Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Data-Driven Modelling of the Van Allen Belts: The 5DRBM Model for Trapped Electrons

Data-Driven Modelling of the Van Allen Belts: The 5DRBM Model for Trapped Electrons The magnetosphere sustained by the rotation of the Earth’s liquid iron core traps charged particles, mostly electrons and protons, into structures referred to as the Van Allen belts. These radiation belts, in which the density of charged energetic particles can be very destructive for sensitive instrumentation, have to be crossed on every orbit of satellites traveling in elliptical orbits around the Earth, as is the case for ESA’s INTEGRAL and XMM-Newton missions. This paper presents the first working version of the 5DRBM-e model, a global, data-driven model of the radiation belts for trapped electrons. The model is based on in-situ measurements of electrons by the radiation monitors on board the INTEGRAL and XMM-Newton satellites along their long elliptical orbits for respectively 16 and 19 years of operations. This model, in its present form, features the integral flux for trapped electrons within energies ranging from 0.7 to 1.75 MeV. Cross-validation of the 5DRBM-e with the well-known AE8min/max and AE9mean models for a low eccentricity GPS orbit shows excellent agreement, and demonstrates that the new model can be used to provide reliable predictions along widely different orbits around Earth for the purpose of designing, planning, and operating satellites with more accurate instrument safety margins. Future work will include extending the model based on electrons of different energies and proton radiation measurement data. Keywords: Van Allen belts, Radiation belt modelling, Trapped particles, Radiation Environment, Space environment, Space weather 1. Introduction Mr. James Van Allen, payload specialist of the Explorer missions, to discover successively: the inner proton belt Only theorised before space exploration began (Explorer 1 & 3), the inner electron belt (Explorer 4) (Stormer, 1937), the Earth Radiation Belts were discov- and the outer electron belt (Pioneer 3) (Van Allen, 1959). ered for the first time in 1958 with the very first US satel- Consequently, these high radiation regions surrounding the lite, Explorer 1. These first in-situ measurements allowed Earth are known as the Van Allen Belts (VAB). These radiation belts are the result of the complex interaction between the Earth’s magnetosphere and the Corresponding author interplanetary medium mainly driven by the solar wind Email addresses: lionel.metrailler@esa.int (Lionel Métrailler), guillaume.belanger@esa.int (Guillaume Bélanger), that moves energetic charge-carrying electrons and pro- peter.kretschmar@esa.int (Peter Kretschmar), tons with it. These particles are trapped in the mag- erik.kuulkers@esa.int (Erik Kuulkers), netosphere forming high radiation torus-shaped regions Ricardo.Perez.Martinez@esa.int (Ricardo Pérez Martínez), jan.uwe.ness@esa.int (Jan-Uwe Ness), pedro.rodriguez@esa.int around the Earth’s magnetic axis. The Van Allen Belts (Pedro Rodriguez), mauro.casale@esa.int (Mauro Casale), extend from altitudes from ∼1,000km (0.2 R ) to more jorge.fauste@esa.int (Jorge Fauste), timothy.finn@esa.int than ∼60,000km (10 R ). (Timothy Finn), celia.sanchez@esa.int (Celia Sanchez), thomas.godard@esa.int (Thomas Godard), Modelling these highly dynamic radiation belts is im- richard.southworth@esa.int (Richard Southworth) portant for space-borne activities, both manned and un- Swiss National Trainee Program, Swiss Space Center, École manned, in the near-Earth environment. Theoretical mod- Polytechnique Fédérale de Lausanne (EPFL), PPH 338, Station 13, els of the belts can be built on simplified physical equations CH-1015 Lausanne, Vaud, Switzerland which describe the movement and behaviour of trapped European Space Agency - ESA/ESAC, Camino Bajo del Castillo, 28692 Villanueva de la Cañada, Madrid, Spain particles. The trapped particles dynamics have already European Space Agency - ESA/ESTEC, Keplerlaan 1, 2201 AZ been well explained in the seventies (Roederer, 1970) as Noordwijk, Netherlands 4 well as in more recent studies using the Van Allen Probes European Space Agency - ESA/ESOC, Robert-Bosch-Strasse 5, D-64293 Darmstadt, Germany showing the presence of a temporary third trapped radi- Preprint submitted to Advances in Space Research July 26, 2019 arXiv:1907.11029v1 [astro-ph.HE] 25 Jul 2019 ation belt (Boyd et al, 2018). The general shape of the to days. VAB is relatively stable except for the occasional appear- Another example of a global model is the Global Radi- ance of the third temporary belt following strong events in ation Earth ENvironment (GREEN) model (Sicard et al, the geomagnetic sphere. However, in practice, the outer- 2018). GREEN is a global model using various global and boundary of the outer electron radiation belt is not con- local models to obtain the most reliable value at each point stant in time (see Figure 4). The VAB can grow rapidly in space. It is very good for detailed local simulations. For when solar eruptions reach the Earth magnetosphere and a global 3D view of the VAB, GREEN might not be the then deflate through various trapped particle losses act- easiest model to use for long term forecasts either. ing at different temporal and spatial scales (Baker et al, NASA’s AE8min/max and AP8min/max models 2018; Vassiliadis et al, 2014; Thorne, 2010). These factors (Sawyer and Vette, 1976; Vette, 1991a) for electrons and make it difficult to simulate the outer part of the trapped protons, respectively, are the most well-known and have electron belt. been used since the 1970s. More recently, the new In this paper, we use radiation flux measurements IRENE-AE9/AP9 models have been issued (O’Brien et al, recorded by radiation monitors on board two spacecraft, 2017; Johnston et al, 2015). In these models, the trapped INTEGRAL and XMM-Newton. Both have been on charged particle populations are treated in McIlwain’s highly elliptical orbits around the Earth for respectively (B, L) or (α , L ) coordinate system (McIlwain, 1961). eq 16 and 19 years probing the Southern and Northern hemi- This system tries to use the symmetries inherent to spheres, respectively. Using these measurements, we con- trapped particles behaviour in a magnetosphere in order struct a dynamical 3D volume model for the VAB based on to increase the measurement sampling and, consequently, the electron radiation detected in the 0.7–1.75 MeV energy have a better statistical validity. This results in a 2D view range. of the Radiation Belts, where only a section of the radia- tion torus is visible. The Electron Slot Region Radiation There are currently a few global dynamical models Environment Model (Sandberg et al, 2014) is the closest for the VAB but they do not easily meet the needs to the model presented in this work, however it also uses that long term planning requires for scientific missions the (α , L ) coordinate system. eq such as INTEGRAL or XMM-Newton. For example, the British Atlantic Survey - Radiation Belt model (BAS- A data-driven model based purely on measurements in RBM) (Glauert et al, 2014) is a global dynamic model that the Cartesian 3D space around the Earth will naturally simulates the high energy electron population (>500 keV) include these asymmetries. The best reference frame for of the radiation belts taking into account effects such as the such a model is one in which the global structure of the changing solar activity and wave-particle interactions. It belts is mostly static. This is the case in the Solar Mag- is mainly based on the Fokker-Planck equation and satel- netic (SM) reference frame where the z-axis is parallel to lite data. This model is very good for simulations of past the Earth’s magnetic dipole (11 degrees tilt with respect VAB states and detailed forecasts on time scales of hours to its rotation axis) with its positive direction towards the INTEGRAL XMM-Newton Figure 1: XMM-Newton and INTEGRAL complementary orbits in the Solar Magnetic (SM) frame. The orbits are not elliptical due to the daily oscillation and yearly rotation of the SM frame with respect to the Geocentric Equatorial Inertial (GEI) frame. The Earth is plotted to scale as a sphere in the centre, and the VAB are shown as the colour-shaded regions giving a qualitative idea of the radiation intensity: the blue corresponds to a low radiation density region and the red is a high density region. Positive x always points towards the Sun. 2 northern hemisphere (south magnetic pole); the x-axis is km, something that could be addressed in the future by defined in the plane given by the z-axis and the Earth-Sun including LEO measurements in the construction of the line with the positive direction towards the Sun; and the model. y-axis is defined to have an orthogonal system. Trajecto- The INTEGRAL Radiation Environment Monitor, ries plotted in this frame will not be elliptical due to the IREM (Hajdas et al, 2003) and the XMM-Newton Euro- daily oscillation and yearly rotation of the SM frame with pean Photon Imaging Camera Radiation Monitor, EPIC respect to the Geocentric Equatorial Inertial (GEI) frame. RM or ERM (Boër et al, 1995) gather in-situ radiation data continuously along the spacecraft’s orbit. Remark- The model presented in this paper uses a new and sim- ably, not only have these two spacecraft gathered more pler approach to visualize the VAB by using the SM 3D reference frame to process the data, create, and use the than 16 years of contemporaneous radiation measure- model. It implies a 3D view of all possible asymmetries ments, but as is shown in Figure 1, their orbits scan dif- detected in the belts. As seen in the xz-section view of ferent parts of the VAB and are thus complementary in the static VAB model presented here (see Figure 3(d)), the increasing the coverage of the belts. belt appears to have a stronger radiative core on the night Table 1 shows the main characteristics of the two ra- side (negative x-direction) and a slightly broader section diation monitors, IREM and ERM, which are in fact sig- on the day side (positive x-direction). This justifies the use nificantly different instruments. One needs to be care- of the 3D SM system instead of McIlwain’s. The caveat of ful when combining the data of both missions. Cross- using the full 3D space in the SM reference frame resides calibration is necessary. The calibrated radiation flux mea- in the sampling of the 3D space which in that case is nat- surements are available through the ESA Open Data In- urally lower than with McIlwain’s reference system. The terface (ODI) server using a simple Python client. The accumulation of 18 years of data in addition to the com- ESA ODI provides ready-to-use space environment data plementarity of INTEGRAL’s and XMM-Newton’s orbital from several different missions, including differential om- configurations (see Figure 1) allow for a global coverage of nidirectional and distinct electron and proton fluxes in the main parts of the VAB as explained in Section 3. different energy ranges for IREM (Mohammadzadeh et al, This paper presents a new, empirical data-driven model 2003; Sandberg et al, 2012) and ERM. of the Earth’s radiation belts denoted 5DRBM-e, where The on-board radiation monitors are intended to trig- 5D represents the model’s five dimensions (three for the ger the shutdown and thus protection of the scien- spatial position, one for time, and one for the intensity of tific payload instruments in case of excessive radiation (Gonzalez-Riestra and Rodriguez-Pascual, 2018). In ev- radiation); RBM stands for Radiation Belts’ Model; and ery revolution, an instrument window is defined during -e stands for electrons. Section 2 presents the basic charac- which the radiation level is expected to be low enough teristics of the radiation monitors from which the data are to use the instruments. The endpoints of this instrument taken, Section 3 describes how the volume model is built, window are predicted based on a simple model of the radia- and Section 4 shows how it can be applied in practice, and tion environment. The safety of scientific payload relies on how it compares to the AE8min/max models. ensuring instruments are operating in safe, low-radiation conditions. The monitors ensure that during the instru- 2. The radiation monitors ment window, the radiation remains below the operational threshold. Modelling radiation accurately is important for There are currently two active ESA missions that cross the safety and lifetime of these instruments, but it also the VAB sampling almost all their structures. They are allows for smoother science operations by relieving the XMM-Newton and INTEGRAL, which were launched re- burden of having to perform re-activation sequences for spectively in 1999 and 2002. These crossings along the 2-3 instruments following an emergency shutdown caused by days orbits of the spacecraft allow for a scanning of the belts from their outer boundaries down to approximately 2,500 km above the Earth’s surface with INTEGRAL data. The ESA Open Data Interface can be accessed directly at This, however, limits reliability at altitudes below 3,000 https://spitfire.estec.esa.int/trac/ODI/wiki/ODIv5 Table 1: Main characteristics of the on-board Si scintillators radiation monitors. The electron energy range corresponds to the one available on the ESA Open Data Interface (ODI) server. Radiation monitors INTEGRAL XMM-Newton Side shield 4.2mm Ta & 5mm Al 5mm Al Front shield 0.65mm Al 0.02mm Be ◦ ◦ FOV/Opening ±20 conical 3mm , 1sr, ±32.8 Thickness 0.5mm Si 0.5mm Si 2 2 Active Area 25mm 85mm e Energy Range 0.65MeV - 2.18MeV 0.13MeV - 1.87MeV 3 unpredicted high radiation. XMM-Newton at [0.7, 0.825, 0.985, 1.2, 1.41, 1.58, 1.75] For XMM-Newton, the prediction is based on a 3D MeV. Hence, an integration from 0.7 to 1.75 MeV is per- surface model of the outer boundary of the electron formed to have one homogeneous data point per position belt developed in 2004 by Mauro Casale and Jorge and time-stamp. Fauste of the XMM-Newton science operations centre The data are then cleaned through simple sigma clip- (Casale and Fauste, 2004). This model uses the first years ping and smoothing to produce a data set with no major of the mission to estimate the outer boundary of the main outliers, which can result from faulty measurements, but electron radiation belt. It works well and is still used today also from a wrong calibration process due to, for exam- for the science payload mission planning. But because it ple, changing in sampling rates (1–2 wrong successive data was tailor-made for XMM-Newton, this model cannot be points, 2 times per orbit for XMM-Newton). Smoothing applied to another mission. In addition, because it is a the data helps to remove the very short variations (of the surface model that gives estimates of the boundary of the order of tens of minutes) or data spikes that we don’t want outer shell of the electron belt, it does not provide radia- to take into account in this model. tion profiles along the length of the crossings through the Positions are converted to the SM reference frame. The belts. data sampling rate is standardised to one measurement For INTEGRAL, the safe belt entry and exit altitudes per minute (averaging the data points every minutes if the are evaluated every month based on the measurements sampling rate is higher) in order to have the same spatial of the two previous years. A simple sinusoidal fit is ap- density of points for both satellites along their trajecto- plied to the measured altitude at a fixed radiation level. ries in the radiation belts. Because the radiation monitors Based on this fit and accounting for some margins, the on each spacecraft are different, radiation measurements available window for observations results from these de- have to be cross-calibrated. Even if the measurements are fined entry and exit altitudes at which the instruments in some manner calibrated on the ODI server, this cross- are switched off and on. Given the orbit’s stability, this calibration is necessary. solution works very well in terms of instrument safety and The cross-calibration is performed separately for the ease of operations. However, due to smoothing of shorter highest and lowest radiation levels that correspond to the term dynamic radiation behaviour, scientific observation inside and outside of the belts, respectively. At the low time is lost. Moreover, the predictions can be unreliable end, the cross-calibration is done by comparing XMM- when the spacecraft crosses different parts of the belts, Newton and INTEGRAL radiation measurements made at something that became more common following the orbit the same time far away from the VAB. This implies assum- changes in 2015 (see Section 4.1). A global 3D volume ing that outside of the VAB the electron flux is isotropic, model could predict the different altitudes for successive which is not strictly accurate, but since the goal is to focus revolutions and thus yield more reliable estimates. The on the belts themselves, this assumption is good enough limitations of the current methods used to ensure safe op- because the radiation levels outside the VAB are orders erations on XMM-Newton and INTEGRAL were the main of magnitudes lower than inside. At the high end, the drivers for the work presented here. cross-calibration is performed by first creating indepen- dent XMM-Newton and INTEGRAL models of the VAB, and then comparing the radiation at corresponding points 3. Building a time-dependent volume model of the in the two mission-specific volume models. In this cross- Van Allen belts calibration process, IREM is taken as the baseline because A reliable simulation of the radiation environment the data show a much smoother behaviour and less scat- around the Earth requires a global understanding of the tering than for ERM. These two cross-calibrations at the radiation belts’ dynamics and the amount of data accumu- low and high ends result in a conversion function to be lated using INTEGRAL and XMM-Newton can help us to applied to ERM measurements. To make the transition have a phenomenological knowledge on this. The 5DRBM- between the two regimes smooth, the transition regions e model is built in a modular way from a static model over are smoothed using local spatial averaging. which time-dependent functions can be applied to account 3.1. The static volume model for the dynamics. Time-dependent deviations from the static model are quantified using the InterQuartile Range After cross-calibration, all data points are plotted in (IQR) of radiation flux measurements in cells over the vol- the SM frame. The 3D grid in units of Earth radii, R , ume of the belts, as defined in the model. The IQRs can is defined on a regular mesh of 0.1 R . The volume of the help to understand where the VAB are strongly variable, grid is centered on the Earth, 14 R in the z direction, and and computing them for different time scales can give an 32 R in both the x and y directions in order to encompass idea of the time evolution of the belts’ general shape. All the differential omnidirectional electron fluxes with The energy binning for each mission is different, and thus an their corresponding positions and time-stamps are taken interpolation in the electron energy spectrum is done to have the from the ODI server: these are measured for INTEGRAL same bins edges resulting in an overlapping range from 0.7 to 1.75 at [0.7, 0.78, 1.125, 1.27, 1.435, 1.615, 1.75] MeV and for MeV instead of 0.65 to 1.87 MeV one can see in Table 1. 4 Figure 2: The left plot shows the real tracks in the x − z plane of the SM frame, INTEGRAL (upwards) and XMM-Newton (downwards), colour-coded with the cross-calibrated flux. The right figure pictures the duplicated tracks after the mirroring step. the whole Van Allen Belts. Knowing the precise position of Figure 3 illustrates the steps used in the construction each measurement within the 3D volume, the value in each of the static model. Values in the resulting 3D matrix are node of the grid can be computed using the nearest data smoothed to obtain the final static volume model of the points. In order to have a full coverage of this volume, all trapped electron belts. A short 3D animation, showing tracks are mirrored with respect to the magnetic equato- the creation of the static model, has been selected for the rial plane, as shown in Figure 2. This assumes an up/down ESA INTEGRAL Picture Of the Month (POM) in July symmetry in the SM reference frame, which is commonly 2018 and at the same time for the NASA High Energy assumed in magnetospheric studies. Time dependencies Astrophysics Picture Of the Week (HEAPOW) . can be added later to take into account up/down asymme- It is this static model that constitutes the basis for the tries. more realistic time-dependent model for which the most Mirroring of the tracks increases the number of nodes important element is the variation due to the solar cycle. in which an average radiation value can be computed. This does not, however, cover the entire volume. Looking at a 3.2. The time dependence 3D rectangle of size 16×16×6 R centered on the Earth, The solar cycle on timescales of a few years is very well covering the main VAB regions below 45,000 km altitude, described by a simple sinusoid with an 11-year period. A 53% of all cells of size 0.1 R contain at least 1 data point, simple sinusoidal fit to the logarithm of the radiation flux and half contain 3 or more data points. This is not enough measured outside the VAB during the past 18 years, com- to build an accurate model down to a resolution of 0.1 R . bining XMM-Newton and INTEGRAL data, describes the Consequently, the remaining voids (nodes without an as- long term variation of the background radiation. The pa- signed value) in the grid need to be filled by iteratively rameters of this fit are determined during the elaboration using the nearest neighbouring nodes. Four iterations are of the model, and can then be used to vary the mean back- performed to fill the full 3D grid. An empty node value is ground radiation of the static model as a function of time found computing the median of all the neighbouring non- 7 when the model is used for simulations or predictions. zero nodes in a 0.8 R box size. In the 3D rectangle of size 16 × 16 × 6 R centered on the Earth, more than 95% of all cells of size 0.8 R contain at least 1 data point INTEGRAL POM July 2018: https://www.cosmos.esa.int/ with a median population per cell of more than 500, which web/integral/pom-archive is enough for a good statistical sampling. HEAPOW 2nd of July 2018: https://heasarc.gsfc.nasa.gov/ docs/objects/heapow/archive/solar_system/vab_integral.html No external parameters are used to include this solar cycle in The box size was optimized as a function of the required resolu- the model. Future work could include a more elaborate fitting of the tion and the processing time. solar cycle, or the use of external solar cycle variables. 5 (a) Real tracks in the SM frame colour-coded with the radiation flux (b) Radiation fluxes assigned to nodes of the 3D grid (c) Filled 3D grid using neighbouring nodes (d) Smoothed 3D grid resulting in the finals 3D static model Figure 3: Main building steps of the 3D static model using real data points. The left side of each figure is the x − y plane top view and the right side is the x − z plane side view. Everything is colour-coded with the cross-calibrated electron flux. The 3D grid has a size of 32x32x14 Earth radii. 6 Figure 4: Interquartile ranges 3D map projections. The left side is the top view of the x − y plane, and the right side is the side view of the x − z plane. High dispersion regions are located mainly at the boundary of the outer belt. The 3D grid has a size of 32x32x14 Earth radii. The background radiation is higher during lower solar ac- 4. Validating the 5DRBM-e volume model tivity and lower during higher solar activity periods. This 4.1. A self-consistency check is because this background radiation is mainly composed of galactic cosmic-rays whose density is strongly modulated A first step is a self-consistency check where the model by the solar cycle in an inversely proportional manner. predictions are compared to the actual data. Naturally, The influence of the solar cycle on the VAB is much more the modelled radiation profile should be consistent with complex and has not, at this stage, been implemented. the data. This comparison is shown in Figure 5 where it is The confidence on the model is estimated via the IQR seen that the model is not only consistent with the data, in 3D cells of 1 R . The IQR corresponds to a measure but that it accurately predicts radiation levels where there of dispersion of a data sample. In each cell, the median are data gaps (black regions in the top plot). radiation level of all data points is computed in order to As mentioned in Section 2, in order to avoid damaging split the values in the cell into two subsamples: one with their radiation-sensitive detectors, spacecraft like XMM- the values below the median, and the other with values Newton and INTEGRAL must turn off the scientific pay- above the median. For each subsample the medians are load instruments before entering and turn them back on computed, and the IQR for the cell is given by the absolute only after having exited the radiation belts. Another check difference between the medians of each subsample. The that can be done is to compare the model-predicted entry confidence interval (uncertainty envelope) on the radiation and exit altitudes to those measured. This is shown in profiles is computed from the IQRs across the modelled Figure 6 where measured altitudes are in blue, and pre- volume. Figure 4 shows the x − y and x − z planes of dicted values are in red. Long-term trends and seasonal the IQR map. Looking at Figure 3(d) it can be seen that variations are well captured, but not quite the peak ampli- the largest dispersions are along the boundary of the VAB tudes. The main reason is the absence of time variations where the effects of the solar wind and solar activity are in the static model which smooths out the general shape expected to be the strongest. The IQRs could also be used of the belts. The cross-calibration is also not perfect. The to explore large time-dependent variations where values split of the single red dotted curve into three in early 2015 are the greatest, and checking time variations in each cell. is the consequence of the change in INTEGRAL’s orbital This will be included in future work. period from 3 to 2.66 days following an orbital adjust- ment (Dietze et al, 2015). With an integer orbital period (3 days), INTEGRAL was scanning a similar part of the belt in each orbit. With the 2.66-day orbit, it needs 8 days 3.3. The 5DRBM-e model (3 orbits) to come back to a similar region of the belt, and hence the three curves. The result is the 5DRBM-e model for trapped 0.7–1.75 4.2. Validation on GPS data MeV electrons. It can be used with any orbit at any epoch to produce a predictive model of the radiation along the Actual validation must be done on data that have not orbit as a function of the solar cycle. The model is con- been used to construct the model. If, in addition, the re- tained in a single FITS file that includes the static model, sult can be compared to the predictions of another model, the sinusoidal fit parameters corresponding to the time- then it is ideal. The GNSS satellites orbits at approxi- dependent solar cycle variations, and the IQR dispersion mately 20,000 km altitude are constantly inside the VAB, which can be used to calculate uncertainties and confi- and thus subject to constant radiation. This makes the ac- dence intervals for any radiation profile. curate prediction of the radiation environment crucial to 7 Figure 5: INTEGRAL mission radiation profiles between 2003 and 2015. Each vertical line corresponds to one revolution with the orbital phase 0 being perigee point. The top figure shows the real radiation data measurements. The second panel shows the radiation intensity extracted from the 5DRBM-e static model along INTEGRAL’s trajectory. Figure 6: Belt entry and exit altitudes for each revolution for the whole INTEGRAL mission and for the expected future trajectory until the planned de-orbiting in 2029. The blue curves are the real measurements, and the red dots are the computed altitudes form the model based on a simple threshold on the electron counts. 8 Figure 7: Radiation profiles for the GPS-2R 12 (Navstar 60, USA 178) satellite. The top figure shows the comparison of the real (blue), the AE8min (dark green), the AE9mean (light green) and the 5DRBM-e (red) radiation data during a low solar activity period. The bottom figure shows the radiation profile for the same satellite but for a high solar activity period and corresponding the AE8max model. The red-shaded region around the red model curve is the 95% confidence interval. the success of any satellite on such orbits. The AE8 and and AE9 models are generally within the 95% confidence AP8 models are widely used for these particular regions of interval (red-shaded region) of the 5DRBM-e model. space as well as the newest AE9 model. Hence, using the data from a GPS satellite and comparing the output of 5. Conclusion the 5DRBM-e model to that of the AE8 and AE9 models constitutes an ideal validation test. Modelling these still rather poorly understood radia- Figure 7 shows a comparison between the static tion belts is important for space-borne activities (manned 5DRBM-e, the AE8min-AE8max and the AE9mean mod- and unmanned) in the near-Earth environment. The ex- els for the GPS-2R 12 (Navstar 60, USA 178) satellite. ploration of predictive methods used to ensure safe oper- The measured radiation data and satellite positions for ations on XMM-Newton and INTEGRAL, together with the dates shown in the figures are extracted from the ESA the benefit of creating global VAB models with great po- ODI server. From these positions, a SPENVIS readable tentials, were the driving motivations for the work pre- trajectory file is produced in order to compute the AE8 and AE9 modelled radiation data. This GPS satellite is sented here. on a very low eccentricity orbit with an average altitude The presented 5DRBM-e model has been built using of 20,200 km and inclination of 55 degrees. Both AE8min INTEGRAL (>16 years) and XMM-Newton (>18 years) (during a solar minimum activity) and AE8max (during a radiation data, crossing the Van Allen Belts in each revo- solar maximum activity) as well as the AE9mean are com- lution. This newly built data-driven model focuses on the pared to the static 5DRBM-e. The GPS data, the AE8 electron belt from approximately 4,000 km up to its outer boundary at around 60,000 km. The radiation flux given 11 by the 5DRBM-e model corresponds to the integrated elec- The AE8 and AE9 models data are taken from the ESA Space Environment Information System (SPENVIS). tron flux in the energy range from 0.7 to 1.75 MeV, the 9 overlapping energy range for the radiation monitors on- The end goal of this project is to provide robust, reli- board the two spacecraft. Remarkably, not only have these able, easy-to-use, data-driven, dynamic, 3D electron and two spacecraft gathered more than 16 years of contempo- proton radiation belt volume models that can be used in raneous radiation measurements, but their orbits scan dif- the space science and engineering community for design- ferent parts of the VAB and are thus complementary in ing, preparing, and running space missions. increasing the coverage of the belts allowing the creation of a relevant global static model. Acknowledgements A reliable prediction of the radiation environment around the Earth requires an understanding of the radi- LM would like to thank the Swiss Space Centre (SSC), and ation belts’ dynamics. Such knowledge is today surpris- the Swiss Space Office (SSO) for the financial support as a ingly limited. In this work, the time-dependent deviations Swiss National Trainee at the European Space Astronomy from the static model are quantified using the interquar- Centre (ESAC/ESA). tile range of radiation flux measurements in 3D cells over the entire volume defined in the model. The IQR gives an excellent idea of the model’s uncertainties on the radiation References intensities, allowing, for example, the computation of con- Baker, D. N., Erickson, P. J., Fennell, J. F. et al., 2018, Space fidence intervals for model-derived quantities such as the Weather Effects in the Earth’s Radiation Belts, Space Sci Rev altitudes at entry and exit points. This first static ver- 214: 17. sion of the 5DRBM-e with dynamic background radiation Boër, M., Naya, J., Chabaud, J.-P. et al., 1995, The CESR Multi- mission Radiation Monitor, IEEE Transactions on Nuclear Sci- modelling, following the solar cycle, shows promising re- ence, Vol. 42, No. 6. sults with respect to the well-known AE8min and AE8max Boyd, A. J., Turner, D. L., Reeves, G. D., et al., 2018, What causes as well as AE9mean models. In addition, a more accurate radiation belt enhancements: A survey of the Van Allen Probes Era, GeophysicalResearch Letters, 45, 5253-5259. modelling of the VAB enhances the predictions of the belts Casale, M., and Fauste, J., 2004, Attempt of Modelling the entry and exit times which contributes to maximize the ob- XMM-Newton Radiation Environment, Technical Note, XMM- servation time and increase the safety during instrument OPS-TN-004, http://xmm2.esac.esa.int/external/xmm_sched/ operation. The simplicity of the 5DRBM-e model and its vischeck/Rad_Model_TN.pd. Dietze, C., Vasconcelos, A., Ziegler, G. et al., 2015, INTEGRAL use of the Solar Magnetic reference frame make it easy End-Of-Life Disposal Manoeuvre Campaign, Proceedings of 25th to visualize in 3D space. Moreover, the model has been International Symposium on Space Flight Dynamics ISSFD, Mu- structured in a step-by-step way with very few dependen- nich. cies from one step to the another which makes it easy to Glauert, S. A., R. B. Horne, and N. P. Meredith, 2014, Three- dimensional electron radiation belt simulations using the BAS Ra- update and modify. If new data are available, a new up- diation Belt Model with new diffusion models for chorus, plasma- dated and improved model can be built in a few hours spheric hiss, and lightning-generated whistlers, J. Geophys. Res. only. This versatility allows for the creation of models Space Physics, 119, 268-289. based on specific data sets with very few modifications to Gonzalez-Riestra, R. and Rodriguez-Pascual, P.M., 2018, The Be- haviour of the XMM-Newton Background: From the beginning the procedure. of the mission until May 2018, XMM-SOC-GEN-TN-0014, issue The intention is to extend this model to include more 3.11, http://xmm2.esac.esa.int/docs/documents/GEN-TN-0014. dynamic features associated with the 11-year solar cycle, pdf. as well as yearly variations that are observed in the data. Hajdas, W., Bühler, P., Eggel, C., et al., 2003, Radiation environ- ment along the INTEGRAL orbit measured with the IREM mon- Short-term variations (days, weeks, months), mainly in- itor, A&A 411, L43-L47. fluenced by short-term solar events, will not be considered Johnston, W. R., O’Brien, T. P., Huston, S. L.et al, 2015, Recent as many existing models already accomplish this, as out- Updates to the AE9/AP9/SPM Radiation Belt and Space Plasma Specification Model IEEE Transactions on Nuclear Science, Vol. lined previously. The electron energy range will also be in- 62, No. 6. creased by using more energy bins from INTEGRAL and McIlwain, C. E., 1961, Coordinates for Mapping the Distribution of XMM-Newton and possibly from other missions. Using Magnetically Trapped Particles, J. Geophys. Res., 66, pp. 3681- more energy bins will require a good knowledge and ex- Mohammadzadeh, A., Evans, H., Nieminen, P. et al, 2003, The ESA trapolation of the energy spectrum seen by both IREM Standard Radiation Environment Monitor Program First Results and ERM instruments at each time-stamp, in order to From PROBA-I and INTEGRAL IEEE Transactions on Nuclear compare the same integrated flux. The expected energy Science, Vol. 50, No. 6. range should start at 0.7 MeV and go up to 2–3 MeV. More O’Brien, T. P., Johnston, W. R., Huston, S. L. et al, 2017, Changes in AE9/AP9-IRENE Version 1.5 IEEE Transactions on Nuclear importantly, a 5DRBM-p model will be constructed based Science, Vol. 65, No. 1. on the proton radiation measurement data also available Roederer, J. G., 1970, Dynamics of Geomagnetically Trapped Radi- from the ODI server. A first model will be built using ation, Physics and Chemistry in Space, Berlin: Springer. only proton fluxes measured by INTEGRAL, because it Sandberg, I., Daglis, I. A., Anastasiadis, A. et al, 2012, Unfolding and Validation of SREM Fluxes, IEEE Transactions on Nuclear has flown with a lower perigee altitude resulting in bet- Science, Vol. 59, No. 4. ter coverage of the inner proton belt. The proton energy Sandberg, I., Daglis, I. A., Heynderickx, D. et al, 2014, Development range will probably start at a few MeV up to several tens and Validation of the Electron Slot Region Radiation Environment of MeV. Model, IEEE Transactions on Nuclear Science, Vol. 61, No. 4. 10 Sawyer, D. M., and J. I. Vette, 1976, AP-8 Trapped Proton Environ- ment for Solar Maximum and Solar Minimum, NSSDC/WDC-A- R&S 76-06. Sicard, A., Boscher, D., Bourdarie, S. et al., 2018, GREEN: the new Global Radiation Earth ENvironment model (beta version). Annales Geophysicae, European Geosciences Union, 2018, 36, pp.953-967. 10.5194/angeo-36-953-2018 . hal-02126965. Stormer, C., 1937, On the Trajectories of Electric Particles in the Field of a Magnetic Dipole with Applications to the Theory of Cosmic Radiation. Sixth Communication. With 17 Figures in the Text, Astrophisica Norvegica, v. 2, p.193. Thorne, R. M., 2010 Radiation belt dynamics: The importance of wave-particle interactions Geophysical research letters, Vol. 37, L22107. Van Allen, James A., 1959, Radiation Belts Around the Earth, Sci. Am., Vol. 200, pp. 39-47, IRE Student Quarterly, pp. 4-15. Vassiliadis D., Klimas A., Fung S. et al., 2004, Structure and Dy- namics of the Outer Radiation Belt. In: Daglis I.A. (eds) Effects of Space Weather on Technology Infrastructure. NATO Science Series II: Mathematics, Physics and Chemistry, vol 176. Springer, Dordrecht. Vette, J. I., 1991a, The AE-8 Trapped Electron Model Environment, NSSDC/WDC-A-R&S 91-24. Walsh, B. M., Kuntz, K. D., Collier, M. R., et al., 2014, Ener- getic particle impact on X-ray imaging with XMM-Newton, Space Weather, 12, 387-394. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Astrophysics arXiv (Cornell University)

Loading next page...
 
/lp/arxiv-cornell-university/data-driven-modelling-of-the-van-allen-belts-the-5drbm-model-for-QzeDggnpwc
ISSN
0273-1177
eISSN
ARCH-3330
DOI
10.1016/j.asr.2019.07.036
Publisher site
See Article on Publisher Site

Abstract

The magnetosphere sustained by the rotation of the Earth’s liquid iron core traps charged particles, mostly electrons and protons, into structures referred to as the Van Allen belts. These radiation belts, in which the density of charged energetic particles can be very destructive for sensitive instrumentation, have to be crossed on every orbit of satellites traveling in elliptical orbits around the Earth, as is the case for ESA’s INTEGRAL and XMM-Newton missions. This paper presents the first working version of the 5DRBM-e model, a global, data-driven model of the radiation belts for trapped electrons. The model is based on in-situ measurements of electrons by the radiation monitors on board the INTEGRAL and XMM-Newton satellites along their long elliptical orbits for respectively 16 and 19 years of operations. This model, in its present form, features the integral flux for trapped electrons within energies ranging from 0.7 to 1.75 MeV. Cross-validation of the 5DRBM-e with the well-known AE8min/max and AE9mean models for a low eccentricity GPS orbit shows excellent agreement, and demonstrates that the new model can be used to provide reliable predictions along widely different orbits around Earth for the purpose of designing, planning, and operating satellites with more accurate instrument safety margins. Future work will include extending the model based on electrons of different energies and proton radiation measurement data. Keywords: Van Allen belts, Radiation belt modelling, Trapped particles, Radiation Environment, Space environment, Space weather 1. Introduction Mr. James Van Allen, payload specialist of the Explorer missions, to discover successively: the inner proton belt Only theorised before space exploration began (Explorer 1 & 3), the inner electron belt (Explorer 4) (Stormer, 1937), the Earth Radiation Belts were discov- and the outer electron belt (Pioneer 3) (Van Allen, 1959). ered for the first time in 1958 with the very first US satel- Consequently, these high radiation regions surrounding the lite, Explorer 1. These first in-situ measurements allowed Earth are known as the Van Allen Belts (VAB). These radiation belts are the result of the complex interaction between the Earth’s magnetosphere and the Corresponding author interplanetary medium mainly driven by the solar wind Email addresses: lionel.metrailler@esa.int (Lionel Métrailler), guillaume.belanger@esa.int (Guillaume Bélanger), that moves energetic charge-carrying electrons and pro- peter.kretschmar@esa.int (Peter Kretschmar), tons with it. These particles are trapped in the mag- erik.kuulkers@esa.int (Erik Kuulkers), netosphere forming high radiation torus-shaped regions Ricardo.Perez.Martinez@esa.int (Ricardo Pérez Martínez), jan.uwe.ness@esa.int (Jan-Uwe Ness), pedro.rodriguez@esa.int around the Earth’s magnetic axis. The Van Allen Belts (Pedro Rodriguez), mauro.casale@esa.int (Mauro Casale), extend from altitudes from ∼1,000km (0.2 R ) to more jorge.fauste@esa.int (Jorge Fauste), timothy.finn@esa.int than ∼60,000km (10 R ). (Timothy Finn), celia.sanchez@esa.int (Celia Sanchez), thomas.godard@esa.int (Thomas Godard), Modelling these highly dynamic radiation belts is im- richard.southworth@esa.int (Richard Southworth) portant for space-borne activities, both manned and un- Swiss National Trainee Program, Swiss Space Center, École manned, in the near-Earth environment. Theoretical mod- Polytechnique Fédérale de Lausanne (EPFL), PPH 338, Station 13, els of the belts can be built on simplified physical equations CH-1015 Lausanne, Vaud, Switzerland which describe the movement and behaviour of trapped European Space Agency - ESA/ESAC, Camino Bajo del Castillo, 28692 Villanueva de la Cañada, Madrid, Spain particles. The trapped particles dynamics have already European Space Agency - ESA/ESTEC, Keplerlaan 1, 2201 AZ been well explained in the seventies (Roederer, 1970) as Noordwijk, Netherlands 4 well as in more recent studies using the Van Allen Probes European Space Agency - ESA/ESOC, Robert-Bosch-Strasse 5, D-64293 Darmstadt, Germany showing the presence of a temporary third trapped radi- Preprint submitted to Advances in Space Research July 26, 2019 arXiv:1907.11029v1 [astro-ph.HE] 25 Jul 2019 ation belt (Boyd et al, 2018). The general shape of the to days. VAB is relatively stable except for the occasional appear- Another example of a global model is the Global Radi- ance of the third temporary belt following strong events in ation Earth ENvironment (GREEN) model (Sicard et al, the geomagnetic sphere. However, in practice, the outer- 2018). GREEN is a global model using various global and boundary of the outer electron radiation belt is not con- local models to obtain the most reliable value at each point stant in time (see Figure 4). The VAB can grow rapidly in space. It is very good for detailed local simulations. For when solar eruptions reach the Earth magnetosphere and a global 3D view of the VAB, GREEN might not be the then deflate through various trapped particle losses act- easiest model to use for long term forecasts either. ing at different temporal and spatial scales (Baker et al, NASA’s AE8min/max and AP8min/max models 2018; Vassiliadis et al, 2014; Thorne, 2010). These factors (Sawyer and Vette, 1976; Vette, 1991a) for electrons and make it difficult to simulate the outer part of the trapped protons, respectively, are the most well-known and have electron belt. been used since the 1970s. More recently, the new In this paper, we use radiation flux measurements IRENE-AE9/AP9 models have been issued (O’Brien et al, recorded by radiation monitors on board two spacecraft, 2017; Johnston et al, 2015). In these models, the trapped INTEGRAL and XMM-Newton. Both have been on charged particle populations are treated in McIlwain’s highly elliptical orbits around the Earth for respectively (B, L) or (α , L ) coordinate system (McIlwain, 1961). eq 16 and 19 years probing the Southern and Northern hemi- This system tries to use the symmetries inherent to spheres, respectively. Using these measurements, we con- trapped particles behaviour in a magnetosphere in order struct a dynamical 3D volume model for the VAB based on to increase the measurement sampling and, consequently, the electron radiation detected in the 0.7–1.75 MeV energy have a better statistical validity. This results in a 2D view range. of the Radiation Belts, where only a section of the radia- tion torus is visible. The Electron Slot Region Radiation There are currently a few global dynamical models Environment Model (Sandberg et al, 2014) is the closest for the VAB but they do not easily meet the needs to the model presented in this work, however it also uses that long term planning requires for scientific missions the (α , L ) coordinate system. eq such as INTEGRAL or XMM-Newton. For example, the British Atlantic Survey - Radiation Belt model (BAS- A data-driven model based purely on measurements in RBM) (Glauert et al, 2014) is a global dynamic model that the Cartesian 3D space around the Earth will naturally simulates the high energy electron population (>500 keV) include these asymmetries. The best reference frame for of the radiation belts taking into account effects such as the such a model is one in which the global structure of the changing solar activity and wave-particle interactions. It belts is mostly static. This is the case in the Solar Mag- is mainly based on the Fokker-Planck equation and satel- netic (SM) reference frame where the z-axis is parallel to lite data. This model is very good for simulations of past the Earth’s magnetic dipole (11 degrees tilt with respect VAB states and detailed forecasts on time scales of hours to its rotation axis) with its positive direction towards the INTEGRAL XMM-Newton Figure 1: XMM-Newton and INTEGRAL complementary orbits in the Solar Magnetic (SM) frame. The orbits are not elliptical due to the daily oscillation and yearly rotation of the SM frame with respect to the Geocentric Equatorial Inertial (GEI) frame. The Earth is plotted to scale as a sphere in the centre, and the VAB are shown as the colour-shaded regions giving a qualitative idea of the radiation intensity: the blue corresponds to a low radiation density region and the red is a high density region. Positive x always points towards the Sun. 2 northern hemisphere (south magnetic pole); the x-axis is km, something that could be addressed in the future by defined in the plane given by the z-axis and the Earth-Sun including LEO measurements in the construction of the line with the positive direction towards the Sun; and the model. y-axis is defined to have an orthogonal system. Trajecto- The INTEGRAL Radiation Environment Monitor, ries plotted in this frame will not be elliptical due to the IREM (Hajdas et al, 2003) and the XMM-Newton Euro- daily oscillation and yearly rotation of the SM frame with pean Photon Imaging Camera Radiation Monitor, EPIC respect to the Geocentric Equatorial Inertial (GEI) frame. RM or ERM (Boër et al, 1995) gather in-situ radiation data continuously along the spacecraft’s orbit. Remark- The model presented in this paper uses a new and sim- ably, not only have these two spacecraft gathered more pler approach to visualize the VAB by using the SM 3D reference frame to process the data, create, and use the than 16 years of contemporaneous radiation measure- model. It implies a 3D view of all possible asymmetries ments, but as is shown in Figure 1, their orbits scan dif- detected in the belts. As seen in the xz-section view of ferent parts of the VAB and are thus complementary in the static VAB model presented here (see Figure 3(d)), the increasing the coverage of the belts. belt appears to have a stronger radiative core on the night Table 1 shows the main characteristics of the two ra- side (negative x-direction) and a slightly broader section diation monitors, IREM and ERM, which are in fact sig- on the day side (positive x-direction). This justifies the use nificantly different instruments. One needs to be care- of the 3D SM system instead of McIlwain’s. The caveat of ful when combining the data of both missions. Cross- using the full 3D space in the SM reference frame resides calibration is necessary. The calibrated radiation flux mea- in the sampling of the 3D space which in that case is nat- surements are available through the ESA Open Data In- urally lower than with McIlwain’s reference system. The terface (ODI) server using a simple Python client. The accumulation of 18 years of data in addition to the com- ESA ODI provides ready-to-use space environment data plementarity of INTEGRAL’s and XMM-Newton’s orbital from several different missions, including differential om- configurations (see Figure 1) allow for a global coverage of nidirectional and distinct electron and proton fluxes in the main parts of the VAB as explained in Section 3. different energy ranges for IREM (Mohammadzadeh et al, This paper presents a new, empirical data-driven model 2003; Sandberg et al, 2012) and ERM. of the Earth’s radiation belts denoted 5DRBM-e, where The on-board radiation monitors are intended to trig- 5D represents the model’s five dimensions (three for the ger the shutdown and thus protection of the scien- spatial position, one for time, and one for the intensity of tific payload instruments in case of excessive radiation (Gonzalez-Riestra and Rodriguez-Pascual, 2018). In ev- radiation); RBM stands for Radiation Belts’ Model; and ery revolution, an instrument window is defined during -e stands for electrons. Section 2 presents the basic charac- which the radiation level is expected to be low enough teristics of the radiation monitors from which the data are to use the instruments. The endpoints of this instrument taken, Section 3 describes how the volume model is built, window are predicted based on a simple model of the radia- and Section 4 shows how it can be applied in practice, and tion environment. The safety of scientific payload relies on how it compares to the AE8min/max models. ensuring instruments are operating in safe, low-radiation conditions. The monitors ensure that during the instru- 2. The radiation monitors ment window, the radiation remains below the operational threshold. Modelling radiation accurately is important for There are currently two active ESA missions that cross the safety and lifetime of these instruments, but it also the VAB sampling almost all their structures. They are allows for smoother science operations by relieving the XMM-Newton and INTEGRAL, which were launched re- burden of having to perform re-activation sequences for spectively in 1999 and 2002. These crossings along the 2-3 instruments following an emergency shutdown caused by days orbits of the spacecraft allow for a scanning of the belts from their outer boundaries down to approximately 2,500 km above the Earth’s surface with INTEGRAL data. The ESA Open Data Interface can be accessed directly at This, however, limits reliability at altitudes below 3,000 https://spitfire.estec.esa.int/trac/ODI/wiki/ODIv5 Table 1: Main characteristics of the on-board Si scintillators radiation monitors. The electron energy range corresponds to the one available on the ESA Open Data Interface (ODI) server. Radiation monitors INTEGRAL XMM-Newton Side shield 4.2mm Ta & 5mm Al 5mm Al Front shield 0.65mm Al 0.02mm Be ◦ ◦ FOV/Opening ±20 conical 3mm , 1sr, ±32.8 Thickness 0.5mm Si 0.5mm Si 2 2 Active Area 25mm 85mm e Energy Range 0.65MeV - 2.18MeV 0.13MeV - 1.87MeV 3 unpredicted high radiation. XMM-Newton at [0.7, 0.825, 0.985, 1.2, 1.41, 1.58, 1.75] For XMM-Newton, the prediction is based on a 3D MeV. Hence, an integration from 0.7 to 1.75 MeV is per- surface model of the outer boundary of the electron formed to have one homogeneous data point per position belt developed in 2004 by Mauro Casale and Jorge and time-stamp. Fauste of the XMM-Newton science operations centre The data are then cleaned through simple sigma clip- (Casale and Fauste, 2004). This model uses the first years ping and smoothing to produce a data set with no major of the mission to estimate the outer boundary of the main outliers, which can result from faulty measurements, but electron radiation belt. It works well and is still used today also from a wrong calibration process due to, for exam- for the science payload mission planning. But because it ple, changing in sampling rates (1–2 wrong successive data was tailor-made for XMM-Newton, this model cannot be points, 2 times per orbit for XMM-Newton). Smoothing applied to another mission. In addition, because it is a the data helps to remove the very short variations (of the surface model that gives estimates of the boundary of the order of tens of minutes) or data spikes that we don’t want outer shell of the electron belt, it does not provide radia- to take into account in this model. tion profiles along the length of the crossings through the Positions are converted to the SM reference frame. The belts. data sampling rate is standardised to one measurement For INTEGRAL, the safe belt entry and exit altitudes per minute (averaging the data points every minutes if the are evaluated every month based on the measurements sampling rate is higher) in order to have the same spatial of the two previous years. A simple sinusoidal fit is ap- density of points for both satellites along their trajecto- plied to the measured altitude at a fixed radiation level. ries in the radiation belts. Because the radiation monitors Based on this fit and accounting for some margins, the on each spacecraft are different, radiation measurements available window for observations results from these de- have to be cross-calibrated. Even if the measurements are fined entry and exit altitudes at which the instruments in some manner calibrated on the ODI server, this cross- are switched off and on. Given the orbit’s stability, this calibration is necessary. solution works very well in terms of instrument safety and The cross-calibration is performed separately for the ease of operations. However, due to smoothing of shorter highest and lowest radiation levels that correspond to the term dynamic radiation behaviour, scientific observation inside and outside of the belts, respectively. At the low time is lost. Moreover, the predictions can be unreliable end, the cross-calibration is done by comparing XMM- when the spacecraft crosses different parts of the belts, Newton and INTEGRAL radiation measurements made at something that became more common following the orbit the same time far away from the VAB. This implies assum- changes in 2015 (see Section 4.1). A global 3D volume ing that outside of the VAB the electron flux is isotropic, model could predict the different altitudes for successive which is not strictly accurate, but since the goal is to focus revolutions and thus yield more reliable estimates. The on the belts themselves, this assumption is good enough limitations of the current methods used to ensure safe op- because the radiation levels outside the VAB are orders erations on XMM-Newton and INTEGRAL were the main of magnitudes lower than inside. At the high end, the drivers for the work presented here. cross-calibration is performed by first creating indepen- dent XMM-Newton and INTEGRAL models of the VAB, and then comparing the radiation at corresponding points 3. Building a time-dependent volume model of the in the two mission-specific volume models. In this cross- Van Allen belts calibration process, IREM is taken as the baseline because A reliable simulation of the radiation environment the data show a much smoother behaviour and less scat- around the Earth requires a global understanding of the tering than for ERM. These two cross-calibrations at the radiation belts’ dynamics and the amount of data accumu- low and high ends result in a conversion function to be lated using INTEGRAL and XMM-Newton can help us to applied to ERM measurements. To make the transition have a phenomenological knowledge on this. The 5DRBM- between the two regimes smooth, the transition regions e model is built in a modular way from a static model over are smoothed using local spatial averaging. which time-dependent functions can be applied to account 3.1. The static volume model for the dynamics. Time-dependent deviations from the static model are quantified using the InterQuartile Range After cross-calibration, all data points are plotted in (IQR) of radiation flux measurements in cells over the vol- the SM frame. The 3D grid in units of Earth radii, R , ume of the belts, as defined in the model. The IQRs can is defined on a regular mesh of 0.1 R . The volume of the help to understand where the VAB are strongly variable, grid is centered on the Earth, 14 R in the z direction, and and computing them for different time scales can give an 32 R in both the x and y directions in order to encompass idea of the time evolution of the belts’ general shape. All the differential omnidirectional electron fluxes with The energy binning for each mission is different, and thus an their corresponding positions and time-stamps are taken interpolation in the electron energy spectrum is done to have the from the ODI server: these are measured for INTEGRAL same bins edges resulting in an overlapping range from 0.7 to 1.75 at [0.7, 0.78, 1.125, 1.27, 1.435, 1.615, 1.75] MeV and for MeV instead of 0.65 to 1.87 MeV one can see in Table 1. 4 Figure 2: The left plot shows the real tracks in the x − z plane of the SM frame, INTEGRAL (upwards) and XMM-Newton (downwards), colour-coded with the cross-calibrated flux. The right figure pictures the duplicated tracks after the mirroring step. the whole Van Allen Belts. Knowing the precise position of Figure 3 illustrates the steps used in the construction each measurement within the 3D volume, the value in each of the static model. Values in the resulting 3D matrix are node of the grid can be computed using the nearest data smoothed to obtain the final static volume model of the points. In order to have a full coverage of this volume, all trapped electron belts. A short 3D animation, showing tracks are mirrored with respect to the magnetic equato- the creation of the static model, has been selected for the rial plane, as shown in Figure 2. This assumes an up/down ESA INTEGRAL Picture Of the Month (POM) in July symmetry in the SM reference frame, which is commonly 2018 and at the same time for the NASA High Energy assumed in magnetospheric studies. Time dependencies Astrophysics Picture Of the Week (HEAPOW) . can be added later to take into account up/down asymme- It is this static model that constitutes the basis for the tries. more realistic time-dependent model for which the most Mirroring of the tracks increases the number of nodes important element is the variation due to the solar cycle. in which an average radiation value can be computed. This does not, however, cover the entire volume. Looking at a 3.2. The time dependence 3D rectangle of size 16×16×6 R centered on the Earth, The solar cycle on timescales of a few years is very well covering the main VAB regions below 45,000 km altitude, described by a simple sinusoid with an 11-year period. A 53% of all cells of size 0.1 R contain at least 1 data point, simple sinusoidal fit to the logarithm of the radiation flux and half contain 3 or more data points. This is not enough measured outside the VAB during the past 18 years, com- to build an accurate model down to a resolution of 0.1 R . bining XMM-Newton and INTEGRAL data, describes the Consequently, the remaining voids (nodes without an as- long term variation of the background radiation. The pa- signed value) in the grid need to be filled by iteratively rameters of this fit are determined during the elaboration using the nearest neighbouring nodes. Four iterations are of the model, and can then be used to vary the mean back- performed to fill the full 3D grid. An empty node value is ground radiation of the static model as a function of time found computing the median of all the neighbouring non- 7 when the model is used for simulations or predictions. zero nodes in a 0.8 R box size. In the 3D rectangle of size 16 × 16 × 6 R centered on the Earth, more than 95% of all cells of size 0.8 R contain at least 1 data point INTEGRAL POM July 2018: https://www.cosmos.esa.int/ with a median population per cell of more than 500, which web/integral/pom-archive is enough for a good statistical sampling. HEAPOW 2nd of July 2018: https://heasarc.gsfc.nasa.gov/ docs/objects/heapow/archive/solar_system/vab_integral.html No external parameters are used to include this solar cycle in The box size was optimized as a function of the required resolu- the model. Future work could include a more elaborate fitting of the tion and the processing time. solar cycle, or the use of external solar cycle variables. 5 (a) Real tracks in the SM frame colour-coded with the radiation flux (b) Radiation fluxes assigned to nodes of the 3D grid (c) Filled 3D grid using neighbouring nodes (d) Smoothed 3D grid resulting in the finals 3D static model Figure 3: Main building steps of the 3D static model using real data points. The left side of each figure is the x − y plane top view and the right side is the x − z plane side view. Everything is colour-coded with the cross-calibrated electron flux. The 3D grid has a size of 32x32x14 Earth radii. 6 Figure 4: Interquartile ranges 3D map projections. The left side is the top view of the x − y plane, and the right side is the side view of the x − z plane. High dispersion regions are located mainly at the boundary of the outer belt. The 3D grid has a size of 32x32x14 Earth radii. The background radiation is higher during lower solar ac- 4. Validating the 5DRBM-e volume model tivity and lower during higher solar activity periods. This 4.1. A self-consistency check is because this background radiation is mainly composed of galactic cosmic-rays whose density is strongly modulated A first step is a self-consistency check where the model by the solar cycle in an inversely proportional manner. predictions are compared to the actual data. Naturally, The influence of the solar cycle on the VAB is much more the modelled radiation profile should be consistent with complex and has not, at this stage, been implemented. the data. This comparison is shown in Figure 5 where it is The confidence on the model is estimated via the IQR seen that the model is not only consistent with the data, in 3D cells of 1 R . The IQR corresponds to a measure but that it accurately predicts radiation levels where there of dispersion of a data sample. In each cell, the median are data gaps (black regions in the top plot). radiation level of all data points is computed in order to As mentioned in Section 2, in order to avoid damaging split the values in the cell into two subsamples: one with their radiation-sensitive detectors, spacecraft like XMM- the values below the median, and the other with values Newton and INTEGRAL must turn off the scientific pay- above the median. For each subsample the medians are load instruments before entering and turn them back on computed, and the IQR for the cell is given by the absolute only after having exited the radiation belts. Another check difference between the medians of each subsample. The that can be done is to compare the model-predicted entry confidence interval (uncertainty envelope) on the radiation and exit altitudes to those measured. This is shown in profiles is computed from the IQRs across the modelled Figure 6 where measured altitudes are in blue, and pre- volume. Figure 4 shows the x − y and x − z planes of dicted values are in red. Long-term trends and seasonal the IQR map. Looking at Figure 3(d) it can be seen that variations are well captured, but not quite the peak ampli- the largest dispersions are along the boundary of the VAB tudes. The main reason is the absence of time variations where the effects of the solar wind and solar activity are in the static model which smooths out the general shape expected to be the strongest. The IQRs could also be used of the belts. The cross-calibration is also not perfect. The to explore large time-dependent variations where values split of the single red dotted curve into three in early 2015 are the greatest, and checking time variations in each cell. is the consequence of the change in INTEGRAL’s orbital This will be included in future work. period from 3 to 2.66 days following an orbital adjust- ment (Dietze et al, 2015). With an integer orbital period (3 days), INTEGRAL was scanning a similar part of the belt in each orbit. With the 2.66-day orbit, it needs 8 days 3.3. The 5DRBM-e model (3 orbits) to come back to a similar region of the belt, and hence the three curves. The result is the 5DRBM-e model for trapped 0.7–1.75 4.2. Validation on GPS data MeV electrons. It can be used with any orbit at any epoch to produce a predictive model of the radiation along the Actual validation must be done on data that have not orbit as a function of the solar cycle. The model is con- been used to construct the model. If, in addition, the re- tained in a single FITS file that includes the static model, sult can be compared to the predictions of another model, the sinusoidal fit parameters corresponding to the time- then it is ideal. The GNSS satellites orbits at approxi- dependent solar cycle variations, and the IQR dispersion mately 20,000 km altitude are constantly inside the VAB, which can be used to calculate uncertainties and confi- and thus subject to constant radiation. This makes the ac- dence intervals for any radiation profile. curate prediction of the radiation environment crucial to 7 Figure 5: INTEGRAL mission radiation profiles between 2003 and 2015. Each vertical line corresponds to one revolution with the orbital phase 0 being perigee point. The top figure shows the real radiation data measurements. The second panel shows the radiation intensity extracted from the 5DRBM-e static model along INTEGRAL’s trajectory. Figure 6: Belt entry and exit altitudes for each revolution for the whole INTEGRAL mission and for the expected future trajectory until the planned de-orbiting in 2029. The blue curves are the real measurements, and the red dots are the computed altitudes form the model based on a simple threshold on the electron counts. 8 Figure 7: Radiation profiles for the GPS-2R 12 (Navstar 60, USA 178) satellite. The top figure shows the comparison of the real (blue), the AE8min (dark green), the AE9mean (light green) and the 5DRBM-e (red) radiation data during a low solar activity period. The bottom figure shows the radiation profile for the same satellite but for a high solar activity period and corresponding the AE8max model. The red-shaded region around the red model curve is the 95% confidence interval. the success of any satellite on such orbits. The AE8 and and AE9 models are generally within the 95% confidence AP8 models are widely used for these particular regions of interval (red-shaded region) of the 5DRBM-e model. space as well as the newest AE9 model. Hence, using the data from a GPS satellite and comparing the output of 5. Conclusion the 5DRBM-e model to that of the AE8 and AE9 models constitutes an ideal validation test. Modelling these still rather poorly understood radia- Figure 7 shows a comparison between the static tion belts is important for space-borne activities (manned 5DRBM-e, the AE8min-AE8max and the AE9mean mod- and unmanned) in the near-Earth environment. The ex- els for the GPS-2R 12 (Navstar 60, USA 178) satellite. ploration of predictive methods used to ensure safe oper- The measured radiation data and satellite positions for ations on XMM-Newton and INTEGRAL, together with the dates shown in the figures are extracted from the ESA the benefit of creating global VAB models with great po- ODI server. From these positions, a SPENVIS readable tentials, were the driving motivations for the work pre- trajectory file is produced in order to compute the AE8 and AE9 modelled radiation data. This GPS satellite is sented here. on a very low eccentricity orbit with an average altitude The presented 5DRBM-e model has been built using of 20,200 km and inclination of 55 degrees. Both AE8min INTEGRAL (>16 years) and XMM-Newton (>18 years) (during a solar minimum activity) and AE8max (during a radiation data, crossing the Van Allen Belts in each revo- solar maximum activity) as well as the AE9mean are com- lution. This newly built data-driven model focuses on the pared to the static 5DRBM-e. The GPS data, the AE8 electron belt from approximately 4,000 km up to its outer boundary at around 60,000 km. The radiation flux given 11 by the 5DRBM-e model corresponds to the integrated elec- The AE8 and AE9 models data are taken from the ESA Space Environment Information System (SPENVIS). tron flux in the energy range from 0.7 to 1.75 MeV, the 9 overlapping energy range for the radiation monitors on- The end goal of this project is to provide robust, reli- board the two spacecraft. Remarkably, not only have these able, easy-to-use, data-driven, dynamic, 3D electron and two spacecraft gathered more than 16 years of contempo- proton radiation belt volume models that can be used in raneous radiation measurements, but their orbits scan dif- the space science and engineering community for design- ferent parts of the VAB and are thus complementary in ing, preparing, and running space missions. increasing the coverage of the belts allowing the creation of a relevant global static model. Acknowledgements A reliable prediction of the radiation environment around the Earth requires an understanding of the radi- LM would like to thank the Swiss Space Centre (SSC), and ation belts’ dynamics. Such knowledge is today surpris- the Swiss Space Office (SSO) for the financial support as a ingly limited. In this work, the time-dependent deviations Swiss National Trainee at the European Space Astronomy from the static model are quantified using the interquar- Centre (ESAC/ESA). tile range of radiation flux measurements in 3D cells over the entire volume defined in the model. The IQR gives an excellent idea of the model’s uncertainties on the radiation References intensities, allowing, for example, the computation of con- Baker, D. N., Erickson, P. J., Fennell, J. F. et al., 2018, Space fidence intervals for model-derived quantities such as the Weather Effects in the Earth’s Radiation Belts, Space Sci Rev altitudes at entry and exit points. This first static ver- 214: 17. sion of the 5DRBM-e with dynamic background radiation Boër, M., Naya, J., Chabaud, J.-P. et al., 1995, The CESR Multi- mission Radiation Monitor, IEEE Transactions on Nuclear Sci- modelling, following the solar cycle, shows promising re- ence, Vol. 42, No. 6. sults with respect to the well-known AE8min and AE8max Boyd, A. J., Turner, D. L., Reeves, G. D., et al., 2018, What causes as well as AE9mean models. In addition, a more accurate radiation belt enhancements: A survey of the Van Allen Probes Era, GeophysicalResearch Letters, 45, 5253-5259. modelling of the VAB enhances the predictions of the belts Casale, M., and Fauste, J., 2004, Attempt of Modelling the entry and exit times which contributes to maximize the ob- XMM-Newton Radiation Environment, Technical Note, XMM- servation time and increase the safety during instrument OPS-TN-004, http://xmm2.esac.esa.int/external/xmm_sched/ operation. The simplicity of the 5DRBM-e model and its vischeck/Rad_Model_TN.pd. Dietze, C., Vasconcelos, A., Ziegler, G. et al., 2015, INTEGRAL use of the Solar Magnetic reference frame make it easy End-Of-Life Disposal Manoeuvre Campaign, Proceedings of 25th to visualize in 3D space. Moreover, the model has been International Symposium on Space Flight Dynamics ISSFD, Mu- structured in a step-by-step way with very few dependen- nich. cies from one step to the another which makes it easy to Glauert, S. A., R. B. Horne, and N. P. Meredith, 2014, Three- dimensional electron radiation belt simulations using the BAS Ra- update and modify. If new data are available, a new up- diation Belt Model with new diffusion models for chorus, plasma- dated and improved model can be built in a few hours spheric hiss, and lightning-generated whistlers, J. Geophys. Res. only. This versatility allows for the creation of models Space Physics, 119, 268-289. based on specific data sets with very few modifications to Gonzalez-Riestra, R. and Rodriguez-Pascual, P.M., 2018, The Be- haviour of the XMM-Newton Background: From the beginning the procedure. of the mission until May 2018, XMM-SOC-GEN-TN-0014, issue The intention is to extend this model to include more 3.11, http://xmm2.esac.esa.int/docs/documents/GEN-TN-0014. dynamic features associated with the 11-year solar cycle, pdf. as well as yearly variations that are observed in the data. Hajdas, W., Bühler, P., Eggel, C., et al., 2003, Radiation environ- ment along the INTEGRAL orbit measured with the IREM mon- Short-term variations (days, weeks, months), mainly in- itor, A&A 411, L43-L47. fluenced by short-term solar events, will not be considered Johnston, W. R., O’Brien, T. P., Huston, S. L.et al, 2015, Recent as many existing models already accomplish this, as out- Updates to the AE9/AP9/SPM Radiation Belt and Space Plasma Specification Model IEEE Transactions on Nuclear Science, Vol. lined previously. The electron energy range will also be in- 62, No. 6. creased by using more energy bins from INTEGRAL and McIlwain, C. E., 1961, Coordinates for Mapping the Distribution of XMM-Newton and possibly from other missions. Using Magnetically Trapped Particles, J. Geophys. Res., 66, pp. 3681- more energy bins will require a good knowledge and ex- Mohammadzadeh, A., Evans, H., Nieminen, P. et al, 2003, The ESA trapolation of the energy spectrum seen by both IREM Standard Radiation Environment Monitor Program First Results and ERM instruments at each time-stamp, in order to From PROBA-I and INTEGRAL IEEE Transactions on Nuclear compare the same integrated flux. The expected energy Science, Vol. 50, No. 6. range should start at 0.7 MeV and go up to 2–3 MeV. More O’Brien, T. P., Johnston, W. R., Huston, S. L. et al, 2017, Changes in AE9/AP9-IRENE Version 1.5 IEEE Transactions on Nuclear importantly, a 5DRBM-p model will be constructed based Science, Vol. 65, No. 1. on the proton radiation measurement data also available Roederer, J. G., 1970, Dynamics of Geomagnetically Trapped Radi- from the ODI server. A first model will be built using ation, Physics and Chemistry in Space, Berlin: Springer. only proton fluxes measured by INTEGRAL, because it Sandberg, I., Daglis, I. A., Anastasiadis, A. et al, 2012, Unfolding and Validation of SREM Fluxes, IEEE Transactions on Nuclear has flown with a lower perigee altitude resulting in bet- Science, Vol. 59, No. 4. ter coverage of the inner proton belt. The proton energy Sandberg, I., Daglis, I. A., Heynderickx, D. et al, 2014, Development range will probably start at a few MeV up to several tens and Validation of the Electron Slot Region Radiation Environment of MeV. Model, IEEE Transactions on Nuclear Science, Vol. 61, No. 4. 10 Sawyer, D. M., and J. I. Vette, 1976, AP-8 Trapped Proton Environ- ment for Solar Maximum and Solar Minimum, NSSDC/WDC-A- R&S 76-06. Sicard, A., Boscher, D., Bourdarie, S. et al., 2018, GREEN: the new Global Radiation Earth ENvironment model (beta version). Annales Geophysicae, European Geosciences Union, 2018, 36, pp.953-967. 10.5194/angeo-36-953-2018 . hal-02126965. Stormer, C., 1937, On the Trajectories of Electric Particles in the Field of a Magnetic Dipole with Applications to the Theory of Cosmic Radiation. Sixth Communication. With 17 Figures in the Text, Astrophisica Norvegica, v. 2, p.193. Thorne, R. M., 2010 Radiation belt dynamics: The importance of wave-particle interactions Geophysical research letters, Vol. 37, L22107. Van Allen, James A., 1959, Radiation Belts Around the Earth, Sci. Am., Vol. 200, pp. 39-47, IRE Student Quarterly, pp. 4-15. Vassiliadis D., Klimas A., Fung S. et al., 2004, Structure and Dy- namics of the Outer Radiation Belt. In: Daglis I.A. (eds) Effects of Space Weather on Technology Infrastructure. NATO Science Series II: Mathematics, Physics and Chemistry, vol 176. Springer, Dordrecht. Vette, J. I., 1991a, The AE-8 Trapped Electron Model Environment, NSSDC/WDC-A-R&S 91-24. Walsh, B. M., Kuntz, K. D., Collier, M. R., et al., 2014, Ener- getic particle impact on X-ray imaging with XMM-Newton, Space Weather, 12, 387-394.

Journal

AstrophysicsarXiv (Cornell University)

Published: Jul 25, 2019

References