1. Introduction
Drought is a progressive and recurrent natural disaster (Dai 2011; Hao and Singh 2015). Drought may produce substantial consequences on ecosystems, the environment, and society (Kallis 2008; Su et al. 2018). In the context of global warming, droughts will be likely to occur more frequently and at more severe magnitudes (Dai 2013). Many droughts and their triggering mechanisms have been studied to date (e.g., Mo et al. 2009; Wood et al. 2015; Cook et al. 2016).
The spatiotemporal characteristics of droughts provide a significant primer for drought research. The characteristics of droughts are different in the temporal dimension and spatial dimension, which are critical to investigate drought development and its underlying mechanism. In contrast to the focus on temporal dimensions in previous drought studies (e.g., Livada and Assimakopoulos 2007; Koutroulis et al. 2011; Li et al. 2015), the spatial characteristics and development of droughts have received increasing attention in the past decade (e.g., Rodriguez et al. 2016; Herrera-Estrada et al. 2017). Climate modes of droughts over a region, which are likely to be produced by the superposition of climate anomalies on interannual and interdecadal scales, are often used to identify homogenous drought regions that can be considered new clues for exploring the spatial characteristics of droughts (Rajsekhar et al. 2013; Fu et al. 2018; Zhou and Liu 2018). For example, Vicente-Serrano et al. (2004) divided the Valencia region into four climate modes and analyzed their temporal variations. Santos et al. (2010) determined three relatively well-defined drought regions and their variabilities in mainland Portugal. Li et al. (2015) identified eight climate divisions of homogeneous drought variation throughout China and investigated their trends and temporal characteristics. Although these works highlight the importance of climate modes in drought studies, most of them pay more attention to the time series analysis of climate modes and mainly report the regional statistical results of droughts (Herrera-Estrada et al. 2017).
Currently, several studies have considered a drought cluster as an object entity and attempted to focus on the spatiotemporal dynamics and characteristics. Lloyd-Hughes (2012) established a three-dimensional (time, longitude, and latitude) structure-based method to investigate the evolution of individual drought events over Europe. Herrera-Estrada et al. (2017) identified 1420 drought clusters during 1979–2009 throughout the world and analyzed the migration of their centroids. Guo et al. (2018) investigated the space–time structure and characteristics of drought events in Central Asia using an improved three-dimensional clustering algorithm. Liu et al. (2019) explored the characteristics of propagation from meteorological to hydrological droughts based on the spatiotemporal linkages between the different independent drought event types in the Yellow River basin. Zhou et al. (2019) proposed an approach to study the migrating characteristics of meteorological droughts based on observations and tested it in the Poyang Lake basin of China. These studies on the spatiotemporal evolutionary processes of droughts offer some implications for understanding drought development. Can the evolution of droughts be characterized and understood from a perspective of climate modes? The issue remains largely unexplored. Investigating the migration rules where droughts change from one mode to another may provide new clues for understanding drought development mechanisms and for drought management.
Hence, this study aims to identify the climate modes of meteorological droughts and to analyze their role in the spatiotemporal evolution of drought events. The Poyang Lake basin, China, is taken as a case study using monthly precipitation data from 73 meteorological stations from 1960 to 2007. The dominant spatial modes of meteorological droughts in the basin are identified using principal component analysis (PCA). The optimal spatial modes are then determined by using K-means cluster analysis (KCA) and one-way analysis of variance (ANOVA) (Scott and Knott 1974; Wilks 2006). Then, an identified spatial mode is viewed as an elementary object entity to characterize drought evolution in space and time. Based on the trajectory migration identification method proposed by Zhou et al. (2019), the spatiotemporal evolutional characteristics of drought events are analyzed from the perspective of climate modes. Finally, the role of spatial modes in drought evolution is investigated and discussed. This study contributes to understanding drought development in both spatial and temporal dimensions.
2. Materials and methods
a. Study area and data
The Poyang Lake basin, covering an area of 1.62 × 105 km2, is located in southeastern China (24°29′–30°04′N and 113°34′–118°28′E). The basin belongs to the humid subtropical climate zone (Fig. 1a). It contains five subbasins (Xiushui, Ganjiang, Fuhe, Xinjiang, and Raohe) and a lake area (Poyang Lake, the largest freshwater lake in China) (Fig. 1b). Water from the five subbasins flows into the lake through rivers. Topographically, the basin consists of mountains and hills, with plains in the northern region of the basin. The surface elevation of the basin ranges from 32 to 2200 m (Fig. 1b). The mean annual precipitation from 1960 to 2007 is approximately 1645 mm. The precipitation is greater in the north than in the south, in the east than in the west, and in the mountainous areas than in the plains.
(a) The geographical location of the Poyang Lake basin, China, and (b) the spatial distribution of the 73 meteorological stations and the elevation across the basin.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0183.1
(a) The geographical location of the Poyang Lake basin, China, and (b) the spatial distribution of the 73 meteorological stations and the elevation across the basin.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0183.1
(a) The geographical location of the Poyang Lake basin, China, and (b) the spatial distribution of the 73 meteorological stations and the elevation across the basin.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0183.1
The basin has a high value in terms of ecosystem services such as water conservation and biodiversity maintenance (Liu et al. 2012). Moreover, in a global context, the basin is a component of the north–south transect of eastern China and is included in the Global Change and Terrestrial Ecosystems project of the International Geosphere–Biosphere Programme (Canadell et al. 2002). In recent decades, extreme hydrometeorological events (i.e., floods and droughts) have occurred more frequently in the basin (Shankman et al. 2006; Ye et al. 2013; Liu and Wu 2016; Liu et al. 2016). For example, in 1978, most parts of the basin suffered from an exceptional drought. The drought event greatly affected approximately 13 million people and resulted in one million tons of reduced agricultural production, and made most of the stored water (e.g., ponds, rivers, and reservoirs) dry up (Yin et al. 2011; Zhang and Zhou 2015). Hence, an investigation of the spatiotemporal evolution of meteorological droughts is significant for drought management in the basin.
The study data can be obtained from the National Meteorological Information Centre of the China Meteorological Administration. The dataset, through strict quality control (China Meteorological Administration 2003), contains monthly precipitation data from 73 meteorological stations for the entire basin from 1960 to 2007 (Fig. 1b).
b. Methodology
1) Overview framework of the study
This study is an extension of our previous works (Zhou and Liu 2018; Zhou et al. 2019). An overview framework of the associated studies is given to emphasize the specific content and importance of this study (Fig. 2).
The framework of this study and its connection with the previous studies. Gray boxes represent basic works, including data processing and analysis. Light blue boxes represent the related works that have been addressed. (a) Gold boxes represent the main contents of this study. (b)–(f) A schematic diagram of drought migration among climate modes.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0183.1
The framework of this study and its connection with the previous studies. Gray boxes represent basic works, including data processing and analysis. Light blue boxes represent the related works that have been addressed. (a) Gold boxes represent the main contents of this study. (b)–(f) A schematic diagram of drought migration among climate modes.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0183.1
The framework of this study and its connection with the previous studies. Gray boxes represent basic works, including data processing and analysis. Light blue boxes represent the related works that have been addressed. (a) Gold boxes represent the main contents of this study. (b)–(f) A schematic diagram of drought migration among climate modes.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0183.1
To explore the spatiotemporal dynamics of meteorological droughts from the perspective of climate modes, two preliminary works have been completed (Fig. 2). One work determined the dominated climate modes of meteorological droughts over the basin and investigated their physical meanings. Since some climate modes with implicit physics can be misleading, interpreting the physical meaning of climate modes appears necessary and significant before further application (Dommenget and Latif 2002; Li et al. 2015). The other work developed an approach to tracking space–time evolutionary processes of meteorological droughts. The two works provide significant support for this study: 1) the dominant climate modes with physical meanings and 2) the approach to quantifying the evolutionary characteristics of drought events. This study investigates the migration characteristics of meteorological droughts based on climate modes. Figures 2b–f describe a drought process migrating among climate modes.
2) Standardized precipitation index
In recent decades, many indices have been proposed to quantify meteorological droughts (Zargar et al. 2011). Among them, the standardized precipitation index (SPI) (McKee et al. 1993) has been widely applied (e.g., Zarch et al. 2015; Stagge et al. 2015; Zhou and Liu 2016) and is recommended by the World Meteorological Organization as a standard index (Hayes et al. 2011). This study applied SPI at a 1-month scale to quantify meteorological droughts. Its detailed procedures are introduced by Zhou and Liu (2016).
The meteorological drought categories based on the SPI ranges are given in Table 1.
Dry period categories based on SPI values.
According to Table 1, an SPI value of −1 is often used as the threshold to identify droughts. For a region, a regional averaged SPI value less than −1 will be considered to be under drought, which is confirmed as an effective scheme when dry conditions occur widely throughout the region (Zhou and Liu 2016). In this study, a drought month is defined as a month in which the SPI value is ≤−1. A drought event is defined as a period in which the SPI value is continuously ≤−1 for at least two consecutive months.
3) Statistical methods for identifying drought climate modes
(i) Principal component analysis
Furthermore, to more clearly identify localization, the eigenvectors are often subjected to a varimax rotation. The rotated PCs can present more stable spatial modes of meteorological droughts. A (climate or spatial) mode is considered a type of distinct spatial distribution of meteorological drought variability (Hannachi et al. 2007).
(ii) K-means cluster analysis
A cluster analysis is used to divide data into homogeneous groups based on the degree of similarity among objects. The K-means cluster analysis (KCA) can guarantee the minimum variability within clusters and the maximum variability among clusters (Hartigan and Wong 1979; Santos et al. 2013). A cluster is regarded as a set of similar objects (i.e., meteorological stations). Hence, combined with the results of the PCA, the KCA method with Euclidean distance is applied to investigate classification groups of meteorological stations. An n-classification group is a set of n clusters containing all stations. For example, a 2-classification group means that there are two clusters in it.
(iii) Analysis of variance
ANOVA is often used to assess whether the difference in the mean values of one factor at different levels is statistically significant or not. In this study, ANOVA was applied to evaluate the appropriateness of classification groups. The distance between a station and its corresponding cluster center is computed, following Eq. (4).
In ANOVA, the total variance is split into two types: the variance between clusters and the variance within clusters. Within clusters, the variance is due to random error whereas systematic differences should explain the variance between clusters. When the ratio of the between-cluster variance to within-cluster variance, or F ratio, becomes larger it is more likely that the means are significantly different. The 5% significance level is chosen to estimate whether the difference of means is significant.
Furthermore, according to the identified optimal classification group, the spatial modes of meteorological droughts over the given region can be obtained.
4) Trajectory migration identification method
In this study, an identified spatial mode is considered an object entity. The trajectory migration identification method (TMIM) proposed by Zhou et al. (2019) is used to investigate the spatiotemporal evolutionary features of meteorological droughts based on topological spatial relations. The approach is applied to identify drought clusters, migration trajectories, and migration directions of meteorological droughts. A drought cluster is defined as a spatially contiguous area under drought; in this study, the area may consist of one or more climate modes.
There is a change in the identification of drought clusters. In the original approach, a threshold of 1° (approximately 100 km) is used as an aggregation criterion when two drought clusters occur. Here, the spatial location–relation between two spatial modes will be judged first. If it is the spatial adjacency relation, the two spatial modes will be merged as a new drought cluster; if it is the spatial disjoint relation, the two spatial modes will be considered as two independent drought clusters.
The migration trajectory and migration direction identification of meteorological drought events are used directly. The detailed procedures of the approach are introduced by Zhou et al. (2019).
3. Results
a. Meteorological drought characteristics over the basin from 1960 to 2007
The spatiotemporal characteristics of meteorological droughts from 1960 to 2007 over the basin were investigated (Fig. 3). As indicated in Fig. 3a, the number of drought months ranged from 76 to 102 months, which accounted for 13.2%–17.7% of the total months (n = 576). More drought months were mainly distributed in the north-central part of the basin. In comparison to the drought months, the number of drought events was much lower, and more drought events mainly occurred in the central part of the basin (Fig. 3b). The mean severity (ranging from 3.09 to 4.11) and mean duration (ranging from 2.00 to 2.64 months) of drought events showed small fluctuating ranges, which indicated that most of them were short-term events (Figs. 3c,d). Moreover, they had a similar spatial pattern characterized by the distribution of the high values in the southeast and low values in the northwest. With respect to the annual mean drought area, the historical series generally revealed a decreasing trend (−1.5% decade−1), but it was statistically insignificant (Fig. 3e). From the series, it was found that the basin suffered from some widespread dry situations, for example, in 1962–64, 1971, and 2007 (e.g., Min et al. 2013).
The spatial and temporal characteristics of meteorological droughts over the Poyang Lake basin. The figure shows the (a) number of drought months, (b) the number of drought events, (c) the mean severity and (d) duration of drought events, and (e) the annual mean drought area.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0183.1
The spatial and temporal characteristics of meteorological droughts over the Poyang Lake basin. The figure shows the (a) number of drought months, (b) the number of drought events, (c) the mean severity and (d) duration of drought events, and (e) the annual mean drought area.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0183.1
The spatial and temporal characteristics of meteorological droughts over the Poyang Lake basin. The figure shows the (a) number of drought months, (b) the number of drought events, (c) the mean severity and (d) duration of drought events, and (e) the annual mean drought area.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0183.1
b. Climate modes identification of meteorological droughts
Figure 4 shows the first five modes and their corresponding contributions to the total spatiotemporal variability. The first four modes individually explained 27.9%, 23.9%, 17.3%, and 13.5% of the total variance, or 82.6% of the variance collectively. The fifth mode explained only 1.7% of the variance. In the study of Zhou and Liu (2018), the physical meaning of these first five climate modes was analyzed. The results showed the four leading climate modes were likely to be produced by multiple climate anomalies with different time scales, while the fifth mode with little variance might be mainly composed of noise, delivering little physical evidence. Hence, the first four climate modes were further applied to identify the spatial modes.
Spatial distribution of the loadings of the first five modes in the Poyang Lake basin.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0183.1
Spatial distribution of the loadings of the first five modes in the Poyang Lake basin.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0183.1
Spatial distribution of the loadings of the first five modes in the Poyang Lake basin.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0183.1
The high loadings (>0.5) shown in Fig. 4 indicate that the respective mode had a significant correlation with the stations. The high loadings in mode 1 were mainly located in the northern lake area, the Xiushui subbasin, and the Raohe subbasin (Fig. 4a). The peak loadings occurred in the northern region of the basin, and they decreased in a southward direction in the basin. With regard to mode 2, high loadings were mainly located in the southern Ganjiang subbasin (Fig. 4b). Nevertheless, the areas with high loadings in mode 1 and mode 2 did not overlap each other. The high loadings in mode 3 and mode 4, however, were mainly located in the Ganjiang subbasin and the Xinjiang subbasin (Figs. 4c,d). The loadings decreased northward and southward for both modes. Note that all the loadings were positive for the first four modes, indicating similar signs of dryness or wetness. In contrast, mode 5 exhibited a complicated pattern with four poles, while the loadings ranged between −0.26 and 0.23, showing that those stations had a limited contribution to mode 5 (Fig. 4e). Overall, the whole basin could be roughly categorized into four spatial regions: the northern, southern, central, and northeastern basins.
Figure 5 shows the 2–5-classification groups and the evaluation of their appropriateness. Combined with Fig. 4, these classification groups were compared to the spatial distribution of the stations with high loadings. It should be noted that the classification groups were similar to the spatial distributions of the high loadings in the modes. For the 2-classification group (Fig. 5a), the region with loadings greater than 0.4 (Fig. 4a) was determined to correspond closely with cluster 1. Similarly, the region with loadings greater than 0.4 (Fig. 4b) was mostly consistent with the spatial coverage of cluster 2. Furthermore, combined with that shown in Figs. 4a–d, the extents of clusters 1–4 in the 4-classification group were in good agreement with the regions, with loadings greater than 0.7, 0.7, 0.5, and 0.5, respectively. However, for the 5-classification group, cluster 1 corresponded to the scope of the loadings from 0.1 to 0.3 (Fig. 4e).
The (a) 2-, (b) 3-, (c) 4-, and (d) 5-classification groups based on the SPI series of the 73 meteorological stations in the Poyang Lake basin from 1960 to 2007. (e)–(h) The distances between each station and the corresponding cluster centers are plotted with box-and-whisker diagrams. The red line in each box represents the median of one cluster. The hollow circle in each box represents the mean value of one cluster. The bottom and top edges of the boxes indicate the 25th and 75th percentiles, respectively; meanwhile, the top and bottom of each whisker represent the most extreme data values.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0183.1
The (a) 2-, (b) 3-, (c) 4-, and (d) 5-classification groups based on the SPI series of the 73 meteorological stations in the Poyang Lake basin from 1960 to 2007. (e)–(h) The distances between each station and the corresponding cluster centers are plotted with box-and-whisker diagrams. The red line in each box represents the median of one cluster. The hollow circle in each box represents the mean value of one cluster. The bottom and top edges of the boxes indicate the 25th and 75th percentiles, respectively; meanwhile, the top and bottom of each whisker represent the most extreme data values.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0183.1
The (a) 2-, (b) 3-, (c) 4-, and (d) 5-classification groups based on the SPI series of the 73 meteorological stations in the Poyang Lake basin from 1960 to 2007. (e)–(h) The distances between each station and the corresponding cluster centers are plotted with box-and-whisker diagrams. The red line in each box represents the median of one cluster. The hollow circle in each box represents the mean value of one cluster. The bottom and top edges of the boxes indicate the 25th and 75th percentiles, respectively; meanwhile, the top and bottom of each whisker represent the most extreme data values.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0183.1
In addition, the distances between stations and their corresponding cluster centers for four classification groups were plotted by box-and-whisker diagrams (Figs. 5e–h). There were few differences between the medians and means of the 2- and 5-classification groups (especially for the clusters 2, 3, and 5 in the 5-classification group), while there was a larger difference between the 3- and 4-classification groups. In comparison to Figs. 5b and 5c, cluster 1 in the 3-classification group was roughly divided into clusters 1 and 3 shown in the 4-classification group, while these two clusters showed large differences in the median and mean values (Fig. 5g).
Furthermore, to assess the optimal classification group, the differences among clusters within classification groups were examined. The significance analysis showed that the 3-classification group (p = 0.012) and the 4-classification group (p = 0.014) passed the significance test (p < 0.05), which indicated that these two classification groups were more appropriate. Combined with this finding with the performance of the PCA, the 4-classification group was selected to obtain the spatial modes. In this classification group, clusters 1–4 included 23, 19, 16, and 15 meteorological stations respectively. Moreover, the four spatial modes could be considered the northern (N), central (C), east-central (EC), and southern (S) regions of the basin, respectively. These identified spatial modes with distinctive physical meaning should be able to contribute to a better understanding of how drought varies regionally (Fovell 1997; Bieniek et al. 2012; Li et al. 2015).
c. Meteorological droughts in spatial modes
The characteristics of meteorological droughts based on the four spatial modes were investigated. A total of 141 drought months and 144 drought clusters were identified from 1960 to 2007. Two drought clusters were identified in November 1971, November 1994, and June 2004, while only one was identified at other times. Figure 6 gives the spatial distribution of the centroids of 144 drought clusters. The number of drought clusters in the 1960s and 1980s was 34 and 33, respectively, which was more than in other periods. The result seemed to indicate that there had a decrease in the number of droughts over the basin. Although there were differences in the number of drought clusters for five consecutive decades, the spatial distributions of their drought centroids, characterized by the zonal pattern from northeast to southwest, were basically consistent.
The spatial distribution of centroids for all drought clusters in different decades. A circle represents the location of the centroid of a drought cluster; N denotes the number of centroids.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0183.1
The spatial distribution of centroids for all drought clusters in different decades. A circle represents the location of the centroid of a drought cluster; N denotes the number of centroids.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0183.1
The spatial distribution of centroids for all drought clusters in different decades. A circle represents the location of the centroid of a drought cluster; N denotes the number of centroids.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0183.1
As indicated in Fig. 7, the high variability of the drought events was shown in the four spatial modes. The drought events in mode N mainly occurred in the 1960s and 1970s, and there were few events after the 1970s. Modes C and EC were basically similar, and most of the events were uniformly distributed during the 1960s to the 1990s. For mode S, drought events were mainly detected in the 1960s and 1990s. Although mode N and mode C were adjacent, they had large differences in drought characteristics. The fewest drought events and the shortest mean drought duration were identified in mode N (EN = 7 and DN = 2.14 ± 0.38 months), while it detected the largest mean drought severity (SN = 3.46 ± 1.10). Mode C was the opposite, displaying the most drought events (EC = 15) and the smallest mean drought severity (SC = 3.25 ± 0.83). In addition, it should be noted that some drought events were recorded by the four modes at the same time, for example, the 1962–63 drought. The distinct differences in the spatial and temporal characteristics of drought events for the four modes provided some implications to investigate the evolutionary processes of droughts in the space–time dimension simultaneously, which is significant for understanding the causes and potential migrating rules.
The temporal evolution of the meteorological drought events in different spatial modes: EN, EC, EEC, and ES denote the number of drought events in modes N, C, EC, and S, respectively; DN, DC, DEC, and DS denote the mean duration (month) of the drought events in modes N, C, EC, and S, respectively; and SN, SC, SEC, and SS denote the mean severity of the drought events in modes N, C, EC, and S, respectively. The red circles represent the moment suffering from drought. The number of circles represents the duration of a drought event.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0183.1
The temporal evolution of the meteorological drought events in different spatial modes: EN, EC, EEC, and ES denote the number of drought events in modes N, C, EC, and S, respectively; DN, DC, DEC, and DS denote the mean duration (month) of the drought events in modes N, C, EC, and S, respectively; and SN, SC, SEC, and SS denote the mean severity of the drought events in modes N, C, EC, and S, respectively. The red circles represent the moment suffering from drought. The number of circles represents the duration of a drought event.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0183.1
The temporal evolution of the meteorological drought events in different spatial modes: EN, EC, EEC, and ES denote the number of drought events in modes N, C, EC, and S, respectively; DN, DC, DEC, and DS denote the mean duration (month) of the drought events in modes N, C, EC, and S, respectively; and SN, SC, SEC, and SS denote the mean severity of the drought events in modes N, C, EC, and S, respectively. The red circles represent the moment suffering from drought. The number of circles represents the duration of a drought event.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0183.1
d. Spatiotemporal evolutional characteristics based on the identified spatial modes
The four spatial modes were used to characterize the evolution of drought events over the basin. The number of drought events with durations greater than or equal to 2 months was 23. Most of them were short events with just lasting 2–3 months. The mean drought duration and severity of these 23 events were 2.60 ± 1.10 months and 2.00 ± 1.43, respectively. Figure 8 reveals the spatiotemporal migration of the 1962–63 (Fig. 8a) and 1964–65 (Fig. 8b) meteorological drought events. The two drought events were selected because they were the two longest duration drought events (lasting 11 months) in the Poyang Lake basin from 1960 to 2007 as reported by Zhou et al. (2019). As indicated in Fig. 8, however, the two events only lasted 7 and 3 months, respectively. It was worth noting that there were no drought events lasting more than 3 months for each spatial mode (Fig. 7), while Fig. 8a exhibited the 1962–63 drought event lasting seven months over the basin. The phenomenon indicated that the different areas of the basin had suffered from precipitation deficits with varying intensities during the period, which mapped the evolutionary processes of a drought event over the basin. During the period of November 1964–January 1965, the spatial modes C, EC, and S all detected drought events that lasted 3 months, except for mode N (Fig. 7). That is, the drought event mainly occurred in the three spatial modes. It was found that the centroids of the drought clusters migrated more than 100 km (Fig. 8b), which indicated that the severity for the same drought area was continually changing.
The evolutional trajectories of the (a) 1962–63 and (b) 1964–65 meteorological drought events in the Poyang Lake basin.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0183.1
The evolutional trajectories of the (a) 1962–63 and (b) 1964–65 meteorological drought events in the Poyang Lake basin.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0183.1
The evolutional trajectories of the (a) 1962–63 and (b) 1964–65 meteorological drought events in the Poyang Lake basin.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0183.1
To further investigate the process, the temporal evolution of the 1962–63 drought event that lasted 11 months is given in Fig. 9. The drought event affected a large area and triggered a serious hydrological drought in 1963 (Yao et al. 2016).
The temporal evolution of the 1962–63 drought in the four spatial modes. The checkmark in the table indicates that there is a drought in a spatial mode. The monthly dry/wet mapping over the Poyang Lake basin is shown. The gray line facilitates the observation of the scope of the four spatial modes.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0183.1
The temporal evolution of the 1962–63 drought in the four spatial modes. The checkmark in the table indicates that there is a drought in a spatial mode. The monthly dry/wet mapping over the Poyang Lake basin is shown. The gray line facilitates the observation of the scope of the four spatial modes.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0183.1
The temporal evolution of the 1962–63 drought in the four spatial modes. The checkmark in the table indicates that there is a drought in a spatial mode. The monthly dry/wet mapping over the Poyang Lake basin is shown. The gray line facilitates the observation of the scope of the four spatial modes.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0183.1
The drought started in December 1962 with mode S, it subsequently extended to the north, and extremely negative SPI values appeared in modes N and EC. Precipitation relieved the drought in mode S (SPI > −0.5), causing the areas of serious drought to reduce to modes N and EC (i.e., the northern region of the basin) in February and March. Then, basin-wide drought occurred again in April, and the serious drought areas followed modes C and S. Modes N and EC shifted to very wet conditions (SPI > 0.5) in May after plentiful precipitation, while an extreme drought occurred primarily in mode S. Relative to the dry/wet pattern in May, the opposite was true in June, and the most serious areas were transformed to mode N. Few stations with modes C and S detected moderate droughts in July. The drought in August was similar to that in April, yet it was less severe. Next, the drought in September occurred with modes N, EC, and C, and mode C was even more serious. Eventually, the 1962–63 drought event diminished in October 1963 in modes N and EC. The driest areas are shown in modes S and N. Moreover, the modes generally matched the areas where drought occurred, especially in December and February–June. The drought event propagated among the four modes, and the center of the drought shifted between the northern and southern basin regions.
As indicated in July 1963, the drought/wet situation was basic normal (the SPI values of most stations mainly ranged from −0.5 to 0.5), according to Table 1. Thus, when the drought threshold of −1 was applied to judge whether a spatial mode was drought, a mismatch would occur. Although the dimensionality reduction discarded some spatial and temporal information, the identified lower-dimension subspaces could effectively illustrate the development processes of the drought events, especially for large-size droughts. This event provided evidence to confirm the importance of spatial modes in understanding the spatial evolution of droughts.
The spatiotemporal migrating characteristics of all 23 meteorological drought events were analyzed (Fig. 10). The drought severity and duration showed a significant positive correlation (R2 = 0.9503, p < 0.05). The slope of 1.747 emphasized drought seriousness. Similarly, the migration displacement (Fig. 10b) and distance (Fig. 10c) increased significantly as the drought duration increased (slope = 43 km month−1, R2 = 0.272, p = 0.0107; slope = 159 km month−1, R2 = 0.8586, p < 0.05). Most short (particularly for 2 months) drought events were likely to migrate near their origin, while relatively long events might be transregional migrations among the spatial modes. This might be why there was significant migration displacement.
The relationships between (a) drought severity, (b) migration displacement, and (c) migration distance, and drought duration.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0183.1
The relationships between (a) drought severity, (b) migration displacement, and (c) migration distance, and drought duration.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0183.1
The relationships between (a) drought severity, (b) migration displacement, and (c) migration distance, and drought duration.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0183.1
The number of meteorological drought events that migrated southward, northward, eastward, and westward accounted for 39.1%, 26.1%, 26.1%, and 8.7% of the total events respectively. Although the relatively small total number of drought events might result in a discrepancy in investigating the migration patterns, the proportions indicated that northward and southward migration might be two main migration patterns in the Poyang Lake basin, which should be acceptable to some extent. The results might be associated with the water vapor flows from main transportation channels. In the future, tracking the net value changes of water vapor transport from different boundaries might be able to explain the reasonability of migration patterns.
4. Discussion
a. Effect of the definition of the drought event on drought migration characteristics
The definition of a drought event used in the study included two key elements: one was a drought threshold of SPI ≤ −1, and the other was a drought duration with at least 2 months. It should be noted that many studies emphasize that a drought event begins when the drought index value first falls below the threshold and ends when it is greater than the threshold (e.g., Xu et al. 2015; Zhai et al. 2016; Fluixá-Sanmartín et al. 2018; Guo et al. 2018; Zhou et al. 2019; Liu et al. 2019). According to this description, the drought events that last one month should also be reasonable. However, at least 3 months of criteria were often used in their analysis because the short events (less than 3 months) were rarely impactful (e.g., Xu et al. 2015; Herrera-Estrada et al. 2017; Schwalm et al. 2017; Guo et al. 2018). However, some studies take drought events that last at least 2 months into account (e.g., Spinoni et al. 2019; Zhou et al. 2019; Liu et al. 2019). In fact, there is no general consensus and explanation on the question of how long a drought event should be considered. Additionally, it should be noted that the mean durations of flash drought events over China are approximately 20–40 days, and in the context of continued warming, the exposure risk of flash drought will increase, which means a longer duration of drought (Yuan et al. 2019). Thus, the 2-month criterion was used in this study. Meanwhile, according to the 2-month criterion, the results were given. There were only nine drought events, and two of them lasted more than 3 months. Except for the migration displacement, the drought severity (slope = 1.8916 km month−1, R2 = 0.963, p < 0.05) and the migration distance (slope = 170 km month−1, R2 = 0.8959, p < 0.05) showed significant correlations with the drought duration. The number of drought events migrating northward and southward accounted for 44.4% and 33.3%, respectively. In comparison to the 2-month criterion, although there were few differences in the migrating characteristics of the drought events, the statistical results needed to be improved because of the sampling size. There might be two reasons for too many short events. One is that there exists considerable variability for SPI at a 1-month scale. The other is the differences in the drought event identification at the point scale and the regional scale because of the high spatiotemporal variability of precipitation. The 1962–63 typical drought event confirmed the issue (Fig. 9). Moreover, Spinoni et al. (2019) noted that it is valid to identify drought events for individual points using the threshold of −1 (i.e., one negative standard deviation), while this threshold is not valid for a country or region; thus, the threshold of 0.75 was selected for regional drought assessment in their study. A similar issue on the assessment of region-scale droughts was reported by Wang et al. (2009), Zhou and Liu (2016), and Fluixá-Sanmartín et al. (2018). To reduce the uncertainty, the threshold of −1 was used for regional drought identification in this study.
b. Reasonability of the spatial modes in characterizing drought clusters
To confirm the role of the spatial modes in representing drought clusters, the results from two processing schemes based on the TMIM were compared. The first scheme was based on the four identified spatial modes (hereafter, TMIM-Modes), and the other scheme was based on observational stations (hereafter, TMIM-Stations). Although the two schemes might produce multiple clusters with various sizes per drought month, the largest one per drought month was selected for comparison.
The characteristics of the drought clusters in the 141 months were compared (Fig. 11). The linear relationships of the spatial coordinates of the two schemes were close to the 1:1 line and showed significant consistency (R2 > 0.79, p < 0.05) (Figs. 11a,b), which indicated that they had a similar spatial distribution. In terms of the mean severity of the drought clusters, the TMIM-Modes and the TMIM-Stations showed a significant linear relationship (R2 = 0.8158, p < 0.05) (Fig. 11c). The slope of the relationship was 1.08, which indicated that there were few differences overall in the mean drought severity. The relationship between the number of drought stations and that of drought modes revealed a significant positive correlation (R2 = 0.85, p < 0.05) (Fig. 11d). It should be noted that the slope of the linear relationship was approximately 15, which suggested that when the number of drought modes was increased by one, the number of drought stations increased by approximately 15. The increment was equivalent to the number of meteorological stations in mode S (Fig. 5c). According to the results of the TMIM-Stations scheme, a total of 139 drought clusters with more than 15 drought stations were identified, which accounted for 30.6% of all drought clusters (n = 454). The number of months with more than 15 drought stations was 124, which accounted for 87.9% of the 141 drought months. Moreover, the analysis indicated that when the number of drought stations was more than 15, the basin was more likely to be only one drought cluster (Zhou et al. 2019). Hence, it was reasonable to assume that the identified spatial modes might capture or represent the drought clusters with more than 15 stations effectively, which indicated that the performance of spatial modes on capturing smaller extent droughts was poor. Most short-term events lasting only two months and the 1962–63 typical event suggested that the performance might be associated with differences in drought assessment at the regional scale and point scale. The results indicated that the performance of spatial modes in characterizing the large-size drought clusters should be valid.
The comparison between the characteristics of the drought clusters obtained from the TMIM-Modes and TMIM-Stations schemes. The comparison includes spatial coordinates, (a) longitude and (b) latitude, and (c) the mean drought severity, as well as (d) the number of stations and the number of drought modes.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0183.1
The comparison between the characteristics of the drought clusters obtained from the TMIM-Modes and TMIM-Stations schemes. The comparison includes spatial coordinates, (a) longitude and (b) latitude, and (c) the mean drought severity, as well as (d) the number of stations and the number of drought modes.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0183.1
The comparison between the characteristics of the drought clusters obtained from the TMIM-Modes and TMIM-Stations schemes. The comparison includes spatial coordinates, (a) longitude and (b) latitude, and (c) the mean drought severity, as well as (d) the number of stations and the number of drought modes.
Citation: Journal of Hydrometeorology 21, 3; 10.1175/JHM-D-19-0183.1
c. The possible role of spatial modes to understand the development of droughts
Meteorological droughts (or precipitation deficit) are regional in nature and occur over almost all climatic zones (Mishra and Singh 2010; Rajsekhar et al. 2013), which are generally caused by anomalous sea surface temperatures and other remote conditions (Dai 2011; Hao et al. 2018). Thus, some studies have explored the connection between the spatiotemporal modes of droughts and global climate variability over a certain region, for example, Turkey, the arid region of China, and the Yangtze River basin, China (Tatli and Turkes 2011; Wang et al. 2015; Xiao et al. 2015; Huang et al. 2018; Zhou and Liu 2018). Although the mechanisms of droughts are very complicated, these studies confirm that the identified spatial modes are characterized by physical meanings to some extent (Schubert et al. 2016; Zhou and Liu 2018). Hence, the spatial modes might be a clue to construct the connection between evolutional characteristics of droughts and the climate variability for understanding drought development and causes.
5. Conclusions
The goal of this study was to explore the spatiotemporal evolution of meteorological droughts from the perspective of spatial modes. This study identified four spatial modes in the Poyang Lake basin. They were the northern (N), central (C), east-central (EC), and southern (S) regions of the basin, which included 23, 19, 16, and 15 meteorological stations, respectively. On the basis of the four spatial modes, a total of 141 drought months, 144 drought clusters, and 23 drought events were identified from 1960 to 2007. Moreover, the spatiotemporal evolutional characteristics of meteorological drought events were analyzed. The results showed that the drought severity (slope = 1.747 month−1), migration displacement (slope = 43 km month−1), and migration distance (slope = 159 km month−1) of the drought events had significant linear relationships with the drought duration (p < 0.05). The number of drought events that migrated northward and southward accounted for 65% of the total events, which might represent two main migration patterns in the basin. Additionally, the comparison in drought cluster identification based on two processing schemes (i.e., TMIM-Modes and TMIM-Stations) confirmed the role of the spatial modes in characterizing the evolution of droughts. The four identified spatial modes might capture or represent drought clusters with more than 15 stations effectively. These findings indicated that the identified spatial modes could effectively characterize the spatiotemporal evolutional characteristics (e.g., severity, migration distance, and migration pattern) of the large-size meteorological drought events in the basin.
Hence, the spatial modes can mostly match the large areas where serious dry/wet conditions occurred and can be used to illustrate drought evolution, which suggests that they can be considered elementary components of the drought events across the basin. From the perspective of spatial modes, investigating drought evolution might provide some implications for understanding drought development and causes.
Acknowledgments
This work was supported by the State Key Program of National Natural Science of China under Grant 41430855, the Hong Kong Scholars Program (XJ201813), the Key Project of Nanjing Institute of Geography and Limnology of Chinese Academy of Sciences (NIGLAS2018GH06), the CRSRI Open Research Program (CKWV2018499/KY), and the National Natural Science Foundation of China (41571514). The authors declare no conflict of interest.
REFERENCES
Abdi, H., and L. J. Williams, 2010: Principal component analysis. Wiley Interdiscip. Rev. Comput. Stat., 2, 433–459, https://doi.org/10.1002/wics.101.
Bieniek, P. A., and Coauthors, 2012: Climate divisions for Alaska based on objective methods. J. Appl. Meteor. Climatol., 51, 1276–1289, https://doi.org/10.1175/JAMC-D-11-0168.1.
Canadell, J. G., W. L. Steffen, and P. S. White, 2002: IGBP/GCTE terrestrial transects: Dynamics of terrestrial ecosystems under environmental change–Introduction. J. Veg. Sci., 13, 298–300, https://doi.org/10.1111/J.1654-1103.2002.TB02054.X.
China Meteorological Administration, 2003: Ground Meteorological Observation Standard. Meteorological Press, 151 pp.
Cook, B. I., K. J. Anchukaitis, R. Touchan, D. M. Meko, and E. R. Cook, 2016: Spatiotemporal drought variability in the Mediterranean over the last 900 years. J. Geophys. Res. Atmos., 121, 2060–2074, https://doi.org/10.1002/2015JD023929.
Dai, A., 2011: Drought under global warming: A review. Wiley Interdiscip. Rev.: Climate Change, 2, 45–65, https://doi.org/10.1002/WCC.81.
Dai, A., 2013: Increasing drought under global warming in observations and models. Nat. Climate Change, 3, 52–58, https://doi.org/10.1038/nclimate1633.
Dommenget, D., and M. Latif, 2002: A cautionary note on the interpretation of EOFs. J. Climate, 15, 216–225, https://doi.org/10.1175/1520-0442(2002)015<0216:ACNOTI>2.0.CO;2.
Fluixá-Sanmartín, J., and Coauthors, 2018: Searching for the optimal drought index and timescale combination to detect drought: A case study from the lower Jinsha River basin, China. Hydrol. Earth Syst. Sci., 22, 889–910, https://doi.org/10.5194/hess-22-889-2018.
Fovell, R. G., 1997: Consensus clustering of U.S. temperature and precipitation data. J. Climate, 10, 1405–1427, https://doi.org/10.1175/1520-0442(1997)010<1405:CCOUST>2.0.CO;2.
Fu, Q., Z. Zhou, T. Li, D. Liu, R. Hou, S. Cui, and P. Yan, 2018: Spatiotemporal characteristics of droughts and floods in northeastern China and their impacts on agriculture. Stochastic Environ. Res. Risk Assess., 32, 2913–2931, https://doi.org/10.1007/s00477-018-1543-z.
Guo, H., A. Bao, F. Ndayisaba, T. Liu, G. Jiapaer, A. M. El-Tantawi, and P. De Maeyer, 2018: Space-time characterization of drought events and their impacts on vegetation in Central Asia. J. Hydrol., 564, 1165–1178, https://doi.org/10.1016/j.jhydrol.2018.07.081.
Hannachi, A., I. T. Jolliffe, and D. B. Stephenson, 2007: Empirical orthogonal functions and related techniques in atmospheric science: A review. Int. J. Climatol., 27, 1119–1152, https://doi.org/10.1002/joc.1499.
Hao, Z. C., and V. P. Singh, 2015: Drought characterization from a multivariate perspective: A review. J. Hydrol., 527, 668–678, https://doi.org/10.1016/j.jhydrol.2015.05.031.
Hao, Z. C., V. P. Singh, and Y. L. Xia, 2018: Seasonal drought prediction: Advances, challenges, and future prospects. Rev. Geophys., 56, 108–141, https://doi.org/10.1002/2016RG000549.
Hartigan, J. A., and M. A. Wong, 1979: Algorithm AS 136: A k-means clustering algorithm. J. Roy. Stat. Soc., 28C, 100–108, https://doi.org/10.2307/2346830.
Hayes, M., M. Svoboda, N. Wall, and M. Widhalm, 2011: The Lincoln declaration on drought indices: Universal meteorological drought index recommended. Bull. Amer. Meteor. Soc., 92, 485–488, https://doi.org/10.1175/2010BAMS3103.1.
Herrera-Estrada, J. E., Y. Satoh, and J. Sheffield, 2017: Spatiotemporal dynamics of global drought. Geophys. Res. Lett., 44, 2254–2263, https://doi.org/10.1002/2016GL071768.
Huang, T., L. Xu, and H. Fan, 2018: Drought characteristics and its response to the global climate variability in the Yangtze River basin, China. Water, 11, 13, https://doi.org/10.3390/w11010013.
Kallis, G., 2008: Droughts. Annu. Rev. Environ. Resour., 33, 85–118, https://doi.org/10.1146/annurev.environ.33.081307.123117.
Koutroulis, A. G., A. E. K. Vrohidou, and I. K. Tsanis, 2011: Spatiotemporal characteristics of meteorological drought for the Island of Crete. J. Hydrometeor., 12, 206–226, https://doi.org/10.1175/2010JHM1252.1.
Li, X., W. Zhou, and Y. D. Chen, 2015: Assessment of regional drought trend and risk over China: A drought climate division perspective. J. Climate, 28, 7025–7037, https://doi.org/10.1175/JCLI-D-14-00403.1.
Liu, Y., and G. Wu, 2016: Hydroclimatological influences on recently increased droughts in China’s largest freshwater lake. Hydrol. Earth Syst. Sci., 20, 93–107, https://doi.org/10.5194/hess-20-93-2016.
Liu, Y., Q. Zhang, J. Liu, and H. Li, 2012: Climatic, Hydrologic and Environmental Change in Poyang Lake Basin. Science Press, 262 pp.
Liu, Y., G. Wu, R. Guo, and R. Wan, 2016: Changing landscapes by damming: The three Gorges Dam causes downstream lake shrinkage and severe droughts. Landscape Ecol., 31, 1883–1890, https://doi.org/10.1007/s10980-016-0391-9.
Liu, Y., Y. Zhu, L. Ren, V. P. Singh, B. Yong, S. Jiang, F. Yuan, and X. Yang, 2019: Understanding the spatiotemporal links between meteorological and hydrological droughts from a three-dimensional perspective. J. Geophys. Res. Atmos., 124, 3090–3109, https://doi.org/10.1029/2018JD028947.
Livada, I., and V. D. Assimakopoulos, 2007: Spatial and temporal analysis of drought in Greece using the Standardized Precipitation Index (SPI). Theor. Appl. Climatol., 89, 143–153, https://doi.org/10.1007/s00704-005-0227-z.
Lloyd-Hughes, B., 2012: A spatio-temporal structure-based approach to drought characterisation. Int. J. Climatol., 32, 406–418, https://doi.org/10.1002/joc.2280.
McKee, T. B., N. J. Doesken, and J. Kleist, 1993: The relationship of drought frequency and duration to time scales. Proc. of the Eighth Conf. on Applied Climatology, Boston, MA, Amer. Meteor. Soc., 179–184.
Min, S., M. Yan, and J. Liu, 2013: Climatic characteristics of the drought in Lake Poyang catchment (in Chinese). Hupo Kexue, 25, 65–72, https://doi.org/10.18307/2013.0109.
Mishra, A. K., and V. P. Singh, 2010: A review of drought concepts. J. Hydrol., 391, 202–216, https://doi.org/10.1016/j.jhydrol.2010.07.012.
Mo, K. C., J. K. E. Schemm, and S. H. Yoo, 2009: Influence of ENSO and the Atlantic multidecadal oscillation on drought over the United States. J. Climate, 22, 5962–5982, https://doi.org/10.1175/2009JCLI2966.1.
Rajsekhar, D., A. K. Mishra, and V. P. Singh, 2013: Regionalization of drought characteristics using an entropy approach. J. Hydrol. Eng., 18, 870–887, https://doi.org/10.1061/(ASCE)HE.1943-5584.0000683.
Rodriguez, R. D., V. P. Singh, F. F. Pruski, and A. T. Calegario, 2016: Using entropy theory to improve the definition of homogeneous regions in the semi-arid region of Brazil. Hydrol. Sci. J., 61, 2096–2109, https://doi.org/10.1080/02626667.2015.1083651.
Santos, J. F., I. Pulido-Calvo, and M. M. Portela, 2010: Spatial and temporal variability of droughts in Portugal. Water Resour. Res., 46, W03503, https://doi.org/10.1029/2009WR008071.
Santos, J. F., M. M. Portela, and I. Pulido-Calvo, 2013: Dimensionality reduction in drought modelling. Hydrol. Processes, 27, 1399–1410, https://doi.org/10.1002/hyp.9300.
Schubert, S. D., and Coauthors, 2016: Global meteorological drought: A synthesis of current understanding with a focus on SST drivers of precipitation deficits. J. Climate, 29, 3989–4019, https://doi.org/10.1175/JCLI-D-15-0452.1.
Schwalm, C. R., and Coauthors, 2017: Global patterns of drought recovery. Nature, 548, 202–205, https://doi.org/10.1038/nature23021.
Scott, A., and M. Knott, 1974: A cluster analysis method for grouping means in the analysis of variance. Biometrics, 30, 507–512, https://doi.org/10.2307/2529204.
Shankman, D., B. D. Keim, and J. Song, 2006: Flood frequency in China’s Poyang Lake region: Trends and teleconnections. Int. J. Climatol., 26, 1255–1266, https://doi.org/10.1002/joc.1307.
Spinoni, J., and Coauthors, 2019: A new global database of meteorological drought events from 1951 to 2016. J. Hydrol. Reg. Stud., 22, 100593, https://doi.org/10.1016/J.EJRH.2019.100593.
Stagge, J. H., L. M. Tallaksen, L. Gudmundsson, A. F. Van Loon, and K. Stahl, 2015: Candidate distributions for climatological drought indices (SPI and SPEI). Int. J. Climatol., 35, 4027–4040, https://doi.org/10.1002/joc.4267.
Su, B., and Coauthors, 2018: Drought losses in China might double between the 1.5°C and 2.0°C warming. Proc. Natl. Acad. Sci. USA, 115, 10 600–10 605, https://doi.org/10.1073/pnas.1802129115.
Tatli, H., and M. Turkes, 2011: Empirical Orthogonal Function analysis of the palmer drought indices. Agric. For. Meteor., 151, 981–991, https://doi.org/10.1016/j.agrformet.2011.03.004.
Vicente-Serrano, S. M., 2006: Differences in spatial patterns of drought on different time scales: An analysis of the Iberian Peninsula. Water Resour. Manage., 20, 37–60, https://doi.org/10.1007/s11269-006-2974-8.
Vicente-Serrano, S. M., J. C. González-Hidalgo, M. de Luis, and J. Raventós, 2004: Drought patterns in the Mediterranean area: The Valencia region (eastern Spain). Climate Res., 26, 5–15, https://doi.org/10.3354/cr026005.
Wang, A., T. J. Bohn, S. P. Mahanama, R. D. Koster, and D. P. Lettenmaier, 2009: Multimodel ensemble reconstruction of drought over the continental United States. J. Climate, 22, 2694–2712, https://doi.org/10.1175/2008JCLI2586.1.
Wang, H., Y. Chen, Y. Pan, and W. Li, 2015: Spatial and temporal variability of drought in the arid region of China and its relationships to teleconnection indices. J. Hydrol., 523, 283–296, https://doi.org/10.1016/j.jhydrol.2015.01.055.
Wilks, D. S., 2006: Statistical Methods in the Atmospheric Sciences. 2nd ed. International Geophysics Series, Vol. 100, Academic Press, 648 pp.
Wood, E. F., S. D. Schubert, A. W. Wood, C. D. Peters-Lidard, K. C. Mo, A. Mariotti, and R. S. Pulwarty, 2015: Prospects for advancing drought understanding, monitoring, and prediction. J. Hydrometeor., 16, 1636–1657, https://doi.org/10.1175/JHM-D-14-0164.1.
Xiao, M. Z., Q. Zhang, and V. P. Singh, 2015: Influences of ENSO, NAO, IOD and PDO on seasonal precipitation regimes in the Yangtze River basin, China. Int. J. Climatol., 35, 3556–3567, https://doi.org/10.1002/joc.4228.
Xu, K., D. W. Yang, H. B. Yang, Z. Li, Y. Qin, and Y. Shen, 2015: Spatio-temporal variation of drought in China during 1961-2012: A climatic perspective. J. Hydrol., 526, 253–264, https://doi.org/10.1016/j.jhydrol.2014.09.047.
Yao, J., Q. Zhang, Y. Li, and M. Li, 2016: Hydrological evidence and causes of seasonal low water levels in a large river-lake system: Poyang Lake, China. Hydrol. Res., 47, 24–39, https://doi.org/10.2166/nh.2016.044.
Ye, X., Q. Zhang, J. Liu, X. Li, and C. Y. Xu, 2013: Distinguishing the relative impacts of climate change and human activities on variation of streamflow in the Poyang Lake catchment, China. J. Hydrol., 494, 83–95, https://doi.org/10.1016/j.jhydrol.2013.04.036.
Yin, J., B. Su, X. Chen, H. Wang, and L. Zheng, 2011: China Climate Change Impact Report: Poyang Lake Basin. China Meteorological Press, 188 pp.
Yuan, X., L. Wang, P. Wu, P. Ji, J. Sheffield, and M. Zhang, 2019: Anthropogenic shift towards higher risk of flash drought over China. Nat. Commun., 10, 4661, https://doi.org/10.1038/s41467-019-12692-7.
Zarch, M. A. A., B. Sivakumar, and A. Sharma, 2015: Droughts in a warming climate: A global assessment of Standardized precipitation index (SPI) and Reconnaissance drought index (RDI). J. Hydrol., 526, 183–195, https://doi.org/10.1016/j.jhydrol.2014.09.071.
Zargar, A., R. Sadiq, B. Naser, and F. I. Khan, 2011: A review of drought indices. Environ. Rev., 19, 333–349, https://doi.org/10.1139/a11-013.
Zhai, J., J. Huang, B. Su, L. Cao, Y. Wang, T. Jiang, and T. Fischer, 2016: Intensity-area-duration analysis of droughts in China 1960–2013. Climate Dyn., 48, 151–168, https://doi.org/10.1007/s00382-016-3066-y.
Zhang, L., and T. Zhou, 2015: Drought over East Asia: A review. J. Climate, 28, 3375–3399, https://doi.org/10.1175/JCLI-D-14-00259.1.
Zhou, H., and Y. Liu, 2016: SPI based meteorological drought assessment over a humid basin: Effects of processing schemes. Water, 8, 373, https://doi.org/10.3390/w8090373.
Zhou, H., and Y. Liu, 2018: Spatio-temporal pattern of meteorological droughts and its possible linkage with climate variability. Int. J. Climatol., 38, 2082–2096, https://doi.org/10.1002/joc.5319.
Zhou, H., Y. Liu, and Y. Liu, 2019: An approach to tracking meteorological drought migration. Water Resour. Res., 55, 3266–3284, https://doi.org/10.1029/2018WR023311.