应对水文序列非一致性变化影响的溯源重构法研究

水文序列的非一致性特征导致多种常用水文分析计算方法在理论上陷入瓶颈,因尚未得到有效的解决,仍需不断探索研究。从因变量与自变量关系的角度看,引起水文变量非一致性变化的原因是影响因子自身的非一致性变化。据此,针对单个影响因子作用于水文变量的规律,提出源函数概念,依据水文变量与影响因子源函数间的统计乘积模型,构建新的函数,并据此函数将非一致性变化的影响因子的作用从原序列中剔除,实现了非一致性序列向新的一致性序列的转化。本文称该方法为溯源重构法。获得的重构序列可用于以一致性序列为基础的分析计算工作。对陕西省佳芦河流域非一致性年最大洪峰流量序列的应用研究表明,溯源重构法不仅能有效消除序列的线性、非线性趋势以及突变等三种非一致性特征,且重构序列的分布更符合原序列的抽样属性。相比于分解合成法仅能去除序列的一阶矩趋势,溯源重构法在去除一、二阶矩的非一致性上均能发挥作用。

Study on mechanism-based reconstruction method to cope with the nonstationary change of hydrological series

The nonstationarity of hydrological series lead to the theoretical bottleneck of many commonly used hydrological analysis and calculation methods, which have not been effectively solved and still needs to be explored and studied continuously. From the point of view of the relationship between depend variable and independent variables, the reason for the nonstationary change of hydrological variable is the nonstationary change of its influencing factors. On this basis,the concept of mechanism function of single influence factor acting on hydrological variable is put forward from the perspective of physical mechanism. According to the statistical multiplication model established between hydrological variable and influence factor mechanism functions, a new function is constructed, and based on this new function, the effect of nonstationary influence factors on hydrological variable is removed from the original series. The reconstruction of hydrological series realizes the transformation from the nonstationary series to a new stationary reconstruction series, so this method is called the Mechanism-based Reconstruction Method (MBRM). The reconstructed series can be used in all the stationary series-based analysis and calculation work. The application research of the nonstationary maximum annual flood peak discharge series in the Jialuhe River basin shows that the MBRM can not only removes the linear, nonlinear trend and abrupt change of hydrological series, but also the distribution of the reconstructed series is more consistent with the sampling characteristics of the original series. The decomposition and synthesis method can only remove the trend in the first moment of the series,while the MBRM can play a role in removing the nonstationarity in the first and second moments.