Smartphone Pond Microbe Biodiversity Monitoring

ISEF Category: Computational Biology and Bioinformatics

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

The Hook

A pond can look clear and still hold a very different microbial community from one just down the road. Those tiny organisms can act like a living water-quality report card. With a phone and a cheap clip-on lens, you can start reading that report yourself. The challenge is turning messy images into useful biodiversity data.

What Is It?

This project studies pond-water microorganisms, the tiny life forms you can see under a simple microscope setup. Think of each image as a crowded street scene. Some organisms are easy to spot, some overlap, and some only show up in certain water conditions.

A self-supervised classifier is a machine learning model that learns patterns from unlabeled images. Instead of teaching it every organism name by hand, you let it find useful visual features on its own. Then you use those features to estimate biodiversity, which is a measure of how many different kinds of organisms appear and how evenly they are represented.

Why This Is a Good Topic

This makes a strong science fair topic because you can collect your own data, build your own image set, and test whether a low-cost phone microscope can track real differences between ponds or sampling dates. You also get a clear problem to solve, how to measure microbial biodiversity without a full research lab. The project teaches image labeling, feature learning, model validation, and basic ecological thinking.

Research Questions

• How does pond location affect the biodiversity index estimated from smartphone microscope images?
• What is the effect of using a self-supervised model instead of a supervised model on classification accuracy for unlabeled pond images?
• Does image magnification level change the number of microorganism types detected per sample?
• To what extent does sampling time across the day change the microbial community signature in the same pond?
• Which image preprocessing step improves biodiversity estimates the most, contrast normalization, background subtraction, or cropping?
• How does repeated sampling of the same site affect the stability of the biodiversity index over time?

Basic Materials

• Smartphone with a good camera.
• Clip-on microscope lens, about $10 to $20.
• Clear slides and cover slips.
• Pipettes or droppers.
• Small labeled sample cups or jars.
• Notebook or spreadsheet for field notes.
• White paper or LED flashlight for consistent lighting.
• Simple microscope stand or stable phone holder.
• Distilled water for rinsing tools.
• Gloves and hand sanitizer.

Advanced Materials

• Smartphone with manual camera controls.
• Clip-on microscope lens or small USB microscope.
• Compound microscope for comparison imaging.
• Computer with Python installed.
• Image labeling software such as Label Studio or CVAT.
• Python libraries for computer vision and machine learning.
• External hard drive or cloud storage for image archives.
• Calibration slide or stage micrometer.
• Field sampling kit with sterile containers.
• GPS-enabled phone or tablet for site tracking.

Software & Tools

• Python: Runs image preprocessing, feature extraction, and model training.
• ImageJ: Measures image quality, organism size, and basic morphology features.
• Google Colab: Lets you train models in the browser without a powerful laptop.
• Label Studio: Helps you organize and review image labels or cluster labels.
• QGIS: Maps pond sampling sites and compares biodiversity patterns across locations.

Experiment Steps

1. Define the biodiversity question you want to answer, such as site differences, seasonal change, or method comparison.
2. Design a consistent imaging setup so your photos stay comparable across sampling days and locations.
3. Plan a labeling strategy, including whether you will hand-label a small subset, use clustering first, or combine both approaches.
4. Build a baseline analysis that turns each image into a numerical signal before you try a self-supervised model.
5. Set up controls that test whether lighting, focus, or water debris changes your biodiversity estimate more than the biology does.
6. Choose a statistical comparison that matches your question, then decide how you will report uncertainty and model confidence.

Common Pitfalls

• Changing lighting between samples, which makes the same microorganism look like a different class.
• Collecting water from very different depths or distances, which mixes sampling variation with true biodiversity change.
• Using blurry or moving images, which causes the model to learn camera motion instead of organism shape.
• Labeling only the easiest organisms, which gives the classifier a skewed view of the sample.
• Treating debris, air bubbles, or pollen as microbes, which inflates the biodiversity index.

What Makes This Competitive

A stronger project would compare more than one pond, more than one imaging method, or more than one model type. You could test whether self-supervised learning beats a small hand-labeled baseline on messy real-world samples. You could also pair the image model with water chemistry data or site metadata to see whether the biodiversity index tracks environmental change. Careful error analysis and clean validation matter a lot here.

Project Variations

• Compare urban pond water, suburban pond water, and farm runoff channels to see whether the biodiversity signal changes across land use.
• Test whether a self-supervised model trained on raw images performs better than one trained on contrast-enhanced images.
• Add a second analysis layer that groups images by organism shape and size, then compare those groups with the biodiversity index.

Learn More

• NASA Earthdata: Search for citizen science and remote sensing resources that explain environmental data collection and validation.
• NIH PubMed: Search for review articles on freshwater microbial diversity, image-based ecology, and self-supervised learning in biology.
• USGS Water Science School: Read background pages on water quality, aquatic ecosystems, and sampling basics.
• NOAA Education Resources: Find free material on freshwater ecosystems, biodiversity, and environmental monitoring.
• MIT OpenCourseWare: Look for introductory machine learning and computer vision course materials that are open to the public.