Conditional Diffusion Models for Microseismic Localization
Generative AI for a classic geophysical inverse problem: locating microseismic events directly from noisy waveform gathers, without arrival-time picking, and with uncertainty estimates from repeated sampling.
- Problem
- Locating microseismic events from noisy recordings is an ill-posed inverse problem. Travel-time methods depend on accurate arrival picks and velocity models, and deterministic deep learning returns a single answer with no measure of reliability.
- Approach
- A conditional diffusion model that generates the event location as two Gaussian heatmaps (X–Z and Y–Z), conditioned on features extracted from the recorded three-component waveforms.
- My role
- Formulated event localization as conditional heatmap generation, combining waveform preprocessing, learned features and quantitative baseline evaluation, and investigated noise sensitivity and repeated stochastic sampling to characterize uncertainty.
- Results
- Mean absolute error below 22 m and R² above 0.81 on each axis across 1,774 semi-synthetic test samples with recorded field noise. Lower MAE than a CNN baseline on every axis, while the CNN had higher R² and lower error standard deviation. A real field event was located within 11.5 m of the catalog solution.
- Conditional diffusion models
- PyTorch
- Convolutional neural networks
- Signal processing

