Urban Heat and Tree Flowering Patterns

ISEF Category: Plant Sciences

Subcategory: Ecology  ·  Difficulty: Advanced  ·  Setup: Home Setup  ·  Time: Full Year

The Hook

City streets can act like tiny heat traps. That extra warmth can push some trees to bloom earlier than trees in cooler neighborhoods. If you can measure that shift, you can connect climate, plants, and real city data in one project. You also get to work with photos, maps, and machine learning.

What Is It?

Phenology is the study of when living things do seasonal events, like budding, flowering, or leaf drop. In this project, you are asking whether street trees in warmer parts of a city reach those stages sooner than trees in cooler parts. Think of it like comparing clocks. One clock sits in a shaded park, and another sits next to asphalt and buildings that hold heat. If the city stays warmer, the tree’s seasonal clock may run faster.

You can build this project with public photos and location data. iNaturalist gives user-uploaded observations of plants, and Google Street View gives past images of the same streets. A CNN, which is a convolutional neural network, is a kind of image model that can sort pictures by visual features. You can train or test a model to label flowering stages, then compare those labels across neighborhoods with different heat levels.

Why This Is a Good Topic

This is a strong science fair topic because you can test a clear question with public data, and you do not need a wet lab. You can connect plant timing to the urban heat-island effect, which matters for city planning, pollinators, and climate adaptation. You will learn data cleaning, image analysis, basic machine learning, and statistics. That mix gives you enough depth for a serious project if you define your tree species and city carefully.

Research Questions

• How does neighborhood surface temperature relate to the first flowering date of a chosen street tree species?
• What is the effect of distance from dense pavement on flowering stage timing in street trees?
• Does the flowering-stage classifier perform better on one tree species than another?
• To what extent do warmer census blocks show earlier peak bloom dates than cooler blocks?
• Which image source, iNaturalist or Google Street View, gives more reliable flowering-stage labels?
• How does tree canopy cover change the link between urban heat and flowering time?
• What is the effect of using local temperature anomaly instead of citywide temperature on model accuracy?

Basic Materials

• Laptop with internet access.
• Spreadsheet software.
• Google Earth or Google Maps for neighborhood scouting.
• iNaturalist observations for your target tree species.
• Google Street View history images.
• Public temperature or land surface temperature data from NOAA or NASA.
• Digital notebook for logging image labels and location metadata.
• Reference photos of the tree species at different flowering stages.

Advanced Materials

• Laptop with a modern GPU or access to a school workstation.
• Python installed with common data science libraries.
• Jupyter Notebook for analysis.
• Image annotation tool for labeling flowering stages.
• OpenStreetMap or GIS software for mapping sample sites.
• Public climate datasets from NOAA, NASA, or USGS.
• Spreadsheet software for metadata cleanup.
• Version control platform for tracking code and labels.

Software & Tools

• Python: Organizes images, runs the classifier, and handles statistical analysis.
• Jupyter Notebook: Lets you document your workflow, code, plots, and notes in one file.
• ImageJ: Helps you inspect images and check whether flowering labels match visible features.
• QGIS: Maps tree locations and compares flowering timing with neighborhood heat patterns.
• Google Earth: Helps you review street-view history and pick consistent sampling sites.

Experiment Steps

1. Define one tree species, one city, and one flowering stage system so your dataset stays consistent.
2. Choose a heat metric, such as surface temperature, tree canopy cover, or distance from major roads.
3. Build a labeling plan that turns photos into a clear flowering-stage dataset.
4. Split your images into training, validation, and test groups so you can check model performance honestly.
5. Link each observation to location and date, then match it with the same neighborhood heat data.
6. Compare flowering timing across warmer and cooler areas, and test whether your pattern holds after you control for sample size and image source.

Common Pitfalls

• Mixing different tree species, which makes your flowering pattern look real when it only reflects species differences.
• Using photos with inconsistent angles or zoom levels, which hides the flower parts your classifier needs.
• Labeling bloom stages too loosely, which blurs early bloom, peak bloom, and late bloom into one category.
• Comparing neighborhoods with very different observation counts, which can bias the heat-island result.
• Ignoring image date or location errors, which can put a tree in the wrong season or the wrong block.

What Makes This Competitive

A competitive version needs more than a simple map of early bloom sites. You can strengthen it by using a clearly defined phenotype, a careful validation set for the classifier, and a neighborhood heat metric that matches your biological question. Strong projects also test more than one explanation, like pavement, canopy cover, and distance from roads. If you can show that your result survives those checks, your project feels much more like research and much less like a class assignment.

Project Variations

• Focus on one common street tree species, like cherry or maple, to reduce biological noise and sharpen the heat signal.
• Compare iNaturalist-only labels with Google Street View labels to test whether one image source misses early bloom more often.
• Swap flowering stage for leaf-out timing, which may be easier to score in cities with fewer flower photos.

Learn More

• NOAA Climate Data Online: Search for local temperature records and urban climate patterns by city.
• NASA Earthdata: Find land surface temperature, land cover, and urban heat island datasets.
• USGS National Map: Use elevation, land cover, and mapping layers to support site comparisons.
• PubMed: Search review articles on urban heat islands, phenology, and plant response to warming.
• Urban Ecosystems: Search this journal for peer-reviewed studies on city trees and seasonal timing.
• MIT OpenCourseWare: Look for free lectures on machine learning, image classification, and data analysis.