1 Introduction

Assessment of water resource variations is a pre-requisite to understand and adopt appropriate management strategies with the aim to avoid adverse environmental effects and reconcile conflicts between users (Xu and Singh 2004). Climate changes, particularly the current well-evidenced global warming and its impacts on hydrological regimes, especially hydrological extremes, e.g. droughts and floods, have become a priority area both for process research and for water management practices (Xu et al. 2005). It was believed that projected global climate changes have the potential to accelerate the global hydrological cycle. Many studies indicated that global warming alters precipitation patterns and results in more frequent extreme weather events, e.g. floods, droughts, and rainstorms (Zhang et al. 2008a; WMO 2003). These results are undoubtedly useful for the better understanding of increasing flood/drought hazards over the world (e.g. Herschy 2002; Mirza 2002). Furthermore, public awareness of extreme climatic events has risen sharply in recent years partly due to the catastrophic nature of floods, droughts, storms and other climatic extremes (e.g. Beniston and Stephenson 2004; Zhang et al. 2006a, b, 2008b). Therefore, it is of scientific and practical merit to better understand changing characteristics of dryness and wetness variations for improving integrated water resource management at the basin scale and human mitigation to hydrological alterations.

Environmental droughts generally include: (1) meteorological drought, (2) hydrological drought; and (3) agricultural drought (e.g. Livada and Assimakopoulos 2007). This study focuses on the meteorological drought which is defined as a lack of precipitation over a region for a period of time. The standardized precipitation index (SPI) (McKee et al. 1993, 1995; Hayes et al. 1999) was widely used to reveal meteorological drought (e.g. Silva et al. 2007; Bordi et al. 2004a; Moreira et al. 2006) and was proven to be a useful tool in the estimation of the intensity and duration of drought events (Bordi et al. 2004a). Livada and Assimakopoulos (2007) used the SPI to analyze drought events in Greece. Wilhite and Glantz (1985) applied the SPI in Nebraska on time scales of 3, 6, 12, 24 and 48 months. In this study, the SPI is calculated based on 1 and 3-month precipitation time series.

The Pearl River is the third largest river in drainage basin area and the second largest river in terms of streamflow in China. However, uneven spatial and temporal distribution of water resources, with 80% of the total discharge occurring in the flooding seasons, i.e. April–September, negatively affects the effective human use of water resource. With booming economic development in the Pearl River basin, pollution-induced water shortage is threatening the security of regional water resources. The East River, a tributary of the Pearl River, bears the heavy responsibility of water supply for Shenzhen and Hong Kong, and about 80% of Hong Kong’s annual water demands come from the East River. Due to the significant role of water resource in regional economic development and conservation of ecological environment in the region, precipitation changes and possible underlying causes have drawn increasing concerns. Wang et al. (2008) explored changing properties of precipitation extremes and streamflow extremes in the East River, one tributary of the Pearl River basin. Dong (2006) indicated close relations between extreme precipitation changes and spatial and temporal distribution of floods in the region. Luo et al. (2008) analyzed precipitation trends in North River basin by using Mann-Kendall trend test and Sen’s T test. In terms of dry and wet changes in China, Bordi et al. (2004b) investigated time–space variations of dry and wet periods in the east China. However, to our best knowledge, no such reports are available in the Pearl River basin so far despite numerous studies of this topic available in the literatures. With this in mind, the objectives of this paper are: (1) to detect changing properties of dry and wet episodes defined by SPI; (2) to analyze trends of frequency and intensity of dry and wet events by using Mann-Kendall trend test; and (3) to explore possible causes behind wet and drought variations in the Pearl River basin with NCAR/NCEP reanalysis dataset.

2 Study region and data

The Pearl River (97°39′E–117°18′E; 3°41′N–29°15′N) (Fig. 1) is the second largest river in terms of streamflow in China with a drainage area of 4.42 × 105 km2 (PRWRC 1991). The Pearl River basin involves three major tributaries: West River, North River and East River. The West River is the largest tributary accounting for 77.8% of the total drainage area of the basin. The North River is the second largest tributary with a drainage area of 46,710 km2. The East River accounts for 6.6% of the total area of the Pearl River. The Pearl River basin is located in the tropical and sub-tropical climate zone with the annual mean temperature ranging between 14–22°C. The precipitation mainly concentrates during April–September (Zhang et al. 2008a), accounting for 72–88% of the annual precipitation (PRWRC 1991).

Fig. 1
figure 1

Study region and rain gauging stations

Full size image

The daily precipitation dataset covering 1 January 1960–31 December 2005 was collected from 42 rain stations in the Pearl River basin. Location of the rain gauging stations can be referred to Fig. 1. Data quality control was made in an earlier study (Zhang et al. 2008a). The NCAR/NCEP reanalysis dataset was used to detect changes of water vapor flux and moisture content in the Pearl River basin with the intention to understand possible causes behind wet and drought changes over the basin.

3 Methodology

3.1 Standardized precipitation index (SPI)

The Standardized precipitation index (SPI) (McKee et al. 1993) was used to quantify the precipitation deficit on multiple time scales. According to McKee et al. (1993), the SPI was defined on each of the time scales as the difference between precipitation on the time series (x i ) and the mean value \(\left( {\bar x} \right)\), divided by the standard deviation (s), i.e.

$$SPI = \frac{{x_i - \bar x}}{s}$$
(1)

The same definition is widely used in the literature (e.g., McKee et al. 1993; Livada and Assimakopoulos 2007). It is well known that very seldom the monthly precipitation time series fits a normal distribution, thus, an initial transformation of the data series is usually done to make SPI a standard normal distribution variable. In this study, different probability distributions have been used to fit precipitation time series on various time scales; gamma, normal and log-normal distributions were selected as candidate distributions; and the Kolmogorov-Smirnov test was performed to evaluate the goodness-of-fit. The results (not shown) indicated that normal distribution had the worst performance. Gamma distribution and log-normal distribution performed equally well for the monthly precipitation series in the basin. Particularly, log-normal distribution had better performance in describing precipitation properties in the rainy season. Gamma distribution performed better in winter (January, February and December) when compared to log-normal distribution. Even so, log-normal distribution also fit the precipitation changes in winter at >95% confidence level well. Therefore, we chose log-normal probability distribution for the monthly precipitation series in the basin. After logarithmic transformation of the dataset, the sample mean and variance of the transformed data will be \(\hat \mu _y \) and \(\hat \sigma _y \), then the SPI becomes

$$SPI = Z = \frac{{\ln \left( x \right) - \hat \mu _y }}{{\hat \sigma _y }}$$
(2)

The drought and wetness severity adopted in this study is defined in Table 1. Due to the fact that precipitation mainly concentrated in the rainy season (April–September; Fig. 2), we studied the changes of SPI for the rainy season. Because the Pearl River basin bears the heavy responsibility of water supply for Hong Kong and the Pearl River Delta, the SPI changes in winter (January, February and December) were also studied. Moreover, the number of months characterized by drought and wetness categories in the rainy season was also analyzed.

Fig. 2
figure 2

Areal monthly mean precipitation in the Pearl River basin (mm/month, average of 42 stations) for the period 1960–2005

Full size image
Table 1 The standardized precipitation index (SPI) categories based on the initial classification of SPI values
Full size table

3.2 Aridity index

To further understand dry and wet variations in terms of agriculture demand, we also calculated the aridity index and compared it with SPI in this study. Aridity index is defined by de Martonne (1926) who used it to study irrigation demands (WMO 1975) and is computed as:

$$I_i = \frac{{12P_i }}{{T_i + 10}}$$
(3)

where P i is the monthly precipitation amount; T i is the monthly mean air temperature. The aridity index aims to identify the months when irrigation is necessary. Generally, irrigation is necessary when I i  < 20. Research results by Livada and Assimakopoulos (2007) indicated that, on a monthly basis, there is a statistically significant exponential correlation between SPI and aridity index (I = I i /12) and significant linear correlation between aridity index and rainfall departures (RD). RD is defined as the difference between the monthly precipitation and its mean values, and then is divided by its standard deviation. They also stated that the exponential relationship between I and SPI should be due to the normalization of the monthly precipitation data prior to SPI estimation (Livada and Assimakopoulos 2007).

3.3 Trend test

There are many statistical methods available for trend detection and each method has its own strength and weakness in trend detection (Zhang et al. 2008a). In this paper, the Mann-Kendall (MK) test (Kendall 1975; Mann 1945), recommended by the World Meteorological Organization (Mitchell et al. 1966), is used to study trends of SPI, intensity of wet and dry months and also the number of months when irrigation is necessary. For the sake of completion and understanding of the results, the procedure of MK trend test performed in the study is briefly introduced as follows.

First the MK test statistic is calculated as

$$S = \sum\limits_{i = 1}^{n - 1} {\sum\limits_{j = i + 1}^n {\operatorname{sgn} \left( {x_j - x_i } \right)} } $$
(4)

where \( \operatorname{sgn} \left( {x_j - x_i } \right) = \left\{ {\begin{array}{*{20}l} { + 1,} \hfill & {x_j > x_i } \hfill \\ {0,} \hfill & {x_j = x_i } \hfill \\ { - 1,} \hfill & {x_j < x_i } \hfill \\ \end{array} } \right. \) and n is the sample size. The statistics S is approximately normally distributed when n ≥ 8, with the mean and the variance as follows:

$$E\left( s \right) = 0$$
(5)
$$V\left( S \right) = \frac{{n\left( {n - 1} \right)\left( {2n + 5} \right) - \sum\limits_{i = 1}^n {t_i i\left( {i - 1} \right)\left( {2i + 5} \right)} }}{{18}}$$
(6)

where t i is the number of ties of extent i.

The standardized statistics (Z) for one-tailed test is formulated as:

$$ Z = \left\{ {\begin{array}{*{20}l} {\frac{{S - 1}} {{\sqrt {Var\left( S \right)} }}} \hfill & {S > 0} \hfill \\ 0 \hfill & {S = 0} \hfill \\ {\frac{{S + 1}} {{\sqrt {Var\left( S \right)} }}} \hfill & {S < 0} \hfill \\ \end{array} } \right. $$
(7)

At the 5% significance level, the null hypothesis of no trend is rejected if |Z| > 1.96.

3.4 Calculation of frequency of wet and dry months in the rainy season

The number of months fell in each categories (frequency of wet and dry months), i.e. from extreme dry to extreme wet (see Table 1 for division) in the rainy season of each year is a useful indicator for the dryness/wetness of the catchment since the rainy season captures 80% of the yearly precipitation total. The trends of these numbers are calculated via the following steps: (1) calculate the SPI for each month in the rainy season and count the number of months falling in each categories within each year, (2) take the numbers in each categories of each year as a time series and calculate the trend of the time series, and (3) spatially interpolate the trend calculated for each station. Due to the fact that the number of extreme dry or wet months is scarce, only, e.g., 75 extreme dry months can be identified in the Pearl River basin for past 46 years account for 4% of total months in the rainy season. For the sake of comparison, the number of extreme dry months will not be analyzed, rather We will only be analyzing the trends of the number of severe (moderate) dry (wet) months in the rainy season.

4 Results and discussion

4.1 Spatial and temporal patterns of SPI

Figure 3 maps the spatial distribution of SPI trends across the Pearl River basin for the rainy season. It can be seen that the major parts of the basin are characterized by decreasing SPI, indicating that drying tendency dominates major parts of the Pearl River basin. However, only a few places are characterized by the decreasing trends of SPI significant at >95% confidence level. A slight increasing trend (not significant at 95% level) of SPI can be identified in a few locations. Thus, generally, wet tendency is mainly identified in southeast parts of the Pearl River basin in the rainy season. As mentioned earlier, the Pearl River basin bears the heavy responsibility of water supply for the Pearl River Delta and Hong Kong—and that is particularly the case for the East River. Figure 4 shows that the entire Pearl River basin is characterized by increasing SPI in the winter season. Thus, the wet tendency prevails over the basin in the winter season, although the increasing trends are not significant at >95% confidence level.

Fig. 3
figure 3

Spatial distribution pattern of trends of SPI in the rainy seasons (April–September) across the Pearl River basin. The numbers in the figure are Z values. Positive values indicate increasing trend and vice versa. If |Z| ≥ 1.96, then the trend is significant at >95% confidence level

Full size image
Fig. 4
figure 4

Spatial distribution pattern of trends of SPI in winter (January, February and December) across the Pearl River basin. The numbers in the figure are Z values. Positive values indicate increasing trend and vice versa. If |Z| ≥ 1.96, then the trend is significant at >95% confidence level

Full size image

4.2 Frequency of wet and dry months in the rainy season

Figure 5 demonstrates the spatial patterns of trends in the number of dry or wet months in the rainy season, i.e. moderate (severe) wet and moderate (severe) dry. Here we decide the number of moderate (severe) wet and moderate (severe) dry months in the following way: (1) the month with SPI > 1.5 (or SPI < −1.5) is defined as severe wet (or dry) month; (2) the month with SPI > 1 (or SPI < −1) is defined as moderate wet (dry) month. Different spatial patterns can be identified for wet or dry episodes of different categories (Fig. 5). Figure 5a indicates a decreasing number of severe wet months in large parts of the Pearl River basin, though this decreasing trend is not statistically significant. Stations with an increasing number of severe wet months distribute sporadically across the basin. Similar phenomenon can be identified in changes of moderate wet months (Fig. 5b). A large area of the Pearl River basin is characterized by a decreasing number of moderate wet months and increasing trends can be observed mainly in the lower East River, lower North River, lower West River and lower Beipan River. Different spatial patterns can be observed in the changes of severe and moderate dry months (Fig. 5c,d). Major parts of the Pearl River basin are dominated by increasing severe and moderate dry months. Similarly, these trends are not statistically significant. It can be seen from Fig. 5c that the increasing number of severe dry months is mainly observed in Beipan River, Nanpan River and Hongshui River and decreasing trends are mainly found in the east parts of the Pearl River basin. In terms of the number of moderate dry months, increasing trends are observed in west and south parts of the basin, e.g. Nanpan River, Beipan River, Zuo River, Yu River, Qian River and Liu River. Decreasing trends in the number of moderate dry months are identified mainly in the east parts of the Pearl River basin (Fig. 5d). Figure 6 illustrates areal average SPI across the Pearl River basin. It can be observed from Fig. 6 that SPI of the basin is increasing before 1980s, and is decreasing after 1980s. An obvious decreasing tendency is found during 1985–1998. SPI in winter, however, is consistently increasing, particularly during 1970–2005.

Fig. 5
figure 5

Spatial distribution pattern of trends of the number of wet and dry months of different categories in the rainy season (April–September). a severe wet; b moderate wet; c severe dry; d moderate dry. The numbers in the figure are Z values. Positive values indicate increasing trend and vice versa. If |Z| ≥ 1.96, then the trend is significant at >95% confidence level

Full size image
Fig. 6
figure 6

Areal average SPI changes in the rainy season (April–September) and winter (DJF) for the Pearl River basin. Dashed lines denote quadratic fit

Full size image

Figure 7 shows trends of frequency of months when irrigation is necessary, i.e. I < 20; hereafter, we will simply call months with I < 20 dry months. It can be seen from Fig. 7 that an increasing number of dry months can be detected mainly in western parts of the Pearl River basin, i.e. Nanpan River, Beipan River, Hongshui River, lower Liu River and upper Zuo River. Major regions in the eastern parts of the basin are characterized by a decreasing frequency of dry months. Figure 7 demonstrates that places characterized by increasing number of dry months largely match with those characterized by an increasing number of severe dry months defined by SPI (Fig. 5c,d). These results imply that, at least in the Pearl River basin, SPI and I index seem to perform similarly well in reflecting wet and dry conditions in the basin. To verify the results of Fig. 5, we further analyzed relationships between I, SPI and rainfall departure (RD), and for illustrative purpose the results for a randomly selected station are shown in Fig. 8a for the rainy season and Fig. 8b for winter. Similar results have been obtained for the relationships between I, SPI and RD when compared to those by Livada and Assimakopoulos (2007) over Greece. Excellent linear relations can be fit between I and RD, and exponential relationships between I and SPI. The exponential relationship between I and SPI can be attributed to the transformation of original monthly precipitation data by using log-normal distribution function.

Fig. 7
figure 7

Trends of the number of months when agricultural irrigation is necessary, i.e. I < 20. The numbers in the figure are Z values. Positive values indicate increasing trend and vice versa. If |Z| ≥ 1.96, then the trend is significant at >95% confidence level

Full size image
Fig. 8
figure 8

Scatter diagram between I, SPI, (open circles) and I, RD (closed circles) and correlation lines for for one station selected randomly from the dataset. a Rainy season; b winter. The example station: Xianning station (26°52′N, 104°17′E)

Full size image

4.3 Water vapor flux analysis

Water vapor flux plays the key role in changes of drought and wet conditions; the water vapor is brought to the Asian continent mainly by the monsoonal flows. Generally, three major low-level monsoonal streams transport water vapor to China (Simmonds et al. 1999; Tian et al. 2004; Chow et al. 2008): (1) the southwesterly flow towards the Indian peninsula and Bengal Bay which is associated with the Indian summer monsoon, and (2) the southerly flow in the South China Sea and (3) the southeasterly flow associated with the southern flange of the north Western Pacific subtropical high (0–20°N, 120–150°E). Routes of the water vapor transport in the rainy season (Fig. 9a) clearly demonstrate the sources of the water vapor: Indian peninsula, Bengal Bay and South China Sea. However, water vapor in the Pearl River is mainly from southwesterly flow. Intensity of water vapor flux in the rainy season is larger than in winter (Fig. 9), indicating considerable importance of water vapor transport on the determination of rainy and dry seasons. To further understand the possible impacts of moisture content on dry and wet conditions in the rainy season and in winter, we also calculated and plotted the standardized total number of wet months (SPI > 0) together with standardized moisture budget and moisture content in the rainy season and in winter (Figs. 10 and 11). It is observed that the relationships between standardized total number of wet months (SPI > 0) with moisture budget and moisture content are remarkably better in winter (Fig. 11, dry season) than in the rainy season (Fig. 10). Furthermore, Fig. 11 indicates an increasing tendency for the standardized total number of wet months, moisture budget and moisture content and this increasing trend is significant at >95% confidence level. This means that, in winter, increasing moisture budget and moisture content results in an increasing number of wet months across the Pearl River basin. This result is in agreement with earlier result that SPI in winter is in increasing tendency (Fig. 4), showing considerable influences of moisture variations on wet and dry conditions in winter. No conclusive relations can be identified between standardized total number of wet months (SPI > 0) and standardized moisture budget and moisture content in the rainy season. Thus, we can say that in the winter (dry) season the wet/dry conditions are mainly influenced by water vapor and moisture content in the air, while more factors than moisture transport influence the wet/dry conditions in the rainy season.

Fig. 9
figure 9

Water vapor flux (kg × m−1 × s−1) in the rainy season (a) and in winter (b) for the Pearl River basin

Full size image
Fig. 10
figure 10

Comparison between standardized areal total amount of wet months (SPI > 0) and moisture content and moisture budget in the rainy season (April–September) across the Pearl River basin

Full size image
Fig. 11
figure 11

Comparison between standardized areal total amount of wet months (SPI > 0) and moisture content and moisture budget in winter (January, February and December) across the Pearl River basin

Full size image

5 Conclusions

In this study, we quantitatively evaluated dry and wet conditions by using the standardized precipitation index (SPI) and aridity index (I) based on monthly precipitation dataset of 42 rain stations in the Pearl River basin for 1960–2005. Mann-Kendall trend test was used to detect trends within the number of months of different dry and wet categories and SPI values. Furthermore, we also attempted to explore possible causes behind changing properties of dry and wet conditions across the Pearl River basin by using NCAR/NCEP dataset. Some interesting conclusions are obtained as follows:

  1. 1.

    Different dry or wet tendencies can be identified in the Pearl River basin in the rainy season and in winter. The Pearl River basin tends to be dryer in the rainy season and to be wetter in winter. However, different parts of the basin show different patterns of dry and wet conditions— a general dry tendency can be observed in major parts of the basin in the rainy season, and a wet tendency in winter can be identified across the entire basin. Most of the stations show decreasing SPI index of rainy season but some are not, which may be due to the extremely inhomogenous spatial distribution of precipitation as a result of stronger convective precipitation and typhoon rain storms which are very common in the rainy season in the Pearl River basin. However, the relatively more homogenous spatial patterns of SPI index in winter may be attributed to relatively stable air masses in the season.

  2. 2.

    In terms of the number of dry or wet months in the rainy season, major parts of the Pearl River basin are characterized by a decreasing frequency of severe and moderate wet months. However, an increasing number of moderate wet months can be observed mainly in southeast parts of the Pearl River basin. Increasing frequency of severe dry months can be observed in regions <108E°, and an increasing number of moderate wet months can be found in regions <107°E and between 108–110°E. Results of aridity index analysis also show a decreasing number of dry months (I < 20) in the rainy season in eastern parts of the Pearl River basin; adverse trends are found in western parts of the basin. These results are in good agreement with changes in the number of dry months in the rainy season. Therefore, aridity index and SPI perform similarly well in reflecting dry conditions over the Pearl River basin. Relationships between I, SPI and RD indicate significant exponential correlation between I and SPI, and significant linear correlation between I and RD. The exponential correlation between I and SPI may be due to the logarithmic transformation of original monthly precipitation data.

  3. 3.

    Moisture flux analysis based on NCAR/NCEP dataset indicates a stronger intensity of water-vapor transport in the rainy season than that in winter (dry season), showing considerable influence of water-vapor flux on dry and wet conditions of the Pearl River basin. The source of water vapor lies mainly in the Indian peninsula and Bengal Bay. Good correlation can be identified between moisture budget, moisture content and number of wet months in winter over the Pearl River basin. Increasing moisture budget and moisture content gives rise to an increasing number of wet months in winter across the basin. Zhang et al. (2008c) indicated decreasing summer precipitation and increasing winter precipitation in the Pearl River basin, and attributed precipitation variations to changes in the East Asian summer monsoon system, which is in good agreement with the results of this study, i.e. drying tendency in the rainy season and wetting tendency in winter. Wang (2001) indicated the weakening of the Asian monsoon circulation after the 1970s which is not beneficial for the northward propagation of the rain belt, and it can explain wet and dry variations in the Pearl River basin. It should be noted here that in comparison with winter, worse correlation can be detected between moisture budget, moisture content and number of wet months in the rainy season, meaning that more factors than moisture budget and moisture content influence the dry and wet conditions in the rainy season, which needs further investigation in the on-going research.

  4. 4.

    To our knowledge, this study is the first that tries to explore spatial and temporal variations of dryness/wetness conditions by using more than one indicator, i.e. SPI and I, over the Pearl River basin and even in China. Comparison was made between dryness/wetness indicators by I and SPI with thee aim to show the applicability of these two techniques and to get a full picture of dryness/wetness variations in the Pearl River basin. Particularly, this study attempted to find out the causes behind the changing properties of dryness/wetness status in the Pearl River basin by using NCAR/NCEP reanalysis dataset. This study is of great scientific and practical merit towards the better understanding of impacts of climate changes, particularly the global warming, on water resources, agricultural activities and other human activities on river basin scales.