基于分位数回归的滦河流域降水及径流变化特性

目前对降水、径流变化特性的研究多集中于序列均值特征上,从分析结果无法了解降水、径流序列在分布尾端的变化情况,且易受异常值干扰。为解决上述问题,本文采用分位数回归方法分析了滦河流域降水、径流的变化特性。首先采用Pettitt变异点检测和Mann-Kendall趋势分析法对序列在均值特征下进行分析,结果表明各子流域年降水序列基本满足一致性,而年径流序列具有非一致性,呈现出突变变异;年降水、径流序列响应关系不明显。然后利用分位数回归方法对年降水、径流序列在高、低分位数水平下的变化特性进行分析,得出大部分子流域年降水序列在不同分位数水平下基本保持一致性,而年径流序列在高、低分位数水平下均表现出减小趋势,且高分位数水平下的减小速率较低分位数大;在高或低分位数水平下,部分子流域年降水、径流序列之间具有一定的响应关系;但对整个流域而言,径流序列主要受人类活动影响,年降水、径流之间的响应关系不明显。该结论可为降水、径流之间变化特性的研究提供新参考。

Temporal variation characteristics of annual precipitation and runoff in Luan River basin based on quantile regression

Most current studies on the temporal variation characteristics of precipitation and runoff series focus on the mean features, and consequently the results provide little insight into the tail parts of distributions and are affected by anomalies easily. To solve this problem, we have applied a quantile regression method to the precipitation and runoff series of the Luan River basin. Pettitt test and Mann-Kendall test were used to analyze the temporal variations in terms of mean features. The results show that the annual precipitation series are essentially stationary while the annual runoff series manifest abrupt changes and hence are non-stationary. Therefore, no response relationship can be found between these two series. Then, we applied a quantile regression method to analyze the same series of precipitation and runoff using high and low levels of quantiles. It demonstrates that at different quantiles, the precipitation series in most sub-basins are stationary while the runoff series in all the sub-basins have a significant descending trend with a faster rate at high quantiles than those at low ones. And in several sub-basins, a good response relationship exists between precipitation and runoff at different quantiles, while in the whole basin, the relationship is not obvious due to the impact of human activities on runoff. This conclusion is useful to peer studies on temporal variations of precipitation and runoff.