Interactive real-time simulation of tidally locked exoplanet climate, hydrology, and biosignatures.
Developed for the International Digital Twins in Astrobiology Hackathon (HACK-4-SAGES 2026).
Version: 1.0.0 (Hackathon release)
Date: March 12, 2026
Authors: Jan Domański, Sandra Zaremba, Kamila Bąk
-
Scientific Framework
1.1 Atmospheric Dynamics
1.2 Hydrological Cycle
1.3 Stellar-Planet Interaction
1.4 Validation -
Technology Stack & XAI
3.1 Explainable AI Copilot
3.2 Numerical Stability
3.3 Data Integrity
Research Question:
How do stellar flares impact the stability of the terminator habitable zone on tidally locked exoplanets, and can methane-based biosignatures persist under such dynamic conditions?
Project Context:
This project is a direct implementation of the "Tidally Locked Planet" challenge proposed in the HACK-4-SAGES materials (Section 3.2, pp. 6-7). We extend the base requirements with a flare simulation module and an XAI Copilot.
What Makes This a Digital Twin:
ExoStress Twin functions as a true digital twin by maintaining a consistent virtual representation of TRAPPIST-1e that evolves in real time based on user inputs, while preserving mass and energy conservation laws. This allows researchers to experiment with "what-if" scenarios in a physically coherent framework.
The core of ExoStress Twin is a custom-built 2D Energy Balance Model (EBM). It bridges the gap between simple 1D models and complex 3D simulations, delivering immediate feedback for parametric exploration.
TRAPPIST-1e is tidally locked, creating permanent day and night hemispheres.
-
Super-rotation simulation:
We implement a numerical “equatorial jet” parameterization. Heat is actively transported eastward usingnp.rollshifts, recreating the asymmetric terminator and “hot-spot shift” described in modern astrophysical literature (Showman & Polvani, 2011). -
Thermal diffusion:
Atmospheric heat redistribution follows a 2D diffusion equation. Users can manipulate the diffusion coefficient ($D$ ) to simulate atmospheres ranging from thin, Mars-like envelopes to thick, Venus-style insulation.
Water is treated as a closed-loop system – a critical requirement for physical validity.
-
Bulk transport closure:
We mathematically guarantee 100% water mass conservation. Water evaporates on the dayside, is transported via parameterized winds, and precipitates on the nightside. -
Phase inventory tracking:
-
Liquid: Maintained within the “habitable zone”
($0,^\circ\mathrm{C} < T < 100,^\circ\mathrm{C}$ ) -
Ice (ice-albedo feedback):
When temperature drops below freezing and water is present, local albedo increases to 0.75. This triggers a cooling loop that can lead to a permanent “Snowball Earth” state. -
Gas:
High temperatures trigger evaporation. Increased water vapour acts as a greenhouse gas, allowing users to observe the onset of a runaway greenhouse effect.
-
M-dwarfs like TRAPPIST-1 are notoriously active. Our engine phenomenologically simulates:
-
Atmospheric stripping:
High-energy flare particles erode the atmosphere. This is modelled as a temporary reduction in greenhouse efficiency and heat diffusion, scaled by the planet’s relative gravity ($g_{\text{rel}}$ ). -
UV photolysis:
Stellar flares trigger rapid destruction of atmospheric methane (CH₄), demonstrating the transient nature of biosignatures on planets around active stars.
-
Terminator temperature:
The baseline Earth-like scenario (albedo = 0.3, diffusion = 0.05) produces a terminator temperature of ~15 °C, consistent with 1D energy balance estimates for TRAPPIST-1e (see Turbet et al. 2016). -
Snowball transition:
The ice-albedo runaway occurs at albedo > 0.6, matching theoretical expectations for global glaciation thresholds. -
Flare response:
Post-flare methane collapse follows first-order kinetics, qualitatively reproducing the photolytic destruction expected under enhanced UV flux.
ExoStress Twin uses biogenic methane as a primary proxy for biological activity.
-
The “sweet spot” heuristic:
CH₄ production is tied to surface habitability, specifically focusing on the terminator zone where liquid water is stable. -
Sub-glacial methanogenesis:
Inspired by Rugheimer et al. (2015), the model accounts for methanogenic activity under global ice sheets. This justifies why a planet in a “Snowball” state might still exhibit detectable biosignatures. -
Methane decay:
A first-order photolysis term removes CH₄, with a loss rate that increases during flare events.
We integrated a deterministic XAI Copilot to assist researchers.
Unlike generic LLMs, this tool:
- Reads live telemetry: habitability fractions, phase shifts, energy flux.
- Generates insights: provides human-readable explanations of complex climate feedbacks (e.g., why a planet entered a stable equilibrium or a drifting state).
- Guarantees stability: rule-based responses avoid the unpredictability of external APIs during live demonstrations.
The frontend is ready to swap in a true LLM endpoint with one line of code.
To keep the simulation robust under extreme user interactions (e.g., 15× flares), we implement the Courant-Friedrichs-Lewy (CFL) condition.
A dynamic time-stepping mechanism subdivides each main step to prevent numerical divergence in the diffusion calculations.
-
NASA Exoplanet Archive:
Core parameters (mass, radius, luminosity) are dynamically read from the archive (local cache included). -
Primary reference:
Baseline parameters follow Agol et al. (2021). -
Fallback:
If CSV files are missing, the code falls back to hard-coded values to ensure the simulation always runs.
The project was built in 48 hours for a hackathon.
Below are the conscious engineering and scientific trade-offs.
Assumption:
A 2D Energy Balance Model with advection parameterization is used instead of a full General Circulation Model (GCM).
Why:
GCMs require supercomputers and hours of simulation time. Our goal is real-time interactivity.
Consequences:
No vertical atmosphere profile, clouds, or realistic 3D circulation.
Assumption:
Water is transported from the dayside to the nightside using np.roll (a 180° longitudinal shift).
Why:
This guarantees mass conservation and mimics the net effect of atmospheric circulation.
Consequences:
No clouds or gradual transport.
Assumption:
CH₄ production scales with the area of the “sweet spot”.
Why:
A full biogeochemical model would require unknown extraterrestrial data.
Consequences:
Methane values (scaled to 2500 ppm) are abstract proxies.
Assumption:
Flares reduce diffusion and greenhouse coefficients.
Why:
The 2D model has no vertical axis.
Consequences:
Atmospheric loss is simulated indirectly.
Assumption:
grid_spacing = 2.0.
Why:
Maintains numerical stability.
Consequences:
Time step not physically calibrated.
Assumption:
Ice albedo fixed at 0.75.
Why:
Captures essential ice-albedo feedback.
Assumption:
Deterministic rules instead of API.
Why:
Guaranteed stability during live demos.
exostress-twin/
├── main.py
├── script.js
├── index.html
├── style.css
├── star_parameters/
├── planet_parameters/
└── requirements.txt
Dependencies
fastapi==0.115.0
uvicorn==0.30.1
numpy==1.26.0
pandas==2.1.0
python-multipart==0.0.6
git clone https://github.com/your-username/exostress-twin.git
cd exostress-twin
pip install -r requirements.txt
python main.pyor
uvicorn main:app --reload --host 0.0.0.0 --port 8000Open:
http://localhost:8000
Requirements:
- Python 3.9+
- Modern browser
Visuals: Vidu AI Voiceover: ElevenLabs AI
Data Sources
- NASA Exoplanet Archive
- Agol et al. (2021)
MIT License.
ExoStress Twin – Because habitability is not just about location, it’s about resilience.