The R/S method has the advantage of determining whether there is a long-term correlation in a time series (Hurst 1951). An existing long-term correlation provides information on future trends in the series. If a time series is detected as persistent behavior, it indicates the future trend of the series will tend to follow that of the past, whereas anti-persistent behavior indicates the opposite trend. If a series is detected as pure random behavior, it is difficult to identify its future trends since there is little correlation between past and future trends (Karakasidis & Liakopoulos 2004). Results of the R/S are presented in Figure 6 and Table 5. The calculated Hurst exponents indicate that the Fenshi, Pingshi and Lishi series exhibited preferably correlated behaviors. Thus, future trends of the annual peak flows in the Fenshi, Pingshi and Lishi stations may follow those of the past. Note that the H value of the Shijiao station data is 0.503, which is very close to 0.5, demonstrating a nearly pure random walk characteristic of the series. Thus, it is unsure whether the future trend of the series will follow that of the past or not. On the other hand, by comparing the H values of the four stations, it can be concluded that correlations between past and future trends gradually weaken from the Fenshi to the Shijiao station.
Table 5

Hurst exponents of the annual peak flows series at the four stations

StationFenshiPingshiLishiShijiao
Hurst exponent 0.687 0.640 0.582 0.503
StationFenshiPingshiLishiShijiao
Hurst exponent 0.687 0.640 0.582 0.503
Figure 6

Results from the R/S analysis of the annual peak flows series of the (a) Fenshi, (b) Pingshi, (c) Lishi and (d) Shijiao stations in BRB. The slope of the straight line obtained by the least square method denotes the H exponent value. R2 is the coefficient of autocorrelation of the time series.

Figure 6

Results from the R/S analysis of the annual peak flows series of the (a) Fenshi, (b) Pingshi, (c) Lishi and (d) Shijiao stations in BRB. The slope of the straight line obtained by the least square method denotes the H exponent value. R2 is the coefficient of autocorrelation of the time series.

Close modal
Close Modal