 1 [59, 60], these months were Different Timescales and ;resholds. *e distribution of still regarded as a single drought event (D � d + d + d ). 2 0 1 2 drought characteristics based on thresholds of −1, −1.5, and *e corresponding magnitude was then defined as −2 for different timescales at different periods, including S � s + s [59, 60]. Intensity stands for the number of 2 1 2 Pre-R (1965–abrupt change year), Pos-R (abrupt change year months in which the drought indicator was lower than −1 +1–2012), and total period (1965–2012), was evaluated in the (−1.5, −2). Drought peak referred to the month in the UHRB and MHRB using boxplots. *e drought charac- “drought event” with the lowest value of the indicator [61]. teristics of SPI-1 and SDI-1 in the MHRB (Table S1) were almost consistent compared with historical records and, for 2.5. Correlation Analysis. Pearson’s correlation coefficient different combinations of timescales, thresholds and periods (PC) was calculated between SDI-1 (one month SDI) and SPI (Table S2) are shown in the supporting information. *e (1, 3, 6, 9, ....., 48 months), monthly (April–October) VCI, drought peak and intensity (Figures S1-S2), duration, and and different timescales of SPI and SDI (1, 3, 6, 9, and 12 severity for different timescales at different thresholds are months) series indicates the characteristics of interannual shown in Figure 3 for MD and in Figure 4 for HD. More propagation at 95% and 99% confidence levels. *e PC drought events were identified at shorter accumulation periods (1–12 months for MD and 1–6 months for HD) ranges from −1 to 1; a PC> 0 indicates a positive correlation and a PC< 0 indicates a negative correlation. In this study, based on thresholds equal to −1.0, −1.5, and −2.0 in the the PC method was also used to detect the intra-annual UHRB (Ylx) and MHRB (Zyx). For MD, the average in- propagation time of the SDI-1 to the different timescales of tensity, duration, and severity of fluctuations increased with SPI and monthly VCI to the different timescales of SDI and cumulative timescale increases between Pre-R and Pos-R SPI conditions. For instance, if the maximum correlation in periods. *e changes were more obvious in Pos-R compared the March SDI and the March series of SPI at the timescales with Pre-R periods for all thresholds in the UHRB. *e of 1, 3, . . ., 48 months is recorded at the 3-month timescale, average peak, intensity, duration, and severity of the MD did this means that the SDI in March is mostly determined by not change much for Pre-R, whereas Pos-R fluctuations SDI/SPI 6 Advances in Meteorology Pre-R Pos-R Total period 80 105 SPI = –1.0 1 3 6 9 12 15 18 21 24 36 48 1 3 6 9 12 15 18 21 24 36 48 1 3 6 9 12 15 18 21 24 36 48 100 100 80 80 60 60 SPI = –1.5 40 40 20 15 20 0 0 1 3 6 9 12 15 18 21 24 36 48 1 3 6 9 12 15 18 21 24 36 48 1 3 6 9 12 15 18 21 24 36 48 100 40 100 80 80 60 60 SPI = –2.0 40 40 20 20 0 0 1 369 12 15 18 21 24 36 48 1 3 6 9 1215182124 1 3 6 9 12 15 18 21 24 36 48 Timescale (SPI) Timescale (SPI) Timescale (SPI) (a) Pre-R Pos-R Total period 50 60 40 45 SPI = –1.0 30 20 20 0 0 1 3 6 9 12 15 18 21 24 36 48 1 369 12 15 18 21 24 36 48 1 369 12 15 18 21 24 36 48 70 50 SPI = –1.5 10 15 0 0 1 3 6 9 12151821243648 1 3 6 9 12151821243648 1 3 6 9 12 15 18 21 24 36 48 50 80 SPI = –2.0 15 20 0 0 1 3 9 12 18 21 24 48 1 3 6 9 12151821243648 1 3 6 9 12151821243648 Timescale (SPI) Timescale (SPI) Timescale (SPI) (b) Figure 3: Continued. Duration (months) Duration (months) Duration (months) Duration (months) Duration (months) Duration (months) Advances in Meteorology 7 Pre-R Pos-R Total period 150 160 SPI = –1.0 0 0 1 3 6 9 12 15 18 21 24 36 48 1 3 6 9 12151821243648 1 3 6 9 12151821243648 100 100 80 80 60 60 SPI = –1.5 30 40 40 20 20 0 0 1 3 6 9 12 15 18 21 24 36 48 1 36 9 12151821243648 1 3 6 9 12151821243648 80 80 70 70 60 60 50 50 40 40 SPI = –2.0 20 30 30 20 20 10 10 0 0 0 1 3 6 9 12 15 18 21 24 36 48 1 3 6 912 15 18 21 24 1 3 6 9 121518212436 48 Timescale (SPI) Timescale (SPI) Timescale (SPI) (c) Pre-R Pos-R Total period 80 80 70 70 60 60 50 50 40 40 SPI = –1.0 40 30 30 20 20 10 10 0 0 1 369 12 15 18 21 24 36 48 1 369 12 15 18 21 24 36 48 1 369 12 15 18 21 24 36 48 80 70 SPI = –1.5 40 20 20 0 0 1 3 6 9 12151821243648 1 3 6 9 12151821243648 1 3 6 9 12 15 18 21 24 36 48 80 80 80 70 70 60 60 50 50 40 40 SPI = –2.0 30 30 20 20 10 10 0 0 1 3 912 18 21 24 48 1 3 6 9 12151821243648 1 3 6 9 12151821243648 Timescale (SPI) Timescale (SPI) Timescale (SPI) (d) Figure 3: Boxplots showing MD duration and severity in the UHRB (a, c) and in the MHRB (b, d) based on SPI using thresholds of −1, −1.5, and −2 for different timescales at different periods (the dot and midline in the boxplots are the mean and median; the upper and lower boundaries of box are the interquartile range). increased with cumulative timescale increases using larger for HD than for MD because a short-term MD may thresholds of −1.0, −1.5, and −2.0. *e average peak, in- not develop into an HD [42] and as the number of events tensity, duration, and severity of value changed little in the decreases, duration lengthens, and severity worsens. *e MHRB. With regard to HD, the duration and severity were average intensity (Figure S2), duration, and severity of the Severity Severity Severity Severity Severity Severity 8 Advances in Meteorology Pre-R Pos-R Total period 40 140 100 30 SDI = –1.0 60 1 39 6 12 1 369 12 1 369 12 80 80 60 20 SDI = –1.5 15 1 39 6 12 1 369 12 1 369 12 SDI = –2.0 15 1 369 12 1 39 6 12 1 369 Timescale (SDI) Timescale (SDI) Timescale (SDI) (a) Pre-R Pos-R Total period 25 100 100 80 80 60 60 SDI = –1.0 40 40 20 20 0 0 1 39 6 12 1 369 12 1 369 12 100 100 80 80 60 60 SDI = –1.5 6 40 40 20 20 0 0 1 36 1 369 12 1 369 12 100 100 80 80 60 60 SDI = –2.0 40 40 20 20 0 0 1 3 1 369 12 1 369 12 Timescale (SDI) Timescale (SDI) Timescale (SDI) (b) Figure 4: Continued. Duration (months) Duration (months) Duration (months) Duration (months) Duration (months) Duration (months) Advances in Meteorology 9 Pre-R Pos-R Total period 60 25 20 50 SDI = –1.0 15 0 0 1 39 6 12 1 369 12 1 369 12 30 75 SDI = –1.5 1 39 6 12 1 369 12 1 369 12 SDI = –2.0 8 30 30 1 369 1 369 12 1 39 6 12 Timescale (SDI) Timescale (SDI) Timescale (SDI) (c) Pre-R Pos-R Total period 16 140 SDI = –1.0 4 60 1 3966 12 1 319 2 1 369 12 SDI = –1.5 3 50 1 36 1 31 6 9 2 1 369 12 140 140 120 120 100 100 80 80 SDI = –2.0 60 60 40 40 20 20 0 0 1 3 1 31 6 9 2 1 369 12 Timescale (SDI) Timescale (SDI) Timescale (SDI) (d) Figure 4: Boxplots showing HD duration and severity for Ylx (a, c) and for Zyx (b, d) based on SDI using thresholds of −1, −1.5, and −2 for different timescales at different periods (the dot and midline in the boxplots are the mean and median; the upper and lower boundaries of box are the interquartile range). Severity Severity Severity Severity Severity Severity 10 Advances in Meteorology available, e.g., Sh-Sd, Sss-Ql, Lyb-Yng, Ljq-Sd, Wfc-Ql, and HD were significantly different in Ylx and Zyx for different periods based on the thresholds −1.0, −1.5, and −2.0. In Ylx, Syk-Zy). Figure 8 indicates that the propagation times were 1, 1, 3, 1, 1, and 1 in the Pre-R period, respectively; 1, 3, 9, 3, the characteristics of HD were similar to MD in that, com- pared to the Pre-R period, the average intensity, duration, and 6, and 6 in the Pos-R period, respectively; and 1, 3, 6, 1, 6, severity of the HD decreased significantly, especially for SDI-3 and 6 in the total period for six tributaries, respectively. and SDI-12. However, in Zyx, the HD were quite inconsistent Figure 9(a) shows that the SDI-1 and SPI had the highest PC with MD, and the average intensity, duration, and severity of (PC � 0.27, p< 0.01) in the Pre-R period for the 9-month the HD increased significantly during the Pos-R period in period, and, in the Pos-R period, the highest PC (PC � 0.46, Zyx, especially for SDI-1, SDI-6, and SDI-12. p< 0.01) was observed for the 6-month period. Figure 9(a) also shows the highest PC for SDI-1 and SPI (PC � 0.38, p< 0.01) was for the 6-month period before the EWDP 3.2. Comparison of MD Characteristics for SPI-1 and HD for period, and, after the EWDP period, the highest PC SDI-1. *e characteristics of MD and HD based on SPI-1 and (PC � 0.48, p< 0.01) was observed for the 6-month period in SDI-1 using thresholds of −1, −1.5, and −2 in the UHRB and Ylx. Figure 9(b) reveals the highest PC (PC � 0.19, p< 0.01) MHRB are shown in Tables 4-5 and Figure 5. More MD was found for the 1-month period Pre-R and the 9-month events occurred at Yng station (names and abbreviations are period (PC � 0.24, p< 0.01) Pos-R. Figure 8(b) also dem- shown in Table 1) in the UHRB (Figure 5) and at Sd station in onstrates the highest PC (PC � 0.17, p< 0.01) was found for the MHRB. More HD events occurred at Zyx (Figure 5). It can the 3 months before the EWDP period and over 36 months be seen from Tables 4-5 that the average, maximum, and (PC � 0.35, p< 0.01) after the EWDP period in Zyx. *ese minimum peak/intensity/duration/severity in Zyx, Yng, and results demonstrate that the changes in the streamflow were Ql stations were identified as longer and more severe based on more sensitive to precipitation in the Pre-R period for short thresholds of −1.0, −1.5, and −2.0 for MD and HD. HD timescales (1–3 months), and Pos-R propagation time in- characteristics based on these thresholds for the average peak, creased (2–6 months) for tributaries, while the propagation maximum peak, and minimum peak in Sh, respectively, were time decreased by 3 months in Ylx and increased by 8 more obvious, and longer average, maximum, and minimum months in Zyx compared to the Pre-R period. *us, res- intensities were found in Ylx, Ljq, and Sh (Table 5). Longer, ervoir operation increased the propagation time of the more severe HD occurred in Ylx, Ljq, Sh, and Wfc. streamflow variables to the SPI during the Pos-R period in Zyx, whereas the propagation time of the streamflow vari- ables to the SPI was decreased during the Pos-R period in 3.3. Variation in VCI. Vegetation condition in the extremely Ylx. *ere was no change in Ylx before and after the EWDP, dry months of April, May, and October (Figure 6) was poor, while after the EWDP, the propagation time of the HD to the especially for alpine vegetation, cultivated vegetation, MD increased on a 33-month timescale compared with grassland, and marshy grassland. From June to September, before the EWDP in Zyx. Figure 9(c) shows that the VCI and the vegetation was in optimal condition for coniferous SPI strongest negative PC was found for the 6-month forests, shrubs, and marshy grassland in the UHRB and for (PC � −0.09) period in the UHRB, and the strongest positive coniferous forests and cultivated vegetation in the MHRB. It PC was found for the 36-month (PC � 0.13) period in the is worth noting that the vegetation condition was poor for MHRB. Figure 9(d) shows that the VCI and SDI highest PC alpine vegetation over the whole growing season. was found for the 6-month period (PC � 0.23, p< 0.01) in We further analyzed the trend and significance level of the the UHRB and the 3-month period (PC � 0.43, p< 0.01) in growing season for VCI. Figure 7 shows that, over the entire the MHRB. *e results indicated that correlation between growing season, alpine vegetation (UHRB) and grassland VCI and MD was not significant in the UHRB and MHRB. (MHRB) showed a significant decreasing trend, while culti- However, the correlation between VCI and HD was sig- vated vegetation (MHRB), coniferous forest, shrubs, and nificant in the UHRB (Ylx) and MHRB (Zyx), while the VCI grassland (UHRB) showed a significant increasing trend. propagation time to HD was found at short timescales. 3.4. Correlation of VCI, HD, and MD 3.4.2. Intra-Annual and Lag Correlations between VCI, HD, and MD at Multiple Scales in Different Periods. *e SDI-1 3.4.1. Interannual Propagation Time between VCI, HD, and MD at Different Timescales in Different Periods. with SPI at different timescales was investigated in relation to the propagation from MD to HD. Figure 10 shows the Figures 8-9 describe the relationship between SDI-1 and VCI-1 with all SPI accumulation periods (1–48 months) and propagation times from MD to HD had significant seasonal characteristics for tributaries in different periods. In spring VCI with all SDI accumulation periods (1–12 months) for (March–May), the drought propagation time was relatively different periods (Pre-R and Pos-R, before and after EWDP, short (1–12 months) in Sss, Ljq, and Wfc, whereas the and total period) in the mainstream for Ylx (SPI calculated propagation time was relatively long (1–48 months) in Sh, using the arithmetic mean of precipitation from Yng and Ql Lyb, and Syk. In summer (June–August), the drought stations), Zyx (SPI calculated using the arithmetic mean of propagation time was relatively long, except for Wfc. precipitation from Gt, Sd, and Zy stations), and tributaries (SPI calculation is based on the area flowing through or the Drought propagation times in autumn (September–No- vember) were relatively short (1–6 months) in Sss, Lyb, and adjacent area because all meteorological stations were not Advances in Meteorology 11 Table 4: *e distribution peak, intensity, duration, and severity for MD characteristics based on SPI-1 using thresholds of −1, −1.5, and −2 in the UHRB and MHRB. Peak Intensity Duration Severity *reshold Mean Max Min Mean Max Min Mean Max Min Mean Max Min −1.0 1.7 3.6 1.1 1.7 4.0 1.0 4.6 15.0 1.0 3.1 10.2 1.1 Ql −1.5 2.0 3.6 1.2 1.3 3.0 1.0 4.6 15.0 1.0 3.6 10.2 1.5 −2.0 2.5 3.6 2.0 1.0 1.0 1.0 4.3 15.0 1.0 4.1 10.2 2.0 −1.0 1.6 3.1 1.7 1.7 5.0 1.0 4.4 23.0 1.0 3.8 12.9 1.0 Yg −1.5 2.2 3.1 2.5 1.6 3.0 1.0 5.5 23.0 1.0 5.4 12.9 1.8 −2.0 2.5 3.1 2.0 1.2 2.0 1.0 6.9 23.0 1.0 5.1 12.5 2.0 −1.0 1.4 2.9 1.0 1.5 4.0 1.0 4.1 14.0 1.0 3.1 8.1 1.0 Gt −1.5 1.8 2.9 1.5 1.1 2.0 1.0 4.5 14.0 1.0 3.6 8.1 1.6 −2.0 2.7 2.9 2.5 1.0 1.0 1.0 2.0 3.0 1.0 3.5 4.0 2.9 −1.0 1.6 2.7 1.0 1.6 6.0 1.0 4.7 30.0 1.0 3.5 18.8 1.0 Zy −1.5 2.0 2.7 1.5 1.5 4.0 1.0 5.5 30.0 1.0 4.4 18.8 1.9 −2.0 2.3 2.7 2.0 1.1 2.0 1.0 4.0 15.0 1.0 3.9 10.0 2.0 −1.0 1.6 3.0 1.0 1.6 5.0 1.0 4.2 17.0 1.0 3.2 9.4 1.0 Sd −1.5 1.9 3.0 1.5 1.2 3.0 1.0 3.5 14.0 1.0 3.4 9.4 1.5 −2.0 2.3 3.0 2.0 1.1 2.0 1.0 3.1 11.0 1.0 3.7 9.4 2.0 Table 5: *e distribution of peak, intensity, duration, and severity for HD characteristics based on SDI-1 using thresholds of −1, −1.5, and −2 in the UHRB and MHRB. Peak Intensity Duration Severity *reshold Mean Max Min Mean Max Min Mean Max Min Mean Max Min −1.0 1.6 2.7 1.1 4.3 18.0 1.0 11.1 31.0 1.0 8.6 30.8 1.1 Ylx −1.5 1.9 2.7 1.5 4.0 18.0 1.0 12.3 31.0 1.0 10.9 30.8 1.5 −2.0 2.3 2.7 2.1 1.8 3.0 1.0 19.5 31.0 10.0 18.0 24.6 8.9 −1.0 1.8 3.3 1.0 3.7 14.0 1.0 12.4 36.0 1.0 8.1 24.7 1.0 Sss −1.5 2.0 3.3 1.5 2.1 5.0 1.0 12.4 30.0 1.0 8.5 22.2 1.5 −2.0 2.6 3.3 2.0 1.3 2.0 1.0 9.5 20.0 2.0 7.7 12.2 2.2 −1.0 1.7 3.0 1.0 3.4 17.0 1.0 8.9 51.0 1.0 7.0 37.0 1.2 Wfc −1.5 2.0 3.0 1.6 2.7 8.0 1.0 12.2 51.0 1.0 9.8 37.0 1.6 −2.0 2.5 3.0 2.0 2.0 4.0 1.0 19.3 54.0 6.0 15.5 40.4 6.6 −1.0 1.8 3.0 1.0 4.1 16.0 1.0 10.2 30.0 1.0 8.7 35.0 1.0 Lyb −1.5 2.1 3.0 1.6 2.8 8.0 1.0 10.4 30.0 1.0 9.9 35.0 1.7 −2.0 2.4 3.0 2.0 1.4 4.0 1.0 8.3 28.0 1.0 9.4 32.0 2.0 −1.0 1.8 5.0 1.0 3.0 21.0 1.0 7.0 40.0 1.0 6.2 41.3 1.1 Zxy −1.5 2.2 5.0 1.5 2.6 12.0 1.0 6.9 38.0 1.0 7.6 38.9 1.5 −2.0 2.7 5.0 2.0 1.6 6.0 1.0 9.5 37.0 1.0 10.2 37.3 2.1 −1.0 1.8 4.0 1.0 3.4 17.0 1.0 7.0 32.0 1.0 6.8 33.7 1.1 Ljq −1.5 2.4 4.0 1.5 3.3 12.0 1.0 10.3 32.0 1.0 10.6 33.7 1.5 −2.0 2.9 4.0 2.1 3.0 12.0 1.0 13.6 32.0 3.0 14.0 29.8 3.8 −1.0 2.1 4.3 1.0 4.3 18.0 1.0 12.9 43.0 1.0 9.7 36.3 1.7 Sh −1.5 2.5 4.3 1.6 3.0 9.0 1.0 8.9 33.0 1.0 9.2 35.3 1.7 −2.0 2.8 4.3 2.0 1.8 5.0 1.0 8.4 32.0 1.0 10.3 33.7 2.3 −1.0 1.7 3.8 1.0 2.4 8.0 1.0 7.6 36.0 1.0 5.2 18.7 1.0 Syk −1.5 2.0 3.8 1.5 1.6 3.0 1.0 7.6 36.0 1.0 5.9 18.7 1.8 −2.0 2.9 3.8 2.4 2.0 3.0 1.0 10.8 35.0 2.0 9.2 18.7 3.2 Syk and relatively long (1–48 months) in Sh, Ljq, and Wfc. drought propagation times in winter did not change and Drought propagation times in winter (December–February) remained long, whereas drought propagation times were were similar to those in summer and were relatively long increased in autumn. Propagation time was longer for Ss, (1–48 months), except for Lyb in the Pre-R period. Drought Lyb, and Syk and shorter for Sss, Ljq, and Wfc in spring; and propagation times in autumn and winter were relatively propagation times in summer were shorter, except for Wfc long, whereas in spring and summer, they were relatively and Ss. It can be clearly seen from Figure 11 that the intra- short in the Pos-R period. *e total period was similar to the annual correlations between the monthly SDI-1 and SPI at Pos-R period. To summarize, in the Pos-R period, the multiple scales and in different periods have strong seasonal 12 Advances in Meteorology N N 15 9 23 35 9 30 10 16 8 6 11 8 3.8 0 60 120 0 60 120 Kilometers Kilometers Events Events Threshold = –1.0 Threshold = –1.0 Threshold = –1.5 Threshold = –1.5 Threshold = –2.0 Threshold = –2.0 (a) (b) Figure 5: Spatial distribution of MD (SPI-1) and HD (SDI-1) events in the UHRB and MHRB. N N 0 60 120 0 60 120 Kilometers Kilometers VCI VCI 4.2 – 15 46 – 60 3 – 15 46 – 60 16 – 30 61 – 65 16 – 30 61 – 90 31 – 45 31 – 45 (a) (b) N N 0 60 120 0 60 120 Kilometers Kilometers VCI VCI 4.2 – 15 46 – 60 2.5 – 15 46 – 60 16 – 30 61 – 91 16 – 30 61 – 95 31 – 45 31 – 45 (c) (d) Figure 6: Continued. Advances in Meteorology 13 N N 0 60 120 0 60 120 Kilometers Kilometers VCI VCI 2.5 – 15 46 – 60 2.8 – 15 46 – 60 16 – 30 61 – 97 16 – 30 61 – 97 31 – 45 31 – 45 (e) (f) 0 60 120 Kilometers VCI 2.6 – 15 46 – 60 16 – 30 61 – 89 31 – 45 (g) Figure 6: *e spatial variation in the average VCI for the growing season (April–October) in the UHRB and MHRB. N N 0 60 120 0 60 120 Kilometers Kilometers Extremely significant decrease Insignificant increase Extremely significant decrease Insignificant increase Significant decrease Significant increase Significant decrease Significant increase Insignificant decrease Extremely significant increase Insignificant decrease Extremely significant increase (a) (b) Figure 7: Continued. 14 Advances in Meteorology N N 0 60 120 0 60 120 Kilometers Kilometers Extremely significant decrease Insignificant increase Extremely significant decrease Insignificant increase Significant decrease Significant increase Significant decrease Significant increase Insignificant decrease Extremely significant increase Insignificant decrease Extremely significant increase (c) (d) N N 0 60 120 0 60 120 Kilometers Kilometers Extremely significant decrease Insignificant increase Extremely significant decrease Insignificant increase Significant decrease Significant increase Significant decrease Significant increase Insignificant decrease Extremely significant increase Insignificant decrease Extremely significant increase (e) (f) 0 60 120 Kilometers Extremely significant decrease Insignificant increase Significant decrease Significant increase Insignificant decrease Extremely significant increase (g) Figure 7: *e spatial distribution of VCI trends for the growing season in the UHRB and MHRB. Advances in Meteorology 15 0.30 0.15 0.00 –0.15 1 3 6 9 12151821243648 Timescale (SPI) Sh Ljq Sss Wfc Lyb Syk (a) 0.30 0.15 0.00 –0.15 1 3 6 9 12151821243648 Timescale (SPI) Sh Ljq Sss Wfc Lyb Syk (b) 0.30 0.15 0.00 –0.15 1 39 6 12 15 18 21 24 36 48 Timescale (SPI) Sh Ljq Sss Wfc Lyb Syk (c) Figure 8: Pearson’s correlation coefficien between the SDI-1 series and the SPI series at different timescales for six tributaries. characteristics in Ylx and Zyx. *e propagation times in autumn, and winter were approximately 9, 9, 3, and 6 months in Zyx for the total period (Figure 11(f)), respec- spring, summer, autumn, and winter were approximately 3, 9, 6 and 12 months in the Ylx for the Pre-R period tively. *ese results show that the propagation time in the (Figure 11(a)), respectively. *e propagation times in the Pos-R period increased in spring, decreased in autumn and spring, summer, autumn and winter were approximately 6, winter, and did not change in summer in Ylx compared to 9, 1, and 6 months in Ylx for the Pos-R period the Pre-R period, whereas in Zyx it increased in spring, (Figure 11(b)), respectively. *e propagation times in summer, and autumn, and it decreased in winter. Com- spring, summer, autumn, and winter were 6 months in Ylx pared with the Pre-R period, the PC was lower on a 1- for the total period (Figure 11(c)). *e propagation times in month timescale in Ylx (Figure 11(b)) and on a 1–3-month spring, summer, autumn, and winter were approximately 6, timescale in Zyx (Figure 11(e)). However, the PC between 1, 1, and 15 months in Zyx for the Pre-R period SDI-1 and different timescale of SPI in the Pos-R period (Figure 11(d)), respectively. *e propagation times in was higher than that in the Pre-R period for long-term timescales (>12 months) in Ylx (except for January and spring, summer, autumn, and winter were 9, 12, 36, and 9 months in Zyx for the Pos-R period (Figure 11(e)), re- June) and Zyx (except for June, July, November, and spectively. *e propagation times in spring, summer, December). It can be seen from Figures S3(a)-S3(b) that, PC PC PC 16 Advances in Meteorology 0.45 0.30 0.30 0.15 0.15 0.00 0.00 –0.15 1 3 6 912 15 18 21 24 36 48 1 361 912 158 21243648 Timescale (SPI) Timescale (SPI) UHRB (1965–2012) Pre-R (1965–1979) MHRB (1965–2012) Pre-R (1965–1979) Pos-R (1980–2012) Before EWDP (1965–1999) Pos-R (1980–2012) Before EWDP (1965–1999) Aer EWDP (2000–2012) Aer EWDP (2000–2012) (a) (b) 0.45 0.15 0.10 0.30 0.05 0.00 0.15 –0.05 –0.10 0.00 –0.15 1 3 6 9 12151821243648 1 369 12 Timescale (SPI) Timescale (SDI) UHRB UHRB MHRB MHRB (c) (d) Figure 9: Pearson’s correlation coefficien between SDI-1 and SPI (a, b) for different timescales at different periods, and VCI (c, d) with different timescales of SPI and SDI in the UHRB (Ylx) and in the MHRB (Zyx). 0.05 0.22 0.39 0.56 0.73 0.89 0.05 0.22 0.39 0.56 0.73 0.89 PC PC 12 12 11 11 10 10 9 9 8 8 7 7 6 6 5 5 4 4 Sh Sss Lyb Ljq Wfc Syk Sh Sss Lyb Ljq Wfc Syk Stations Stations (a) (b) Figure 10: Continued. Pre-R PC PC Month PC Pos-R PC Month 29 Advances in Meteorology 17 0.05 0.22 0.39 0.56 0.73 0.89 1 10 20 39 48 PC Months 12 12 11 11 10 10 9 9 8 8 7 7 6 6 5 5 4 4 3 3 2 2 1 1 Sh Sss Lyb Ljq Wfc Syk Sh Sss Lyb Ljq Wfc Syk Stations Stations (c) (d) 110 20 39 48 1 10 20 39 48 Months Months 12 12 11 11 10 10 9 9 8 8 7 7 6 6 5 5 4 4 3 3 2 2 1 1 Sh Sss Lyb Ljq Wfc Syk Sh Sss Lyb Ljq Wfc Syk Stations Stations (e) (f) Figure 10: Intra-annual strongest PC (a–c) and the drought propagation time in months (e–f) for SPI and SDI-1 at six tributaries. 1–48 months for different timescales at different periods after the EWDP, the PC of SDI-1 and the timescale of SPI were higher in spring and autumn and lower in summer (Pre-R, Pos-R, and total), and the strongest correlation was (except for August) and winter than before the EWDP in detected at a lag of 0 months (i.e., no lag) in Ylx and Zyx Ylx. However, Figures S3(c)-S3(d) show that, after the (Figure 13). We also found no lag before and after the EWDP EWDP, the PC of SDI and SPI was higher in spring, (Figure S5), whereas there was a higher PC at 3–12 months summer, autumn, and winter, specifically in spring, in Zyx. before the EWDP and after the EWDP at 6–15 months in In different periods (Pre-R and Pos-R and before and after Ylx. *e highest PC was at 1–12 months before the EWDP EWDP) and different timescales, the PC of the Pos-R and after EWDP at 9–48 months in Zyx. We also analyzed period and after the EWDP was higher in Ylx (except for 1, the lag correlation of the SPI with VCI-1 and SDI (accu- mulation periods of 1–12 months) with VCI-1 (Figure 14), 36, and 48 months) and Zyx, especially in Zyx than in the Pre-R period and before the EWDP. Figure 12(a) shows and PC was calculated for the VCI-1 series, which were lagged by 0 to 6 months after the SPI and SDI series. *e that the higher PC between VCI and SPI in April at 6 months was 0.63 (p< 0.01), whereas the lower PC in May at results demonstrated that the highest PC (PC � 0.21, 1 month and 48 months was −0.06. *e highest PC between p< 0.01) was found at a 6-month lag between VCI and SPI, VCI and SDI in May at 12 months was 0.78 (p< 0.01), while the strongest correlation was at a lag of 0 months (i.e., whereas the lowest PC in March at 3 months was −0.69 no lag) between VCI and SDI in the UHRB (Ylx) during (p< 0.01) in the UHRB (Figure 12(b)). Figure 12(c) shows 1999–2012. *e highest PC (PC � 0.26, p< 0.01) was ob- that the highest PC between VCI and SPI was 0.84 served at the 4-month lag scale between VCI and SPI, and (p< 0.01) in July at 9 months, whereas the lowest PC was the strongest PC (PC � 0.47, p< 0.01) was found at a lag of 1 −0.19 in October at 1 month. *e highest PC between VCI month between VCI and SDI during 1999–2012 in the MHRB (Zyx). As mentioned above, it can be concluded that and SDI was observed at 0.71 in May at 6 months, and the lowest PC was −0.23 in March at 6 months in the MHRB there was no lag between SDI and SPI; however, VCI with (Figure 12(d)). SPI had significant lag correlations in the short term in the We further analyzed whether there was a lag between the UHRB and MHRB. Additionally, the VCI with SDI had a different timescales of SPI with SDI-1. *e PC between SDI- significant 1-month lag correlation in the short term in the 1 lagged by 0–6 months, with SPI accumulation periods of MHRB (Zyx). Total period Month Month Month Month –0.04 18 Advances in Meteorology PC PC 0.78 0.83 11 11 10 10 0.54 0.62 9 9 8 8 0.29 0.41 7 7 6 6 0.05 0.21 5 5 4 4 –0.20 0.00 3 3 2 2 1 1 –0.45 –0.21 1 369 12 15 18 21 24 36 48 1 369 12 15 18 21 24 36 48 Timescale (SPI) Timescale (SPI) (a) (b) PC PC 0.75 0.51 12 12 11 11 10 10 0.58 0.28 9 9 8 8 0.40 0.05 7 7 6 6 0.23 –0.18 5 5 4 4 0.06 –0.40 3 3 2 2 1 1 –0.12 –0.63 1 369 12 15 18 21 24 36 48 1 369 12 15 18 21 24 36 48 Timescale (SPI) Timescale (SPI) (c) (d) PC PC 0.46 0.54 12 12 11 11 10 10 0.29 0.34 9 9 8 8 0.13 0.13 7 7 6 6 –0.08 5 5 4 4 –0.20 –0.29 3 3 2 2 1 1 –0.37 –0.50 1 369 12 15 18 21 24 36 48 1 369 12 15 18 21 24 36 48 Timescale (SPI) Timescale (SPI) (e) (f) Figure 11: Heat map showing Intra-annual PC of SPI accumulation periods of 1–48 months with SDI-1 in the UHRB (Ylx) and MHRB (Zyx) (note: the x-axis represents the correlation between the SDI-1 and SPI at different timescales and the y-axis represents the 12 months of the year; the up arrow represents the correlation coefficients that increased in the postreservoir period (Pos-R) in the Ylx and Zyx compared to the period of prereservoir (Pre-R), or vice versa); (a) (1965–1979, PC≥ 0.52, p< 0.05), (b) (1980–2012, PC≥ 0.35, p< 0.05), and (c) (1965–2012, PC≥ 0.29, p< 0.05) in the UHRB (Ylx), and (d) (1965–1984, PC≥ 0.45 or≤− 0.47, p< 0.05), (e) (1985–2012, PC≥ 0.38 or≤−0.48, p< 0.05), and (f ) (1965–2012, PC≥ 0.29 or ≤ −0.33, p< 0.05) in the MHRB (Zyx). Total period Pre-R Month Month Month Pos-R Month Month Month 0.15 0.19 –0.01 Advances in Meteorology 19 UHRB (VCI-SPI) PC MHRB (VCI-SPI) PC 0.63 0.85 9 9 0.49 0.64 8 8 0.35 0.43 7 7 0.22 0.22 6 6 5 0.08 5 0.02 4 4 –0.06 –0.19 136 9 12 15 18 21 24 36 48 136 9 12 15 18 21 24 36 48 Timescale (SPI) Timescale (SPI) (a) (b) UHRB (VC-SDI) PC MHRB (VCI-SDI) PC 0.78 0.71 9 0.62 9 0.61 8 8 0.46 0.52 7 7 0.31 0.42 6 6 0.33 5 5 4 4 0.23 16 3 9 12 16 3 9 12 Timescale (SDI) Timescale (SDI) (c) (d) Figure 12: Heat map showing growing season Pearson’s correlation coefficients between VCI and different timescale of SPI, VCI, and different timescale of SDI for the period of 1999–2012 in the UHRB (Ylx). (a) (PC≥ 0.53, p< 0.05), (b) (PC≥ 0.55 or≤ −0.69, p< 0.01) and MHRB (Zyx), (c) (PC≥ 0.55, p< 0.05), and (d) (PC≥ 0.54, p< 0.05). PC PC 0.27 0.46 6 6 5 0.18 5 0.37 4 4 0.09 0.28 3 3 –0.00 1 –0.10 1 0.11 0 0 –0.19 0.02 1 3 6 9 12 15 18 21 24 36 48 1 3 6 9 12 15 18 21 24 36 48 Timescale (SPI) Timescale (SPI) (a) (b) Figure 13: Continued. Month Month Pre-R Lag-months Month Pos-R Month Lag-months 2 20 Advances in Meteorology PC PC 0.44 0.19 6 6 5 5 0.37 0.10 4 4 0.30 0.02 3 3 0.22 –0.06 1 0.16 1 –0.14 0 0 0.09 –0.22 1 3 6 9 12 15 18 21 24 36 48 1 3 6 9 12 15 18 21 24 36 48 Timescale (SPI) Timescale (SPI) (c) (d) PC PC 0.24 0.18 6 6 5 0.19 5 0.13 4 4 0.13 0.08 3 3 0.08 0.03 1 0.03 1 –0.02 0 0 –0.02 –0.07 1 3 6 9 12 15 18 21 24 36 48 1 3 6 9 12 15 18 21 24 36 48 Timescale (SPI) Timescale (SPI) (e) (f) Figure 13: Heat maps showing PC between SDI-1 lagged by 0–6 months and SPI accumulation periods of 1–48 months for different timescales at different periods in the UHRB (Ylx) and MHRB (Zyx). (a) (1965–1979, PC≥ 0.15 or≤‒0.16, p< 0.05), (b) (1980–2012, PC≥ 0.14, p<0.01), and (c) (1965–2012, PC≥ 0.11, p<0.01) in the UHRB (Ylx) and (d) (1965–1984, PC≥ 0.16 or ≤‒0.11, p< 0.05), (e) (1985–2012, PC≥ 0.12, p< 0.01), and (f) (1965–2012, PC≥ 0.10, p< 0.05) in the MHRB (Zyx). For the total period (Table S2), HD events were more fre- 4. Discussion quent at Zyx than at Ylx for different timescales based on Our results indicated that MD were more serious in Pos-R thresholds of −1.0, −1.5, and −2.0. *ese results were than in Pre-R in the UHRB, there was little change in the consistent with Ma et al. [44]. Figure 1(c) and Table 6 show MHRB, and MD were more serious in the UHRB than in the that the main vegetation type in the UHRB was grassland, MHRB for duration and severity based on thresholds of and there was an increasing trend to cultivated land for −1.0, −1.5, and −2.0 for the total period. *ese results are 2000–2010. *e UHRB had fewer reservoirs and less arable inconsistent with [44], probably because the precipitation land than the MHRB, and the drought duration and severity data types and SPI probability distribution function were were more likely to be related to climate change. For ex- different. In our study, precipitation data types were based ample, Ma et al. [44], Qiu et al. [52], Gao and Zhang [62], on meteorological stations, and the SPI probability distri- Yang et al. [63], Cong et al. [64], and Cheng et al. [65] bution function was a three-parameter log-logistic distri- showed that an increase in streamflow was caused by climate bution. However, Ma et al. [44] used 0.5 resolution change (increased precipitation) factors rather than human precipitation data and their SPI probability distribution activities in the UHRB. At Zyx, the duration and severity of function was an empirical Gringorten plotting position. *e drought were greater than those in the Pre-R period. More Pos-R period had a significant positive impact on the extreme MD events occurred in 1985, 1992, 2000, and 2005 evolution of different timescales based on thresholds of −1.0, (i.e., Pos-R) [66], and the decrease in precipitation may have −1.5, and −2.0 in Ylx, which was consistent with Wu et al. led to a decrease in streamflow and thus more serious HD [60], who showed that Pos-R period in the HD had reduced events compared with the Pre-R period (Table S2). *e area duration and severity compared with Pre-R period. How- of cultivated land increased, and unused land and grassland ever, there were more serious duration and severity, spe- decreased from 1975 to 2010. *e total irrigation volume in 8 3 cifically at 6 and 12 months, than in the Pre-R period in Zyx. the area increased by 0.57 ×10 m /year in 1990–2010 in the Total-period Lag-months Lag-months Lag-months Lag-months Advances in Meteorology 21 between VCI and SPI, consistent with Li et al. [19]. MHRB [54]. Along with the rapid increase in the number of pumping wells, groundwater irrigation has increased Meanwhile increasing drought timescales (≥12 months) 8 3 suggested positive correlations in the UHRB and MHRB 2.11 × 10 m /year [54]. It is clear that human activities have changed the land surface and altered hydrological processes, (Figure 8(c)). Niu et al. [45] also found that variations in the including evapotranspiration, infiltration, surface runoff, Enhanced Vegetation Index (EVI) generally followed the and storage of water and; in this way, they affect the de- evolution of soil moisture in the top 1 m of the soil profile. velopment of drought and thereby exacerbate drought [67]. *ere were both positive and negative time lags. For ex- Zhu et al. [68] also pointed out that human activities, such as ample, during a 2001 drought event, it was observed that the the expansion of irrigation, rapid population growth, and Soil Moisture Anomaly Index reached a negative peak value. However, the EVI did not show a consistent response, in- socioeconomic development, have deeply modified hydro- logical processes in the MHRB. *erefore, the above- dicating it was not affected by natural drought events be- cause it was maintained by irrigation activities in cropland in mentioned climate change and human activities may affect changes in drought characteristics between the different the MHRB. *e strongest PC values between VCI and SPI were for 6-month and 36-month timescales for the growing periods in the UHRB and MHRB. Previous research has examined the propagation time season in the UHRB and MHRB, respectively. *ese results and lags of HD to MD in different regions of the world were inconsistent with Vicente-Serrano et al. [41], who [4, 23, 43, 44]. In this study, we found that the PC at Ylx was found that, in arid and humid regions, vegetation responded larger than that at Zyx. *e correlation with climate change to MD at short timescales as well as in semiarid and sub- and human activities changed from positive to negative at humid regions at long timescales. *e main reason may be Zyx. Ma et al. [44] found that climate change was inclined to the relationship with different vegetation types, anthropo- genic activity, and climate change, although the UHRB is a increase streamflow and propagation time, contributing from −57% to 63% in the UHRB, whereas human activities semiarid region. In Figure 1(c), there is little coniferous forest, and the region is mainly grassland, shrub, and alpine played a dominant role in water consumption with a con- tribution rate greater than −89% to further alter HD vegetation. However, cultivated vegetation dominates in the MHRB, and there are 24 irrigation districts (Figure S5) and characteristics and propagation time in the MHRB. In this study, we found the MD propagation to HD was different in 71 reservoirs (four under construction) with thousands of different catchments and had a significant seasonal dis- canals and over 6000 pumping wells for irrigation [71]. *e crepancy at different timescales for different periods. *e total irrigation volume in the area has increased in the reservoirs increased the propagation time of the hydrological MHRB [52]. *us, crop improvements and water-saving variables to MD during the Pos-R period at Zyx, and the irrigation technologies would increase water use efficiency results were in line with Wu et al. [60]. *e PC values and drought tolerance capacity [70]. Our findings also differed from Zhang et al. [18], who found significant annual between SDI and SPI in the Pos-R period were lower on a short-term scale (1–3 months) than those in the Pre-R high PC values in arid and semiarid regions, and the cor- responding drought timescales were 3–6 months (Januar- period, whereas they were higher for the long-term timescale (>12 months) at Ylx and Zyx, which was also consistent with y–December). *is may be related to the different timescales (April–October). We further found a significant positive and Wu et al. [60]. *e propagation time in the Pos-R period increased in spring, decreased in autumn and winter, and high PC value (PC≥ 0.53, p≤ 0.05) between VCI and SPI on did not change in summer at Ylx compared with the Pre-R an intra-annual scale, implying that water availability is the period, while it increased in spring, summer, and autumn critical factor for vegetation’s various spatiotemporal ac- and decreased in winter at Zyx. *e possible causes of the tivities [52] in the UHRB and MHRB, especially in the uneven spatial variations of precipitation are temperature MHRB (PC≥ 0.70, p≤ 0.01). Our results also revealed that, and melting of snow and glaciers in the Qilian Mountains with an increasing drought timescale, the PC values between VCI and SPI were higher in the MHRB, specifically in when the temperature is rising, which may lead to the differences in HD variations [44, 62]. *e propagation time August (the PC values fluctuated from 0.43 to 0.82 for 1–48 months) and September (the PC values fluctuated from 0.07 for tributaries would also vary in different periods in relation to natural and social environmental factors [4, 23, 69]. For to 0.83 for 1–48 months). *ese results suggested that instance, Deng et al. [70] demonstrated that the 33 croplands can suffer from extreme and prolonged drought tributaries in the MHRB no longer joined the mainstream conditions [10] in the MHRB. Furthermore, different veg- after 1980s, and they gradually disappeared and formed etation types have different propagation times for MD; the independent irrigation oases. *ese may diverge in terms of grassland and cultivated vegetation at a 3-month timescale different climate and human activity impacts. *e lack of lag and a 12-month timescale for the shrubland, coniferous in the SDI and SPI was in line with Barker et al. [23] at Ylx forest, and broadleaf forest [8, 21]. Figures 14 and 6 dem- and Zyx. onstrate that alpine vegetation and grassland were poor (especially in April and May) and there was an extremely *e high correlation between monthly NDVI and the MD index at different timescales is an indicator that can significant decrease for the growing season, whereas culti- vated vegetation was good, and there was an extremely describe the impact of drought on vegetation, and the month of highest correlation means the greatest sensitivity of significant increase for the growing season in the UHRB and vegetation to drought [10, 41]. Our results exhibited a MHRB. Additionally, VCI with SPI had a significant lag negative correlation on short timescales (≤9 months) correlation in the short term in the UHRB (6-month lag) and 2 22 Advances in Meteorology UHRB (VCI-SPI) PC MHRB (VCI-SPI) PC 0.21 0.26 6 6 5 5 0.15 0.18 4 4 0.09 0.11 3 3 0.02 0.03 1 –0.04 1 –0.04 0 0 –0.10 –0.12 1 3 6 9 12 15 18 21 24 36 48 1 3 6 9 12 15 18 21 24 36 48 Timescale (SPI) Timescale (SDI) (a) (b) PC MHRB (VCI-SDI) VCI–SPI (VCI-SDI) PC 0.25 0.48 6 6 5 5 0.15 0.34 0.05 0.20 0.07 0.05 –0.07 1 –0.15 1 0 0 –0.24 –0.21 1 369 12 1 369 12 Timescale (SDI) Timescale (SDI) (c) (d) Figure 14: Heat maps showing PC between VCI lagged by 0–6 months and SPI (1-48)/SDI (1-12) for the period of 1999–2012 in the UHRM (Ylx). (a) PC≥ 0.16, p< 0.05), (b) PC≥ 0.18 or≤‒0.15, p< 0.05) and in the MHRB (Zyx), (c) PC≥ 0.15, p< 0.05), and (d) PC≥ 0.16 or≤‒0.18, p< 0.05). Table 6: Changes in the area of the primary types of land use and transition matrix and the primary types of land use changes in the UHRB 4 2 and MHRB (10 hm ). Period Land use types Cultivated Forest Grassland Water area Built-up area Unused land Cultivated 0.10 0.02 0.16 0.01 0.02 0.00 Forest 0.05 1.84 16.78 0.48 0.00 1.81 Grassland 0.29 0.65 42.13 1.27 0.03 6.13 UHRB 2000–2011 Water area 0.04 0.06 0.96 0.91 0.01 0.63 Built land 0.03 0.00 0.04 0.00 0.02 0.00 Unused land 0.00 0.08 7.83 0.57 0.00 16.87 Cultivated 14.41 0.04 0.26 0.05 0.28 0.15 Forest 0.42 1.20 0.21 0.01 0.01 0.12 Grassland 3.97 0.26 12.86 0.16 0.06 1.70 MRHB 1975–2010 Water area 0.53 0.02 0.34 1.86 0.01 0.19 Built land 0.06 0.00 0.00 0.00 1.54 0.00 Unused land 1.84 0.38 1.46 0.14 0.22 62.76 1975–1987 14.40 1.52 15.33 2.22 1.60 65.52 1987–1992 15.25 1.21 14.49 2.04 1.34 66.88 MRHB 1992–2001 16.48 1.23 14.29 2.11 1.41 67.04 2000–2010 17.88 1.25 13.03 1.80 1.71 63.87 MHRB (4-month lag), and the VCI with SDI had a sig- timescale of the HD and VCI with the different timescales of nificant 1-month lag correlation in the MHRB, similar to Lin the MD did not fully reflect their relationships. *e reason is et al. [35]. Additionally, the correlation between a single that the streamflow and vegetation propagation times with Lag-months Lag-months Lag-months Lag-months Advances in Meteorology 23 precipitation differ at different timescales the average peak, intensity, duration, and severity of [10, 18, 19, 21, 72, 73]. HD decreased at Ylx, whereas the effects were more Our study had some limitations. *e quantification of serious at Zyx. *e PC values were lower at short-term drought is a difficult task and drought is intrinsically scales (1–3 months) than at long-term scales for the multiscalar. *ere is no unique physical variable we can Pos-R period at Ylx and Zyx. *e PC values for MD measure to quantify drought intensity. Vegetation propa- and HD after the EWDP were higher than those gation time to droughts is still an open scientific problem before the EWDP in the UHRB and MHRB, especially due to the complexity of droughts and limited knowledge of in the MHRB. Human activities at different timescales physiological processes. However, understanding the rela- (Pre-R period and Pos-R period and before EWDP tionship between these mechanisms and the characteristics and after EWDP) may affect the correlation between of droughts is crucial for improving our knowledge of drought and the timescale of MD lag to HD. vegetation vulnerability to climate fluctuations and climate (2) *e propagation time decreased 3 months at Ylx and change [41]. Meanwhile, with global warming, the propa- increased 8 months at Zyx compared to the Pre-R gation time of VCI and HD, VCI and MD, and HD and MD period and between VCI and SPI at 6 months in the to climate change also has changed [5], which poses new UHRB and 9 months in the MHRB. However, VCI challenges that need further research. Furthermore, in- with SDI at 6 months at Ylx and 3 months at Zyx did creasing human activity has dramatically changed natural significantly affect the growing season. On an intra- droughts; to manage drought effectively, we need to ac- annual scale, for Ylx, Zyx, and six tributaries, knowledge that human influence is as integral to drought as propagation times were different in the Pre-R and natural climate variability. *e complex interactions be- Pos-R periods, and there were seasonal differences tween natural and human processes need to be considered between Pos-R and Pre-R periods. [67]. In addition, there are still some issues worthy of in- (3) *e SDI and SPI showed no lag at Ylx and Zyx. depth discussion. Often, people are considered as passive However, VCI with MD had a significant lag corre- recipients at the end of the propagation cascade from MD lation at the short-term scale in the UHRB (6-month via soil moisture drought to HD [74] but, in fact, people lag) and MHRB (4-month lag), while the VCI with actively influence drought propagation [67]. Anthropogenic HD has significant 1-month lag correlation at Zyx. changes to the land surface alter hydrological processes, (4) Overall, the average peak/intensity/duration/severity including evapotranspiration, infiltration, surface stream- of MD and HD are weakening and the propagation flow, and storage of water [67, 75]. Understanding how time from MD to HD is also reduced for Pos-R irrigation [76] (the effect of long-term irrigation in the HRB period compared with Pre-R period in the UHRB. on the evolution of drought [77, 78]), land use/land cover Constructing reservoirs has had a positive effect, change [79], and construction of reservoirs or dams prolonging MD propagation to HD times, but the [42, 63, 64, 80, 81] impact MD, HD, and VCI characteristics drought characteristics (average peak/intensity/du- and MD propagation to HD and VCI need further research. ration/severity) have increased from the perspective Trnka et al. [82] presented a set of 60 priority questions and of climatology. *erefore, positive drought preven- considered interdisciplinary characteristics to optimize tion measures are necessary to consider the char- future drought research, primarily covering the following acteristics of propagation between MD and HD as drought-related topics: monitoring, impacts, drought well as seasonal differences in the MHRB. forecasts, climatology, adaptation, and planning. *erefore, in future drought research, it is necessary to comprehen- sively consider priority questions and interdisciplinary Data Availability characteristics, especially for irrigated agricultural areas in *e data used to support the findings of this study are arid regions. available from the corresponding author upon request. 5. Conclusions Conflicts of Interest Drought affects land surface dynamics, and the strongest *e authors declare that there are no conflicts of interest in impact of drought on vegetation occurs in arid and semiarid this paper. areas. *erefore, understanding the evolutionary charac- teristics and propagation of different drought types may Acknowledgments provide useful information about appropriate adaptation and mitigation strategies against the effects of drought on *is research was supported by the Strategic Priority Re- agricultural production and vegetation production and even search Program of the Chinese Academy of Sciences (Grant alleviate the losses caused by drought. Our results highlight nos. XDA19070502, XDA20100104, and XDA19040500), the the following points: National Natural Science Foundation of China (Grant nos. (1) More drought events occurred at shorter accumula- 41571516 and 41690144), and the Fundamental Research tion periods based on thresholds � −1.0, −1.5, and Funds for the Central Universities (Grant no. −2.0 for MD and HD. Compared to the Pre-R period, 2019jbkyjd013). 24 Advances in Meteorology [5] A. Dai, “Drought under global warming: a review,” Wiley Supplementary Materials Interdisciplinary Reviews: Climate Change, vol. 2, no. 1, pp. 45–65, 2011. Figure S1: boxplots showing MD peak and intensity in the [6] L. Wang and W. Chen, “A CMIP5 multimodel projection of UHRB ((a) and (c)) and in the MHRB ((b) and (d)) based future temperature, precipitation, and climatological drought on SPI using thresholds of −1, −1.5, and −2 for different in China,” International Journal of Climatology, vol. 34, no. 6, timescales at different periods (the dot and midline in the pp. 2059–2078, 2014. boxplots are the mean and median; the upper and lower [7] G. Leng, Q. Tang, and S. Rayburg, “Climate change impacts on boundaries of box are the interquartile ranges). Figure S2: meteorological, agricultural and hydrological droughts in boxplots showing HD peak and intensity at Ylx ((a) and (c)) China,” Global and Planetary Change, vol. 126, pp. 23–34, and at Zyx ((b) and (d)) based on SDI using thresholds of −1, −1.5, and −2 for different timescales at different periods [8] Z. Li, T. Zhou, X. Zhao et al., “Assessments of drought impacts (the dot and midline in the boxplots are the mean and on vegetation in China with the optimal time scales of the median; the upper and lower boundaries of box are the climatic drought index,” International Journal of Environ- interquartile ranges). Figure S3: the growing season spatial mental Research and Public Health, vol. 12, no. 7, pp. 7615– distribution of VCI during 2000–2001 in the UHRB and 7634, 2015. [9] J. Sheffield and E. F. Wood, “Projected changes in drought MHRB. Figure S4: heat map showing interannual Pearson’s occurrence under future global warming from multi-model, correlation coefficient of SPI accumulation periods of 1–48 multi-scenario, IPCC AR4 simulations,” Climate Dynamics, months with SDI-1 in the UHRB (Ylx) and MHRB (Zyx) vol. 31, no. 1, pp. 79–105, 2008. (note: the x-axis represents the correlation between the [10] H.-J. Xu, X.-P. Wang, C.-Y. Zhao, and X.-M. Yang, “Diverse SDI-1 and SPI at different timescales and the y-axis rep- responses of vegetation growth to meteorological drought resents the 12 months of the year; the up arrow represents across climate zones and land biomes in northern China from that the correlation coefficients are increased in the period 1981 to 2014,” Agricultural and Forest Meteorology, vol. 262, after EWDP in the UHRB (Ylx) and MHRB (Zyx) com- pp. 1–13, 2018. pared to the period before EWDP, or vice versa); (a) [11] M. Yu, Q. Li, M. J. Hayes, M. D. Svoboda, and R. R. Heim, (1965–1999, PC≥ 0.35, p< 0.05), (b) (2000–2012, PC≥ 0.57, “Are droughts becoming more frequent or severe in China p< 0.05) notes in the Ylx and (c) (1965–1999, PC≥ 0.35 based on the Standardized Precipitation Evapotranspiration Index: 1951–2010?” International Journal of Climatology, or≤ −0.38, p< 0.05), (d) (2000–2012, PC≥ 0.56, p< 0.05) vol. 34, no. 3, pp. 545–558, 2014. notes in the Zyx. Figure S5: heat maps showing PC between [12] L. Zhang, J. Xiao, Y. Zhou, Y. Zheng, J. Li, and H. Xiao, VCI lagged by 0–6 months and SPI (1–48)/SDI (1–12) “Drought events and their effects on vegetation productivity before (1965–1999) and after (2000–2012) EWDP; (a) in China,” Ecosphere, vol. 7, no. 12, Article ID e01591, 2016. (1965–1999, PC≥ 0.10, p< 0.05), (b) (2000–2012, PC≥ 0.16, [13] R. R. HeimJr., “Review of twentieth-century drought indices p< 0.05) in the UHRM (Ylx) and (c) (1965–1999, PC≥ 0.11 used in the United States,” Bulletin of the American Meteo- or PC≤ −0.11, p< 0.05), (d) (2000–2012, PC≥ 0.17, rological Society, vol. 83, pp. 1149–1166, 2002. p< 0.05) in the MHRB (Zyx). Figure S6: reservoirs under [14] T. J. Chang, “Investigation of precipitation droughts by use of construction and construction (a) irrigation system and kriging method,” Journal of Irrigation and Drainage Engi- area (b) in the MHRB. Table S1: identification of MD neering, vol. 117, no. 6, pp. 935–943, 1991. (threshold � −1, SPI-1) and HD (Zyx) characteristics [15] E. A. B. Eltahir, “Drought frequency analysis of annual rainfall series in central and western Sudan,” Hydrological Sciences (threshold � −1, SDI-1) based on run theory with com- Journal, vol. 37, no. 3, pp. 185–199, 1992. parison with historical drought records in MHRB. Table S2: [16] J. A. Dracup, K. S. Lee, and E. G. Paulson Jr., “On the statistical MD and HD drought characteristics for different timescales characteristics of drought events,” Water Resources Research, for different threshold for prereservoir period, post- vol. 16, no. 2, pp. 289–296, 1980. reservoir period, and total period in UHRB (Ylx) and [17] S. M. Quiring and S. Ganesh, “Evaluating the utility of the MHRB (Zyx). (Supplementary Materials) Vegetation Condition Index (VCI) for monitoring meteo- rological drought in Texas,” Agricultural and Forest Meteo- rology, vol. 150, no. 3, pp. 330–339, 2010. References [18] Q. Zhang, D. Kong, V. P. Singh, and P. Shi, “Response of vegetation to different time-scales drought across China: [1] S. E. Nicholson, C. J. Tucker, and M. B. Ba, “Desertification, spatiotemporal patterns, causes and implications,” Global and drought, and surface vegetation: an example from the west Planetary Change, vol. 152, pp. 1–11, 2017. african sahel,” Bulletin of the American Meteorological Society, [19] C. Li, W. Leal Filho, J. Yin et al., “Assessing vegetation re- vol. 79, no. 5, pp. 815–829, 1998. sponse to multi-time-scale drought across inner Mongolia [2] A. T. DeGaetano, “A temporal comparison of drought im- plateau,” Journal of Cleaner Production, vol. 179, pp. 210–216, pacts and responses in the New York city metropolitan area,” Climatic Change, vol. 42, no. 3, pp. 539–560, 1999. [20] Y. Zhang, J. Gao, L. Liu, Z. Wang, M. Ding, and X. Yang, [3] A. K. Mishra and V. P. Singh, “A review of drought concepts,” “NDVI-based vegetation changes and their responses to cli- Journal of Hydrology, vol. 391, no. 1-2, pp. 202–216, 2010. mate change from 1982 to 2011: a case study in the Koshi [4] S. Huang, P. Li, Q. Huang, G. Leng, B. Hou, and L. Ma, “*e River Basin in the middle Himalayas,” Global and Planetary propagation from meteorological to hydrological drought and Change, vol. 108, pp. 139–148, 2013. its potential influence factors,” Journal of Hydrology, vol. 547, [21] A. Zhao, A. Zhang, S. Cao, X. Liu, J. Liu, and D. Cheng, pp. 184–195, 2017. “Responses of vegetation productivity to multi-scale drought Advances in Meteorology 25 in Loess Plateau, China,” CATENA, vol. 163, pp. 165–171, basin, China,” Water Resources Management, vol. 28, no. 10, 2018. pp. 3095–3110, 2014. [22] B. Lloyd-Hughes, “*e impracticality of a universal drought [38] A. Mondal and P. P. Mujumdar, “Return levels of hydrologic definition,” ;eoretical and Applied Climatology, vol. 117, droughts under climate change,” Advances in Water Re- no. 3-4, pp. 607–611, 2014. sources, vol. 75, pp. 67–79, 2015. [23] L. J. Barker, J. Hannaford, A. Chiverton, and C. Svensson, [39] J. Niu, J. Chen, and L. Sun, “Exploration of drought evolution “From meteorological to hydrological drought using stand- using numerical simulations over the Xijiang (West River) ardised indicators,” Hydrology and Earth System Sciences, basin in South China,” Journal of Hydrology, vol. 526, vol. 20, no. 6, pp. 2483–2505, 2016. pp. 68–77, 2015. [24] B. T. Mckee, J. Nolan, and J. Kleist, “*e relationship of ´ [40] J. Lorenzo-Lacruz, S. Vicente-Serrano, J. Gonzalez-Hidalgo, drought frequency and duration to time scales,” in Pro- J. Lopez-Moreno, ´ and N. Cortesi, “Hydrological drought ceedings of the 8th Conference on Applied Climatology, response to meteorological drought in the Iberian Peninsula,” vol. 17–22, pp. 179–184, Anaheim, CA, USA, January 1993. Climate Research, vol. 58, no. 2, pp. 117–131, 2013. [25] D. Wang, M. Hejazi, X. Cai, and A. J. Valocchi, “Climate [41] S. M. Vicente-Serrano, C. Gouveia, J. J. Camarero et al., change impact on meteorological, agricultural, and hydro- “Response of vegetation to drought time-scales across global logical drought in central Illinois,” Water Resources Research, land biomes,” Proceedings of the National Academy of Sci- vol. 47, no. 9, 2011. ences, vol. 110, no. 1, pp. 52–57, 2013. [26] A. AghaKouchak and N. Nakhjiri, “A near real-time satellite- [42] A. F. Van Loon, M. H. J. Van Huijgevoort, and based global drought climate data record,” Environmental H. A. J. Van Lanen, “Evaluation of drought propagation in an Research Letters, vol. 7, no. 4, p. 044037, 2012. ensemble mean of large-scale hydrological models,” Hy- [27] J. H. Stagge, L. M. Tallaksen, L. Gudmundsson, drology and Earth System Sciences, vol. 16, no. 11, pp. 4057– A. F. Van Loon, and K. Stahl, “Candidate distributions for 4078, 2012. climatological drought indices (SPI and SPEI),” International [43] M. Peña-Gallardo, S. M. Vicente-Serrano, J. Hannaford et al., Journal of Climatology, vol. 35, no. 13, pp. 4027–4040, 2015. “Complex influences of meteorological drought time-scales [28] I. Nalbantis and G. Tsakiris, “Assessment of hydrological on hydrological droughts in natural basins of the contiguous drought revisited,” Water Resources Management, vol. 23, Unites States,” Journal of Hydrology, vol. 568, pp. 611–625, no. 5, pp. 881–897, 2009. 2019. [29] S. Li, L. Xiong, L. Dong, and J. Zhang, “Effects of the three [44] F. Ma, L. Luo, A. Ye, and Q. Duan, “Drought characteristics gorges reservoir on the hydrological droughts at the down- and propagation in the semiarid Heihe river basin in stream yichang station during 2003–2011,” Hydrological northwestern China,” Journal of Hydrometeorology, vol. 20, Processes, vol. 27, no. 26, pp. 3981–3993, 2013. no. 1, pp. 59–77, 2019. [30] H. Tabari, J. Nikbakht, and P. Hosseinzadeh Talaee, “Hy- [45] J. Niu, S. Kang, X. Zhang, and J. Fu, “Vulnerability analysis drological drought assessment in northwestern Iran based on based on drought and vegetation dynamics,” Ecological In- streamflow drought index (SDI),” Water Resources Man- dicators, vol. 105, pp. 329–336, 2019. agement, vol. 27, no. 1, pp. 137–151, 2013. [46] R. P. Pandey and K. S. Ramasastri, “Relationship between the [31] E. Rimkus, E. Stonevicius, ˇ V. Korneev, J. Kazys, ˇ common climatic parameters and average drought fre- G. Valiuˇskevicius, ˇ and A. Pakhomau, “Dynamics of meteo- quency,” Hydrological Processes, vol. 15, no. 6, pp. 1019–1032, rological and hydrological droughts in the Neman river ba- 2001. sin,” Environmental Research Letters, vol. 8, no. 4, p. 045014, [47] X. Li, G. Cheng, Y. Ge et al., “Hydrological cycle in the Heihe river basin and its implication for water resource management [32] T. Fischer, M. Gemmer, B. Su, and T. Scholten, “Hydrological in endorheic basins,” Journal of Geophysical Research: At- long-term dry and wet periods in the Xijiang River basin, mospheres, vol. 123, no. 2, pp. 890–914, 2018. [48] N. Zhao, T. Yue, C. Chen, M. Zhao, and Z. Fan, “An improved South China,” Hydrology and Earth System Sciences, vol. 17, no. 1, pp. 135–148, 2013. statistical downscaling scheme of Tropical Rainfall Measuring [33] X. Hong, S. Guo, Y. Zhou, and L. Xiong, “Uncertainties in Mission precipitation in the Heihe River basin, China,” In- assessing hydrological drought using streamflow drought ternational Journal of Climatology, vol. 38, no. 8, pp. 3309– index for the upper Yangtze River basin,” Stochastic Envi- 3322, 2018. ronmental Research and Risk Assessment, vol. 29, no. 4, [49] A. Zhang, C. Zheng, S. Wang, and Y. Yao, “Analysis of pp. 1235–1247, 2015. streamflow variations in the Heihe River Basin, northwest [34] A.-A. Belal, H. R. El-Ramady, E. S. Mohamed, and China: trends, abrupt changes, driving factors and ecological A. M. Saleh, “Drought risk assessment using remote sensing influences,” Journal of Hydrology: Regional Studies, vol. 3, and GIS techniques,” Arabian Journal of Geosciences, vol. 7, pp. 106–124, 2015. no. 1, pp. 35–53, 2014. [50] J. Niu, X.-G. Zhu, M. A. J. Parry et al., “Environmental [35] Q. Lin, Z. Wu, V. P. Singh, S. H. R. Sadeghi, H. He, and G. Lu, burdens of groundwater extraction for irrigation over an “Correlation between hydrological drought, climatic factors, inland river basin in Northwest China,” Journal of Cleaner reservoir operation, and vegetation cover in the Xijiang Basin, Production, vol. 222, pp. 182–192, 2019. South China,” Journal of Hydrology, vol. 549, pp. 512–524, [51] Y. Ge, X. Li, C. Huang, and Z. Nan, “A decision support 2017. system for irrigation water allocation along the middle reaches [36] I. Cordery and M. McCall, “A model for forecasting drought of the Heihe river basin, northwest China,” Environmental from teleconnections,” Water Resources Research, vol. 36, Modelling & Software, vol. 47, pp. 182–192, 2013. no. 3, pp. 763–768, 2000. [52] L. Qiu, D. Peng, Z. Xu, and W. Liu, “Identification of the [37] S. Huang, J. Chang, Q. Huang, and Y. Chen, “Spatio-temporal impacts of climate changes and human activities on runoff in changes and frequency analysis of drought in the wei river the upper and middle reaches of the Heihe River basin, 26 Advances in Meteorology China,” Journal of Water and Climate Change, vol. 7, no. 1, [69] J. I. Lopez-Moreno, ´ S. M. Vicente-Serrano, J. Zabalza et al., “Hydrological response to climate variability at different time pp. 251–262, 2015. [53] X. Deng and C. Zhao, “Identification of water scarcity and scales: a study in the Ebro basin,” Journal of Hydrology, vol. 477, pp. 175–188, 2013. providing solutions for adapting to climate changes in the Heihe river basin of China,” Advances in Meteorology, [70] X.-P. Deng, L. Shan, H. Zhang, and N. C. Turner, “Improving agricultural water use efficiency in arid and semiarid areas of vol. 2015, Article ID 279173, 13 pages, 2015. [54] M. Zhang, S. Wang, B. Fu, G. Gao, and Q. Shen, “Ecological China,” Agricultural Water Management, vol. 80, no. 1–3, pp. 23–40, 2006. effects and potential risks of the water diversion project in the [71] X. Xu, Y. Jiang, M. Liu, Q. Huang, and G. Huang, “Modeling Heihe River Basin,” Science of the Total Environment, vol. 619- and assessing agro-hydrological processes and irrigation 620, pp. 794–803, 2018. water saving in the middle Heihe River basin,” Agricultural [55] A. A. Paulo, L. S. Pereira, and P. G. Matias, “Analysis of local Water Management, vol. 211, pp. 152–164, 2019. and regional droughts in southern Portugal using the theory [72] J. Niu, J. Chen, K. Wang, and B. Sivakumar, “Multi-scale of runs and the standardised precipitation index,” in Tools for streamflow variability responses to precipitation over the Drought Mitigation in Mediterranean Regions, G. Rossi, headwater catchments in southern China,” Journal of Hy- A. Cancelliere, L. S. Pereira, T. Oweis, M. Shatanawi, and drology, vol. 551, pp. 14–28, 2017. A. Zairi, Eds., Springer, Dordrecht, Netherlands, 2003. [73] H. Liu, M. Zhang, Z. Lin, and X. Xu, “Spatial heterogeneity of [56] F. Kogan and J. Sullivan, “Development of global drought- the relationship between vegetation dynamics and climate watch system using NOAA/AVHRR data,” Advances in Space change and their driving forces at multiple time scales in Research, vol. 13, no. 5, pp. 219–222, 1993. Southwest China,” Agricultural and Forest Meteorology, [57] J. Spinoni, T. Antofie, P. Barbosa et al., “An overview of vol. 256-257, pp. 10–21, 2018. drought events in the Carpathian Region in 1961–2010,” [74] R. Orth and G. Destouni, “Drought reduces blue-water fluxes Advances in Science and Research, vol. 10, no. 1, pp. 21–32, more strongly than green-water fluxes in Europe,” Nature Communications, vol. 9, no. 1, p. 3602, 2018. [58] V. M. Yevjevich, An Objective Approach to Definitions and [75] N. S. Diffenbaugh, D. L. Swain, and D. Touma, “Anthropo- Investigations of Continental Hydrologic Droughts, Colorado genic warming has increased drought risk in California,” State University, Fort Collins, CO, USA, 1967. Proceedings of the National Academy of Sciences, vol. 112, [59] Y. L. Zhou, P. Zhou, J. L. Jin, and J. Li, “Establishment of no. 13, pp. 3931–3936, 2015. hydrological drought index on sources of regional water [76] P. Droogers, W. G. M. Bastiaanssen, M. Beyazgul, ¨ Y. Kayam, supply and its application to drought frequency analysis for G. W. Kite, and H. Murray-Rust, “Distributed agro-hydro- Kunming,” Journal of Hydraulic Engineering, vol. 45, logical modeling of an irrigation system in western Turkey,” pp. 1038–1047, 2014. Agricultural Water Management, vol. 43, no. 2, pp. 183–202, [60] J. Wu, Z. Liu, H. Yao et al., “Impacts of reservoir operations on multi-scale correlations between hydrological drought and [77] Y. Chen, J. Niu, S. Kang, and X. Zhang, “Effects of irrigation meteorological drought,” Journal of Hydrology, vol. 563, on water and energy balances in the Heihe River basin using pp. 726–736, 2018. VIC model under different irrigation scenarios,” Science of [61] S. Mitra and P. Srivastava, “Spatiotemporal variability of ;e Total Environment, vol. 645, pp. 1183–1193, 2018. meteorological droughts in southeastern USA,” Natural [78] J. Niu, Q. Liu, S. Kang, and X. Zhang, “*e response of crop Hazards, vol. 86, no. 3, pp. 1007–1038, 2017. water productivity to climatic variation in the upper-middle [62] L. Gao and Y. Zhang, “Spatio-temporal variation of hydro- reaches of the Heihe River basin, Northwest China,” Journal of logical drought under climate change during the period Hydrology, vol. 563, pp. 909–926, 2018. 1960–2013 in the Hexi Corridor, China,” Journal of Arid [79] B. I. Cook, R. L. Miller, and R. Seager, “Amplification of the Land, vol. 8, no. 2, pp. 157–171, 2016. North American “Dust Bowl” drought through human-in- [63] L. Yang, Q. Feng, Z. Yin et al., “Identifying separate impacts of duced land degradation,” Proceedings of the National Acad- climate and land use/cover change on hydrological processes emy of Sciences, vol. 106, no. 13, pp. 4997–5001, 2009. in upper stream of Heihe River, Northwest China,” Hydro- [80] Q. Huang, Z. Sun, C. Opp, T. Lotz, J. Jiang, and X. Lai, logical Processes, vol. 31, no. 5, pp. 1100–1112, 2017. “Hydrological drought at dongting lake: its detection, char- [64] Z. Cong, M. Shahid, D. Zhang, H. Lei, and D. Yang, “At- acterization, and challenges associated with three gorges dam tribution of runoff change in the alpine basin: a case study of in central yangtze, China,” Water Resources Management, the Heihe Upstream Basin, China,” Hydrological Sciences vol. 28, no. 15, pp. 5377–5388, 2014. Journal, vol. 62, no. 6, pp. 1013–1028, 2017. [81] S. Rangecroft, A. F. Van Loon, H. Maureira, K. Verbist, and [65] Q. Cheng, X. Zuo, F. Zhong, L. Gao, and S. Xiao, “Runoff D. M. Hannah, “Multi-method assessment of reservoir effects variation characteristics, association with large-scale circu- on hydrological droughts in an arid region,” Earth System lation and dominant causes in the Heihe River Basin, Dynamics Discussion, pp. 1–32, 2016. Northwest China,” Science of the Total Environment, vol. 688, [82] M. Trnka, M. Hayes, F. Jurecka ˇ et al., “Priority questions in pp. 361–379, 2019. multidisciplinary drought research,” Climate Research, [66] K. G. Wen, China Meteorological Disaster (Gansu Volume), vol. 75, no. 3, pp. 241–260, 2018. China Meteorological, Beijing, China, 2006, In Chinese. [67] A. F. Van Loon, T. Gleeson, J. Clark et al., “Drought in the anthropocene,” Nature Geoscience, vol. 9, no. 2, pp. 89–91, [68] G. Zhu, Y. Su, C. Huang, Q. Feng, and Z. Liu, “Hydro- geochemical processes in the groundwater environment of Heihe River Basin, northwest China,” Environmental Earth Sciences, vol. 60, no. 1, pp. 139–153, 2010. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Advances in Meteorology Hindawi Publishing Corporation

Meteorological Drought, Hydrological Drought, and NDVI in the Heihe River Basin, Northwest China: Evolution and Propagation

Loading next page...
 
/lp/hindawi-publishing-corporation/meteorological-drought-hydrological-drought-and-ndvi-in-the-heihe-8M2tjW5ipo

References

References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.

Publisher
Hindawi Publishing Corporation
Copyright
Copyright © 2020 Fanglei Zhong et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
ISSN
1687-9309
eISSN
1687-9317
DOI
10.1155/2020/2409068
Publisher site
See Article on Publisher Site

Abstract

Hindawi Advances in Meteorology Volume 2020, Article ID 2409068, 26 pages https://doi.org/10.1155/2020/2409068 Research Article Meteorological Drought, Hydrological Drought, and NDVI in the Heihe River Basin, Northwest China: Evolution and Propagation 1 1,2 3 Fanglei Zhong , Qingping Cheng , and Ping Wang Northwest Institute of Eco-Environment and Resources, Chinese Academy of Sciences, Lanzhou 730000, China University of the Chinese Academy of Sciences, Beijing 100039, China School of Tourism and Geographical Sciences, Yunnan Normal University, Kunming 650500, China Correspondence should be addressed to Qingping Cheng; qpchengtyli@foxmail.com Received 24 October 2019; Revised 10 February 2020; Accepted 22 February 2020; Published 9 June 2020 Academic Editor: Budong Qian Copyright © 2020 Fanglei Zhong et al. *is is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. Understanding the evolution and propagation of different drought types is crucial to reduce drought hazards in arid and semiarid regions. Here, Standardized Precipitation Index (SPI), Streamflow Drought Index (SDI), and Vegetation Condition Index (VCI) were used to investigate the spatiotemporal variation of different drought types and correlations between Pre (Pre-R)/post (Pos-R)- reservoir. Results showed that the average peak/intensity/duration/severity of meteorological droughts (MD) were greater in the Pre- R than in the Pos-R period in the upstream Heihe River Basin (UHRB), while there was little change between the Pre-R and Pos-R periods in the midstream Heihe River Basin (MHRB). *e average peak/intensity/duration/severity of hydrological drought (HD) decreased in the mainstream for Yingluoxia (Ylx) but increased for Zhengyixia (Zyx) station in the Pos-R period. Propagation time decreased by 3 months (negative effect) in Ylx and increased by 8 months (positive effect) in Zyx compared with the Pre-R period. In the Pos-R period, propagation time increased (1–3 months) for tributaries (positive effect). Propagation times for the mainstream and tributaries varied for different seasons and time periods. Pearson’s correlation coefficient values were lower at short timescales (1–3 months) but higher at long timescales for the Pos-R period in Ylx and Zyx for SDI-1 with different timescales of SPI. *e SDI and SPI had no lag in the UHRB and MHRB. However, VCI with SPI had a significant lag correlation at short timescales in the UHRB (lag 6 months) and MHRB (lag 4 months), and the VCI with SDI had a significant lag correlation for 1 month in the MHRB. *e propagation time from MD to HD has been reduced for Pos-R in the UHRB. *ere was a positive effect (prolonged MD propagation HD time) in Pos-R but still faces serious drought stress in the MHRB. and central portions of northern China since the late 1. Introduction 1990s [11, 12]. Drought is recognized as the world’s most costly and In general, drought can be classified into four types: pressing natural hazard, giving rise to significant losses in meteorological, hydrological, agricultural, and socioeco- the fields of the economy, ecology, and environment (e.g., nomic [13]. Meteorological droughts (MD) are the water crop losses, degradation and desertification, urban water shortages caused by an imbalance in precipitation and supply shortages, and forest fires) [1–4]. *e damage from evaporation; precipitation is commonly used for MD droughts is expected to increase in severity [5]. Compared analysis [14, 15]. Hydrological droughts (HD) are related to a with other natural hazards, the spatial extent of drought is period with inadequate surface and subsurface water re- very large and its time of influence is commonly much sources for established water uses in a given water resource longer. Under global warming, more frequent and severe management system. Streamflow data are widely applied for droughts have been projected in the 21st century [6–8], HD analysis [14, 16]. Because the influence of drought on particularly in the mid-latitudes [9, 10]. Severe and ex- terrestrial ecosystems is becoming increasingly acute, the treme droughts have become more frequent in the eastern vegetation response to drought is a crucial topic in the 2 Advances in Meteorology intensity, severity, and peak, while detecting the spatio- domain of climate research [17–19]. *e Normalized Dif- ference Vegetation Index (NDVI) is a good measure to temporal characteristics of VCI; (2) to investigate VCI and HD propagation times based on MD at different timescales estimate green biomass, leaf area index, and patterns of productivity and has been widely used to assess vegetation and VCI propagation times based on HD at different degradation, ecosystem features, and the physiological timescales; and (3) to explore the potential factors influ- drought conditions of vegetation [20, 21]. encing drought characteristics and propagation times. Drought indices are the best ways to monitor drought and drought events at present. Lloyd-Hughes [22] counted 2. Materials and Methods over 100 drought indicators that have been developed for different types of drought [23]. A commonly used index is 2.1. Study Area. *e Heihe River Basin (HRB) is the second the Standardized Precipitation Index (SPI) [24]. Its calcu- largest inland river basin in the arid zone of northwest lation is simple because it only needs precipitation as an China. *e HRB flows through Qinghai Province, Gansu input, and it can be calculated over different timescales, Province, and the Inner Mongolia Autonomous Region. *e which allows SPI to monitor both short-term droughts, such HRB is located along the land Silk Roads (“One Belt, One ° ° ° ° as agricultural droughts, and long-term droughts, such as Road”) between 97.1 E–102.0 E and 37.7 N–42.7 N with a 5 2 HD [3, 25–27]. For HD index, the Streamflow Drought total area of approximately 1.43 ×10 km [47, 48]. *e HRB Index (SDI) proposed by Nalbantis and Tsakiris [28] includes three sections from south to north: the upstream overcomes the problems of predicting drought onset and Heihe River Basin (UHRB) from the Qilian Mountains to duration using cumulative streamflow volumes and the areal Yingluoxia station (outlet of the mountains) belongs to the extent of drought using a spatially integrated streamflow at a subhumid and semiarid temperate continental monsoon basin outlet; it is also widely used in different regions of the climate zone; the midstream Heihe River Basin (MHRB) world [28–33]. *e Vegetation Condition Index (VCI) is a running from Yingluoxia station to Zhengyixia station be- normalization of the NDVI, which filters out the contri- longs to the arid climate zone; and downstream HRB ter- bution of local geographic resources to the spatial variability minates in the Juyan Lakes (east and west branches, of the NDVI [17, 34]. It reflects agricultural (vegetation) respectively) [49]. From upstream to downstream, the ele- drought accurately in many parts of the world and has been vation decreases, the water availability decreases, and the applied in daily drought monitoring by the National Oceanic landscape changes from glaciers and alpine biomes in the and Atmospheric Administration of the United States UHRB to steppes and agricultural ecosystems in the MHRB (NOAA) and the National Satellite Meteorological Centre of to riparian ecosystems and vast areas of desert in the China [35]. downstream area [47]. In this study, the UHRB and MHRB Many researchers in related fields have focused on were selected as the study area. Figure 1 shows the study area drought evolution, drought prediction, and developing re- and the hydrometeorological stations. Analyses were also liable drought indices using observed and simulated data conducted at the sub-basin level in the UHRB and MHRB. In [4, 36–39]. Recently, studying the propagation time of VCI the MHRB, the annual average precipitation, potential and HD based on MD using Pearson’s or Spearman’s evapotranspiration, and streamflow were 226 mm, 924 mm, correlation coefficients has become a research focus and 88 mm, respectively. Over 90% of the population, grain [8, 10, 18, 19, 21, 23, 40–46]. *ese two kinds of droughts can production, and major industries are concentrated in the reflect the different stages of drought development. Com- MHRB. Approximately 84% of the total available water was monly, MD develops and ends relatively quickly, while HD is consumed for irrigation. On average, a decline in ground- the result of MD [4]. *e corresponding propagation time water of about 1.86 m is attributable to water consumption depends on local landscape conditions [46]. *erefore, in- for irrigation of the farmland area in Heihe River from 1981 vestigating the propagation time from MD to HD [4] and to 2010 [50] with demand constantly increasing [51, 52]. from MD to VCI along with its potential influencing factors *ere are a series of reservoirs (one large reservoir, 9 me- is necessary for establishing an effective monitoring and dium size reservoirs, and 89 small reservoirs) and a relatively warning system for HD and VCI based on MD. complete irrigation system consisting of more than 893 main However, little research has been done into the propa- canals and branch canals [53, 54]. Land use and land cover gation of MD to HD and VCI and HD propagation to VCI, have also significantly changed in the HRB. Urban, culti- in particular in subarid and arid watersheds, which is one of vated, and forest land have increased, while water area and the aims of this paper. *e investigation of VCI and HD grassland have become increasingly degraded. propagation time from MD at different timescales under the influence of global warming and human activities (e.g., construction of reservoirs or dams, land use, and land cover 2.2. Data. Daily precipitation datasets from 1961 to 2012 change) will help to elucidate the impacts of droughts on were collected by five meteorological stations (Table 1) and terrestrial ecosystems and inform planning and optimizing derived from the China Meteorological Data Sharing Service water resource allocations during droughts. It is important System of National Meteorological Information Centre V3.0 to identify the drought propagation process and mecha- (http://www.nmic.gov.cn/). Rigorous quality control was nisms to help establish a drought early warning system. *e conducted by the China National Meteorological Infor- primary objectives of this study are as follows: (1) to inspect mation Centre before the data were released. Monthly ob- the spatiotemporal characteristics of MD and HD duration, served streamflow data, which were also of high quality, for Advances in Meteorology 3 98°E 100°E 102°E 98°E 100°E N N 40°N 40°N 38°N 38°N 38°N Elevation (m) 38°N High : 5297 0 60 120 0 60 120 Kilometers Kilometers Low : 1277 98°E 100°E 102°E 98°E 100°E 102°E Hydrological station LUCC Meteorology station Forest Built-up area River Grassland Unused land Water area Cultivated land (a) (b) 98°E 100°E 38°N 38°N 0 60 120 Kilometers 98°E 100°E Coniferous forest Marshy grassland Alpine vegetation Swamp Cultivated vegetation Glacial snow and salinized land Shrubs Residential land Desert River system Grassland (c) Figure 1: *e upstream and midstream in the Heihe River Basin with meteorological and hydrological stations and digital elevation map (a), land use and land cover (LULC) in 2011 (b), and vegetation types (c). Table 1: Information on the meteorological stations. ° ° Sub-basin Station Longitude ( E) Latitude ( N) Elevation (m) Data period ° ° Qilian (Ql) 100.24 38.19 2787.4 1961–2012 UHRB ° ° Yeniugou (Yng) 99.58 38.41 3286.0 1961–2012 ° ° Zhangye (Zy) 100.46 38.91 1482.7 1961–2012 ° ° MHRB Shandan (Sd) 101.08 38.77 1764.6 1961–2012 ° ° Gaotai (Gt) 99.79 39.36 1332.2 1961–2012 eight hydrological stations (Table 2) were collected from the SPOT/VEGETATION NDVI and MODIS from 1999 to Hydrological Bureau of Gansu Province. Yingluoxia (Ylx) 2012. *e data were obtained from the resource and envi- and Zhengyixia (Zyx) stations were located in the main ronmental science data center of the Chinese Academy of stream of the HRB. *e monthly NDVI (1 km spatial res- Sciences (http://www.resdc.cn), and LULC data in the olution) dataset was the satellite remote sensing data of UHRB (2000 and 2010) and MHRB (1975, 1985, 1995, 2000, 4 Advances in Meteorology Table 2: Information on the hydrological stations. Longitude Latitude Catchment area Elevation Data Abrupt Sub-basin Station River ° ° ( E) ( N) (km ) (m) period year Yingluoxia (Ylx) Heihe River 100.18 38.82 10009 1700 1945–2012 1979 Hongshui Shuangshusi (Sss) 100.83 38.32 578 2490 1957–2012 1974 River UHRB Wafangcheng Dazhuma 100.48 38.43 334 2440 1957–2012 1985 (Wfc) River Liyuanbao (Lyb) Liyuan River 100.00 38.97 2641 1760 1956–2012 1976 Zhengyixia (Zyx) Heihe River 99.47 39.82 35634 1280 1957–2012 1984 Lijiaqiao (Ljq) Maying River 101.13 38.52 1205 2150 1957–2012 1986 MHRB Haichaoba Shunhua (Sh) 100.72 38.59 191.8 1948 1957–2012 1995 River Suyoukou (Syk) Suyou River 100.58 38.79 473.4 1565 1957–2012 1971 Abrupt change year was detected by the Pettitt test. Table 3: Classification of SPI, SDI, and VCI. and 2010) were obtained from the Cold and Arid Regions Science Data Centre (http://westdc.westgis.ac.cn/). *e Value Drought grades historical disaster records were derived from the China SPI/SDI≤ −2.0, VCI< 15 Extreme drought Meteorological Disaster Yearbook (Gansu volume). Zhang −2.0< SPI/SDI≤ −1.5, 15≤ VCI< 30 Severe drought et al. [49] used Pettitt tests to determine abrupt change −1.5< SPI/SDI≤ −1.0, 30≤ VCI< 45 Moderate drought points in the annual streamflow series from 1945 to 2012 and −1.0< SPI/SDI≤ 0, 45≤ VCI< 60 Mild drought from 1957 to 2012 at the Ylx and Zyx stations to ascertain the impact of the reservoir activities in the HRB. *e results SPI. It is worth noting that the procedure should be per- indicated only one significant abrupt change point in 1979 at formed separately for each month and can be calculated for Ylx station and in 1984 at Zyx station [49]. *e impact of any timescale (e.g., 1, 3, 6, 9, and 12 months). *e total human activities is more evident after the Chinese Gov- streamflow X in a given month j and year i depends on the ernment initiated the Ecological Water Diversion Project i,j chosen timescale k and can be calculated as [55] (EWDP) in mainstream Heihe River for 2000 [54]. *ere- fore, in this paper, we focused on the mainstream Ylx and X � 􏽘 V + 􏽘 V , if j< k; Zyx stations, between the Pre-R and Pos-R periods, before i,j i−1,l i,l l�13−k+l i�1 and after EWDP. *e abrupt change points for the (1) tributaries for six stations are shown in Table 2 (Pettitt tests). X � 􏽘 V , if j≥ k, i,j i,l i�j−k+l 2.3. Methods where V and V denote streamflow volumes in years i − 1 i-1,l i,l 2.3.1. Standardized Precipitation Index (SPI). *e SPI index and i, respectively. Due to the three-parameter log-logistic was developed by McKee et al. [24]. Within a certain distribution used to calculate the SPI, we also chose this geographic area, the precipitation usually fluctuates regu- probability distribution to standardize the observed larly. If the precipitation is less than the average annual streamflow data used to get the SDI to facilitate comparison precipitation, a drought may occur. If precipitation exceeds in the HRB. It is emphasized here that positive SDI values the annual average, flooding may occur. To calculate the SPI, reflect wet conditions, while negative values indicate a hy- a frequency distribution function is first constructed from a drological drought. series of long-term precipitation observations. In this paper, a three-parameter log-logistic distribution was used because the frequencies of precipitation accumulated at different 2.3.3. Vegetation Condition Index (VCI). Kogan and Sulli- van [56] proposed the VCI index in 1993. It is here assumed timescales are well modeled using these statistical distri- butions. Because SPI is based on the cumulative probability that the change of the VCI index is only affected by weather factors; in this way, the ecosystem factors and extreme of a given timescale, here the total amounts of precipitation in the current month and previous i months (i � 1, 2, 3, ......) weather factors are separated, the impact of extreme weather on vegetation is evaluated, and agricultural drought is were used to calculate SPI on a timescale of i + 1 month. In our study, we used the SPI (1, 3, 6, 9, ......, 48 months) to monitored. A smaller VCI index correlates to a greater analyze the characteristics of drought change in the HRB. drought. *e VCI index is calculated as follows: For drought classifications, please see Table 3. NDVI − NDVI J min VCI � × 100, (2) NDVI − NDVI max min 2.3.2. Streamflow Drought Index (SDI). *e SDI was de- veloped by Nalbantis and Tsakiris [28] to characterize hy- where NDVI, NDVI , and NDVI are monthly NDVI, max min drological drought based on developing the concepts of the multiyear maximum NDVI, and multiyear minimum NDVI Advances in Meteorology 5 (in this study, 14 years), respectively, for each grid cell. *e D: duration, S: severity, P: peak, D : nondrought duration, T : initiation time, T : termination time, and I: intensity VCI changes from 0 to 100 corresponding to changes in i e 2.5 vegetation condition from extremely unfavorable to optimal. 2.0 In the case of an extremely dry month, the vegetation condition is poor, and the VCI is close to or equal to zero. A 1.5 VCI of 50 reflects fair vegetation conditions. At optimal D = d + d + d 1.0 2 0 1 2 vegetation conditions, the VCI is close to 100 [34]. 0.5 d d d According to t-test, the trend was divided into extremely D 0 2 0.0 significant increase (slope> 0, p< 0.01), significant increase S S 1 T 2 (slope> 0, 0.01≤ p< 0.05), insignificant change (p≥ 0.05), T –0.5 e extremely significant decrease (slope< 0, p< 0.01), and S –1.0 significant decrease (slope< 0, 0.01≤ p< 0.05). –1.5 –2.0 2.4. Drought Variables and Run ;eory. According to McKee S = s + s 2 1 2 –2.5 et al. [24] and Spinoni et al. [57], the derived drought variables based on run theory [58] follow certain definitions –3.0 (Figure 2). When the monthly SPI/SDI values were below Figure 2: Definition of drought characteristics including peak, −1, the corresponding month was potentially identified as a intensity, duration, and magnitude using the SPI/SDI and run drought event (e.g., the four drought events for D , D , d , 0 1 0 theory. and d shown in Figure 2). *resholds of −1 (moderate drought), −1.5 (severe drought), and −2 (extreme drought) were used to identify drought events and drought charac- the cumulative water balance occurring from January to teristics. Drought duration was defined as the number of March [19, 21, 23, 41]. Meanwhile, to determine whether months from the first month in which the indicator goes there is a lag between the SPI (accumulation periods of 1–48 lower than −1 to the last month with a negative value before months) and SDI-1, cross-correlations were calculated for the indicator returns to positive values. If the duration of a SDI-1 series which were lagged by 0 to 6 months after the SPI drought event only contained one month where SPI/ series. In this case, the SPI accumulation period with the SDI< −1, this month was regarded as a single drought event strongest correlation with SDI-1 was denoted as the lagged (e.g., D with a magnitude S ). Drought severity was defined 1 1 SPI-n (1, 3, 6, 9, ....., 48 months) [23]. *e same interpre- as the sum of the monthly absolute values of the index when tation applies to the lag correlation of the VCI with different the index was≤−1.0 (−1.5, −2). A drought event may contain timescales SPI and VCI with different timescales SDI. a few consecutive months with negative SPI/SDI (e.g., du- ration D and its magnitude S ). If the duration of a drought 0 0 3. Results event contained two branches, such as d and d , and the 0 2 interruption period d between d and d was less than 6 1 0 2 3.1. MD and HD for the Mainstream Characteristics between months in which 0< SPI/SDI< 1 [59, 60], these months were Different Timescales and ;resholds. *e distribution of still regarded as a single drought event (D � d + d + d ). 2 0 1 2 drought characteristics based on thresholds of −1, −1.5, and *e corresponding magnitude was then defined as −2 for different timescales at different periods, including S � s + s [59, 60]. Intensity stands for the number of 2 1 2 Pre-R (1965–abrupt change year), Pos-R (abrupt change year months in which the drought indicator was lower than −1 +1–2012), and total period (1965–2012), was evaluated in the (−1.5, −2). Drought peak referred to the month in the UHRB and MHRB using boxplots. *e drought charac- “drought event” with the lowest value of the indicator [61]. teristics of SPI-1 and SDI-1 in the MHRB (Table S1) were almost consistent compared with historical records and, for 2.5. Correlation Analysis. Pearson’s correlation coefficient different combinations of timescales, thresholds and periods (PC) was calculated between SDI-1 (one month SDI) and SPI (Table S2) are shown in the supporting information. *e (1, 3, 6, 9, ....., 48 months), monthly (April–October) VCI, drought peak and intensity (Figures S1-S2), duration, and and different timescales of SPI and SDI (1, 3, 6, 9, and 12 severity for different timescales at different thresholds are months) series indicates the characteristics of interannual shown in Figure 3 for MD and in Figure 4 for HD. More propagation at 95% and 99% confidence levels. *e PC drought events were identified at shorter accumulation periods (1–12 months for MD and 1–6 months for HD) ranges from −1 to 1; a PC> 0 indicates a positive correlation and a PC< 0 indicates a negative correlation. In this study, based on thresholds equal to −1.0, −1.5, and −2.0 in the the PC method was also used to detect the intra-annual UHRB (Ylx) and MHRB (Zyx). For MD, the average in- propagation time of the SDI-1 to the different timescales of tensity, duration, and severity of fluctuations increased with SPI and monthly VCI to the different timescales of SDI and cumulative timescale increases between Pre-R and Pos-R SPI conditions. For instance, if the maximum correlation in periods. *e changes were more obvious in Pos-R compared the March SDI and the March series of SPI at the timescales with Pre-R periods for all thresholds in the UHRB. *e of 1, 3, . . ., 48 months is recorded at the 3-month timescale, average peak, intensity, duration, and severity of the MD did this means that the SDI in March is mostly determined by not change much for Pre-R, whereas Pos-R fluctuations SDI/SPI 6 Advances in Meteorology Pre-R Pos-R Total period 80 105 SPI = –1.0 1 3 6 9 12 15 18 21 24 36 48 1 3 6 9 12 15 18 21 24 36 48 1 3 6 9 12 15 18 21 24 36 48 100 100 80 80 60 60 SPI = –1.5 40 40 20 15 20 0 0 1 3 6 9 12 15 18 21 24 36 48 1 3 6 9 12 15 18 21 24 36 48 1 3 6 9 12 15 18 21 24 36 48 100 40 100 80 80 60 60 SPI = –2.0 40 40 20 20 0 0 1 369 12 15 18 21 24 36 48 1 3 6 9 1215182124 1 3 6 9 12 15 18 21 24 36 48 Timescale (SPI) Timescale (SPI) Timescale (SPI) (a) Pre-R Pos-R Total period 50 60 40 45 SPI = –1.0 30 20 20 0 0 1 3 6 9 12 15 18 21 24 36 48 1 369 12 15 18 21 24 36 48 1 369 12 15 18 21 24 36 48 70 50 SPI = –1.5 10 15 0 0 1 3 6 9 12151821243648 1 3 6 9 12151821243648 1 3 6 9 12 15 18 21 24 36 48 50 80 SPI = –2.0 15 20 0 0 1 3 9 12 18 21 24 48 1 3 6 9 12151821243648 1 3 6 9 12151821243648 Timescale (SPI) Timescale (SPI) Timescale (SPI) (b) Figure 3: Continued. Duration (months) Duration (months) Duration (months) Duration (months) Duration (months) Duration (months) Advances in Meteorology 7 Pre-R Pos-R Total period 150 160 SPI = –1.0 0 0 1 3 6 9 12 15 18 21 24 36 48 1 3 6 9 12151821243648 1 3 6 9 12151821243648 100 100 80 80 60 60 SPI = –1.5 30 40 40 20 20 0 0 1 3 6 9 12 15 18 21 24 36 48 1 36 9 12151821243648 1 3 6 9 12151821243648 80 80 70 70 60 60 50 50 40 40 SPI = –2.0 20 30 30 20 20 10 10 0 0 0 1 3 6 9 12 15 18 21 24 36 48 1 3 6 912 15 18 21 24 1 3 6 9 121518212436 48 Timescale (SPI) Timescale (SPI) Timescale (SPI) (c) Pre-R Pos-R Total period 80 80 70 70 60 60 50 50 40 40 SPI = –1.0 40 30 30 20 20 10 10 0 0 1 369 12 15 18 21 24 36 48 1 369 12 15 18 21 24 36 48 1 369 12 15 18 21 24 36 48 80 70 SPI = –1.5 40 20 20 0 0 1 3 6 9 12151821243648 1 3 6 9 12151821243648 1 3 6 9 12 15 18 21 24 36 48 80 80 80 70 70 60 60 50 50 40 40 SPI = –2.0 30 30 20 20 10 10 0 0 1 3 912 18 21 24 48 1 3 6 9 12151821243648 1 3 6 9 12151821243648 Timescale (SPI) Timescale (SPI) Timescale (SPI) (d) Figure 3: Boxplots showing MD duration and severity in the UHRB (a, c) and in the MHRB (b, d) based on SPI using thresholds of −1, −1.5, and −2 for different timescales at different periods (the dot and midline in the boxplots are the mean and median; the upper and lower boundaries of box are the interquartile range). increased with cumulative timescale increases using larger for HD than for MD because a short-term MD may thresholds of −1.0, −1.5, and −2.0. *e average peak, in- not develop into an HD [42] and as the number of events tensity, duration, and severity of value changed little in the decreases, duration lengthens, and severity worsens. *e MHRB. With regard to HD, the duration and severity were average intensity (Figure S2), duration, and severity of the Severity Severity Severity Severity Severity Severity 8 Advances in Meteorology Pre-R Pos-R Total period 40 140 100 30 SDI = –1.0 60 1 39 6 12 1 369 12 1 369 12 80 80 60 20 SDI = –1.5 15 1 39 6 12 1 369 12 1 369 12 SDI = –2.0 15 1 369 12 1 39 6 12 1 369 Timescale (SDI) Timescale (SDI) Timescale (SDI) (a) Pre-R Pos-R Total period 25 100 100 80 80 60 60 SDI = –1.0 40 40 20 20 0 0 1 39 6 12 1 369 12 1 369 12 100 100 80 80 60 60 SDI = –1.5 6 40 40 20 20 0 0 1 36 1 369 12 1 369 12 100 100 80 80 60 60 SDI = –2.0 40 40 20 20 0 0 1 3 1 369 12 1 369 12 Timescale (SDI) Timescale (SDI) Timescale (SDI) (b) Figure 4: Continued. Duration (months) Duration (months) Duration (months) Duration (months) Duration (months) Duration (months) Advances in Meteorology 9 Pre-R Pos-R Total period 60 25 20 50 SDI = –1.0 15 0 0 1 39 6 12 1 369 12 1 369 12 30 75 SDI = –1.5 1 39 6 12 1 369 12 1 369 12 SDI = –2.0 8 30 30 1 369 1 369 12 1 39 6 12 Timescale (SDI) Timescale (SDI) Timescale (SDI) (c) Pre-R Pos-R Total period 16 140 SDI = –1.0 4 60 1 3966 12 1 319 2 1 369 12 SDI = –1.5 3 50 1 36 1 31 6 9 2 1 369 12 140 140 120 120 100 100 80 80 SDI = –2.0 60 60 40 40 20 20 0 0 1 3 1 31 6 9 2 1 369 12 Timescale (SDI) Timescale (SDI) Timescale (SDI) (d) Figure 4: Boxplots showing HD duration and severity for Ylx (a, c) and for Zyx (b, d) based on SDI using thresholds of −1, −1.5, and −2 for different timescales at different periods (the dot and midline in the boxplots are the mean and median; the upper and lower boundaries of box are the interquartile range). Severity Severity Severity Severity Severity Severity 10 Advances in Meteorology available, e.g., Sh-Sd, Sss-Ql, Lyb-Yng, Ljq-Sd, Wfc-Ql, and HD were significantly different in Ylx and Zyx for different periods based on the thresholds −1.0, −1.5, and −2.0. In Ylx, Syk-Zy). Figure 8 indicates that the propagation times were 1, 1, 3, 1, 1, and 1 in the Pre-R period, respectively; 1, 3, 9, 3, the characteristics of HD were similar to MD in that, com- pared to the Pre-R period, the average intensity, duration, and 6, and 6 in the Pos-R period, respectively; and 1, 3, 6, 1, 6, severity of the HD decreased significantly, especially for SDI-3 and 6 in the total period for six tributaries, respectively. and SDI-12. However, in Zyx, the HD were quite inconsistent Figure 9(a) shows that the SDI-1 and SPI had the highest PC with MD, and the average intensity, duration, and severity of (PC � 0.27, p< 0.01) in the Pre-R period for the 9-month the HD increased significantly during the Pos-R period in period, and, in the Pos-R period, the highest PC (PC � 0.46, Zyx, especially for SDI-1, SDI-6, and SDI-12. p< 0.01) was observed for the 6-month period. Figure 9(a) also shows the highest PC for SDI-1 and SPI (PC � 0.38, p< 0.01) was for the 6-month period before the EWDP 3.2. Comparison of MD Characteristics for SPI-1 and HD for period, and, after the EWDP period, the highest PC SDI-1. *e characteristics of MD and HD based on SPI-1 and (PC � 0.48, p< 0.01) was observed for the 6-month period in SDI-1 using thresholds of −1, −1.5, and −2 in the UHRB and Ylx. Figure 9(b) reveals the highest PC (PC � 0.19, p< 0.01) MHRB are shown in Tables 4-5 and Figure 5. More MD was found for the 1-month period Pre-R and the 9-month events occurred at Yng station (names and abbreviations are period (PC � 0.24, p< 0.01) Pos-R. Figure 8(b) also dem- shown in Table 1) in the UHRB (Figure 5) and at Sd station in onstrates the highest PC (PC � 0.17, p< 0.01) was found for the MHRB. More HD events occurred at Zyx (Figure 5). It can the 3 months before the EWDP period and over 36 months be seen from Tables 4-5 that the average, maximum, and (PC � 0.35, p< 0.01) after the EWDP period in Zyx. *ese minimum peak/intensity/duration/severity in Zyx, Yng, and results demonstrate that the changes in the streamflow were Ql stations were identified as longer and more severe based on more sensitive to precipitation in the Pre-R period for short thresholds of −1.0, −1.5, and −2.0 for MD and HD. HD timescales (1–3 months), and Pos-R propagation time in- characteristics based on these thresholds for the average peak, creased (2–6 months) for tributaries, while the propagation maximum peak, and minimum peak in Sh, respectively, were time decreased by 3 months in Ylx and increased by 8 more obvious, and longer average, maximum, and minimum months in Zyx compared to the Pre-R period. *us, res- intensities were found in Ylx, Ljq, and Sh (Table 5). Longer, ervoir operation increased the propagation time of the more severe HD occurred in Ylx, Ljq, Sh, and Wfc. streamflow variables to the SPI during the Pos-R period in Zyx, whereas the propagation time of the streamflow vari- ables to the SPI was decreased during the Pos-R period in 3.3. Variation in VCI. Vegetation condition in the extremely Ylx. *ere was no change in Ylx before and after the EWDP, dry months of April, May, and October (Figure 6) was poor, while after the EWDP, the propagation time of the HD to the especially for alpine vegetation, cultivated vegetation, MD increased on a 33-month timescale compared with grassland, and marshy grassland. From June to September, before the EWDP in Zyx. Figure 9(c) shows that the VCI and the vegetation was in optimal condition for coniferous SPI strongest negative PC was found for the 6-month forests, shrubs, and marshy grassland in the UHRB and for (PC � −0.09) period in the UHRB, and the strongest positive coniferous forests and cultivated vegetation in the MHRB. It PC was found for the 36-month (PC � 0.13) period in the is worth noting that the vegetation condition was poor for MHRB. Figure 9(d) shows that the VCI and SDI highest PC alpine vegetation over the whole growing season. was found for the 6-month period (PC � 0.23, p< 0.01) in We further analyzed the trend and significance level of the the UHRB and the 3-month period (PC � 0.43, p< 0.01) in growing season for VCI. Figure 7 shows that, over the entire the MHRB. *e results indicated that correlation between growing season, alpine vegetation (UHRB) and grassland VCI and MD was not significant in the UHRB and MHRB. (MHRB) showed a significant decreasing trend, while culti- However, the correlation between VCI and HD was sig- vated vegetation (MHRB), coniferous forest, shrubs, and nificant in the UHRB (Ylx) and MHRB (Zyx), while the VCI grassland (UHRB) showed a significant increasing trend. propagation time to HD was found at short timescales. 3.4. Correlation of VCI, HD, and MD 3.4.2. Intra-Annual and Lag Correlations between VCI, HD, and MD at Multiple Scales in Different Periods. *e SDI-1 3.4.1. Interannual Propagation Time between VCI, HD, and MD at Different Timescales in Different Periods. with SPI at different timescales was investigated in relation to the propagation from MD to HD. Figure 10 shows the Figures 8-9 describe the relationship between SDI-1 and VCI-1 with all SPI accumulation periods (1–48 months) and propagation times from MD to HD had significant seasonal characteristics for tributaries in different periods. In spring VCI with all SDI accumulation periods (1–12 months) for (March–May), the drought propagation time was relatively different periods (Pre-R and Pos-R, before and after EWDP, short (1–12 months) in Sss, Ljq, and Wfc, whereas the and total period) in the mainstream for Ylx (SPI calculated propagation time was relatively long (1–48 months) in Sh, using the arithmetic mean of precipitation from Yng and Ql Lyb, and Syk. In summer (June–August), the drought stations), Zyx (SPI calculated using the arithmetic mean of propagation time was relatively long, except for Wfc. precipitation from Gt, Sd, and Zy stations), and tributaries (SPI calculation is based on the area flowing through or the Drought propagation times in autumn (September–No- vember) were relatively short (1–6 months) in Sss, Lyb, and adjacent area because all meteorological stations were not Advances in Meteorology 11 Table 4: *e distribution peak, intensity, duration, and severity for MD characteristics based on SPI-1 using thresholds of −1, −1.5, and −2 in the UHRB and MHRB. Peak Intensity Duration Severity *reshold Mean Max Min Mean Max Min Mean Max Min Mean Max Min −1.0 1.7 3.6 1.1 1.7 4.0 1.0 4.6 15.0 1.0 3.1 10.2 1.1 Ql −1.5 2.0 3.6 1.2 1.3 3.0 1.0 4.6 15.0 1.0 3.6 10.2 1.5 −2.0 2.5 3.6 2.0 1.0 1.0 1.0 4.3 15.0 1.0 4.1 10.2 2.0 −1.0 1.6 3.1 1.7 1.7 5.0 1.0 4.4 23.0 1.0 3.8 12.9 1.0 Yg −1.5 2.2 3.1 2.5 1.6 3.0 1.0 5.5 23.0 1.0 5.4 12.9 1.8 −2.0 2.5 3.1 2.0 1.2 2.0 1.0 6.9 23.0 1.0 5.1 12.5 2.0 −1.0 1.4 2.9 1.0 1.5 4.0 1.0 4.1 14.0 1.0 3.1 8.1 1.0 Gt −1.5 1.8 2.9 1.5 1.1 2.0 1.0 4.5 14.0 1.0 3.6 8.1 1.6 −2.0 2.7 2.9 2.5 1.0 1.0 1.0 2.0 3.0 1.0 3.5 4.0 2.9 −1.0 1.6 2.7 1.0 1.6 6.0 1.0 4.7 30.0 1.0 3.5 18.8 1.0 Zy −1.5 2.0 2.7 1.5 1.5 4.0 1.0 5.5 30.0 1.0 4.4 18.8 1.9 −2.0 2.3 2.7 2.0 1.1 2.0 1.0 4.0 15.0 1.0 3.9 10.0 2.0 −1.0 1.6 3.0 1.0 1.6 5.0 1.0 4.2 17.0 1.0 3.2 9.4 1.0 Sd −1.5 1.9 3.0 1.5 1.2 3.0 1.0 3.5 14.0 1.0 3.4 9.4 1.5 −2.0 2.3 3.0 2.0 1.1 2.0 1.0 3.1 11.0 1.0 3.7 9.4 2.0 Table 5: *e distribution of peak, intensity, duration, and severity for HD characteristics based on SDI-1 using thresholds of −1, −1.5, and −2 in the UHRB and MHRB. Peak Intensity Duration Severity *reshold Mean Max Min Mean Max Min Mean Max Min Mean Max Min −1.0 1.6 2.7 1.1 4.3 18.0 1.0 11.1 31.0 1.0 8.6 30.8 1.1 Ylx −1.5 1.9 2.7 1.5 4.0 18.0 1.0 12.3 31.0 1.0 10.9 30.8 1.5 −2.0 2.3 2.7 2.1 1.8 3.0 1.0 19.5 31.0 10.0 18.0 24.6 8.9 −1.0 1.8 3.3 1.0 3.7 14.0 1.0 12.4 36.0 1.0 8.1 24.7 1.0 Sss −1.5 2.0 3.3 1.5 2.1 5.0 1.0 12.4 30.0 1.0 8.5 22.2 1.5 −2.0 2.6 3.3 2.0 1.3 2.0 1.0 9.5 20.0 2.0 7.7 12.2 2.2 −1.0 1.7 3.0 1.0 3.4 17.0 1.0 8.9 51.0 1.0 7.0 37.0 1.2 Wfc −1.5 2.0 3.0 1.6 2.7 8.0 1.0 12.2 51.0 1.0 9.8 37.0 1.6 −2.0 2.5 3.0 2.0 2.0 4.0 1.0 19.3 54.0 6.0 15.5 40.4 6.6 −1.0 1.8 3.0 1.0 4.1 16.0 1.0 10.2 30.0 1.0 8.7 35.0 1.0 Lyb −1.5 2.1 3.0 1.6 2.8 8.0 1.0 10.4 30.0 1.0 9.9 35.0 1.7 −2.0 2.4 3.0 2.0 1.4 4.0 1.0 8.3 28.0 1.0 9.4 32.0 2.0 −1.0 1.8 5.0 1.0 3.0 21.0 1.0 7.0 40.0 1.0 6.2 41.3 1.1 Zxy −1.5 2.2 5.0 1.5 2.6 12.0 1.0 6.9 38.0 1.0 7.6 38.9 1.5 −2.0 2.7 5.0 2.0 1.6 6.0 1.0 9.5 37.0 1.0 10.2 37.3 2.1 −1.0 1.8 4.0 1.0 3.4 17.0 1.0 7.0 32.0 1.0 6.8 33.7 1.1 Ljq −1.5 2.4 4.0 1.5 3.3 12.0 1.0 10.3 32.0 1.0 10.6 33.7 1.5 −2.0 2.9 4.0 2.1 3.0 12.0 1.0 13.6 32.0 3.0 14.0 29.8 3.8 −1.0 2.1 4.3 1.0 4.3 18.0 1.0 12.9 43.0 1.0 9.7 36.3 1.7 Sh −1.5 2.5 4.3 1.6 3.0 9.0 1.0 8.9 33.0 1.0 9.2 35.3 1.7 −2.0 2.8 4.3 2.0 1.8 5.0 1.0 8.4 32.0 1.0 10.3 33.7 2.3 −1.0 1.7 3.8 1.0 2.4 8.0 1.0 7.6 36.0 1.0 5.2 18.7 1.0 Syk −1.5 2.0 3.8 1.5 1.6 3.0 1.0 7.6 36.0 1.0 5.9 18.7 1.8 −2.0 2.9 3.8 2.4 2.0 3.0 1.0 10.8 35.0 2.0 9.2 18.7 3.2 Syk and relatively long (1–48 months) in Sh, Ljq, and Wfc. drought propagation times in winter did not change and Drought propagation times in winter (December–February) remained long, whereas drought propagation times were were similar to those in summer and were relatively long increased in autumn. Propagation time was longer for Ss, (1–48 months), except for Lyb in the Pre-R period. Drought Lyb, and Syk and shorter for Sss, Ljq, and Wfc in spring; and propagation times in autumn and winter were relatively propagation times in summer were shorter, except for Wfc long, whereas in spring and summer, they were relatively and Ss. It can be clearly seen from Figure 11 that the intra- short in the Pos-R period. *e total period was similar to the annual correlations between the monthly SDI-1 and SPI at Pos-R period. To summarize, in the Pos-R period, the multiple scales and in different periods have strong seasonal 12 Advances in Meteorology N N 15 9 23 35 9 30 10 16 8 6 11 8 3.8 0 60 120 0 60 120 Kilometers Kilometers Events Events Threshold = –1.0 Threshold = –1.0 Threshold = –1.5 Threshold = –1.5 Threshold = –2.0 Threshold = –2.0 (a) (b) Figure 5: Spatial distribution of MD (SPI-1) and HD (SDI-1) events in the UHRB and MHRB. N N 0 60 120 0 60 120 Kilometers Kilometers VCI VCI 4.2 – 15 46 – 60 3 – 15 46 – 60 16 – 30 61 – 65 16 – 30 61 – 90 31 – 45 31 – 45 (a) (b) N N 0 60 120 0 60 120 Kilometers Kilometers VCI VCI 4.2 – 15 46 – 60 2.5 – 15 46 – 60 16 – 30 61 – 91 16 – 30 61 – 95 31 – 45 31 – 45 (c) (d) Figure 6: Continued. Advances in Meteorology 13 N N 0 60 120 0 60 120 Kilometers Kilometers VCI VCI 2.5 – 15 46 – 60 2.8 – 15 46 – 60 16 – 30 61 – 97 16 – 30 61 – 97 31 – 45 31 – 45 (e) (f) 0 60 120 Kilometers VCI 2.6 – 15 46 – 60 16 – 30 61 – 89 31 – 45 (g) Figure 6: *e spatial variation in the average VCI for the growing season (April–October) in the UHRB and MHRB. N N 0 60 120 0 60 120 Kilometers Kilometers Extremely significant decrease Insignificant increase Extremely significant decrease Insignificant increase Significant decrease Significant increase Significant decrease Significant increase Insignificant decrease Extremely significant increase Insignificant decrease Extremely significant increase (a) (b) Figure 7: Continued. 14 Advances in Meteorology N N 0 60 120 0 60 120 Kilometers Kilometers Extremely significant decrease Insignificant increase Extremely significant decrease Insignificant increase Significant decrease Significant increase Significant decrease Significant increase Insignificant decrease Extremely significant increase Insignificant decrease Extremely significant increase (c) (d) N N 0 60 120 0 60 120 Kilometers Kilometers Extremely significant decrease Insignificant increase Extremely significant decrease Insignificant increase Significant decrease Significant increase Significant decrease Significant increase Insignificant decrease Extremely significant increase Insignificant decrease Extremely significant increase (e) (f) 0 60 120 Kilometers Extremely significant decrease Insignificant increase Significant decrease Significant increase Insignificant decrease Extremely significant increase (g) Figure 7: *e spatial distribution of VCI trends for the growing season in the UHRB and MHRB. Advances in Meteorology 15 0.30 0.15 0.00 –0.15 1 3 6 9 12151821243648 Timescale (SPI) Sh Ljq Sss Wfc Lyb Syk (a) 0.30 0.15 0.00 –0.15 1 3 6 9 12151821243648 Timescale (SPI) Sh Ljq Sss Wfc Lyb Syk (b) 0.30 0.15 0.00 –0.15 1 39 6 12 15 18 21 24 36 48 Timescale (SPI) Sh Ljq Sss Wfc Lyb Syk (c) Figure 8: Pearson’s correlation coefficien between the SDI-1 series and the SPI series at different timescales for six tributaries. characteristics in Ylx and Zyx. *e propagation times in autumn, and winter were approximately 9, 9, 3, and 6 months in Zyx for the total period (Figure 11(f)), respec- spring, summer, autumn, and winter were approximately 3, 9, 6 and 12 months in the Ylx for the Pre-R period tively. *ese results show that the propagation time in the (Figure 11(a)), respectively. *e propagation times in the Pos-R period increased in spring, decreased in autumn and spring, summer, autumn and winter were approximately 6, winter, and did not change in summer in Ylx compared to 9, 1, and 6 months in Ylx for the Pos-R period the Pre-R period, whereas in Zyx it increased in spring, (Figure 11(b)), respectively. *e propagation times in summer, and autumn, and it decreased in winter. Com- spring, summer, autumn, and winter were 6 months in Ylx pared with the Pre-R period, the PC was lower on a 1- for the total period (Figure 11(c)). *e propagation times in month timescale in Ylx (Figure 11(b)) and on a 1–3-month spring, summer, autumn, and winter were approximately 6, timescale in Zyx (Figure 11(e)). However, the PC between 1, 1, and 15 months in Zyx for the Pre-R period SDI-1 and different timescale of SPI in the Pos-R period (Figure 11(d)), respectively. *e propagation times in was higher than that in the Pre-R period for long-term timescales (>12 months) in Ylx (except for January and spring, summer, autumn, and winter were 9, 12, 36, and 9 months in Zyx for the Pos-R period (Figure 11(e)), re- June) and Zyx (except for June, July, November, and spectively. *e propagation times in spring, summer, December). It can be seen from Figures S3(a)-S3(b) that, PC PC PC 16 Advances in Meteorology 0.45 0.30 0.30 0.15 0.15 0.00 0.00 –0.15 1 3 6 912 15 18 21 24 36 48 1 361 912 158 21243648 Timescale (SPI) Timescale (SPI) UHRB (1965–2012) Pre-R (1965–1979) MHRB (1965–2012) Pre-R (1965–1979) Pos-R (1980–2012) Before EWDP (1965–1999) Pos-R (1980–2012) Before EWDP (1965–1999) Aer EWDP (2000–2012) Aer EWDP (2000–2012) (a) (b) 0.45 0.15 0.10 0.30 0.05 0.00 0.15 –0.05 –0.10 0.00 –0.15 1 3 6 9 12151821243648 1 369 12 Timescale (SPI) Timescale (SDI) UHRB UHRB MHRB MHRB (c) (d) Figure 9: Pearson’s correlation coefficien between SDI-1 and SPI (a, b) for different timescales at different periods, and VCI (c, d) with different timescales of SPI and SDI in the UHRB (Ylx) and in the MHRB (Zyx). 0.05 0.22 0.39 0.56 0.73 0.89 0.05 0.22 0.39 0.56 0.73 0.89 PC PC 12 12 11 11 10 10 9 9 8 8 7 7 6 6 5 5 4 4 Sh Sss Lyb Ljq Wfc Syk Sh Sss Lyb Ljq Wfc Syk Stations Stations (a) (b) Figure 10: Continued. Pre-R PC PC Month PC Pos-R PC Month 29 Advances in Meteorology 17 0.05 0.22 0.39 0.56 0.73 0.89 1 10 20 39 48 PC Months 12 12 11 11 10 10 9 9 8 8 7 7 6 6 5 5 4 4 3 3 2 2 1 1 Sh Sss Lyb Ljq Wfc Syk Sh Sss Lyb Ljq Wfc Syk Stations Stations (c) (d) 110 20 39 48 1 10 20 39 48 Months Months 12 12 11 11 10 10 9 9 8 8 7 7 6 6 5 5 4 4 3 3 2 2 1 1 Sh Sss Lyb Ljq Wfc Syk Sh Sss Lyb Ljq Wfc Syk Stations Stations (e) (f) Figure 10: Intra-annual strongest PC (a–c) and the drought propagation time in months (e–f) for SPI and SDI-1 at six tributaries. 1–48 months for different timescales at different periods after the EWDP, the PC of SDI-1 and the timescale of SPI were higher in spring and autumn and lower in summer (Pre-R, Pos-R, and total), and the strongest correlation was (except for August) and winter than before the EWDP in detected at a lag of 0 months (i.e., no lag) in Ylx and Zyx Ylx. However, Figures S3(c)-S3(d) show that, after the (Figure 13). We also found no lag before and after the EWDP EWDP, the PC of SDI and SPI was higher in spring, (Figure S5), whereas there was a higher PC at 3–12 months summer, autumn, and winter, specifically in spring, in Zyx. before the EWDP and after the EWDP at 6–15 months in In different periods (Pre-R and Pos-R and before and after Ylx. *e highest PC was at 1–12 months before the EWDP EWDP) and different timescales, the PC of the Pos-R and after EWDP at 9–48 months in Zyx. We also analyzed period and after the EWDP was higher in Ylx (except for 1, the lag correlation of the SPI with VCI-1 and SDI (accu- mulation periods of 1–12 months) with VCI-1 (Figure 14), 36, and 48 months) and Zyx, especially in Zyx than in the Pre-R period and before the EWDP. Figure 12(a) shows and PC was calculated for the VCI-1 series, which were lagged by 0 to 6 months after the SPI and SDI series. *e that the higher PC between VCI and SPI in April at 6 months was 0.63 (p< 0.01), whereas the lower PC in May at results demonstrated that the highest PC (PC � 0.21, 1 month and 48 months was −0.06. *e highest PC between p< 0.01) was found at a 6-month lag between VCI and SPI, VCI and SDI in May at 12 months was 0.78 (p< 0.01), while the strongest correlation was at a lag of 0 months (i.e., whereas the lowest PC in March at 3 months was −0.69 no lag) between VCI and SDI in the UHRB (Ylx) during (p< 0.01) in the UHRB (Figure 12(b)). Figure 12(c) shows 1999–2012. *e highest PC (PC � 0.26, p< 0.01) was ob- that the highest PC between VCI and SPI was 0.84 served at the 4-month lag scale between VCI and SPI, and (p< 0.01) in July at 9 months, whereas the lowest PC was the strongest PC (PC � 0.47, p< 0.01) was found at a lag of 1 −0.19 in October at 1 month. *e highest PC between VCI month between VCI and SDI during 1999–2012 in the MHRB (Zyx). As mentioned above, it can be concluded that and SDI was observed at 0.71 in May at 6 months, and the lowest PC was −0.23 in March at 6 months in the MHRB there was no lag between SDI and SPI; however, VCI with (Figure 12(d)). SPI had significant lag correlations in the short term in the We further analyzed whether there was a lag between the UHRB and MHRB. Additionally, the VCI with SDI had a different timescales of SPI with SDI-1. *e PC between SDI- significant 1-month lag correlation in the short term in the 1 lagged by 0–6 months, with SPI accumulation periods of MHRB (Zyx). Total period Month Month Month Month –0.04 18 Advances in Meteorology PC PC 0.78 0.83 11 11 10 10 0.54 0.62 9 9 8 8 0.29 0.41 7 7 6 6 0.05 0.21 5 5 4 4 –0.20 0.00 3 3 2 2 1 1 –0.45 –0.21 1 369 12 15 18 21 24 36 48 1 369 12 15 18 21 24 36 48 Timescale (SPI) Timescale (SPI) (a) (b) PC PC 0.75 0.51 12 12 11 11 10 10 0.58 0.28 9 9 8 8 0.40 0.05 7 7 6 6 0.23 –0.18 5 5 4 4 0.06 –0.40 3 3 2 2 1 1 –0.12 –0.63 1 369 12 15 18 21 24 36 48 1 369 12 15 18 21 24 36 48 Timescale (SPI) Timescale (SPI) (c) (d) PC PC 0.46 0.54 12 12 11 11 10 10 0.29 0.34 9 9 8 8 0.13 0.13 7 7 6 6 –0.08 5 5 4 4 –0.20 –0.29 3 3 2 2 1 1 –0.37 –0.50 1 369 12 15 18 21 24 36 48 1 369 12 15 18 21 24 36 48 Timescale (SPI) Timescale (SPI) (e) (f) Figure 11: Heat map showing Intra-annual PC of SPI accumulation periods of 1–48 months with SDI-1 in the UHRB (Ylx) and MHRB (Zyx) (note: the x-axis represents the correlation between the SDI-1 and SPI at different timescales and the y-axis represents the 12 months of the year; the up arrow represents the correlation coefficients that increased in the postreservoir period (Pos-R) in the Ylx and Zyx compared to the period of prereservoir (Pre-R), or vice versa); (a) (1965–1979, PC≥ 0.52, p< 0.05), (b) (1980–2012, PC≥ 0.35, p< 0.05), and (c) (1965–2012, PC≥ 0.29, p< 0.05) in the UHRB (Ylx), and (d) (1965–1984, PC≥ 0.45 or≤− 0.47, p< 0.05), (e) (1985–2012, PC≥ 0.38 or≤−0.48, p< 0.05), and (f ) (1965–2012, PC≥ 0.29 or ≤ −0.33, p< 0.05) in the MHRB (Zyx). Total period Pre-R Month Month Month Pos-R Month Month Month 0.15 0.19 –0.01 Advances in Meteorology 19 UHRB (VCI-SPI) PC MHRB (VCI-SPI) PC 0.63 0.85 9 9 0.49 0.64 8 8 0.35 0.43 7 7 0.22 0.22 6 6 5 0.08 5 0.02 4 4 –0.06 –0.19 136 9 12 15 18 21 24 36 48 136 9 12 15 18 21 24 36 48 Timescale (SPI) Timescale (SPI) (a) (b) UHRB (VC-SDI) PC MHRB (VCI-SDI) PC 0.78 0.71 9 0.62 9 0.61 8 8 0.46 0.52 7 7 0.31 0.42 6 6 0.33 5 5 4 4 0.23 16 3 9 12 16 3 9 12 Timescale (SDI) Timescale (SDI) (c) (d) Figure 12: Heat map showing growing season Pearson’s correlation coefficients between VCI and different timescale of SPI, VCI, and different timescale of SDI for the period of 1999–2012 in the UHRB (Ylx). (a) (PC≥ 0.53, p< 0.05), (b) (PC≥ 0.55 or≤ −0.69, p< 0.01) and MHRB (Zyx), (c) (PC≥ 0.55, p< 0.05), and (d) (PC≥ 0.54, p< 0.05). PC PC 0.27 0.46 6 6 5 0.18 5 0.37 4 4 0.09 0.28 3 3 –0.00 1 –0.10 1 0.11 0 0 –0.19 0.02 1 3 6 9 12 15 18 21 24 36 48 1 3 6 9 12 15 18 21 24 36 48 Timescale (SPI) Timescale (SPI) (a) (b) Figure 13: Continued. Month Month Pre-R Lag-months Month Pos-R Month Lag-months 2 20 Advances in Meteorology PC PC 0.44 0.19 6 6 5 5 0.37 0.10 4 4 0.30 0.02 3 3 0.22 –0.06 1 0.16 1 –0.14 0 0 0.09 –0.22 1 3 6 9 12 15 18 21 24 36 48 1 3 6 9 12 15 18 21 24 36 48 Timescale (SPI) Timescale (SPI) (c) (d) PC PC 0.24 0.18 6 6 5 0.19 5 0.13 4 4 0.13 0.08 3 3 0.08 0.03 1 0.03 1 –0.02 0 0 –0.02 –0.07 1 3 6 9 12 15 18 21 24 36 48 1 3 6 9 12 15 18 21 24 36 48 Timescale (SPI) Timescale (SPI) (e) (f) Figure 13: Heat maps showing PC between SDI-1 lagged by 0–6 months and SPI accumulation periods of 1–48 months for different timescales at different periods in the UHRB (Ylx) and MHRB (Zyx). (a) (1965–1979, PC≥ 0.15 or≤‒0.16, p< 0.05), (b) (1980–2012, PC≥ 0.14, p<0.01), and (c) (1965–2012, PC≥ 0.11, p<0.01) in the UHRB (Ylx) and (d) (1965–1984, PC≥ 0.16 or ≤‒0.11, p< 0.05), (e) (1985–2012, PC≥ 0.12, p< 0.01), and (f) (1965–2012, PC≥ 0.10, p< 0.05) in the MHRB (Zyx). For the total period (Table S2), HD events were more fre- 4. Discussion quent at Zyx than at Ylx for different timescales based on Our results indicated that MD were more serious in Pos-R thresholds of −1.0, −1.5, and −2.0. *ese results were than in Pre-R in the UHRB, there was little change in the consistent with Ma et al. [44]. Figure 1(c) and Table 6 show MHRB, and MD were more serious in the UHRB than in the that the main vegetation type in the UHRB was grassland, MHRB for duration and severity based on thresholds of and there was an increasing trend to cultivated land for −1.0, −1.5, and −2.0 for the total period. *ese results are 2000–2010. *e UHRB had fewer reservoirs and less arable inconsistent with [44], probably because the precipitation land than the MHRB, and the drought duration and severity data types and SPI probability distribution function were were more likely to be related to climate change. For ex- different. In our study, precipitation data types were based ample, Ma et al. [44], Qiu et al. [52], Gao and Zhang [62], on meteorological stations, and the SPI probability distri- Yang et al. [63], Cong et al. [64], and Cheng et al. [65] bution function was a three-parameter log-logistic distri- showed that an increase in streamflow was caused by climate bution. However, Ma et al. [44] used 0.5 resolution change (increased precipitation) factors rather than human precipitation data and their SPI probability distribution activities in the UHRB. At Zyx, the duration and severity of function was an empirical Gringorten plotting position. *e drought were greater than those in the Pre-R period. More Pos-R period had a significant positive impact on the extreme MD events occurred in 1985, 1992, 2000, and 2005 evolution of different timescales based on thresholds of −1.0, (i.e., Pos-R) [66], and the decrease in precipitation may have −1.5, and −2.0 in Ylx, which was consistent with Wu et al. led to a decrease in streamflow and thus more serious HD [60], who showed that Pos-R period in the HD had reduced events compared with the Pre-R period (Table S2). *e area duration and severity compared with Pre-R period. How- of cultivated land increased, and unused land and grassland ever, there were more serious duration and severity, spe- decreased from 1975 to 2010. *e total irrigation volume in 8 3 cifically at 6 and 12 months, than in the Pre-R period in Zyx. the area increased by 0.57 ×10 m /year in 1990–2010 in the Total-period Lag-months Lag-months Lag-months Lag-months Advances in Meteorology 21 between VCI and SPI, consistent with Li et al. [19]. MHRB [54]. Along with the rapid increase in the number of pumping wells, groundwater irrigation has increased Meanwhile increasing drought timescales (≥12 months) 8 3 suggested positive correlations in the UHRB and MHRB 2.11 × 10 m /year [54]. It is clear that human activities have changed the land surface and altered hydrological processes, (Figure 8(c)). Niu et al. [45] also found that variations in the including evapotranspiration, infiltration, surface runoff, Enhanced Vegetation Index (EVI) generally followed the and storage of water and; in this way, they affect the de- evolution of soil moisture in the top 1 m of the soil profile. velopment of drought and thereby exacerbate drought [67]. *ere were both positive and negative time lags. For ex- Zhu et al. [68] also pointed out that human activities, such as ample, during a 2001 drought event, it was observed that the the expansion of irrigation, rapid population growth, and Soil Moisture Anomaly Index reached a negative peak value. However, the EVI did not show a consistent response, in- socioeconomic development, have deeply modified hydro- logical processes in the MHRB. *erefore, the above- dicating it was not affected by natural drought events be- cause it was maintained by irrigation activities in cropland in mentioned climate change and human activities may affect changes in drought characteristics between the different the MHRB. *e strongest PC values between VCI and SPI were for 6-month and 36-month timescales for the growing periods in the UHRB and MHRB. Previous research has examined the propagation time season in the UHRB and MHRB, respectively. *ese results and lags of HD to MD in different regions of the world were inconsistent with Vicente-Serrano et al. [41], who [4, 23, 43, 44]. In this study, we found that the PC at Ylx was found that, in arid and humid regions, vegetation responded larger than that at Zyx. *e correlation with climate change to MD at short timescales as well as in semiarid and sub- and human activities changed from positive to negative at humid regions at long timescales. *e main reason may be Zyx. Ma et al. [44] found that climate change was inclined to the relationship with different vegetation types, anthropo- genic activity, and climate change, although the UHRB is a increase streamflow and propagation time, contributing from −57% to 63% in the UHRB, whereas human activities semiarid region. In Figure 1(c), there is little coniferous forest, and the region is mainly grassland, shrub, and alpine played a dominant role in water consumption with a con- tribution rate greater than −89% to further alter HD vegetation. However, cultivated vegetation dominates in the MHRB, and there are 24 irrigation districts (Figure S5) and characteristics and propagation time in the MHRB. In this study, we found the MD propagation to HD was different in 71 reservoirs (four under construction) with thousands of different catchments and had a significant seasonal dis- canals and over 6000 pumping wells for irrigation [71]. *e crepancy at different timescales for different periods. *e total irrigation volume in the area has increased in the reservoirs increased the propagation time of the hydrological MHRB [52]. *us, crop improvements and water-saving variables to MD during the Pos-R period at Zyx, and the irrigation technologies would increase water use efficiency results were in line with Wu et al. [60]. *e PC values and drought tolerance capacity [70]. Our findings also differed from Zhang et al. [18], who found significant annual between SDI and SPI in the Pos-R period were lower on a short-term scale (1–3 months) than those in the Pre-R high PC values in arid and semiarid regions, and the cor- responding drought timescales were 3–6 months (Januar- period, whereas they were higher for the long-term timescale (>12 months) at Ylx and Zyx, which was also consistent with y–December). *is may be related to the different timescales (April–October). We further found a significant positive and Wu et al. [60]. *e propagation time in the Pos-R period increased in spring, decreased in autumn and winter, and high PC value (PC≥ 0.53, p≤ 0.05) between VCI and SPI on did not change in summer at Ylx compared with the Pre-R an intra-annual scale, implying that water availability is the period, while it increased in spring, summer, and autumn critical factor for vegetation’s various spatiotemporal ac- and decreased in winter at Zyx. *e possible causes of the tivities [52] in the UHRB and MHRB, especially in the uneven spatial variations of precipitation are temperature MHRB (PC≥ 0.70, p≤ 0.01). Our results also revealed that, and melting of snow and glaciers in the Qilian Mountains with an increasing drought timescale, the PC values between VCI and SPI were higher in the MHRB, specifically in when the temperature is rising, which may lead to the differences in HD variations [44, 62]. *e propagation time August (the PC values fluctuated from 0.43 to 0.82 for 1–48 months) and September (the PC values fluctuated from 0.07 for tributaries would also vary in different periods in relation to natural and social environmental factors [4, 23, 69]. For to 0.83 for 1–48 months). *ese results suggested that instance, Deng et al. [70] demonstrated that the 33 croplands can suffer from extreme and prolonged drought tributaries in the MHRB no longer joined the mainstream conditions [10] in the MHRB. Furthermore, different veg- after 1980s, and they gradually disappeared and formed etation types have different propagation times for MD; the independent irrigation oases. *ese may diverge in terms of grassland and cultivated vegetation at a 3-month timescale different climate and human activity impacts. *e lack of lag and a 12-month timescale for the shrubland, coniferous in the SDI and SPI was in line with Barker et al. [23] at Ylx forest, and broadleaf forest [8, 21]. Figures 14 and 6 dem- and Zyx. onstrate that alpine vegetation and grassland were poor (especially in April and May) and there was an extremely *e high correlation between monthly NDVI and the MD index at different timescales is an indicator that can significant decrease for the growing season, whereas culti- vated vegetation was good, and there was an extremely describe the impact of drought on vegetation, and the month of highest correlation means the greatest sensitivity of significant increase for the growing season in the UHRB and vegetation to drought [10, 41]. Our results exhibited a MHRB. Additionally, VCI with SPI had a significant lag negative correlation on short timescales (≤9 months) correlation in the short term in the UHRB (6-month lag) and 2 22 Advances in Meteorology UHRB (VCI-SPI) PC MHRB (VCI-SPI) PC 0.21 0.26 6 6 5 5 0.15 0.18 4 4 0.09 0.11 3 3 0.02 0.03 1 –0.04 1 –0.04 0 0 –0.10 –0.12 1 3 6 9 12 15 18 21 24 36 48 1 3 6 9 12 15 18 21 24 36 48 Timescale (SPI) Timescale (SDI) (a) (b) PC MHRB (VCI-SDI) VCI–SPI (VCI-SDI) PC 0.25 0.48 6 6 5 5 0.15 0.34 0.05 0.20 0.07 0.05 –0.07 1 –0.15 1 0 0 –0.24 –0.21 1 369 12 1 369 12 Timescale (SDI) Timescale (SDI) (c) (d) Figure 14: Heat maps showing PC between VCI lagged by 0–6 months and SPI (1-48)/SDI (1-12) for the period of 1999–2012 in the UHRM (Ylx). (a) PC≥ 0.16, p< 0.05), (b) PC≥ 0.18 or≤‒0.15, p< 0.05) and in the MHRB (Zyx), (c) PC≥ 0.15, p< 0.05), and (d) PC≥ 0.16 or≤‒0.18, p< 0.05). Table 6: Changes in the area of the primary types of land use and transition matrix and the primary types of land use changes in the UHRB 4 2 and MHRB (10 hm ). Period Land use types Cultivated Forest Grassland Water area Built-up area Unused land Cultivated 0.10 0.02 0.16 0.01 0.02 0.00 Forest 0.05 1.84 16.78 0.48 0.00 1.81 Grassland 0.29 0.65 42.13 1.27 0.03 6.13 UHRB 2000–2011 Water area 0.04 0.06 0.96 0.91 0.01 0.63 Built land 0.03 0.00 0.04 0.00 0.02 0.00 Unused land 0.00 0.08 7.83 0.57 0.00 16.87 Cultivated 14.41 0.04 0.26 0.05 0.28 0.15 Forest 0.42 1.20 0.21 0.01 0.01 0.12 Grassland 3.97 0.26 12.86 0.16 0.06 1.70 MRHB 1975–2010 Water area 0.53 0.02 0.34 1.86 0.01 0.19 Built land 0.06 0.00 0.00 0.00 1.54 0.00 Unused land 1.84 0.38 1.46 0.14 0.22 62.76 1975–1987 14.40 1.52 15.33 2.22 1.60 65.52 1987–1992 15.25 1.21 14.49 2.04 1.34 66.88 MRHB 1992–2001 16.48 1.23 14.29 2.11 1.41 67.04 2000–2010 17.88 1.25 13.03 1.80 1.71 63.87 MHRB (4-month lag), and the VCI with SDI had a sig- timescale of the HD and VCI with the different timescales of nificant 1-month lag correlation in the MHRB, similar to Lin the MD did not fully reflect their relationships. *e reason is et al. [35]. Additionally, the correlation between a single that the streamflow and vegetation propagation times with Lag-months Lag-months Lag-months Lag-months Advances in Meteorology 23 precipitation differ at different timescales the average peak, intensity, duration, and severity of [10, 18, 19, 21, 72, 73]. HD decreased at Ylx, whereas the effects were more Our study had some limitations. *e quantification of serious at Zyx. *e PC values were lower at short-term drought is a difficult task and drought is intrinsically scales (1–3 months) than at long-term scales for the multiscalar. *ere is no unique physical variable we can Pos-R period at Ylx and Zyx. *e PC values for MD measure to quantify drought intensity. Vegetation propa- and HD after the EWDP were higher than those gation time to droughts is still an open scientific problem before the EWDP in the UHRB and MHRB, especially due to the complexity of droughts and limited knowledge of in the MHRB. Human activities at different timescales physiological processes. However, understanding the rela- (Pre-R period and Pos-R period and before EWDP tionship between these mechanisms and the characteristics and after EWDP) may affect the correlation between of droughts is crucial for improving our knowledge of drought and the timescale of MD lag to HD. vegetation vulnerability to climate fluctuations and climate (2) *e propagation time decreased 3 months at Ylx and change [41]. Meanwhile, with global warming, the propa- increased 8 months at Zyx compared to the Pre-R gation time of VCI and HD, VCI and MD, and HD and MD period and between VCI and SPI at 6 months in the to climate change also has changed [5], which poses new UHRB and 9 months in the MHRB. However, VCI challenges that need further research. Furthermore, in- with SDI at 6 months at Ylx and 3 months at Zyx did creasing human activity has dramatically changed natural significantly affect the growing season. On an intra- droughts; to manage drought effectively, we need to ac- annual scale, for Ylx, Zyx, and six tributaries, knowledge that human influence is as integral to drought as propagation times were different in the Pre-R and natural climate variability. *e complex interactions be- Pos-R periods, and there were seasonal differences tween natural and human processes need to be considered between Pos-R and Pre-R periods. [67]. In addition, there are still some issues worthy of in- (3) *e SDI and SPI showed no lag at Ylx and Zyx. depth discussion. Often, people are considered as passive However, VCI with MD had a significant lag corre- recipients at the end of the propagation cascade from MD lation at the short-term scale in the UHRB (6-month via soil moisture drought to HD [74] but, in fact, people lag) and MHRB (4-month lag), while the VCI with actively influence drought propagation [67]. Anthropogenic HD has significant 1-month lag correlation at Zyx. changes to the land surface alter hydrological processes, (4) Overall, the average peak/intensity/duration/severity including evapotranspiration, infiltration, surface stream- of MD and HD are weakening and the propagation flow, and storage of water [67, 75]. Understanding how time from MD to HD is also reduced for Pos-R irrigation [76] (the effect of long-term irrigation in the HRB period compared with Pre-R period in the UHRB. on the evolution of drought [77, 78]), land use/land cover Constructing reservoirs has had a positive effect, change [79], and construction of reservoirs or dams prolonging MD propagation to HD times, but the [42, 63, 64, 80, 81] impact MD, HD, and VCI characteristics drought characteristics (average peak/intensity/du- and MD propagation to HD and VCI need further research. ration/severity) have increased from the perspective Trnka et al. [82] presented a set of 60 priority questions and of climatology. *erefore, positive drought preven- considered interdisciplinary characteristics to optimize tion measures are necessary to consider the char- future drought research, primarily covering the following acteristics of propagation between MD and HD as drought-related topics: monitoring, impacts, drought well as seasonal differences in the MHRB. forecasts, climatology, adaptation, and planning. *erefore, in future drought research, it is necessary to comprehen- sively consider priority questions and interdisciplinary Data Availability characteristics, especially for irrigated agricultural areas in *e data used to support the findings of this study are arid regions. available from the corresponding author upon request. 5. Conclusions Conflicts of Interest Drought affects land surface dynamics, and the strongest *e authors declare that there are no conflicts of interest in impact of drought on vegetation occurs in arid and semiarid this paper. areas. *erefore, understanding the evolutionary charac- teristics and propagation of different drought types may Acknowledgments provide useful information about appropriate adaptation and mitigation strategies against the effects of drought on *is research was supported by the Strategic Priority Re- agricultural production and vegetation production and even search Program of the Chinese Academy of Sciences (Grant alleviate the losses caused by drought. Our results highlight nos. XDA19070502, XDA20100104, and XDA19040500), the the following points: National Natural Science Foundation of China (Grant nos. (1) More drought events occurred at shorter accumula- 41571516 and 41690144), and the Fundamental Research tion periods based on thresholds � −1.0, −1.5, and Funds for the Central Universities (Grant no. −2.0 for MD and HD. Compared to the Pre-R period, 2019jbkyjd013). 24 Advances in Meteorology [5] A. Dai, “Drought under global warming: a review,” Wiley Supplementary Materials Interdisciplinary Reviews: Climate Change, vol. 2, no. 1, pp. 45–65, 2011. Figure S1: boxplots showing MD peak and intensity in the [6] L. Wang and W. Chen, “A CMIP5 multimodel projection of UHRB ((a) and (c)) and in the MHRB ((b) and (d)) based future temperature, precipitation, and climatological drought on SPI using thresholds of −1, −1.5, and −2 for different in China,” International Journal of Climatology, vol. 34, no. 6, timescales at different periods (the dot and midline in the pp. 2059–2078, 2014. boxplots are the mean and median; the upper and lower [7] G. Leng, Q. Tang, and S. Rayburg, “Climate change impacts on boundaries of box are the interquartile ranges). Figure S2: meteorological, agricultural and hydrological droughts in boxplots showing HD peak and intensity at Ylx ((a) and (c)) China,” Global and Planetary Change, vol. 126, pp. 23–34, and at Zyx ((b) and (d)) based on SDI using thresholds of −1, −1.5, and −2 for different timescales at different periods [8] Z. Li, T. Zhou, X. Zhao et al., “Assessments of drought impacts (the dot and midline in the boxplots are the mean and on vegetation in China with the optimal time scales of the median; the upper and lower boundaries of box are the climatic drought index,” International Journal of Environ- interquartile ranges). Figure S3: the growing season spatial mental Research and Public Health, vol. 12, no. 7, pp. 7615– distribution of VCI during 2000–2001 in the UHRB and 7634, 2015. [9] J. Sheffield and E. F. Wood, “Projected changes in drought MHRB. Figure S4: heat map showing interannual Pearson’s occurrence under future global warming from multi-model, correlation coefficient of SPI accumulation periods of 1–48 multi-scenario, IPCC AR4 simulations,” Climate Dynamics, months with SDI-1 in the UHRB (Ylx) and MHRB (Zyx) vol. 31, no. 1, pp. 79–105, 2008. (note: the x-axis represents the correlation between the [10] H.-J. Xu, X.-P. Wang, C.-Y. Zhao, and X.-M. Yang, “Diverse SDI-1 and SPI at different timescales and the y-axis rep- responses of vegetation growth to meteorological drought resents the 12 months of the year; the up arrow represents across climate zones and land biomes in northern China from that the correlation coefficients are increased in the period 1981 to 2014,” Agricultural and Forest Meteorology, vol. 262, after EWDP in the UHRB (Ylx) and MHRB (Zyx) com- pp. 1–13, 2018. pared to the period before EWDP, or vice versa); (a) [11] M. Yu, Q. Li, M. J. Hayes, M. D. Svoboda, and R. R. Heim, (1965–1999, PC≥ 0.35, p< 0.05), (b) (2000–2012, PC≥ 0.57, “Are droughts becoming more frequent or severe in China p< 0.05) notes in the Ylx and (c) (1965–1999, PC≥ 0.35 based on the Standardized Precipitation Evapotranspiration Index: 1951–2010?” International Journal of Climatology, or≤ −0.38, p< 0.05), (d) (2000–2012, PC≥ 0.56, p< 0.05) vol. 34, no. 3, pp. 545–558, 2014. notes in the Zyx. Figure S5: heat maps showing PC between [12] L. Zhang, J. Xiao, Y. Zhou, Y. Zheng, J. Li, and H. Xiao, VCI lagged by 0–6 months and SPI (1–48)/SDI (1–12) “Drought events and their effects on vegetation productivity before (1965–1999) and after (2000–2012) EWDP; (a) in China,” Ecosphere, vol. 7, no. 12, Article ID e01591, 2016. (1965–1999, PC≥ 0.10, p< 0.05), (b) (2000–2012, PC≥ 0.16, [13] R. R. HeimJr., “Review of twentieth-century drought indices p< 0.05) in the UHRM (Ylx) and (c) (1965–1999, PC≥ 0.11 used in the United States,” Bulletin of the American Meteo- or PC≤ −0.11, p< 0.05), (d) (2000–2012, PC≥ 0.17, rological Society, vol. 83, pp. 1149–1166, 2002. p< 0.05) in the MHRB (Zyx). Figure S6: reservoirs under [14] T. J. Chang, “Investigation of precipitation droughts by use of construction and construction (a) irrigation system and kriging method,” Journal of Irrigation and Drainage Engi- area (b) in the MHRB. Table S1: identification of MD neering, vol. 117, no. 6, pp. 935–943, 1991. (threshold � −1, SPI-1) and HD (Zyx) characteristics [15] E. A. B. Eltahir, “Drought frequency analysis of annual rainfall series in central and western Sudan,” Hydrological Sciences (threshold � −1, SDI-1) based on run theory with com- Journal, vol. 37, no. 3, pp. 185–199, 1992. parison with historical drought records in MHRB. Table S2: [16] J. A. Dracup, K. S. Lee, and E. G. Paulson Jr., “On the statistical MD and HD drought characteristics for different timescales characteristics of drought events,” Water Resources Research, for different threshold for prereservoir period, post- vol. 16, no. 2, pp. 289–296, 1980. reservoir period, and total period in UHRB (Ylx) and [17] S. M. Quiring and S. Ganesh, “Evaluating the utility of the MHRB (Zyx). (Supplementary Materials) Vegetation Condition Index (VCI) for monitoring meteo- rological drought in Texas,” Agricultural and Forest Meteo- rology, vol. 150, no. 3, pp. 330–339, 2010. References [18] Q. Zhang, D. Kong, V. P. Singh, and P. Shi, “Response of vegetation to different time-scales drought across China: [1] S. E. Nicholson, C. J. Tucker, and M. B. Ba, “Desertification, spatiotemporal patterns, causes and implications,” Global and drought, and surface vegetation: an example from the west Planetary Change, vol. 152, pp. 1–11, 2017. african sahel,” Bulletin of the American Meteorological Society, [19] C. Li, W. Leal Filho, J. Yin et al., “Assessing vegetation re- vol. 79, no. 5, pp. 815–829, 1998. sponse to multi-time-scale drought across inner Mongolia [2] A. T. DeGaetano, “A temporal comparison of drought im- plateau,” Journal of Cleaner Production, vol. 179, pp. 210–216, pacts and responses in the New York city metropolitan area,” Climatic Change, vol. 42, no. 3, pp. 539–560, 1999. [20] Y. Zhang, J. Gao, L. Liu, Z. Wang, M. Ding, and X. Yang, [3] A. K. Mishra and V. P. Singh, “A review of drought concepts,” “NDVI-based vegetation changes and their responses to cli- Journal of Hydrology, vol. 391, no. 1-2, pp. 202–216, 2010. mate change from 1982 to 2011: a case study in the Koshi [4] S. Huang, P. Li, Q. Huang, G. Leng, B. Hou, and L. Ma, “*e River Basin in the middle Himalayas,” Global and Planetary propagation from meteorological to hydrological drought and Change, vol. 108, pp. 139–148, 2013. its potential influence factors,” Journal of Hydrology, vol. 547, [21] A. Zhao, A. Zhang, S. Cao, X. Liu, J. Liu, and D. Cheng, pp. 184–195, 2017. “Responses of vegetation productivity to multi-scale drought Advances in Meteorology 25 in Loess Plateau, China,” CATENA, vol. 163, pp. 165–171, basin, China,” Water Resources Management, vol. 28, no. 10, 2018. pp. 3095–3110, 2014. [22] B. Lloyd-Hughes, “*e impracticality of a universal drought [38] A. Mondal and P. P. Mujumdar, “Return levels of hydrologic definition,” ;eoretical and Applied Climatology, vol. 117, droughts under climate change,” Advances in Water Re- no. 3-4, pp. 607–611, 2014. sources, vol. 75, pp. 67–79, 2015. [23] L. J. Barker, J. Hannaford, A. Chiverton, and C. Svensson, [39] J. Niu, J. Chen, and L. Sun, “Exploration of drought evolution “From meteorological to hydrological drought using stand- using numerical simulations over the Xijiang (West River) ardised indicators,” Hydrology and Earth System Sciences, basin in South China,” Journal of Hydrology, vol. 526, vol. 20, no. 6, pp. 2483–2505, 2016. pp. 68–77, 2015. [24] B. T. Mckee, J. Nolan, and J. Kleist, “*e relationship of ´ [40] J. Lorenzo-Lacruz, S. Vicente-Serrano, J. Gonzalez-Hidalgo, drought frequency and duration to time scales,” in Pro- J. Lopez-Moreno, ´ and N. Cortesi, “Hydrological drought ceedings of the 8th Conference on Applied Climatology, response to meteorological drought in the Iberian Peninsula,” vol. 17–22, pp. 179–184, Anaheim, CA, USA, January 1993. Climate Research, vol. 58, no. 2, pp. 117–131, 2013. [25] D. Wang, M. Hejazi, X. Cai, and A. J. Valocchi, “Climate [41] S. M. Vicente-Serrano, C. Gouveia, J. J. Camarero et al., change impact on meteorological, agricultural, and hydro- “Response of vegetation to drought time-scales across global logical drought in central Illinois,” Water Resources Research, land biomes,” Proceedings of the National Academy of Sci- vol. 47, no. 9, 2011. ences, vol. 110, no. 1, pp. 52–57, 2013. [26] A. AghaKouchak and N. Nakhjiri, “A near real-time satellite- [42] A. F. Van Loon, M. H. J. Van Huijgevoort, and based global drought climate data record,” Environmental H. A. J. Van Lanen, “Evaluation of drought propagation in an Research Letters, vol. 7, no. 4, p. 044037, 2012. ensemble mean of large-scale hydrological models,” Hy- [27] J. H. Stagge, L. M. Tallaksen, L. Gudmundsson, drology and Earth System Sciences, vol. 16, no. 11, pp. 4057– A. F. Van Loon, and K. Stahl, “Candidate distributions for 4078, 2012. climatological drought indices (SPI and SPEI),” International [43] M. Peña-Gallardo, S. M. Vicente-Serrano, J. Hannaford et al., Journal of Climatology, vol. 35, no. 13, pp. 4027–4040, 2015. “Complex influences of meteorological drought time-scales [28] I. Nalbantis and G. Tsakiris, “Assessment of hydrological on hydrological droughts in natural basins of the contiguous drought revisited,” Water Resources Management, vol. 23, Unites States,” Journal of Hydrology, vol. 568, pp. 611–625, no. 5, pp. 881–897, 2009. 2019. [29] S. Li, L. Xiong, L. Dong, and J. Zhang, “Effects of the three [44] F. Ma, L. Luo, A. Ye, and Q. Duan, “Drought characteristics gorges reservoir on the hydrological droughts at the down- and propagation in the semiarid Heihe river basin in stream yichang station during 2003–2011,” Hydrological northwestern China,” Journal of Hydrometeorology, vol. 20, Processes, vol. 27, no. 26, pp. 3981–3993, 2013. no. 1, pp. 59–77, 2019. [30] H. Tabari, J. Nikbakht, and P. Hosseinzadeh Talaee, “Hy- [45] J. Niu, S. Kang, X. Zhang, and J. Fu, “Vulnerability analysis drological drought assessment in northwestern Iran based on based on drought and vegetation dynamics,” Ecological In- streamflow drought index (SDI),” Water Resources Man- dicators, vol. 105, pp. 329–336, 2019. agement, vol. 27, no. 1, pp. 137–151, 2013. [46] R. P. Pandey and K. S. Ramasastri, “Relationship between the [31] E. Rimkus, E. Stonevicius, ˇ V. Korneev, J. Kazys, ˇ common climatic parameters and average drought fre- G. Valiuˇskevicius, ˇ and A. Pakhomau, “Dynamics of meteo- quency,” Hydrological Processes, vol. 15, no. 6, pp. 1019–1032, rological and hydrological droughts in the Neman river ba- 2001. sin,” Environmental Research Letters, vol. 8, no. 4, p. 045014, [47] X. Li, G. Cheng, Y. Ge et al., “Hydrological cycle in the Heihe river basin and its implication for water resource management [32] T. Fischer, M. Gemmer, B. Su, and T. Scholten, “Hydrological in endorheic basins,” Journal of Geophysical Research: At- long-term dry and wet periods in the Xijiang River basin, mospheres, vol. 123, no. 2, pp. 890–914, 2018. [48] N. Zhao, T. Yue, C. Chen, M. Zhao, and Z. Fan, “An improved South China,” Hydrology and Earth System Sciences, vol. 17, no. 1, pp. 135–148, 2013. statistical downscaling scheme of Tropical Rainfall Measuring [33] X. Hong, S. Guo, Y. Zhou, and L. Xiong, “Uncertainties in Mission precipitation in the Heihe River basin, China,” In- assessing hydrological drought using streamflow drought ternational Journal of Climatology, vol. 38, no. 8, pp. 3309– index for the upper Yangtze River basin,” Stochastic Envi- 3322, 2018. ronmental Research and Risk Assessment, vol. 29, no. 4, [49] A. Zhang, C. Zheng, S. Wang, and Y. Yao, “Analysis of pp. 1235–1247, 2015. streamflow variations in the Heihe River Basin, northwest [34] A.-A. Belal, H. R. El-Ramady, E. S. Mohamed, and China: trends, abrupt changes, driving factors and ecological A. M. Saleh, “Drought risk assessment using remote sensing influences,” Journal of Hydrology: Regional Studies, vol. 3, and GIS techniques,” Arabian Journal of Geosciences, vol. 7, pp. 106–124, 2015. no. 1, pp. 35–53, 2014. [50] J. Niu, X.-G. Zhu, M. A. J. Parry et al., “Environmental [35] Q. Lin, Z. Wu, V. P. Singh, S. H. R. Sadeghi, H. He, and G. Lu, burdens of groundwater extraction for irrigation over an “Correlation between hydrological drought, climatic factors, inland river basin in Northwest China,” Journal of Cleaner reservoir operation, and vegetation cover in the Xijiang Basin, Production, vol. 222, pp. 182–192, 2019. South China,” Journal of Hydrology, vol. 549, pp. 512–524, [51] Y. Ge, X. Li, C. Huang, and Z. Nan, “A decision support 2017. system for irrigation water allocation along the middle reaches [36] I. Cordery and M. McCall, “A model for forecasting drought of the Heihe river basin, northwest China,” Environmental from teleconnections,” Water Resources Research, vol. 36, Modelling & Software, vol. 47, pp. 182–192, 2013. no. 3, pp. 763–768, 2000. [52] L. Qiu, D. Peng, Z. Xu, and W. Liu, “Identification of the [37] S. Huang, J. Chang, Q. Huang, and Y. Chen, “Spatio-temporal impacts of climate changes and human activities on runoff in changes and frequency analysis of drought in the wei river the upper and middle reaches of the Heihe River basin, 26 Advances in Meteorology China,” Journal of Water and Climate Change, vol. 7, no. 1, [69] J. I. Lopez-Moreno, ´ S. M. Vicente-Serrano, J. Zabalza et al., “Hydrological response to climate variability at different time pp. 251–262, 2015. [53] X. Deng and C. Zhao, “Identification of water scarcity and scales: a study in the Ebro basin,” Journal of Hydrology, vol. 477, pp. 175–188, 2013. providing solutions for adapting to climate changes in the Heihe river basin of China,” Advances in Meteorology, [70] X.-P. Deng, L. Shan, H. Zhang, and N. C. Turner, “Improving agricultural water use efficiency in arid and semiarid areas of vol. 2015, Article ID 279173, 13 pages, 2015. [54] M. Zhang, S. Wang, B. Fu, G. Gao, and Q. Shen, “Ecological China,” Agricultural Water Management, vol. 80, no. 1–3, pp. 23–40, 2006. effects and potential risks of the water diversion project in the [71] X. Xu, Y. Jiang, M. Liu, Q. Huang, and G. Huang, “Modeling Heihe River Basin,” Science of the Total Environment, vol. 619- and assessing agro-hydrological processes and irrigation 620, pp. 794–803, 2018. water saving in the middle Heihe River basin,” Agricultural [55] A. A. Paulo, L. S. Pereira, and P. G. Matias, “Analysis of local Water Management, vol. 211, pp. 152–164, 2019. and regional droughts in southern Portugal using the theory [72] J. Niu, J. Chen, K. Wang, and B. Sivakumar, “Multi-scale of runs and the standardised precipitation index,” in Tools for streamflow variability responses to precipitation over the Drought Mitigation in Mediterranean Regions, G. Rossi, headwater catchments in southern China,” Journal of Hy- A. Cancelliere, L. S. Pereira, T. Oweis, M. Shatanawi, and drology, vol. 551, pp. 14–28, 2017. A. Zairi, Eds., Springer, Dordrecht, Netherlands, 2003. [73] H. Liu, M. Zhang, Z. Lin, and X. Xu, “Spatial heterogeneity of [56] F. Kogan and J. Sullivan, “Development of global drought- the relationship between vegetation dynamics and climate watch system using NOAA/AVHRR data,” Advances in Space change and their driving forces at multiple time scales in Research, vol. 13, no. 5, pp. 219–222, 1993. Southwest China,” Agricultural and Forest Meteorology, [57] J. Spinoni, T. Antofie, P. Barbosa et al., “An overview of vol. 256-257, pp. 10–21, 2018. drought events in the Carpathian Region in 1961–2010,” [74] R. Orth and G. Destouni, “Drought reduces blue-water fluxes Advances in Science and Research, vol. 10, no. 1, pp. 21–32, more strongly than green-water fluxes in Europe,” Nature Communications, vol. 9, no. 1, p. 3602, 2018. [58] V. M. Yevjevich, An Objective Approach to Definitions and [75] N. S. Diffenbaugh, D. L. Swain, and D. Touma, “Anthropo- Investigations of Continental Hydrologic Droughts, Colorado genic warming has increased drought risk in California,” State University, Fort Collins, CO, USA, 1967. Proceedings of the National Academy of Sciences, vol. 112, [59] Y. L. Zhou, P. Zhou, J. L. Jin, and J. Li, “Establishment of no. 13, pp. 3931–3936, 2015. hydrological drought index on sources of regional water [76] P. Droogers, W. G. M. Bastiaanssen, M. Beyazgul, ¨ Y. Kayam, supply and its application to drought frequency analysis for G. W. Kite, and H. Murray-Rust, “Distributed agro-hydro- Kunming,” Journal of Hydraulic Engineering, vol. 45, logical modeling of an irrigation system in western Turkey,” pp. 1038–1047, 2014. Agricultural Water Management, vol. 43, no. 2, pp. 183–202, [60] J. Wu, Z. Liu, H. Yao et al., “Impacts of reservoir operations on multi-scale correlations between hydrological drought and [77] Y. Chen, J. Niu, S. Kang, and X. Zhang, “Effects of irrigation meteorological drought,” Journal of Hydrology, vol. 563, on water and energy balances in the Heihe River basin using pp. 726–736, 2018. VIC model under different irrigation scenarios,” Science of [61] S. Mitra and P. Srivastava, “Spatiotemporal variability of ;e Total Environment, vol. 645, pp. 1183–1193, 2018. meteorological droughts in southeastern USA,” Natural [78] J. Niu, Q. Liu, S. Kang, and X. Zhang, “*e response of crop Hazards, vol. 86, no. 3, pp. 1007–1038, 2017. water productivity to climatic variation in the upper-middle [62] L. Gao and Y. Zhang, “Spatio-temporal variation of hydro- reaches of the Heihe River basin, Northwest China,” Journal of logical drought under climate change during the period Hydrology, vol. 563, pp. 909–926, 2018. 1960–2013 in the Hexi Corridor, China,” Journal of Arid [79] B. I. Cook, R. L. Miller, and R. Seager, “Amplification of the Land, vol. 8, no. 2, pp. 157–171, 2016. North American “Dust Bowl” drought through human-in- [63] L. Yang, Q. Feng, Z. Yin et al., “Identifying separate impacts of duced land degradation,” Proceedings of the National Acad- climate and land use/cover change on hydrological processes emy of Sciences, vol. 106, no. 13, pp. 4997–5001, 2009. in upper stream of Heihe River, Northwest China,” Hydro- [80] Q. Huang, Z. Sun, C. Opp, T. Lotz, J. Jiang, and X. Lai, logical Processes, vol. 31, no. 5, pp. 1100–1112, 2017. “Hydrological drought at dongting lake: its detection, char- [64] Z. Cong, M. Shahid, D. Zhang, H. Lei, and D. Yang, “At- acterization, and challenges associated with three gorges dam tribution of runoff change in the alpine basin: a case study of in central yangtze, China,” Water Resources Management, the Heihe Upstream Basin, China,” Hydrological Sciences vol. 28, no. 15, pp. 5377–5388, 2014. Journal, vol. 62, no. 6, pp. 1013–1028, 2017. [81] S. Rangecroft, A. F. Van Loon, H. Maureira, K. Verbist, and [65] Q. Cheng, X. Zuo, F. Zhong, L. Gao, and S. Xiao, “Runoff D. M. Hannah, “Multi-method assessment of reservoir effects variation characteristics, association with large-scale circu- on hydrological droughts in an arid region,” Earth System lation and dominant causes in the Heihe River Basin, Dynamics Discussion, pp. 1–32, 2016. Northwest China,” Science of the Total Environment, vol. 688, [82] M. Trnka, M. Hayes, F. Jurecka ˇ et al., “Priority questions in pp. 361–379, 2019. multidisciplinary drought research,” Climate Research, [66] K. G. Wen, China Meteorological Disaster (Gansu Volume), vol. 75, no. 3, pp. 241–260, 2018. China Meteorological, Beijing, China, 2006, In Chinese. [67] A. F. Van Loon, T. Gleeson, J. Clark et al., “Drought in the anthropocene,” Nature Geoscience, vol. 9, no. 2, pp. 89–91, [68] G. Zhu, Y. Su, C. Huang, Q. Feng, and Z. Liu, “Hydro- geochemical processes in the groundwater environment of Heihe River Basin, northwest China,” Environmental Earth Sciences, vol. 60, no. 1, pp. 139–153, 2010.

Journal

Advances in MeteorologyHindawi Publishing Corporation

Published: Jun 9, 2020

References