Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “Ground truth”

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 685 records · Page 38
All sourcesNASA NTRSDOE OSTI
DOE OSTI2023

SNM Radiation Signature Classification Using Different Semi-Supervised Machine Learning Models

The timely detection of special nuclear material (SNM) transfers between nuclear facilities is an important monitoring objective in nuclear nonproliferation. Persistent monitoring enabled by successful detection and characterization of radiological material movements could greatly enhance the nuclear nonproliferation mission in a range of applications. Supervised machine learning can be used to signal detections when material is present if a model is trained on sufficient volumes of labeled measurements. However, the nuclear monitoring data needed to train robust machine learning models can be costly to label since radiation spectra may require strict scrutiny for characterization. Therefore, this work investigates the application of semi-supervised learning to utilize both labeled and unlabeled data. As a demonstration experiment, radiation measurements from sodium iodide (NaI) detectors are provided by the Multi-Informatics for Nuclear Operating Scenarios (MINOS) venture at Oak Ridge National Laboratory (ORNL) as sample data. Anomalous measurements are identified using a method of statistical hypothesis testing. After background estimation, an energy-dependent spectroscopic analysis is used to characterize an anomaly based on its radiation signatures. In the absence of ground-truth information, a labeling heuristic provides data necessary for training and testing machine learning models. Supervised logistic regression serves as a baseline to compare three semi-supervised machine learning models: co-training, label propagation, and a convolutional neural network (CNN). In each case, the semi-supervised models outperform logistic regression, suggesting that unlabeled data can be valuable when training and demonstrating value in semi-supervised nonproliferation implementations.

98 NUCLEAR DISARMAMENT, SAFEGUARDS, AND PHYSICAL P↗
DOE OSTI2023

A Data-Driven Framework for Direct Local Tensile Property Prediction of Laser Powder Bed Fusion Parts

This article proposes a generalizable, data-driven framework for qualifying laser powder bed fusion additively manufactured parts using part-specific in situ data, including powder bed imaging, machine health sensors, and laser scan paths. To achieve part qualification without relying solely on statistical processes or feedstock control, a sequence of machine learning models was trained on 6299 tensile specimens to locally predict the tensile properties of stainless-steel parts based on fused multi-modal in situ sensor data and a priori information. A cyberphysical infrastructure enabled the robust spatial tracking of individual specimens, and computer vision techniques registered the ground truth tensile measurements to the in situ data. The co-registered 230 GB dataset used in this work has been publicly released and is available as a set of HDF5 files. The extensive training data requirements and wide range of size scales were addressed by combining deep learning, machine learning, and feature engineering algorithms in a relay. The trained models demonstrated a 61% error reduction in ultimate tensile strength predictions relative to estimates made without any in situ information. Lessons learned and potential improvements to the sensors and mechanical testing procedure are discussed.

36 MATERIALS SCIENCE↗
DOE OSTI2021

Open Data and Deep Semantic Segmentation for Automated Extraction of Building Footprints

Advances in machine learning and computer vision, combined with increased access to unstructured data (e.g., images and text), have created an opportunity for automated extraction of building characteristics, cost-effectively, and at scale. These characteristics are relevant to a variety of urban and energy applications, yet are time consuming and costly to acquire with today’s manual methods. Several recent research studies have shown that in comparison to more traditional methods that are based on features engineering approach, an end-to-end learning approach based on deep learning algorithms significantly improved the accuracy of automatic building footprint extraction from remote sensing images. However, these studies used limited benchmark datasets that have been carefully curated and labeled. How the accuracy of these deep learning-based approach holds when using less curated training data has not received enough attention. The aim of this work is to leverage the openly available data to automatically generate a larger training dataset with more variability in term of regions and type of cities, which can be used to build more accurate deep learning models. In contrast to most benchmark datasets, the gathered data have not been manually curated. Thus, the training dataset is not perfectly clean in terms of remote sensing images exactly matching the ground truth building’s foot-print. A workflow that includes data pre-processing, deep learning semantic segmentation modeling, and results post-processing is introduced and applied to a dataset that include remote sensing images from 15 cities and five counties from various region of the USA, which include 8,607,677 buildings. The accuracy of the proposed approach was measured on an out of sample testing dataset corresponding to 364,000 buildings from three USA cities. The results favorably compared to those obtained from Microsoft’s recently released US building footprint dataset.

97 MATHEMATICS AND COMPUTING↗
DOE OSTI2022

Implementing Spatio-Temporal 3D-Convolution Neural Networks and UAV Time Series Imagery to Better Predict Lodging Damage in Sorghum

Unmanned aerial vehicle (UAV)-based remote sensing is gaining momentum in a variety of agricultural and environmental applications. Very-high-resolution remote sensing image sets collected repeatedly throughout a crop growing season are becoming increasingly common. Analytical methods able to learn from both spatial and time dimensions of the data may allow for an improved estimation of crop traits, as well as the effects of genetics and the environment on these traits. Multispectral and geometric time series imagery was collected by UAV on 11 dates, along with ground-truth data, in a field trial of 866 genetically diverse biomass sorghum accessions. We compared the performance of Convolution Neural Network (CNN) architectures that used image data from single dates (two spatial dimensions, 2D) versus multiple dates (two spatial dimensions + temporal dimension, 3D) to estimate lodging detection and severity. Lodging was detected with 3D-CNN analysis of time series imagery with 0.88 accuracy, 0.92 Precision, and 0.83 Recall. This outperformed the best 2D-CNN on a single date with 0.85 accuracy, 0.84 Precision, and 0.76 Recall. The variation in lodging severity was estimated by the best 3D-CNN analysis with 9.4% mean absolute error (MAE), 11.9% root mean square error (RMSE), and goodness-of-fit (R2) of 0.76. This was a significant improvement over the best 2D-CNN analysis with 11.84% MAE, 14.91% RMSE, and 0.63 R2. The success of the improved 3D-CNN analysis approach depended on the inclusion of “before and after” data, i.e., images collected on dates before and after the lodging event. The integration of geometric and spectral features with 3D-CNN architecture was also key to the improved assessment of lodging severity, which is an important and difficult-to-assess phenomenon in bioenergy feedstocks such as biomass sorghum. This demonstrates that spatio-temporal CNN architectures based on UAV time series imagery have significant potential to enhance plant phenotyping capabilities in crop breeding and Precision agriculture applications.

3D-convolution neural networks↗
DOE OSTI2022

Deep Convolutional Neural Networks Exploit High-Spatial- and -Temporal-Resolution Aerial Imagery to Phenotype Key Traits in Miscanthus

Miscanthus is one of the most promising perennial crops for bioenergy production, with high yield potential and a low environmental footprint. The increasing interest in this crop requires accelerated selection and the development of new screening techniques. New analytical methods that are more accurate and less labor-intensive are needed to better characterize the effects of genetics and the environment on key traits under field conditions. We used persistent multispectral and photogrammetric UAV time-series imagery collected 10 times over the season, together with ground-truth data for thousands of Miscanthus genotypes, to determine the flowering time, culm length, and biomass yield traits. We compared the performance of convolutional neural network (CNN) architectures that used image data from single dates (2D-spatial) versus the integration of multiple dates by 3D-spatiotemporal architectures. The ability of UAV-based remote sensing to rapidly and non-destructively assess large-scale genetic variation in flowering time, height, and biomass production was improved through the use of 3D-spatiotemporal CNN architectures versus 2D-spatial CNN architectures. The performance gains of the best 3D-spatiotemporal analyses compared to the best 2D-spatial architectures manifested in up to 23% improvements in R2, 17% reductions in RMSE, and 20% reductions in MAE. The integration of photogrammetric and spectral features with 3D architectures was crucial to the improved assessment of all traits. In conclusion, our findings demonstrate that the integration of high-spatiotemporal-resolution UAV imagery with 3D-CNNs enables more accurate monitoring of the dynamics of key phenological and yield-related crop traits. This is especially valuable in highly productive, perennial grass crops such as Miscanthus, where in-field phenotyping is especially challenging and traditionally limits the rate of crop improvement through breeding.

47 OTHER INSTRUMENTATION↗
DOE OSTI2024

Mapping Geothermal Indicator Minerals Using Fusion of Target Detection Algorithms

Mineral mapping from satellite images provides valuable insights into subsurface mineral alteration for geothermal exploration. In previous studies, eight fundamental algorithms were used for mineral mapping utilizing USGS spectra, a collection of reflectance spectra containing samples of minerals, rocks, and soils created by the USGS. We used an ASD FieldSpec 4 Hi-RES NG portable spectrometer to collect spectra for analyzing ASTER images of the Coso Geothermal Field. Then, we established the ground-truth information and the spectral library by analyzing 97 samples. Samples collected from the field were analyzed using the CSIRO TSG (The Spectral Geologist of the Commonwealth Scientific and Industrial Research Organization). Based on the mineralogy study, multiple high-purity spectra of geothermal alteration minerals were selected from collected data, including alunite, chalcedony, hematite, kaolinite, and opal. Eight mineral spectral target detection algorithms were applied to the preprocessed satellite data with a proposed local spectral library. We measured the highest overall accuracy of 87% for alunite, 95% for opal, 83% for chalcedony, 60% for hematite, and 96% for kaolinite out of these eight algorithms. Three, four, five, and eight algorithms were fused to extract mineral alteration with the obtained target detection results. The results prove that the fusion of algorithms gives better results than using individual ones. In conclusion, this paper discusses the significance of evaluating different mapping algorithms. It proposes a robust fusion approach to extract mineral maps as an indicator for geothermal exploration.

Cavur, Mahmut↗
DOE OSTI2025

Acoustic Tomography of the Atmosphere: A Large-Eddy Simulation Sensitivity Study

Accurate measurement of atmospheric turbulent fluctuations is critical for understanding environmental dynamics and improving models in applications such as wind energy. Advanced remote sensing technologies are essential for capturing instantaneous velocity and temperature fluctuations. Acoustic tomography (AT) offers a promising approach that utilizes sound travel times between an array of transducers to reconstruct turbulence fields. This study presents a systematic evaluation of the time-dependent stochastic inversion (TDSI) algorithm for AT using synthetic travel-time measurements derived from large-eddy simulation (LES) fields under both neutral and convective atmospheric boundary-layer conditions. Unlike prior work that relied on field observations or idealized fields, the LES framework provides a ground-truth atmospheric state, enabling quantitative assessment of TDSI retrieval reliability, sensitivity to travel-time measurement noise, and dependence on covariance model parameters and temporal data integration. A detailed sensitivity analysis was conducted to determine the best-fit model parameters, identify the tolerance thresholds for parameter mismatch, and establish a maximum spatial resolution. The TDSI algorithm successfully reconstructed large-scale velocity and temperature fluctuations with root mean square errors ( RMSE s) below 0.35 m/s and 0.12 K, respectively. Spectral analysis established a maximum spatial resolution of approximately 1.4 m, and reconstructions remained robust for travel-time measurement uncertainties up to 0.002 s. These findings provide critical insights into the operational limits of TDSI and inform future applications of AT for atmospheric turbulence characterization and system design.

17 WIND ENERGY↗
DOE OSTI2023

A Field Study of Nonintrusive Load Monitoring Devices and Implications for Load Disaggregation

Evaluations of nonintrusive load monitoring (NILM) algorithms and technologies have mostly occurred in constrained, artificial environments. However, few field evaluations of NILM products have taken place in actual buildings under normal operating conditions. This paper describes a field evaluation of a state-of-the-art NILM product, tested in eight homes. The match rate metric—a technique recommended by a technical advisory group—was used to measure the NILM’s success in identifying specific loads and the accuracy of the energy consumption estimates. A performance assessment protocol was also developed to address common issues with NILM mislabeling and ground-truth comparisons that have not been sufficiently addressed in past evaluations. The NILM product’s estimates were compared to the submetered consumption of eight major appliances. Overall, the product had good performance in disaggregating the energy consumption of the electric water heaters, which included both electric resistance and heat-pump water heaters, but only a fair accuracy with refrigerators, dryers, and air conditioners. The performance was poor for cooking equipment, furnace fans, clothes washers, and dishwashers. Moreover, the product was often unable to detect major loads in homes. Typically, two or more appliances were not detected in a home. At least two dryers, furnace fans, and air conditioners went undetected across the eight homes. On the other hand, the dishwasher was detected in all homes where available or monitored. The key findings were qualitatively compared to those of past field evaluations. Potential areas for improvement in NILM product performance were determined along with areas where complementary technologies may be able to aid in load-disaggregation applications.

47 OTHER INSTRUMENTATION↗
DOE OSTI2025

Bolide Infrasound Signal Morphology and Yield Estimates: A Case Study of Two Events Detected by a Dense Acoustic Sensor Network

Two bolides (2016 June 2 and 2019 April 4) were detected at multiple regional infrasound stations, with many of the locations receiving multiple detections. Analysis of the received signals was used to estimate the yield, location, and trajectory, as well as the type of shock that produced the received signal. The results from the infrasound analysis were compared with ground-truth information that was collected through other sensing modalities. This multimodal framework offers an expanded perspective on the processes governing bolide shock generation and propagation. The majority of signal features showed reasonable agreement between the infrasound-based interpretation and the other observational modalities, though the yield estimate from the 2019 bolide was significantly lower using the infrasound detections. There was also evidence suggesting that one of the detections was from a cylindrical shock that was initially propagating upward, which is unusual though not impossible.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2020

Inpainting Galactic Foreground Intensity and Polarization Maps Using Convolutional Neural Networks

The Deep Convolutional Neural Networks (DCNNs) have been a popular tool for image generation and restoration. In this work, we applied DCNNs to the problem of inpainting non-Gaussian astrophysical signal, in the context of Galactic diffuse emissions at the millimetric and submillimetric regimes, specifically Synchrotron and Thermal Dust emissions. Both signals are affected by contamination at small angular scales due to extragalactic radio sources (the former) and dusty star-forming galaxies (the latter). Furthermore, we compare the performance of the standard diffusive inpainting with that of two novel methodologies relying on DCNNs, namely Generative Adversarial Networks and Deep-Prior. We show that the methods based on the DCNNs are able to reproduce the statistical properties of the ground-truth signal more consistently with a higher confidence level.

79 ASTRONOMY AND ASTROPHYSICS↗
DOE OSTI2022

LensingETC: A Tool to Optimize Multifilter Imaging Campaigns of Galaxy-scale Strong Lensing Systems

Imaging data is the principal observable required to use galaxy-scale strong lensing in a multitude of applications in extragalactic astrophysics and cosmology. In this paper, we develop Lensing Exposure Time Calculator (LensingETC; to optimize the efficiency of telescope-time usage when planning multifilter imaging campaigns for galaxy-scale strong lenses. This tool simulates realistic data tailored to specified instrument characteristics and then automatically models them to assess the power of the data in constraining lens model parameters. We demonstrate a use case of this tool by optimizing a two-filter observing strategy (in the IR and ultraviolet-visual (UVIS)) within the limited exposure time per system allowed by a Hubble Space Telescope (HST) Snapshot program. We find that higher resolution is more advantageous to gain constraining power on the lensing observables, when there is a trade-off between signal-to-noise ratio and resolution; for example, between the UVIS and IR filters of the HST. We also find that, whereas a point-spread function (PSF) with sub-Nyquist sampling allows the sample mean for a model parameter to be robustly recovered for both galaxy–galaxy and point-source lensing systems, a sub-Nyquist-sampled PSF introduces a larger scatter than a Nyquist-sampled one in the deviation from the ground truth for point-source lens systems.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗
DOE OSTI2025

Leveraging Open-Source Satellite-Derived Building Footprints for Height Inference

At a global scale, cities are growing and characterizing the built environment is essential for deeper understanding of human population patterns, urban development, energy usage, climate change impacts, among others. Buildings are a key component of the built environment and significant progress has been made in recent years to scale building footprint extractions from satellite datum and other remotely sensed products. Billions of building footprints have recently been released by companies such as Microsoft and Google at a global scale. However, research has shown that depending on the methods leveraged to produce a footprint dataset, discrepancies can arise in both the number and shape of footprints produced. Therefore, each footprint dataset should be examined and used on a case-by-case study. In this work, we find through two experiments on Oak Ridge National Laboratory and Microsoft footprints within the same geographic extent that our approach of inferring height from footprint morphology features is source agnostic. Regardless of the differences associated with the methods used to produce a building footprint dataset, our approach of inferring height was able to overcome these discrepancies between the products and generalize, as evidenced by 98% of our results being within 3m of the ground-truthed height. This signifies that our approach can be applied to the billions of open-source footprints which are freely available to infer height, a key building metric. This work impacts the broader domain of urban science in which building height is a key, and limiting factor.

Stipek, Clinton [ORNL] (ORCID:0000000280501096)↗
DOE OSTI2023

Development of a Heavy-Duty Electric Vehicle Integration and Implementation (HEVII) Tool

As demand for consumer electric vehicles (EVs) has drastically increased in recent years, manufacturers have been working to bring heavy-duty EVs to market to compete with Class 6-8 diesel-powered trucks. Many high-profile companies have committed to begin electrifying their fleet operations, but have yet to implement EVs at scale due to their limited range, long charging times, sparse charging infrastructure, and lack of data from in-use operation. Thus far, EVs have been disproportionately implemented by larger fleets with more resources. To aid fleet operators, it is imperative to develop tools to evaluate the electrification potential of heavy-duty fleets. However, commercially available tools, designed mostly for light-duty vehicles, are inadequate for making electrification recommendations tailored to a fleet of heavy-duty vehicles. The main challenge is that light-duty tools do not estimate real-time vehicle mass, a factor that has a disproportionate impact on the energy consumption of large commercial vehicles. The Heavy-Duty Electric Vehicle Integration and Implementation (HEVII) tool advances the state of the art in evaluating electrification potential and infrastructure requirements for fleets of commercial vehicles. In this work, the HEVII tool is demonstrated with non-uniformly sampled telematics data from an existing fleet to assess the suitability for electrification of each individual vehicle, determine optimal locations for charging infrastructure to support a fleet of EVs and analyze associated costs. Payload mass is predicted using sparse ground-truth data for all input drive cycles and an initial data analysis is conducted to assess the characteristics driving behaviors and energy consumption of the fleet using an adaptable vehicle model. Battery size requirements are determined by applying a novel charger placement algorithm to maximize routes that are viable for EVs and balance time delays with infrastructure development costs. This work details and demonstrates the different aspects of the HEVII tool, presenting preliminary results from an example use case.

ADVANCED PROPULSION SYSTEMS↗
DOE OSTI2022

PRMI: A Dataset of Minirhizotron Images for Diverse Plant Root Study

Understanding a plant's root system architecture (RSA) is crucial for a variety of plant science problem domains including sustainability and climate adaptation. Minirhizotron (MR) technology is a widely-used approach for phenotyping RSA non-destructively by capturing root imagery over time. Precisely segmenting roots from the soil in MR imagery is a critical step in studying RSA features. In this paper, we introduce a large-scale dataset of plant root images captured by MR technology. In total, there are over 72K RGB root images across six different species including cotton, papaya, peanut, sesame, sunflower, and switchgrass in the dataset. The images span a variety of conditions including varied root age, root structures, soil types, and depths under the soil surface. All of the images have been annotated with weak image-level labels indicating whether each image contains roots or not. The image-level labels can be used to support weakly supervised learning in plant root segmentation tasks. In addition, 63K images have been manually annotated to generate pixel-level binary masks indicating whether each pixel corresponds to root or not. These pixel-level binary masks can be used as ground truth for supervised learning in semantic segmentation tasks. By introducing this dataset, we aim to facilitate the automatic segmentation of roots and the research of RSA with deep learning and other image analysis algorithms.

Xu, Weihuang↗
DOE OSTI2025

Classifying thermodynamic cloud phase using machine learning models

Vertically resolved thermodynamic cloud-phase classifications are essential for studies of atmospheric cloud and precipitation processes. The Department of Energy (DOE) Atmospheric Radiation Measurement (ARM) Thermodynamic Cloud Phase (THERMOCLDPHASE) value-added product (VAP) uses a multi-sensor approach to classify the thermodynamic cloud phase by combining lidar backscatter and depolarization, radar reflectivity, Doppler velocity, spectral width, microwave-radiometer-derived liquid water path, and radiosonde temperature measurements. The measured pixels are classified as ice, snow, mixed phase, liquid (cloud water), drizzle, rain, and liq_driz (liquid+drizzle). We use this product as the ground truth to train three machine learning (ML) models to predict the thermodynamic cloud phase from multi-sensor remote sensing measurements taken at the ARM North Slope of Alaska (NSA) observatory: a random forest (RF), a multi-layer perceptron (MLP), and a convolutional neural network (CNN) with a U-Net architecture. Evaluations against the outputs of the THERMOCLDPHASE VAP with 1 year of data show that the CNN outperforms the other two models, achieving the highest test accuracy, F1 score, and mean intersection over union (IOU). Analysis of ML confidence scores shows that ice, rain, and snow have higher confidence scores, followed by liquid, while mixed, drizzle, and liq_driz have lower scores. Feature importance analysis reveals that the mean Doppler velocity and vertically resolved temperature are the most influential data streams for ML thermodynamic cloud-phase predictions. Lidar measurements exhibit lower feature importance due to rapid signal attenuation caused by the frequent presence of persistent low-level clouds at the NSA site. The ML models' generalization capacity is further evaluated by applying them at another Arctic ARM site in Norway using data taken during the ARM Cold-Air Outbreaks in the Marine Boundary Layer Experiment (COMBLE) field campaign. The models demonstrated similar performance to that observed at the NSA site. Finally, we evaluate the ML models' response to simulated instrument outages and signal degradation and show that a CNN U-Net model trained with input channel dropouts performs better when input fields are missing.

ARM Aerial Facility↗
DOE OSTI2023

CPC data from TAMU TRACER campaign in the Houston, TX region from July to September 2022

During TRACER, the Texas A&M Rapid Onsite Atmospheric Measurements Van (ROAM-V) was deployed to capture airmasses behind (maritime) and ahead (continental) of the passage of the sea-breeze front through Houston. On select sampling days, ROAM-V sampled in the morning/mid-day on the coast and then transited to a second inland site for the afternoon/evening. The suite of instruments deployed on ROAM-V included a Condensation Particle Counter (CPC; GRIMM Model 5.403 CPC), Scanning Mobility Particle Sizer (SMPS; TSI 3750 detector, TSI 3082 classifier, TSI 3088 neutralizer, TSI 3081A Differential Mobility Analyzer), Cloud Condensation Nuclei counter (Droplet Measurement Technologies CCN Counter), micro pulse lidar (Droplet Measurement Technologies Micro Pulse LiDAR (miniMPL)), and a Davis Rotating Uniform size-cut Monitor (DRUM; DRUMAir 4-DRUM). Before sampling at each location, the latitude and longitude were recorded using the GPS on the phone application “My Altitude”. Onboard the ROAM-V, aerosol samples are drawn through a shared isokinetic inlet at a flow rate ranging from 3.5 to 7.0 LPM. A portion of this flow is directed through a cyclone impactor (Brechtel, Inc. Model SCC 0.732) and 0.3 LPM is directed to the CPC. To correct the data for particle losses, we used a two-step method. First, the measured SMPS size distributions were used to calculate total particle loss through the inlet for every SMPS scan at a single deployment location. Second, the CPC data was corrected for particle losses using the average of the total losses per scan. This calculation was done separately for each deployment location due to changes in the measured size distributions between locations. Particle loss from diffusion (based on Kesten, 1991 and Gormley, 1949), inertial impaction in 90-degree bends (based on Aerosol Measurement, 2011 and Crane, 1977), and cyclone impactor efficiency (based on Dirgo, 1985) were included in the loss calculation. When the SMPS was not sampling at a location (in the case of an instrument malfunction or operator error), the reported CPC data was corrected with an average of the total losses for the entire campaign at the specified deployment location (e.g., if we needed to correct Galveston data, then the average of all calculated losses at Galveston was taken). These flatline corrections were used for all data on 22/07/13, 22/07/20, 22/07/22, and the data from Galveston on 22/08/09. An additional flatline correction of 14% was applied to all CPC data based on laboratory calibration prior to and after the campaign. In laboratory calibrations, we identified that for the same sample of air the field CPC (GRIMM Model 5.403 CPC) undercounted the total concentration of particles by 14% compared to the ground truth laboratory CPC (TSI Model 3750). This data was collected for ARM Field Campaign AFC07055 and supported by DOE ASR grant DE-SC0021047. For any further questions, please feel free to contact the instrument PI, Sarah D. Brooks, sbrooks@tamu.edu. Kesten et. al. Calibration of a TSI Model 3025 Ultrafine Condensation Particle Counter. Aerosol Science and Technology, 15:2, 107-111, 1991. Gormley et. al. Diffusion from a Stream Flowing through a Cylindrical Tube. Proceedings of the Royal Irish Academy, Vol 52, 163-169, 1948. Aerosol Measurement: Principles, Techniques, and Applications, Third Edition. John Wiley & Sons, Inc, 2011. Crane et. al. Inertial Deposition of Particles in a Bent Pipe. Journal of Aerosol Science, Vol 8, 161-170, 1977. Dirgo et. al. Cyclone Collection Efficiency: Comparison of Experimental Results with Theoretical Predictions. Aerosol Science and Technology, 4:4, 401-415, 1985.

54 ENVIRONMENTAL SCIENCES↗
DOE OSTI2025

Thermodynamic Cloud Phase Classifications Using Machine Learning at NSA and ANX

Vertically resolved thermodynamic cloud phase classifications are essential for studies of atmospheric cloud and precipitation processes. The Department of Energy (DOE) Atmospheric Radiation Measurement (ARM) THERMOCLDPHASE Value-Added Product (VAP) uses a multi-sensor approach to classify thermodynamic cloud phase by combining lidar backscatter and depolarization, radar reflectivity, Doppler velocity, spectral width, microwave radiometer-derived liquid water path, and radiosonde temperature measurements. The measured voxels are classified as ice, snow, mixed-phase, liquid (cloud water), drizzle, rain, and liq_driz (liquid+drizzle). We use this product as the ground truth to train three machine learning (ML) models to predict the thermodynamic cloud phase from multi-sensor remote sensing measurements taken at the ARM North Slope of Alaska (NSA) observatory: a random forest (RF), a multilayer perceptron (MLP), and a convolutional neural network (CNN) with a U-Net architecture. Evaluations against the outputs of the THERMOCLDPHASE VAP with one year of data show that the CNN outperforms the other two models, achieving the highest test accuracy, F1-score, and mean Intersection over Union (IOU). Analysis of ML confidence scores shows ice, rain, and snow have higher confidence scores, followed by liquid, while mixed, drizzle, and liq_driz have lower scores. Feature importance analysis reveals that the mean Doppler velocity and vertically resolved temperature are the most influential datastreams for ML thermodynamic cloud phase predictions. The ML models’ generalization capacity is further evaluated by applying them at another Arctic ARM site in Norway using data taken during the ARM Cold-Air Outbreaks in the Marine Boundary Layer Experiment (COMBLE) field campaign. Finally, we evaluate the ML models’ response to simulated instrument outages and signal degradation.

54 ENVIRONMENTAL SCIENCES↗