HYDROLOGICAL PROCESSES Hydrol. Process. 22, 4997– 5003 (2008) Published online 26 August 2008 in Wiley InterScience (www.interscience.wiley.com) DOI: 10.1002/hyp.7119 Multifractal detrended fluctuation analysis of streamflow series of the Yangtze River basin, China Qiang Zhang,1,2 * Chong-Yu Xu,3 Yongqin David Chen4 and Zuguo Yu5,6 1 State Key Laboratory of Lake Science and Environment, Nanjing Institute of Geography and Limnology, Chinese Academy of Sciences, Nanjing 210008, China 2 Institute of Space and Earth Information Science, The Chinese University of Hong Kong, Hong Kong, China 3 Department of Geosciences, University of Oslo, Norway 4 Department of Geography and Resource Management, The Chinese University of Hong Kong, Hong Kong, China 5 School of Mathematical Sciences, Queensland University of Technology, GPO Box 2434, Brisbane, Q4001, Australia 6 School of Mathematics and Computational Science, Xiangtan University, Hunan 411105, China Abstract: Scaling and multifractal properties of the hydrological processes of the Yangtze River basin were explored by using a multifractal detrended fluctuation analysis (MF-DFA) technique. Long daily mean streamflow series from Cuntan, Yichang, Hankou and Datong stations were analyzed. Using shuffled streamflow series, the types of multifractality of streamflow series was also studied. The results indicate that the discharge series of the Yangtze River basin are non-stationary. Different correlation properties were identified within streamflow series of the upper, the middle and the lower Yangtze River basin. The discharge series of the upper Yangtze River basin are characterized by short memory or anti-persistence; while the streamflow series of the lower Yangtze River basin is characterized by long memory or persistence. hq vs q curves indicate multifractality of the hydrological processes of the Yangtze River basin. hq curves of shuffled streamflow series suggest that the multifractality of the streamflow series is mainly due to the correlation properties within the hydrological series. This study may be of practical and scientific importance in regional flood frequency analysis and water resource management in different parts of the Yangtze River basin. Copyright 2008 John Wiley & Sons, Ltd. KEY WORDS multifractal property; multifractal detrended fluctuation analysis; hydrological processes; the Yangtze River basin Received 10 February 2008; Accepted 26 June 2008 INTRODUCTION As one of the major components of the hydrological cycle, streamflow analysis and modelling have received considerable attention in hydrological science in recent decades. The last 20 years or so has witnessed great progress in studying the scaling behaviour of some geophysical fields, including streamflow, rainfall, temperature, etc.(Lovejoy and Mandelbrot, 1985; Pandey et al., 1998). It is accepted that there exists persistence in climatological and hydrological series over a wide range of time scales. Generally, the persistence on long time scales is larger than that on short time scales (Pelletier and Turcotte, 1997). Hurst (1951) and Hurst et al. (1965) demonstrated persistence using the rescaled-range technique, identifying a power-law rescale-range plot in the climatological and hydrological series with an average exponent of 0Ð73. Actually, many signals present complex behaviour that can exhibit long-range power law correlation and/or nonstationary trends, e.g. DNA sequences (Yu et al., 2001; Yu et al., 2004) and meteorological measurements (Olsson, 1996; Lin and Fu, 2008). This complex behaviour can be characterized by the famous * Correspondence to: Qiang Zhang, Institute of Space and Earth Information Science, The Chinese University of Hong Kong, Hong Kong, China. E-mail: zhangqnj@gmail.com Copyright 2008 John Wiley & Sons, Ltd. Hurst exponent, or scaling exponent, which quantifies the correlation properties of a signal. It is feasible to characterize these diverse phenomena by using critical exponents and thus to identify similarities between different systems (Chianca et al., 2005). The currently well-evidenced global warming is believed to accelerate the hydrological cycle (Menzel and Bürger, 2002; Xu and Singh, 2005; Zhang et al., 2008). Therefore, increasing attention has been paid to sustainability and hydro-environmental protection, which require modelling of dynamic processes such as runoffinduced wash-off from impermeable surfaces and flood prediction from ungauged basins. Hydrologists and meteorologists come to realize the importance of investigating scaling properties of hydro-meteorological series in that good understanding of scaling properties of hydrological system is of great importance for hydrological modelling, regionalization of flood frequency and assessment of hydrological conditions in ungauged area based on gauged regions. Peng et al. (1994) introduced the detrended fluctuation analysis (DFA), which has since then been widely used to detect the long-range correlations in stationary and nonstationary time series (Maraun et al., 2004; Bunde et al., 2006). The DFA method has been applied successfully in diverse fields such as DNA and protein sequences, heart rate dynamics, weather records (Kantelhardt et al., 2002; Yu et al., 4998 Q. ZHANG ET AL. 2006). The multifractal detrended fluctuation analysis (MF-DFA) proposed by Kantelhardt et al. (2002) is a modified version of DFA to detect multifractal properties of time series. It allows a reliable multifractal characterization of nonstationary time series such as geophysical phenomena (Kantelhardt et al., 2002). It is now known that multifractal is the appropriate framework for scaling fields and time series and thus can provide the natural framework for analysing and modelling various geophysical processes (Pandey et al., 1998). Gupta et al. (1994) reported that streamflows are multiscaling with basin area. Tessier et al. (1996) analysed multifractal properties of river flow series from 30 small river basins in France. These researches provided a broader framework in modelling the rainfall-runoff processes, such as topography and river network, that generate and modify the streamflow processes through the basin (Gupta et al., 1996; Movahed and Hermanis, 2008). The Yangtze River (Changjiang) (91 ° E–122 ° E, ° 25 N–35 ° N) is the longest river in China and the third longest river in the world. It has adrainage area of 1 808 500 km2 with mean annual discharge of 23 400 m3 s1 measured at Hankou Station. The Yangtze River originates in the Qinghai-Tibet Plateau and flows about 6300 km eastwards to the East China Sea (Zhang et al., 2006). The upper Yangtze River basin is from the origin to Yichang, with a length of 4504 km and drainage area of about 1Ð0 ð 106 km2 . The river reach between Yichang and Hukou (the outlet of Poyang Lake) is the middle Yangtze River basin, with a length of 955 km and drainage area of 0Ð68 ð 106 km2 . The lower Yangtze River basin is from Hukou to the river mouth, with a length of 938 km and drainage area of 0Ð12 ð 106 km2 (CWRC, 2000). The river is located in the monsoon region of East Asia subtropical zone, and has a mean annual precipitation of about 1090 mm (Zhang et al., 2005). Moreover, the Yangtze River basin plays a considerable role in the socio-economic development of China. However, frequent floods have exerted tremendous impacts on the local ecological environment and on human society in the Yangtze River basin. Historical flood records (CWRC, 2000) showed that, during the past 200 years, eight floods occurred in the third cold period of the Little Ice Age, and 19 floods occurred in the warm 20th century. 1990–2000 is the warmest period in the past 1000 years and seven floods occurred in that period. Besides flood events, drought hazards in the Yangtze River basin have also given increasing concerns. Gemmer et al. (2007) indicated that drought frequency increased in the middle Yangtze River Basin in May, September and October, which is mainly reflected by downward trends in the monthly streamflow, especially in May. During these months, flood and drought hazards in the Yangtze River Basin have been aggravated. Therefore, introducing sound measures for basin-scale water resource management will rely heavily on a good understanding of scaling properties of hydrological series of the Yangtze River basin based on long streamflow series. However, little research has been found on this topic. Undoubtedly, a Copyright 2008 John Wiley & Sons, Ltd. better understanding of the statistical properties of hydrological series in terms of multifractal parameters is of scientific and practical interest. Therefore, the objective of this work is to investigate the complex behaviour of the streamflow series over space and time scales, using the multifractal framework. The paper is organized as follows. First we give a description of the river flow data from four hydrological stations located along the mainstream of the Yangtze River. The methods used in this study are then introduced in detail, followed by a discussion of the results of this study. Concise conclusions are listed in the final section. DATA AND METHODOLOGY Data Long daily mean streamflow series extracted from four hydrological stations, i.e. Cuntan, Yichang, Hankou and Datong, along the mainstream of the Yangtze River basin have been analysed (Figure 2). The location of these four stations is shown in Figure 1. Detailed information about the hydrological dataset is given in Table I. The streamflow data represent the hydrological conditions of the upper, middle and lower Yangtze River basins, respectively. The quality of the streamflow series is controlled by the Changjiang Water Resources Commission, Ministry of Water Resources, China. The length of the data series is more than 50 years, with some more than 100 years. Methodology Multifractal detrended fluctuation analysis (MF-DFA) is a generalization of standard DFA by identifying the scaling of the qth-order moments of the time series, which may be non-stationary (Kantelhardt et al., 2002). Movahed et al. (2006) described the procedure of MFDFA. Actually, the first three steps are the same as those in the conventional DFA procedure. Assuming xk is a time series, k D 1, . . . , N. Step 1. Determine the ‘profile’ Yi D i [xk < x >], kD1 i D 1, . . . , N, where < x > is the mean of xk . Step 2. Divide the profile Yi into Ns D intN/s nonoverlapping segments of equal length s; here intN/s is Table I. Detailed hydrological records of stations along the mainstream of the Yangtze River basin Station name Cuntan station Yichang station Hankou station Datong station Time interval Drainage area (ð104 km2 ) 86Ð66 100Ð55 148Ð80 170Ð54 1 1 1 1 Jan. Jan. Jan. Jan. 1893–31 1946–31 1952–31 1950–31 Dec. Dec. Dec. Dec. 2004 2004 2001 2000 Hydrol. Process. 22, 4997– 5003 (2008) DOI: 10.1002/hyp 4999 MULTIFRACTAL DETRENDED FLUCTUATION ANALYSIS OF STREAMFLOW SERIES Figure 1. Location of the Yangtze River basin and hydrological stations 10 x 104 Cuntan 5 0 10 0 x 5000 10000 15000 20000 25000 30000 35000 40000 45000 104 Yichang Streamflow (m3/s) 5 0 10 0 5000 10000 15000 20000 25000 x 104 Hankou 5 0 10 0 x 2000 4000 6000 8000 10000 12000 14000 16000 18000 20000 18000 20000 104 Datong 5 0 0 2000 4000 6000 8000 10000 12000 14000 16000 Time (days) Figure 2. Streamflow series of Cuntan, Yichang, Hankou and Datong stations of the Yangtze River basin Copyright 2008 John Wiley & Sons, Ltd. Hydrol. Process. 22, 4997– 5003 (2008) DOI: 10.1002/hyp 5000 Q. ZHANG ET AL. the integer part of N/s. Since the length N of the series is often not a multiple of the timescale s considered, a short part at the end of the profile may remain. To retain this part of the series, the same procedure is repeated starting from the opposite end. Thereby, 2Ns segments are obtained. Step 3. Calculate the local trend for each of the 2Ns segments by a least squares fit of the series. Then determine the variance 1 F s, D fY[ 1s C i] y ig2 s iD1 and surrogate time series. The random shuffling of time series destroys the long range correlation. Therefore, if the multifractality is only due to the long range correlation, hshuf q D 0Ð5. The multifractality nature due to the broadening of the PDF signals is not affected by the shuffling procedure. However, if both kinds of multifractality are present, the shuffled series will show weaker multifractality when compared with the original time series. s 2 for each segment , 1 s iD1 s D 1, . . . , Ns , and F2 s, D fY[N Ns s C i] y ig2 , for D Ns C 1, . . . , 2Ns . Here, y i is the fitting polynomial in segment . Linear, quadratic, cubic or higher order polynomials can be used in the fitting procedure (DFA1, DFA2, DFA3, . . DFAm). Step 4. Average over all segments to obtain qth-order the2N s 2 1 fluctuation function, defined as Fq s D 2N [F s, s D1 1/q ]q/2 , where q 6D 0, s ½ m C 2. Step 5. Determine the scaling behaviour Fq s / shq of the fluctuation functions by the log–log plot of Fq s versus s for each value of q. For stationary time series, the exponent h2 for small time scales is identical to the well-known Hurst exponent H. For non-stationary signal, the relation between the exponent h2 for small scales and the Hurst exponent H is H D h2 1 (Hu et al., 2001). For small scales where the effect of the sinusoidal trend is not pronounced, h2 > 1 indicates that the time series is nonstationary (Movahed et al., 2006). It is well known that for uncorrelated series, the scaling exponent H equals to 0Ð5; 0Ð5 < H < 1 indicates long memory or persistence; 0 < H < 0Ð5 indicates short memory or anti-persistence. Hence we can use the value of h2 to determine whether a time series is stationary or nonstationary and detects its correlations. In general, two different types of multifractality in time series can be distinguished (Movahed et al., 2006): (1) multifractality due to a broadening of the probability density function (PDF) of the time series. In this case the multifractality cannot be removed by shuffling the series; (2) multifractality due to different correlations in small and large scale fluctuations. In this case the time series have a PDF with finite moments. Therefore the shuffled time series will show mono-fractal scaling because all long-term correlations are destroyed by the shuffling procedure. The easiest way to clarify the type of multifractality is by analysing the corresponding shuffled Copyright 2008 John Wiley & Sons, Ltd. RESULTS AND DISCUSSION Figure 3 displays the DFA1 exponent of the streamflow series of Cuntan, Yichang, Hankou and Datong stations of the Yangtze River basin. Based on log-log plots of Fq s versus s of the streamflow series, one crossover point can be detected for the three curves of Fq s versus s. The timing for these crossover points is between 346 and 398 days, and is due to annual periodicity. To determine the statistical properties of the fluctuations of streamflow series, we compute the three scaling exponents for smaller time scales. The h2 values of the streamflow series of these four stations are 1Ð3027 š 0Ð0141 (Cuntan station), 1Ð3637 š 0Ð0095 (Yichang station), 1Ð4911 š 0Ð0063 (Hankou station) and 1Ð6659 š 0Ð0132 (Datong station) respectively. The h2 values of the streamflow series of Cuntan, Yichang, Hankou and Datong station are all larger than 1. These numerical results indicate that these four streamflow series are nonstationary signals. Therefore, based on the relationship between the Hurst exponent and the exponent h2 for small scales, i.e. H D h2 1, we obtained associated Hurst exponents for the streamflow series of Cuntan, Yichang, Hankou and Datong stations as 0Ð3027 š 0Ð0141, 0Ð3637 š 0Ð0095, 0Ð4911 and 0Ð6659 š 0Ð0132. Therefore, for Cuntan, Yichang and Hankou stations, the streamflow fluctuations are characterized by short memory or anti-persistence. Moreover, the Hurst exponent of the hydrological series from the Hankou station is close to 0Ð5, implying that the streamflow series of the Hankou station is close to being an uncorrelated series. For the streamflow fluctuations at Datong station in the lower Yangtze River basin, the numerical results suggest long memory or persistence. The above results indicate that the streamflow series of the four stations of the Yangtze River basin are nonstationary processes. Furthermore, the streamflow series of the Cuntan, Yichang and Hankou stations are characterized by short memory, and the streamflow series of the Datong stations by long memory. The result of the MF-DFA procedure is the family of the generalized Hurst exponents hq (Figure 4). For an actual multifractal signal, hq is a decreasing function of q; while for a monofractal signal, hq is of a constant value. It can be seen from Figure 4 that hq vs q curves of original streamflow series indicate strong dependence of hq on q, suggesting that the streamflow series of the four hydrological stations in the Yangtze River basin are characterized by multfractality. Here we are also interested in the possible source of multifractality. To clarify the Hydrol. Process. 22, 4997– 5003 (2008) DOI: 10.1002/hyp 5001 MULTIFRACTAL DETRENDED FLUCTUATION ANALYSIS OF STREAMFLOW SERIES 6.5 6.5 The slope is 0.2598 ± 0.0134 6 5.5 5 log10 F(s) log10 F(s) 5.5 Crossover point P1 Cuntan station 4.5 4 5 Crossover point P2 Yichang station 4.5 4 The slope is 1.3027 ± 0.0141 The slope is 1.3637 ± 0.0095 3.5 3.5 3 The slope is 0.2193 ± 0.0110 6 0 1 2 3 4 3 5 0 1 2 log10 s 3 4 5 log10 s 7 6.5 The slope is 0.2364 ± 0.0099 6 The slope is 0.3340 ± 0.0132 6 log10 F(s) log10 F(s) 5.5 Hankou station 5 Crossover point P3 4.5 5 Datong station Crossover point P4 4 4 3 0 1 2 3 The slope is 1.6659 ± 0.0132 3 The slope is 1.4911 ± 0.0063 3.5 4 2 5 0 1 2 log10 s 3 4 5 log10 s 6.5 6 Crossover points log10 F(s) 5.5 5 4.5 Cuntan station 4 Hankou station 3.5 Datong station Yichang station 3 2.5 0 1 2 3 4 5 log10 s Figure 3. Log–log plots of Fq s versus s of streamflow series of Cuntan, Yichang, Hankou and Datong stations in the Yangtze River basin 1.6 Cuntan Yichang Hankou Datong Cuntan (shuffled) Yichang (shuffled) Hankou (shuffled) Datong (shuffled) 1.5 1.4 1.3 1.2 h (q) 1.1 1 0.9 0.8 0.7 0.6 0.5 0.4 0 2 4 6 8 10 q Figure 4. Generalized Hurst exponent hq as a function of q for original and shuffled streamflow series of Cuntan, Yichang, Hankou and Datong stations Copyright 2008 John Wiley & Sons, Ltd. type of multifractality, we used a shuffled streamflow series. Figure 4 shows hq vs q curves of the shuffled streamflow series: it can be seen that hq vs q curves of the shuffled streamflow series indicate almost independence of hq on q. The hq values are largely equal to 0Ð5. If only the breadth of the PDF is responsible for the multifractality, hq D hshuf q D 0 can be expected. However, if only correlation multifractality is present, one can expect hshuf q D 0Ð5 (Movahed et al., 2006). It can be seen from Figure 4 that the hq vs q curves of shuffled streamflow series are almost independent of q values and hq values are mostly equal to 0Ð5. Therefore, we can conclude that the multifractality of the streamflow series of the Yangtze River basin is mainly due to the correlation properties (short-term correlation for Cuntan, Yichang and Hankou stations; long-term correlation for Datong station). Hydrol. Process. 22, 4997– 5003 (2008) DOI: 10.1002/hyp 5002 Q. ZHANG ET AL. Streamflow series are the combined results of the precipitation phenomena and the impact of other basin factors such as soil moisture, plant coverage, and channel geometry. Therefore the gauged streamflow series are the results of the overall complex interactions between precipitation input and the basin factors that modify it (Pandey et al., 1998). It can be seen from Figure 3 that similarities of h2 values can be identified between Cuntan and Yichang stations, but different h2 values between Cuntan, Yichang, Hankou and Datong station. Gupta et al. (1996) exploited the hypothesis of statistical self-similarity, or scaling invariance, in the spatial variability of rainfall, channel network structures and floods and also the distributed rainfall-landform-runoff relationships. Xu et al. (2006) divided the whole Yangtze River basin into three parts along the longitude from west to east, which correspond well with the decrease in altitude. The upper region (<¾104 ° E) has a mean altitude of 2551 m above sea level (m.a.s.l), and the middle (>¾104 ° E and <¾112 ° E) and lower regions (>¾112 ° E) have mean altitudes of 627 and 113 m.a.s.l, respectively. Cuntan and Yichang stations are located in the upper Yangtze River basin, being characterized by similar topography and channel network structures. Hankou station is located in the middle Yangtze River basin. Datong station is located in the lower part of the Yangtze River basin, and the topographical properties are different from those of the middle and upper Yangtze River. The different scaling properties of the streamflow series from Cuntan, Yichang, Hankou and Datong stations are probably due to these factors mentioned above. CLOSING REMARKS The scaling and multifractal properties of long streamflow series of the Yangtze River basin have been analysed by using a multifractal detrended fluctuation analysis technique. Furthermore, using a shuffling procedure, the types of multifractality of the streamflow series of the Yangtze River basin have been established. Several interesting findings have been obtained as follows: (1) Log-log plots of Fq s versus s of streamflow series show that the h2 values of the streamflow series of these four stations are 1Ð3027 š 0Ð0141 (Cuntan station), 1Ð3637 š 0Ð0095 (Yichang station), 1Ð4911 š 0Ð0063 (Hankou station) and 1Ð6659 š 0Ð0132 (Datong station) respectively. All these h2 values suggest that the hydrological processes of the Yangtze River basin are non-stationary. Moreover, the h2 values together with the relationship between Hurst exponent and the exponent h2 for small scales, i.e. H D h2 1, indicate that the streamflow series of the upper and middle Yangtze River basin (Xu et al., 2006) are characterized by short memory or antipersistence; while the streamflow series of the lower Yangtze River basin is characterized by long memory or persistence. Copyright 2008 John Wiley & Sons, Ltd. (2) To decide the types of multifractality of the streamflow series, the original streamflow series was shuffled. Comparison of hq curves of original and shuffled streamflow series indicates that the multifractality properties of the streamflow series are mainly due to correlation characteristics within the hydrological series of the Yangtze River basin. (3) Different scaling properties for Cuntan, Yichang, Hankou and Datong are probably due to the different topographical characteristics in the upper, the middle and the lower parts of the Yangtze River basin. Different rainfall–landform–runoff relationships in the upper, middle and lower Yangtze River basin may be the major factors inducing different scaling properties of the hydrological processes of the Yangtze River basin. This research is helpful for a better understanding of the scaling properties of hydrological processes in the Yangtze River basin and for regional flood frequency analysis. ACKNOWLEDGEMENTS The work described in this paper was supported financially by National Natural Science Foundation of China (Grant No.: 40701015 and 30570426), the Innovative Project of Nanjing Institute of Geography and Limnology, CAS (CXNIGLAS200814), the Outstanding Overseas Chinese Scholars Fund from CAS (The Chinese Academy of Sciences) and by the Fok Ying Tung Education Foundation of China (Grant No.: 101004). Cordial thanks are extended to two anonymous reviewers and the editor-in-chief, Professor Malcolm G Anderson, for their valuable comments and suggestions, which greatly improved the quality of this paper. REFERENCES Bunde EK, Kantelhardt JW, Braun P, Bunde A, Havlin S. 2006. Longterm persistence and multifractality of river runoff records: Detrended fluctuation studies. Journal of Hydrology 322: 120–137. Changjiang Water Resources Commission (Ministry of Water Resources, China) (CWRC). 2000. Hydrological records of the Yangtze River. Beijing: Cyclopaedia Press of China (in Chinese). Chianca CV, Ticona A, Penna TJP. 2005. Fourier-detrended fluctuation analysis. Physica A 357: 447–454. Gemmer M, Jiang T, Su BD, Kundzewicz ZW. 2007. Seasonal precipitation changes in the wet season and their influence on flood/drought hazards in the Yangtze River basin, China. Quaternary International 186(1): 12–21. DOI: 10Ð1016/j.quaint. 2007Ð10Ð001. Gupta VK, Castro SL, Over TM. 1996. On scaling exponents of spatial peak flows from rainfall and river network geometry. Journal of Hydrology 187: 81–104. Gupta VK, Messa OJ, Dawdy DR. 1994. Multi scaling theory of flood peaks: regional quantile analysis. Water Resources Research 30: 3405– 3421. Hu K, Ivanov PC, Chen Z, Carpena P, Stanley HE. 2001. Effect of trends on detrended fluctuation analysis. Physical Review E 64: 011114. Hurst HE. 1951. Long-term storage capacity of reservoirs. Transactions of the American Society of Civil Engineering 116: 770– 808. Hurst HE, Black RP, Simaika YM. 1965. Long-term Storage: An Experimental Study. Constable: London. Kantelhardt JW, Zschiegner SA, Koscielny-Bunde E, Bunde A, Havlin S, Stanley HE. 2002. Multifractal detrended fluctuation analysis of nonstationary time series. Physica A 316: 87–114. Hydrol. Process. 22, 4997– 5003 (2008) DOI: 10.1002/hyp MULTIFRACTAL DETRENDED FLUCTUATION ANALYSIS OF STREAMFLOW SERIES Lin GX, Fu ZT. 2008. A universal model to characterize different multifractal behaviours of daily temperature records over China. Physica A 387: 573–579. Lovejoy S, Mandelbrot B. 1985. Fractal properties of rain and a fractal model. Tellus 37A: 209–232. Maraun D, Rust HW, Timmer J. 2004. Tempting long-memory-on the interpretation of DFA results. Nonlinear Processes in Geophysics 11: 495– 503. Menzel L, Bürger G. 2002. Climate change scenarios and runoff response in the Mulde catchment (Southern Elbe, Germany). Journal of Hydrology 267: 53–64. Movahed MS, Jafari GR, Ghasemi F, Rahvar S, Tabar MRR. 2006. Multifractal detrended fluctuation analysis of sunspot time series, Journal of Statistical Mechanics; P02003. Movahed MS, Hermanis E. 2007. Fractal analysis of river flow fluctuations. Physica A 387: 915– 932. DOI:10Ð1016/j.physa.2007Ð10Ð007. Olsson J. 1996. Validity and applicability of a scale-independent multifractal relationship for rainfall. Atmospheric Research 42: 53– 65. Pandey G, Lovejoy S, Schertzer D. 1998. Multifractal analysis of daily river flows including extremes for basins of five to two million square kilometers, one day to 75 years. Journal of Hydrology 208: 62– 81. Pelletier JD, Turcotte DL. 1997. Long-range persistence in climatological and hydrological time series: analysis, modelling and application to drought hazard assessment. Journal of Hydrology 203: 198– 208. Peng C-K, Buldyrev SV, Havlin S, Simons M, Stanley HE, Goldberger AL. 1994. Mosaic organization of DNA nucleotides. Physical Revue E 49: 1685– 1689. Tessier Y, Lovejoy S, Hubert P, Schertzer D, Pecknold S. 1996. Multifractal analysis and modelling of rainfall and river flows and Copyright 2008 John Wiley & Sons, Ltd. 5003 scaling, causal transfer functions. Journal of Geophysical Research 101(D21): 26427– 26440. Xu C-Y, Singh VP. 2005. Evaluation of three complementary relationship evapotranspiration models by water balance approach to estimate actual regional evapotranspiration in different climatic regions. Journal of Hydrology 308: 105– 121. Xu C-Y, Gong L-B, Jiang T, Chen DL. 2006. Analysis of spatial distribution and temporal trend of reference evapotranspiration and pan evaporation in Changjiang (Yangtze River) catchment. Journal of Hydrology 327: 81–93. Yu ZG, Anh V, Wang B. 2001. Correlation property of length sequences based on global structure of complete genome. Physical Revue E 63: 011903. Yu ZG, Anh V, Lau KS. 2004. Fractal analysis of measure representation of large proteins based on the detailed HP model. Physica A 337: 171– 184. Yu ZG, Anh VV, Lau KS, Zhou LQ. 2006. Fractal and multifractal analysis of hydrophobic free energies and solvent accessibilities in proteins. Physical Revue E 63: 031920. Zhang Q, Jiang T, Gemmer M, Becker S. 2005. Precipitation, temperature and discharge analysis from 1951 to 2002 in the Yangtze Catchment, China. Hydrological Sciences Journal 50(1): 65–80. Zhang Q, Liu C-L, Xu C-Y, Xu Y-P, Jiang T. 2006. Observed trends of annual maximum water level and streamflow during past 130 years in the Yangtze River basin, China. Journal of Hydrology 324: 255–265. Zhang Q, Xu C-Y, Zhang ZX, Chen YQ, Liu CL. 2007. Spatial and temporal variability of extreme precipitation during 1960– 2005 in the Yangtze River basin and possible association with largescale circulation. Journal of Hydrology 353: 215–227. DOI: 101016/j.jhydrol.2007-11-023. Hydrol. Process. 22, 4997– 5003 (2008) DOI: 10.1002/hyp