Scholarlink

Purpose <p>The uncertainty in suspended sediment load (SSL) observations is an important limitation in sediment research. Errors stemming from various methodological factors (i.e., sampling scheme and frequency, discharge (Q) and suspended sediment concentration (SSC) measurements, load calculation method, measuring period length, and interannual aggregation method) have been estimated before, but their integrated effects on SSL calculations remain mostly unquantified.</p> Methods <p>Using daily Q and SSC time series from 168 stations (7 km<sup>2</sup> – 475,400 km<sup>2</sup>; 2,112 catchment-years), we (i) simulated SSL loads under 22,176 unique combinations of methodological factors through a Monte Carlo approach and computed their errors relative to reference loads, and (ii) developed the SedUCE (<b>Sed</b>iment <b>U</b>n<b>C</b>ertainty <b>E</b>stimation) machine learning model to predict the expected errors on SSL observations.</p> Results <p>On average, annual simulated loads deviated from reference loads by up to a factor of 3.1 (underestimation) and a factor of 2.0 (overestimation) within a 95% confidence interval. Total SSL uncertainty was mostly influenced by the sampling frequency, whereas the expected bias was primarily affected by the load calculation method and sampling scheme. Interactions between methodological factors could also counteract or exacerbate the individual factors’ influences. SedUCE showed good overall performance in predicting SSL errors (R<sup>2</sup> = 0.71, MAE = 0.17, RMSE = 0.26), using only the methodological factors and catchment area as training variables.</p> Conclusion <p>SedUCE is a low-input, freely available tool for obtaining realistic uncertainty estimates on SSL observations. Quantifying uncertainty can help assess the quality and reliability of sediment observations and make more informed decisions during the collection, analysis, and interpretation of sediment data.</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

A wide-scope approach to quantifying and predicting total uncertainty in (multi)annual suspended sediment load observations

  • 期刊论文

  • 发表日期：

    2026年3月28日

查看全文
  • Florence Tan,
  • Pasquale Borrelli,
  • Hugo de Oliveira Fagundes,
  • Matthias Vanmaercke

摘要

Purpose

The uncertainty in suspended sediment load (SSL) observations is an important limitation in sediment research. Errors stemming from various methodological factors (i.e., sampling scheme and frequency, discharge (Q) and suspended sediment concentration (SSC) measurements, load calculation method, measuring period length, and interannual aggregation method) have been estimated before, but their integrated effects on SSL calculations remain mostly unquantified.

Methods

Using daily Q and SSC time series from 168 stations (7 km2 – 475,400 km2; 2,112 catchment-years), we (i) simulated SSL loads under 22,176 unique combinations of methodological factors through a Monte Carlo approach and computed their errors relative to reference loads, and (ii) developed the SedUCE (Sediment UnCertainty Estimation) machine learning model to predict the expected errors on SSL observations.

Results

On average, annual simulated loads deviated from reference loads by up to a factor of 3.1 (underestimation) and a factor of 2.0 (overestimation) within a 95% confidence interval. Total SSL uncertainty was mostly influenced by the sampling frequency, whereas the expected bias was primarily affected by the load calculation method and sampling scheme. Interactions between methodological factors could also counteract or exacerbate the individual factors’ influences. SedUCE showed good overall performance in predicting SSL errors (R2 = 0.71, MAE = 0.17, RMSE = 0.26), using only the methodological factors and catchment area as training variables.

Conclusion

SedUCE is a low-input, freely available tool for obtaining realistic uncertainty estimates on SSL observations. Quantifying uncertainty can help assess the quality and reliability of sediment observations and make more informed decisions during the collection, analysis, and interpretation of sediment data.