vplanet_inference

Python tools for statistical inference with VPLanet — a planetary system evolution simulator.

vplanet_inference provides a high-level Python interface for:

• Running VPLanet forward models with unit-aware parameter substitution

• Performing parameter sweeps and sensitivity analysis

• Running Bayesian inference (MCMC / nested sampling) constrained by observational data

Getting Started

• Installation
  • Requirements
  • Setting up a conda environment
  • Installing vplanet_inference
  • Optional dependency: alabi
  • Verifying the installation
• Quickstart
  • Running a forward model
  • Tracking time evolution
  • Two-body tidal evolution example

Tutorials

• Tutorial Overview
• Setting Up and Running VPLanet Models
  • 1. VPLanet Input Files: The Manual Way
    • 1.1 The Primary Input File (vpl.in)
    • 1.2 Body Files (star.in)
  • 2. Running VPLanet from the Command Line
  • 3. Understanding VPLanet Output Files
    • 3.1 The Log File
    • 3.2 The Forward File
  • 4. Running the Same Model Through vplanet_inference
    • 4.1 Declaring Input Parameters (inparams)
    • 4.2 Declaring Output Parameters (outparams)
    • 4.3 Running a Single Model
    • 4.4 Varying Input Parameters
    • 4.5 Tracking Time Evolution
  • Summary
• Global Sensitivity Analysis with vplanet_inference
  • What we cover
    • 1. Sobol Sensitivity Indices
  • Workflow
    • 2. Setting Up the VPLanet Model
  • 2.1 Declaring Input Parameters
  • 2.2 Initializing VplanetModel
  • 2.3 Defining a Model Wrapper
    • 3. Running the Sensitivity Analysis
  • 3.1 Generating the Saltelli Sample
  • 3.2 Evaluating the Model
  • 3.3 Computing Sobol Indices
    • 4. Visualizing Results
  • Interpreting the Results
    • 5. Bar Chart of Sensitivity by Parameter
    • 6. Using AnalyzeVplanetModel (YAML Workflow)
    • Summary
• MCMC with VPLanet and alabi
  • Problem: inferring TRAPPIST-1 stellar properties
    • 1. Configure the VPLanet Forward Model
  • Sanity check: single run at known TRAPPIST-1 values
    • 2. Observational Data and Priors
    • 3. Custom Likelihood and Prior Functions
  • 3.1 Custom Log-Likelihood
  • 3.2 Log-Prior
  • 3.3 Log-Posterior
    • 4. Option A — Direct MCMC with dynesty
    • 5. Option B — Direct MCMC with emcee
    • 6. Option C — Surrogate MCMC with alabi
  • Installation
  • Workflow
  • 6.1 Setting Up the Prior Sampler and Transform
  • 6.2 Initialising the Surrogate Model
  • 6.3 Active Learning
  • 6.4 MCMC on the Surrogate
  • 6.5 Reloading a Saved Model
    • Summary

Case Studies

• Case Studies Overview
• Maximum Likelihood Estimation for Earth Interior Evolution
  • 1. Earth Model Setup
    • 1.1 VPLanet Input Files
    • 1.2 Input Parameters
    • 1.3 Output Parameters and Observational Constraints
    • 1.4 Initializing VplanetModel
  • 2. The Log-Likelihood Function
  • 3. Fiducial Model
    • 3.1 Visualizing the Thermal Evolution at Fiducial Parameters
  • 4. Single Optimization Run
  • 5. Multi-Start MLE
  • 6. Analyzing the MLE Results
    • 6.1 Best-Fit Parameters
    • 6.2 Model Outputs vs. Observations
    • 6.3 Analyze Local Sensitivity Around the MLE Solution
    • 6.4 Compute Fischer Information Matrix
• Custom Rotation-Activity-XUV Model
  • 1. The Johnstone XUV Model
  • 2. StellarEvolutionModel Setup
  • 3. Single Model Evaluation
  • 4. Parameter Sweeps
  • 5. Configuring run_alabi.py for a New Star
  • 6. Testing Different Constraint Combinations
  • 7. Prior and Likelihood Setup
  • 8. Running alabi Surrogate MCMC
  • 9. Reloading and Extending a Run
  • Summary

API Reference

• API Reference
  • vplanet_inference.model
  • vplanet_inference.analysis
  • vplanet_inference.parameters
  • vplanet_inference.fischer
  • vplanet_inference.info_vis