Global Ionosphere–Thermosphere Model (GITM)

• GITM is a three-dimensional, nonhydrostatic model that simulates the coupled dynamics, chemistry, energetics, and electrodynamics of the upper atmosphere.
• It uses an altitude-based grid to resolve rapid vertical accelerations, storm-induced changes, and detailed interactions among neutral flows, plasma, and radiation.
• GITM’s flexible framework has been extended to Mars (M-GITM) and exoplanet studies, addressing ion escape, NO emissions, and launch-induced ionospheric effects.

The Global Ionosphere–Thermosphere Model (GITM) is a self-consistent, three-dimensional, coupled ionosphere–thermosphere general circulation model for the neutral and ionized upper atmosphere. In terrestrial applications it is a nonlinear, nonhydrostatic model that solves the full set of time-dependent governing equations from about 100 km to roughly 600 km altitude, with explicit coupling among neutral dynamics, plasma dynamics, composition, energy, and electrodynamics (Yiğit et al., 2015). In Mars-adapted form, M-GITM, the same general framework is modified for Martian parameters, ion–neutral chemistry, and radiative processes from the surface to about 250 km (Dong et al., 2018, Li et al., 27 Apr 2026). GITM is normally operated as a 3D global model on a spherical, altitude-based grid without the hydrostatic assumption, but it can also be run in a 1D column mode using the same physics and parameterization (Hanson et al., 29 Aug 2025).

1. Core architecture and dynamical formulation

GITM’s defining architectural feature is its nonhydrostatic treatment of the upper atmosphere. In the terrestrial storm configuration, the neutral momentum equation retains the vertical acceleration term rather than imposing hydrostatic balance, which allows vertical accelerations and acoustic–gravity waves to be represented and enables rapid adjustment to strong, transient forcing (Yiğit et al., 2015). A representative form used in the cited terrestrial work is

$$\rho \left(\frac{\partial \mathbf{v}}{\partial t} + \mathbf{v}\cdot\nabla \mathbf{v}\right) = -\nabla p + \rho \mathbf{g} - 2\rho \,\mathbf{\Omega}\times \mathbf{v} + \nabla\cdot \boldsymbol{\tau} + \sum_s \rho_s \nu_{sn}(\mathbf{v}_s-\mathbf{v}),$$

where $\rho$ is neutral mass density, $\mathbf{v}$ is the neutral wind, $p$ is pressure, $\mathbf{g}$ is gravity, $\mathbf{\Omega}$ is Earth’s rotation vector, $\boldsymbol{\tau}$ is viscous stress, and the final term represents ion–neutral drag (Yiğit et al., 2015).

The same modeling lineage is retained across planetary adaptations. M-GITM is described as a 3-D, time-dependent, nonhydrostatic general circulation model derived from the terrestrial GITM core and modified for Martian parameters, ion–neutral chemistry, and radiative processes (Li et al., 27 Apr 2026). In the Mars upper-atmosphere escape study, M-GITM preserves GITM’s nonhydrostatic, three-dimensional, time-dependent general circulation framework while replacing Earth-specific physics with Mars parameters, chemistry, and radiative processes (Dong et al., 2018). In the exoplanet study, GITM is characterized as a self-consistent, multi-fluid code that solves the coupled dynamics, energetics, chemistry, and electrodynamics of the upper atmosphere, with both 3D global and 1D column configurations implemented from the same physics base (Hanson et al., 29 Aug 2025).

The model grid is altitude-based rather than pressure-based. Representative terrestrial storm runs used a latitude–longitude grid of $2.5^\circ \times 5^\circ$, 54 vertical levels with spacing of one-third of the local scale height, and a time step of 2–4 s over roughly 100–600 km (Yiğit et al., 2015). M-GITM production runs described for Mars employed $5^\circ \times 5^\circ$ longitude–latitude grids and 2.5 km vertical resolution, corresponding to about 0.25 scale height (Dong et al., 2018). In the Earth-like exoplanet study, the 3D configuration used a full-globe $80\times48$ horizontal mesh, a lower boundary at 100 km, and vertical cell thickness set to a fraction of the local scale height, specifically 0.3 for EUV factors $\rho$0 and 0.2 for factors of 20 and 50 (Hanson et al., 29 Aug 2025).

This architecture positions GITM between hydrostatic thermosphere models and plasma-only frameworks. A plausible implication is that its explicit treatment of neutral momentum and altitude-based expansion makes it especially suited to problems in which rapid forcing, vertical coupling, and composition–dynamics feedbacks are central.

2. Electrodynamics, chemistry, and energy closure

A central feature of GITM is that electrodynamics is treated self-consistently. In the terrestrial storm study, this includes the low-latitude dynamo electric field, while high-latitude forcing is supplied by empirical magnetospheric electric potentials and particle precipitation patterns that ingest storm inputs (Yiğit et al., 2015). The specific high-latitude drivers in that configuration are the IMF-dependent Weimer (1996) electric potential model, using IMF $\rho$1, $\rho$2, and dipole tilt, and the Fuller-Rowell and Evans (1987) precipitation specification using the hemispheric power index (Yiğit et al., 2015). In the exoplanet application, GITM likewise computes ionospheric conductivities and resulting Joule heating self-consistently from local plasma-neutral properties and the prescribed electric field (Hanson et al., 29 Aug 2025).

The neutral energy equation in the terrestrial formulation includes compressional work, radiative terms, conduction, and multiple heating sources:

$\rho$3

with Joule heating $\rho$4, auroral particle heating $\rho$5, solar EUV heating $\rho$6, radiative cooling $\rho$7, and thermal conduction $\rho$8 (Yiğit et al., 2015). Joule heating in the neutral frame is given by

$\rho$9

and the Hall term does not contribute to $\mathbf{v}$0 because it is orthogonal to $\mathbf{v}$1 (Yiğit et al., 2015). The exoplanet study gives the local Pedersen and Hall conductivities as

$\mathbf{v}$2

with the current density written as $\mathbf{v}$3 (Hanson et al., 29 Aug 2025).

Chemistry is similarly integral rather than auxiliary. The nitric oxide study updated the lower-thermospheric NO scheme in GITM by revising rates and branching ratios and adding an explicit $\mathbf{v}$4 metastable channel,

$\mathbf{v}$5

with $\mathbf{v}$6 (Lin et al., 2018). In that formulation, the local NO 5.3 $\mathbf{v}$7m volume emission rate is parameterized as

$\mathbf{v}$8

and global integrated NO emission agrees within 20% of the SABER-derived empirical proxy over most simulated conditions (Lin et al., 2018). This establishes GITM not only as a circulation model but also as a thermospheric energy-budget model in which photochemistry, ion chemistry, and infrared cooling are dynamically consequential.

3. Geomagnetic-storm dynamics in the terrestrial thermosphere–ionosphere

A canonical terrestrial application of GITM is the simulation of the 5–6 August 2011 major geomagnetic storm. In that study, storm-time interplanetary inputs were taken from ACE observations and ingested as 30-minute smoothed time series of IMF components, solar wind density, and solar wind speed to emphasize large-scale variations; a steady benchmark run used August 1–5 averages as a quiet-control state (Yiğit et al., 2015). Much of the analysis focused on 400 km altitude between 5 August 0600 UT and 7 August 0000 UT (Yiğit et al., 2015).

The simulated ionospheric response was immediate. Within two hours of storm onset, high-latitude mean horizontal ion flows at 400 km increased from about 240 m s$\mathbf{v}$9 to about 500 m s$p$0 in the Southern Hemisphere and to about 560 m s$p$1 in the Northern Hemisphere. During the main phase near 2200 UT, the high-latitude mean reached about 630 m s$p$2 in the Southern Hemisphere and more than 700 m s$p$3 in the Northern Hemisphere, corresponding to about 150–180% enhancement overall and up to about 200% in the Northern Hemisphere. Under sustained southward IMF, ion speeds exceeded 1600 m s$p$4 in polar cap and auroral regions, and elevated flows of about 1100 m s$p$5 persisted into the recovery phase (Yiğit et al., 2015).

The neutral atmosphere responded more slowly because of neutral inertia, but the thermal and dynamical effects were large. The global mean neutral temperature at 400 km rose by about 140 K, from about 820 K to about 960 K, an increase of about 15%. The winter Southern Hemisphere showed a larger fractional response than the summer Northern Hemisphere: a peak high-latitude mean temperature increase of about 40% or about 300 K in the Southern Hemisphere, versus about 20% or about 170 K in the Northern Hemisphere, both peaking near 0500 UT on 6 August, roughly 11 hours after onset. The global mean Joule heating increased by more than a factor of three, from about 300 K day$p$6 to over 1000 K day$p$7, and local instantaneous high-latitude Joule heating could exceed the mean solar heat input of about 1500 K day$p$8 (Yiğit et al., 2015).

Neutral circulation intensified correspondingly. The global mean horizontal neutral wind magnitude increased by more than 50%, from about 130 to about 200 m s$p$9. Polar maps showed regions of neutral flow exceeding 700 m s$\mathbf{g}$0 during the main phase, with enhanced flows persisting into recovery (Yiğit et al., 2015). Hemispheric asymmetry was quantified by the conjugate difference $\mathbf{g}$1 between Southern and Northern Hemisphere horizontal winds. In the quiet run, $\mathbf{g}$2 was typically within $\mathbf{g}$3 m s$\mathbf{g}$4, whereas during the storm $\mathbf{g}$5 exceeded $\mathbf{g}$6 m s$\mathbf{g}$7, with peak differences increasing by factors of 4–5 relative to pre-storm values (Yiğit et al., 2015).

The simulations identified several mechanisms acting together: seasonal background state, geomagnetic–geographic pole offsets and magnetic-field asymmetries, conductance and ion drag differences, nonlinear advection, and IMF $\mathbf{g}$8 effects (Yiğit et al., 2015). In particular, the horizontal advective forcing diagnostic $\mathbf{g}$9 was enhanced during the storm and displayed pronounced hemispheric asymmetry, especially in the main and recovery phases, supporting a causal role for advection in sustaining and amplifying circulation differences after the initial ion-drag energization (Yiğit et al., 2015). This is one of the clearest demonstrations in the cited literature that GITM’s nonhydrostatic, tightly coupled formulation is not merely a numerical detail but materially affects interpretation of storm-time hemispheric structure.

4. Chemistry-driven and event-specific terrestrial extensions

GITM has also been extended to process-specific studies in which chemistry and transport dominate the observable response. In the nitric oxide emission study, three chemistry configurations were compared: the old chemistry, a new chemistry with updated rate and branching-ratio data, and the new chemistry plus explicit $\mathbf{\Omega}$0 production of NO. Switching from the old to the new scheme increased global integrated 5.3 $\mathbf{\Omega}$1m emission by up to a factor of three, while the inclusion of $\mathbf{\Omega}$2 contributed an additional 5–25% depending on solar and geomagnetic activity. Under a moderate geomagnetic increase, GITM simulated localized enhancement of about 70% in NO column density and a factor of three in column emission, and the global integrated emission agreed within $\mathbf{\Omega}$3 of the SABER-derived Thermosphere Climate Index over most simulated conditions (Lin et al., 2018). Because NO is described there as a major thermospheric “thermostat,” these updates directly affect modeled heating–cooling closure rather than only minor-species realism (Lin et al., 2018).

The rocket-launch study extended GITM in a different direction by adding exhaust chemistry, diffusion, and moving source terms to simulate large-scale ionospheric depletions generated by Falcon 9 second-stage exhaust. In that work, the dominant F-region depletion mechanism was the rapid conversion of long-lived $\mathbf{\Omega}$4 into molecular ions through charge exchange with $\mathbf{\Omega}$5, $\mathbf{\Omega}$6, and $\mathbf{\Omega}$7, followed by fast dissociative recombination. Under the instantaneous-recombination approximation, the net electron sink was written as

$\mathbf{\Omega}$8

with the added charge-exchange rates and recombination rates taken from the sources specified in the study (Bowden et al., 2020).

That implementation used GITM in regional mode over $\mathbf{\Omega}$9 N–$\boldsymbol{\tau}$0 N and $\boldsymbol{\tau}$1 W–$\boldsymbol{\tau}$2 W with approximate horizontal resolution of $\boldsymbol{\tau}$3, initialized by IRI and MSIS, with winds consistent with HWM07 in the background climatology (Bowden et al., 2020). For Jason-3, the simulation produced a maximum TEC depletion of 5.26 TECU at 19:50 UT and a maximum depletion area of $\boldsymbol{\tau}$4 km$\boldsymbol{\tau}$5. For FORMOSAT-5, the corresponding values were 11.5 TECU at 21:40 UT and $\boldsymbol{\tau}$6 km$\boldsymbol{\tau}$7 at 23:20 UT, with modeled electron removal above 200 km about 6.8 times larger than for Jason-3 (Bowden et al., 2020). The simulated FORMOSAT-5 drift was about 72 m s$\boldsymbol{\tau}$8 toward $\boldsymbol{\tau}$9, compared with about 66 m s$2.5^\circ \times 5^\circ$0 toward $2.5^\circ \times 5^\circ$1 from Swarm observations, and sensitivity tests using different exhaust compositions changed peak electron depletion by less than 6% (Bowden et al., 2020).

These studies show that GITM can absorb targeted physical extensions without abandoning its general-circulation character. The common pattern is that chemistry sets source and sink rates, while advection, diffusion, and background circulation govern morphology, timing, and persistence.

5. Mars implementation: M-GITM, ion escape, and thermospheric polar warming

M-GITM is the Mars-adapted implementation of GITM. It retains the nonhydrostatic, three-dimensional, time-dependent general circulation framework but replaces Earth-specific physics with Mars parameters, chemistry, and radiative processes to simulate the thermal, compositional, and dynamical structure of the Martian atmosphere from the ground to the exobase, stated as 0–250 km in one study and from the surface to about 250 km in another (Dong et al., 2018, Li et al., 27 Apr 2026). In the escape study, M-GITM solved for neutral and ion densities around the globe, including neutral species such as $2.5^\circ \times 5^\circ$2, CO, O, $2.5^\circ \times 5^\circ$3, $2.5^\circ \times 5^\circ$4, Ar, and He, and photochemical ions $2.5^\circ \times 5^\circ$5, $2.5^\circ \times 5^\circ$6, $2.5^\circ \times 5^\circ$7, $2.5^\circ \times 5^\circ$8, and $2.5^\circ \times 5^\circ$9 (Dong et al., 2018).

In the Mars solar-wind interaction study, M-GITM provided the cold thermospheric background to a four-species multifluid MHD model, supplying neutral densities, neutral temperatures, neutral winds, and top-of-atmosphere photoionization frequencies under APHMOD conditions with $5^\circ \times 5^\circ$0 and $5^\circ \times 5^\circ$1 (Dong et al., 2018). A major result was that fully 3D M-GITM inputs were required to reproduce the observed ionospheric structure along MAVEN tracks, even though total ion escape rates were similar between 1D and 3D atmospheres. The 3D thermosphere captured strong diurnal and latitudinal variability, including dayside enhancement of $5^\circ \times 5^\circ$2, a nightside bulge of atomic O produced by day-to-night transport above about 130 km, and enhanced exobase temperatures near mid-afternoon and the evening terminator (Dong et al., 2018). With 1D thermosphere inputs, the MF-MHD model slightly overestimated molecular ion densities and underestimated $5^\circ \times 5^\circ$3; with 3D M-GITM inputs, the model better matched MAVEN LPW electron densities and NGIMS ion densities (Dong et al., 2018).

The same study used M-GITM together with the AMPS hot oxygen exosphere to examine ion escape. The outflow-only ion escape rate at 6 $5^\circ \times 5^\circ$4 was 2.30, 2.28, 2.38, and $5^\circ \times 5^\circ$5 s$5^\circ \times 5^\circ$6 for the four cases combining 1D or 3D thermosphere and exosphere inputs, including a case with no hot O corona (Dong et al., 2018). The key trend was compositional rather than total: 3D thermosphere increased $5^\circ \times 5^\circ$7 escape and decreased molecular-ion escape relative to 1D, while the 3D hot oxygen corona reduced $5^\circ \times 5^\circ$8 and $5^\circ \times 5^\circ$9 escape through mass loading of the solar wind, leaving $80\times48$0 escape similar. Under the relatively weak Solar Cycle 24 APHMOD conditions, cold thermospheric oxygen was the primary neutral source for $80\times48$1 escape, and the molecular-to-atomic escape ratio $80\times48$2 fell from 2.17 with no hot O to 1.67 with 3D hot O (Dong et al., 2018).

A different use of M-GITM addressed Mars Aphelion Thermospheric Polar Warming. In that study, M-GITM incorporated solar fluxes from MAVEN/EUVM through FISM-M, dust radiative effects based on Mars Climate Sounder Version 6.1 dust distributions, and a spectral, nonlinear non-orographic gravity-wave parameterization that launches at the top of the mean planetary boundary layer (Li et al., 27 Apr 2026). The simulations showed that the standard M-GITM configuration underestimates the observed dawn polar warming: MAVEN EUVM climatologies indicated a latitudinal temperature difference $80\times48$3 of about 40–60 K between roughly $80\times48$4S and $80\times48$5S at 150–180 km, whereas the standard M-GITM run with solar minimum forcing and gravity waves on produced only about 10–15 K at 150 km (Li et al., 27 Apr 2026). Solar maximum forcing increased the modeled $80\times48$6 to about 20–30 K, and turning gravity waves off increased it further to about 30–35 K, but the no-gravity-wave case also produced unrealistically strong winds approaching about 250 m s$80\times48$7 rather than about 100 m s$80\times48$8 with gravity waves on (Li et al., 27 Apr 2026).

The dynamical diagnosis is especially informative. Vertical dynamical heating terms were near zero, while horizontal adiabatic heating and horizontal hydrodynamic heating controlled the polar warming magnitude. Local dust storms had little impact, and increasing horizontal resolution from $80\times48$9 to $\rho$00 left the polar $\rho$01 essentially unchanged (Li et al., 27 Apr 2026). This suggests that, within the tested M-GITM configurations, the main uncertainty lies not in dust forcing or nominal horizontal resolution but in the representation of subgrid-scale gravity waves and the associated wave–mean flow interaction.

6. Exoplanet applications, dimensionality, and model limits

GITM has also been used outside the Solar System as a model for Earth-like exoplanet upper atmospheres under enhanced stellar EUV forcing. In that study, the model was run with EUV flux factors from 1 to 50 times the mean solar value at Earth, using an Earth-like magnetic field, a constant solar wind speed of 400 km s$\rho$02, IMF $\rho$03 nT, no auroral precipitation, neutral heating efficiency 0.05, photoelectron heating efficiency 0.0, and eddy diffusion coefficient 50 (Hanson et al., 29 Aug 2025). The baseline spectrum was specified by $\rho$04 sfu and converted to a 59-bin EUV spectrum by SERF and EUVAC (Hanson et al., 29 Aug 2025).

The principal outcome was a transition from an Earth-like upper atmosphere to a rapidly escaping one. The study reported a threshold EUV flux level for a stable atmosphere around a factor of 10 times the baseline, corresponding to about 0.316 AU for a planet orbiting the Sun. At EUV levels above this threshold, the atmosphere was described as rapidly escaping (Hanson et al., 29 Aug 2025). The Jeans parameter used for diagnosis was

$\rho$05

with $\rho$06 indicating hydrodynamic escape and $\rho$07 indicating gravitationally bound gas (Hanson et al., 29 Aug 2025). At 20$\rho$08 EUV, the model showed cyclic growth and depletion because a constant lower-bound neutral supply intermittently refilled the expanding upper atmosphere, with hydrodynamic escape above about 5000 km. At 50$\rho$09 EUV, continuous depletion occurred, and $\rho$10 above about 2000 km nearly everywhere, indicating pervasive hydrodynamic blow-off (Hanson et al., 29 Aug 2025).

The conductivity and escape diagnostics emphasize why dimensionality matters. At 1000 km, the Pedersen conductivity profile was about $\rho$11 S m$\rho$12 for 1$\rho$13, about $\rho$14 S m$\rho$15 for 5$\rho$16, and greater than $\rho$17 S m$\rho$18 for 10$\rho$19 EUV, with the altitude where $\rho$20 fell below $\rho$21–$\rho$22 S m$\rho$23 moving from about 1000 km to about 2500 km and then above 4000 km (Hanson et al., 29 Aug 2025). The 3D global upward $\rho$24 flux was about $\rho$25 s$\rho$26 at baseline and about $\rho$27 s$\rho$28 at 10$\rho$29, while 1D column fluxes differed from the 3D global value by 1–2 orders of magnitude depending on location and EUV level (Hanson et al., 29 Aug 2025). At 9$\rho$30, some 1D runs failed to converge, behaving similarly to the extreme-EUV cases because all dynamics were forced into a single vertical column without lateral relief (Hanson et al., 29 Aug 2025).

This limitation of 1D is echoed in the Mars studies. In the Mars escape problem, 1D and 3D atmospheres gave similar total escape rates, but only 3D thermosphere inputs reproduced MAVEN ionospheric structure (Dong et al., 2018). In the exoplanet problem, 1D columns misestimated densities and fluxes by orders of magnitude at high EUV (Hanson et al., 29 Aug 2025). A common misconception is therefore that similar integrated escape rates imply dynamical sufficiency of spherically symmetric backgrounds; the cited studies show that this is not generally true for morphology, composition, or local energetics.

The main limitations described across the cited literature are empirical forcing, incomplete coupling, and parameterization uncertainty. In the August 2011 storm study, high-latitude forcing was specified by climatological Weimer and Fuller-Rowell–Evans patterns that do not include realistic short-time variability, and IMF and solar-wind inputs were smoothed at 30-minute cadence (Yiğit et al., 2015). In the Mars MF-MHD coupling, the thermosphere and exosphere were one-way coupled into the plasma model, so plasma feedback on neutrals was neglected; hot oxygen distributions and escape remained uncertain despite improved collision physics (Dong et al., 2018). In the exoplanet runs, the host-star spectrum was an Earth-based SERF/EUVAC spectrum scaled uniformly, explicit polar-wind acceleration physics was not included, and escape above the top boundary was inferred from fluxes and the Jeans parameter rather than by tracking particle fates (Hanson et al., 29 Aug 2025). In the rocket-launch application, the plume was approximated as diffusive expansion from rest, condensation was neglected, and regional boundaries limited long-time fidelity (Bowden et al., 2020). In the Mars TPW study, the gravity-wave source spectrum or launch level may be too strong or too uniformly applied at high latitudes, and polar filters on latitude–longitude grids may remove high-frequency variability important for the warming signal (Li et al., 27 Apr 2026).

Taken together, these studies define GITM as a model family rather than a single fixed configuration: a nonhydrostatic, multi-fluid, altitude-coordinate framework that can be specialized to terrestrial storms, thermospheric cooling, launch-induced ionospheric holes, Mars thermosphere–ionosphere coupling, solar-wind interaction and escape, and Earth-like exoplanet blow-off. The recurring methodological conclusion is that the model’s value lies in retaining coupled dynamics, energetics, chemistry, and electrodynamics in three dimensions. Where those couplings are reduced to empirical closure, one-way forcing, or 1D symmetry, the resulting solutions may still reproduce some integrated quantities, but they systematically lose causal fidelity in structure, timing, and transport.