Urban Night Cooling Trends With MODIS Data

ISEF Category: Earth and Environmental Sciences

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

The Hook

Cities do not cool the same way as forests or farmland. Asphalt, concrete, trees, and water all change how heat leaves the surface after sunset. You can measure that difference from space and ask which land-cover patterns slow cooling the most. That gives you a real climate question with real data.

What Is It?

This project looks at nighttime land surface temperature, or LST, from MODIS satellites. LST means the temperature of the ground or surface skin, not the air temperature you see in a weather app. At night, surfaces that held heat during the day cool at different rates. Dark pavement, dry soil, grass, and tree-covered areas all behave differently, like different kinds of cookware left on the counter after a hot meal.

Your job is to compare those cooling patterns across cities over many years. You are not just asking, “Are cities warmer?” You are asking, “How fast do they cool, and how much of that change comes from land cover versus broader climate warming?” Causal forests are a machine learning method that helps estimate how one factor changes an outcome while accounting for other factors. In simple terms, it helps you separate the effect of city form from the background climate signal.

Why This Is a Good Topic

This is a strong science fair topic because it is measurable, data-rich, and tied to a real problem. Urban heat affects energy use, sleep, health, and heat risk. You can test the idea with free satellite data, land-cover maps, and public climate records, so you do not need a lab bench. You also get to learn remote sensing, data cleaning, trend analysis, and causal thinking, which are all useful research skills.

Research Questions

• How does nighttime urban cooling rate change across cities with different tree cover levels?
• What is the effect of impervious surface fraction on nighttime cooling rate after sunset?
• Does nighttime cooling trend differ between urban cores and nearby suburban areas?
• To what extent does background regional warming explain changes in urban nighttime LST over two decades?
• Which land-cover variable best predicts slower nighttime cooling, tree cover, water cover, or impervious surface?
• How does nighttime cooling differ between humid and dry climate zones after controlling for land cover?
• To what extent do urban form variables reduce the apparent warming trend in MODIS nighttime LST?

Basic Materials

• Computer with reliable internet access.
• External hard drive or cloud storage for large satellite files.
• Spreadsheet software for data screening and plotting.
• Python installed with pandas, numpy, matplotlib, and scikit-learn.
• Google Earth Engine account for screening MODIS scenes and pulling land-cover data.
• Public land-cover dataset such as NLCD for U.S. cities or ESA WorldCover for global cities.
• NOAA or NASA climate records for background temperature comparison.

Advanced Materials

• Computer with high RAM and storage for large raster stacks.
• Python with xarray, rasterio, geopandas, statsmodels, and scikit-learn.
• QGIS for map inspection and spatial masking.
• MODIS LST night products from NASA Earthdata.
• MODIS land-cover products or a comparable global land-cover raster.
• City boundary shapefiles from census or municipal GIS portals.
• Reanalysis or station climate data for covariate control.
• Optional access to a workstation or university server for repeated model runs.

Software & Tools

• Google Earth Engine: Screens satellite scenes, builds city-scale composites, and helps match dates across years.
• Python: Cleans, joins, and models the time series and spatial predictors.
• QGIS: Checks city boundaries, land-cover masks, and map outputs.
• ImageJ: Measures image-based validation plots if you compare false-color map exports.
• R: Runs alternative statistics and makes publication-style graphics.

Experiment Steps

1. Define the city set and the exact time window you will compare.
2. Choose one nighttime LST product and one land-cover dataset so your inputs stay consistent.
3. Build a clean city-by-year table that links temperature trends to land-cover fractions and climate covariates.
4. Decide how you will separate local land-cover effects from background warming with a causal model or matched comparison design.
5. Plan validation checks that test whether your result changes when you switch cities, years, or masking rules.
6. Pre-register the main outcome you will report, such as cooling rate, trend slope, or model-attributed effect size.

Common Pitfalls

• Mixing daytime and nighttime MODIS products, which makes cooling trends impossible to compare.
• Using city boundaries that change over time, which can fake a trend that comes from geometry, not climate.
• Comparing raw LST values without masking clouds, water, or missing pixels, which adds noise and bias.
• Treating land cover and regional climate as the same effect, which hides whether the city itself or the broader climate drives the trend.
• Running a causal forest on too few cities or too few years, which makes the model unstable and hard to interpret.

What Makes This Competitive

A stronger project will go past simple trend plots. You can separate urban form from background warming, test whether the pattern holds across climate zones, and compare multiple model specs. The best versions also check uncertainty, not just point estimates. If you can show that one land-cover factor keeps predicting slower nighttime cooling after careful controls, your project starts to look like real research.

Project Variations

• Compare nighttime cooling trends across cities with different tree canopy percentages and see whether greener cities cool faster over time.
• Test whether coastal cities show different nighttime cooling patterns than inland cities after controlling for land cover.
• Swap in a different remote sensing product, such as Landsat-based surface temperature, and compare whether the trend direction matches MODIS.

Learn More

• NASA Earthdata: Search for MODIS land surface temperature products and user guides.
• NASA MODIS Land Products: Find product descriptions, quality flags, and data documentation on NASA sites.
• NOAA National Centers for Environmental Information: Use climate normals, station records, and long-term temperature context.
• USGS EarthExplorer: Explore land-cover, elevation, and related geospatial datasets.
• PubMed: Search for review articles on urban heat islands, land cover, and surface temperature.
• MIT OpenCourseWare: Look for free courses in remote sensing, climate science, and environmental data analysis.