IAC-19-A6.2.4 Page 1 of 11 IAC-19-A6.2.4 Long-Term Environmental Effects of Deploying the Oneweb Satellite Constellation

IAC-19-A6.2.4

Long-Term Environmental Effects of Deploying the OneWeb Satellite Constellation

H.G. Lewisa*, T. Maclayb, JP Sheehanc and M. Lindsayd a Faculty of Engineering and Physical Sciences, University of Southampton, Southampton SO17 1BJ, UK, [email protected] b OneWeb, 1785 Greensboro Station Place, Tower 3, Suite 600, McLean, VA 22102, USA, [email protected] c OneWeb, 1785 Greensboro Station Place, Tower 3, Suite 600, McLean, VA 22102, USA, [email protected] d OneWeb, WestWorks Building, 195 Wood Lane, London W12 7FQ, UK, [email protected] * Corresponding Author

Abstract Large constellations of satellites offering global communications services are being planned for Low Earth Orbit (LEO). The deployment of such constellations would transform space traffic in this important environment, leading to questions about the sustainability of this activity in the LEO region. This paper reports the results of a study using the Debris Analysis and Monitoring Architecture to the Geosynchronous Environment (DAMAGE) evolutionary model to assess environmental sensitivity to the design & operational parameters of the OneWeb constellation, and quantifies the long-term environmental effects of deploying the constellation. The study results indicate that environmental degradation can be mitigated through use of high post-mission disposal reliability, short deorbit times, and the separation of orbital planes. Further, if two constellations are to be deployed in the same vicinity, the inter-constellation collision probability can be reduced significantly by introducing a separation between the two systems.

Keywords: Orbital Debris, Debris Models, New Space, Large Constellations

environmental sensitivity to the design & operational 1. Introduction parameters of the OneWeb constellation; (2) illustrate the In the last decade a ‘New Space’ age has emerged as long-term environmental effects of the OneWeb a result of technological advances in space launch and constellation; (3) investigate the implications of a second space systems, and widespread private sector large constellation operating in proximity to the OneWeb investments. This ‘New Space’ has been characterised by constellation; and (4) investigate the implications of a deployments of small satellites and fleets of small hypothetical collision within the OneWeb constellation. satellites into the LEO region. These space systems offer The study incorporated planned design and operational an appealing, low-cost route for start-ups, and enable parameters from a genuine satellite constellation and the opportunities for novel space-based services that would DAMAGE model was extensively upgraded with have been unachievable under the traditional ‘Old Space’ features required to simulate those characteristics. In paradigm. In particular, large constellations of satellites addition, a novel collision prediction method developed offering global broadband communications services, specifically for the assessment of large constellations was which have previously been provided from employed to overcome the limitations of existing geosynchronous orbit, are being planned for LEO. This methods applied in this context. solution represents an opportunity to disrupt the traditional communications sector, thanks to the 2. Method significant reduction in latency and potential for true 2.1 The DAMAGE model global coverage, especially in the domain of internet The DAMAGE debris model is a three-dimensional connectivity. Consequently, a number of companies, computational model of the full LEO to GEO Earth Orbit including OneWeb, have begun the development of debris environment. It includes source models for objects constellations of satellites to deliver broadband internet down to 10 cm but is capable of modelling populations services to the world. However, the deployment of such of objects down to 1 mm over short projection periods for constellations would transform space traffic in this a limited set of target objects. Large objects are treated important environment and questions arise about the individually, whereas smaller objects are sampled sustainability of this use of the LEO region [1]. This statistically. DAMAGE does incorporate a flexible de- paper reports the results of a study using the DAMAGE sampling algorithm enabling the generation and evolutionary model to quantify the long-term treatment of discrete objects from previously sampled environmental effects of deploying the OneWeb satellite population files. constellation. The aims of this study were to (1) assess

IAC-19-A6.2.4 Page 1 of 11 70th International Astronautical Congress (IAC), Washington D.C., United States, 21-25 October 2019. Copyright ©2019 by the authors. Published by the IAF with permission and released to the IAF to publish in all forms.

DAMAGE is supported by a fast, semi-analytical to adjust the post-service orbit of the satellites. In general, orbital propagator [2], a breakup model [3,4], several post-mission disposal for constellations is modelled as a collision prediction algorithms including a method based three-step process: (a) descent to a circular orbit below on the cube approach [5] adopted in NASA’s LEO-to- the constellation; (b) a coast period of up to 30 days, and; GEO Environment Debris model (LEGEND) [6]. (c) lowering of the perigee altitude (and apogee altitude DAMAGE also includes a deterministic collisions if desired) to achieve a specified disposal orbit and/or assessment (DCA) method which was added to address residual lifetime. If electric propulsion is specified, the issues related specifically to collision assessment for DAMAGE orbital propagator will compute the constellations. The method relies on two geometrical pre- appropriate trajectories to achieve steps (a) and (c), and filters introduced by Hoots et al. [7] to firstly screen the the user can adjust the low-thrust burn angles to achieve object population to remove from the potential a desired descent time. The duration of the coast phase is conjunction list any pairs of objects with orbits that do selected at random and was implemented to ensure that not intersect (the “perigee-apogee” test) and secondly to the disposal orbits achieved by satellites from the same locate points of closest approach between pairs of orbits. orbital plane are separated in RAAN and altitude to Once these points have been found the DCA method then reduce collision risk. performs a distance check and, if passed, effectively DAMAGE is also able to simulate constellations with determines if the pair of objects will be at one of those operational orbital planes that are separated by a user- locations at the same time; if they are, a collision is specified altitude increment. When this option is assumed to take place. selected, a small adjustment is made to the inclination of A constellation module has been added to allow the each plane to ensure that the orbital precession induced investigation of a wide range of constellations and the by the Earth zonal gravity perturbations affect the orbital variety of mitigation measures they may employ. This planes in the constellation equally, thus maintaining the module was used as part of the European “mini- constellation geometry. campaign” in 2015 and for a more recent European Space Agency (ESA) study, both investigating the possible 2.2 Simulation cases impacts of large constellations on the LEO environment This study provided an opportunity to understand the [1]. The constellation module has undergone a substantial environmental impacts of the OneWeb constellation and, upgrade to enable the deployment and operation of in particular, to address the aims set out in section 2 specific constellation systems, such as the OneWeb above. This was achieved through the use of simulation constellation, to be simulated. cases focused on each of the objectives, and a case used A relatively simple process is used within the to understand the evolution of the space debris constellation module to build and subsequently replenish environment without the deployment of the OneWeb constellations within simulations. This process is based constellation (the “reference” case). on a launch schedule comprising the number of launches The reference case used for this study corresponds to per year, the number of satellites on each launcher, and the current reference case adopted by the Inter-Agency the duration over which the build and replenishment are Space Debris Coordination Committee (IADC) and to the to take place. reference case reported in Lewis et al. [1]: A low altitude deployment from the launcher can be specified and the DAMAGE orbital propagator will • A 1 January 2013 epoch with an initial compute the ascent trajectories for the constellation population corresponding to all objects ≥ 10 cm satellites, incorporating a user-specified ascent time. residing within or crossing the LEO protected Throughout this period, the user can indicate whether the region satellites are capable of collision avoidance. DAMAGE • Launch traffic was assumed to be represented by will also compute the required phasing in the Right the repetition of recent launches (1 January 2005 Ascension of the Ascending Node (RAAN) to deliver the to 31 December 2012) satellites into the correct orbital plane following their • New spacecraft and rocket upper stages were ascent. Satellites launched via the replenishment assumed to achieve a 90% success rate with schedule replace the corresponding satellites in the respect to post-mission disposal (uncontrolled re- constellation, and the older satellites are retired, entry within 25 years) following the user-specified post-mission disposal • Vehicle passivation was assumed to be 100% behaviour. Once in service, the satellites maintain their successful such that no explosions were within-plane (true anomaly) spacing and the RAAN permitted within the projection period. spacing of the orbital planes for the chosen constellation, subject only to Earth zonal gravity perturbations. This reference case was modified slightly to DAMAGE can apply a variety of post-mission incorporate collision avoidance for all active spacecraft disposal options, including the use of electric propulsion

IAC-19-A6.2.4 Page 2 of 11

70th International Astronautical Congress (IAC), Washington D.C., United States, 21-25 October 2019. Copyright ©2019 by the authors. Published by the IAF with permission and released to the IAF to publish in all forms. in the background population. The baseline OneWeb operational parameters of the OneWeb constellation; (2) constellation was created by simulating the following: the evaluation of the long-term environmental effects of the OneWeb constellation, including the reference case; • Walker-star constellation comprising 882 (3) the investigation of the implications of a second large satellites in 18 orbital planes all at the same 1200 constellation operating in proximity to the OneWeb km altitude and inclined at 87.9° constellation; and (4) the investigation the implications • Constellation spacecraft design lifetime between of a collision within the OneWeb constellation. 6 and 7 years, 150 kg and 3.5 sq. metres

• Constellation build-up phase from 1 Jan 2018 to 3.1 Environmental sensitivity to OneWeb constellation 1 Jan 2023 with 5 launches per year and 34 design and operational parameters spacecraft per launch, and with an additional The aim of this study was to understand the launch carrying 32 spacecraft every 5 years implications of key constellation design and operational • Constellation replenishment phase from 1 Jan parameters for the environmental impact of the OneWeb 2023 to 1 Jan 2094 with the same launch pattern constellation. The following parameters were as the build-up phase. investigated: • Deployment from launch vehicle to a circular

orbit at 500 km followed by a 5-day checkout 1. Orbital plane separation (altitude): 0 km, 1 km, period, then a low-thrust ascent to 1200 km in 2 km, 3 km, 4 km, 5 km, and 8 km approximately 4.5 months. Immediate de-orbit of 2. Deorbit Time: 25 yrs, 5 yrs, 1 yr, and 0 yr (re- orbital rocket stages following deployment. entry under power) • Post-mission disposal (PMD) of 95% of 3. Post-mission disposal Reliability: 75%, 90%, constellation spacecraft to a 200 × 1100 km orbit 95%, and 99% (assumed to be consistent with re-entry) within 4. Active Removal Rate: 0/yr, 1/yr, 5/yr, and 10/yr approximately 8 months using electric

propulsion. The analysis aimed to demonstrate the sensitivity of • Collision avoidance at 100% success rate for the evaluation metrics to the constellation design and active OneWeb satellites (i.e. no active OneWeb operational practices employed by OneWeb. In general, satellites were permitted to collide with any the cases identified above represented the minimum tracked object). number of cases to investigate and the set was increased

where appropriate to gain an improved understanding of Simulations of the reference and all OneWeb the sensitivities. Here, the factors above were varied with constellation cases did not include objects smaller than respect to the baseline OneWeb constellation case 10 cm, but all debris-on-debris collisions involving described in section 2 above. objects ≥ 10 cm (including fragments) were permitted.

Unless otherwise specified below, the results describe the 3.1.1 Orbital plane separation outcome of 100 MC runs for each simulation case, in In previous studies by Lewis et al. [1] and Bastida which the mean anomalies of all active and failed Virgili et al. [8] it was found that failed constellation constellation objects were propagated, and the mean satellites contributed in a significant way to the long-term anomalies for all constellation satellites conducting post- environmental impacts of large constellations of satellites. mission disposal manoeuvres were randomised. In In these studies, the number of failed satellites continued addition, the results include background population to increase monotonically while the constellation was objects (described in the reference case, above) unless operational as a result of continuing failures and the lack specified otherwise. Due to the 100% successful collision of atmospheric drag. In addition, collisions between avoidance assumption for the constellation, this meant failed satellites – which cannot be actively avoided – that only collisions between inactive satellites or debris generated additional fragmentation debris even after the were permitted. end of the constellation operations. Separating orbital For cases where the orbital planes of the OneWeb planes in altitude provides a way to reduce the number of constellation were separated, the results described below collision events involving failed constellation satellites. were derived using the DCA method and did not include This simulation case was used to quantify the the background population, thereby isolating the effects effectiveness of this measure. Here, DAMAGE applied of plane separation. the chosen plane separation to the semi-major axes of the

orbital planes, and also ensured that the separation 3. Results and discussion between active satellites in neighbouring/consecutive Results from the simulations are presented in the planes was always maintained at the separation value following order in the sections below: (1) the assessment specified. of the environmental sensitivity to the design &

IAC-19-A6.2.4 Page 3 of 11

The DAMAGE predictions of the average number of the perigee altitude of the disposal orbit. Disposal orbit objects for 100 MC runs of the plane separation cases are perigee altitudes were selected to span residual orbital shown in Fig. 1. lifetimes, after decommissioning, from zero (powered, random re-entry) to 25 years. Table 1 shows the values of perigee altitude used for this study and the corresponding residual lifetimes.

Table 1. Relationship between disposal orbit perigee altitude and approximate average residual lifetime, for the OneWeb satellites and a disposal orbit apogee altitude of 1100 km. Perigee altitude Approximate residual lifetime (km) (years) 200 0 (re-entry under power) 300 1 350 3 375 5

Fig. 1. Average number of objects predicted by the DCA 400 10 450 17 method for plane separation cases without the 500 25 background population.

The DAMAGE predictions of the average number of The results show that the rate of growth of objects for 100 MC runs of these post-mission disposal constellation objects was reduced, on average, by lifetime cases are shown in Fig. 2. The oscillations seen incorporating a plane separation into the constellation in the results were due to the solar cycle; actual orbit design. Small differences in the number of intact objects lifetimes varied above and below the values listed in Tab. were observed (a maximum difference of approximately 1, depending on solar activity level, affecting the number 60 objects by the end of the projection period). The of satellites on-orbit at any point in time. beneficial effects become apparent with even small plane separations, and results suggest diminishing returns such that plane separations larger than 3 km or 4 km did not provide any additional reduction in the number of objects produced in the simulations. Similar results were found for the number of catastrophic collisions occurring by the year 2094 and for the number of collision fragments generated.

3.1.2 Deorbit time Lewis et al. [1] and Bastida Virgili et al. [8] found that long deorbit times for constellation satellites were a leading cause of increased negative environmental impacts. In particular, constellation satellites that followed the IADC guideline and remained on-orbit for Fig. 2. Average number of objects predicted by the cube 25 years after decommissioning were found to be algorithm for disposal orbit perigee (background involved in a high number of collisions with the population included). background population of objects, even if the post- mission disposal reliability was near-perfect. Both Unsurprisingly, the results in Fig. 2 show that the studies determined that short deorbit times were required impact on the environment, in terms of the number of to reduce these negative consequences. objects, can be reduced by selecting lower perigee For the present study, satellites being deorbited from altitudes for the disposal. In the best case, when the the constellation were first lowered from 1200 km to an OneWeb satellites continue to lower the perigee altitude 1100 km circular orbit to remove them from the until re-entry (i.e. 0 years residual lifetime, or powered constellation. Then their perigees were lowered to a re-entry), the number of objects in the environment in disposal orbit where atmospheric drag would reduce the year 2094 was at a level 80% of the corresponding apogee altitude and cause the satellite to re-enter the number for the 25-year residual lifetime/500 km perigee atmosphere. The effect of post-mission deorbit time for altitude case. These results therefore provide the OneWeb constellation was investigated by varying considerable support to the idea that reducing the residual

IAC-19-A6.2.4 Page 4 of 11

70th International Astronautical Congress (IAC), Washington D.C., United States, 21-25 October 2019. Copyright ©2019 by the authors. Published by the IAF with permission and released to the IAF to publish in all forms. lifetimes of constellation satellites, below the 25 years rate. So, it was the collision fragment population that recommended by existing mitigation guidelines, will accounted for the exponential growth in the total number benefit the environment. of objects seen in Fig. 3. Similar behaviour was seen for the collision For relatively low PMD rates, the large numbers of fragments and the number of catastrophic collisions. fragments were generated by a large number of With respect to the number of catastrophic collisions, the catastrophic collisions. For the 75% PMD case, there change from a 25-year residual lifetime solution to a were 39.91 catastrophic collisions, on average (c.f. the powered re-entry (i.e. 0 years residual lifetime) resulted baseline 90% PMD case with 17.75 catastrophic in 52% of the number of catastrophic collisions taking collisions, on average). Note that non-catastrophic place, on average. collisions (i.e. those with energies less than 40 J/g) were The spatial densities at the constellation altitude considered in the simulations, but only for objects ≥ 10 remained unchanged under different disposal orbit cm. Increasing PMD to levels above 90% resulted in a solutions, but the key benefits emerged at altitudes below further reduction in the number of catastrophic collisions. about 800 km. This is important when the background At 99% reliability the number of catastrophic collisions population is taken into account, because that particular fell to 14.54. region is heavily populated and is the likely destination 75% post-mission disposal reliability resulted in altitude for future small satellite missions. In each nearly an order of magnitude higher spatial density at the disposal orbit case except the powered re-entry, however, constellation altitude, compared with the baseline case. a high spatial density arose at the constellation disposal In addition, the fragments generated in collisions were orbit perigee altitude. likely to have been ejected into eccentric orbits, with apogees above the constellation and perigees below the 3.1.3 Post-mission disposal reliability constellation. As a result, the spatial density at altitudes The DAMAGE predictions of the average number of between 800 km and 2000 km also increased. objects for 100 MC runs of the post-mission disposal reliability cases are shown in Fig. 3. The results show that 3.1.4 Active removal rate the impact on the environment, in terms of the number of As described in Lewis et al. [1] and Bastida Virgili et objects, can be reduced by increasing the reliability of the al. [8], failed constellation satellites represent the primary post-mission disposal operation. The number of objects reason for negative environmental effects arising from in the environment at year 2094 increased exponentially the deployment of large constellations of satellites in with decreases in reliability. LEO. Active removal of these satellites might provide a way to mitigate the negative environmental effects. This simulation case was performed to quantify the potential benefits of active removal for the OneWeb constellation. The DAMAGE predictions of the average number of objects for 100 MC runs of the active debris removal cases are shown in Fig. 4. The results show that increasing the ADR rate from zero to 10 per year produced a corresponding decrease in the number of objects in the LEO environment from 20,763 objects to 19,466 objects, on average, at the end of the simulated time period. This equates to an effective reduction factor of 1.73 for the highest removal rate (i.e. for every object removed by ADR, the number of objects in LEO decreased by 1.73 on average) and an effective reduction Fig. 3. Average number of objects predicted by the cube factor of 2.99 for five removals per year. The increase in algorithm for disposal reliability cases with the effectiveness for lower removal rates is in line with the background population. results of previous ADR studies that illustrated a similar trend of benefit with diminishing returns as ADR rate is Decreasing the post-mission disposal reliability from increased. the baseline level of 95% resulted in an increase in the number of failed satellites and a corresponding increase in the number of fragments generated by collisions involving those satellites. Whereas the number of failed satellites was inversely proportional to the disposal reliability rate, the number of collision fragments increased exponentially with decreases in the reliability

IAC-19-A6.2.4 Page 5 of 11

The results have shown that careful selection of OneWeb design and operational parameters can mitigate environmental impacts. These parameters were: separation of orbital planes in altitude, short deorbit times while maintaining control of satellites until re-entry, and high reliability of the post-mission disposal functions. In addition, active removal of failed satellites can be considered as a means of augmenting post-mission disposal reliability. The following section reports on a study where these particular OneWeb parameters were simulated.

3.2 Long-term environmental effects of the OneWeb Fig. 4. Average number of objects predicted by the cube constellation algorithm for active removal rate cases (background Results from the sensitivity study described above population included). show the relative environmental benefits of afforded by careful selection of design and operational parameters for 3.1.5 Conclusions drawn from the sensitivity study large LEO constellations. The purpose of the following Four sensitivity studies were performed, aimed at study was to understand more specifically the impact that understanding the influence of key design and the choices made by OneWeb may have on the long-term operational parameters on the environmental impact of evolution of the LEO space debris environment. This the OneWeb constellation. All of these cases were study was therefore simple in design and involved the investigated using a minimum of 100 Monte Carlo comparison of a simulation featuring the deployment of simulations, performed using the DAMAGE code. The the planned OneWeb constellation with a simulation results from these cases have shown that: without the constellation (the reference case). Both of these simulations incorporated the existing background • Separating the orbital planes within a population of objects and future background launch constellation is an effective means of reducing traffic, following the method outlined in section 2 above. the probability of collision among failed In addition, the number of MC runs was increased to 300 satellites, and therefore, the impact on the in order to improve the robustness of the study. The key environment. results from the reference case are described first. • Reducing the residual orbital lifetime of constellation satellites below the currently 3.2.1 Reference case: no constellation recommended 25-year limit provides substantial The DAMAGE prediction of the average number of benefits with respect to the environment. Short ≥10 cm objects in LEO for 300 MC runs of the reference residual orbital lifetimes also prevent the build- (no constellation) case is shown in Fig. 5 with the +/- up of high concentrations of inactive satellites at standard deviations. The short-term oscillation seen in low altitudes (resulting from the co-location of the Figure is due to the ~11-year solar cycle. In terms of orbital perigees). the long-term behaviour, the average number of objects • Post-mission disposal reliability of the satellites showed a relatively slow increase from the start of the was found to be a key influence on the projection period to the year 2050, followed by a environment impact of the constellation. The corresponding slow decrease (ignoring the short-term number of objects can grow exponentially, even oscillation). The spread captured by the standard without the addition of the background deviation was approximately 20% of the average value at population, unless reliability rates are the end of the projection period and reached 10% of the sufficiently high or other mitigation measures are average as early as the year 2020. employed (e.g. the separation of orbital planes). • Active removal of failed constellation satellites provided a way to reduce the environmental impact of the OneWeb constellation. Ultimately, the benefit of removing a satellite is approximately the same whether it is removed via post-mission disposal (i.e. under its own power) or via ADR.

IAC-19-A6.2.4 Page 6 of 11

The average catastrophic collision rate for all objects (i.e. including the background population) was 0.18 per year for the OneWeb case. The rate for collisions involving an object from the background population and a OneWeb object was an order of magnitude lower: 0.022/yr. The rate at which self-induced collisions occurred – i.e. only between OneWeb satellites – was another order of magnitude lower than this: 0.002/yr. Similar trends were seen in the number of fragments that were generated within these categories. The results of this study also underscored the importance of PMD reliability. Over long simulation periods, failed satellites accumulate, providing additional Fig. 5. Average number of objects predicted by opportunities for such objects to collide with each other DAMAGE with and without the OneWeb constellation. and the background environment. When coupled with an The dashed lines indicate the standard deviation about the assumed growth in the background debris population, mean of the no-constellation case. lower PMD reliability rates correspond to scenarios in which the percentage of catastrophic collisions involving The probability distribution derived from the Monte a constellation satellite increase over time. That is, the Carlo simulations revealed a non-Gaussian nature for the lower the PMD reliability, the more likely it is that probability distributions, with considerable probability catastrophic collisions will involve a constellation density for high numbers of objects in the orbiting satellite. population. In addition, the spread is greater than Another interesting result of the study highlighted the suggested by the standard deviation: the range at the end benefits of smaller satellites. Fragment production is a of the projection period was 12,000-27,000 objects. The function of the mass involved in the collision, and at only uncertainty represented by this range arises solely as a 150 kg, collisions involving OneWeb satellites did not result of the stochastic nature of the collision prediction, tend to produce large numbers of fragments compared to and the generation of resulting fragments from the collisions involving more massive objects. Normal distributions in the NASA standard break-up model described in Johnson et al. [3, 4]. It is also possible 3.2.3 Conclusions from the study for DAMAGE projections to result in fewer objects in the In a comparison with a reference case, the impact of LEO environment than the number existing at the the OneWeb constellation on the environment (as beginning of the projection period. The probability of this measured by the number of objects and the cumulative occurring in any MC run was approximately 20%. number of catastrophic collisions) was not significant, presuming effective collision avoidance measures are 3.2.2 The OneWeb case employed by active constellation satellites. Even if no The OneWeb constellation incorporated in the collision avoidance maneuvering is presumed, there was simulation was the same as the baseline case described in an approximately even chance that the OneWeb section 2 except for one modification: the orbital planes constellation case would produce MC runs that were were each separated by 4 km altitude to reflect the actual indistinguishable from the runs obtained for the reference OneWeb constellation design. Due to the computational case. burden associated with the inclusion of the background The collision rate in the background population was population and the use of the DCA method, which would an order of magnitude higher than the corresponding rate ideally have been used to investigate this constellation for collisions involving OneWeb objects, and when arrangement, the cube method was used to perform this collisions involving OneWeb satellites did occur, the simulation. The DAMAGE predictions of the average relatively small number of fragments produced by these number of objects for 100 MC runs of the reference (no events further limited the impact of the constellation on constellation) case and the OneWeb case are shown in the environment. Fig. 5. As can be seen, the average number of objects in the LEO environment was higher when the OneWeb 3.3 Implications of a second large constellation constellation was included. However, the count includes operating in proximity to OneWeb approximately 1200 active OneWeb satellites in all years This section describes the results of a sensitivity study from 2023 to the end of the projection period: those at the to investigate the effects of operating two large mission altitude, those replacement satellites ascending constellations in proximity to one another. In particular, to the mission altitude, and those retired satellites the focus is on the benefits of separation distance in descending from the mission altitude. reducing inter-constellation collision risk, which has a

IAC-19-A6.2.4 Page 7 of 11

70th International Astronautical Congress (IAC), Washington D.C., United States, 21-25 October 2019. Copyright ©2019 by the authors. Published by the IAF with permission and released to the IAF to publish in all forms. direct bearing on the long-term sustainability of altitude in 5 months. Immediate de-orbit of constellations, and in LEO operations in general. orbital rocket stages following deployment. • PMD of 95% of constellation spacecraft to an 3.3.1 Sensitivity study orbit with 200 km perigee and apogee 100 km The simulation considers the deployment of two large below the mission altitude (assumed to be constellations (A and B) in LEO, with varying amounts consistent with re-entry) within less than 8 of altitude separation between them. To isolate the effect months using electric propulsion of separation on inter-constellation collision, only • Collision avoidance at 100% success rate for spacecraft from two constellations were simulated; no active constellation B satellites (i.e. no active objects from the background LEO population and no constellation B satellites were permitted to non-constellation traffic were included. The spacecraft collide with any tracked object, including those were propagated over an 80-year projection period from from constellation A). 1 January 2018 at intervals of 5 days. To account for the random aspects of the model and simulation, 100 runs As described above, it was assumed that were performed for each case outlined below. The mean constellations A and B adopted space debris mitigation anomalies of all objects were propagated, and all practices that were more rigorous that those outlined in conjunctions between spacecraft were identified using an the Inter-Agency Space Debris Coordination Committee implementation of the cube algorithm and were recorded Space Debris Mitigation Guidelines. for subsequent analysis. For each of the nine simulation cases, the collision The simulation study had two objectives: firstly to rate over the 2035-2060 period was estimated from the look at the implications of proximity without a number of collisions predicted from all conjunctions fragmentation, and secondly to consider the implications recorded using the cube algorithm. This collision rate, as of a fragmentation in constellation B. The first objective a function of the separation between the two was met by analysing the collision risk between intact constellations, was subsequently used to quantify the objects from each constellation, and the second objective effects of the constellation separation. In particular, two was met by analysing the collision risk between categories of collision were used for the analysis: (1) fragments from a catastrophic collision involving two collisions involving an intact, failed spacecraft from spacecraft from constellation B (simulated in every constellation A and an intact, failed spacecraft from Monte Carlo run on 1 January 2033). Fragmentation constellation B (labelled as “intact v intact” below); and debris from the simulated collision was generated using (2) collisions involving an intact, failed spacecraft from an implementation of the NASA standard breakup model constellation A and a fragment from a collision in [3, 4] and added to the simulation. constellation B presumed to occur on 1 January 2033 Constellation A was created by simulating the (labelled as “fragment v intact” below). baseline OneWeb case as described in section 2. Nine The number of collisions increased non-linearly with instances of constellation B were created and simulated time due to the increasing number of failed satellites from to account for nine different separations from continuing constellation launch activity over the full constellation A, with the following characteristics: projection period. For the 1200 km case, the collision rate for the period 2035-2060 was 14.72 × 10-3 per year, • Walker-delta constellation comprising 1600 whereas for the period 2060-2095 it was 60.23 × 10-3 per satellites in 32 orbital planes all at the same year; approximately four times greater. The altitude. Separation from constellation A was corresponding collision rates for all of the constellation achieved by placing constellation B at different separation cases are shown in Table 2 below. To altitudes, in 25 km increments between 1,200 normalise the collision rates, they were divided by the (complete overlap) to 1,000 km. Constellation B collision rate for the 0 km separation case. The resulting was given an inclination of 53°. normalised collision rates were found to be the same • Constellation spacecraft design lifetime between whether using the 2035-2060 gradient or the 2065-2090 6 and 7 years, 400 kg and 5 sq. metres gradient (as illustrated in Fig. 6). At a separation of 100 • Constellation build-up phase from 1 Jan 2018 to km between the two constellations, the “fragment v 1 Jan 2023 with 20 launches per year and 16 intact” collision rate was reduced to one-fifth of the rate spacecraft per launch observed for the no-separation case. The beneficial • Constellation replenishment phase from 1 Jan effects of separation for the “fragment v intact” case 2023 to 1 Jan 2094 with 21 launches per year and diminished as the separation increased. 16 spacecraft per launch. • Deployment from launch vehicle to a circular orbit at 500 km followed by a 5-day checkout period, then a low-thrust ascent to the mission

IAC-19-A6.2.4 Page 8 of 11

Table 2. Average catastrophic collision rate estimated by DAMAGE for the 2035-2060 and 2065-2090 periods and over 100 Monte Carlo runs as a function of the separation of constellations A and B. The rate of collisions involving an intact, failed spacecraft from constellation A and an intact, failed spacecraft from constellation B are shown as “Intact v Intact”. The rate of collisions involving an intact, failed spacecraft from constellation A and a fragment from the 1 January 2033 collision of two spacecraft from constellation B are shown as “Fragment v Intact”. Collision rate 2035- Collision rate 2065- 2060 2090 (× 103 #/year) (× 103 #/year) Fig. 6. Fraction of the collision rates computed by Separ- Intact v Fragment Intact v Fragment DAMAGE for 2035-2060 as a function of the separation ation Intact v Intact Intact v Intact of two constellations A and B. (km) 0 14.72 1.34 60.2 2.35 25 0.12 0.67 0.67 1.2 3.4 Implications of a collision within the OneWeb 50 0.07 0.45 0.25 0.82 constellation 75 0.08 0.34 0.27 0.60 The purpose of this final portion of the study was to 100 0.08 0.25 0.41 0.48 investigate the environmental implications of the 125 0.16 0.17 0.67 0.37 occurrence of one catastrophic collision between two 150 0.16 0.20 0.56 0.37 OneWeb satellites at the operational altitude. As before, 175 0.16 0.16 0.53 0.32 the OneWeb baseline constellation was used, as defined 200 0.13 0.16 0.38 0.26 in section 2, and in order to isolate the characteristics of interest, the background population of objects was not Inspection of the “fragment v intact” conjunctions considered. A catastrophic collision involving two revealed that most events occurred at, or near, the altitude spacecraft from the OneWeb constellation was simulated of constellation A with the number of conjunctions at to take place on 1 January 2033 in every Monte Carlo run. altitudes below the constellation being at least an order Fragmentation debris from this event was generated of magnitude lower. using an implementation of the NASA standard breakup model [3, 4] and added to the simulation. 3.3.2 Conclusions from the study The DAMAGE predictions of the average number of The results show that when large constellations objects for 100 MC runs of the presumed collision case overlap, the collision probability among failed satellites are shown in Fig. 7 and are compared with the equivalent goes up by a factor of 100. Unsurprisingly, if a collision results from the baseline case. The presumed collision occurs in one constellation, the fragments are ejected into event can be observed at 1 January 2033 as a sudden orbits that affect a wide region, so subsequent collision increase in the number of objects. Approximately 350 risk can be reduced with increased separation. By adding fragments were generated by this event and these a separation of 25 km, the likelihood of a fragment-intact accounted for the majority of the collision fragments in collision drops by half, and with 100 km separation, the the environment at the end of the projection period. risk drops to only 20% (compared to being at the same Inspection of the cumulative number of catastrophic altitude). collisions for the presumed collision case and for the As this is a parametric sensitivity study focused on baseline case reveals that, at least through the year 2075, the benefits of separating constellations in altitude, some there was no significant difference between the two cases. types of collisions were not considered. Consequently, In other words, an initial collision did not have a the absolute numbers in Table 2 should not be taken at noticeable effect on the subsequent collision rate face value, but the relative trend of risk reduction with observed. increased constellation separation is valid. Furthermore, absolute collision rates will be dependent on how rigorously constellations implement their debris mitigation plans (e.g., post-mission disposal success rate), but here again, the benefits of constellation separation seen in the results of this study remain valid, as shown in Fig. 6.

IAC-19-A6.2.4 Page 9 of 11

The study results demonstrate that separating the orbital planes within the constellation, targeting high reliability of post-mission disposal, and reducing the residual orbital lifetime of constellation satellites (below the currently recommended 25-year limit) provide substantial benefits with respect to the environment. These measures are incorporated into the design and operation of the OneWeb constellation, and as a result, the impact of the OneWeb constellation on the simulated environment was insignificant (as measured by the number of objects and the cumulative number of catastrophic collisions with respect to a reference case without the constellation). The collision rate in the Fig. 7. Average number of objects predicted by the cube background population was an order of magnitude higher algorithm for presumed collision and baseline cases. than the corresponding rate for collisions involving OneWeb objects, and when collisions involving OneWeb 3.4.1 Conclusions from the study satellites did occur, the relatively small number of A self-induced collision event was simulated in all fragments produced by these events further limited the 100 Monte Carlo simulations performed by DAMAGE. impact of the constellation on the environment. The dominant effect of this event was an increase in the The results also showed that the collision risk for two number of objects. These objects – collision fragments – large constellations orbiting in close proximity can be were distributed at a range of altitudes around the reduced by adding separation. It was found that a constellation. There was no significant effect on the separation of 25 km reduced the likelihood of a fragment- subsequent collision rates after this self-induced event. It intact collision by half, and a 100 km separation reduced should be noted that the limited environmental effects of the likelihood to only 20% (compared to being at the this event are due, in large part, to the small mass of the same altitude). Finally, a self-induced collision event in satellites involved in the presumed collision. The mass the OneWeb constellation was found to increase the of a OneWeb satellite is only 150 kg, and the number of objects overall, but had no significant effect environmental impact of a collision in a constellation of on the subsequent collision rate in the environment. larger satellites would obviously scale with the mass involved (as well as the number of satellites in the Acknowledgements constellation). The authors gratefully acknowledge the provision of population data by colleagues from the ESA Space 4. Conclusions Debris Office and their support for this work. The study described in this paper was designed to understand the potential environmental impacts of References deploying the OneWeb satellite constellation in LEO and [1] H.G. Lewis, J. Radtke, A. Rossi, J. Beck, M. built upon similar studies conducted by Lewis et al.[1] Oswald, P. Anderson, B. Bastida Virgili, H. Krag, and others, which did not incorporate genuine design and Sensitivity of the space debris environment to operational parameters from a planned satellite large constellations and small satellites, J. British constellation. The aims of this new study were to: Interplanetary Soc. 70 (2017) 105-117. [2] H.G. Lewis, A. Saunders, G.G. Swinerd, R.J. • Assess environmental sensitivity to the design & Newland, Effect of thermospheric contraction on operational parameters of the OneWeb remediation of the near-Earth space debris constellation; environment, J. Geophys. Res. 116 (2011) • Illustrate the long-term environmental effects of A00H08. deploying the OneWeb constellation; [3] N. L. Johnson, P. H. Krisko, J. C. Liou, P. D. Anz- • Investigate the implications of a second large Meador, NASA’s new breakup model of EVOLVE constellation operating in proximity to the 4.0, Adv. Space Res. 28 (2001) 1377–1384. OneWeb constellation; and [4] P.H. Krisko, Proper implementation of the 1998 • Investigate the implications of a collision within NASA break up model, Orbital Debris Quarterly the OneWeb constellation. Newsletter 15 (2011) 4-5. [5] J.-C. Liou, D.J. Kessler, M. Matney, G. Stansbery, A series of simulation studies, performed using the A new approach to evaluate collision probabilities DAMAGE code, were used to address each of these among asteroids, comets and Kuiper Belt objects, objectives.

IAC-19-A6.2.4 Page 10 of 11

Lunar and Planetary Science XXXIV, 2003, paper no. 1823. [6] J.-C. Liou, D.T. Hall, P.H. Krisko, J.N. Opiela, LEGEND – a three dimensional LEO-to-GEO debris evolutionary model, Adv. Space Res. 34

(2004) 981-986. [7] F.R. Hoots, L.L. Crawford, R.L. Roehrich, An analytic method to determine future close approaches between satellites, Celestial Mechanics 33 (1984) 143-158. [8] B. Bastida Virgili, J.C. Dolado, H.G. Lewis, J. Radtke, H. Krag, B. Revelin, C. Cazaux, C. Colombo, R. Crowther, M. Metz, Risk to space sustainability from large constellations of satellites, Acta Astronautica 126 (2016) 154-162.

IAC-19-A6.2.4 Page 11 of 11