MMA processing to date Currently generating >5000 90-minute models in two year-files: SW_OPER_MMA_SHA_2F_20131126T000000_20140101T000000_0101 SW_OPER_MMA_SHA_2F_20140101T000000_2014****T******_0101 Input data & models: Level1b Swarm ABC 0301 & 0302. CHAOS4+V4 for core field-model (with predictive fields to 2015.0). AUX_LIT for lithospheric field. CM5 ionospheric model (with F10.7 currently downloaded from NGDC) Processing: Select mid-/low-latitude Swarm data. Remove estimate of core-, lithospheric, and ionospheric-fields. Fit degree 1 internal and external model per orbit in GEO frame: Order 0 (dominant) coefficient by inversion Order 1 coefficients by mean-over-orbit Separate using 1-yr Q-response filter based on 1-D conductivity (Utada et al, 2003)
Comparison with Est & Ist (from Dst) Est & Ist (mean=0): Derived by NGDC Using Nov 2013 – Nov 2014 Using version with mean Ist = 0 Fairly good agreement with MMA: Compare in MAG frame Est: Correlation coeff = 0.84 RMS diff = 11.2 nT Ist: Correlation coeff = 0.90 RMS diff = 2.2 nT But some step-like changes (highlighted) Not clear yet why this is happening: diagnostics of MMA don’t indicate an obvious problem probably Dst baselines Year 2013.9 2014.9 MMA - Est Est (nT) -MMA q10 (nT)
Comparison with VMD – order 0 coeffs VMD index (Thomson & Lesur, 2007): Minute-mean, mid-low latitude obs data Long-term trends subtracted 20 minute cadence. Internal & ext. deg 1 coeffs in GEO frame. Currently computed to Aug 2014 MMA-VMD agreement good for dominant order 0 coeffs: Compare in GEO frame q10 coeff: Correlation = 0.94, RMS = 13.9 nT (mostly offset) g10 coeff: Correlation = 0.91, RMS = 1.5 nT Any step changes in difference with VMD not so obvious, compared to difference with Est – better baseline control in VMD Year2013.92014.9 2014.7 Steps? MMA - Est MMA - VMD VMD q10 MMA q10 2014.7 2013.9
Comparison with VMD – order 1 coeffs Much poorer agreement between order 1 coeffs: CoeffCorrelation MMA-VMDRMS (nT) q11/s110.4/0.64.1/4.3 g11/h110.4/0.61.4/1.3 Comparison between MMA and TDS-1 during testing was much better. (corr coeff >~ 0.95) Scale of differences vary periodically: Local-time dependence? Noon-midnight orbits worst? MMA uses all LTs + ionospheric model: TDS-1 testing => better than night-only (Hamilton, 2012) But ionospheric model now used for real data RMS differences between model & input magnetospheric field residuals follow same pattern: Trying to fit > degree 1 fields? E.g. Ionospheric, magnetospheric higher degree fields RMS model-data Swarm A & B LTs MMA - VMD VMD q11 (nT) MMA q11 (nT) Year2013.9 2014.7 2014.9
Fields from degrees > 1 MMA fits degree 1 fields only For order 1 terms, degrees >1 may dominate (e.g. Lühr & Maus, 2010) Plot below shows magnetospheric residual data (black) and degree 1 fit (red): Orbit over night-side half of orbit on a quiet day from all 3 Swarm Clearly more structure than degree 1 field can fit Structure may be local-time dependent ~ 45 minutes
Satellite separation Any evidence of Swarm A and B seeing asymmetries in field due to LT dependence & FACs (e.g. Balasis et al., 2004)? Swarm A & B orbits now ~15 deg separated Re-ran MMA but using either Swarm A or B only Look for changes in differences as orbits diverge: Caused by differences in fields sampled Apparent overall growth in differences Could be due to LT-dependent field structure Results should be clearer as satellites further diverge and main field model updated Year 2013.9 2014.9 s11: MMA A – MMA B q10: MMA A – MMA B Swarm A & B LTs
MMA Summary Robust daily automatic operation Dominant order 0 terms seem fairly robust (good agreement with VMD and Dst) Order 1 terms do not look very robust – possible day-side ionospheric and/or magnetospheric contamination? No clear longitudinal structure visible in Swarm A vs Swarm B models but still early in mission References Utada, H., T. Koyama, H. Shimizu, and A. Chave, A semi-global reference model for electrical conductivity in the mid-mantle beneath the north Pacific region, Geophys. Res. Lett., 30(4), 1194, 2003. Hamilton, B., Rapid modelling of the large-scale magnetospheric field from Swarm satellite data, Earth Planets Space, 65, 1295–1308, 2013. Luhr, H. and S. Maus, Solar cycle dependence of quiet-time magnetospheric currents and a model of their near-Earth magnetic fields, Earth Planets Space, 62, 843–848, 2010. Balasis, G., G. Egbert, and S. Maus, Local time effects in satellite estimates of electromagnetic induction transfer functions, Geophys. Res. Lett., 31, 16, L16610, 2004.