Smartphone Iron Testing for Well Water

ISEF Category: Chemistry

Subcategory: Analytical Chemistry  ·  Difficulty: Intermediate  ·  Setup: School Lab  ·  Time: 1 to 2 Months

The Hook

Iron in well water can stain sinks, ruin laundry, and signal bigger water quality issues. A phone camera can spot color changes, but room light and cloudy water can throw it off. Your project can test whether a cheap reagent and a CNN can make those readings much more reliable. That gives you a real problem, a real signal, and a real chance to improve it.

What Is It?

This project measures dissolved iron by turning it into a color change. A reagent like thiocyanate or ferrozine reacts with iron and makes a colored product. The darker or stronger the color, the more iron is present. Your phone camera acts like a simple detector, and the RGB values tell you how intense the color looks.

Think of it like judging how much tea is in a cup. A pale cup and a dark cup look different, but a shadow or yellow light can fool your eye. A CNN, short for convolutional neural network, can learn patterns in images and help correct for messy lighting, glare, and turbidity, which means cloudiness in the sample. That gives you a better estimate than raw camera color alone.

Why This Is a Good Topic

This is a strong science fair topic because you can test a clear cause and effect, then improve the measurement itself. You are not just asking whether color changes, you are asking how well a phone can quantify iron under realistic conditions. That connects to groundwater testing, low-cost sensing, and access to safe drinking water. You can learn calibration, image analysis, statistics, and model validation without needing a professional lab.

Research Questions

• How does ambient lighting affect smartphone RGB estimates of dissolved iron concentration?
• What is the effect of sample turbidity on colorimetric iron readings from a phone camera?
• Does a CNN improve iron concentration predictions compared with raw RGB features alone?
• To what extent does reagent choice, thiocyanate versus ferrozine, change the linear range of detection?
• Which color channel, red, green, or blue, best predicts iron concentration under controlled lighting?
• How does container material affect measured color intensity in iron test samples?
• What is the effect of using a reference color card on between-day consistency of iron measurements?

Basic Materials

• Smartphone with a rear camera and manual exposure control app.
• White light source with stable brightness.
• Small clear sample cups or cuvettes.
• Distilled water.
• Iron standard solution or a safe teacher-prepared iron dilution set.
• Thiocyante reagent or ferrozine reagent, depending on availability.
• White cardstock or a simple light box for consistent background.
• Color reference card or printed calibration target.
• Digital kitchen scale with 0.1 g accuracy.
• Pipettes or disposable droppers.
• Notebook or spreadsheet for recording RGB values and sample labels.

Advanced Materials

• Smartphone with RAW photo capture or manual camera controls.
• Desktop or laptop with Python installed.
• Spectrophotometer for comparison data.
• Laboratory glassware for preparing iron standards.
• Certified iron standard solution.
• Ferrozine or thiocyanate reagents from a chemistry lab.
• Centrifuge or filtration setup to manage turbidity, if needed.
• Controlled light box or imaging enclosure.
• Turbidity standard such as suspended silica or clay, if approved by the lab.
• ImageJ for extracting color metrics from images.
• A labeled training set of sample images for CNN development.

Software & Tools

• Python: Trains a CNN, processes image data, and compares model predictions with simple RGB calibration.
• ImageJ: Measures color intensity from photos and helps you extract consistent image features.
• Google Colab: Lets you train a small CNN in a free cloud notebook if your computer is slow.
• Google Sheets: Organizes standards, replicate trials, and calibration data.
• MIT OpenCourseWare: Offers free background material on analytical chemistry and data analysis concepts.

Experiment Steps

1. Define the iron range you want to measure and choose one reagent system to start with.
2. Build a stable image setup so every photo uses the same camera position, background, and light geometry.
3. Prepare a calibration set with known iron levels and collect labeled images under several lighting and turbidity conditions.
4. Extract RGB features and compare simple calibration curves against a machine learning model.
5. Test the model on new samples that were not used for training and check error, bias, and repeatability.
6. Compare your final method against a reference measurement or a second imaging setup to see where it performs best and where it fails.

Common Pitfalls

• Using tap water or dirty glassware for standards, which adds iron contamination and skews the calibration.
• Training the CNN on too few images, which makes the model memorize the training set instead of learning real patterns.
• Changing the phone, camera app, or exposure settings between trials, which breaks consistency in RGB readings.
• Ignoring turbidity effects, which causes cloudy samples to look darker even when iron levels stay the same.
• Skipping a true test set, which makes the model look accurate during training but weak on new well-water samples.

What Makes This Competitive

A competitive version of this project does more than make a phone read color. It compares a simple RGB method with a model-based correction, then proves which one handles lighting and turbidity better. Strong entries also test multiple sample conditions, use a clean validation set, and report error with real statistics, not just one graph. If you can show where the phone method breaks and how much the CNN repairs it, your project gets much stronger.

Project Variations

• Test the same method on iron-spiked bottled water, river water, or school lab water instead of well water.
• Compare thiocyanate and ferrozine side by side to see which reagent gives a cleaner phone-based signal.
• Use a color reference card or gray card correction instead of a CNN and compare both approaches.

Learn More

• PubMed: Search for review articles on ferrozine iron assays, thiocyanate colorimetry, and smartphone-based analytical chemistry.
• NIH RePORTER: Look for funded projects on low-cost water testing and image-based sensing.
• NASA Earthdata: Use background on water quality, remote sensing, and measurement uncertainty to think about sensor design.
• MIT OpenCourseWare: Search for analytical chemistry and machine learning courses that cover calibration, regression, and validation.
• USGS Water Data: Find information on groundwater and well-water quality, including iron and related water chemistry topics.
• Analytical Chemistry: Search recent journal articles for smartphone colorimetry, CNN correction methods, and portable metal ion detection.