Search NASA⌕ Search

SEARCH · Search NASA

Results for “Surrogate Mode”

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 19 records

Enabling in-time Prognostics with Surrogate Modeling through Physics-enhanced Dynamic Mode Decomposition Method

Computational models provide essential quantitative tools for assessing and predicting the health and performance of physical systems. However, high-fidelity models are rarely used in real-time operations or large optimization loops, due to their time-intensive nature. A common approach to improving computational efficiency of prognosis is to employ surrogate models. Such models can significantly decrease computation time for some accuracy loss. In this context, use of Dynamic Mode Decomposition (DMD) is proposed to generate surrogate models for lithium-ion (Li-ion) battery discharge. DMD has been suggested and used successfully in the area of fluid dynamics for over a decade, but it has not been applied to the PHM domain, where far-ahead prediction of nonlinear behavior is crucial to propagate faults or predict Remaining Useful Life (RUL). For Li-ion battery health management, the standard application of DMD using only the observable quantities of interest was unable to capture the nonlinear discharge of batteries exhibited in lab testing. The Koopman theory, however, provides a mechanism to tradeoff low dimensional nonlinear models with high-dimensional linear ones in a DMD framework, by augmenting nonlinear state variables into the system representation. In this way, DMD allows for configurable simulation accuracy dependent on the dimensionality of the Koopman operator. For battery health management, we augmented the observable variables with the hidden states of a higher-fidelity physics model to build the DMD surrogate. In comparison to a high-fidelity model, the surrogate improved computational efficiency with only a minimal loss of accuracy, and enabled long-term prognostics horizons. A generalized method for this was implemented in the prog models python package.

prognostics and health management↗

Improving Accuracy of Structural Dynamic Modification with Augmented Residual Vectors

It is often important to perform sensitivity analysis to determine how a structural model will be impacted by design changes. Often, the structural analysts will manually make changes to the finite element model (FEM) to determine the effects. But when dealing with a large FEM with millions of degrees of freedom these manual changes can be cumbersome and calculation of the effects can computationally expensive. Therefore, it is desirable to determine the effects of model changes through approximation methods. One common technique is to determine the analytical sensitivity of the FEM model with respect to the given change. These analytical sensitivities are valid when small changes are made to the structural model, but invalid if large changes need to be assessed. Another approach is to use Structural Dynamic Modification (SDM) to create a surrogate model to analyze model changes. SDM is a widely-used sensitivity method and is used in applications of model updating, uncertainty quantification, and model design studies. SMD is valid for moderate (10-20 percent) changes in the structural model, but model approximations are often needed for large parameter changes (greater than 20 percent). Structural Dynamic Modification can be improved by using residual vectors to augment the surrogate model formulation from SDM. Adding the residual modes increases the fidelity of the surrogate model while keeping the computational cost low. This paper discusses the application and limitations of the augmented residual modes method to two structures: the Integrated Spacecraft and Payload Element (ISPE) of the Space Launch System (SLS) and the full SLS as it is configured during its Integrated Modal Test (IMT).

Stewart, Eric C.↗

A Priori Analysis of Flamelet-Based Modeling for a Dual-Mode Scramjet Combustor

An a priori investigation of the applicability of flamelet-based combustion models to dual-mode scramjet combustion was performed utilizing Reynolds-averaged simulations (RAS). For this purpose, the HIFiRE Direct Connect Rig (HDCR) flowpath, fueled with a JP-7 fuel surrogate and operating in dual- and scram-mode was considered. The chemistry of the JP-7 fuel surrogate was modeled using a 22 species, 18-step chemical reaction mechanism. Simulation results were compared to experimentally-obtained, time-averaged, wall pressure measurements to validate the RAS solutions. The analysis of the dual-mode operation of this flowpath showed regions of predominately non-premixed, high-Damkohler number, combustion. Regions of premixed combustion were also present but associated with only a small fraction of the total heat-release in the flow. This is in contrast to the scram-mode operation, where a comparable amount of heat is released from non-premixed and premixed combustion modes. Representative flamelet boundary conditions were estimated by analyzing probability density functions for temperature and pressure for pure fuel and oxidizer conditions. The results of the present study reveal the potential for a flamelet model to accurately model the combustion processes in the HDCR and likely other high-speed flowpaths of engineering interest.

Quinlan, Jesse R.↗

Uncertainty Quantification using Deep Ensembles for Decision Making in Cyber-Physical-Human Systems

In this paper and its companion, Differential Equation Approximation Using Gradient-Boosted Quantile Regression, Robison et al., we examine an approach to quantifying model uncertainty with the aim of increasing the trustworthiness of computational models in human-machine interactions. In Differential Equation Approximation Using Gradient-Boosted Quantile Regression, we focus on gradient-boosted decision trees, while in this one, we give more details about deep ensembles. Uncertainty quantification is crucial for building trustworthy autonomous decision-making agents in human-machine teams. There are two types of uncertainties: aleatoric and epistemic. The former is related to the inherent stochasticity (noise) of the process, whereas the latter is associated with the lack of knowledge or representation capability of models, such as neural networks. By lack of knowledge, we mean the model’s inability to accurately predict outputs for all possible inputs. The aleatory uncertainty can be estimated fairly easily with, for example, filters, whereas epistemic uncertainty is challenging to compute. This paper uses deep ensembles to quantify both aleatory and epistemic uncertainty. It can act as an uncertainty-aware surrogate transition model for decision-making frameworks. "Uncertainty-aware" means that the surrogate transition model should make predictions along with confidence in those predictions. In the context of decision-making, the transition models are ordinary differential equations (ODEs). Since ODEs can be simulated to make one-step or multi-step predictions, a good surrogate model for them should perform reasonably well in both modes. In a multi-step approach, the trajectory sampling method TS∞ was used to propagate uncertainty over multiple steps. The cartpole dynamical system was selected to demonstrate the ability of deep ensembles as good surrogate transition models for decision-making frameworks. The deep ensembles modeled the dynamics of cartpole ODEs and made uncertainty-aware predictions in single-step and multi-step transition modes.

CPH systems↗

X-Ray Diffraction Techniques for a Field Instrument: Patterns of Lithologic Provences

Future exploration of Mars will attempt to shed light on the mineralogy of surface materials. Instruments deployed from remote platforms should have the capability to conduct both intensive analyses as well as rapid, reconnaissance surveys while they function in the martian environment as surrogate geologists. In order to accommodate the reconnaissance mode of analysis and to compensate for analytical limitations imposed by the space-flight conditions, data analysis methods are being developed that will permit interpretation of data by recognition of signatures or "fingerprints". Specifically, we are developing a technique which will allow interpretation of diffraction patterns by recognition of characteristic signatures of different lithologic provences. This technique allows a remote vehicle to function in a rapid-scan mode using the lithologic signature to determine where a more thorough analysis is needed. An x-ray diffraction pattern is characterized by the angular positions of diffracted x-rays, x-ray intensity levels and background radiation levels. These elements may be used to identify a generalized x-ray signature. Lithologic signatures are being developed in two ways. A signature is composed using the ideal powder diffraction indices from the mineral assembledge common to a specific lithologic provence. This is then confirmed using a laboratory diffraction pattern of a whole rock powder. Preliminary results comparing the diffraction signatures of the major mineral assembledges common to basalt, carbonate, and evaporite basin deposits indicate that lithologies are differentiable as a "fingerprint". Statistical analyses are being performed to establish the confidence levels of this technique.

Marshall, J.↗

Quasi-geostrophic wave-CISK in an unbounded baroclinic shear

The free-shear problem, a minimal version of baroclinic, quasi-geostrophic wave-CISK is analyzed. The basic state consists of a zonal flow, unbounded above and below, with constant vertical shear and Brunt-Vaisala frequency and zero meridional gradient of the potential vorticity. Convective heating is parameterized in terms of the convergence below an arbitrary level. Consideration of the potential vorticity dynamics of the unstable modes illustrates how heating may act as a dynamical surrogate for potential vorticity gradients. Although highly idealized, the free-shear problem also explains much of the behavior of more general wave-CISK models.

Snyder, Chris↗

Forming Human-Robot Teams Across Time and Space

NASA pushes telerobotics to distances that span the Solar System. At this scale, time of flight for communication is limited by the speed of light, inducing long time delays, narrow bandwidth and the real risk of data disruption. NASA also supports missions where humans are in direct contact with robots during extravehicular activity (EVA), giving a range of zero to hundreds of millions of miles for NASA s definition of "tele". . Another temporal variable is mission phasing. NASA missions are now being considered that combine early robotic phases with later human arrival, then transition back to robot only operations. Robots can preposition, scout, sample or construct in advance of human teammates, transition to assistant roles when the crew are present, and then become care-takers when the crew returns to Earth. This paper will describe advances in robot safety and command interaction approaches developed to form effective human-robot teams, overcoming challenges of time delay and adapting as the team transitions from robot only to robots and crew. The work is predicated on the idea that when robots are alone in space, they are still part of a human-robot team acting as surrogates for people back on Earth or in other distant locations. Software, interaction modes and control methods will be described that can operate robots in all these conditions. A novel control mode for operating robots across time delay was developed using a graphical simulation on the human side of the communication, allowing a remote supervisor to drive and command a robot in simulation with no time delay, then monitor progress of the actual robot as data returns from the round trip to and from the robot. Since the robot must be responsible for safety out to at least the round trip time period, the authors developed a multi layer safety system able to detect and protect the robot and people in its workspace. This safety system is also running when humans are in direct contact with the robot, so it involves both internal fault detection as well as force sensing for unintended external contacts. The designs for the supervisory command mode and the redundant safety system will be described. Specific implementations were developed and test results will be reported. Experiments were conducted using terrestrial analogs for deep space missions, where time delays were artificially added to emulate the longer distances found in space.

Hambuchen, Kimberly↗

Toward Transition Modeling in a Hypersonic Boundary Layer at Flight Conditions

An accurate physics-based transition prediction method integrated with computational fluid dynamics (CFD) solvers is pursued for hypersonic boundary layer flows over slender hypersonic vehicles at flight conditions. The geometry and flow conditions are selected to match relevant trajectory locations from the ascent phase of the HIFiRE-1 flight experiment, namely, a 7-degree half-angle cone with 2.5 mm nose radius, freestream Mach numbers in the range of 3.8 – 5.5 and freestream unit Reynolds numbers in the range of 3.3 × 10(exp 6) – 21.4 × 10(exp 6) m(exp -1). Earlier research had shown that the onset of transition during the HIFiRE-1 flight experiment correlated with an amplification factor of N ≈ 13.5 for the planar Mack modes. However, to incorporate the N-factor correlations into a CFD code, we investigate surrogate models for disturbance amplification that avoid the direct computation of stability characteristics. A commonly used approach for low-speed flows is based on an a priori database of stability characteristics for locally similar profiles. However, the results presented in this paper demonstrate that the application of this approach to hypersonic boundary layers over blunt spherical nose-tip cones leads to large, unacceptable errors in the predictions of amplification factors, mainly due to its failure in accounting for the effects of the entropy layer on the boundary-layer profiles along the length of the model. We propose and demonstrate an alternate approach that employs the stability computations for a canonical set of blunt cone configurations to train a physics-informed convolutional neural network model that is shown to provide substantially improved transition predictions for hypersonic flow configurations with entropy-layer effects. Furthermore, the excellent performance of the neural network model is also confirmed for cone configurations with nose radius and half-angle values that do not correspond to those used to build the database. Finally, the convolutional neural network model is shown to outperform the linear stability calculations for underresolved basic states.

Pedro Paredes↗

Diagnosis of Magnetic Structures and Intermittency in Space Plasma Turbulence using the Method of Surrogate Data

Several observations in space plasmas have reported the presence of coherent structures at different plasma scales. Structure formation is believed to be a direct consequence of nonlinear interactions between the plasma modes, which depend strongly on phase synchronization of those modes. Despite this important role of the phases in turbulence, very limited work has been however devoted to study the phases as a potential tracers of nonlinearities in comparison with the wealth of literature on power spectra of turbulence where phases are totally missed. We present a method based on surrogate data to systematically detect coherent structures in turbulent signals. The new method has been applied successfully to magnetosheath turbulence (Sahraoui, Phys. Rev. E, 2008, in press), where the relationship between the identified phase coherence and intermittency (classically identified as non Gaussian tails of the PDFs) as well as the energy cascade has been studied. Here we review the main results obtained in that study and show further applications to small scale solar wind turbulence. Implications of the results on theoretical modelling of space turbulence (applicability of weak/wave turbulence, its validity limits and its connection to intermittency) will be discussed.

Sahraoui, Fouad↗

HIFIRE Flight 2 Overview and Status Update 2011

A collaborative international effort, the Hypersonic International Flight Research Experimentation (HIFiRE) Program aims to study basic hypersonic phenomena through flight experimentation. HIFiRE Flight 2 teams the United States Air Force Research Lab (AFRL), NASA, and the Australian Defence Science and Technology Organisation (DSTO). Flight 2 will develop an alternative test technique for acquiring high enthalpy scramjet flight test data, allowing exploration of accelerating hydrocarbon-fueled scramjet performance and dual-to-scram mode transition up to and beyond Mach 8 flight. The generic scramjet flowpath is research quality and the test fuel is a simple surrogate for an endothermically cracked liquid hydrocarbon fuel. HIFiRE Flight 2 will be a first of its kind in contribution to scramjets. The HIFiRE program builds upon the HyShot and HYCAUSE programs and aims to leverage the low-cost flight test technique developed in those programs. It will explore suppressed trajectories of a sounding rocket propelled test article and their utility in studying ramjet-scramjet mode transition and flame extinction limits research. This paper describes the overall scramjet flight test experiment mission goals and objectives, flight test approach and strategy, ground test and analysis summary, development status and project schedule. A successful launch and operation will present to the scramjet community valuable flight test data in addition to a new tool, and vehicle, with which to explore high enthalpy scramjet technologies.

Jackson, Kevin R.↗

Supervised Remote Robot with Guided Autonomy and Teleoperation (SURROGATE): A Framework for Whole-Body Manipulation

The use of the cognitive capabilties of humans to help guide the autonomy of robotics platforms in what is typically called "supervised-autonomy" is becoming more commonplace in robotics research. The work discussed in this paper presents an approach to a human-in-the-loop mode of robot operation that integrates high level human cognition and commanding with the intelligence and processing power of autonomous systems. Our framework for a "Supervised Remote Robot with Guided Autonomy and Teleoperation" (SURROGATE) is demonstrated on a robotic platform consisting of a pan-tilt perception head, two 7-DOF arms connected by a single 7-DOF torso, mounted on a tracked-wheel base. We present an architecture that allows high-level supervisory commands and intents to be specified by a user that are then interpreted by the robotic system to perform whole body manipulation tasks autonomously. We use a concept of "behaviors" to chain together sequences of "actions" for the robot to perform which is then executed real time.

Hebert, Paul↗

Multiple periodicities in the solar magnetic field - Possible origin in a multiple-mode solar dynamo

The solar magnetic field is generated in an oscillatory mode with a 22 yr full period and gives rise to the 11 yr sunspot cycle. However, analyses of contemporary solar records, as well as other surrogate indicators of solar activity, suggest the presence also of longer term periodicities in the solar magnetic cycle. This paper suggests that the solar dynamo can operate in a multiply periodic state, with several periodicites being generated simultaneously at different depths in the convection zone. A simple two-layer model of the solar convection zone is used to illustrate the physical mechanism of spatially localized, multiple-periodicity-mode dynamo regeneration. The two layers are characterized by differences in their respective turbulent magnetic diffusivities. Although the magnetic modes interact with one another, each mode is produced large in one layer or the other, and has an oscillation period approximately equal to the time characteristic of magnetic diffusion across the layer. The observed complicated periodicity pattern in the solar magnetic field could be a combination of two (or more) dynamo modes generated in this manner. The calculations are carried out using a differential rotation model consistent with recent helioseismological measurements, illustrating the challenge to dynamo theory raised by those observational results.

Boyer, D. W.↗

Time Resolved Visualization of Liquid Jet Interaction With H2-Air Rotating Detonations Using MHz Rate Diesel PLIF

The characterization of the dynamic response of liquid jets to transient detonation wave passage is critical for optimization and modeling of liquid fueled rotating detonation combustors. In this work, a rotating detonation combustor (RDC) is operated on hydrogen and air to sustain stable detonation waves that acts as a detonation driver and interact in a one-way coupled manner with a single liquid fuel jet that propagates into the combustion chamber with cycle periods of ~ 250 μs. Diesel is utilized as a realistic fuel surrogate with higher aromatic compounds to enable fluorescence excitation, using the 355 nm thirdharmonic output of a burst-mode Nd:YAG laser, imaged at rates up to 1 MHz. By optimizing the technique to accommodate orders of magnitude variations in the fuel density throughout the injection process, the PLIF data enable quantitative measurements including the refill time, the relative recovery between liquid and gaseous jets and jet trajectory across various momentum flux ratio. As the passage of the detonation wave imparts significant changes in the momentum flux ratio, the qualitative liquid break-up process and spatial distribution varies significantly in time. As the injection system recovers ~ 70% of the cycle period and return to a quasi-steady position and allow comparisons with theoretical jet trajectories. These data, enabled by ultra-high-speed PLIF imaging, represent some of the first detailed measurements for quantifying the dynamic response and recovery of liquid jets exposed to periodic detonations in an operating RDC.

Propulsion↗

A Priori Analysis of a Compressible Flamelet Model using RANS Data for a Dual-Mode Scramjet Combustor

In an effort to make large eddy simulation of hydrocarbon-fueled scramjet combustors more computationally accessible using realistic chemical reaction mechanisms, a compressible flamelet/progress variable (FPV) model was proposed that extends current FPV model formulations to high-speed, compressible flows. Development of this model relied on observations garnered from an a priori analysis of the Reynolds-Averaged Navier-Stokes (RANS) data obtained for the Hypersonic International Flight Research and Experimentation (HI-FiRE) dual-mode scramjet combustor. The RANS data were obtained using a reduced chemical mechanism for the combustion of a JP-7 surrogate and were validated using avail- able experimental data. These RANS data were then post-processed to obtain, in an a priori fashion, the scalar fields corresponding to an FPV-based modeling approach. In the current work, in addition to the proposed compressible flamelet model, a standard incompressible FPV model was also considered. Several candidate progress variables were investigated for their ability to recover static temperature and major and minor product species. The effects of pressure and temperature on the tabulated progress variable source term were characterized, and model coupling terms embedded in the Reynolds- averaged Navier-Stokes equations were studied. Finally, results for the novel compressible flamelet/progress variable model were presented to demonstrate the improvement attained by modeling the effects of pressure and flamelet boundary conditions on the combustion.

Quinlan, Jesse R.↗

Investigation of Liquid Fuel Injection Processes in a Rotating Detonation Combustor using Megahertz Planar Laser-Induced Fluorescence Imaging

Direct imaging of liquid fuel injection is performed within an optically accessible rotating detonation combustor (RDC) using planar laser-induced fluorescence (PLIF) to spatio-temporally resolve the highly dynamic spray characteristics at rates up to 1 MHz. The RDC is operated on air and hydrogen to sustain a stable, cyclically propagating detonation wave with cycle periods of up to ∼250 μs. One of the hydrogen fuel injection sites is replaced with a liquid jet to introduce a diesel spray as a fuel surrogate to enable tracer-free fluorescence excitation using the 355 nm third-harmonic output of a burst-mode Nd:YAG laser. The time-resolved PLIF measurements reveal the evolution of the highly unsteady spray, including the dwell period after the arrival of the detonation wave, the jet recovery, breakup and entrainment into the supersonic crossflow of air, propagation into the detonation chamber, and interaction with the detonation wave. Following passage of the detonation wave, the spray characteristics reveal significant changes in the momentum flux ratio between the liquid and air streams, altering the jet trajectory and temporarily halting fuel delivery to the detonation channel. This temporary cessation of fuel spray injection into the combustor is quantified, along with the fill rate as a function of time. As the injection system recovers, the fuel spray eventually returns to a quasi-steady position prior to the arrival of the detonation wave, allowing qualitative comparisons with theoretical jet trajectories for a range of air mass-flux conditions. These data, enabled by ultra-high-speed PLIF imaging, represent some of the first detailed measurements for characterizing and quantifying the interactions of liquid jets and detonations in an operating RDC.

Propulsion↗

Investigation of Liquid Fuel Refill Dynamics in a Rotating Detonation Combustor using Megahertz Planar Laser-Induced Fluorescence

Direct imaging of liquid fuel injection is performed within an optically accessible rotating detonation combustor (RDC) using planar laser-induced fluorescence (PLIF) to spatio-temporally resolve the highly dynamic spray characteristics at rates up to 1 MHz. The RDC is operated on air and hydrogen to sustain a stable, cyclically propagating detonation wave with cycle periods of up to ∼250 μs. One of the hydrogen fuel injection sites is replaced with a liquid jet to introduce a diesel spray as a fuel surrogate to enable tracer-free fluorescence excitation using the 355 nm third-harmonic output of a burst-mode Nd:YAG laser. The time-resolved PLIF measurements reveal the evolution of the highly unsteady spray, including the dwell period after the arrival of the detonation wave, the jet recovery, breakup and entrainment into the supersonic crossflow of air, propagation into the detonation chamber, and interaction with the detonation wave. Following passage of the detonation wave, the spray characteristics reveal significant changes in the momentum flux ratio between the liquid and air streams, altering the jet trajectory and temporarily halting fuel delivery to the detonation channel. This temporary cessation of fuel spray injection into the combustor is quantified, along with the fill rate as a function of time. As the injection system recovers, the fuel spray eventually returns to a quasi-steady position prior to the arrival of the detonation wave, allowing qualitative comparisons with theoretical jet trajectories for a range of air mass-flux conditions. These data, enabled by ultra-high-speed PLIF imaging, represent some of the first detailed measurements for characterizing and quantifying the interactions of liquid jets and detonations in an operating RDC.

Propulsion↗

Time Resolved Visualization of A Liquid Jet in an RDE Using MHz Rate Diesel PLIF

The characterization of the dynamic response of liquid jets to transient detonation wave passage is critical for optimization and modeling of liquid fueled rotating detonation combustors. In this work, a rotating detonation combustor (RDC) is operated on hydrogen and air to sustain stable detonation waves that acts as a detonation driver and interact in a one-way coupled manner with a single liquid fuel jet that propagates into the combustion chamber with cycle periods of ~ 250 μs. Diesel is utilized as a realistic fuel surrogate with higher aromatic compounds to enable fluorescence excitation, using the 355 nm third harmonic output of a burst-mode Nd:YAG laser, imaged at rates up to 1 MHz. By optimizing the technique to accommodate orders of magnitude variations in the fuel density throughout the injection process, the PLIF data enable quantitative measurements including the refill time, the relative recovery between liquid and gaseous jets and jet trajectory across various momentum flux ratio. As the passage of the detonation wave imparts significant changes in the momentum flux ratio, the qualitative liquid break-up process and spatial distribution varies significantly in time. As the injection system recovers ~ 70% of the cycle period and return to a quasi-steady position and allow comparisons with theoretical jet trajectories. These data, enabled by ultra-high-speed PLIF imaging, represent some of the first detailed measurements for quantifying the dynamic response and recovery of liquid jets exposed to periodic detonations in an operating RDC.

Propulsion↗

Simulation of the Aerosol Size Distribution Using a Neural Network Surrogate for the Modal Aerosol Module (MAM7)

One objective of atmospheric simulations is to quantify the distribution of aerosols and their properties. Accurate parameterizations of the processes governing aerosol mass, particle number, and particle size distribution are important for predicting the Earth’s net radiative balance and aerosol-cloud interactions. The Modal Aerosol Module (MAM7) is a two-moment aerosol model that simulates mass, number, and size distribution of seven modes comprised of internally mixed aerosol species. The two-moment scheme adds significant computational expense but allows for the prediction of varying particle size distribution relative to the bulk method which predicts only total mass. In this work, we developed a neural network surrogate model for MAM7 (MAMnet) to predict the aerosol number concentration in NASA’s Global Earth Observing System (GEOS) without adding prohibitive computational expense. MAMnet, can be driven by output from a single moment, mass-based, aerosol scheme (Goddard Chemistry Aerosol and Radiation model (GOCART)) or from reanalysis products (Modern-Era Retrospective analysis for Research and Applications, Version 2 (MERRA-2)). MAMnet was trained using number concentrations from a 5-year GEOS/MAM7 simulation at 1-degree horizontal resolution and using the total mass calculated across modes as inputs, as well as temperature and air density. The model architecture for MAMnet was based on AlexNet, the 2012 winner of the ImageNet Large Scale Visual Recognition Challenge. While some modifications were necessary to accommodate our problem, important aspects of the network were preserved. MAMnet was able to reproduce zonal dynamics and spatial distributions of the aerosol number concentration however predictability in the upper troposphere was poor.

Katherine H Breen↗