US Grid Heat And Wind Risk By 2050

ISEF Category: Earth and Environmental Sciences

Subcategory: Climate Science  ·  Difficulty: Advanced  ·  Setup: University Lab  ·  Time: Full Year

The Hook

A hot day can raise electricity demand just as wind power drops. That double hit can stress the grid fast. Climate models let you ask where that risk could grow by 2050. You can turn a big climate question into a sharp, data-driven project.

What Is It?

This project studies compound events, which are two or more weather extremes that happen together. In your case, the key pair is heat waves and low-wind days. People sometimes call that combo a dunkelflaute, a German term for dark, still weather that can hurt wind and solar output. Think of it like a car engine that needs more fuel right when the fuel line gets weaker.

You are not just asking, "Will it get hotter?" You are asking, "How often will hot days and weak-wind days overlap in the same region?" That matters because electricity demand often rises during heat waves, while renewable supply can fall when winds slow. By comparing present-day patterns with CMIP6 climate model ensembles, you can estimate how the risk changes by 2050 across US grid regions.

Why This Is a Good Topic

This is a strong science fair topic because you can frame it as a clear probability problem. You can measure how often two variables overlap, compare regions, and test whether model output changes under different bias-correction choices. The topic connects to real grid planning, energy reliability, and climate adaptation. You can also learn climate data handling, spatial analysis, and basic risk statistics without needing a wet lab.

Research Questions

• How does the probability of simultaneous heat-wave and low-wind days change across US grid regions by 2050?
• What is the effect of using different CMIP6 ensemble members on the estimated compound-event risk?
• Does bias correction change the ranking of grid regions by dunkelflaute risk?
• To what extent do coastal and inland grid regions differ in future overlap between heat waves and low-wind days?
• Which month or season shows the strongest increase in compound-event frequency by 2050?
• What is the effect of defining heat waves with a fixed threshold versus a percentile-based threshold on compound-event counts?

Basic Materials

• Laptop with enough storage for climate datasets.
• Spreadsheet software or Python installed on the laptop.
• Internet access for downloading NOAA, NASA, or CMIP6 climate data.
• External hard drive or cloud storage for large files.
• Region map of US power grid zones or NERC regions.
• Notebook for logging data decisions, variable definitions, and quality checks.

Advanced Materials

• Access to a university workstation or high-memory computer.
• Python with xarray, pandas, numpy, scipy, and matplotlib installed.
• Geospatial tools such as GeoPandas or QGIS.
• CMIP6 ensemble datasets and bias-corrected climate products.
• Reanalysis or observational wind and temperature datasets for validation.
• A gridded US power system region shapefile or boundary dataset.

Software & Tools

• Python: Processes gridded climate data, computes event overlap, and runs summary statistics.
• Jupyter Notebook: Keeps your analysis, plots, and notes in one place.
• QGIS: Maps regional compound-event risk and compares grid zones.
• xarray: Handles large netCDF climate files without flattening them first.
• ImageJ: Not needed for this topic, so skip it unless you later add map image analysis.

Experiment Steps

1. Define your compound event carefully, including the temperature threshold, wind threshold, and the regional unit you will analyze.
2. Choose one baseline period and one future window so your comparison stays consistent across all models.
3. Select a bias-correction approach and decide whether you will compare corrected and uncorrected outputs.
4. Build an event-count method that turns daily climate data into probabilities for each grid region.
5. Plan a validation check against observed or reanalysis data so you know whether your baseline matches reality.
6. Design the comparison logic for regions, seasons, and ensemble members before you start generating figures.

Common Pitfalls

• Using different definitions of heat waves across regions, which makes your probability map hard to compare.
• Counting windy and hot days separately instead of measuring their overlap on the same day.
• Mixing model output grids with power region boundaries without checking spatial alignment.
• Ignoring seasonality, which can hide the months when compound risk spikes most.
• Treating one climate model as the answer instead of comparing ensemble spread and uncertainty.

What Makes This Competitive

A stronger project will not stop at a simple map of future risk. You can compare several thresholds, several models, and several bias-correction methods, then test whether the regional ranking stays stable. You can also add uncertainty bands and a sensitivity analysis, which makes your result much more useful than a single forecast. If you connect the climate signal to grid relevance, your work becomes more than a weather exercise.

Project Variations

• Use only ERCOT, PJM, or CAISO to compare one grid market in more detail.
• Replace wind speed with low renewable output potential and test whether the compound risk pattern changes.
• Compare CMIP6 results with a reanalysis-based historical baseline to test how well models reproduce observed dunkelflaute days.

Learn More

• NASA Earthdata: Search for climate and reanalysis datasets, plus tutorials on downloading gridded data.
• NOAA Climate Data Online: Find observed temperature and wind records for baseline checks.
• USGS Climate Adaptation Science Centers: Read applied climate-risk reports and regional adaptation examples.
• NOAA National Centers for Environmental Information: Look for climate normals, extremes, and data documentation.
• PubMed: Search review articles on compound climate extremes and heat-related energy demand.
• MIT OpenCourseWare: Find free courses on climate, data analysis, and probability modeling.