← Back to Research

Lithosphere Structure in Tibet

Timeline May 2026-Present
Supervisor Dr. Ajay Kumar
Stack
Python ASPECT

Objective

The lithospheric structure underneath Tibet remains a major open question due to the highly unusual plate collision dynamics in the region. Because of this complexity, there are currently several contradicting tomographic models, each supported by different geophysical reasoning. The goal of this work is to evaluate these competing models and establish a reliable benchmark structure for the Tibetan lithosphere.

Pipeline Setup

To test these models, we take the leading tomographic datasets and pass them through an automated preprocessing pipeline. First, the raw velocity information is converted into temperature and density. The pipeline then clips this data to our specific region of interest, converts it into the appropriate units, and formats it so it can be fed directly into the ASPECT Stokes solver.

Simulation & Evaluation

Once the data is formatted, we run an instantaneous solve within ASPECT using appropriate parameters to see how the system reaches equilibrium. We then extract output variables (specifically topography, strain rate, stress, and velocities) and compare them against observed real-world numbers to determine which structural model is the most viable. To handle the evaluation phase, I built a custom CLI tool (post-process) to rapidly plot and analyze the generated data.

Limitations

Modeling a region this complex requires several compromises. For instance, the temperature profile currently assumes a linear gradient for the top 80km, while the deeper profile is calculated via Gibbs free energy minimization. We also rely on CRUST 1.0 for crustal density and a DMM inversion for the mantle.

There are also technical hurdles in the simulation itself: we use a specific workaround to successfully load the density data into ASPECT, and the model does not yet account for absolute velocities at the boundaries. Despite these assumptions, the pipeline provides a functional foundation to iteratively improve the models.

Code

  • aspect-input-data-scripts
  • post-process (CLI Interface)