基于Generalized LASSO模型的长江寸滩站年均径流突变研究

为了区分水文时间序列的趋势和跳跃变异,将基于L_1范数正则化技术的generalized LASSO模型应用于水文序列变异识别。经过识别,发现长江寸滩水文站年平均流量序列在1969年发生了向下的均值跃变。此外,趋势分析表明无论是跃变前后的子序列还是剔除跳跃成分的整个序列,均未检出显著的趋势,这说明对寸滩水文站年平均流量序列的跳跃变异假设是合理的。基于generalized LASSO模型的结果与其他突变检测方法结果进行比较,所得结论是一致的。

Abrupt change of annual mean streamflow at Cuntan Station on Yangtze River based on generalized LASSO

In order to distinguish trends and abrupt changes in hydrological time series, the generalized LASSO model based on L₁-norm regularization was used to detect changes in hydrological series. It is found that the annual mean streamflow series at the Cuntan Station shows a downward step change in 1969. Furthermore, trend analysis shows that both the sub-series before/after 1969 and the whole series after removing step change components show no significant trend, indicating that the assumption of the step change in the annual mean streamflow series at the Cuntan Station is reasonable. The conclusions obtained from the generalized LASSO method are consistent with those obtained from other change detection methods.