根据三水、马口水文站1959-2004水文年实测流量资料的月均序列,应用经验模式分解(EMD)方法分析三水站分流比水文年序列的周期变化及趋势变化,应用启发式分割算法(BG)识别变异点,并采用1K/S分析方法探讨分流比变化的持续性特征,结果表明:近46年来,三水站分流比序列具有3.2a、6.6a、15.3a、30.7a的波动周期,表现出复杂的多时间尺度性;三水站分流比的变化趋势是显著增加,增加速率为0.0022/a,并在1992年发生了突变,且为均值向上跳跃;该序列具有非常强的状态持续性,存在长期记忆性.未来仍将呈显著的增加趋势。
Based on the streamflow data at Sanshui and Makou stations in the upper West and North River network from 1959 to 2004, the empirical mode decomposition(EMD), rescaled range analysis(R/S) and heuristic segmentation(BG) methods were used to find out the variation law of annual average streamflow ratio of Sanshui/(Makou + Sanshui). The empirical mode decomposition method was used to obtain intrinsic mode functions (IMFs), by which variations of the original series were analyzed with various time scales. The results of intrinsic mode functions by EMD indicate that there are obvious periodic variations with scales 3.2a/ 6.6a/15.3a/30.Ta for annual average streamflow ratio of Sanshui/(Makou + Sanshui), and scale 3.2a is the first period. The results of res. by EMD show that there are significant change trend and abruptly jumping point in 1992 for annual average diversion ratio of flow discharge at Sanshui hydrological station. The computed Hurst exponents by R/S analysis method indicate that a long-term memory characteristic exists in the annual streamflow ratio of Sanshui/(Makou + Sanshui). The Hurst exponent was found to be greater than 0.5, implying that the future tendency of diversion ratio of flow discharge is consistent with those of the past.