GNSS Slow-Slip Detection With Bayesian Modeling

ISEF Category: Earth and Environmental Sciences

Subcategory: Geosciences  ·  Difficulty: Advanced  ·  Setup: Home Setup  ·  Time: 1 to 2 Months

The Hook

Some faults move without earthquakes you can feel. They creep in tiny steps over days or months, like a zipper sliding one tooth at a time. You can find those hidden shifts with real GNSS data and a smart change-point model. That gives you a serious geoscience project without digging up a single rock.

What Is It?

GNSS stands for Global Navigation Satellite System. These are the satellite data streams that let scientists track how the ground moves by millimeters. If a fault changes behavior, the station positions can start to drift, bend, or jump. A Bayesian change-point model looks for the moment when the data stop following the old pattern and start following a new one.

Think of it like watching a slow leak in a tire. At first, the pressure looks normal. Then the trend shifts, but only a little, and you need a careful method to spot when that shift began. In this project, you use publicly available Nevada Geodetic Lab GNSS time series to test whether a fault zone shows slow-slip or post-seismic deformation signals. Slow-slip events happen when a fault releases stress quietly. Post-seismic deformation happens after a big quake, when the crust keeps adjusting. Both can leave patterns in the data that a change-point model can detect.

Why This Is a Good Topic

This is a strong science fair topic because it uses public data, a clear signal detection problem, and a method you can test in a repeatable way. You can compare different stations, time windows, or model settings and ask which ones detect fault motion best. The project connects to earthquake science, hazard monitoring, and geodesy, which makes the real-world value easy to explain. You can also learn data cleaning, model fitting, uncertainty, and how scientists decide whether a pattern is real or just noise.

Research Questions

• How does the choice of GNSS station affect the ability to detect a slow-slip signal??
• What is the effect of using daily versus weekly sampling on change-point detection in fault motion data??
• Does a Bayesian change-point model detect post-seismic deformation earlier than a simple linear trend model??
• To what extent do nearby reference stations reduce false change-point detections in GNSS time series??
• Which fault zone time series shows the strongest evidence of a shift in ground motion??
• How does removing seasonal signals change the number and timing of detected change points??

Basic Materials

• Computer with internet access and enough storage for time series files.
• Spreadsheet software or Python installed.
• Free GNSS time series from Nevada Geodetic Lab.
• Fault location and earthquake context from USGS earthquake catalog.
• Notebook for tracking station choices, model settings, and results.
• Basic graphing tool for plotting displacement over time.

Advanced Materials

• Computer with Python, R, or both.
• GNSS time series from Nevada Geodetic Lab.
• USGS earthquake catalog data for event timing and magnitude.
• Station metadata and coordinate files.
• Change-point modeling package in Python or R.
• ImageJ or another plotting tool if you need to inspect exported figures.
• Access to a reference paper on GNSS noise models and seasonal correction.

Software & Tools

• Python: Fits Bayesian change-point models, cleans time series, and makes publication-style plots.
• R: Handles time series analysis and model comparison if you prefer that workflow.
• Jupyter Notebook: Keeps code, notes, and figures in one place while you test ideas.
• pandas: Organizes GNSS station data and helps you filter, merge, and compare time windows.
• matplotlib: Plots displacement trends, change points, and model output clearly.

Experiment Steps

1. Choose one fault zone and a small set of GNSS stations that span the area you want to test.
2. Decide which motion signal you will measure first, such as east, north, or vertical displacement.
3. Build a clean baseline by removing obvious gaps, outliers, and station records that are too noisy.
4. Select one Bayesian change-point model and define the comparison models you will use as a check.
5. Plan a way to test whether detected shifts match known earthquake timing, seasonal motion, or quiet-slip intervals.
6. Compare results across stations or time windows and decide how you will judge a true signal versus noise.

Common Pitfalls

• Using stations with too many gaps, which makes the model treat missing data like real motion.
• Mixing stations from different tectonic settings, which hides the fault signal you want to measure.
• Forgetting to remove seasonal effects, which can look like a fake slow-slip event.
• Treating one sharp jump as slow slip when the shift actually comes from an earthquake or a data glitch.
• Comparing model outputs without a clear ground truth, which makes it hard to defend your conclusion.

What Makes This Competitive

A strong version of this project does more than spot a change in one station. You could compare several faults, several station types, or several model assumptions and show which setup produces the most reliable detections. You could also test false alarm rates, uncertainty bounds, or how well the model matches known earthquake histories. That kind of careful analysis turns a data exercise into a real research question.

Project Variations

• Compare slow-slip detection across two fault zones with different earthquake histories.
• Test whether vertical GNSS motion or horizontal motion gives cleaner change-point results.
• Add seasonal correction before modeling and see how much the detection picture changes.

Learn More

• Nevada Geodetic Lab: Search for free GNSS time series, station metadata, and Nevada geodesy data products.
• USGS Earthquake Catalog: Find earthquake timing, location, and magnitude data for context and validation.
• NASA Earthdata: Search for background material on GNSS, crustal motion, and Earth observation data.
• NOAA National Centers for Environmental Information: Look for geophysical data guides and background on Earth systems analysis.
• PubMed: Search review articles on slow-slip events, post-seismic deformation, and geodetic signal detection.
• MIT OpenCourseWare: Find free course material on geophysics, statistics, and time series analysis.