Similarity, difference and correlation of meteorological and hydrological drought indices in a humid climate region – the Poyang Lake catchment in China

Variability of water resources has become a serious issue facing many communities and nations around the world within the background of global climate change. It is reported that the well-evidenced global warming over the last century has seriously altered hydrological regimes at regional to global scales, resulting in more and more severe floods and droughts all over the world (Huntington 2006; Jung et al. 2012; Xiong et al. 2013; Gosling 2014; Emam et al. 2015; Yan et al. 2016). This situation looks to continue at present and in the near future according to the report of IPCC (2013). Although public awareness of extreme climatic events has risen sharply and significant progress in science and technology for environmental management has been achieved during the past decades, our society still continues to suffer from the consequences of these meteorological and hydrological hazards worldwide (e.g., Nafarzadegana et al. 2012; Tao et al. 2013). From a scientific and practical point of view, understanding the changing characteristics of drought and wetness variations is of essential importance for improving integrated water resources management at the catchment scale and human mitigation of hydrological alterations.

Drought is a natural phenomenon of near surface water shortage caused by regional moisture deficit or unbalance of water supply and demand in a region during a certain period of time. This phenomenon is usually considered as a common, widespread, and recurring climate-related hazard and can occur virtually in all climate zones (Riebsame et al. 1991; Moradi et al. 2011; Portela et al. 2015). Due to complex underlying causes and different perspectives that have been focused on, drought classification and monitoring methods are usually not consistent. Commonly, environmental drought generally includes (1) meteorological drought, (2) hydrological drought, (3) agricultural drought, and (4) socio-economic drought (Heim 2002). However, the first two are most attractive to scientists and researchers. Meteorological drought is the basic precondition for the other types of environmental drought, and usually defines a precipitation deficit over a region for a period of time. Hydrological drought is the result of long-term meteorological drought, which is characterized by the quick decrease of stream flow or the abnormal low stage of river or lake water level. Drought not only directly affects people's water security and food security; continuous drought may also change surface landscapes (such as land cover and soil properties), and further affects the processes of runoff yield and flow concentration, causing a series of ecological environmental problems (Weng & Yang 2010). As the occurrence of drought hazard is not unexpected, research on the occurrence and development processes of drought should be given sufficient attention.

In order to quantitatively study the intensity and duration of drought events, various drought indices have been developed over the past few decades (Palmer 1965; McKee et al. 1993; Heim 2002; Rahmat et al. 2015; Zhang et al. 2015; Zhu et al. 2016). The Palmer drought severity index (PDSI) (Palmer 1965) and standardized precipitation index (SPI) (McKee et al. 1993; Rahmat et al. 2013) are the two typical indicators for assessing the conditions of meteorological drought. Calculated from a simple water–balance model forced by monthly precipitation and temperature data (Palmer 1965), the PDSI is effectively applied in large-scale drought assessments (Sheffield et al. 2012). However, lack of multi-scalar character makes the PDSI an unreliable index for identifying different types of drought (Sheffield et al. 2012; Tao et al. 2013). The SPI has been widely used to reveal meteorological drought and was proven to be a useful tool in the estimation of the intensity and duration of drought events (Bordi et al. 2004; Zhang et al. 2009; Nafarzadegana et al. 2012). The SPI calculation is based only on precipitation deficiency, but neglects other critical variables that may affect drought conditions. Therefore, it is not an ideal tool for drought assessment in those areas with obvious climate change. For those reasons, Vicente-Serrano et al. (2010) proposed a new drought index: the standardized precipitation–evapotranspiration index (SPEI) based on precipitation and potential evapotranspiration, which effectively combines the physical principles of the PDSI with the multi-scalar character of the SPI, and has become widely used in recent years.

With respect to meteorological conditions, river discharge directly reflects the amount of water resources in a region that may have impacts on local society. For the assessment of hydrological drought, several indices have been developed, including Palmer hydrological drought severity index, surface water supply index, and standardized runoff index (SRI), etc. (Shukla & Wood 2008; Marengo et al. 2011). Among these, the SRI is a natural extension of SPI used to describe the anomalies of stream flow and river or lake stage. It, however, has more appeal than the SPI as it incorporates hydrological and meteorological processes that influence the volume and timing of streamflow. Due to the effects of underlying landscape conditions as well as human activities, the variation of hydrological drought index may be desynchronized from the response of meteorological drought index.

Assessment of the variability of hydro-meteorological drought in a region is essential for local water resources management and for the prevention of the risk of natural hazards. In addition, understanding the characteristics of different drought indices will provide insight into the causes and impacts of different indices. As a consequence of climate change, seasonal hydro-meteorological drought has occurred more frequently in recent decades, even in humid regions. Associated with this context, the main goals of this study are: (1) to evaluate the potential capability and performance of SPEI and SRI in assessing drought variability; and (2) to examine the correlation and differences between hydrological droughts and meteorological droughts at different timescales.

The catchment belongs to a subtropical climate zone with an average annual temperature of 17.5 °C and average annual precipitation of 1,665 mm (Figure 1(b)). Due to the dominant effect of the Southeast Asian Monsoon, 45% of annual precipitation is concentrated in the flood season from April to June (Figure 1(b)).

In this study, a complete data set of daily precipitation and temperature from 14 standard national weather stations inside the catchment was obtained from the National Climate Centre of China Meteorological Administration (CMA). The data set of all weather stations covers the period 1960–2010 with no missing data on the variables. Data quality control was done by CMA before delivery. Locations of these weather stations can be referred to in Figure 1.

Observed daily stream flows for 1960–2010 at five gauging stations (Figure 1) were collected from the Hydrological Bureau of Jiangxi Province, China. The drainage area of the five gauging stations accounts for 74.5% of the total Poyang Lake catchment. Among the five stations, Waizhou, Lijiadu, and Meigang located at the lower reaches of the Ganjiang, Fuhe, and Xinjiang Rivers were considered as the outlet for these sub-catchments. Dukengfeng and Wanjiabu are the two stations with relative small drainage areas and are located at the branches of the Raohe and Xiushui rivers. The basic features of these gauging stations are listed in Table 1.

Although there are several methods for meteorological drought estimation, the SPEI was selected to estimate meteorological drought in this study because it combines the physical principles of the PDSI with the multi-scalar character of the SPI (Tao et al. 2013). The process of SPEI estimation is completely consistent with SPI, but adds the effect of potential evapotranspiration.

According to the value of SPEI, meteorological drought can be classified into three categories (Vicente-Serrano et al. 2010). Table 2 lists the SPEI classification and corresponding cumulative probabilities of drought occurrences. More detailed information can be referred to in Vicente-Serrano et al. (2010) and Tao et al. (2013).

The concept employed by McKee et al. (1993) for SPI was applied in defining SRI. The calculation of SRI follows the same procedure as for SPI, but the input for this index is monthly runoff data rather than monthly precipitation. The fundamental idea of using SRI is to examine drought from a hydrological perspective with comparison to traditional drought indices, such as SPI and PDSI. In addition, the SRI has the same classification as that of SPI.

In this study, the nonparametric Mann–Kendall (MK) statistical test was applied to estimate the change trend of seasonal hydro-meteorological drought. The method is widely used for trend detection in hydrological and meteorological series due to its robustness against non-normal distributions and insensitiveness to missing values (e.g., Wang et al. 2008; Zhang et al. 2009; Ehsanzadeh & Adamowski 2010; Li et al. 2012). The null hypothesis H₀ of the MK test is that there is no trend of the calculated drought conditions from which the data set X (x₁, x₂, x₃ … x_n) is drawn. The null hypothesis H₀ should be rejected if statistic parameter |Z| ≥ 1.96 at 5% significance level. The Z value is a standard normal variable that represents the significance level of a specific trend. A positive value of Z indicates increasing trend, and a negative value of Z indicates decreasing trend.

The authors of this paper have applied this method in the study of streamflow and climatic series of the Poyang Lake catchment (Ye et al. 2013; Zhang et al. 2014). Details about the method can be referred to in the above-mentioned literature.

The continuous wavelet transform (CWT) is a mathematical method that was developed for signal processing. Through CWT analysis, hydro-meteorological series can be decomposed into time–frequency space to determine both the dominant modes of variability and how those modes vary in time (Torrence & Compo 1998). Up to now, the CWT method has been widely used for analyzing localized variations of power within a geophysical time series (Grinsted et al. 2004; Zhang et al. 2009). The concept of the method is thoroughly explained and discussed by Torrence & Compo (1998). In this study, as a complement to the MK method which studies the trend of the drought indices, the CWT method was applied to study the periodicity of hydro-meteorological series of the Poyang Lake catchment, and Morlet wavelet was chosen as a basic wavelet because it provides good balance between time and frequency localization.

On consideration of the distribution patterns of wavelet power spectrum in Figure 4(a), the 22 years primary periodicity of the catchment is remarkably evident. During the study period 1960–2010, the Poyang Lake catchment experienced a periodic evolution process of meteorological drought, wet, drought, wet, and drought. The latest meteorological drought episode began in about 2003, and it is now entering into the later period. According to the distribution of wavelet power spectrum, it seems that the latest drought episode lasted to the end of 2013, and then a wet episode sets in. On the scale of 7 years periodicity, a distinct drought and wet alteration period was found from the middle 1960s to the middle 1980s. However, two distinct drought and wet alteration periods were found on the scale of 4 years periodicity, the first period was from the earlier 1960s to the middle of the 1970s, and the second period was from the earlier 1990s to the middle of the 2000s.

Change trend of seasonal SRI is roughly consistent with that of SPEI in January, May, July, and August (Figure 5(a)), but clear differences are seen in other months. A decreasing trend of SRI was mainly found in May and June. In contrast with SPEI, SRI in March, September, and November shows increasing trends. Especially, the increasing trend of SRI in September is significant at a 0.05 significance level. In addition, the increasing trend of SRI in December and February are much more obvious than for SPEI. It is worth noting that according to the change trends of monthly SPEI and SRI, cross-correlation analysis shows that the two indices reach a maximum correlation coefficient at 1-month time lag (Figure 5(b)). This result indicates that changes of hydrological drought in the Poyang Lake catchment are usually delayed for 1 month compared with meteorological drought.

The Poyang Lake catchment is one of the most sensitive regions of climate change in the Yangtze River basin; especially the rising temperature and increasing frequency of extreme precipitation have been remarkably obvious since the 1990s (Ye et al. 2013). Wang & Chen (2012) pointed out that the impact of temperature anomaly on the drought/wetness variability cannot be neglected. Our recent study, by comparing the performance of SPEI and SPI in the catchment, also concluded that positive temperature anomaly since the 1990s plays an important role that intensifies the drought process, while negative temperature anomaly during 1965–1986 can alleviate droughts (Ye et al. 2015). In this study, the estimated frequency distribution of meteorological drought by SPEI reflects well the impacts of both precipitation and temperature, and the results are somewhat different from those estimations based just on precipitation (e.g., Liu et al. 2012b; Min et al. 2013). Seasonally, extreme meteorological drought seldom occurs in autumn, while extreme hydrological drought is rare in winter. Min et al. (2013) investigated the climatic characteristics of drought in the catchment by using the Z index method, and revealed that extreme meteorological droughts mainly occurred in spring.

As hydrological drought, to a large extent, is the result of long-term meteorological drought, the fluctuations of SPEI and SRI time series at different timescales are somewhat consistent and correlated and the correlation increases as timescales increase. Furthermore, the performance of the SRI series shows an obvious hydrologic delay of about 1–2 months in response to SPEI at timescales >12 months. The desynchronization of the two indices is in agreement with the findings of other studies (e.g., Shukla & Wood 2008; Liu et al. 2012a). As is already known, the volume and variation of river discharge are not only determined by the local meteorological conditions, but also affected by the underlying landscape characteristics. The size of the drainage area plays an important role that leads to hydrologic delays in the form of soil moisture, groundwater, and even surface streamflow routines (Raudkiui 1979). The larger the drainage area, the longer the hydrologic delays of soil moisture, groundwater, and streamflow discharges to the catchment outlet. The results of correlation analysis between SRI and SPEI series showed that the maximum correlation coefficient was found between 1-month SRI and SPEI at a timescale of 2 months. This result is consistent with some previous studies, such as those of Vicente-Serrano & López-Moreno (2005) and Du et al. (2012), but differs from other studies (e.g., McKee et al. 1993; Zhai et al. 2010). Of the five gauging stations Lijiadu station has the smallest drainage area, thus river discharge is more associated with the current meteorological conditions than those for other stations. However, for those rivers with relatively larger drainage areas, river discharges are more associated with the current and previous meteorological conditions of shorter periods than of longer periods.

SPEI and SRI in the Poyang Lake catchment showed a primary periodicity of 22 years and two secondary periodicities of 4 and 7 years. This result indicates that on an annual basis, climatic anomalies in the subtropical catchments such as the Poyang Lake catchment are the dominant factor that control the variability of local river discharge. Also, the result of periodicity analysis revealed that the Poyang Lake catchment is now entering into a wet episode. As the change of water resources of the catchment is particularly important for the eco-environmental conservation of the Poyang Lake wetland as well as regional economic development of the lower Yangtze River, outcomes of this study are expected to assist in understanding the medium- and long-term variability of catchment water resources in the near future. Additionally, as the variation of SRI can be desynchronized from the response of SPEI, joint consideration of the relevance of meteorological drought and hydrological drought under various timescales is useful in surface water resources management and hazard protection.

In this study, the performance of SPEI and SRI in assessing drought variability and the correlation and differences between hydrological droughts and meteorological droughts at different timescales was investigated using hydro-meteorological data from a humid region – the Poyang Lake basin in China. The study concludes that in the Poyang Lake catchment, seasonal hydro-meteorological droughts have occurred frequently in recent decades. Annual variability of both indices showed a primary periodicity of 22 years and two secondary periodicities of 4 and 7 years. SRI is less variable than SPEI at shorter timescales and it shows an obvious hydrologic delay of about 1–2 months in response to SPEI at timescales larger than 12 months. For the desynchronized variation of the two indices at different timescales, the influences of geophysical characteristics of river catchments that moderate the hydrological response should be considered. With respect to the study catchment, the 2-month timescale of SPEI is more appropriate for drought event monitoring, especially for those rivers with drainage areas larger than 0.35 × 10⁴ km².

This work was financially supported by the National Basic Research Program of China (2012CB417003), Key Laboratory of Watershed Geographic Sciences, Nanjing Institute of Geography and Limnology, Chinese Academy of Sciences (WSGS2015003), Collaborative Innovation Center for Major Ecological Security Issues of Jiangxi Province and Monitoring Implementation (No. JXS-EW-00), Fundamental Research Funds for the Central Universities (XDJK2016C093), and National Natural Science Foundation of China (41201026).