A satellite placed in space is constantly affected by the space environment, resulting in various impacts from tempo - rary faults to permanent failures depending on factors such as satellite orbit, solar and geomagnetic activities, satellite local time, and satellite construction material. Anomaly events commonly occur during periods of high geomag- netic activity that also trigger plasma variation in the low Earth orbit (LEO) environment. In this study, we diagnosed anomalies in LEO satellites using electron data from the Medium Energy Proton and Electron Detector onboard the National Oceanic and Atmospheric Administration (NOAA)-15 satellite. In addition, we analyzed the fluctuation of electron flux in association with geomagnetic disturbances 3 days before and after the anomaly day. We selected 20 LEO anomaly cases registered in the Satellite News Digest database for the years 2000–2008. Satellite local time, an important parameter for anomaly diagnosis, was determined using propagated two-line element data in the SGP4 simplified general perturbation model to calculate the longitude of the ascending node of the satellite through the position and velocity vectors. The results showed that the majority of LEO satellite anomalies are linked to low-energy electron fluxes of 30–100 keV and magnetic perturbations that had a higher correlation coefficient (~ 90%) on the day of the anomaly. The mean local time calculation for the anomaly day with respect to the nighttime migration of energetic electrons revealed that the majority of anomalies (65%) occurred on the night side of Earth during the dusk- to-dawn sector of magnetic local time. Keywords: LEO satellite anomaly, Low-energy charged particles, Geomagnetic activity launch to avoid damage during operation; nevertheless, Introduction failures arise owing to many factors such as command Placing a satellite into space is a challenge owing not errors, mechanical and electrical faults, and design or only to the technical aspects of the mission requirements manufacturing problems as well as environmental effects but also because of the space environment of the satel- on the satellite (Vampola 1994). lite placement and operation. The variability of the space The causes of satellite failures are generally difficult to environment around the satellite can lead to multiple diagnose accurately; thus, they are often referred to as effects such as loss of performance in satellite subsys - anomalies. Documentation of numerous anomaly cases tems. Although some effects present low-level and tem - and knowledge about the proximate causes has led to an porary risks from which the satellite can recover, others understanding that plasma variation around satellites also present high-level risks that can result in notorious fail- plays an important role in anomalies through interaction ures that can permanently stop the satellite operation. between the plasma and satellite system. This interac - The reliability of satellites is tested and proved well before tion gives rise to various impacts on satellites depending on satellite orbit, relative position in space, satellite local time (SLT), solar and geomagnetic activities, and materi- *Correspondence: firstname.lastname@example.org; email@example.com Graduate School of System Informatics, Kobe University, 1-1 als in the satellite structure (Hastings and Garret 1996). Rokkodai-cho, Nada-ku, Kobe, Japan Full list of author information is available at the end of the article © The Author(s) 2018. This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creat iveco mmons .org/licen ses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. Ahmad et al. Earth, Planets and Space (2018) 70:91 Page 2 of 16 It is believed that satellite anomalies occur predomi- It is a challenge to study anomaly cases for satellites in the nantly during periods of high geomagnetic activity, which LEO environment because they were the most numerous can change the plasma properties abruptly and accelerate among active satellites in 2015, with a total of 696 (53.33%), electrons and ions to energies on the order of kiloelec- whereas satellites in GEO numbered about 481 (36.86%; tronvolts. Electrons and ions with energies above tens UCS 2015). That is, the population of LEO satellites is sig - of kiloelectronvolts are known to contribute greatly to nificantly higher than that of GEO satellites (SIA 2015). spacecraft charging phenomena (Lai 2012). Because the In this study, we undertook the diagnosis of the proxi- thermal speed of electrons is much larger than that of mate causes of anomalies on LEO satellites recorded on ions, the electron impact on satellites is of prime inter- the Satellite News Digest website (SND 2017) and evalu- est for satellite anomaly diagnosis, especially in low Earth ated their relationship with space environment variations orbit (LEO) environments below an altitude of 1000 km. during a specific period before and after the anomaly. Regarding LEO satellites, the effects of charging by The space environment was obtained primarily from low- plasma variation are not critical except in some regions energy electrons detected by the National Oceanic and such as polar regions and a small equatorial region Atmospheric Administration (NOAA)-15 satellite (NOAA known as the South Atlantic Anomaly. Plasma in those NCEI 2017) as well as by geomagnetic parameters repre- regions is cold and dense with typical energies less than sented by two indices: Kp and Dst (NASA GSFC 2016). 10 keV, although it can change rapidly during periods To acquire the SLT, an important factor for anomaly diag- of high geomagnetic activity. Rapid changes in plasma nosis, we applied the Simplified General Perturbation-4 properties around a satellite can be problematic for some (SGP4) code to two-line element (TLE) data for each sat- sensors and even for the satellite itself through charg- ellite obtained from Space-Track (2015). All of these steps ing effects. Plasma variations in LEO can be destruc - enabled us to diagnose the relationship between low- tive because the collective plasma can release sufficient energy electron fluxes and their associated magnetic per - energy to the satellite to cause effects such as surface turbations with regard to their effect on LEO satellites. charging, detector contamination, and surface chemi- cal reactions (Hastings 1995). Of interest is that satellite Data and method charging occurs in finite time until an equilibrium state NOAA‑15/MEPED electron data is reached (Lai 2012). Moreover, the current balance on The NOAA-15 satellite was launched on May 13, 1998, a satellite owing to its interaction with plasma depends and was placed at an altitude of 807 km with a polar incli- strongly on material properties as well as the flux of ener - nation of around 98.8°. This satellite is equipped with getic particles along the satellite trajectory. Space Environment Monitor 2 instruments including the One of the challenges in diagnosing satellite anomalies Medium Energy Proton and Electron Detector (MEPED), arises from the limited amount of anomaly data because which is a set of solid-state energetic particle detectors some anomalies have not been formally or thoroughly that measure the flux of protons and electrons within an documented (Koons et al. 2000). In addition, it is diffi - energy range of 30 keV to 200 MeV. To measure particles cult to maintain a comprehensive database in which all from different directions, the detectors are orientated in anomalies are categorized by, for example, satellite orbit the zenith (0°) and horizontal/perpendicular-to-zenith type, material properties, position on the anomaly day, (90°) directions and are referred to here as the 0° and 90° SLT, type of anomaly, and space weather at the time of detectors, respectively. The MEPED also has omnidirec - the anomaly. In addition, the database should contain tional detectors (Evans and Greer 2004), although these detailed descriptions of anomalies, including the initial data were not used in the present study. The MEPED presumption of cause, and it should be up to date. instrument measures only the electron fluxes in three Numerous studies have been published regarding sat- energy channels that span 30–2500 keV. Because the ellite anomaly events, including the estimation of causes, electron detectors are also sensitive to protons, as shown but the majority of research is more concerned with in Table 1, all electron data must be examined and cor- geosynchronous Earth orbit (GEO) satellites; few stud- rected to obtain better accuracy. ies have examined anomalies in LEO and medium Earth orbit satellites (Fennell et al. 2001; Belov et al. 2005; Table 1 MEPED electron energy channel detection Pilipenko et al. 2006; Iucci et al. 2006; Patil et al. 2008; and proton contamination ranges O’Brien 2009; Choi et al. 2011; Thomsen et al. 2013). In Channel Range (keV) Contaminant range (keV) addition, in the attempt to find anomaly causes related E1 30–100 210–2700 to environmental change, most analyses have used high- E2 100–300 280–2700 energy particle data at energies on the order of megaelec- E3 300–2500 440–2700 tronvolts for both electrons and protons. Ahmad et al. Earth, Planets and Space (2018) 70:91 Page 3 of 16 Table 1 shows the proton energy ranges at which each electron channels. The rates of CP for the same four electron channel is subject to contamination by protons anomaly cases are shown in Table 2. (CP) with an energy range of 210–2700 keV for E1 chan- The biggest CP occurred in channel E3 for both detec - nel. The E2 electron channel is sensitive to protons over tors and all cases. A higher energy corresponded to a the energy range of 280–2700 keV; and the E3 channel is larger CP. Moreover, it appears that the magnitude of affected by protons in the range of 440–2700 keV (Evans contamination is independent of satellite orbit. The data and Greer 2004). Protons with energies in these ranges from both equatorial and polar satellites showed similar contaminate the electron detector channels and affect magnitudes of CP. electron flux measurement; thus, their effects must be To confirm that the electron data were usable, the sec - removed from the data. Several methods can remove CP ond method was applied. The electron fluxes in each from electron data, as introduced by Lam et al. (2010), energy channel should be larger than twice the CP, here- Rodger et al. (2010), Asikainen and Mursula (2013), and after referred to as 2CP, because one contaminant will Peck et al. (2015). Because we used only short-term data result in one incorrect electron flux (Rodger et al. 2010). over the 7 days straddling the anomaly day, the electron The 2CP values are indicated by the light blue curves in data were examined by comparing the flux variations Fig. 1. Only electrons in channel E3 for both the 0° and between electrons and the CP within specific intervals, as 90° detectors were affected by CP. The increases in elec - done by Tadokoro et al. (2007). When the trend of elec- tron flux were followed by proton flux enhancement with tron fluxes differed from the proton fluxes, the electron similar patterns. The other channels, E1 and E2, were not data were regarded as contaminant-free. Furthermore, totally affected by the contaminant. Based on this confir - we also adopted the method introduced by Rodger et al. mation of usability, we removed the CP from the electron (2010), in which the electron data are deemed to have data using the aforementioned methods. acceptable quality if the electron flux in each energy Particle data from the 0° and 90° detectors were used in channel is larger than twice the CP flux. We applied these this study for the following reasons. When the NOAA- methods for removing CP from electron channels for 15 satellite passes through a high-latitude region, the 0° some anomaly cases, as shown in Fig. 1. detector measures precipitated particles, whereas the In the present study, we focus on the variation of elec- 90° detector counts trapped particles. Conversely, during tron flux during the 7-day interval straddling the anom - low-latitude passage, the 0° detector measures trapped aly day for four cases: two equatorial satellites, ASCA particles, whereas the 90° detector counts precipitated and FUSE (1), and two polar satellites, ERS-1 and Radar- particles (Asikainen and Mursula 2008). Owing to the sat-1(1). The other 16 cases are presented in the “ Appen- orthogonal orientation of the detectors, we attempted dix”. We used the hourly averages of electron flux to to reconcile the electron data from both detectors to match the magnetic field data. Figure 1 shows the elec- simplify the diagnosis process. In this study, reconcilia- tron flux variations for channels E1 (red lines), E2 (blue tion was performed by using the following steps. First, lines), and E3 (green lines). The trend of CP on the elec - hourly averages of electron data were selected from the tron fluxes is indicated by the black lines. Overall, the 0° and the 90° detectors. Second, the CP was removed protons affected only a small portion of the electron flux from all electron channels in both detectors by deduct- data: day of year (DOY) 197-198 for ASCA (Fig. 1a) and ing or subtracting the proton fluxes from the electron DOY 328-329 for FUSE (1) (Fig. 1b) for both the 0° and flux data. This subtraction was appropriate because one 90° detectors. Significant CP was not found in the other flux of contaminant will produce one incorrect electron two cases (Fig. 1c, d) in channel E1, whereas CP sig- flux (Rodger et al. 2010). This means that the presence nificantly affected the electron data in channels E2 and of proton flux in the electron detector will lead to incor - E3. It should be noted that CP did not occur exactly in rect measurement of electron flux; thus, it needs to be conjunction with the increase in electron flux: The pat - removed. In most cases, the electron fluxes were twice tern of the peaks was totally divergent. The average CP the size of the CP fluxes except for channel E3, indicat - in the electron channels was obtained from the following ing the domination of electron flux in the electron detec - formula: tor. Because we used only short intervals of 7 days, the trend lines showed that the contaminant affected only a small portion of the electron data, such as on DOY × 100% ≈ CR 197-198 and DOY 328-329 in Fig. 1a, b, respectively. ¯ ¯ where Γ and Γ are the average flux of protons and elec - u Th s, not all electron data during the selected interval p e trons, respectively, within the selected interval, and CR were contaminated significantly by protons, as shown is the percentage of proton flux contamination in the in Fig. 1. Third, we calculated the magnitude or result of Ahmad et al. Earth, Planets and Space (2018) 70:91 Page 4 of 16 x10 x10 2.5 ab 0° 0° E1 E2 1.5 E3 1 CP 0.5 2CP x10 x10 15 15 90° 90° 10 10 5 5 0 0 194195 196197 198199 200201 326 327 328 329 330 331 332 333 Days of Year 2000 Days of Year 2001 x10 x10 2.5 cd 0° 0° 1.5 0.5 0 0 x10 x10 6 4 90° 90° 0 0 67 68 69 70 71 72 73 74 328 329 330 331 332 333 334 335 Days of Year 2000 Days of Year 2002 Fig. 1 Variation of electron and proton fluxes during the 7-day interval straddling the anomaly day on a ASCA, b FUSE (1), c ERS-1, and d Radar - sat-1(1). The red, blue, and green curves represent electron fluxes for channels E1, E2, and E3, respectively, and the black and light blue curves indicate the variation of contaminating protons obtained from methods 1 and 2 in sequence Table 2 Proton contamination in each energy channel as a Kp and Dst indices as diagnosis parameters percentage of electron fluxes within ± 3 days of the anom- Geomagnetic data provide fundamental parameters for aly day for each case environmentally induced satellite anomaly research. These data, along with the energetic particle data, have Channel Proton contamination in 0°/90° detectors (%) been used to diagnose anomalies on LEO and GEO sat- ASCA FUSE (1) ERS‑1 Radarsat‑1(1) ellites. In this study, we used only the geomagnetic data E1 3.3/1.5 4.9/1.7 2.2/0.7 1.6/0.7 represented by the Kp and Dst indices because these E2 7.9/3.6 8.4/3.1 16.5/2.2 6.8/1.8 data have shown good agreement with anomaly events, E3 33.9/16.2 39.5/11.7 37.1/10.9 28.2/7.5 as reported by Belov et al. (2005), Patil et al. (2008), and Choi et al. (2011). The Kp index is an indicator of global geomagnetic dis - turbance counted by a scale of 0 to 9. In general, a larger the corrected electron flux data from the second step: Kp index corresponds to more anomaly occurrences. Far- 2 2 Γ = sqrt(Γ + Γ ) , where Γ corr 0 deg corr 0 deg corr 90 deg corr thing et al. (1982) showed that anomalies on the GOES and Γ are corrected electron fluxes from the 90 deg corr satellite series are clearly linked to Kp index values rang- 0° and 90° detectors, respectively. These data were used ing from 1 to 6. Their finding showed that the overall with the magnetic field data to diagnose the anomalies on tendency of anomalies increased with higher Kp. Fennell the LEO satellites. et al. (2001) also found that highly elliptical orbit satellite anomalies increased with an increase in Kp. Furthermore, Electron Fluxes [electrons/cm2/s/ster] Ahmad et al. Earth, Planets and Space (2018) 70:91 Page 5 of 16 Choi et al. (2011) found that although some GEO satellite Failure descriptions provided by the SND can be the anomalies were linked to lower Kp values, the number of initial step for anomaly diagnosis. The descriptions con - GEO anomalies increased with an increase in Kp. tain information about failure status, such as partial loss, The level of geomagnetic disturbance can also be indi - total loss, or contact loss, as well the satellite subsystem cated by the Dst index, which quantifies plasma changes sustaining damage, such as solar array failure, power in the ring current owing to those disturbances. Dur- drop, or reaction wheel failure. Failure descriptions, ing magnetic storms, the Dst index drops rapidly as the along with the anomaly time, were used to construct the energetic electron flux increases substantially (Fennell criteria to select the LEO satellites for analysis. As crite- et al. 2001). Several studies, such as Belov et al. (2005), ria used in this analysis, all failures owing to space debris pointed out that a numerous satellites in different groups were ignored. In addition, satellites were used only if the are linked to magnetic storms through the Dst index. difference between the perigee and apogee of their orbit Pilipenko et al. (2006) also found a similar relation- was less than 100 km or if their orbital eccentricity was ship between GEO anomalies registered in the National almost circular. As a result, all LEO anomaly cases in this Geophysical Data Center databases in conjunction with study involve satellites having orbits similar to that of the Dst index variation. They noted that not all anomalies NOAA-15 satellite, which assures the relevance of the occurred precisely at the time of significant drop in the electron data used in this study. Dst index. It should be noted that not all sources of LEO satellite In this study, a Dst index less than − 30 nT (Gonzales failure used in this study were explicitly stated; thus, we et al. 1994) and a Kp index in the range of 3 to 9 indi- presumed that the anomalies with unknown causes are cated a high level of geomagnetic activity. Such high associated with plasma variations triggered by geomag- levels are generally associated with geomagnetic storms netic disturbances. This is consistent with other studies and substorms. In some events, a geomagnetic substorm that attributed anomalies to environmental changes by occurs during the main storm phase (Wu et al. 2004) default when analysis from telemetry data was infeasible and can lead to partial ring current formation (Gonzales (Vampola 1994; Gubby and Evans 2002). The LEO satel - et al. 1994). The use of these bounds in the present study lite anomaly cases used in this paper are listed in Table 4. is intended to simply identify the levels of geomagnetic Some information in Table 4, such as altitude and incli- activity regarding anomaly occurrences. We also adopted nation, are somewhat different from the existing data an interval of 3 days before the anomaly day because in SND. We preferred to use satellite orbital data from Choi et al. (2011) found that such an interval has good Space-Track because it also provided the TLE data used agreement with the occurrence rate of the anomalies. We for SLT calculation for each satellite. As previously men- appended the interval of 3 days to examine the charac- tioned, we attempted to confirm the anomaly day for teristic environment after the anomaly day. Charged par- each satellite and found that only the ICESat anomaly day ticles can remain in the satellite orbital path during the had a discrepancy. ICESat was reported to suffer failure recovery phase of magnetic storms (Choi et al. 2011). on March 29, 2003, around 9:58 a.m. Eastern Standard Although Choi et al. (2011) focused on GEO satellites, we Time (EST), when the Geoscience Laser Altimeter Sys- found similar features in LEO (“Results and discussion” tem transmitter stopped emitting laser pulses (Kichak section). 2011). Hence, in this study, we preferred to use the date reported by the National Aeronautics and Space Admin- SND database of anomalies istration (NASA) for the ICESat anomaly. All other Numerous studies, such as Robertson and Stoneking anomaly days were obtained from the SND database. (2003) and Belov et al. (2005), have examined anomalies We also found that two satellites, FUSE and Radarsat-1, on LEO satellites using various databases. The former incurred anomalies more than once. Therefore, the first study used only a small number of LEO anomaly cases, anomalies were designated as FUSE (1) and Radar- whereas the latter study applied Kosmos data on anom- sat-1(1), and the second were FUSE (2) and Radarsat-1(2). alies attributed to high-energy particles. It is difficult to inventory the LEO satellite anomaly data exhaustively Results and discussion because most anomalies are not very accessible. In this Relationship between electron flux and magnetic flux study, we used only SND data but excluded some fail- We initially examined the strengths of geomagnetic dis- ures attributed to space debris. Moreover, we rechecked turbances affecting the properties of plasma in space. A and confirmed some SND data in terms of anomaly date previous study showed that the charge buildup on satel- and failure descriptions by using other sources, and the lites as a trigger of anomalies did not originate from the appropriate adjustments were made. magnetic perturbations but rather from energetic elec- trons affecting the satellite surfaces (Lam and Hruska Ahmad et al. Earth, Planets and Space (2018) 70:91 Page 6 of 16 Table 3 Comparison of mean SLT derived from SND index value was multiplied by 10 (Kp × 10) for scaling and the extracted TLE for selected cases purposes. We initially expected that the correlation would be Case UTC (SND) Local time Difference significantly higher during the period of solar maxi - SND conversion Extracted TLE mum in 2000 and 2001 because the electrons trapped #7 (Aqua) 1500 13:35:58 13:34:45 0:01:13 (−) in the magnetosphere are related to solar activity (Lam #11 (ICEsat) 2349 2:12:40 2:10:59 0:01:41 (−) et al. 2010). The anomalies ASCA, FUSE (1), and ERS-1 #12 (Midori) 1234 22:28:48 22:29:38 0:01:50 (+) (Fig. 2a–c, respectively) occurred during the solar maxi- #17 (HST ) 0258 2:52:32 2:51:42 0:01:50 (−) mum. Geomagnetic disturbances occurred several times, as indicated by the Kp and Dst indices. For the anomalies in these three satellites, the maximum Kp index values 1991). The properties of energetic particles and plasma were 9, 8.3, and 4.3, respectively, whereas the Dst index around Earth change in association with magnetic per- dropped to − 300, − 221, and − 51 nT, respectively. We turbations. In addition, some studies have shown that presumed that the relationship was not always individu- lower-energy electrons of <100 keV have fluctuated more ally consistent for every event owing to the different data - strongly with geomagnetic variability compared with sets. Furthermore, high-level geomagnetic disturbances higher-energy electrons (Pilipenko et al. 2006; Choi et al. were not always followed by increased electron flux. A 2011). The first study used the Dst index in association plausible explanation for the time lag between geomag- with electron fluxes of >1 MeV and >2 MeV, whereas the netic disturbance and electron flux enhancement rang - latter applied the Kp index corresponding to electron ing from hours to days is that it is driven by a mechanism fluxes of 50 keV to 1.5 MeV. Because the magnetic per - such as radial transport diffusion or pitch-angle scatter - turbations are believed to be indicative of plasma changes ing (Tadokoro et al. 2007). However, the overall trend for in space, we examined the relationship between both each anomaly case showed that lower-energy electrons indices together with electron flux variations within the (>30 keV, channel E1) had a good correlation with Kp and 7-day interval. We first evaluated the relationship of Kp, Dst values, as shown in Fig. 3. Dst, and electron flux in each detector energy channel In Fig. 3, the x-axes represent a sequence of seven of (E1, E2, and E3) and then identified the energy channel the satellite anomaly cases listed in Table 4. Because the most strongly associated with magnetic perturbations FUSE (2) and Yohkoh anomalies occurred within 5 days represented by both indices. It should be noted that the of each other, we merged their data for simplicity. Hence, electron fluxes from all three channels have the same in the diagnosis steps, both satellites were designated as number of datasets for each event. Because we can access Additional file 1: case #5, using #5a for FUSE (2) and #5b all magnetic data from NASA GSFC (2016), we preferred for Yohkoh. In general, the correlation trend was consist- to use the hourly averaged data for Kp and Dst and to ent for each anomaly case, where R for E1 was larger than adjust the 1-h resolution for electron flux data to match that for E2 and E3. Furthermore, the right-hand panel the resolution of the selected Kp and Dst data. The rela - shows a correlation with the Dst index, where for some tionship of each energy channel is expressed by a correla- cases the level of disturbance was less than − 100 nT tion coefficient, designated as R, which is shown in Fig. 2 (intense storm), such as in case #4 (Fig. 4b). For some for the same four anomaly cases as those given in Fig. 1. events, such as Additional file 2: case #14, no storm was Because we adopted the seven-day interval, which is present. We also noted that electron flux had a negative essentially short-term data, the images in Fig. 2 were cre- correlation with Dst because the ring current energy con- ated by averaging all data points (Dst and E1, E2, and E3 taining electrons and ions is inversely proportional to the data) corresponding to various Kp bins at intervals of 10: magnitude of Dst (Lohmeyer et al. 2012). Overall, the 0, 10, 20, …, 90. We then fitted the line for the averages of trend in Fig. 3 shows that the majority of lower-energy the bins rather than for the entire dataset. The red (blue) electrons are highly responsive to geomagnetic distur- spots in the figure represent a scatter diagram for the bance. Hence, in the following section, we diagnosed averaged Kp (Dst) index bins. The R values of each energy LEO satellite anomalies using only lower-energy electron channel are given at the top and bottom of each panel. data (channel E1). The R values show that the electrons with lower energy (E1) obviously have a stronger relationship with magnetic Diagnosis of LEO satellite anomalies perturbations than those with higher energy (E2 and E3). Satellite failure descriptions can provide clues for trac- This strong correlation was evident for both Kp and Dst ing the source of anomalies. The SND data used in this indices. Overall, the R values for channel E1 were larger study did not contain any assertive statements to defini - than those for channels E2 and E3. In this study, the Kp tively identify the cause of any anomalies. In addition, our Ahmad et al. Earth, Planets and Space (2018) 70:91 Page 7 of 16 ab R = 0.73 R = 0.79 R = 0.41R 100 = 0.05 R = 0.49 R = -0.02 100 100 0 40 0 -50 -50 -50 -100 -100 -40 -40 -100 -40 -150 -150 R = -0.2 R = -0.21 1 -80 -80 R = -0.31 R = -0.31 -150 R = -0.59 R = -0.59 -200 -80 -200 -120 -120 -200 -250 -250 R = -0.56 R = -0.56 R = -0.4 R = -0.48 8 -120 R = -0.3 R = -0.34 4 -160 -160 -300 -250 -300 10 11 12 13 14 9101112136 8101210111213149 10 11 12 13 78 91011 cd R = 0.77 R = 0.28 R = -0.27 50 50 50 60 50 R = 0.37 R = -0.46 R = 0.79 40 40 40 40 30 30 20 20 20 20 20 10 10 10 10 0 0 -10 -10 -10 -10 -10 -20 -20 -20 -20 -20 R = 0.34 R = 0.34 -20 -30 -40 -30 -30 -30 -30 -40 R = -0.86 R = -0.86 R = -0.42 R = -0.42 R = 0.1 R = 0.17 7 R = -0.7 R = -0.79 9 R = -0.6 R = -0.63 3 -40 -40 -40 -60 10 10.5 10 11.5 12 99.5 10 10.5 7.588.5 910.5111 11.5 2 10 10.5 11 11.5 8.599.5 10 Log electron fluxes [electrons/cm2/sec/ster] Fig. 2 Relationship of geomagnetic variability (Kp and Dst indices) and electron fluxes for channels E1 (left), E2 (middle), and E3 (right) around the day of anomaly for a ASCA, b FUSE (1), c ERS-1, and d Radarsat-1(1). The red and blue spots indicate the relationship with Kp and Dst data, respec- tively investigation to find the relationship between LEO anoma - • Radarsat-1 was discontinued by the Canadian Space lies and environmental changes was hindered by the limited Agency owing to a deteriorating ACS. This satellite amount of data. For example, we obtained only five anom - had previously incurred excessive friction and tem- aly cases for equatorial satellites, whereas the remaining perature in the primary momentum wheel in Sep- cases involved polar satellites. Furthermore, the majority of tember 1999. Its back-up wheel had a similar prob- anomalies were presented as technical failures, such as atti- lem in November 2002. No information about the tude loss and power drop, without identifying the cause of cause of the anomaly is available. those conditions, such as environment, debris, or phantom • Midori-2 (ADEOS-2) switched to “light load” mode command. The descriptions are summarized below. owing to an unknown anomaly. The power level fell from 6 kW to 1 kW. It was presumed that a rela- tionship might exist between the accident and solar • ERS-1 operated for almost nine years before being flares. Further investigation is needed. terminated by the European Space Agency owing to • No definite information was available on the causes failure in its attitude control system (ACS); no report of failure for satellites in the SND database (SND about the cause of the ACS failure was prepared. 2017) other than those detailed in this paper; there- • ASCA (Astro-D) lost its attitude, and a power drop fore, further study is needed. coincided with increased solar activity. Increas- ing incident solar radiation was suspected to impart torque to the satellite, which increased its drag. Because we were unable to obtain much conclusive Atmospheric drag strongly affects satellite orbit, information about the causes of anomalies from the rather than onboard systems, so we concluded that SND database or other sources (http://www.astro nauti the cause of this anomaly remains unknown. x.com and http://space fligh t101.com/space craft /satel • FUSE was reported to experience several reaction lite-catal og), we attributed these anomalies to environ- wheel failures in quick succession. Although it ini- mental changes by default. As the first step, we traced the tially recovered, the third and fourth reaction wheels changes in environmental conditions using the Kp and incurred damage several times until they failed com- Dst indices as well as the electron flux data from channel pletely in August 2007. No explanation of the failure E1, as shown in Fig. 4. cause is available. Dst [nT] Kp Dst [nT] Kp Ahmad et al. Earth, Planets and Space (2018) 70:91 Page 8 of 16 1 0.4 0.8 0.2 0.6 0 E1 0.4 -0.2 E2 E3 0.2 -0.4 0 -0.6 -0.2 -0.8 -0.4 -1 14 811121418 148 11 12 14 18 Case Number Fig. 3 Correlation coefficient between electron fluxes and (left) Kp index and (right) Dst index for channels E1 (red), E2 (blue), and E3 (light blue) for selected cases In the figure, the left y-axes represent the magnitude of as shown in Additional file 1: case #5b (Yohkoh; “Appen- geomagnetic disturbance for both the Kp (red curves) and dix”), whereas other events showed that the anomaly was Dst indices (blue curves), whereas the right y-axes des- related to Dst, as in case #4 (FUSE (1), Fig. 4b). Thus, the ignate the electron fluxes for channel E1 (green curves). use of both indices in this paper increased our ability to As previously discussed, the lower-energy electrons (E1) associate anomalies with geomagnetic phenomena. correlated significantly to magnetic perturbations. The Previous attempts to link anomalies to magnetic activ- interval inside the black dashed lines indicates the day ity, such as Fennell et al. (2001), showed that the majority on which the anomaly occurred. The first two panels of anomalies occurred during substorms, although not (Fig. 4a, b) show data for equatorial satellites (ASCA and all substorms led to anomalies. During a geomagnetic FUSE (1)), whereas the last two panels (Fig. 4c, d) show substorm, a numerous energetic particles driven by solar data for polar satellites (ERS-1 and Radarsat-1(1)). wind interact with local plasma in the geomagnetic tail It is of interest that the “peaks” of Kp and “valleys” of on the night side of Earth, resulting in plasma configu - Dst did not occur simultaneously owing to their differ - ration changes. The increased plasma energies from this ent complex mechanisms. A time lag existed sometimes process are injected toward Earth through plasma sheets. between two events. For example, in Fig. 4a, the magnetic Any satellites that cross these regions potentially expe- activity increased 2 days prior to the anomaly day (DOY rience charge buildup. Moreover, it is well established 195), although a decline in Dst (down to -43 nT) was pre- that an increase in particle flux associated with magnetic ceded by an increase in Kp (up to 6.3). We also found this activity occurs during a period of increasing anomalies pattern in other cases, as described in the “Appendix”. The (Welling 2010), although only 15% of anomalies occurred Kp index globally measures the magnetic activity at high during magnetically disturbed conditions. In this study, geomagnetic latitudes in which plasma in the magneto- we investigated the LEO satellite anomalies listed in tail is heated and is then transported earthward, whereas Table 4 according to three patterns: the Dst is measured at low latitudes linked to ring current disturbances. We attributed the time lag to differences 1. The anomaly coincided with the main phase of a in the observation method. In addition, the differences magnetic storm (pattern 1). could be related to the transport mechanism in which the 2. The anomaly occurred during the recovery phase of a energetic particles are first accelerated at high latitudes magnetic storm (pattern 2). and are then transported to drift into the ring current 3. High geomagnetic activities occurred during the (Prolss 2004). The transport mechanism likely leads to 3 days before or after the anomaly day (pattern 3). measurement time lags between Kp and Dst. The exist - ence of a time lag between Kp and Dst, especially at the These patterns were used in the present study to evalu - peak of geomagnetic disturbances, affects the identifica - ate anomalies on LEO satellites. However, only 60% of tion of satellite anomaly occurrences in association with anomalies (12 cases) fit one of the three patterns; 40% (8 Kp enhancement and Dst depression simultaneously. In cases) showed weak correlation to magnetic variability. some events, the anomaly had good correlation with Kp, Correlation Coefficient Ahmad et al. Earth, Planets and Space (2018) 70:91 Page 9 of 16 ab x10 x10 Asca Satellite AnomalyFuse1 Satellite Anomaly 100 2.5 100 Anomaly day Anomaly day 2.5 -50 2 1.5 -50 E1 -100 1.5 Kp Index -150 -100 Dst Index [nT] -200 -150 -250 0.5 0.5 -200 -300 -350 0 -250 0 194195 196197 198199 200201 326327 328329 330331 332333 Days of Year 2000 Days of Year 2001 Ers1 Satellite Anomaly x10 Radarsat 1(1) Satellite Anomaly x10 c d 60 6 60 4 Anomaly day Anomaly day 3.5 40 5 20 4 2.5 0 3 2 -20 1.5 -20 -40 -40 1 -60 0.5 -60 0 -80 0 67 68 69 70 71 72 73 74 328329 330331 332333 334335 Days of Year 2000 Days of Year 2002 Fig. 4 Variation of electron fluxes and geomagnetic disturbances during the 7-day interval straddling the anomaly day for a ASCA, b FUSE 1(1), c ERS-1, and d Radarsat-1(1). The red, blue, and green curves are the Kp index, Dst index, and electron flux in channel E1, respectively For the ASCA, FUSE (1), ERS-1, and Radarsat-1(1) anomaly occurred during the recovery phase of a mag- anomalies in Fig. 4, we found that all the above criteria netic storm (pattern 2). The disturbance of geomagnetic were satisfied by these cases collectively because mag - field strength occurred a few days prior to the anomaly netic storms followed by an increase in electron flux day and was severe on DOY 328, which is 1 day before occurred prior to and up to the anomaly day. In Fig. 4a the FUSE (1) anomaly day. The maximum Kp was 8.3, (ASCA), a magnetic storm began on DOY 195 (Kp 7 and and the minimum Dst was − 221 nT on DOY 328. We Dst − 43 nT). It is clear that electron flux enhancement found that the time lag between the Kp and Dst changes occurred gradually following the increase in Kp (~9) and was insignificant, as in the ASCA case. Therefore, we pre - decrease in Dst (~ − 289 nT) on the anomaly day (DOY sumed that energetic electrons were accelerated in the 197). This occurred during the main phase of a magnetic magnetotail plasma sheet and were then transported into storm (pattern 1). We did not observe a significant time the ring current. The insignificant time lag between Kp lag during this event; thus, we concluded that the activity and Dst changes for the ASCA and FUSE (1) anomalies was caused likely by the mechanism explained by Gon- confirm that both indices had good agreement with elec - zales et al. (1994), in which a large injection of energetic tron flux enhancement during periods of highly disturbed particles from the magnetotail into the ring current and geomagnetic activity. We also found this pattern in other high-latitude region is proportionally constant. Moreo- cases, such as Additional file 3: cases #12 (DART) and ver, the increased electron flux affected ASCA on the Additional file 4: cases #13 (Monitor-E; “Appendix”). anomaly day, which coincided with the increase in Kp Although FUSE is an equatorial satellite and the other and decrease in Dst. two satellites (DART and Monitor-E) are polar satellites, In contrast to the ASCA case, case #4 (FUSE (1), it appears that pattern 2 is independent of satellite orbit. Fig. 4b) appears to have different features, in which the Geomagnetic indices Geomagnetic indices Local Time Local Time Electron fluxes [electrons/cm2/s/ster] Electron fluxes [electrons/cm2/s/ster] Geomagnetic indices Geomagnetic indices Local Time Local Time Electron fluxes [electrons/cm2/s/ster] Electron fluxes [electrons/cm2/s/ster] Ahmad et al. Earth, Planets and Space (2018) 70:91 Page 10 of 16 Pattern 3 was found in the ERS-1 anomaly (Fig. 4c), DOY 328 that was followed by increased electron flux. where multiple storms occurred around the anomaly day. However, the impact on FUSE (1) was delayed. Because The strongest, on DOY 68 (Kp 4 and Dst − 51 nT), weak - several storms occurred prior to the anomaly day, it is ened to a similar maximum Kp but lower Dst strength still problematic to precisely determine which pattern (~ − 35 nT) on the anomaly day (DOY 70). In general, contributed to the anomaly. the electron flux fluctuated in step with the Kp variation, It remains unknown which energies are primary con- indicating that electron streams were injected earthward tributors to satellite anomalies (Choi et al. 2011). Gen- from the magnetotail and reached LEO with temporal erally, however, the use of the lower-energy channel fluctuation. For the Radarsat-1(1) anomaly in Fig. 4d, sev- (30–100 keV, channel E1) is relevant in this study owing eral magnetic storms also occurred on DOY 329 (Kp 3.7 to its sensitivity to geomagnetic disturbances, as demon- and Dst − 60 nT) and became stronger on the anomaly strated by the R values presented in Fig. 2. day (Kp 5 and Dst − 64 nT). A similar pattern was found To support this relationship, we adopted a method used in other cases, such as Terra (#3), Yohkoh (#5b), Radar- by Choi et al. (2011) to calculate the anomaly occurrence sat-1(2) (#8), Landsat 7 (#9), ICESat (#10), and Midori rate for LEO satellites. We selected the maximum Kp (#11; “Appendix”). Most LEO anomalies in this study are value, for example on the day of anomaly, for each anom- associated with pattern 3. aly case in Table 4. We counted the number of anomalies It is interesting that no anomalies occurred prior to the corresponding to each maximum Kp as well as the total reported anomaly day because several magnetic storms number of days of maximum Kp within the period 2000 obviously occurred before that day, as shown in Fig. 4 and to 2008. The anomaly occurrence rate was obtained by in the “Appendix”. To explain this phenomenon, we pro- dividing the number of anomalies by days for each maxi- pose the following scenarios. mum Kp. The same operation was performed for 3, 2, and 1 days prior to the anomaly day. We found that the anom- aly occurrence rate correlation coefficient (R) was highest 1. Anomalies occurred with a time delay between storm on the day of the anomaly, as shown in Fig. 5. occurrence and incoming electrons near the satellite, In contrast to that reported by Choi et al. (2011), in as shown in Fig. 4b, c. This was also found in some which the 2-day window (3 days prior to anomaly day) anomaly cases studied by Farthing et al. (1982), Lam had good correlation with the Kp index, we found that and Hruska (1991), and Iucci et al. (2006). the 0-day window (on the anomaly day) had higher cor- 2. The link between LEO anomalies and environmen - relation with Kp (R ~ 90%), as shown in Fig. 5 (panel 4). tal change was weak in some cases, specifically #5a We presumed that despite the orbital dependence, the [FUSE (2)], #6 (Aqua), and from case #14 (Kirari) to correlation also varied case by case. As shown in Fig. 5, case #19 (Orbcomm) in Table 4 (Additional files 2, especially panel 4, it is obvious that a higher Kp index 11, 12, 13, 14, and 15) in Appendix. Weak linkage was corresponded overall to a higher number of satellite also found in GEO anomaly cases studied by Gubby anomalies. and Evans (2002). 3. Anomalies were not promptly logged and docu- mented by the satellite operator when the anomaly -3 days occurred. The anomaly occurrence was logged and R = 0.7855 & slope = 0.0653 0.5 documented several days after the event, resulting in an inaccurate local time for the anomaly. This evi - dence was found by one of the authors during analy- R = 0.7681 & slope = 0.087 0.5 sis of some anomaly cases using unpublished data -2 days from a satellite operator. 0 R = 0.8415 & slope = 0.0837 As another interesting point, we found that some 0.5 -1 day anomalies occurred on the cuspate gradients, where the curve of electron flux and Kp and Dst indices change steeply. This feature was also found by Gubby and Evans Anomaly day R = 0.9015 & slope = 0.0602 0.5 (2002). In Fig. 4a, the sharp drop in Dst on DOY 197 indi- cates an abrupt change in electron streams, especially in 012345678 9 the ring current. The electron flux fluctuated rapidly dur - Kp index ing this period. A similar pattern is shown in Fig. 4c on Fig. 5 LEO satellite anomaly occurrence rates for (from top) 3 days DOY 70 and in Fig. 4d on DOY 331. Contrary to these (panel 1), 2 days (panel 2), and 1 day (panel 3) prior to the anomaly cases, Dst in Fig. 4b shows a clear cuspate gradient on day and on the anomaly day (panel 4) Occurrence rate of LEO anomalies Ahmad et al. Earth, Planets and Space (2018) 70:91 Page 11 of 16 We have shown that a strong relationship occurred between the electron flux in channel E1 and magnetic perturbations through Kp and Dst indices, for which the average R values were around 75% and 60%, respectively (Fig. 2). We have also demonstrated a strong relationship between the anomaly occurrence rate and the Kp index, which was higher on anomaly days. Hence, we inferred that lower-energy electrons also play an important role in satellite anomalies, owing primarily to satellite charging (Fennell et al. 2001). It is of interest that anomalies on the DMSP F6 and F7 satellites, at an altitude of 840 km, were linked to elec- trons with energies over 14 keV (Gussenhoven et al. 1985). In that study, the satellite potential was strongly related to the variation in electron flux with energies on the order of tens of kiloelectronvolts, which dropped to a potential level of − 462 V. Other simulations on the DMSP F13 satellite also found the same malfunc- tion, known as electrostatic discharge, which subjected the satellite to energetic electrons of 31.3 keV (Ander- Fig. 6 Procedure for LEO satellite local time calculation son and Koons 1996). We chose the low-energy electron channel (30–100 keV) in this study for the following rea- sons. First, it is sensitive to geomagnetic disturbances, as shown in “Relationship between electron flux and determined the relative position of a satellite with respect magnetic flux ” section. In addition, the electron data to the Sun through the longitude of ascending node from NOAA-15 are readily accessible; this energy range (LAN) parameter. This parameter indicates the longitude is partly attributed to surface charging (Anderson 2012) of a satellite in an equatorial plane eastward, where the or internal charging (Fennell et al. 2001). Moreover, Lam Sun is offset to that plane (Vallado and McClain 1997). et al. (2010) reported consistency in local time distribu- The satellite reaches the ascending node at a particular tion between the precipitation and injection of electrons SLT every day. The important point is that the calcula - (channel E1) in the nightside region owing to a mecha- tion of SLT using LAN will not vary significantly owing nism known as whistler mode chorus wave resonance. to the small nodal regression rate, which is less than 0.5° u Th s, the use of the lower electron channel in this study in our case or around 2 min/day assuming that 1 solar is appropriate for the diagnosis process. hour equals 15°. Hence, although LAN oscillates, it does not affect the SLT variance significantly at an hourly time SLT scale. The SLT calculation was performed only for the The SLT is defined in many anomaly cases because of its satellite anomalies given in Table 4, for which the SLT connection with magnetic field fluctuations (Vampola was not provided in the SND data. The anomaly times 1994); its dynamical process has been detailed by Vam- were provided in SND only for Aqua (1500 UTC), Midori pola (1994) and Lai (2012). The majority of anomalies, (2349 UTC), HST (1234 UTC), and ICESat (0258 UTC). especially in GEO satellites, occur during the midnight- The last time was obtained from Kichak (2011). to-dawn sector of magnetic local time (MLT). SND The TLE dataset from Space-Track is very important provides only a limited number of anomaly times and in this technique because it provides orbital parameters. is given mostly in Coordinated Universal Time (UTC); The use of TLE data corresponding to the anomaly day is therefore, it is important to determine the SLT because of the primary concern for the calculation. Detailed explana- its relationship with the migration of energetic particles tions of the TLE data are provided at www.space -track . owing to magnetic perturbations. It is more complicated org. For the first step, we extracted TLE data using the to find anomaly tendencies associated with SLT for LEO SGP4 code to obtain the position and velocity vectors anomalies owing to relative satellite position changes of each satellite. The SGP4 is an orbital propagator that over time. However, we applied a method using celestial computes drag effects and estimates orbital parameters, mechanics for this calculation, as shown in Fig. 6. as discussed in detail by Vallado et al. (2006). Both posi- Because the earthward injection of energetic particles tion and velocity vectors are crucial for nodal vector cal- from the magnetotail occurs on the nightside, we initially culations as well as the satellite LAN, i.e., the angle in the Ahmad et al. Earth, Planets and Space (2018) 70:91 Page 12 of 16 equatorial plane where the satellite crosses from south to for the diagnosis process. By considering the estimated north (Vallado and McClain 1997). In parallel, we can also time resolution for actual local time, that is, 2 min, we derive the satellite epoch in UTC from the extracted TLE presumed that the time resolution of SLT obtained from data. Both the LAN and epoch are then used to calculate TLE data extraction is approximately 4 min. Further- the SLT for the anomaly. It is important to note that the more, we located the satellite positions in Table 4 during SLT obtained from this method was the mean SLT when the anomaly day using the propagated TLE data obtained the satellite crossed the LAN on the day of the anomaly. from the SGP4 code. Thus, we obtained the latitude and The vertical blue dashed lines in Fig. 4 show the mean longitude of the satellites and then converted them into SLT for these four anomaly cases, as obtained from the the geomagnetic frame using the converter at http:// aforementioned method using the extracted TLE data wdc.kugi.kyoto -u.ac.jp/igrf/gggm. By taking the average for the day of the satellite anomaly. Because the TLE data period of satellites per orbit (Table 4) as around 90 min, for satellites were recorded mostly during the ascending we estimated the resolution of the satellite position to be phase, the TLE epoch can be expressed as the time at approximately 8°. Figure 7 summarizes the MLT sectors which the satellite passes through the LAN in the equato- and approximate satellite locations on the day of their rial plane. Therefore, we can estimate the position of the anomaly. The red spots represent the approximate loca - satellite with respect to the local time sector (midnight, tions of the corresponding satellites listed in Table 4 with dawn, noon, or dusk). respect to the geomagnetic frame. As previously discussed, discrepancies exist among the The largest numbers of anomalies appeared in two local time of magnetic storm events, increased electron sectors: the noon-to-dusk and dusk-to-midnight sec- flux, and anomaly occurrence on the satellite. We cannot tors. This pattern is similar to that reported by Lam and always expect that satellite anomalies occur immediately Hruska (1991), who found anomalies mostly distributed as the magnetic activity becomes very active. Figure 4b within the noon-to-dusk and evening sectors. As previ- makes this evident: The FUSE (1) anomaly (DOY 329) ously discussed, the time delay between magnetic storms occurred after the most severe part of a magnetic storm and anomalies may explain why the MLT of anomalies (DOY 328). Hence, we assumed that time delays also was predominantly spread within the above two sectors. existed for local times of anomalies in other cases. This In addition, we also suggest that the anomaly occurrences time delay is likely also related to satellite position during in LEO are slightly different from those of GEO. Dur - magnetically disturbed conditions. For example, the time ing magnetic storms, the accelerated electrons and ions delay can be very short if the satellite is located near the reach the GEO altitude at around midnight (Lai 2012). ionospheric magnetic foot point where the substorm is It is clear that the majority of GEO anomalies occurred initiated. Conversely, the time delay can be large when the from the midnight-to-dawn sector of MLT. In contrast satellite is located in the ring current because buildup of to the GEO satellite anomalies, the LEO satellite anoma- the ring current takes hours or days. Moreover, the local lies are generally affected by geomagnetic activity and time of anomalies in this study tended to coincide with the the precipitation of electron fluxes in the upper atmos - increasing phase of electron flux, as shown in Fig. 4 a, c, d. phere. Because the times of high geomagnetic activity We then further examined the local times given in and electron flux do not always coincide, we suspect that the SND database for Aqua, Midori, HST, and ICE- this leads to discrepancies in the local time distribution Sat and compared them with the extracted TLE data of LEO satellite anomalies. It should be noted that the used for the other cases. To convert the UTC into SLT, distribution of pre-midnight anomalies in Fig. 7 appears the LAN parameter and UTC were used together to be related to scattered electrons associated with whis- (SLT = LAN/15 + UTC). Because most cases have tler mode chorus wave resonance (Lam et al. 2010), in very small nodal regression, the time resolution can be which the wave intensities and the precipitating electron estimated at around 2 min. Table 3 compares the SLTs flux increase simultaneously when geomagnetic activity obtained by UTC conversion and those calculated with increases. Moreover, Wing et al. (2013) reported that the the extracted TLE. pre-midnight pattern is likely also associated with mono- As shown in the table, the differences between the energetic electrons that are accelerated by the electric mean SLT obtained from the SND–UTC time conver- field or low-frequency Alfven waves. They showed that sion and that derived from extracted TLE were less than the aurora electron flux is distributed predominantly in 2 min. The negative or positive sign in the bracket in the the pre-midnight sector and that it dramatically increases “Difference” column indicates whether the TLE-derived after the substorm onset. SLT was ahead or behind the actual local time. Because In addition, we compared the local time morphology the time differences were less than 2 min, we concluded using the anomaly cases in our study to that reported by that the SLT calculation can be applied to the other cases Anderson (2012). The following results were noted. Ahmad et al. Earth, Planets and Space (2018) 70:91 Page 13 of 16 6 MLT 10° 30° 50° 70° 12 MLT 0 MLT 19 3 18 MLT Fig. 7 Distribution of local time for LEO satellites within the anomaly day Fig. 8 Distribution of local time of LEO DMSP satellite anomalies (Anderson 2012) • Our study used multiple LEO satellites with different orbits (Table 4). storm (pattern 2), as shown in FUSE (1) (case #4; Fig. 4b), • Anderson used a series of LEO DMSP satellites at the DART (Additional file 3: case #12), and Monitor-E (Addi- same Sun-synchronous orbits (840 km altitude and tional file 4: case #13); or they were attributed to multiple 99° inclination). storms prior to and after the anomaly day (pattern 3), such • Our study related LEO anomalies to lower-energy as that in Radarsat-1(1) (case #7; Fig. 4d), Terra (Addi- electron fluxes (30–100 keV) and magnetic perturba - tional file 5: case #3), Yohkoh (Additional file 1: case #5b), tions (indicated by Kp and Dst indices). Radarsat-1(2) (Additional file 6: case #8), Landsat 7 (Addi- • Anderson attributed anomalies to charging and used tional file 7: case #9), ICESat (Additional file 8: case #10), lower-energy electrons (30–31.3 keV). and Midori (Additional file 9: case #11). We also noted • Our study found that the proximate local times of that among these three patterns, the LEO anomalies were anomalies were distributed mostly from dusk to mid- linked most often to pattern 3 (40%). For the remaining night (Fig. 7). cases, such as FUSE (2) (Additional file 1: case #5a), Aqua (Additional file 10: case #6), and from case #14(Kirari) to • Anderson found that the local times of DMSP anom- case #19 (Orbcomm) (Additional files 2 , 11, 12, 13, 14, alies occurred from dusk to midnight (Fig. 8). and 15) in Table 4, the anomaly occurrences were weakly linked to the geophysical parameters used in this study. Nevertheless, the contribution of lower-energy electron Figures 7 and 8 show similar morphologies of the local fluxes and their associated geomagnetic disturbances have time of anomalies, with distribution mostly within the played a role in these cases. In particular, the anomaly on dusk-to-midnight sector. Although this morphology var- FUSE (2) (Additional file 1: case #5a) resembled the FUSE ies depending on the case, both our study and Anderson (1) (case #4; Fig. 4b) malfunction and might represent (2012) showed good correlation between LEO anomalies worsening of the initial damage to the FUSE satellite. and lower-energy electrons. The determination of SLT using the LAN parameter resulted in small deviations between local time derived Conclusion from SND data and that extracted from TLE data. The We investigated the proximate cause of anomalies on LEO deviations were less than 2 min, as shown in Table 3. satellites and found that around 60% of anomalies in this Although the LAN parameter changed over time, its study were strongly related to lower-energy electron fluxes oscillation per day was very small, resulting in discrep- of 30–100 keV and to associated magnetic perturbations ancies of minutes and seconds but not hours (Table 3, through Kp and Dst indices. The anomalies tended to fol - column 4). Because this parameter can represent the low three patterns: They occurred during the main phase position of the satellite relative to the Sun, it is relevant of a magnetic storm (pattern 1), as incurred by ASCA (case for determining the mean SLT with respect to the Sun in #2; Fig. 4a); they coincided with the recovery phase of a Ahmad et al. Earth, Planets and Space (2018) 70:91 Page 14 of 16 and electron detector; MLT: Magnetic local time; SGP4: Simplified general association with energetic particle injection in the night- perturbations-4 (model); SLT: Satellite local time; SND: Satellite news Digest; side magnetotail. TLE: Two-line element; UTC : Coordinated universal time. Figure 7 shows that the local times of LEO anomalies Authors’ contributions were scattered primarily within two sectors: pre-dusk NA was involved in research of LEO satellite anomalies associated with space and pre-midnight. Because we intended to relate the weather parameters and contributed to the manuscript and figure prepara- anomalies to the migration of low-energy electrons in tion. DH was responsible for space weather-related anomaly research and wrote the manuscript. TD prepared the analyses of satellite orbits, applied the the nightside of the magnetosphere, we inferred that the SGP4 code used in this research, and assisted with the manuscript revision. HU LEO anomalies in this study occurred predominantly in and YM were involved in the spacecraft and plasma interaction research and the dusk-to-dawn sector of MLT. This phenomenon was provided comments and suggestions for the manuscript. All authors read and approved the final manuscript. noted in 65% of the occurrences. During geomagnetic disturbances, energetic electrons accelerated in the mag- Authors’ information netotail plasma sheet and drifted into the ring current. Nizam Ahmad: The author works for Indonesian National Institute of Aeronautics and Space (LAPAN) in the field of spacecraft anomaly and celestial mechanics A large portion of these energetic electrons, which have research. Currently, the author is taking a Ph.D. course in the Graduate School complex motions, were lost, precipitated into the upper of System Informatics, Kobe University, Japan, and conducting research related atmosphere, and immersed LEO satellites in electron to spacecraft and plasma interaction through the Electro-magnetic Spacecraft Environment Simulator (EMSES). The Ph.D. study program has been registered fluxes. This was evident in the flux variation in lower- under sponsorship of the Ministry of Research and Technology, Secretariat energy channels as a result of magnetic perturbations of Project Management Office (PMO) Research and Innovation in Science represented by the Kp and Dst indices. and Technology Project (RISET-PRO) (REf.No.:7/RISET-Pro/SFS/I/2014). Dhani Herdiwijaya: Currently, the author serves as senior researcher and lecturer at the Additional files Institut Teknologi Bandung (ITB), Indonesia, in the field of astronomy, specifically in solar physics and space plasma. Thomas Djamaluddin: The author serves as chair of the Indonesian National Institute of Aeronautics and Space (LAPAN). The Additional file 1. Variation of geophysical parameters around the author also serves as professor in the field of astronomy and celestial mechanics. anomaly day of FUSE (2) (case #5a) and Yohkoh (case #5b) satellites. Hideyuki Usui: The author serves as professor in the Graduate School of System Informatics, Kobe University, Japan. The author has also conducted research Additional file 2. Variation of geophysical parameters around the in the field of space plasma including its interaction with spacecraft as well as anomaly day of Kirari satellite (case #14). wave–particle interactions in space plasmas. Yohei Miyake: The author serves Additional file 3. Variation of geophysical parameters around the as associate professor in the Graduate School of System Informatics, Kobe anomaly day of DART satellite (case #12). University, Japan (H. Usui’s research group) and has written and published many papers related to spacecraft and plasma interaction. Additional file 4. Variation of geophysical parameters around the anomaly day of Monitor-E satellite (case #13). Author details Additional file 5. Variation of geophysical parameters around the Graduate School of System Informatics, Kobe University, 1-1 Rokkodai-cho, anomaly day of Terra satellite (case #3). Nada-ku, Kobe, Japan. National Institute of Aeronautics and Space (LAPAN), Additional file 6. Variation of geophysical parameters around the Jl. Dr. Djunjunan 133, Bandung, Indonesia. Astronomy Research Division anomaly day of Radarsat 1(2) satellite (case #8). and Bosscha Observatory, Bandung Institute of Technology, Jl. Ganesha No. 10, Bandung, Indonesia. National Institute of Aeronautics and Space (LAPAN), Jl. Additional file 7. Variation of geophysical parameters around the Pemuda Persil No. 1, Jakarta, Indonesia. anomaly day of Landsat 7 satellite (case #9). Additional file 8. Variation of geophysical parameters around the Acknowledgements anomaly day of ICEsat satellite (case #10). The authors would like to thank Kobe University, Institut Teknologi Bandung, and the Indonesian National Institute of Aeronautics and Space for provid- Additional file 9. Variation of geophysical parameters around the ing facilities and other support for this research. The main author specifically anomaly day of Midori satellite (case #11). thank the Ministry of Research and Technology under Secretariat of Project Additional file 10. Variation of geophysical parameters around the Management Office (PMO) Research and Innovation in Science and Technology anomaly day of Aqua satellite (case #6). Project (RISET-PRO) for supporting and facilitating activities during the Ph.D. study program. Also appreciated are the U.S. National Oceanic and Atmos- Additional file 11. Variation of geophysical parameters around the pheric Administration, U.S. National Aeronautics and Space Administration, and anomaly day of KOMPASS 2 satellite (case #15). Space-Track, which provided much of the data used in this study. Special thanks Additional file 12. Variation of geophysical parameters around the are extended to Dr. Kelso, of Celestrak, for his instruction on applying SGP4 to anomaly day of HST satellite (case #16). local time calculations. In addition, David Eagle, of the MathWorks community, is appreciated for sharing the orbital and celestial mechanics codes. Finally, the Additional file 13. Variation of geophysical parameters around the authors express gratitude to the reviewers for providing helpful suggestions. anomaly day of MetOp-A satellite (case #17). Additional file 14. Variation of geophysical parameters around the Competing interests anomaly day of Orbview 3 satellite (case #18). The authors declare that they have no competing interests. Additional file 15. Variation of geophysical parameters around the anomaly day of Orbcomm satellite (case #19). Consent for publication Not applicable. Ethics approval and consent to participate Not applicable. Abbreviations ACS: Attitude control system; CP: Contamination by protons; DOY: Day of year; Funding EST: Eastern standard time; GEO: Geosynchronous earth orbit; LAN: Longitude The submission of the manuscript was funded by Kobe University. of ascending node; LEO: Low Earth orbit; MEPED: Medium energy proton Ahmad et al. Earth, Planets and Space (2018) 70:91 Page 15 of 16 Appendix Variation of electron fluxes and Kp and Dst indices within selected intervals for other anomaly cases given in Table 4. Table 4 LEO satellite anomalies during the period 2000–2008 including brief description of failure Case Satellite Anomaly Alt. Incl. Anomaly description Name Date (km) (deg) #1 ERS 1 03/10/00 772 98 Total loss #2 ASCA 07/15/00 570 31 Safe mode, total loss #3 Terra 10/26/00 702 98 Telemetry monitor error #4 FUSE (1) 11/25/01 752 24 X-axis reaction wheel error #5a FUSE (2) 12/10/01 752 24 Y-axis reaction wheel error #5b Yohkoh 12/15/01 575 31 Loss of control #6 Aqua 06/27/02 702 98 Single event upset #7 Radarsat 1(1) 11/27/02 792 98 Loss of attitude #8 Radarsat 1(2) 12/30/02 792 98 Attitude control problem #9 Landsat 7 05/31/03 702 98 Thematic mapper failure #10 ICEsat 03/29/03 595 94 One of three lasers aboard failed #11 Midori 10/24/03 805 98 Total loss #12 DART 04/15/05 554 96 Navigational error #13 Monitor-E 10/18/05 527 97 Loss of attitude control #14 Kirari 11/24/05 593 97 One of four reaction wheels failed #15 KOMPASS 2 05/29/06 422 78 Various malfunctions #16 HST 06/30/06 564 28 ACS instrument fail #17 MetOp-A 11/04/06 821 98 Temporary payload shutdown #18 Orbview 3 03/04/07 707 97 Stopped sending usable imagery #19 Orbcomm 11/10/08 758 98 Satellite operation problems infrastructure. Part of the NATO sciences series II: mathematic, phys- ics and chemistry book series, vol 176. Kluwer Academic Publishers, Publisher’s Note Dordrecht, pp 147–163 Springer Nature remains neutral with regard to jurisdictional claims in pub- Choi HS, Lee J, Cho KS, Kwak YS, Cho IH, Park YD, Kim YH, Baker DN, Reeves lished maps and institutional affiliations. GD, Lee DK (2011) Analysis of GEO spacecraft anomalies: space weather relationship. Space Weather 9(6):1–12. https ://doi.org/10.1029/2010S Received: 19 February 2018 Accepted: 25 April 2018W0005 97 Evans DS, Greer MS (2004) Polar orbiting environmental satellite space envi- ronment monitor-2: instrument description and archive data documen- tation. NOAA technical memorandum OAR SEC 93. Available at ftp:// satda t.ngdc.noaa.gov/sem/poes/docs/sem2_docs/2006/SEM2v 2.0.pdf. Accessed 21 November 2017 References Farthing WH, Brown JP, Bryant WC (1982) Differential spacecraft charging on Anderson PC (2012) Characteristics of spacecraft charging in low Earth orbit. J the geostationary operational environmental satellites. NASA technical Geophys Res 117:A07308.https ://doi.org/10.1029/2011J A0168 75 memorandum 83908. Available at https ://ntrs.nasa.gov/archi ve/nasa/ Anderson PC, Koons HC (1996) Spacecraft charging anomaly on a low-altitude casi.ntrs.nasa.gov/19820 01848 0.pdf. Accessed 7 November 2017 satellite in an aurora. J Spacecraft Rockets 33(5):734–738. https ://doi. Fennell JF, Koons HC, Roeder L, Blake JB (2001) Spacecraft charging: observa- org/10.2514/3.26828 tions and relationship to satellite anomalies. In: Harris RA (ed) Proceeding Asikainen T, Mursula K (2008) Energetic electron flux behaviour at low of the seventh international conference, ESTEC, Noordwijk, the Nether- L-shells and its relation to the South Atlantic Anomaly. J Atmos Terr Phys lands, 23–27 April 2001 70:532–538 Gonzales WD, Joselyn JA, Kamide Y, Kroehl HW, Rostoker G, Tsurutani BT, Asikainen T, Mursula K (2013) Correcting the NOAA/MEPED energetic electron Vasyliunas VM (1994) What is geomagnetic storm? J Geophys Res fluxes for detector efficiency and proton contamination. J Geophys Res 99(A4):5771–5792. https ://doi.org/10.1029/93JA0 2867 118(10):6500–6510. https ://doi.org/10.1002/jgra.50584 Gubby R, Evans J (2002) Space environment effects and satellite design. Belov A, Dorman L, Iucci N, Kryakunova O, Ptitsyna N (2005) The relation J Atmos Terr Phys 64(16):1723–1733. https ://doi.org/10.1016/S1364 of high- and low-orbit satellite anomalies to different geophysical -6826(02)00122 -0 parameters. In: Daglis IA (ed) Eec ff ts of space weather on technology Ahmad et al. Earth, Planets and Space (2018) 70:91 Page 16 of 16 Gussenhoven MS, Hardy DA, Rich F, Burke WJ, Yeh HC (1985) High-level space- Pilipenko V, Yagova N, Romanova N, Allen J (2006) Statistical relationship craft charging in the low-altitude polar auroral environment. J Geophys between satellite anomalies at geostationary orbit and high-energy Res 90(A11):11009–11023. https ://doi.org/10.1029/JA090 iA11p 11009 particles. Adv Space Res 37(6):1192–1205. https ://doi.org/10.1016/j. Hastings DE (1995) A review of plasma interactions with spacecraft in asr.2005.03.152 low Earth orbit. J Geophys Res 100(A8):14457–14483. https ://doi. Prolss GW (2004) Physics of the Earth’s space environment: an introduction. org/10.1029/94JA0 3358 Springer, Berlin Hastings D, Garret H (1996) Spacecraft environment interaction. Cambridge Robertson B, Stoneking E (2003) Satellite GN&C anomaly trends. In: 2003 AAS University Press, Cambridge guidance and control conference, American Astronautical Society guid- Iucci N, Dorman LI, Levitin AE, Belov AV, Eroshenko EA, Ptitsyna NG, Villoresi G, ance and control conference, Breckenridge, United States, 5–9 February Chizhenkov GV, Gromova LI, Parisi M, Tyasto MI, Yanke VG (2006) Space- 2003 craft operational anomalies and space weather impact hazards. Adv Rodger CJ, Clilverd MA, Green JC, Lam MM (2010) Use of POES SEM-2 observa- Space Res 37(1):184–190. https ://doi.org/10.1016/j.asr.2005.03.028 tions to examine radiation belt dynamics and energetic electron precipi- Kichak RA (2011) Independent GLAS anomaly review board executive sum- tation into the atmosphere. J Geophys Res. https ://doi.org/10.1029/2008J mary. Available at https ://icesa t.gsfc.nasa.gov/icesa t/docs/IGARB .pdf. A0140 23 Accessed 25 July 2017 Satellite News Digest (SND) (2017) Satellite outages and failures. https ://www. Koons HC, Mazur JE, Selesnick RS, Blake JB, Fennell JF, Roeder JL, Anderson sat-nd.com. Accessed 2 August 2017 PC (2000) The impact of the space environment on space systems. In: SIA (2015) State of the satellite industry report – September 2015. Available at Proceedings of the 6th spacecraft charging conference, AFRL science http://www.sia.org Accessed 22 Desember 2015 center, 1 September 2000 Space-Track (2015) Two-line element set database. https ://www.space -track . Lai ST (2012) Fundamentals of spacecraft charging: spacecraft interactions org. Accessed 18 November 2015 with space plasmas. Princeton University Press, Princeton Tadokoro H, Tsuchiya F, Miyoshi Y, Misawa H, Morioka A, Evans DS (2007) Elec- Lam HL, Hruska J (1991) Magnetic signatures for satellite anomalies. J Space- tron flux enhancement in the inner radiation belt during moderate mag- craft 28(1):93–99. https ://doi.org/10.2514/3.26214 netic storms. Ann Geophys 25(6):1359–1364. https ://doi.org/10.5194/ Lam MM, Horne RB, Meredith NP, Glauert SA, Griffin TM, Green JC (2010) Origin angeo -25-1359-2007 of energetic electron precipitation >30 keV into the atmosphere. J Geo- Thomsen MF, Henderson MG, Jordanova VK (2013) Statistical properties of the phys Res 115(A4):A00F08. https ://doi.org/10.1029/2009J A0146 19 surface-charging environment at geosynchronous orbit. Space Weather Lohmeyer W, Cahoy K, Baker DN (2012) Correlation of GEO communication 11(5):237–244. https ://doi.org/10.1002/swe.20049 satellite anomalies and space weather phenomena: improved satellite Union of Concerned Scientists (UCS) (2015) UCS satellite database. https :// performance and risk mitigation. In: 30th AIAA international commu-www.ucsus a.org. Accessed 22 December 2015 nications satellite system conference (ICSSC), Ottawa, Canada, 24–27 Vallado DA, McClain WD (1997) Fundamentals of astrodynamics and applica- September 2012. https ://doi.org/10.2514/6.2012-15083 tions. McGraw-Hill, New York NASA’s Goddard Space Flight Center (NASA GSFC)(2016) OMNIweb data docu- Vallado DA, Crawford P, Hujsak R, Kelso TS (2006) Revisiting Spacetrack report mentation. https ://omniw eb.gsfc.nasa.gov/html/ow_data.html. Accessed #3. In: AIAA/AAS astrodynamics specialist conference and exhibit, Key- 15 September 2016 stone, Colorado, 21-24 August 2006. https ://doi.org/10.2514/6.2006-6753 NOAA’s National Center for Environmental Information (NOAA NCEI) (2017) Vampola AL (1994) Analysis of environmentally induced spacecraft anomalies. POES space environment monitor data and documentation. http://satda J Spacecraft Rockets 31(2):154–159. https ://doi.org/10.2514/3.26416 t.ngdc.noaa.gov/sem/poes/data. Accessed 17 April 2017 Welling DT (2010) The long-term effects of space weather on satellite opera- O’Brien TP (2009) SEAS-GEO: a spacecraft environmental anomalies expert tions. Ann Geophys 28(6):1361–1367. https ://doi.org/10.5194/angeo system for geosynchronous orbit. Space Weather 7(9):1. https ://doi. -28-1361-2010 org/10.1029/2009S W0004 73 Wing S, Gkioulidou M, Johnson JR, Newell PT, Wang CP (2013) Auroral particle Patil CG, Rajaram G, Prasad MYS (2008) Correlation of GSO satellite anomalies precipitation characterized by the substorm cycle. J Geophys Res with space weather data. Acta Astronaut 63(1–4):458–470. https ://doi. 118(3):1022–1039. https ://doi.org/10.1002/jgra.50160 org/10.1016/j.actaa stro.2007.12.050 Wu CC, Liou K, Lepping RP, Meng CI (2004) Identification of subtorms within Peck ED, Randall CE, Green JC, Rodriguez JV, Rodger CJ (2015) POES MEPED storms. J Atmos Terr Phys 66(2):125–132. https ://doi.org/10.1016/j. differential flux retrievals and electron channel contamination correction. jastp .2003.09.012 J Geophys Res 120(6):4596–4612. https ://doi.org/10.1002/2014J A0208 17
Earth, Planets and Space – Springer Journals
Published: May 29, 2018
It’s your single place to instantly
discover and read the research
that matters to you.
Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.
All for just $49/month
Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly
Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.
Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.
Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.
All the latest content is available, no embargo periods.
“Hi guys, I cannot tell you how much I love this resource. Incredible. I really believe you've hit the nail on the head with this site in regards to solving the research-purchase issue.”Daniel C.
“Whoa! It’s like Spotify but for academic articles.”@Phil_Robichaud
“I must say, @deepdyve is a fabulous solution to the independent researcher's problem of #access to #information.”@deepthiw
“My last article couldn't be possible without the platform @deepdyve that makes journal papers cheaper.”@JoseServera