Scholarlink

<p>The accuracy of cross-time-scale runoff prediction is affected by data characteristics, and accuracy improvement is challenging. This study examined 18,250 global hydrological stations, identified the multi-scale effect of runoff time series (MSER), and proposed an MSER-based improved prediction method (MSEIP). It introduced models, such as multiple linear regression (MLR) and Gaussian process regression (GPR), and evaluation metrics, including optimization proportion (OP) and optimization efficiency (OE), for comparative analysis. The results showed that MSER is applicable to over 73% of hydrological stations, and its applicability increases with larger flow rates. The improvement effect of MSEIP is negatively correlated with time scales (weekly to yearly scale, OPMAE: 0.99–0.60) and positively correlated with flow rates (from less than 100 to more than 2000 m<sup>3</sup>/s, OPQR: 0.6–0.85). MLR is suitable for identifying MSER at small scales (OPMAE of 1 at the weekly scale), while GPR performs better at large scales (seasonally and yearly scales, GPR’s OPQR is 0.67 and 0.58, respectively, higher than MLR’s 0.29 and 0.21). MSER explains differences in runoff prediction accuracy across time scales from data characteristics, and MSEIP provides technical support and a reference for improving cross-scale prediction accuracy.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Multi-scale effects of runoff time series and its improved prediction methods

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2025年8月29日

查看全文
  • Zhongzheng He,
  • Jiahao Lu,
  • Yongqiang Wang,
  • Taixin Zhang,
  • Chao Wang,
  • Jun Guo,
  • Chen Ji,
  • Hui Qin

摘要

The accuracy of cross-time-scale runoff prediction is affected by data characteristics, and accuracy improvement is challenging. This study examined 18,250 global hydrological stations, identified the multi-scale effect of runoff time series (MSER), and proposed an MSER-based improved prediction method (MSEIP). It introduced models, such as multiple linear regression (MLR) and Gaussian process regression (GPR), and evaluation metrics, including optimization proportion (OP) and optimization efficiency (OE), for comparative analysis. The results showed that MSER is applicable to over 73% of hydrological stations, and its applicability increases with larger flow rates. The improvement effect of MSEIP is negatively correlated with time scales (weekly to yearly scale, OPMAE: 0.99–0.60) and positively correlated with flow rates (from less than 100 to more than 2000 m3/s, OPQR: 0.6–0.85). MLR is suitable for identifying MSER at small scales (OPMAE of 1 at the weekly scale), while GPR performs better at large scales (seasonally and yearly scales, GPR’s OPQR is 0.67 and 0.58, respectively, higher than MLR’s 0.29 and 0.21). MSER explains differences in runoff prediction accuracy across time scales from data characteristics, and MSEIP provides technical support and a reference for improving cross-scale prediction accuracy.