skip to main content

Attention:

The NSF Public Access Repository (NSF-PAR) system and access will be unavailable from 11:00 PM ET on Thursday, May 23 until 2:00 AM ET on Friday, May 24 due to maintenance. We apologize for the inconvenience.


Title: Deep Learning for Subgrid‐Scale Turbulence Modeling in Large‐Eddy Simulations of the Convective Atmospheric Boundary Layer
Abstract

In large‐eddy simulations, subgrid‐scale (SGS) processes are parameterized as a function of filtered grid‐scale variables. First‐order, algebraic SGS models are based on the eddy‐viscosity assumption, which does not always hold for turbulence. Here we apply supervised deep neural networks (DNNs) to learn SGS stresses from a set of neighboring coarse‐grained velocity from direct numerical simulations of the convective boundary layer at friction Reynolds numbersReτup to 1243 without invoking the eddy‐viscosity assumption. The DNN model was found to produce higher correlation between SGS stresses compared to the Smagorinsky model and the Smagorinsky‐Bardina mixed model in the surface and mixed layers and can be applied to different grid resolutions and various stability conditions ranging from near neutral to very unstable. The DNN model can capture key statistics of turbulence ina posteriori(online) tests when applied to large‐eddy simulations of the atmospheric boundary layer.

 
more » « less
Award ID(s):
2019625 2028644
NSF-PAR ID:
10368452
Author(s) / Creator(s):
 ;  ;  ;  ;  ;  ;  ;  ;  ;  ;  
Publisher / Repository:
DOI PREFIX: 10.1029
Date Published:
Journal Name:
Journal of Advances in Modeling Earth Systems
Volume:
14
Issue:
5
ISSN:
1942-2466
Format(s):
Medium: X
Sponsoring Org:
National Science Foundation
More Like this
  1. Abstract

    We present a new ensemble of 36 numerical experiments aimed at comprehensively gauging the sensitivity of nested large-eddy simulations (LES) driven by large-scale dynamics. Specifically, we explore 36 multiscale configurations of the Weather Research and Forecasting (WRF) Model to simulate the boundary layer flow over the complex topography at the Perdigão field site, with five nested domains discretized at horizontal resolutions ranging from 11.25 km to 30 m. Each ensemble member has a unique combination of the following input factors: (i) large-scale initial and boundary conditions, (ii) subgrid turbulence modeling in thegray zoneof turbulence, (iii) subgrid-scale (SGS) models in LES, and (iv) topography and land-cover datasets. We probe their relative importance for LES calculations of velocity, temperature, and moisture fields. Variance decomposition analysis unravels large sensitivities to topography and land-use datasets and very weak sensitivity to the LES SGS model. Discrepancies within ensemble members can be as large as 2.5 m s−1for the time-averaged near-surface wind speed on the ridge and as large as 10 m s−1without time averaging. At specific time points, a large fraction of this sensitivity can be explained by the different turbulence models in the gray zone domains. We implement a horizontal momentum and moisture budget routine in WRF to further elucidate the mechanisms behind the observed sensitivity, paving the way for an increased understanding of the tangible effects of the gray zone of turbulence problem.

    Significance Statement

    Several science and engineering applications, including wind turbine siting and operations, weather prediction, and downscaling of climate projections, call for high-resolution numerical simulations of the lowest part of the atmosphere. Recent studies have highlighted that such high-resolution simulations, coupled with large-scale models, are challenging and require several important assumptions. With a new set of numerical experiments, we evaluate and compare the significance of different assumptions and outstanding challenges in multiscale modeling (i.e., coupling large-scale models and high-resolution atmospheric simulations). The ultimate goal of this analysis is to put each individual assumption into the wider perspective of a realistic problem and quantify its relative importance compared to other important modeling choices.

     
    more » « less
  2. The surface wind structure and vertical turbulent transport processes in the eyewall of hurricane Isabel (2003) are investigated using six large-eddy simulations (LESs) with different horizontal grid spacing and three-dimensional (3D) sub-grid scale (SGS) turbulent mixing models and a convection permitting simulation that uses a coarser grid spacing and one-dimensional vertical turbulent mixing scheme. The mean radius-height distribution of storm tangential wind and radial flow, vertical velocity structure, and turbulent kinetic energy and momentum fluxes in the boundary layer generated by LESs are consistent with those derived from historical dropsonde composites, Doppler radar, and aircraft measurements. Unlike the convection permitting simulation that produces storm wind fields lacking small-scale disturbances, all LESs are able to produce sub-kilometer and kilometer scale eddy circulations in the eyewall. The inter-LES differences generally reduce with the decrease of model grid spacing. At 100-m horizontal grid spacing, the vertical momentum fluxes induced by the model-resolved eddies and the associated eddy exchange coefficients in the eyewall simulated by the LESs with different 3D SGS mixing schemes are fairly consistent. Although with uncertainties, the decomposition in terms of eddy scales suggests that sub-kilometer eddies are mainly responsible for the vertical turbulent transport within the boundary layer (~1 km depth following the conventional definition) whereas eddies greater than 1 km become the dominant contributors to the vertical momentum transport above the boundary layer in the eyewall. The strong dependence of vertical turbulent transport on eddy scales suggests that the vertical turbulent mixing parameterization in mesoscale simulations of tropical cyclones is ultimately a scale-sensitive problem. 
    more » « less
  3. In a tropical cyclone (TC), turbulence not only exists in the planetary boundary layer (PBL) but also can be generated above the PBL by the cloud processes in the eyewall and rainbands. It is found that the Hurricane Analysis and Forecast System (HAFS), a new multi-scale operational model for TC prediction, fails to capture the intense turbulent mixing in eyewall and rainband clouds due to a poor estimation of static stability in clouds. The problem is fixed by including the effects of multi-phase water in the stability calculation. Simulations of 21 TCs and tropical storms in the North Atlantic basin of 2016–2019 hurricane seasons totaling 118 forecast cycles show that the stability correction substantially improves HAFS's skill in predicting storm track and intensity. Analyses of HAFS's simulations of Hurricane Michael (2018) show that the positive tendency of vortex's tangential wind resulting from the radially inward transport of absolute vorticity dominates the eddy correlation tendencies induced by the model-resolved asymmetric eddies and serves as a main mechanism for the rapid intensification of Michael. The sub-grid scale (SGS) turbulent transport above the PBL in the eyewall plays a pivotal role in initiating a positive feedback among the eyewall convection, mean secondary overturning circulation, vortex acceleration via the inward transport of absolute vorticity, surface evaporation, and radial convergence of moisture in the PBL. Without the SGS transport above the PBL, the model-resolved vertical transport alone may not be sufficient in initiating the positive feedback underlying the rapid intensification of TCs. 
    more » « less
  4. Abstract. The fundamental mechanism underlying tropical cyclone (TC) intensification may be understood from the conservation of absolute angular momentum, where the primary circulation of a TC is driven by the torque acting on air parcels resulting from asymmetric eddy processes, including turbulence. While turbulence is commonly regarded as a flow feature pertaining to the planetary boundary layer (PBL), intense turbulent mixing generated by cloud processes also exists above the PBL in the eyewall and rainbands. Unlike the eddy forcing within the PBL that is negative definite, the sign of eyewall/rainband eddy forcing above the PBL is indefinite and thus provides a possible mechanism to spin up a TC vortex. In this study, we show that the Hurricane Weather Research & forecasting (HWRF) model, one of the operational models used for TC prediction, is unable to generate appropriate sub-grid-scale (SGS) eddy forcing above the PBL due to lack of consideration of intense turbulent mixing generated by the eyewall and rainband clouds. Incorporating an in-cloud turbulent mixing parameterization in the PBL scheme notably improves HWRF's skills on predicting rapid changes in intensity for several past major hurricanes. While the analyses show that the SGS eddy forcing above the PBL is only about one-fifth of the model-resolved eddy forcing, the simulated TC vortex inner-core structure and the associated model-resolved eddy forcing exhibit a substantial dependence on the parameterized SGS eddy processes. The results highlight the importance of eyewall/rainband SGS eddy forcing to numerical prediction of TC intensification, including rapid intensification at the current resolution of operational models.

     
    more » « less
  5. We developed a novel autonomously dynamic nonlocal turbulence model for the large and very large eddy simulation (LES, VLES) of homogeneous isotropic turbulent flows. The model is based on a generalized (integer-to-noninteger)-order Laplacian of the filtered velocity field, and a novel dynamic model has been formulated to avoid the need for tuning the model constant. Three data-driven approaches were introduced for the determination of the fractional-order to have a model that is totally free of any tuning parameter. Our analysis includes both the a priori and the a posteriori tests. In the former test, using a high-fidelity and well-resolved dataset from direct numerical simulations (DNSs), we computed the correlation coefficients for the stress components of the subgrid-scale (SGS) stress tensor and the one we get directly from the DNS results. Moreover, we compared the probability density function of the ensemble-averaged SGS forces for different filter sizes. In the latter, we employed our new model along with other conventional models including the static and dynamic Smagorinsky models into our pseudo-spectral solver and tested the final predicted quantities. The results of the newly developed model exhibit an expressive agreement with the ground-truth DNS results in all components of the SGS stress and forces. Also, the model exhibits promising results in the VLES region as well as the LES region, which could be remarkably important for cost-efficient nonlocal turbulence modeling, e.g., in meteorological and environmental applications.

     
    more » « less