Scholarlink

<p>Accurate estimation of hourly precipitation is critical for hydrological forecasting and hydroelectric operations. This study develops a multivariate geostatistical framework that integrates irregularly positioned meteorological station measurements with the regularly gridded ERA5-Land reanalysis data for southern Quebec, Canada. The workflow addresses the zero-inflation challenge by separating precipitation occurrences from intensity. The occurrences are modeled using sequential indicator simulation (SIS), while the precipitation intensity is simulated conditionally through multivariate turning bands simulation (TBS) with ERA5-Land as an auxiliary collocated variable. The method generates 100 scenarios per hour, allowing uncertainty quantification via probability and quantile maps. Cross-validation results show coefficients of 0.65 to 0.82 for the model of occurrences and mean absolute errors of 0.8 to 1.5 mm for hourly intensities. Integrating ERA5-Land reduces kriging variance by 25% to 40% relative to univariate approaches, particularly where station coverage is sparse. The 9&#xa0;km resolution probabilistic outputs offer operational value for regional hydrological modeling and hydropower risk management.</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

Probabilistic Modeling of Hourly Precipitation Jointly with ERA5-Land: Application of Geostatistical Simulations

  • 期刊论文

  • 发表日期：

    2026年3月19日

查看全文
  • Dominique Tapsoba,
  • Raphael Rousseau-Rizzi,
  • Pedram Masoudi,
  • Ophélie Lemarchand

摘要

Accurate estimation of hourly precipitation is critical for hydrological forecasting and hydroelectric operations. This study develops a multivariate geostatistical framework that integrates irregularly positioned meteorological station measurements with the regularly gridded ERA5-Land reanalysis data for southern Quebec, Canada. The workflow addresses the zero-inflation challenge by separating precipitation occurrences from intensity. The occurrences are modeled using sequential indicator simulation (SIS), while the precipitation intensity is simulated conditionally through multivariate turning bands simulation (TBS) with ERA5-Land as an auxiliary collocated variable. The method generates 100 scenarios per hour, allowing uncertainty quantification via probability and quantile maps. Cross-validation results show coefficients of 0.65 to 0.82 for the model of occurrences and mean absolute errors of 0.8 to 1.5 mm for hourly intensities. Integrating ERA5-Land reduces kriging variance by 25% to 40% relative to univariate approaches, particularly where station coverage is sparse. The 9 km resolution probabilistic outputs offer operational value for regional hydrological modeling and hydropower risk management.