Satellite Methane Detection from Landfills

ISEF Category: Environmental Engineering

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

The Hook

Landfills can leak methane, and you cannot see most of it from the ground. Satellites can spot gas patterns from space, almost like reading a heat map for invisible pollution. That gives you a real chance to compare what sites report with what the sky shows. Your project can ask a real environmental question with public data.

What Is It?

This project uses satellite data to look for methane plumes above landfills. Methane is a greenhouse gas, and a plume is a moving cloud of higher gas concentration. In this case, you are not watching smoke. You are reading a satellite signal that changes when methane in the air is higher than the background.

Sentinel-5P carries the TROPOMI sensor, which measures gases in the atmosphere across large areas. Think of it like a very wide camera that does not capture color, but instead captures gas patterns. A convolutional neural network, or CNN, is a machine learning model that learns image patterns. You can use it to sort scenes or flag likely plume shapes, then compare those patterns with reported landfill emissions.

Why This Is a Good Topic

This topic is strong because you can test a real-world pollution question with public data. You do not need a wet lab to start, and you can still do original work by comparing satellite observations, reported inventories, and site features. The project connects clean air, climate science, and machine learning. You can also learn data cleaning, image analysis, and model evaluation, which makes the final project much stronger than a simple chart.

Research Questions

• How does reported landfill methane emissions compare with satellite-observed methane signals over the same sites?
• What is the effect of landfill size on the strength of the detected methane signal?
• Does distance from major urban areas change the chance that a landfill shows a clear methane plume?
• To what extent do seasonal changes affect methane detection at landfill sites?
• Which satellite image features best predict whether a landfill has a high observed methane signal?
• How does a CNN classifier compare with a simple threshold rule for identifying likely methane plumes?

Basic Materials

• Laptop or desktop computer with internet access.
• External hard drive or cloud storage for large satellite files.
• Spreadsheet software, such as Google Sheets or Excel.
• Python installed through Anaconda or another free distribution.
• Jupyter Notebook for data cleaning and analysis.
• Satellite or map data viewer, such as NASA Earthdata Search or Google Earth Pro.
• Public landfill emissions inventory data from EPA or other government sources.

Advanced Materials

• Access to Sentinel-5P TROPOMI Level-2 methane data.
• Python environment with geospatial libraries such as xarray, rasterio, geopandas, and scikit-learn.
• GPU access for CNN training if your image set is large.
• GIS software such as QGIS for spatial checks and map overlays.
• Ground truth or reference datasets for methane hotspots, if available from published studies.
• High-resolution land cover or landfill boundary data from government or municipal sources.

Software & Tools

• Python: Cleans satellite data, builds features, and runs the model.
• Jupyter Notebook: Organizes code, notes, and visual outputs in one place.
• QGIS: Lets you map landfill locations and compare them with satellite scenes.
• Google Earth Pro: Helps you inspect sites, check surroundings, and verify coordinates.
• ImageJ: Measures image intensity patterns if you export satellite scenes as images.

Experiment Steps

1. Define the exact comparison you want, such as reported emissions versus observed methane signal at landfill sites.
2. Choose a small set of landfills and decide how you will match each site to satellite scenes in the same time window.
3. Design the signal rule you will use, whether that is a CNN classifier, a plume score, or a simple threshold baseline.
4. Plan controls that rule out bad weather, cloud contamination, and nearby sources that could confuse your results.
5. Build a clean dataset and keep site metadata, image dates, and emission values in one consistent table.
6. Decide how you will judge success, such as accuracy, correlation, or agreement between reported and observed emissions.

Common Pitfalls

• Using scenes with clouds or bad retrieval quality, which can hide methane signals and distort your labels.
• Mixing up reported inventory values with satellite retrieval units, which makes your comparison meaningless.
• Training a CNN on too few landfill examples, which causes the model to memorize sites instead of learning patterns.
• Ignoring nearby oil, gas, or industrial sources, which can create false positives around a landfill.
• Comparing data from different dates without matching weather and season, which can make emissions look higher or lower for the wrong reason.

What Makes This Competitive

A strong version of this project does more than map hotspots. It tests a clear method, compares it with a baseline, and reports how well the method works on sites it did not train on. You can push the work further by checking whether the model handles cloud cover, seasons, and nearby sources correctly. A careful error analysis and a clean validation plan can make the project feel much more like real environmental research.

Project Variations

• Compare methane signals across active, closed, and capped landfills to see whether site status changes detectability.
• Replace the CNN with a simpler classifier or regression model, then compare performance and interpretability.
• Add a spatial analysis layer that tests whether landfill methane signals cluster near specific land use types or population centers.

Learn More

• NASA Earthdata: Search for Sentinel-5P TROPOMI methane data, tutorials, and data access tools.
• NOAA Global Monitoring Laboratory: Review background material on methane as a greenhouse gas and atmospheric measurements.
• EPA Landfill Methane Outreach Program: Find landfill emissions context and reporting background from a government source.
• PubMed: Search for review articles on satellite detection of methane plumes and landfill emissions.
• MIT OpenCourseWare, Introduction to Machine Learning: Use course materials to learn the basics of classification and model evaluation.
• QGIS Documentation: Learn free GIS tools for mapping sites and checking satellite overlays.