Extreme value analysis of streamflow time series in Poyang Lake Basin, China

Peng TIAN, Guang-ju ZHAO, Jing LI*, Kun TIAN

1. College of Water Resources and Architectural Engineering, Northwest A&F University, Yangling 712100, P. R. China

2. Institute of Soil and Water Conservation, Northwest A&F University, Yangling 712100, P. R. China

3. Institute of Soil and Water Conservation, Chinese Academy of Sciences and Ministry of Water Resources, Yangling 712100, P. R. China

4. State Key Laboratory of Hydroscience and Engineering, Tsinghua University, Beijing 100084, P. R. China

Extreme value analysis of streamflow time series in Poyang Lake Basin, China

Peng TIAN1, Guang-ju ZHAO2,3, Jing LI*1, Kun TIAN4

1. College of Water Resources and Architectural Engineering, Northwest A&F University, Yangling 712100, P. R. China

2. Institute of Soil and Water Conservation, Northwest A&F University, Yangling 712100, P. R. China

3. Institute of Soil and Water Conservation, Chinese Academy of Sciences and Ministry of Water Resources, Yangling 712100, P. R. China

4. State Key Laboratory of Hydroscience and Engineering, Tsinghua University, Beijing 100084, P. R. China

Extreme meteorological and hydrological events may cause major disasters and heavy social and economic losses. Therefore, more and more studies have focused on extreme hydro-meteorological events in various climates and geographic regions. Based on nearly 50 years of observed records of the Poyang Lake Basin, the occurrence and changing trends of extreme streamflow indices, including the annual maximum flow, annual peak-over-threshold flows, and low flows, were analyzed for ten hydrologicalstations. The results indicate that most annual maximum flows occurred from April to July, highly attributed to the Southeast Asian summer monsoons, whereas the annual minimum flows were concentrated between January and February. As for the low flow indices (the annual minimum flow, annual minimum 7-d flow, and annual minimum 30-d flow), a significant increasing trend was detected in most parts of the Poyang Lake Basin. The trends illustrate the potential effects of climate change and human activities on the hydrological cycle over the Poyang Lake Basin.

Poyang Lake Basin; extreme streamflow event; Mann-Kendall test; directional statistics

1 Introduction

Extreme meteorological and hydrological events all over the world may lead to major disasters and result in heavy social and economic losses, and even severe health consequences. Concerns over these extreme events have been increasing in recent decades (Houghton et al. 2001). In previous studies, significant changes of extreme streamflow were reported in some regions through investigation of the observed data. In Europe, significant positive flood trends were detected in the northern part (Kundzewicz and Robson 2004; Lindstr?m and Bergstr?m 2004). Douglas et al. (2000) found increasing trends in low flow from the mid-west towards the northeast of the USA. In China, Zhang et al. (2005) evaluated the relations betweentemperature, precipitation, and streamflow of the Yangtze River Basin from 1951 to 2002, which suggests that the present global warming would intensify the flood hazards in the basin. Yang et al. (2010) investigated the low flow characteristics of the Yellow River, and found that the low flow regime of the middle and lower Yellow River was decreasing overwhelmingly. On the other hand, some research showed no significant change in extreme streamflow. For instance, there is no regional flood trends detected in the UK, as well as most of the USA (Kundzewicz and Robson 2004). According to these studies, the spatial patterns of changes in floods and low flows were complicated and various around the world (Zhai et al. 2010). The variation of extreme meteorological and hydrological events needs to be clarified.

The Poyang Lake Basin has been faced with serious water resources problems. Both droughts and floods have occurred frequently in the basin in recent decades. Moreover, floods have increased in their severity since 1990 (Guo and Jiang 2008). The most severe floods of the last five decades in Poyang Lake occurred in 1998, 1996, and 1995. The runoff coefficients of the Poyang Lake Basin increased significantly and changed abruptly in the late 1980s and the early 1990s (Guo et al. 2006). Zhao et al. (2010) found that the average annual streamflow of the Poyang Lake Basin increased significantly over the last five decades and streamflow is more sensitive to the changes in precipitation than evaporation. Although some efforts have been undertaken to detect the changes in extreme meteorological and hydrological events, the occurrence and changing trends of extreme streamflow in the Poyang Lake Basin have not been sufficiently and systematically addressed by using a suitable approach. In addition, the driving forces for the changes of the extreme hydrological events have not been clearly identified.

The objective of this study was to analyze the spatial and temporal variation of extreme streamflow in the Poyang Lake Basin in order to identify whether the extreme streamflow had significantly changed during the period from 1957 to 2003 under the impacts of climate change and human activities. In detail, we would like to quantify the occurrence and changing trends of extreme events by using the directional statistics and the Mann-Kendall test.

2 Study area and data

The Poyang Lake Basin, located in the middle reaches of the Yangtze River Basin (Fig. 1), China, covers an area of 162 200 km2, which accounts for nearly 97% of the whole area of Jiangxi Province. Poyang Lake is a narrow outlet into the Yangtze River, which lies on the northern border of the province. As shown in Fig. 1, the five major rivers in Jiangxi Province flowing into Poyang Lake are the Ganjiang River, the Xiushui River, the Fuhe River, the Raohe River, and the Xinjiang River. The rivers downstream flow into the relatively broad alluvial valleys surrounding Poyang Lake. The Ganjiang River, extending 750 km, is the longest river in the region, which contributes more than 50% of the total discharge into Poyang Lake (Shankman et al. 2006). The whole basin is dominated by the Southeast Asian monsoon in summer, with the rainy season starting from April.

Fig. 1 Location of study area and hydrological stations in Poyang Lake Basin

In this study, the daily streamflow records from ten hydrological stations (shown in Fig. 1) in the five main tributaries of the Poyang Lake Basin were selected for investigation according to their lengths and continuities. The detailed information about the ten hydrological stations can be found in Table 1. The daily time series were validated by identifying and rectifying sequences of anomalous flows, and checking their completeness. Missing data were filled up by hydrological simulations. Of these stations, Hushan Station located in the Raohe River sub-basin has a two-year gap (1977-1978), which was filled up by a simple hydrological model implemented in PCRaster (Zhao et al. 2010).

Table 1 Detailed information of hydrological stations in Poyang Lake Basin

In this study, we selected six indices to describe the characteristics of the extreme streamflow including both low flow and high flow: the annual minimum flow (Qmin(m3/s)), annual minimum 7-d flow (Q7-dmin(m3/s)), annual minimum 30-d flow (Q30-dmin(m3/s)),annual maximum flow (Qmax(m3/s)), and annual peak-over-threshold flows (QPOT95(m3/s) andQPOT99(m3/s)).

Qmaxrepresents the maximum value of the flood events in each year, and only one value is detected every year. It represents the largest streamflow value of several observed high discharge values in wet years, whereas a relatively low flow value that may not necessarily be characterized as a flood at all in dry years will be selected (Svensson et al. 2005). A more representative way to describe the frequency of high flows in a continuous record is using a peak-over-threshold (POT) series consisting of a series of independent daily mean flows that exceed a certain threshold (Svensson et al. 2005). The thresholds with 95 percent (QPOT95) and 99 percent (QPOT99) values of the daily discharge records at each station were selected to describe the magnitude and frequency of the extreme high flows.

Qmin,Q7-dmin, andQ30-dminwere selected for low flow analysis.Q7-dmin, known as dry weather flow, is commonly used for low flow analysis (Yang et al. 2010). Comparably, the analysis ofQ7-dminandQ30-dminbased on a time series of long-duration low flows is less sensitive to measurement errors.

3 Methodologies

3.1 Directional statistics

Directional statistics is concerned mainly with observations that are unit vectors in the plane or in the three-dimensional space (Mardia 1999). The sample space is typically a circle or a sphere, so that the occurrence ofQmaxandQmincan be assumed to be an object in an annual circle, and thus be described by means of directional statistics. So far, directional statistics has been widely applied in the hydro-climatic field (Bayliss and Jones 1993; Cunderlik and Ouarda 2009). This approach transforms the original occurrence dates of individual extreme events into a directional variable. TakingQmaxas an example,θiis an angular value that denotes the Julian date (Di) on whichQmaxoccurs in theith year, and can be converted by

wheredis the number of days in a year equal to 365, or 366 for a leap year.

To identify the characteristics of the occurrence and magnitude ofQmax, the regularity ofQmaxoccurrence was introduced. The mean direction (θ) represents a directional location of a sample consisting of dates of flood occurrence. A convenient method for describing the variability of the individual dates of flood occurrence around the mean value can be defined as (Bayliss and Jones 1993)

whereNis the number of flood events.

mrepresents a directional location measuring a series of dates for flood occurrence, and can be defined as

3.2 Trend detection

A large number of parametric and non-parametric methods have been successfully applied in hydro-climatic series analysis (Li et al. 2009; Zhao et al. 2009). In this study, we analyzed the temporal trends of extreme streamflow indices by using both the least square test and the non-parametric Mann-Kendall (MK) test (Mann 1945; Kendall 1975). The MK test has been widely applied to detecting a trend in hydro-climatic time series (Yue and Wang 2002; Novotny and Stefan 2007; Li et al. 2009). The details of the method have been described in previous literature (Novotny and Stefan 2007; Zhao et al. 2009).

The MK test requires the time series data to be serially independent (Yue and Wang 2002). The present study undertook the serial correlation detection by examining the autocorrelation coefficients of the time series. As shown in Fig. 2, almost all of the coefficients are lower than the 95% confidence level, which means that the streamflow series (Qmax,Qmin,Q7-dmin, andQ30-dmin) are time independent.

Fig. 2 Autocorrelation analysis of streamflow time series at Xiashan Station during period from 1957 to 2003

4 Results and discussion

The occurrence and magnitude ofQmaxandQminevents were extracted from the continuous daily streamflow records at ten hydrological stations in the Poyang Lake Basin. The occurrence of the extreme events was detected by directional statistics. Both the simple linear test and the MK test were used for assessment of the variability of trends in extreme flow indices.

4.1 Extreme event occurrence

Fig. 3 shows the distribution of dates forQmaxandQminevents at Waizhou Station during the period from 1957 to 2003 (as points plotted on a unit circle, respectively). From Fig. 3(a), it can be seen thatQmaxmostly occurred from April to July. The Poyang Lake Basin lies in a wet climate zone, which is dominated by the Southeast Asian monsoon in summer. The summer monsoon brings atmospheric moisture into the basin, leading to large amounts of rainfall during the wet season. According to the monthly precipitation distribution, rainfall was mainly concentrated between March and June, which causedQmaxto occur in the wet season (Zhao et al. 2010). Fig. 3(b) indicates thatQminoccurred from October to March in the following year.

Fig. 3 Occurrence ofQmaxandQminat Waizhou Station in Poyang Lake Basin

Generally speaking, the extreme events obviously illustrated the seasonality of precipitation.Qmaxoccurred in the wet season in a year from April to August, whereasQminoccurred in the dry season from October to March in the following year. On the other hand, the occurrence dates ofQmaxwere concentrated in a shorter period (three to four months) than those ofQmin(about six months). This may be associated with the regional climate patterns and monsoon activities.

Fig. 4 shows the seasonality and flow regularity ofQmaxandQmin. It reveals a high degree of seasonality of floods and droughts, since there is a comparatively small range of the values for the mean dates of the occurrence ofQmaxandQmin. All the mean dates of the occurrence fall within May 20 to June 20 forQmax, and December 25 to February 2 forQmin.Previous studies have also found the highest flood frequency in June in the Poyang Lake Basin (Guo et al. 2006). Also, the occurrence ofQmaxandQminis fairly regular in the Poyang Lake Basin, implying a low variability in the occurrence dates ofQmaxandQmin, with values ofranging from 0.635 to 0.855, and 0.405 to 0.785, respectively. Through comparison, the broader range of low flow regularity indicates a higher variability of extreme drought events.

Fig. 4 Seasonality ofQmaxandQmin, and flow regularities at ten stations

4.2 Trend analysis

Both the MK test and the least square test were applied to examining the temporal variation of extreme streamflow series (Qmax,Qmin,Q7-dmin, andQ30-dmin) at ten hydrological stations in the Poyang Lake Basin. Considering the homogeneity of some stations in the same sub-catchment, we chose six stations (Xiashan, Waizhou, Hushan, Meigang, Wanjiabu, and Lijiadu) in five sub-catchments for the trend analysis.

The annual variations ofQ7-dmintrends estimated by the simple linear method in the Poyang Lake Basin are shown in Fig. 5. Notable upward trends were detected in the time series ofQ7-dminamong most of the stations, except Lijiadu Station.Q7-dminat Waizhou Station downstream of the Ganjiang River shows an evident upward trend with a mean annual increase of 3.70 m3/s. In contrast,Q7-dminat Lijiadu Station shows a downward trend with a mean annual decrease of 0.27 m3/s (as shown in Table 2). This finding is consistent with the average annual streamflow trends in the Poyang Lake Basin (Zhao et al. 2010). The main reason for the abnormal trends of extreme flow events at Lijiadu Station may result from the non-significant changes of precipitation in this catchment.

Table 2 displays the linear trends ofQmax,Qmin,Q7-dmin, andQ30-dmin. The positive values indicate that the series have increasing trends, and vice versa. In general,Qmaxdid not show evident trends at most stations for the study period, and the changing trends ranged from–11.79 to 29.15 m3/s. As for low flows, notable results can be found at most stations except for Lijiadu, illustrating upward trends forQmin,Q7-dmin, andQ30-dmin. As shown in Table 2, the trends at Waizhou and Xiajiang stations in the Ganjiang River sub-basin are more significantwith upward trends from 2.51 to 5.15 m3/s, which is attributed to large daily average discharges at these two stations (Fig. 5). Similarly, Guo et al. (2006) found significant increasing trends forQ7-dminat most stations across the whole basin. According to our analysis, the increasing trends for the extreme low flows were due to both the increasing annual precipitation and anthropologic activities. There is no evident trend forQmax, although most studies indicated significant increasing trends in both annual precipitation and summer rainfall (Guo and Jiang 2008; Zhao et al. 2010). Many studies have investigated whether extreme high or low flows are associated with climate change or land-use/land-cover change and concluded that both climate change and human activities (like reservoir operation) should be considered in assessing the impacts of environmental changes on streamflow processes (Wang et al. 2008).

Fig. 5 Variation ofQ7-dminfor six hydrological stations of Poyang Lake Basin

The trends of the extreme streamflow series (Qmax,Qmin,Q7-dmin, andQ30-dmin) for each hydrological station in the Poyang Lake Basin were analyzed with the MK test method at a significance level of 0.05. The statisticZreflecting the time series changing trends was calculated (Fig. 6). TheZvalue is equal to ±1.96 at the significance level of 0.05.

Table 2 Mean annual increment of extreme streamflow series in Poyang Lake Basin (m3/s)

Fig. 6 Extreme flow trends detected by MK method in Poyang Lake Basin

As shown in Fig. 6,Zvalues forQmaxrange from –1.00 to 0.90, which illustrate that there are no significant trends forQmaxat all stations. Generally, the MK test results are consistent with linear trends as displayed in Table 2, and only three stations (Bashang, Waizhou, Xiashan) in the Ganjiang River sub-basin show downward trends, but not significantly. Comparably, the results of trend analysis for low flows are most insightful, indicating similar trends obtained by examiningQmin,Q7-dmin, andQ30-dmin(Fig. 6). Among the ten hydrological stations, only two stations (Lijiadu and Bashang) did not show significant increasing trend at the 95% confidence level. The increasing precipitation, to some degree, is associated with the increasing trends of the extreme low flows. According to previous studies, Zhao et al. (2010) found that the streamflow increased significantly at a 95% confidence level except for Lijiadu Station in the dry season, which might be due to the non-significant changes of the precipitation at this station.

Similar linear trends were found inQPOT95andQPOT99in the Poyang Lake Basin. Here we only present the variation ofQPOT95and the annual number of days aboveQPOT95(Fig. 7). The results indicate that there are no significant trends for all the stations, which is similar toQmax, as described above. As shown in the figure, a relatively wet period is detectable according to both the mean discharge values and the annual number of days above theQPOT95since the beginning of the 1990s. This is also consistent with the investigation result in Guo and Jiang (2008) in the Poyang Lake Basin: they found that the frequency of extreme precipitation had become increasingly significant since 1990 and extreme events were becoming more concentrated in shorter periods. This might be attributed to the weakened thermal contrast between the tropical ocean and the Asian inland region, which causes a weaker summer monsoon that stays longer in the Yangtze River Basin and results in quasi-stationary fronts with heavy rainfall systems from late spring to midsummer (Wang and Zhou 2005).

5 Conclusions

In this study, we selected six extreme streamflow indices,Qmax,Qmin,Q7-dmin,Q30-dmin,QPOT95, andQPOT99, to examine their occurrence and spatio-temporal trends in the Poyang Lake Basin. The main conclusions are as follows:

(1) A good seasonality of floods and droughts can be clearly found in the Poyang Lake Basin. The occurrence dates ofQmaxandQminare concentrated in the period from May to June, and from December to February in the following year, respectively. The occurrence ofQmaxandQminis fairly regular in the Poyang Lake Basin, implying low variability in the dates ofQmaxandQmin, with extreme flow regularity ranging from 0.635 to 0.855 and from 0.405 to 0.785, respectively.

(2) There is generally a good agreement between the results of the trend analysis using the linear regression method and the MK method. The major hydrological characteristics ofQmax,QPOT95, andQPOT99are shown to have no statistically significant trends at most stationsin the whole basin. The extreme low flows have significant increasing trends, particularly at Waizhou and Xiashan stations. The MK test shows thatQmin,Q7-dmin, andQ30-dminin most parts of the Poyang Lake Basin displayed significant increasing trends.

Fig. 7 Mean discharge aboveQPOT95and number of days aboveQPOT95at six hydrological stations in Poyang Lake Basin

(3) The main factors leading to complicated trends of extreme streamflow may result from spatial and temporal differences of precipitation, together with the impacts of intensive human activities on hydrological processes. Further investigation should be undertaken to thoroughly explain how these factors influence the extreme streamflow in the Poyang Lake Basin.

Bayliss, A. C., and Jones, R. C. 1993.Peaks-over-Threshold Flood Database: Summary Statistics and Seasonality.Wallingford: Institute of Hydrology.

Cunderlik, J. M., and Ouarda, T. B. M. J. 2009. Trends in the timing and magnitude of floods in Canada.Journal of Hydrology, 375(3-4), 471-480. [doi:10.1016/j.jhydrol.2009.06.050]

Douglas, E. M., Vogel, R. M., and Kroll, C. N. 2000. Trends in floods and low flows in the United States: Impact of spatial correlation.Journal of Hydrology, 240(1-2), 90-105. [doi:10.1016/S0022-1694(00) 00336-X]

Guo, H., Jiang, T., Wang, G. J., Su, B. D., and Wang, Y. J. 2006. Observed trends and jumps of climate change over Lake Poyang Basin, China: 1961-2003.Journal of Lake Sciences, 18(5), 443-451 (in Chinese).

Guo, H., Hu, Q., and Jiang, T. 2008. Annual and seasonal streamflow responses to climate and land-cover changes in the Poyang Lake basin, China.Journal of Hydrology, 355(1-4), 106-122. [doi:10.1016/j. jhydrol.2008.03.020]

Guo, H., and Jiang, T. 2008. Trend analysis of flood peak/dry season flow rate in Poyang Lake Basin.Journal of Natural Disasters, 17(3), 75-80. (in Chinese)

Houghton, J. T., Ding, Y., Griggs, D. J., Noguer, M., van der Linen, P. J., and Dai, X. 2001. Climate change 2001: The scientific basis.Contribution of Working Group 1 to the Third Assessment Report of the Intergovernmental Panel on Climate Change(IPCC), 944. Cambridge: Cambridge University Press.

Kendall, M. G. 1975.Rank Correlation Methods. London: Griffin.

Kundzewicz, Z. W., and Robson, A. J. 2004. Change detection in hydrological records: A review of the methodology.Hydrological Sciences Journal, 49(1), 7-19. [doi:10.1623/hysj.49.1.7.53993]

Li, Z., Liu, W. Z., Zhang, X. C., and Zheng, F. L. 2009. Impacts of land use change and climate variability on hydrology in an agricultural catchment on the Loess Plateau of China.Journal of Hydrology, 377(1-2), 35-42. [doi:10.1016/j.jhydrol.2009.08.007]

Lindstr?m, G., and Bergstr?m, S. 2004. Runoff trends in Sweden 1807-2002.Hydrological Sciences Journal, 49(1), 69-83. [doi:10.1623/hysj.49.1.69.54000]

Mann, H. B. 1945. Nonparametric tests against trend.Econometrica, 13(3), 245-259.

Mardia, K. V. 1999. Directional statistics and shape analysis.Journal of Applied Statistics, 26(8), 949-957. [doi:10.1080/02664769921954]

Novotny, E. V., and Stefan, H. G. 2007. Streamflow in Minnesota: Indicator of climate change.Journal of Hydrology, 334(3-4), 319-333. [doi:10.1016/j.jhydrol.2006.10.011]

Shankman, D., Keim, B. D., and Song, J. 2006. Flood frequency in China’s Poyang Lake Region: Trends and teleconnections.International Journal of Climatology, 26(9), 1255-1266. [doi:10.1002/joc.1307]

Svensson, C., Kundzewicz, Z. W., and Maurer, T. 2005. Trend detection in river flow series: 2. Flood and low-flow index series.Hydrological Sciences Journal, 50(5), 811-824. [doi:10.1623/hysj.2005.50.5.797]

Wang, W., Chen, X., Shi, P., and van Gelder, P. H. A. J. M. 2008. Detecting changes in extreme precipitation and extreme streamflow in the Dongjiang River Basin in southern China.Hydrology and Earth System Sciences, 12(1), 207-221. [doi:10.5194/hess-12-207-2008]

Wang, Y, Q., and Zhou, L. 2005. Observed trends in extreme precipitation events in China during 1961-2001 and the associated changes in large-scale circulation.Geophysical Research Letters, 32, L09707. [doi: 10.1029/2005GL022574]

Yang, T., Xu, C. Y., Shao, Q. X., Chen, X., Lu, G. H., and Hao, Z. C. 2010. Temporal and spatial patterns of low-flow changes in the Yellow River in the last half century.Stochastic Environmental Research and Risk Assessment, 24(2), 297-309. [doi:10.1007/s00477-009-0318-y]

Yue, S., and Wang, C. Y. 2002. Applicability of prewhitening to eliminate the influence of serial correlation on the Mann-Kendall test.Water Resources Management, 38(6), 1068. [doi:10.1029/2001WR000861]

Zhai, J. Q., Su, B. D., Krysanova, V., Vetter, T., Gao, C., and Jiang, T. 2010. Spatial variation and trends in PDSI and SPI indices and their relation to streamflow in 10 large regions of China.Journal of Climate, 23(3), 649-663. [doi:10.1175/2009JCLI2968.1]

Zhang, Q., Jiang, T., Gemmer, M., and Becker, S. 2005. Precipitation, temperature and runoff analysis from 1950 to 2002 in the Yangtze Basin, China.Hydrological Sciences Journal, 50(1), 65-80. [doi:10.1623/ hysj.50.1.65.56338]

Zhao, G. J., H?rmann, G., Fohrer, N., Gao, J. F., Zhai, J. Q., and Zhang, Z. X. 2009. Spatial and temporal characteristics of wet spells in the Yangtze River Basin from 1961 to 2003.Theoretical and Applied Climatology, 98(1-2), 107-117. [doi:10.1007/s00704-008-0099-0]

Zhao, G. J., H?rmann, G., Fohrer, N., Zhang, Z. X., and Zhai, J. Q. 2010. Streamflow trends and climate variability impacts in Poyang Lake Basin, China.Water Resources Management, 24(4), 689-706. [doi:10.1007/s11269-009-9465-7]

This work was supported by the National Basic Research Program of China (the 973 Program, Grant No. 2007CB407203), and the Chinese Postdoctoral Science Fund Project (Grant No. 20110490402).

*Corresponding author (e-mail:lijinglee@tom.com)

Received Jul. 9, 2010; accepted Apr. 21, 2011