Articles | Volume 14, issue 6
https://doi.org/10.5194/gmd-14-3769-2021
© Author(s) 2021. This work is distributed under
the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
https://doi.org/10.5194/gmd-14-3769-2021
© Author(s) 2021. This work is distributed under
the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Development of a large-eddy simulation subgrid model based on artificial neural networks: a case study of turbulent channel flow
Robin Stoffer
CORRESPONDING AUTHOR
Meteorology and Air Quality Group, Wageningen University, Wageningen, the Netherlands
Caspar M. van Leeuwen
SURFsara, Amsterdam, the Netherlands
Damian Podareanu
SURFsara, Amsterdam, the Netherlands
Valeriu Codreanu
SURFsara, Amsterdam, the Netherlands
Menno A. Veerman
Meteorology and Air Quality Group, Wageningen University, Wageningen, the Netherlands
Martin Janssens
Meteorology and Air Quality Group, Wageningen University, Wageningen, the Netherlands
Oscar K. Hartogensis
Meteorology and Air Quality Group, Wageningen University, Wageningen, the Netherlands
Chiel C. van Heerwaarden
Meteorology and Air Quality Group, Wageningen University, Wageningen, the Netherlands
Related authors
Luuk D. van der Valk, Adriaan J. Teuling, Luc Girod, Norbert Pirk, Robin Stoffer, and Chiel C. van Heerwaarden
The Cryosphere, 16, 4319–4341, https://doi.org/10.5194/tc-16-4319-2022, https://doi.org/10.5194/tc-16-4319-2022, 2022
Short summary
Short summary
Most large-scale hydrological and climate models struggle to capture the spatially highly variable wind-driven melt of patchy snow cover. In the field, we find that 60 %–80 % of the total melt is wind driven at the upwind edge of a snow patch, while it does not contribute at the downwind edge. Our idealized simulations show that the variation is due to a patch-size-independent air-temperature reduction over snow patches and also allow us to study the role of wind-driven snowmelt on larger scales.
Francisca Aguirre-Correa, Oscar Hartogensis, Pedro Bonacic-Vera, and Francisco Suárez
EGUsphere, https://doi.org/10.5194/egusphere-2025-2984, https://doi.org/10.5194/egusphere-2025-2984, 2025
This preprint is open for discussion and under review for Hydrology and Earth System Sciences (HESS).
Short summary
Short summary
We studied an arid, endorheic basin in the Chilean Altiplano to understand how rainfall and evaporation affect groundwater and water availability. Using a rainfall-runoff model and 40 years of satellite data, we found that much of the water evaporates and less reaches the aquifers than expected. Our results challenge the idea that the basin is fully closed and suggest that current water budget estimates may need revision, an urgent task under a changing climate.
Job I. Wiltink, Hartwig Deneke, Chiel C. van Heerwaarden, and Jan Fokke Meirink
Atmos. Meas. Tech., 18, 3917–3936, https://doi.org/10.5194/amt-18-3917-2025, https://doi.org/10.5194/amt-18-3917-2025, 2025
Short summary
Short summary
Global horizontal irradiance retrievals from satellite observations are affected by spatial displacements due to parallax and cloud shadows. We assess different approaches to correct for these displacements and quantify their added value by comparison with a network of ground-based pyranometer observations. The corrections are found to become increasingly important at higher spatial resolutions and are most relevant for variable cloud types.
Mary Rose Mangan, Jordi Vilà-Guerau de Arellano, Bart J. H. van Stratum, Marie Lothon, Guylaine Canut-Rocafort, and Oscar K. Hartogensis
Atmos. Chem. Phys., 25, 8959–8981, https://doi.org/10.5194/acp-25-8959-2025, https://doi.org/10.5194/acp-25-8959-2025, 2025
Short summary
Short summary
Using observations and high-resolution turbulence modeling, we examine the influence of irrigation-driven surface heterogeneity on the atmospheric boundary layer (ABL). We use a multi-scale approach for characterizing surface heterogeneity to explore how its influence on the ABL within a grid cell would change with higher-resolution models. We find that the height of the ABL is variable across short distances and that the surface heterogeneity is felt least strongly in the middle of the ABL.
Marc Castellnou Ribau, Mercedes Bachfischer, Marta Miralles Bover, Borja Ruiz, Laia Estivill, Jordi Pages, Pau Guarque, Brian Verhoeven, Zisoula Ntasiou, Ove Stokkeland, Chiel Van Herwaeeden, Tristan Roelofs, Martin Janssens, Cathelijne Stoof, and Jordi Vilà-Guerau de Arellano
EGUsphere, https://doi.org/10.5194/egusphere-2025-1923, https://doi.org/10.5194/egusphere-2025-1923, 2025
This preprint is open for discussion and under review for Atmospheric Measurement Techniques (AMT).
Short summary
Short summary
Firefighter entrapments can occur when wildfires escalate suddenly due to fire-atmosphere interactions. This study presents a method to analyze this in real-time using two weather balloon measurements: ambient and in-plume conditions. Researchers launched 156 balloons during wildfire seasons in Spain, Chile, Greece, and the Netherlands. This methodology detects sudden changes in fire behavior by comparing ambient and in-plume data, ultimately enhancing research on fire-atmosphere interactions.
Luuk D. van der Valk, Oscar K. Hartogensis, Miriam Coenders-Gerrits, Rolf W. Hut, and Remko Uijlenhoet
EGUsphere, https://doi.org/10.5194/egusphere-2025-1128, https://doi.org/10.5194/egusphere-2025-1128, 2025
Short summary
Short summary
Commercial microwave links (CMLs), part of mobile phone networks, transmit comparable signals as instruments specially designed to estimate evaporation. Therefore, we investigate if CMLs could be used to estimate evaporation, even though they have not been designed for this purpose. Our results illustrate the potential of using CMLs to estimate evaporation, especially given their global coverage, but also outline some major drawbacks, often a consequence of unfavourable design choices for CMLs.
Wouter Mol and Chiel van Heerwaarden
Atmos. Chem. Phys., 25, 4419–4441, https://doi.org/10.5194/acp-25-4419-2025, https://doi.org/10.5194/acp-25-4419-2025, 2025
Short summary
Short summary
Sunlight varies often and quickly under broken cloud cover, and every cloud field creates a unique pattern of sunlight on the surface below. These variations affect many processes in the Earth system, from photosynthesis and chemistry to cloud formation itself. The exact way in which cloud particles interact with sunlight is complex and expensive to calculate. We demonstrate a simplified framework which explains how sunlight changes for potentially any cloud field.
Robbert Petrus Johannes Moonen, Getachew Agmuas Adnew, Jordi Vilà-Guerau de Arellano, Oscar Karel Hartogensis, David Joan Bonell Fontas, Shujiro Komiya, Sam P. Jones, and Thomas Röckmann
EGUsphere, https://doi.org/10.5194/egusphere-2025-452, https://doi.org/10.5194/egusphere-2025-452, 2025
Short summary
Short summary
Understory ejections are distinct turbulent features emerging in prime tall forest ecosystems. We share a method to isolate understory ejections based on H2O-CO2 anomalie quadrants. From these, we calculate the flux contributions of understory ejections and all flux quadrants. In addition we show that a distinctly depleted isotopic composition can be found in the ejected water vapour. Finally, we explored the role of clouds as a potential trigger for understory ejections.
Hans Segura, Xabier Pedruzo-Bagazgoitia, Philipp Weiss, Sebastian K. Müller, Thomas Rackow, Junhong Lee, Edgar Dolores-Tesillos, Imme Benedict, Matthias Aengenheyster, Razvan Aguridan, Gabriele Arduini, Alexander J. Baker, Jiawei Bao, Swantje Bastin, Eulàlia Baulenas, Tobias Becker, Sebastian Beyer, Hendryk Bockelmann, Nils Brüggemann, Lukas Brunner, Suvarchal K. Cheedela, Sushant Das, Jasper Denissen, Ian Dragaud, Piotr Dziekan, Madeleine Ekblom, Jan Frederik Engels, Monika Esch, Richard Forbes, Claudia Frauen, Lilli Freischem, Diego García-Maroto, Philipp Geier, Paul Gierz, Álvaro González-Cervera, Katherine Grayson, Matthew Griffith, Oliver Gutjahr, Helmuth Haak, Ioan Hadade, Kerstin Haslehner, Shabeh ul Hasson, Jan Hegewald, Lukas Kluft, Aleksei Koldunov, Nikolay Koldunov, Tobias Kölling, Shunya Koseki, Sergey Kosukhin, Josh Kousal, Peter Kuma, Arjun U. Kumar, Rumeng Li, Nicolas Maury, Maximilian Meindl, Sebastian Milinski, Kristian Mogensen, Bimochan Niraula, Jakub Nowak, Divya Sri Praturi, Ulrike Proske, Dian Putrasahan, René Redler, David Santuy, Domokos Sármány, Reiner Schnur, Patrick Scholz, Dmitry Sidorenko, Dorian Spät, Birgit Sützl, Daisuke Takasuka, Adrian Tompkins, Alejandro Uribe, Mirco Valentini, Menno Veerman, Aiko Voigt, Sarah Warnau, Fabian Wachsmann, Marta Wacławczyk, Nils Wedi, Karl-Hermann Wieners, Jonathan Wille, Marius Winkler, Yuting Wu, Florian Ziemen, Janos Zimmermann, Frida A.-M. Bender, Dragana Bojovic, Sandrine Bony, Simona Bordoni, Patrice Brehmer, Marcus Dengler, Emanuel Dutra, Saliou Faye, Erich Fischer, Chiel van Heerwaarden, Cathy Hohenegger, Heikki Järvinen, Markus Jochum, Thomas Jung, Johann H. Jungclaus, Noel S. Keenlyside, Daniel Klocke, Heike Konow, Martina Klose, Szymon Malinowski, Olivia Martius, Thorsten Mauritsen, Juan Pedro Mellado, Theresa Mieslinger, Elsa Mohino, Hanna Pawłowska, Karsten Peters-von Gehlen, Abdoulaye Sarré, Pajam Sobhani, Philip Stier, Lauri Tuppi, Pier Luigi Vidale, Irina Sandu, and Bjorn Stevens
EGUsphere, https://doi.org/10.5194/egusphere-2025-509, https://doi.org/10.5194/egusphere-2025-509, 2025
Short summary
Short summary
The nextGEMS project developed two Earth system models that resolve processes of the order of 10 km, giving more fidelity to the representation of local phenomena, globally. In its fourth cycle, nextGEMS performed simulations with coupled ocean, land, and atmosphere over the 2020–2049 period under the SSP3-7.0 scenario. Here, we provide an overview of nextGEMS, insights into the model development, and the realism of multi-decadal, kilometer-scale simulations.
Jolanda J. E. Theeuwen, Sarah N. Warnau, Imme B. Benedict, Stefan C. Dekker, Hubertus V. M. Hamelers, Chiel C. van Heerwaarden, and Arie Staal
EGUsphere, https://doi.org/10.5194/egusphere-2025-289, https://doi.org/10.5194/egusphere-2025-289, 2025
Short summary
Short summary
The Mediterranean Basin is prone to drying. This study uses a simple model to explore how forests affect the potential for rainfall by analyzing the lowest part of the atmosphere. Results show that forestation amplifies drying in dry areas and boosts rainfall potential in wet regions, where it also promotes cooling. These findings suggest that the impact of forestation varies with soil moisture, and may possibly mitigate or intensify future drying.
Luuk D. van der Valk, Oscar K. Hartogensis, Miriam Coenders-Gerrits, Rolf W. Hut, Bas Walraven, and Remko Uijlenhoet
EGUsphere, https://doi.org/10.5194/egusphere-2024-2974, https://doi.org/10.5194/egusphere-2024-2974, 2025
Short summary
Short summary
Commercial microwave links (CMLs), part of mobile phone networks, transmit comparable signals as instruments specially designed to estimate evaporation. Therefore, we investigate if CMLs could be used to estimate evaporation, even though they have not been designed for this purpose. Our results illustrate the potential of using CMLs to estimate evaporation, especially given their global coverage, but also outline some major drawbacks, often a consequence of unfavourable design choices for CMLs.
Job I. Wiltink, Hartwig Deneke, Yves-Marie Saint-Drenan, Chiel C. van Heerwaarden, and Jan Fokke Meirink
Atmos. Meas. Tech., 17, 6003–6024, https://doi.org/10.5194/amt-17-6003-2024, https://doi.org/10.5194/amt-17-6003-2024, 2024
Short summary
Short summary
Meteosat Spinning Enhanced Visible and Infrared Imager (SEVIRI) global horizontal irradiance (GHI) retrievals are validated at standard and increased spatial resolution against a network of 99 pyranometers. GHI accuracy is strongly dependent on the cloud regime. Days with variable cloud conditions show significant accuracy improvements when retrieved at higher resolution. We highlight the benefits of dense network observations and a cloud-regime-resolved approach in validating GHI retrievals.
Mirjam Tijhuis, Bart J. H. van Stratum, and Chiel C. van Heerwaarden
Atmos. Chem. Phys., 24, 10567–10582, https://doi.org/10.5194/acp-24-10567-2024, https://doi.org/10.5194/acp-24-10567-2024, 2024
Short summary
Short summary
Radiative transfer in the atmosphere is a 3D processes, which is often modelled in 1D for computational efficiency. We studied the differences between using 1D and 3D radiative transfer. With 3D radiation, larger clouds that contain more liquid water develop. However, they cover roughly the same part of the sky, and the average total radiation at the surface is nearly unchanged. The increase in cloud size might be important for weather models, as it can impact the formation of rain, for example.
Raquel González-Armas, Jordi Vilà-Guerau de Arellano, Mary Rose Mangan, Oscar Hartogensis, and Hugo de Boer
Biogeosciences, 21, 2425–2445, https://doi.org/10.5194/bg-21-2425-2024, https://doi.org/10.5194/bg-21-2425-2024, 2024
Short summary
Short summary
This paper investigates the water and CO2 exchange for an alfalfa field with observations and a model with spatial scales ranging from the stomata to the atmospheric boundary layer. To relate the environmental factors to the leaf gas exchange, we developed three equations that quantify how many of the temporal changes of the leaf gas exchange occur due to changes in the environmental variables. The novelty of the research resides in the capacity to dissect the dynamics of the leaf gas exchange.
Robbert P. J. Moonen, Getachew A. Adnew, Oscar K. Hartogensis, Jordi Vilà-Guerau de Arellano, David J. Bonell Fontas, and Thomas Röckmann
Atmos. Meas. Tech., 16, 5787–5810, https://doi.org/10.5194/amt-16-5787-2023, https://doi.org/10.5194/amt-16-5787-2023, 2023
Short summary
Short summary
Isotope fluxes allow for net ecosystem gas exchange fluxes to be partitioned into sub-components like plant assimilation, respiration and transpiration, which can help us better understand the environmental drivers of each partial flux. We share the results of a field campaign isotope fluxes were derived using a combination of laser spectroscopy and eddy covariance. We found lag times and high frequency signal loss in the isotope fluxes we derived and present methods to correct for both.
Bert G. Heusinkveld, Wouter B. Mol, and Chiel C. van Heerwaarden
Atmos. Meas. Tech., 16, 3767–3785, https://doi.org/10.5194/amt-16-3767-2023, https://doi.org/10.5194/amt-16-3767-2023, 2023
Short summary
Short summary
This paper presents a new instrument for fast measurements of solar irradiance in 18 wavebands (400–950 nm): GPS perfectly synchronizes 10 Hz measurement speed to universal time, low-cost (< EUR 200) complete standalone solution for realizing dense measurement grids to study cloud-shading dynamics, 940 nm waveband reveals atmospheric moisture column information, 11 wavebands to study photosynthetic active radiation and light interaction with vegetation, and good reflection spectra performance.
Wouter B. Mol, Wouter H. Knap, and Chiel C. van Heerwaarden
Earth Syst. Sci. Data, 15, 2139–2151, https://doi.org/10.5194/essd-15-2139-2023, https://doi.org/10.5194/essd-15-2139-2023, 2023
Short summary
Short summary
We describe a dataset of detailed measurements of sunlight reaching the surface, recorded at a rate of one measurement per second for 10 years. The dataset includes detailed information on direct and scattered sunlight; classifications and statistics of variability; and observations of clouds, atmospheric composition, and wind. The dataset can be used to study how the atmosphere influences sunlight variability and to validate models that aim to predict this variability with greater accuracy.
Luuk D. van der Valk, Adriaan J. Teuling, Luc Girod, Norbert Pirk, Robin Stoffer, and Chiel C. van Heerwaarden
The Cryosphere, 16, 4319–4341, https://doi.org/10.5194/tc-16-4319-2022, https://doi.org/10.5194/tc-16-4319-2022, 2022
Short summary
Short summary
Most large-scale hydrological and climate models struggle to capture the spatially highly variable wind-driven melt of patchy snow cover. In the field, we find that 60 %–80 % of the total melt is wind driven at the upwind edge of a snow patch, while it does not contribute at the downwind edge. Our idealized simulations show that the variation is due to a patch-size-independent air-temperature reduction over snow patches and also allow us to study the role of wind-driven snowmelt on larger scales.
Felipe Lobos-Roco, Oscar Hartogensis, Francisco Suárez, Ariadna Huerta-Viso, Imme Benedict, Alberto de la Fuente, and Jordi Vilà-Guerau de Arellano
Hydrol. Earth Syst. Sci., 26, 3709–3729, https://doi.org/10.5194/hess-26-3709-2022, https://doi.org/10.5194/hess-26-3709-2022, 2022
Short summary
Short summary
This research brings a multi-scale temporal analysis of evaporation in a saline lake of the Atacama Desert. Our findings reveal that evaporation is controlled differently depending on the timescale. Evaporation is controlled sub-diurnally by wind speed, regulated seasonally by radiation and modulated interannually by ENSO. Our research extends our understanding of evaporation, contributing to improving the climate change assessment and efficiency of water management in arid regions.
Anja Ražnjević, Chiel van Heerwaarden, and Maarten Krol
Atmos. Meas. Tech., 15, 3611–3628, https://doi.org/10.5194/amt-15-3611-2022, https://doi.org/10.5194/amt-15-3611-2022, 2022
Short summary
Short summary
We evaluate two widely used observational techniques (Other Test Method (OTM) 33A and car drive-bys) that estimate point source gas emissions. We performed our analysis on high-resolution plume dispersion simulation. For car drive-bys we found that at least 15 repeated measurements were needed to get within 40 % of the true emissions. OTM 33A produced large errors in estimation (50 %–200 %) due to its sensitivity to dispersion coefficients and underlying simplifying assumptions.
Anja Ražnjević, Chiel van Heerwaarden, Bart van Stratum, Arjan Hensen, Ilona Velzeboer, Pim van den Bulk, and Maarten Krol
Atmos. Chem. Phys., 22, 6489–6505, https://doi.org/10.5194/acp-22-6489-2022, https://doi.org/10.5194/acp-22-6489-2022, 2022
Short summary
Short summary
Mobile measurement techniques (e.g., instruments placed in cars) are often employed to identify and quantify individual sources of greenhouse gases. Due to road restrictions, those observations are often sparse (temporally and spatially). We performed high-resolution simulations of plume dispersion, with realistic weather conditions encountered in the field, to reproduce the measurement process of a methane plume emitted from an oil well and provide additional information about the plume.
Carlos Román-Cascón, Marie Lothon, Fabienne Lohou, Oscar Hartogensis, Jordi Vila-Guerau de Arellano, David Pino, Carlos Yagüe, and Eric R. Pardyjak
Geosci. Model Dev., 14, 3939–3967, https://doi.org/10.5194/gmd-14-3939-2021, https://doi.org/10.5194/gmd-14-3939-2021, 2021
Short summary
Short summary
The type of vegetation (or land cover) and its status influence the heat and water transfers between the surface and the air, affecting the processes that develop in the atmosphere at different (but connected) spatiotemporal scales. In this work, we investigate how these transfers are affected by the way the surface is represented in a widely used weather model. The results encourage including realistic high-resolution and updated land cover databases in models to improve their predictions.
Felipe Lobos-Roco, Oscar Hartogensis, Jordi Vilà-Guerau de Arellano, Alberto de la Fuente, Ricardo Muñoz, José Rutllant, and Francisco Suárez
Atmos. Chem. Phys., 21, 9125–9150, https://doi.org/10.5194/acp-21-9125-2021, https://doi.org/10.5194/acp-21-9125-2021, 2021
Short summary
Short summary
We investigate the influence of regional atmospheric circulation on the evaporation of a saline lake in the Altiplano region of the Atacama Desert through a field experiment and regional modeling. Our results show that evaporation is controlled by two regimes: (1) in the morning by local conditions with low evaporation rates and low wind speed and (2) in the afternoon with high evaporation rates and high wind speed. Afternoon winds are connected to the regional Pacific Ocean–Andes flow.
Cited articles
Abadi, M., Barham, P., Chen, J., Chen, Z., Davis, A., Dean, J., Devin, M., Ghemawat, S., Irving, G., Isard, M., Kudlur, M., Levenberg, J., Monga, R., Moore, S., Murray, D. G., Steiner, B., Tucker, P., Vasudevan, V., Warden, P., Wicke, M., Yu, Y., and Zheng, X.: TensorFlow: A system for
large-scale machine learning, in: 12th (USENIX) Symposium on Operating
Systems Design and Implementation (OSDI 16), 265–283, 2016. a
Bardina, J., Ferziger, J., and Reynolds, W.: Improved subgrid-scale models for
large-eddy simulation, in: 13th fluid and plasmadynamics conference, 1357,
https://doi.org/10.2514/6.1980-1357, 1980. a
Bolton, T. and Zanna, L.: Applications of deep learning to ocean data inference
and subgrid parameterization, J. Adv. Model. Earth Sy.,
11, 376–399, https://doi.org/10.1029/2018MS001472, 2019. a
Bou-Zeid, E., Meneveau, C., and Parlange, M.: A scale-dependent Lagrangian
dynamic model for large eddy simulation of complex turbulent flows, Phys. Fluids, 17, 025105, https://doi.org/10.1063/1.1839152, 2005. a
Breiman, L.: Random forests, Mach. Learn., 45, 5–32,
https://doi.org/10.1023/A:1010933404324, 2001. a
Brenowitz, N. D. and Bretherton, C. S.: Spatially extended tests of a neural
network parametrization trained by coarse-graining, J. Adv.
Model. Earth Sy., 11, 2728–2744, https://doi.org/10.1029/2019MS001711, 2019. a
Brunton, S. L., Noack, B. R., and Koumoutsakos, P.: Machine learning for fluid
mechanics, Annu. Rev. Fluid Mech., 52, 477–508,
https://doi.org/10.1146/annurev-fluid-010719-060214, 2020. a
Chow, F. K. and Moin, P.: A further study of numerical errors in large-eddy
simulations, J. Comput. Phys., 184, 366–380,
https://doi.org/10.1016/S0021-9991(02)00020-7, 2003. a, b
Clark, R. A., Ferziger, J. H., and Reynolds, W. C.: Evaluation of subgrid-scale
models using an accurately simulated turbulent flow, J. Fluid
Mech., 91, 1–16, https://doi.org/10.1017/S002211207900001X, 1979. a, b
Denaro, F. M.: What does Finite Volume-based implicit filtering really resolve
in Large-Eddy Simulations?, J. Comput. Phys., 230,
3849–3883, https://doi.org/10.1016/j.jcp.2011.02.011, 2011. a, b
Duraisamy, K., Iaccarino, G., and Xiao, H.: Turbulence modeling in the age of
data, Annu. Rev. Fluid Mech., 51, 357–377,
https://doi.org/10.1146/annurev-fluid-010518-040547, 2019. a
Fisher, A., Rudin, C., and Dominici, F.: All Models are Wrong, but Many are
Useful: Learning a Variable's Importance by Studying an Entire Class of
Prediction Models Simultaneously, J. Mach. Learn. Res., 20,
1–81, 2019. a
Gamahara, M. and Hattori, Y.: Searching for turbulence models by artificial
neural network, Phys. Rev. Fluids, 2, 054604,
https://doi.org/10.1103/PhysRevFluids.2.054604, 2017. a, b, c, d
Ghosal, S.: An analysis of numerical errors in large-eddy simulations of
turbulence, J. Comput. Phys., 125, 187–206,
https://doi.org/10.1006/jcph.1996.0088, 1996. a
Giacomini, B. and Giometto, M. G.: On the suitability of second-order accurate finite-volume solvers for the simulation of atmospheric boundary layer flow, Geosci. Model Dev., 14, 1409–1426, https://doi.org/10.5194/gmd-14-1409-2021, 2021. a
Guan, Y., Chattopadhyay, A., Subel, A., and Hassanzadeh, P.: Stable a
posteriori LES of 2D turbulence using convolutional neural networks:
Backscattering analysis and generalization to higher Re via transfer
learning, arXiv preprint arXiv:2102.11400,
available at: https://arxiv.org/pdf/2102.11400.pdf (last access: 1 March 2021), 2021. a, b, c, d, e, f
He, K., Zhang, X., Ren, S., and Sun, J.: Delving Deep into Rectifiers:
Surpassing Human-Level Performance on ImageNet Classification, in: The IEEE
International Conference on Computer Vision (ICCV), 1026–1034,
https://doi.org/10.1109/ICCV.2015.123, 2015. a
Hornik, K., Stinchcombe, M., and White, H.: Multilayer feedforward networks are
universal approximators., Neural Networks, 2, 359–366,
https://doi.org/10.1016/0893-6080(89)90020-8, 1989. a
Jimenez, J. and Moser, R. D.: Large-eddy simulations: where are we and what can
we expect?, AIAA journal, 38, 605–612, https://doi.org/10.2514/2.1031, 2000. a
Kaandorp, M. L. A. and Dwight, R. P.: Data-driven modelling of the Reynolds
stress tensor using random forests with invariance, Comput. Fluids, 202,
104497, https://doi.org/10.1016/j.compfluid.2020.104497, 2020. a
Kravchenko, A. G. and Moin, P.: On the effect of numerical errors in large eddy
simulations of turbulent flows, J. Comput. Phys., 131,
310–322, https://doi.org/10.1006/jcph.1996.5597, 1997. a
Kutz, J. N.: Deep learning in fluid dynamics, J. Fluid Mech., 814,
1–4, https://doi.org/10.1017/jfm.2016.803, 2017. a
Langford, J. A. and Moser, R. D.: Optimal LES formulations for isotropic
turbulence, J. Fluid Mech., 398, 321–346,
https://doi.org/10.1017/S0022112099006369, 1999. a, b
Langford, J. A. and Moser, R. D.: Breakdown of continuity in large-eddy
simulation, Phys. Fluids, 13, 1524–1527, https://doi.org/10.1063/1.1358876, 2001. a
Lilly, D. K.: The representation of small-scale turbulence in numerical
simulation experiments, in: Proceedings of the IBM Scientific Computing
Symposium on Environmental Sciences, 195–210, https://doi.org/10.5065/D62R3PMM,
1967. a, b, c
Ling, J., Jones, R., and Templeton, J.: Machine learning strategies for systems
with invariance properties, J. Comput. Phys., 318, 22–35,
https://doi.org/10.1016/j.jcp.2016.05.003, 2016a. a
Ling, J., Kurzawski, A., and Templeton, J.: Reynolds averaged turbulence
modelling using deep neural networks with embedded invariance, J.
Fluid Mech., 807, 155–166, https://doi.org/10.1017/jfm.2016.615,
2016b. a, b
Liu, S., Meneveau, C., and Katz, J.: On the properties of similarity
subgrid-scale models as deduced from measurements in a turbulent jet, J. Fluid Mech., 275, 83–119, https://doi.org/10.1017/S0022112094002296, 1994. a, b, c
Maulik, R., San, O., Rasheed, A., and Vedula, P.: Subgrid modelling for
two-dimensional turbulence using neural networks, J. Fluid Mech.,
858, 122–144, https://doi.org/10.1017/jfm.2018.770, 2019. a, b, c, d
McMillan, O. J. and Ferziger, J. H.: Direct testing of subgrid-scale models,
AIAA Journal, 17, 1340–1346, https://doi.org/10.2514/3.61313, 1979. a, b
Milano, M. and Koumoutsakos, P.: Neural network modeling for near wall
turbulent flow, J. Comput. Phys., 182, 1–26,
https://doi.org/10.1006/jcph.2002.7146, 2002. a, b
Molnar, C.: Interpretable Machine Learning, lulu.com,
available at: https://christophm.github.io/interpretable-ml-book/ (last access: 14 April 2021), 2019. a
Moser, R. D., Kim, J., and Mansour, N. N.: Direct numerical simulation of
turbulent channel flow up to Re τ= 590, Phys. Fluids, 11, 943–945,
https://doi.org/10.1063/1.869966, 1999. a
Nadiga, B. T. and Livescu, D.: Instability of the perfect subgrid model in
implicit-filtering large eddy simulation of geostrophic turbulence, Phys.
Rev. E, 75, 046 303, https://doi.org/10.1103/PhysRevE.75.046303, 2007. a
Rasp, S.: Coupled online learning as a way to tackle instabilities and biases in neural network parameterizations: general algorithms and Lorenz 96 case study (v1.0), Geosci. Model Dev., 13, 2185–2196, https://doi.org/10.5194/gmd-13-2185-2020, 2020. a, b
Rumelhart, D. E., Hinton, G. E., and Williams, R. J.: Learning representations
by back-propagating errors, Nature, 323, 533–536, https://doi.org/10.1038/323533a0,
1986. a
Sarghini, F., De Felice, G., and Santini, S.: Neural networks based subgrid
scale modeling in large eddy simulations, Comput. Fluids, 32, 97–108,
https://doi.org/10.1016/S0045-7930(01)00098-6, 2003. a, b, c
Schmitt, F. G.: About Boussinesq's turbulent viscosity hypothesis: historical
remarks and a direct evaluation of its validity, Comptes Rendus
Mécanique, 335, 617–627, https://doi.org/10.1016/j.crme.2007.08.004, 2007. a
Singh, A. P., Duraisamy, K., and Zhang, Z. J.: Augmentation of turbulence
models using field inversion and machine learning, in: 55th AIAA Aerospace
Sciences Meeting, 0993, https://doi.org/10.2514/6.2017-0993, 2017. a
Smagorinsky, J.: General circulation experiments with the primitive equations:
I. The basic experiment, Mon. Weather Rev., 91, 99–164,
https://doi.org/10.1175/1520-0493(1963)091<0099:GCEWTP>2.3.CO;2, 1963. a, b
Stoffer, R.: robinstoffer/microhh2: Rep corresponding to GMD publication
Stoffer et al. (2021) (Version ANN_SGS_v1.2-alpha) [code], Zenodo,
https://doi.org/10.5281/zenodo.4767902, 2021. a
Van Driest, E. R.: On turbulent flow near a wall, J. Aeronaut.
Sci., 23, 1007–1011, https://doi.org/10.2514/8.3713, 1956. a, b
van Heerwaarden, C. C., van Stratum, B. J. H., and Heus, T.: microhh/microhh:
1.0.0 (Version 1.0.0) [code], Zenodo, https://doi.org/10.5281/zenodo.822842, 2017a. a
van Heerwaarden, C. C., van Stratum, B. J. H., Heus, T., Gibbs, J. A., Fedorovich, E., and Mellado, J. P.: MicroHH 1.0: a computational fluid dynamics code for direct numerical simulation and large-eddy simulation of atmospheric boundary layer flows, Geosci. Model Dev., 10, 3145–3165, https://doi.org/10.5194/gmd-10-3145-2017, 2017b. a, b, c, d, e, f
Völker, S., Moser, R. D., and Venugopal, P.: Optimal large eddy simulation
of turbulent channel flow based on direct numerical simulation statistical
data, Phys. Fluids, 14, 3675–3691, https://doi.org/10.1063/1.1503803, 2002. a
Vollant, A., Balarac, G., and Corre, C.: Subgrid-scale scalar flux modelling
based on optimal estimation theory and machine-learning procedures, J. Turbulence, 18, 854–878, https://doi.org/10.1080/14685248.2017.1334907, 2017. a, b
Wang, J., Wu, J., and Xiao, H.: Physics-informed machine learning approach for
reconstructing Reynolds stress modeling discrepancies based on DNS data,
Phys. Rev. Fluids, 2, 034603, https://doi.org/10.1103/PhysRevFluids.2.034603,
2017. a
Wang, Z., Luo, K., Li, D., Tan, J., and Fan, J.: Investigations of data-driven
closure for subgrid-scale stress in large-eddy simulation, Phys. Fluids,
30, 125101, https://doi.org/10.1063/1.5054835, 2018. a, b, c
Wu, J., Xiao, H., and Paterson, E.: Physics-informed machine learning approach
for augmenting turbulence models: A comprehensive framework, Phys. Rev.
Fluids, 3, 074602, https://doi.org/10.1103/PhysRevFluids.3.074602, 2018.
a
Xie, C., Wang, J., Li, K., and Ma, C.: Artificial neural network approach to
large-eddy simulation of compressible isotropic turbulence, Phys. Rev.
E, 99, 053113, https://doi.org/10.1103/PhysRevE.99.053113, 2019. a, b, c
Yuval, J. and O'Gorman, P. A.: Stable machine-learning parameterization of
subgrid processes for climate modeling at a range of resolutions, Nat.
Commun., 11, 1–10, https://doi.org/10.1038/s41467-020-17142-3, 2020. a
Zhou, Z., He, G., Wang, S., and Jin, G.: Subgrid-scale model for large-eddy
simulation of isotropic turbulent flows using an artificial neural network,
Comput. Fluids, 195, 104319, https://doi.org/10.1016/j.compfluid.2019.104319,
2019. a, b, c, d
Short summary
Turbulent flows are often simulated with the large-eddy simulation (LES) technique, which requires subgrid models to account for the smallest scales. Current subgrid models often require strong simplifying assumptions. We therefore developed a subgrid model based on artificial neural networks, which requires fewer assumptions. Our data-driven SGS model showed high potential in accurately representing the smallest scales but still introduced instability when incorporated into an actual LES.
Turbulent flows are often simulated with the large-eddy simulation (LES) technique, which...