Search NASA⌕ Search

SEARCH · Search NASA

Results for “subgrid”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 127 records · Page 7
All sourcesNASA NTRSDOE OSTI
NASA NTRS2012

Correction of Excessive Precipitation over Steep Mountains in a General Circulation Model (GCM)

Excessive precipitation over steep and high mountains (EPSM) is a well-known problem in GCMs and regional climate models even at a resolution as high as 19km. The affected regions include the Andes, the Himalayas, Sierra Madre, New Guinea and others. This problem also shows up in some data assimilation products. Among the possible causes investigated in this study, we found that the most important one, by far, is a missing upward transport of heat out of the boundary layer due to the vertical circulations forced by the daytime subgrid-scale upslope winds, which in turn is forced by heated boundary layer on the slopes. These upslope winds are associated with large subgrid-scale topographic variance, which is found over steep mountains. Without such subgrid-scale heat ventilation, the resolvable-scale upslope flow in the boundary layer generated by surface sensible heat flux along the mountain slopes is excessive. Such an excessive resolvable-scale upslope flow in the boundary layer combined with the high moisture content in the boundary layer results in excessive moisture transport toward mountaintops, which in turn gives rise to excessive precipitation over the affected regions. We have parameterized the effects of subgrid-scale heated-slope-induced vertical circulation (SHVC) by removing heat from the boundary layer and depositing it in the layers higher up when topographic variance exceeds a critical value. Test results using NASA/Goddard's GEOS-5 GCM have shown that the EPSM problem is largely solved.

Chao, Winston C.↗
DOE OSTI2024

Methods to Evaluate Subcolumn Profiles Based on Two-Point Diagnostics

In atmospheric models, stochastic generation of subgrid-scale profiles or “subcolumns” has been used for a variety of purposes. Such subcolumns can be generated from subgrid probability density functions (PDFs) at different vertical levels, when such PDFs are available. To do so, the generator needs to decide how strongly points should be correlated in the vertical, that is, how much the values should be overlapped. This is sometimes called “PDF overlap.” To assess vertical correlation in a simplified, observable setting, here the vertical correlation of vertical velocity in subcloud layers is examined. Doppler lidar is used to evaluate the vertical profiles of vertical velocity produced by a large-eddy simulation (LES) model and the Subgrid Importance Latin Hypercube Sampler (SILHS) subcolumn generator. In order to diagnose unrealistic features in subcolumn profiles, various statistical diagnostics are examined here, including the bivariate PDF of vertical velocity at two separated points (i.e., altitudes), the two-point velocity correlation, the integral correlation length, the PDF of two-point velocity differences, and the skewness and kurtosis of two-point velocity differences. The profiles produced by LES match lidar well, except that they are too smooth at small scales. The profiles produced by SILHS exhibit sharp jumps from updraft to downdraft that are not observed in the lidar data. To reduce the generation of these unrealistically sharp jumps, the SILHS sampling method is revised. The diagnostics confirm that the revised sampling method reduces the overprediction of sharp jumps.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2025

Simulating Atmospheric Processes in Earth System Models and Quantifying Uncertainties With Deep Learning Multi‐Member and Stochastic Parameterizations

Abstract Deep learning is a powerful tool to represent subgrid processes in climate models, but many application cases have so far used idealized settings and deterministic approaches. Here, we develop stochastic parameterizations with calibrated uncertainty quantification to learn subgrid convective and turbulent processes and surface radiative fluxes of a superparameterization embedded in an Earth System Model (ESM). We explore three methods to construct stochastic parameterizations: (a) a single Deep Neural Network (DNN) with Monte Carlo Dropout; (b) a multi‐member parameterization; and (c) a Variational Encoder Decoder with latent space perturbation. We show that the multi‐member parameterization improves the representation of convective processes, especially in the planetary boundary layer, compared to individual DNNs. The respective uncertainty quantification illustrates that methods (b) and (c) are advantageous compared to a dropout‐based DNN parameterization regarding the spread of convective processes. Hybrid simulations with our best‐performing multi‐member parameterizations remained challenging and crash within the first days. Therefore, we develop a pragmatic partial coupling strategy relying on the superparameterization for condensate emulation. Partial coupling reduces the computational efficiency of hybrid Earth‐like simulations but enables model stability over 5 months with our multi‐member parameterizations. However, our hybrid simulations exhibit biases in thermodynamic fields and differences in precipitation patterns. Despite this, the multi‐member parameterizations enable improvements in reproducing tropical extreme precipitation compared to a traditional convection parameterization. Despite these challenges, our results indicate the potential of a new generation of multi‐member machine learning parameterizations leveraging uncertainty quantification to improve the representation of stochasticity of subgrid effects.

Behrens, Gunnar [Deutsches Zentrum für Luft‐ und R↗
DOE OSTI2025

Implementation of Sub‐Grid Scale Temperature Perturbations Induced by Non‐Orographic Gravity Waves in WACCM6

Abstract Atmospheric gravity waves can play a significant role on atmospheric chemistry through temperature fluctuations. A recent modeling study introduced a method to implement subgrid‐scale orographic gravity‐wave‐induced temperature perturbations in the Whole Atmosphere Community Climate Model (WACCM). The model with a wave‐induced temperature parameterization was able to reproduce for example, the influence of mountain wave events on atmospheric chemistry, as highlighted in previous literature. Here we extend the subgrid‐scale wave‐induced temperature parameterization to also include non‐orographic gravity waves arising from frontal activity and convection. We explore the impact of these waves on middle atmosphere chemistry, particularly focusing on reactions that are strongly sensitive to temperature. The non‐orographic gravity waves increase the variability of chemical reaction rates, especially in the lower mesosphere. As an example, we show that this, in turn, leads to increases in the daytime ozone variability. To demonstrate another impact, we briefly investigate the role of non‐orographic gravity waves in cirrus cloud formation in this model. Consistent with findings from the previous study focusing on orographic gravity waves, non‐orographic waves also enhance homogeneous nucleation and increase cirrus clouds. The updated method used enables the global chemistry‐climate model to account for both orographic and non‐orographic gravity‐wave‐induced subgrid‐scale dynamical perturbations in a consistent manner.

Yook, Simchan [Department of Earth Atmospheric and↗
DOE OSTI2026

Exposing and Reducing Biases of Simulating Mixed-Phase Clouds in the Convection-Permitting E3SM Atmosphere Model: Lessons From an Arctic Cold-Air Outbreak

Mixed-phase clouds modulate the water and energy cycles of high-latitude regions, yet their liquid-ice phase partitioning has long been poorly simulated in climate models. Here, simulations of Arctic mixed-phase clouds by the Simple Cloud-Resolving E3SM Atmosphere Model (SCREAM) are assessed against large-eddy simulations, satellite data, and ground-based observations during the Cold-Air Outbreaks in the Marine Boundary Layer Experiment field campaign. SCREAM simulates nearly completely frozen clouds, which is attributed largely to the unreasonably strong Wegener–Bergeron–Findeisen (WBF) process that converts liquid to ice excessively and partly to the early over-abundant ice production at cold temperatures from a temperature-deterministic deposition ice nucleation scheme. Assuming no subgrid variation for the WBF process in the original formulation particularly conflicts with the instantaneous saturation adjustment assumption in the condensation scheme that assumes subgrid variability, leading to exaggerated WBF process rates. A proposed simple physically-based improvement on the treatment of subgrid cloud overlap substantially increases supercooled liquid water content and notably improves cloud-top phase partitioning, aligning better with observations. Improvement of supercooled liquid water content also converges with increasing horizontal resolution. The deposition ice nucleation scheme is found responsible for a falsely-produced ice cloud aloft that is not observed, biasing the simulated cloud radiative effects and top-of-atmosphere radiative fluxes. This study identifies key deficiencies in cloud parameterizations that continue to challenge convection-permitting models.

Geosciences↗
DOE OSTI2024

WRF-Comfort: simulating microscale variability in outdoor heat stress at the city scale with a mesoscale model

Abstract. Urban overheating and its ongoing exacerbation due to global warming and urban development lead to increased exposure to urban heat and increased thermal discomfort and heat stress. To quantify thermal stress, specific indices have been proposed that depend on air temperature, mean radiant temperature (MRT), wind speed, and relative humidity. While temperature and humidity vary on scales of hundreds of meters, MRT and wind speed are strongly affected by individual buildings and trees and vary on the meter scale. Therefore, most numerical thermal comfort studies apply microscale models to limited spatial domains (commonly representing urban neighborhoods with building blocks) with resolutions on the order of 1 m and a few hours of simulation. This prevents the analysis of the impact of city-scale adaptation and/or mitigation strategies on thermal stress and comfort. To solve this problem, we develop a methodology to estimate thermal stress indicators and their subgrid variability in mesoscale models – here applied to the multilayer urban canopy parameterization BEP-BEM within the Weather Research and Forecasting (WRF) model. The new scheme (consisting of three main steps) can readily assess intra-neighborhood-scale heat stress distributions across whole cities and for timescales of minutes to years. The first key component of the approach is the estimation of MRT in several locations within streets for different street orientations. Second, mean wind speed and its subgrid variability are downscaled as a function of the local urban morphology based on relations derived from a set of microscale LES and RANS simulations across a wide range of realistic and idealized urban morphologies. Lastly, we compute the distributions of two thermal stress indices for each grid square, combining all the subgrid values of MRT, wind speed, air temperature, and absolute humidity. From these distributions, we quantify the high and low tails of the heat stress distribution in each grid square across the city, representing the thermal diversity experienced in street canyons. In this contribution, we present the core methodology as well as simulation results for Madrid (Spain), which illustrate strong differences between heat stress indices and common heat metrics like air or surface temperature both across the city and over the diurnal cycle.

Geology↗
DOE OSTI2026

Slow Wake Recovery and Low Turbulence Behind Wind Farms Parameterized in Mesoscale Simulations

Numerical weather prediction (NWP) and climate models equipped with wind-farm parameterizations (WFPs) can simulate cluster wake effects affecting downstream wind farms in both onshore and offshore environments. This study evaluates wake recovery behind a wind farm represented by the NWP-WFP approach in the Weather Research and Forecasting (WRF) model using either the Fitch et al. (2012) or Ma et al. (2022a, b) WFPs. Results are benchmarked against large-eddy simulations (LES) of an idealized offshore wind farm with aligned and staggered layouts under neutral atmospheric stability. Near-farm wake recovery is underestimated in NWP-WFP simulations due to its representation on a coarse mesoscale grid. This limitation leads to slow wake recovery through two interconnected mechanisms: (i) spatial gradients in the wind velocity field are weaker compared to LES and (ii) turbulence kinetic energy (TKE) remains low not because of excessive dissipation but due to insufficient shear production caused by these weakened gradients. For the scenario considered here, a wind-speed bias develops in the near-farm wake and persists into the far wake. Differences between the NWP-WFP simulations and LES emerge within a short distance downstream of the farm exit, where the mesoscale simulations recover too slowly. This reduced recovery contributes approximately 0.15-0.50 m s-1 to the near-farm wind-speed bias. The bias established in this region is not subsequently compensated for downstream but instead propagates into the far wake, where wind-speed differences of approximately 0.4-0.6 m s-1 remain up to 50 km downstream. Higher-resolution mesoscale simulations partially reduce this bias. Increasing turbine-added TKE or including subgrid wake effects provides additional improvement, but neither fully addresses the underlying cause. The slow wake recovery is not caused by limitations of the WFPs themselves, as it also occurs outside their region of influence, and adding subgrid wake effects does not significantly impact recovery. Rather, the slow wake recovery is a consequence of mesoscale flow representation. This behavior is not limited to regions downstream of the wind farm but is less visible within the farm, where wake recovery occurs simultaneously with turbine-induced momentum extraction. These results highlight the need for improved representations of wake recovery both within and downstream of wind farms. While enhanced subgrid modeling, shear-driven TKE production, and refined WFP formulations may improve intra-farm dynamics, accurately capturing near-farm wake recovery downstream remains challenging, as WFPs do not act in this region.

17 WIND ENERGY↗
NASA NTRS1990

Quantized-"Gray-Scale" Electronic Synapses

Proposed array of programmable synaptic connections for electronic neural network applications offers multiple quantized levels of connection strength using only simple, two-terminal, binary microswitch devices. Subgrids in fine grid of programmable resistive connections connected externally in parallel to form coarser synaptic grid. By selection of pattern of connections in each subgrid, connection strength of synaptic node represented by that subgrid set at quantized "gray level". Device structures promise implementations of quantized-"gray-scale" synaptic arrays with very high density.

Lamb, James L.↗
NASA NTRS1993

Large eddy simulation and direct numerical simulation of high speed turbulent reacting flows

The objective of this research is to make use of Large Eddy Simulation (LES) and Direct Numerical Simulation (DNS) for the computational analyses of high speed reacting flows. Our efforts in the first phase of this research conducted within the past three years have been directed in several issues pertaining to intricate physics of turbulent reacting flows. In our previous 5 semi-annual reports submitted to NASA LaRC, as well as several technical papers in archival journals, the results of our investigations have been fully described. In this progress report which is different in format as compared to our previous documents, we focus only on the issue of LES. The reason for doing so is that LES is the primary issue of interest to our Technical Monitor and that our other findings were needed to support the activities conducted under this prime issue. The outcomes of our related investigations, nevertheless, are included in the appendices accompanying this report. The relevance of the materials in these appendices are, therefore, discussed only briefly within the body of the report. Here, results are presented of a priori and a posterior analyses for validity assessments of assumed Probability Density Function (PDF) methods as potential subgrid scale (SGS) closures for LES of turbulent reacting flows. Simple non-premixed reacting systems involving an isothermal reaction of the type A + B yields Products under both chemical equilibrium and non-equilibrium conditions are considered. A priori analyses are conducted of a homogeneous box flow, and a spatially developing planar mixing layer to investigate the performance of the Pearson Family of PDF's as SGS models. A posteriori analyses are conducted of the mixing layer using a hybrid one-equation Smagorinsky/PDF SGS closure. The Smagorinsky closure augmented by the solution of the subgrid turbulent kinetic energy (TKE) equation is employed to account for hydrodynamic fluctuations, and the PDF is employed for modeling the effects of scalar fluctuations. The implementation of the model requires the knowledge of the local values of the first two SGS moments. These are provided by additional modeled transport equations. In both a priori and a posteriori analyses, the predicted results are appraised by comparison with subgrid averaged results generated by DNS. Based on these results, the paths to be followed in future investigations are identified.

Adumitroaie, V.↗
NASA NTRS1996

Parallel Simulation of Unsteady Turbulent Flames

Time-accurate simulation of turbulent flames in high Reynolds number flows is a challenging task since both fluid dynamics and combustion must be modeled accurately. To numerically simulate this phenomenon, very large computer resources (both time and memory) are required. Although current vector supercomputers are capable of providing adequate resources for simulations of this nature, the high cost and their limited availability, makes practical use of such machines less than satisfactory. At the same time, the explicit time integration algorithms used in unsteady flow simulations often possess a very high degree of parallelism, making them very amenable to efficient implementation on large-scale parallel computers. Under these circumstances, distributed memory parallel computers offer an excellent near-term solution for greatly increased computational speed and memory, at a cost that may render the unsteady simulations of the type discussed above more feasible and affordable.This paper discusses the study of unsteady turbulent flames using a simulation algorithm that is capable of retaining high parallel efficiency on distributed memory parallel architectures. Numerical studies are carried out using large-eddy simulation (LES). In LES, the scales larger than the grid are computed using a time- and space-accurate scheme, while the unresolved small scales are modeled using eddy viscosity based subgrid models. This is acceptable for the moment/energy closure since the small scales primarily provide a dissipative mechanism for the energy transferred from the large scales. However, for combustion to occur, the species must first undergo mixing at the small scales and then come into molecular contact. Therefore, global models cannot be used. Recently, a new model for turbulent combustion was developed, in which the combustion is modeled, within the subgrid (small-scales) using a methodology that simulates the mixing and the molecular transport and the chemical kinetics within each LES grid cell. Finite-rate kinetics can be included without any closure and this approach actually provides a means to predict the turbulent rates and the turbulent flame speed. The subgrid combustion model requires resolution of the local time scales associated with small-scale mixing, molecular diffusion and chemical kinetics and, therefore, within each grid cell, a significant amount of computations must be carried out before the large-scale (LES resolved) effects are incorporated. Therefore, this approach is uniquely suited for parallel processing and has been implemented on various systems such as: Intel Paragon, IBM SP-2, Cray T3D and SGI Power Challenge (PC) using the system independent Message Passing Interface (MPI) compiler. In this paper, timing data on these machines is reported along with some characteristic results.

Menon, Suresh↗
NASA NTRS2001

The Numerical Analysis of a Turbulent Compressible Jet

A numerical method to simulate high Reynolds number jet flows was formulated and applied to gain a better understanding of the flow physics. Large-eddy simulation was chosen as the most promising approach to model the turbulent structures due to its compromise between accuracy and computational expense. The filtered Navier-Stokes equations were developed including a total energy form of the energy equation. Subgrid scale models for the momentum and energy equations were adapted from compressible forms of Smagorinsky's original model. The effect of using disparate temporal and spatial accuracy in a numerical scheme was discovered through one-dimensional model problems and a new uniformly fourth-order accurate numerical method was developed. Results from two- and three-dimensional validation exercises show that the code accurately reproduces both viscous and inviscid flows. Numerous axisymmetric jet simulations were performed to investigate the effect of grid resolution, numerical scheme, exit boundary conditions and subgrid scale modeling on the solution and the results were used to guide the three-dimensional calculations. Three-dimensional calculations of a Mach 1.4 jet showed that this LES simulation accurately captures the physics of the turbulent flow. The agreement with experimental data was relatively good and is much better than results in the current literature. Turbulent intensities indicate that the turbulent structures at this level of modeling are not isotropic and this information could lend itself to the development of improved subgrid scale models for LES and turbulence models for RANS simulations. A two point correlation technique was used to quantify the turbulent structures. Two point space correlations were used to obtain a measure of the integral length scale, which proved to be approximately 1/2 D(sub j). Two point space-time correlations were used to obtain the convection velocity for the turbulent structures. This velocity ranged from 0.57 to 0.71 U(sub j).

DeBonis, James R.↗
NASA NTRS2006

Computing Flows Using Chimera and Unstructured Grids

DRAGONFLOW is a computer program that solves the Navier-Stokes equations of flows in complexly shaped three-dimensional regions discretized by use of a direct replacement of arbitrary grid overlapping by nonstructured (DRAGON) grid. A DRAGON grid (see figure) is a combination of a chimera grid (a composite of structured subgrids) and a collection of unstructured subgrids. DRAGONFLOW incorporates modified versions of two prior Navier-Stokes-equation-solving programs: OVERFLOW, which is designed to solve on chimera grids; and USM3D, which is used to solve on unstructured grids. A master module controls the invocation of individual modules in the libraries. At each time step of a simulated flow, DRAGONFLOW is invoked on the chimera portion of the DRAGON grid in alternation with USM3D, which is invoked on the unstructured subgrids of the DRAGON grid. The USM3D and OVERFLOW modules then immediately exchange their solutions and other data. As a result, USM3D and OVERFLOW are coupled seamlessly.

Liou, Meng-Sing↗
NASA NTRS2012

Correction of Excessive Precipitation Over Steep and High Mountains in a General Circulation Model

Excessive precipitation over steep and high mountains (EPSM) is a well-known problem in GCMs and meso-scale models. This problem impairs simulation and data assimilation products. Among the possible causes investigated in this study, we found that the most important one, by far, is a missing upward transport of heat out of the boundary layer due to the vertical circulations forced by the daytime upslope winds, which are forced by the heated boundary layer on subgrid-scale slopes. These upslope winds are associated with large subgrid-scale topographic variation, which is found over steep and high mountains. Without such subgridscale heat ventilation, the resolvable-scale upslope flow in the boundary layer generated by surface sensible heat flux along the mountain slopes is excessive. Such an excessive resolvablescale upslope flow combined with the high moisture content in the boundary layer results in excessive moisture transport toward mountaintops, which in turn gives rise to EPSM. Other possible causes of EPSM that we have investigated include 1) a poorly-designed horizontal moisture flux in the terrain-following coordinates, 2) the condition for cumulus convection being too easily satisfied at mountaintops, 3) the presence of conditional instability of the computational kind, and 4) the absence of blocked flow drag. These are all minor or inconsequential. We have parameterized the ventilation effects of the subgrid-scale heated-slope-induced vertical circulation (SHVC) by removing heat from the boundary layer and depositing it in layers higher up when the topographic variance exceeds a critical value. Test results using NASA/Goddard's GEOS-S GCM have shown that this largely solved the EPSM problem.

Chao, Winston C.↗
NASA NTRS2023

Interactions Between Thresholds and Spatial Discretizations of Snow: Insights From Estimates of Wolverine Denning Habitat in the Colorado Rocky Mountains

Thresholds can be used to interpret environmental data in a way that is easily communicated and useful for decision making purposes. However, thresholds are often developed for specific data products and time periods, changing findings when the same threshold is applied to datasets or periods with different characteristics. Here, we test the impact of different spatial discretizations of snow on annual estimates of wolverine denning opportunities in the Colorado Rocky Mountains, defined using a snow water equivalent (SWE) threshold (0.20 m) and threshold date (15 May) from previous habitat assessments. Annual potential wolverine denning area (PWDA) was thresholded from a 36-year (1985 – 2020) snow reanalysis model with three different spatial discretizations: 1) 480 m grid cells (D480), 2) 90 m grid cells (D90), and 3) 480 m grid cells with implicit representations of subgrid snow spatial heterogeneity (S480). Relative to the D480 and S480 discretizations, D90 resolved shallower snow deposits on slopes between 3050 and 3350 m elevation, decreasing PWDA by 10%, on average. In years with warmer and/or drier winters, S480 discretizations with subgrid representations of snow heterogeneity increased PWDA, even within grid cells where mean 15 May SWE was less than the SWE threshold. These simulations increased PWDA by upwards of 30% in low snow years, as compared to the D480 and D90 simulations without subgrid snow heterogeneity. Despite PWDA sensitivity to different snow spatial discretizations, PWDA was controlled more by annual variations in winter precipitation and temperature. However, small changes to the SWE threshold (± 0.07 m) and threshold date (± 2 weeks) also affected PWDA by as much as 82%. Across these threshold ranges, PWDA was approximately 18% more sensitive to the SWE threshold than the threshold date. However, the sensitivity to the threshold date was larger in years with late spring snowfall, when PWDA depended on whether modeled SWE was thresholded before, during, or after spring snow accumulation. Our results demonstrate that snow thresholds are useful but may not always provide a complete picture of the annual variability in snow-adapted wildlife denning opportunities. Studies thresholding spatiotemporal datasets could be improved by including 1) information about the fidelity of thresholds across multiple spatial discretizations, and 2) uncertainties related to ranges of realistic thresholds.

Justin M. Pflug↗
DOE OSTI2025

Overset-Grid Method with Smooth Orbital Partitioning for Molecular Scattering Calculations

To solve molecular photoionization and electron scattering problems, we use an overset-grid representation of electronic continuum functions, which has an extended central spherical grid that overlaps small spherical grids (subgrids) centered on each atom of a polyatomic molecule. Here, in this work, we present an improved algorithm that smoothly partitions the total wave function between the central grid and the atomic subgrids. The smooth partitioning allows one to use approximately one-fourth the number of partial waves on the central grid compared to our previous implementation with switching functions. The resulting numerical method for treating electron scattering and photoionization of polyatomic molecules combines the accuracy and flexibility of pure numerical grid representations with the rapid convergence of hybrid combinations of atom-centered basis-set expansions and grid methods. The overset-grid representation is implemented using the complex Kohn variational principle for scattering and photoionization amplitudes. The faster convergence with respect to the number of central grid partial waves is demonstrated and accuracy is verified by comparisons with the previous implementation and with far more computationally demanding single-center numerical expansions in electron-molecule scattering and photoionization calculations on the neon dimer (Ne 2 ) system, carbon tetrafluoride (CF 4 ) molecule, and the pyridine (C 5 H 5 N) molecule in the static-exchange approximation.

Molecules↗
DOE OSTI2025

Navigating the Noise: Bringing Clarity to ML Parameterization Design With O $\boldsymbol{\mathcal{O}}$(100) Ensembles

Abstract Machine‐learning (ML) parameterizations of subgrid processes (here of turbulence, convection, and radiation) may one day replace conventional parameterizations by emulating high‐resolution physics without the cost of explicit simulation. However, uncertainty about the relationship between offline and online performance (i.e., when integrated with a large‐scale general circulation model) hinders their development. Much of this uncertainty stems from limited sampling of the noisy, emergent effects of upstream ML design decisions on downstream online hybrid simulation. Our work rectifies the sampling issue via the construction of a semi‐automated, end‐to‐end pipeline for size ensembles of hybrid simulations, revealing important nuances in how systematic reductions in offline error manifest in changes to online error and online stability. For example, removing dropout and switching from a Mean Squared Error to a Mean Absolute Error loss both reduce offline error, but they have opposite effects on online error and online stability. Other design decisions, like incorporating memory, converting moisture input from specific humidity to relative humidity, using batch normalization, and training on multiple climates do not come with any such compromises. Finally, we show that ensemble sizes of may be necessary to reliably detect causally relevant differences online. By enabling rapid online experimentation at scale, we can empirically settle debates regarding subgrid ML parameterization design that would have otherwise remained unresolved in the noise.

Lin, Jerry [Department of Earth System Sciences Un↗
DOE OSTI2024

Bimodality in simulated precipitation frequency distributions and its relationship with convective parameterizations

Abstract Bimodality in precipitation frequency distributions is often evident in atmospheric models, but rarely in observations. This study i) proposes a metric to objectively quantify the bimodality in precipitation distributions, ii) evaluates model simulations contributed to the Coupled Model Intercomparison Project (CMIP) phase 5 (CMIP5), phase 6 (CMIP6), and the DYnamics of the Atmospheric general circulation Modeled On Non-hydrostatic Domains (DYAMOND) project by comparing them to satellite-based and reanalysis precipitation products, and iii) investigates possible origins of bimodal precipitation distributions. Our results reveal that about 83% (20 out of 24) of CMIP5 and 70% (21 out of 30) of CMIP6 models used in this study exhibit bimodal distributions. The few DYAMOND models that use a deep convective parameterization also show bimodal distributions, while most DYAMOND models do not. Predictably, the bimodality originates from the separation of precipitation process between resolved grid-scale and parameterized subgrid-scale. However, in a larger number of models bimodality arises from the parameterized subgrid-scale convective precipitation alone.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2025

Evaluating quenching in cosmological simulations of galaxy formation with spectral covariance in the optical window

ABSTRACT Cosmological hydrodynamical simulations provide valuable insights on galaxy evolution when coupled with observational data. Comparisons with real galaxies are typically performed via scaling relations of the observables. Here, we follow an alternative approach based on the spectral covariance in a model-independent way. We build upon previous work by Sharbaf et al. that studied the covariance of high-quality SDSS (Sloan Digital Sky Survey) continuum-subtracted spectra in a relatively narrow range of velocity dispersion ($\sigma \in [100,150]$ km s$^{-1}$). Here, the same analysis is applied to synthetic data from the eagle and IllustrisTNG100 simulations, to assess the ability of these runs to mimic real galaxies. The real and simulated spectra are consistent regarding spectral covariance, although with subtle differences that can inform the implementation of subgrid physics. Spectral fitting done a posteriori on stacks segregated with respect to latent space reveals that the first principal component (PC1) is predominantly influenced by the stellar age distribution, with an underlying age–metallicity degeneracy. Good agreement is found regarding star formation prescriptions but there is disagreement with active galactic nucleus (AGN) feedback, that also affects the subset of quiescent galaxies. We show a substantial difference in the implementation of the AGN subgrid prescriptions, regarding central black hole seeding, that could lead to the mismatch. Differences are manifest between these two simulations in the star formation histories stacked with respect to latent space. We emphasize that this methodology only relies on the spectral covariance to assess whether simulations provide a true representation of galaxy formation.

Sharbaf, Z. (ORCID:0009000450545946)↗