Search NASA⌕ Search

SEARCH · Search NASA

Results for “cloud platform”

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 901 records · Page 50

Science Writer's Guide to Landsat 7

The Earth Observing System (EOS), the centerpiece of NASA's Earth science program, is a suite of spacecraft and interdisciplinary science investigations dedicated to advancing our understanding of global change. The flagship EOS satellite, Terra (formerly EOS AM-1), scheduled for launch in July 1999, will provide key measurements of the physical and radiative properties of clouds; air-land and air-sea exchanges of energy, carbon, and water; trace gases; and volcanoes. Flying in formation with Terra, Landsat 7 will make global high spatial resolution measurements of land surface and surrounding coastal regions. Other upcoming EOS missions and instruments include QuikSCAT, to collect sea surface wind data; the Stratospheric Gas and Aerosol Experiment (SAGE III), to create global profiles of key atmospheric gases; and the Active Cavity Radiometer Irradiance Monitors (ACRIM) to measure the energy output of the Sun. The second of the major, multi-instrument EOS platforms, PM-1, is scheduled for launch in 2000. Interdisciplinary research projects sponsored by EOS use specific Earth science data sets for a broader investigation into the function of Earth systems. Current EOS research spans a wide range of sciences, including atmospheric chemistry, hydrology, land use, and marine ecosystems. The EOS program has been managed since 1990 by the Goddard Space Flight Center in Greenbelt, Md., for NASA's Office of Earth Science in Washington, D. C. Additional information on the program can be found on the EOS Project Science Office Web site (http://eospso.gsfc.nasa.gov).

Source record↗

Assessment of Clear Sky Radiative Forcing in the Caribbean Region Using an Aerosol Dispersion Model and Ground Radiometry During Puerto Rico Dust Experiment

This study investigates the surface and top of the atmosphere solar radiative forcing by long-range transport of Saharan dust. The calculations of radiative forcing are based on measurements collected in the Puerto Rico Dust Experiment (PRIDE) carried out during July, 2000. The purpose of the experiment was the characterization of the Saharan dust plume, which frequently reaches the Caribbean region during the summer. The experiment involved the use of three approaches to study the plume: space and ground based remote sensing, airborne and ground based in-situ measurements and aerosol dispersion modeling. The diversity of measuring platforms provides an excellent opportunity for determination of the direct effect of dust on the clear sky radiative forcing. Specifically, comparisons of heating rates, surface and TOA fluxes derived from the Navy global aerosol dispersion model NAAPS (NRL Aerosol Analysis and Prediction System) and actual measurements of fluxes from ground and space based platforms are shown. In addition, the direct effect of dust on the clear sky radiative forcing is modeled. The extent and time of evolution of the radiative properties of the plume are computed with the aerosol concentrations modeled by NAAPS. Standard aerosol parameterizations, as well as in-situ composition and size distributions measured during PRIDE, are utilized to compute the aerosol optical depth, single scattering albedo and asymmetry factor. Radiative transfer computations are done with an in-house modified spectral radiative transfer code (Fu-Liou). The code includes gas absorption and cloud particles (ice and liquid phase) and it allows the input of meteorological data. The code was modified to include modules for the aerosols contribution to the calculated fluxes. This comparison study helps to narrow the current uncertainty in the dust direct radiative forcing, as recently reported in the 2001 IPCC assessment.

Gasso, Santiago↗

Evolving HPC and Application Design Toward a Coupled Data Assimilation System at NASA Suitable for Emerging Exascale Platforms

The prediction capabilities of global models have continuously evolved from the traditional medium-range global weather prediction application to span scales in support of hourly prediction of convective scale storms to seasonal Earth system prediction. This evolution has increased the demands on the system infrastructure design and workflow to achieve the required performance on modern high-performance computing (HPC) platforms. The planned evolution of the Goddard Earth Observing System (GEOS) modeling and assimilation system will stress the capabilities of conventional HPC overwhelming the available compute cycles at the NASA Center for Climate Simulation (NCCS) at the NASA Goddard Space Flight Center in the coming 5-10 years. This has led to the re-design of key elements of the assimilation and modeling systems to achieve significant gains in performance on anticipated Exacale platforms. The transition of the assimilation system to the Joint Effort for Data assimilation Integration (JEDI) framework has positioned GEOS to exploit new efficient algorithms for data assimilation (DA) in a fully-coupled Earth system context. The suitability of the GEOS model to leverage a domain specific language (DSL) approach and artificial intelligence (AI) is being explored to accelerate computational performance and data exchange efficiency of the coupled Earth system model. The storage and processing of large data volumes produced by these advance systems is being redesigned with a data-centric cloud-based approach. We will highlight the recent efforts in these areas and emphasize the demand for further development and re-design to achieve the science objectives in support of NASA's Earth system modeling and assimilation missions.

Putman, Bill↗

Contributions of Nimbus 7 TOMS Data to Volcanic Study and Hazard Mitigation

Nimbus TOMS data have led to advancements among many volcano-related scientific disciplines, from the initial ability to quantify SO2 clouds leading to derivations of eruptive S budgets and fluxes, to tracking of individual clouds, assessing global volcanism and atmospheric impacts. Some of the major aspects of TOMS-related research, listed below, will be reviewed and updated: (1) Measurement of volcanic SO2 clouds: Nimbus TOMS observed over 100 individual SO2 clouds during its mission lifetime; large explosive eruptions are now routinely and reliably measured by satellite. (2) Eruption processes: quantification of SO2 emissions have allowed assessments of eruption sulfur budgets, the evaluation of "excess" sulfur, and inferences of H2S emissions. (3) Detection of ash: TOMS data are now used to detect volcanic particulates in the atmosphere, providing complementary analyses to infrared methods of detection. Paired TOMS and AVHRR studies have provided invaluable information on volcanic cloud compositions and processes. (4) Cloud tracking and hazard mitigation: volcanic clouds can be considered gigantic tracers in the atmosphere, and studies of the fates of these clouds have led to new knowledge of their physical and chemical dispersion in the atmosphere for predictive models. (5) Global trends: the long term data set has provided researchers an unparalleled record of explosive volcanism, and forms a key component in assessing annual to decadal trends in global S emissions. (6) Atmospheric impacts: TOMS data have been linked to independent records of atmospheric change, in order to compare cause and effect processes following a massive injection of SO2 into the atmosphere. (7) Future TOMS instruments and applications: Nimbus TOMS has given way to new satellite platforms, with several wavelength and resolution modifications. New efforts to launch a geostationary TOMS could provide unprecedented observations of volcanic activity.

Krueger, Arlin J.↗

Embedded Real-Time Linux for Instrument Control and Data Logging

When I moved to the west. coast to take a job at NASA's Ames Research Center in Mountain View, CA, I was impressed with the variety of equipment and software which scientists at the center use to conduct their research. was happy to find that I was just as likely to see a machine running Lenox as one running Windows in the offices and laboratories of NASA Ames (although many people seem to use Moos around here). I was especially happy to find that the particular group with whom I was going to work, the Atmospheric Physics Branch at Ames, relied almost entirely on Lenox machines for their day-to-day work. So it was no surprise that when it was time to construct a new control system for one of their most important pieces of hardware, a switch from an unpredictable DOS-based platform to an Embedded Linux-based one was a decision easily made. The system I am working on is called the Solar Spectral Flux Radiometer (SSFR), a PC-104 based system custom-built by Dr. Warren Gore at Ames. Dr. Gore, Dr. Peter Pilewskie, Dr. Maura Robberies and Larry Pezzolo use the SSFR in their research. The team working on the controller project consists of Dr. Gore, John Pommier, and myself. The SSFR is used by the ,cities Atmospheric Radiation Group to measure solar spectral irradiance at moderate resolution to determine the radiative effect of clouds, aerosols, and gases on climate, and also to infer the physical properties of aerosols and clouds. Two identical SSFR's have been built and successfully deployed in three field missions: 1) the Department of Energy Atmospheric Radiation Measurement (ARM) Enhanced Shortwave Experiment (ARESE) II in February/March, 2000; 2) the Puerto Rico Dust Experiment (PRIDE) in July, 2000; and 3) the South African Regional Science Initiative (SAFARI) in August/September, 2000. Additionally, the SSFR was used to acquire water vapor spectra using the Ames Diameter base-path multiple-reflection absorption cell in a laboratory experiment.

Clanton, Sam↗

An Update on Oceanic Precipitation Rate and its Zonal Distribution in Light of Advanced Observations from Space

This study contributes to the estimation of the global mean and zonal distribution of oceanic precipitation rate using complementary information from advanced precipitation measuring sensors and provides an independent reference to assess current precipitation products. Precipitation estimates from the Tropical Rainfall Measuring Mission (TRMM) precipitation radar (PR) and CloudSat cloud profiling radar (CPR) were merged, as the two complementary sensors yield an unprecedented range of sensitivity to quantify rainfall from drizzle through the most intense rates. At higher latitudes, where TRMM PR does not exist, precipitation estimates from Aqua's Advanced Microwave Scanning Radiometer for Earth Observing System (AMSR-E) complemented CloudSat CPR to capture intense precipitation rates. The high sensitivity of CPR allows estimation of snow rate, an important type of precipitation at high latitudes, not directly observed in current merged precipitation products. Using the merged precipitation estimate from the CloudSat, TRMM, and Aqua platforms (this estimate is abbreviated to MCTA), the authors' estimate for 3-yr (2007-09) nearglobal (80degS-80degN) oceanic mean precipitation rate is approx. 2.94mm/day. This new estimate of mean global ocean precipitation is about 9% higher than that of the corresponding Climate Prediction Center (CPC) Merged Analysis of Precipitation (CMAP) value (2.68mm/day) and about 4% higher than that of the Global Precipitation Climatology Project (GPCP; 2.82mm/day). Furthermore, MCTA suggests distinct differences in the zonal distribution of precipitation rate from that depicted in GPCPand CMAP, especially in the Southern Hemisphere.

profiling radar↗

CLARREO Approach for Reference Intercalibration of Reflected Solar Sensors: On-Orbit Data Matching and Sampling

The implementation of the Climate Absolute Radiance and Refractivity Observatory (CLARREO) mission was recommended by the National Research Council in 2007 to provide an on-orbit intercalibration standard with accuracy of 0.3% (k = 2) for relevant Earth observing sensors. The goal of reference intercalibration, as established in the Decadal Survey, is to enable rigorous high-accuracy observations of critical climate change parameters, including reflected broadband radiation [Clouds and Earth's Radiant Energy System (CERES)], cloud properties [Visible Infrared Imaging Radiometer Suite (VIIRS)], and changes in surface albedo, including snow and ice albedo feedback. In this paper, we describe the CLARREO approach for performing intercalibration on orbit in the reflected solar (RS) wavelength domain. It is based on providing highly accurate spectral reflectance and reflected radiance measurements from the CLARREO Reflected Solar Spectrometer (RSS) to establish an on-orbit reference for existing sensors, namely, CERES and VIIRS on Joint Polar Satellite System satellites, Advanced Very High Resolution Radiometer and follow-on imagers on MetOp, Landsat imagers, and imagers on geostationary platforms. One of two fundamental CLARREO mission goals is to provide sufficient sampling of high-accuracy observations that are matched in time, space, and viewing angles with measurements made by existing instruments, to a degree that overcomes the random error sources from imperfect data matching and instrument noise. The data matching is achieved through CLARREO RSS pointing operations on orbit that align its line of sight with the intercalibrated sensor. These operations must be planned in advance; therefore, intercalibration events must be predicted by orbital modeling. If two competing opportunities are identified, one target sensor must be given priority over the other. The intercalibration method is to monitor changes in targeted sensor response function parameters: effective offset, gain, nonlinearity, optics spectral response, and sensitivity to polarization. In this paper, we use existing satellite data and orbital simulationmethods to determinemission requirements for CLARREO, its instrument pointing ability, methodology, and needed intercalibration sampling and data matching for accurate intercalibration of RS radiation sensors on orbit.

radiometry↗

Enabling the democratization of the genomics revolution with a fully integrated web-based bioinformatics platform

Continued advancements in sequencing technologies have fueled the development of new sequencing applications and promise to flood current databases with raw data. A number of factors prevent the seamless and easy use of these data, including the breadth of project goals, the wide array of tools that individually perform fractions of any given analysis, the large number of associated software/hardware dependencies, and the detailed expertise required to perform these analyses. To address these issues, we have developed an intuitive web-based environment with a wide assortment of integrated and cutting-edge bioinformatics tools in pre-configured workflows. These workflows, coupled with the ease of use of the environment, provide even novice next-generation sequencing users with the ability to perform many complex analyses with only a few mouse clicks and, within the context of the same environment, to visualize and further interrogate their results. As a result, this bioinformatics platform is an initial attempt at Empowering the Development of Genomics Expertise (EDGE) in a wide range of applications for microbial research.

59 BASIC BIOLOGICAL SCIENCES↗

Cybersecurity Platform and Certification Framework Development for Extreme Fast Charging (XFC)-Integrated Charging Ecosystem (Final Project Report)

This report summarizes a pioneering effort in Electric Vehicle charging infrastructure ecosystem cybersecurity requirements, assessment methodologies, functional verification, as well as embodiment of the key technologies in the form of hardware and software tools being made available to the public. EPRI led a team of experts, as well as a stakeholder coalition encompassing all key actors in the EV charging infrastructure ecosystem that includes eXtreme Fast Charging (XFC) equipment (defined as 200kW or above). EV charging infrastructure in the United States is a patchwork of networks that have continued to grow organically and have been designed to serve the charging needs of the EV owners, who are their customers. In doing so, each network provider, as well as their connected entities such as the cloud Electric Vehicle Service Providers or EVSPs, utility back office, utility AMI networks, payment networks, as well as Original Equipment Manufacturer (EV manufacturer) telematics networks, have designed systems that may work well individually, but no single entity is responsible for the entire ecosystem to be secure in terms of data exchange. Furthermore, there is no uniformity in how each actor has implemented the cybersecurity requirements since no system-wide cybersecurity requirements existed prior to this project. The final project report describes the technical approach guided by the EV charging infrastructure cybersecurity working group, convened specifically for this project. The technical approach included definition of requirements at the ecosystem level, treated as a ‘system of systems’, and then passed down to individual systems (EVSE, EV, cloud EVSP, utility, and the payment networks), followed by developing the cybersecurity risk and vulnerability assessment methods, that were later applied to real-world cyber-physical systems at EPRI, ANL, and NREL laboratories, to validate both the process and the results. Finally, in a spotlight over the most vulnerable equipment, which is the EV charge station (AC or DC), the team developed a multi-layer cybersecurity implementation in the embedded domain embodied by the open-source Secure Network Interface Card (SNIC) demonstrating the various ways in which the infrastructure can be secured protecting against the identified attack surfaces. Finally, the entire process of EV infrastructure cybersecurity assessment was encapsulated in the Electric Vehicle Charging Cybersecurity Management (EVC2M) online GUI-based tool, that is expected to be released to the public. The report presents the objectives, the technical approach, the key results, and recommendations for future work.

24 POWER TRANSMISSION AND DISTRIBUTION↗

Highlights from a Decade of OMI-TOMS Total Ozone Observations on EOS Aura

Total ozone measurements from OMI have been instrumental in meeting Aura science objectives. In the last decade, OMI has extended the length of the TOMS total ozone record to over 35 years to monitor stratospheric ozone recovery. OMI-TOMS total ozone measurements have also been combined synergistically with measurements from other Aura instruments and MLS in particular, which provides vertically resolved information that complements the total O3 mapping capability of OMI. With this combined approach, the EOS Aura platform has produced more accurate and detailed measurements of tropospheric ozone. This has led in turn to greater understanding of the sources and transport of tropospheric ozone as well as its radiative forcing effect. The combined use of OMI and MLS data was also vital to the analysis of the severe Arctic ozone depletion event of 2011. The quality of OMI-TOMS total O3 data used in these studies is the result of several factors: a mature and well-validated algorithm, the striking stability of the OMI instrument, and OMI's hyperspectral capabilities used to derive cloud pressures. The latter has changed how we think about the effects of clouds on total ozone retrievals. We will discuss the evolution of the operational V8.5 algorithm and provide an overview and motivation for V9. After reviewing results and developments of the past decade, we finally highlight how ozone observations from EOS Aura are playing an important role in new ozone mapping missions.

Ozone↗

Atmospheric Research 2018 Technical Highlights

Atmospheric research in the Earth Sciences Division (610) consists of research and technology development programs dedicated to advancing knowledge and understanding of the atmosphere and its interaction with the climate of Earth. The Division’s goals are to improve understanding of the dynamics and physical properties of precipitation, clouds, and aerosols; atmospheric chemistry, including the role of natural and anthropogenic trace species on the ozone balance in the stratosphere and the troposphere; and radiative properties of Earth’s atmosphere and the influence of solar variability on the Earth’s climate. Major research activities are carried out in the Mesoscale Atmospheric Processes Laboratory, the Climate and Radiation Laboratory, the Atmospheric Chemistry and Dynamics Laboratory, and the Wallops Field Support Office. The overall scope of the research covers an end-to-end process, starting with the identification of scientific problems, leading to observation requirements for remote sensing platforms, technology and retrieval algorithm development; followed by flight projects and satellite missions; and eventually, resulting in data processing, analyses of measurements, and dissemination from flight projects and missions.

Highlights↗

SU-E-T-186: Cloud-Based Quality Assurance Application for Linear Accelerator Commissioning

Purpose: To identify anomalies and safety issues during data collection and modeling for treatment planning systems Methods: A cloud-based quality assurance system (AQUIRE - Automated QUalIty REassurance) has been developed to allow the uploading and analysis of beam data aquired during the treatment planning system commissioning process. In addition to comparing and aggregating measured data, tools have also been developed to extract dose from the treatment planning system for end-to-end testing. A gamma index is perfomed on the data to give a dose difference and distance-to-agreement for validation that a beam model is generating plans consistent with the beam data collection. Results: Over 20 linear accelerators have been commissioning using this platform, and a variety of errors and potential saftey issues have been caught through the validation process. For example, the gamma index of 2% dose, 2mm DTA is quite sufficient to see curves not corrected for effective point of measurement. Also, data imported into the database is analyzed against an aggregate of similar linear accelerators to show data points that are outliers. The resulting curves in the database exhibit a very small standard deviation and imply that a preconfigured beam model based on aggregated linear accelerators will be sufficient in most cases. Conclusion: With the use of this new platform for beam data commissioning, errors in beam data collection and treatment planning system modeling are greatly reduced. With the reduction in errors during acquisition, the resulting beam models are quite similar, suggesting that a common beam model may be possible in the future. Development is ongoing to create routine quality assurance tools to compare back to the beam data acquired during commissioning. I am a medical physicist for Alzyen Medical Physics, and perform commissioning services.

60 APPLIED LIFE SCIENCES↗

Inter-Calibration and Concatenation of Climate Quality Infrared Cloudy Radiances from Multiple Instruments

A change in climate is not likely captured from any single instrument, since no single instrument can span decades of time. Therefore, to detect signals of global climate change, observations from many instruments on different platforms have to be concatenated. This requires careful and detailed consideration of instrumental differences such as footprint size, diurnal cycle of observations, and relative biases in the spectral brightness temperatures. Furthermore, a common basic assumption is that the data quality is independent of the observed scene and therefore can be determined using clear scene data. However, as will be demonstrated, this is not necessarily a valid assumption as the globe is mostly cloudy. In this study we highlight challenges in inter-calibration and concatenation of infrared radiances from multiple instruments by focusing on the analysis of deep convective or anvil clouds. TRMM/VIRS is potentially useful instrument to make correction for observational differences in the local time and foot print sizes, and thus could be applied retroactively to vintage instruments such as AIRS, IASI, IRIS, AVHRR, and HIRS. As the first step, in this study, we investigate and discuss to what extent AIRS and VIRS agree in capturing deep cloudy radiances at the same local time. The analysis also includes comparisons with one year observations from CrIS. It was found that the instruments show calibration differences of about 1K under deep cloudy scenes that can vary as a function of land type and local time of observation. The sensitivity of footprint size, view angle, and spectral band-pass differences cannot fully explain the observed differences. The observed discrepancies can be considered as a measure of the magnitude of issues which will arise in the comparison of legacy data with current data.

cloudy radiance↗

Retrieval of Aerosol Properties from Multi-Spectral Extinction Measurements

The direct-beam spectral extinction of solar radiation contains information on atmospheric composition in a form that is essentially free from the data analysis complexities that often arise from multiple scattering. Ground based Multi-Filter Shadowband Radiometer (MFRSR) measurements provide such information for the vertical atmospheric column path, while solar occultation measurements from a satellite platform provide horizontal slices through the atmosphere. We describe application of a Multi-Spectral Atmospheric Column Extinction (MACE) analysis technique used to analyze MFRSR data also to occultation measurements made by SAGE II. For analysis, we select the 1985 Nevado del Ruiz volcanic eruption period to retrieve atmospheric profiles of ozone and NO2, and changes in the stratospheric aerosol size and optical depth. The time evolution of volcanic aerosol serves as a passive tracer to study stratospheric dynamics, and changes in particle size put constraints on the sulfur chemistry modeling of volcanic aerosols. Paper presented at The '99 Kyoto Aerosol-Cloud Workshop, held Dec 1-3, 1999, Kyoto, Japan

Lacis, Andrew A.↗

Design of a Far-Infrared Spectrometer for Atmospheric Thermal Emission Measurements

Global measurements of far infrared emission from the upper troposphere are required to test models of cloud radiative forcing, water vapor continuum emission, and cooling rates. Spectra with adequate resolution can also be used for retrieving atmospheric temperature and humidity profiles, and yet there are few spectrally resolved measurements of outgoing longwave flux at wavelengths longer than 16 m. It has been difficult to make measurements in the far infrared due to the need for liquid-helium cooled detectors and large optics to achieve adequate sensitivity and bandwidth. We review design considerations for infrared Fourier transform spectrometers, including the dependence of system performance on basic system parameters, and discuss the prospects for achieving useful sensitivity from a satellite platform with a lightweight spectrometer using uncooled detectors.

Johnson, David G.↗

FPGA-Accelerated Machine Learning Inference as a Service for Particle Physics Computing

Large-scale particle physics experiments face challenging demands for high-throughput computing resources both now and in the future. New heterogeneous computing paradigms on dedicated hardware with increased parallelization, such as Field Programmable Gate Arrays (FPGAs), offer exciting solutions with large potential gains. The growing applications of machine learning algorithms in particle physics for simulation, reconstruction, and analysis are naturally deployed on such platforms. We demonstrate that the acceleration of machine learning inference as a web service represents a heterogeneous computing solution for particle physics experiments that potentially requires minimal modification to the current computing model. As examples, we retrain the ResNet-50 convolutional neural network to demonstrate state-of-the-art performance for top quark jet tagging at the LHC and apply a ResNet-50 model with transfer learning for neutrino event classification. Using Project Brainwave by Microsoft to accelerate the ResNet-50 image classification model, we achieve average inference times of 60 (10) ms with our experimental physics software framework using Brainwave as a cloud (edge or on-premises) service, representing an improvement by a factor of approximately 30 (175) in model inference latency over traditional CPU inference in current experimental hardware. A single FPGA service accessed by many CPUs achieves a throughput of 600–700 inferences per second using an image batch of one, comparable to large batch-size GPU throughput and significantly better than small batch-size GPU throughput. Deployed as an edge or cloud service for the particle physics computing model, coprocessor accelerators can have a higher duty cycle and are potentially much more cost-effective.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

FPGAs as a Service to Accelerate Machine Learning Inference [PowerPoint]

Large-scale particle physics experiments face challenging demands for high-throughput computing resources both now and in the future. New heterogeneous computing paradigms on dedicated hardware with increased parallelization, such as Field Programmable Gate Arrays (FPGAs), offer exciting solutions with large potential gains. The growing applications of machine learning algorithms in particle physics for simulation, reconstruction, and analysis are naturally deployed on such platforms. We demonstrate that the acceleration of machine learning inference as a web service represents a heterogeneous computing solution for particle physics experiments that requires minimal modification to the current computing model. As examples, we retrain the ResNet-50 convolutional neural network to demonstrate state-of-the-art performance for top quark jet tagging at the LHC, and apply a ResNet-50 model with transfer learning for neutrino event classification. Using Project Brainwave b y Microsoft to accelerate the ResNet-50 image classification model, we achieve average inference times of 60 (10) milliseconds with our experimental physics software framework using Brainwave as a cloud (edge or on-premises) service, representing an improvement by a factor of approximately 30 (175) in model inference latency over traditional CPU inference in current experimental hardware. A single FPGA service accessed by many CPUs achieves a throughput of 600--700 inferences per second using an image batch of one, comparable to large batch-size GPU throughput and significantly better than small batch-size GPU throughput. Deployed as an edge or cloud service for the particle physics computing model, coprocessor accelerators can have a higher duty cycle and are potentially much more cost-effective.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

Multi-Temporal Predictive Modelling of Sorghum Biomass Using UAV-Based Hyperspectral and LiDAR Data

High-throughput phenotyping using high spatial, spectral, and temporal resolution remote sensing (RS) data has become a critical part of the plant breeding chain focused on reducing the time and cost of the selection process for the “best” genotypes with respect to the trait(s) of interest. In this paper, the potential of accurate and reliable sorghum biomass prediction using visible and near infrared (VNIR) and short-wave infrared (SWIR) hyperspectral data as well as light detection and ranging (LiDAR) data acquired by sensors mounted on UAV platforms is investigated. Predictive models are developed using classical regression-based machine learning methods for nine experiments conducted during the 2017 and 2018 growing seasons at the Agronomy Center for Research and Education (ACRE) at Purdue University, Indiana, USA. The impact of the regression method, data source, timing of RS and field-based biomass reference data acquisition, and the number of samples on the prediction results are investigated. R2 values for end-of-season biomass ranged from 0.64 to 0.89 for different experiments when features from all the data sources were included. Geometry-based features derived from the LiDAR point cloud to characterize plant structure and chemistry-based features extracted from hyperspectral data provided the most accurate predictions. Evaluation of the impact of the time of data acquisition during the growing season on the prediction results indicated that although the most accurate and reliable predictions of final biomass were achieved using remotely sensed data from mid-season to end-of-season, predictions in mid-season provided adequate results to differentiate between promising varieties for selection. The analysis of variance (ANOVA) of the accuracies of the predictive models showed that both the data source and regression method are important factors for a reliable prediction; however, the data source was more important with 69% significance, versus 28% significance for the regression method.

09 BIOMASS FUELS↗