<p>Mesoscale-to-microscale coupling (MMC) extends the capability of large-eddy simulations by introducing spatially and temporally varying mesoscale conditions in order to simultaneously model the large-scale atmospheric dynamics and the small-scale turbulent processes. This coupling is important for a range of applications, including wind energy forecasting, wildfire spread prediction, and urban climate modeling. In this study, we introduce two new offline MMC methods within the class of profile assimilation techniques. The first method, wavelet-based profile assimilation (WPA) belongs to the class of error-based approaches, which applies multi-resolution decomposition using wavelet basis function to compute the internal forcing for the momentum and temperature equations, while the second method is a hybrid approach that combines a physics-based geostrophic balance for momentum source terms with wavelet-based profile assimilation for temperature source terms. Unlike previous profile assimilation methods, such as direct and indirect profile assimilation (DPA, IPA), which apply error-based corrections to both momentum and temperature equations, the proposed hybrid method avoids momentum error forcing, which simplifies the MMC algorithm and implementation. To demonstrate our approach, a composite mesoscale dataset derived from the American WAKE experimeNt (AWAKEN) is used to add mesoscale forcing to large-eddy simulations conducted over flat terrain for a full diurnal cycle. The composite dataset combines experimental observations from scanning lidar measurements spanning heights from 100 to <InlineEquation ID="IEq1"> <EquationSource Format="TEX">$1100~\textrm{m}$</EquationSource> <EquationSource Format="MATHML"><math> <mrow> <mn>1100</mn> <mspace width="3.33333pt" /> <mtext>m</mtext> </mrow> </math></EquationSource> </InlineEquation>, virtual tower data, and mesoscale output from Weather Research and Forecasting (WRF) model simulations. Our results show that the new assimilation strategies yield mean flow profiles and turbulence statistics that achieve substantially better accuracy relative to previous profile assimilation techniques. The proposed framework offers a flexible and physically consistent method for coupling mesoscale information into large-eddy simulations across a range of atmospheric boundary layer conditions.</p>

Development of Profile Assimilation Methods for Data-Driven Large Eddy Simulations

期刊论文
发表日期：

2026年3月10日

查看全文

Arjun Ajay,
Jagdeep Singh,
Sebastiano Stipa,
Philippe Beaucage,
Joshua Brinkerhoff

摘要

Mesoscale-to-microscale coupling (MMC) extends the capability of large-eddy simulations by introducing spatially and temporally varying mesoscale conditions in order to simultaneously model the large-scale atmospheric dynamics and the small-scale turbulent processes. This coupling is important for a range of applications, including wind energy forecasting, wildfire spread prediction, and urban climate modeling. In this study, we introduce two new offline MMC methods within the class of profile assimilation techniques. The first method, wavelet-based profile assimilation (WPA) belongs to the class of error-based approaches, which applies multi-resolution decomposition using wavelet basis function to compute the internal forcing for the momentum and temperature equations, while the second method is a hybrid approach that combines a physics-based geostrophic balance for momentum source terms with wavelet-based profile assimilation for temperature source terms. Unlike previous profile assimilation methods, such as direct and indirect profile assimilation (DPA, IPA), which apply error-based corrections to both momentum and temperature equations, the proposed hybrid method avoids momentum error forcing, which simplifies the MMC algorithm and implementation. To demonstrate our approach, a composite mesoscale dataset derived from the American WAKE experimeNt (AWAKEN) is used to add mesoscale forcing to large-eddy simulations conducted over flat terrain for a full diurnal cycle. The composite dataset combines experimental observations from scanning lidar measurements spanning heights from 100 to $1100~\textrm{m}$ $1100 m$ , virtual tower data, and mesoscale output from Weather Research and Forecasting (WRF) model simulations. Our results show that the new assimilation strategies yield mean flow profiles and turbulence statistics that achieve substantially better accuracy relative to previous profile assimilation techniques. The proposed framework offers a flexible and physically consistent method for coupling mesoscale information into large-eddy simulations across a range of atmospheric boundary layer conditions.