Tree-Ring Drought Reconstruction With ML

ISEF Category: Earth and Environmental Sciences

Subcategory: Climate Science  ·  Difficulty: Advanced  ·  Setup: Home Setup  ·  Time: Full Year

The Hook

A tree can hold decades of climate history in its rings. Each ring is like a yearly receipt for water stress, heat, and growth. If you can match ring patterns to real drought records, you can estimate past drought even where weather stations did not exist. That turns old wood into a climate archive.

What Is It?

This project uses tree rings as a proxy, which means you use one natural signal to estimate another. Trees often grow wider rings in wet years and thinner rings in dry years. If you compare ring patterns with measured drought data, you can train a model to learn that relationship.

Think of it like teaching a computer to read a handwritten note. The measured drought record is the answer key, and the tree rings are the clue. Once the model learns the pattern, you can apply it to older ring series and estimate drought severity before modern instruments were available. Public tree-ring data from the International Tree-Ring Data Bank, or ITRDB, makes this project possible without collecting your own samples.

Why This Is a Good Topic

This is a strong science fair topic because you can ask a clear question, test it with public data, and compare model performance across time periods or sites. It connects to water supply, wildfire risk, agriculture, and climate history. You can learn data cleaning, feature selection, model validation, and how scientists turn indirect evidence into climate estimates.

Research Questions

• How does the choice of drought index affect the accuracy of a tree-ring proxy model?
• What is the effect of using one watershed's tree-ring chronologies versus multiple nearby chronologies on reconstruction skill?
• Does a random forest model outperform linear regression for estimating historical drought severity from ring-width data?
• To what extent do early 20th-century reconstructions differ when you train the model on different calibration periods?
• Which tree-ring chronologies contribute the most to predicting drought in the target watershed?
• How does excluding low-quality or short chronologies change model uncertainty?

Basic Materials

• Laptop or desktop computer with internet access.
• Spreadsheet software such as Google Sheets or Excel.
• Python installed through Anaconda or a similar free distribution.
• Public tree-ring chronology data from the ITRDB.
• Public drought or climate data from NOAA or USGS.
• Notebook for tracking data cleaning decisions and model tests.

Advanced Materials

• Laptop or desktop computer with internet access.
• Python with pandas, scikit-learn, matplotlib, seaborn, and statsmodels.
• Jupyter Notebook for reproducible analysis.
• Public tree-ring chronology data from the ITRDB.
• Public drought or climate data from NOAA, USGS, or NOAA NCEI.
• GIS software such as QGIS for mapping chronology sites and watershed boundaries.
• Statistical comparison tools for cross-validation and uncertainty analysis.

Software & Tools

• Python: Runs data cleaning, model training, prediction, and validation.
• Jupyter Notebook: Keeps code, notes, and results together in one reproducible file.
• pandas: Organizes tree-ring and climate tables for merging and filtering.
• scikit-learn: Trains proxy models and tests how well they predict drought.
• QGIS: Maps chronology locations against your target watershed and climate region.

Experiment Steps

1. Define your target watershed and choose one drought metric you can obtain for the calibration period.
2. Collect matching tree-ring chronologies from the ITRDB and screen them for overlap, length, and site relevance.
3. Align the tree-ring series with instrumental drought data and decide how you will handle missing years, standardization, and lagged effects.
4. Build a baseline model first, then compare it with a more flexible ML model so you can judge whether extra complexity helps.
5. Plan a validation strategy that tests the model on held-out years and, if possible, on a separate time block.
6. Decide how you will present uncertainty, model importance, and differences between reconstructed and observed drought patterns.

Common Pitfalls

• Mixing tree-ring series with different chronology lengths, which creates hidden gaps and unfair comparisons.
• Using the full record for training and testing, which makes the model look better than it really is.
• Ignoring calibration period choice, which can change the relationship between rings and drought.
• Comparing raw ring widths without standardizing tree age effects, which can blur the climate signal.
• Treating nearby chronologies as independent when they may reflect the same climate pattern.

What Makes This Competitive

A competitive project goes beyond making one reconstruction. You can compare multiple models, multiple drought metrics, and multiple chronology subsets, then test which choices hold up under validation. Strong projects also report uncertainty clearly and explain why certain sites or years matter more than others. If you can show a careful method that improves reliability or interpretation, your project becomes much stronger.

Project Variations

• Use summer drought instead of annual drought so you can test whether ring growth tracks the season that matters most for trees.
• Compare one watershed to a nearby watershed to see whether the proxy model generalizes across space.
• Swap ring-width chronologies for density or isotope-based records, if you can find public data with the same climate target.

Learn More

• International Tree-Ring Data Bank (ITRDB): Search the ITRDB database for public tree-ring chronologies and metadata for your study region.
• NOAA National Centers for Environmental Information: Search for drought, precipitation, and climate reconstructions and instrumental records.
• USGS Water Data for the Nation: Find watershed streamflow and hydrology data that can help define local drought impacts.
• NOAA Paleoclimatology: Look for review papers, datasets, and proxy climate records related to tree rings.
• MIT OpenCourseWare, Introduction to Machine Learning: Use free course materials to learn model training, validation, and error metrics.
• PubMed: Search for review articles on dendroclimatology, proxy modeling, and drought reconstruction.