Space Objects In Orbit Research Articles

A recurrent neural network-based approach for ballistic coefficient estimation of resident space objects in low earth orbit

Characterizing Resident Space Objects (RSOs) has become paramount for several Space Situational Awareness functions, such as accurate orbit prediction, collision avoidance and sensor tasking. Due to the huge and diverse amount of data to handle and fuse for this purpose, there is an increasing interest in Machine Learning (ML) approaches to retrieve physical parameters of RSOs in a cost-efficient manner. Among different ML architectures, Recurrent Neural Networks (RNNs) are particularly suitable to handle sequential or time-series data. In this context, this paper demonstrates the possibility to use Recurrent Neural Networks to estimate the ballistic coefficient of RSOs in Low Earth Orbit from time series of orbital elements. A particular RNN architecture has been adopted, i.e., the Gated Recurrent Unit, as it has been demonstrated to be cost efficient and to have good performance for the required application. The sensitivity of RNNs with respect to atmospheric model uncertainty has been analysed. The algorithm can handle unevenly spaced time-series data. The proposed neural network has been demonstrated to be robust with respect to orbital state errors. The applicability of the presented approach is tested and discussed using synthetic datasets generated with a high-accuracy numerical orbital propagator, reaching a mean percentage error of 10% in the estimation of the ballistic coefficient.

CLOWN: The PASO Cloud Detection for Optimization of Automatic Optical Surveys

Orbiting space objects have become in the last decade a major nuisance impacting ground astronomy and orbiting space assets, from observatories to satellites and space stations. In particular with the rise of the satellite population in Low Earth Orbits, space objects are becoming an even bigger threat and a strong problem to astronomical observations. To tackle these threats, several coordinated surveillance networks composed of dedicated sensors (telescopes, radars, and laser ranging facilities) track and survey space objects, from debris to active satellites. As part of the European Space Surveillance & Tracking network, Portugal is developing the Pampilhosa da Serra Space Observatory, with both radio and optical telescopes dedicated to the Space Situational Awareness domain, deployed at a Dark Sky destination. To optimize telescope survey time, we developed CLOud Watcher at Night (CLOWN), an application interface that automatically monitors clouds in real time. This software can correctly trace cloud positions in the sky and provide accurate pointing information to the observation planning of the optical telescope to avoid cloudy areas. CLOWN only requires the use of an all-sky camera, which is already a norm in observatories with optical telescopes and can be used with any camera, including those for which no information about its model specification do exist. CLOWN does not require great computing power, and it does not require the installation of additional equipment. CLOWN results are very promising and confirm that the app can correctly identify clouds in a variety of different conditions and cloud types.

Predicting RSO populations using a neighbouring orbits technique

Abstract The determination of the full population of Resident Space Objects (RSOs) in Low Earth Orbit (LEO) is a key issue in the field of space situational awareness that will only increase in importance in the coming years. We endeavour to describe a novel method of inferring the population of RSOs as a function of orbital height and inclination for a range of magnitudes. The method described uses observations of an orbit of known height and inclination to detect RSOs on neighbouring orbits. These neighbouring orbit targets move slowly relative to our tracked orbit, and are thus detectable down to faint magnitudes. We conduct simulations to show that, by observing multiple passes of a known orbit, we can infer the population of RSOs within a defined region of orbital parameter space. Observing a range of orbits from different orbital sites will allow for the inference of a population of LEO RSOs as a function of their orbital parameters and object magnitude.

Effects of phase angle and sensor properties on on-orbit debris detection using commercial star trackers

The recent proliferation of resident space objects (RSOs) in low Earth orbit (LEO) threatens the sustainability of space as a resource and requires persistent monitoring to avoid collisions involving valuable space assets. State-of-the-art ground-based space surveillance techniques, due to their susceptibility to atmosphere, weather, and lighting conditions, tend to focus on RSOs with characteristic length greater than 10 cm or 1 dm. Consequently, millions of smaller LEO RSOs remain untracked by ground-based methods, which reduces overall space situational awareness. Onboard satellite sensors offer a space-based method for tracking RSOs. Prior research has investigated the feasibility of using commercial star trackers (CSTs) — optical sensors prevalent on most active spacecraft — to observe, detect, and estimate the position and velocity of RSOs larger than 10 cm. In a recent effort, we expanded on these feasibility studies by assessing the capabilities of CSTs to detect debris particles smaller than 10 cm in characteristic length. We modeled the particles as Lambertian spheres with zero phase angle and ten percent reflectivity and found that typical CSTs can detect properly illuminated debris of characteristic length between 1 cm and 10 cm even at distances of tens of kilometers. More sensitive CSTs can characterize decimeter-scale RSOs hundreds of kilometers away; alternatively, they can track centimeter-scale and smaller RSOs at closer distances. In this paper, we summarize these key results and extend our previous study by relaxing its zero-phase-angle assumption and characterizing the effect of Sun-RSO-CST phase angle on debris detection range. We identify a number of representative CSTs with publicly available optical characteristics and consider the effects of properties such as pixel pitch, focal length, aperture diameter, and field of view (FOV) on the capability of each CST to detect debris at a given distance and relative velocity (in the form of streak speed). We find that, for debris particles modeled as diffuse Lambertian spheres, Sun-RSO-CST phase angles as high as 57° result in no more than 20% reduction to the useful RSO-CST detection range. In addition, we find that pixel pitch and focal length, rather than aperture diameter, tend to determine the capability of a given CST to resolve two distinct RSOs. Furthermore, streak speed may serve as a stronger limiting factor for detection of smaller debris particles than for larger ones. Despite these limitations, the overall results indicate that CSTs have the potential to substantially enhance space-based debris detection capability.

Space Domain Awareness Observations Using the Buckland Park VHF Radar

There is increasing interest in space domain awareness worldwide, motivating investigation of the use of non-traditional sensors for space surveillance. One such class of sensor is VHF wind profiling radars, which have a low cost relative to other radars typically applied to this task. These radars are ubiquitous throughout the world and may potentially offer complementary space surveillance capabilities to the Space Surveillance Network. This paper updates an initial investigation on the use of Buckland Park VHF wind profiling radars for observing resident space objects in low Earth orbit to further investigate the space surveillance capabilities of the sensor class. The radar was operated during the Australian Defence “SpaceFest” 2019 activity, incorporating new beam scheduling and signal processing functionality that extend upon the capabilities described in the initial investigation. The beam scheduling capability used two-line element propagations to determine the appropriate beam direction to use to observe transiting satellites. The signal processing capabilities used a technique based on the Keystone transform to correct for range migration, allowing the development of new signal processing modes that allow the coherent integration time to be increased to improve the SNR of the observed targets, thereby increasing the detection rate. The results reveal that 5874 objects were detected over 10 days, with 2202 unique objects detected, representing a three-fold increase in detection rate over previous single-beam direction observations. The maximum detection height was 2975.4 km, indicating a capability to detect objects in medium Earth orbit. A minimum detectable RCS at 1000 km of −10.97 dBm2 (0.09 m2) was observed. The effects of Faraday rotation resulting from the use of linearly polarised antennae are demonstrated. The radar’s utility for providing total electron content (TEC) measurements is investigated using a high-range resolution mode and high-precision ephemeris data. The short-term Fourier transform is applied to demonstrate the radar’s ability to investigate satellite rotation characteristics and monitor ionospheric plasma waves and instabilities.

Dynamic Model Fidelity Effects on Covariance Based Track Association

This paper examines the performance of covariance-based track association with varying levels of dynamic model fidelity. To minimize the dynamic nonlinearities, the mean modified equinoctial (MME) elements are used. Ground-based radar measurements are generated by propagating the osculating orbits of 100 space objects in low Earth orbit using a 70×70 spherical harmonic gravity model (JGM-3). Atmospheric drag, solar radiation, and third-body gravity are not considered. To associate two tracks, the two sets of states and covariances are propagated to a common time. The Mahalanobis distance between the two tracks is then used to determine whether they associate. A higher-fidelity dynamic model for state propagation improves association performance, while the fidelity of the dynamic model for covariance propagation does not have a significant impact. The MME-based track association demonstrates excellent performance when incorporating J2, J22, J4, and J6 in the MME state propagation as well as the MME covariance propagation with J2. It achieves an accuracy of 99.92%, a true positive sensitivity of 92.43%, a false negative sensitivity of 7.57%, a false positive sensitivity of 0%, and a true negative sensitivity of 100% when associating tracks separated by 10 days.

Variable-fidelity sensors and observer uncertainty using touring multi-body periodic orbits to conduct cislunar SSA: preliminary study

An accelerating interest in cislunar space and lunar orbit for civilian, commercial, and scientific missions requires a space situational awareness (SSA) architecture extending beyond geosynchronous orbit to promote space traffic management and safety. Space-based SSA in cislunar space is challenging due to difficulties associated with accurately estimating the position of the surveillance satellite, which is a foundational requirement for effectively performing the general SSA mission. Using multiple surveillance satellites with lower-fidelity sensors helps alleviate these concerns by aggregating multiple data sets with higher variance to achieve the same level or potentially improved accuracy as compared to fewer higher-quality sensors. A subset of Earth–Moon periodic orbits, herein identified as “touring” orbits, are used for an optical surveillance constellation with a target resident space object (RSO) in a L1 Halo orbit. Angles-only measurement data are processed utilizing an extended Kalman filter to estimate the position of the RSO. The analysis focuses on assessing the effectiveness of different numbers of surveillance satellites using touring cislunar periodic orbits for conducting the SSA mission relative to L1. Overall, this study finds that the use of an SSA constellation with low-fidelity sensors can match the performance achieved by a constellation featuring higher-fidelity sensors and reduced observer uncertainty for the observer orbits examined.

ATLAS: Latest Advancements and First Observations

The increasing amount of space debris poses a significant threat to operational satellites and space-based services. This article updates the community on the current status of the development of ATLAS, a tracking radar that is part of the EUSST network and aims to detect space objects in low Earth orbits. This article focuses on the latest activities performed: calibration of the pointing system and initial observations of space objects. The calibration procedure consisted of cross-scanning the Solar disk and yielded great results, obtaining an offset of 5.3° in azimuth and 0.10° in elevation. The first observation campaign resulted in 33 range detections of the International Space Station (ISS) with a probability of false alarm of 10−9. The observations were then used to readjust the radar equation to assess the real-world performance of the system.

Long baseline bistatic radar imaging of tumbling space objects for enhancing space domain awareness

AbstractLong baseline bistatic radar systems herald enhanced sensitivity and metric accuracy for space objects in geosynchronous orbits and beyond. Radio telescopes are ideal participants in such a system; in particular, they often feature large apertures with low‐noise temperatures and have stable, synchronised clocks. Pairing radio telescopes with high‐power radars creates new methodologies for Space Domain Awareness. This paper describes long baseline bistatic measurements using the Millstone Hill Radar in the USA, the Tracking and Imaging Radar in Germany, multiple receivers of the enhanced multi‐element remotely linked interferometer network array in the United Kingdom, and the Westerbork Synthesis Radio Telescope in the Netherlands. The authors, a Research Task Group formed by the NATO Science and Technology Organisation Sensors and Electronic Technology Panel (SET‐293), performed novel bistatic and monostatic radar imaging experiments with real on‐orbit tumbling rocket bodies. These experiments on tumbling objects at near‐geosynchronous orbits highlight successful demonstrations of advanced bistatic Doppler characterisation across diverse imaging geometries. Specialised Doppler processing on tumbling targets, such as the Doppler superpulse algorithm, enables high‐fidelity rotation period estimation and determination of minimum target size.

REALIZATION OF THE TECHNOLOGY OF DUAL USE OF OUTER SPACE CONTROL MEANS

Context. The radar means of outer space control are monofunctional systems, despite their potential multifunctionality. To ensure the effective implementation of target functions, only some of the total number of potential functions are used, other existing functions are not studied and as a result, don’t used. The target functions of most existing domestic and foreign radiolocation stations on space control RLS SC are practically reduced only to the control of space objects in different orbits. The obtained information is not fully used, so the new target doesn’t form. As the history of the development of defense complexes shows, in most industrialized countries there are examples of the use of military developments for civilian purposes (spin-off) and civilian developments for military purposes (spin-on). As a result of these synergy processes emphasis on the strategy of double technologies and double innovation increased. Objective. The aim of the work is to study the possibility of introducing an ionospheric channel into the domestic radar station 5N86 Dnipro (Hen House) and expanding its intended use by using most of its functional systems to implement the ionosphere control function. Method. The paper uses a comparative analysis of the main functional systems and technical characteristics of the 5N86 Dnipro radar and non-coherent scatter radars (PHR) of the global ionosphere control network. Results. The main characteristic features of the RLS 5N86 are analyzed and, taking into account the characteristics of the signals, the possibility of using the multifunctionality of the radar to form a new target function for monitoring outer space is substantiated. From an applied point of view, a number of specific scientific and practical solutions are given, aimed at the realization of dual-purpose technology in the implementation of the ionosphere control function by a radar station - both for solving the problems of increasing the own efficiency and for the interests of fundamental science. It is shown that the creation of new target RLS is based on both the use of already existing systems and the introduction of new ones. The importance of the scientific task on the implementation of the ionosphere control function and the possibility of integrating into the global ionosphere control network is substantiated. Conclusions. The presence of scientific and technical developments and the practical experience of the domestic RLS SC developer makes the new target function realization absolutely realistic. The implementation of the double-purpose technology will ensure an effective solution for both applied and fundamental scientific tasks.

A Case Study on the Effect of Atmospheric Density Calibration on Orbit Predictions with Sparse Angular Data

Accurately modeling the density of atmospheric mass is critical for orbit determination and prediction of space objects. Existing atmospheric mass density models (ADMs) have an accuracy of about 15%. Developing high-precision ADMs is a long-term goal that requires a better understanding of atmospheric density characteristics, more accurate modeling methods, and improved spatiotemporal data. This study proposes a method for calibrating ADMs using sparse angular data of space objects in low-Earth orbit over a certain period of time. Applying the corrected ADM not only improves the accuracy of orbit determination, but also enhances the accuracy of orbit prediction beyond the correction period. The study compares the impact of two calibration methods: atmospheric mass density model coefficient (ADMC) calibration and high precision satellite drag model (HASDM) calibration on the accuracy of orbit prediction of space objects. One month of ground-based telescope array angular data is used to validate the results. Space objects are classified as calibration objects, participating in ADM calibration, and verification objects, inside and outside the calibration orbit region, respectively. The results show that applying the calibrated ADM can significantly increase the accuracy of orbit prediction. For objects within the calibration orbit region, the calibration object’s orbit prediction error was reduced by about 55%, while that of verification objects was reduced by about 45%. The reduction in orbit prediction error outside this region was about 30%. This proposed method contributes significantly to the development of more reliable ADMs for orbit prediction of space objects with sparse angular data and can provide significant academic value in the field of space situational awareness.

Comparative Accuracies of Models for Drag Prediction During Geomagnetically Disturbed Periods: A First Principles Model Versus Empirical Models

AbstractWe examine the accuracy of density prediction by the first principles model Thermosphere Ionosphere Electrodynamics General Circulation Model (TIEGCM) developed by the National Center for Atmospheric Research and compare it to the accuracy of three empirical models: Jacchia 71, the Naval Research Laboratory Mass Spectrometer Incoherent Scatter Extended 2000 (NRLMSIS), Jacchia 1971, and Jacchia‐Bowman 2008. Comparisons are made for three large storms: the October 2003 storm, the March 2013 storm, and the March 2015 storm. To evaluate the accuracy of these models we use tracking data for nine space objects in low Earth orbit. Additionally, we evaluate the accuracy of the TIEGCM and NRLMSIS with data from high precision accelerometers on the Challenging Minisatellite Payload (CHAMP) and Gravity field and Circulation Explorer (GOCE) satellites. The goal is to assess the use of a first principles model as a potential tool for forecasting satellite drag during large magnetic storms. For the storms considered, we found the TIEGCM, JB2008, and NRLMSIS models to be substantially more accurate than the Jacchia 71 model. The accuracies of the TIEGCM and JB2008 models were similar, but overall, the TIEGCM was more accurate. We found smaller differences for TIEGCM versus CHAMP than for NRLMIS for the Halloween Storm, and smaller differences than results published for JB2008 and the assimilative model HASDM. The empirical models are at present more practical for operational purposes, but the TIEGCM, developed as a research model, with a greater focus on operational use offers the potential for improved utility during stressing conditions.

Aggregate effects of proliferating low-Earth-orbit objects and implications for astronomical data lost in the noise

The rising population of artificial satellites and associated debris in low-altitude orbits is increasing the overall brightness of the night sky, threatening ground-based astronomy as well as a diversity of stakeholders and ecosystems reliant on dark skies. We present calculations of the potentially large rise in global sky brightness from space objects in low Earth orbit, including qualitative and quantitative assessments of how professional astronomy may be affected. Debris proliferation is of special concern: we calculate that all log-decades in debris size contribute approximately the same amount of night sky radiance, so debris-generating events are expected to lead to a rapid rise in night sky brightness along with serious collision risks for satellites from centimetre-sized objects. This increase in low-Earth-orbit traffic will lead to loss of astronomical data and diminish opportunities for ground-based discoveries as faint astrophysical signals become increasingly lost in the noise. Lastly, we discuss the broader consequences of brighter skies for a range of sky constituencies, equity/inclusion and accessibility for Earth- and space-based science, and cultural sky traditions. Space and dark skies represent an intangible heritage that deserves intentional preservation and safeguarding for future generations. Each space launch is assessed for various risks, but not its wider impacts. This Perspective shows how the aggregate effects of space launches, plus the attendant rise of space debris, affect the darkness of our night sky now and in the future.

Novel Source–Sink Model for Space Environment Evolution with Orbit Capacity Assessment

The increasing number of anthropogenic space objects (ASOs) in low Earth orbit (LEO) poses a threat to the safety and sustainability of the space environment. Multiple companies are planning to launch large constellations of hundreds to thousands of satellites in the near future, increasing the probability of collisions and debris generation. This paper analyzes the long-term evolution of the LEO ASO population with the goal of estimating LEO orbital capacity. This is carried out by introducing a new probabilistic source–sink model. The developed source–sink model is a multishell multispecies model, which includes different object species, such as active and derelict satellites, and debris. Furthermore, debris are divided into the following two subgroups: trackable and nontrackable debris, the last ones representing a significant hazard for active satellites. In addition, the proposed model accounts for collision events and atmospheric drag effects, which include the influence of solar activity. Indeed, the Jacchia–Bowman 2008 thermospheric density model is exploited. The results prove that considering untracked debris within the model produces more collisions, and therefore a smaller population of active satellites affecting the safety of LEO and its orbital capacity.

Методика оценивания оперативности информационного обеспечения автоматизированных систем предупреждения об опасных ситуациях в околоземном космическом пространстве

This article presents the structure and content of the main stages of the methodology for assessing the efficiency of information support for automated warning systems about dangerous situations in near-Earth outer space when observing space objects in high orbits by optical means of observing outer space with the use of a software and hardware information collection complex. Problem statement: development of a technique for detecting deviations of space objects from the calculated orbits of their movement by ground-based means of observing outer space, taking into account various options for constructing orbital groupings and a network of ground-based optical means of observation. Results: A methodology has been developed for assessing the efficiency of information support for automated warning systems about dangerous situations in near-Earth outer space, taking into account various ways of organizing an information support system and allowing dynamically evaluating the efficiency of detecting deviations of space objects from calculated orbits. Practical significance: the proposed methodology makes it possible to identify optimal ways of organizing information support for automated warning systems about dangerous situations in near-Earth outer space based on an assessment of the efficiency of detecting deviations of space objects from calculated orbits with dynamic changes in the space environment and the state of operation of space surveillance facilities and to form recommendations for improving information support for automated warning systems about dangerous situations in near-Earth outer space. Discussion: the novelty of the proposed formulation of the problem lies in the fact that the structure of the developed methodology allows us to take into account various ways of organizing the information support system and to build probabilistic estimates of the efficiency of its functioning on the basis of dynamic changes in the space situation and the state of ground-based means of observing outer space.

Long-Term Evolution of the Space Environment Considering Constellation Launches and Debris Disposal

Increasingly frequent launch activities, as well as the development of mega constellations, would cause a drastic increase in the number of space objects, which will then alter the evolution of outer space. To reveal this long-term change, an accurate space-environment model is required. There are two main approaches to building this model, one of which is to track the state of space objects individually, which will use significant computing resources; the other is to take macroscopic variables, such as spatial density, as the state variable to depict a group of space debris, which requires less computational effort. In this study, a space debris environment evolution model with spatial density as the state variable is established, which considers the nonzero eccentricity of the debris orbit and utilizes the NASA breakup model to ensure accuracy. In addition, the Gaussian mixture model (GMM) is applied to take the uncertainty of launch activities into account. The long-term impacts of mega constellations and their post-mission disposal (PMD) on the debris environment are discussed based on the evolution model. It was found that constellations with high orbit altitude, such as OneWeb, will lead to an exponential increase in space objects in low Earth orbit (LEO). In addition, deorbit time is the main factor affecting the PMD efficiency, followed by deorbit strategies.

Super diffraction limit spectral imaging detection and material type identification of distant space objects.

The image information of distant objects shows a diffuse speckle pattern due to diffraction limit, non-uniform scattering, etc., which is difficult to achieve object discrimination. In this study, we have developed a staring spectral video imaging system mounted on a ground-based telescope observation platform to detect the high orbit space objects and gain their spectral images for six groups of GEO targets. The speckle remains basically the same characteristic as the projection structure of the object due to "the balloon inflation phenomenon of near parallel light during long-distance atmospheric transmission" under the premise of considering the bi-directional reflection distribution function (BRDF), Rayleigh scattering theory, and the memory effect. Based on this phenomenon, a mathematical model of remote target scattering spectrum imaging is established where the speckle can be treated as both a global speckle and speckle combination of texture blocks caused by various components of the target. The radial basis function (RBF) neural network is separately used to invert the global speckle and the speckle combination of the texture blocks on account of the typical target material database. The results show that the target materials are of relatively fewer kinds in the global inversion with only including gallium arsenide panel (GaAs) and carbon fiber (CF), for which the highest goodness of curve fitting is only 77.97. An improved algorithm makes their goodness of fit reach 90.29 and 93.33, respectively, in view of one conjecture that the target surface contains unknown materials. The spectral inversion result of the texture blocks shows that the types of materials in each target texture block increase significantly, and that the area ratio of different materials inverted in the block is different from each other. It is further confirmed that the speckle image contains the overall projection structure of distant target and the spectral image projection of each component is relatively fixed, which is the result of the comprehensive action of various mechanisms of ultra-long-haul atmospheric transmission and optical system focusing imaging after BRDF spectral scattering. The spectral image fine inversion is expected to restore the clear structure of the target. This discovery provides important support for the remote imaging and identification of distant and ultra-diffractive targets.

Ground Photometric Measurements Inversion Methods for Space Object Characteristics Based on Non-linear Filtering: A Short Review

: The ground space optical observation system considered as one of the main measurement methods to observe space object in Medium Earth Orbit (MEO), Highly Elliptical Orbit (HEO) and Geostationary Orbit (GEO) faces difficulties in forming high-solution images for objects located in GEO that only appears several light spots with limited pixels due to the impacts of observing distance, resolution ratio, and other atmospheric conditions. The light curve is the time history of an object’s observed brightness obtained from the ground-based optical observation system through light curves, right ascension, and declination of space objects. The characteristic of space objects such as position, velocity, size, shape, and material can be inverted. This paper has analyzed the principles, applications, merits, and demerits of several non-linear filtering methods in detail. Besides, the scientific description of inversion for characteristics of a non-resolved space object from ground photometric measurements and the essence of the non-linear filtering inversion method has also been clarified. A selection principle of the non-linear filtering inversion method for different space object characteristics is then proposed, and the developing direction of such inversion methods is also described in the end.

Track detection of high-velocity resident space objects in Low Earth Orbit

The world’s economy has become heavily dependent on the services provided by satellites. With the exponential increase in satellite launches, the population of defunct or inactive hardware in space has grown substantially. This is especially true in sensitive orbits such as the Low Earth Orbit (LEO) and Geostationary Earth Orbit (GEO) regimes. These objects, collectively known as orbital debris, can reach speeds of up to 28 000km h−1 in LEO. At these orbital speeds, even the smallest of objects can pose a considerable threat to operational satellites or astronauts. This makes the monitoring, and detection, of these objects of the utmost importance. This work describes the latest detection strategy used in one of Europe’s largest Space Situational Awareness (SSA) installation; the BIstatic RAdar for LEo Survey (BIRALES) space debris radar. We present a novel bottom-up approach that makes use of single-linkage clustering to identify faint radar streaks in spectrogram data. Tests on synthetic data have shown that the detection strategy presented in this study obtains a higher detection rate when it is compared against existing methods. Unlike other approaches, this detection strategy, using the Multi-beam streak detection strategy (MSDS) algorithm, was still able to recall 90% of the track information at an Signal-to-Noise Ratio (SNR) of 2dB.

Uncertainty quantification techniques for data-driven space weather modeling: thermospheric density application

Machine learning (ML) has been applied to space weather problems with increasing frequency in recent years, driven by an influx of in-situ measurements and a desire to improve modeling and forecasting capabilities throughout the field. Space weather originates from solar perturbations and is comprised of the resulting complex variations they cause within the numerous systems between the Sun and Earth. These systems are often tightly coupled and not well understood. This creates a need for skillful models with knowledge about the confidence of their predictions. One example of such a dynamical system highly impacted by space weather is the thermosphere, the neutral region of Earth’s upper atmosphere. Our inability to forecast it has severe repercussions in the context of satellite drag and computation of probability of collision between two space objects in low Earth orbit (LEO) for decision making in space operations. Even with (assumed) perfect forecast of model drivers, our incomplete knowledge of the system results in often inaccurate thermospheric neutral mass density predictions. Continuing efforts are being made to improve model accuracy, but density models rarely provide estimates of confidence in predictions. In this work, we propose two techniques to develop nonlinear ML regression models to predict thermospheric density while providing robust and reliable uncertainty estimates: Monte Carlo (MC) dropout and direct prediction of the probability distribution, both using the negative logarithm of predictive density (NLPD) loss function. We show the performance capabilities for models trained on both local and global datasets. We show that the NLPD loss provides similar results for both techniques but the direct probability distribution prediction method has a much lower computational cost. For the global model regressed on the Space Environment Technologies High Accuracy Satellite Drag Model (HASDM) density database, we achieve errors of approximately 11% on independent test data with well-calibrated uncertainty estimates. Using an in-situ CHAllenging Minisatellite Payload (CHAMP) density dataset, models developed using both techniques provide test error on the order of 13%. The CHAMP models—on validation and test data—are within 2% of perfect calibration for the twenty prediction intervals tested. We show that this model can also be used to obtain global density predictions with uncertainties at a given epoch.