Storm Drain Discharge Detection

ISEF Category: Earth and Environmental Sciences

Subcategory: Water Science  ·  Difficulty: Advanced  ·  Setup: School Lab  ·  Time: Full Year

The Hook

A storm drain can look empty and still carry a pollution event you never see. One sudden jump in conductivity can hint that wastewater, road runoff, or another odd source entered the system. Your job is to catch that jump before it gets lost in the noise. This project turns a simple sensor into an early warning tool.

What Is It?

This project looks for illicit-discharge events in storm drains. An illicit discharge means something entered the drain system when it should not have, such as wastewater, soap, or industrial runoff. You are not trying to see the pollutant directly. You are watching for a pattern in water quality that changes too fast to ignore.

Think of the drain like a heartbeat monitor. If the signal stays steady, the system looks normal. If conductivity, which measures how well water carries electricity, suddenly rises or falls, that can point to a new source of dissolved ions. Temperature helps too, because some events bring water that is warmer or cooler than the surrounding flow. A change-point detector such as CUSUM watches for shifts in the data instead of waiting for a huge spike, so it can flag subtle events that basic threshold rules miss.

Your project can use a small logger, like an ESP32, to collect data over time. Then you can compare normal conditions with known or simulated anomaly events. The goal is to see whether your detection method can find changes quickly, cleanly, and with few false alarms.

Why This Is a Good Topic

This is a strong science fair topic because you can test a real environmental problem with measurable data. Stormwater pollution affects rivers, lakes, and public health, and cities need better ways to spot hidden discharge events. You can build a clear comparison between normal flow and anomaly periods, then judge how well your detector performs. That gives you real engineering, real data analysis, and a result you can explain without a huge lab setup.

Research Questions

• How does a conductivity-based change-point detector compare with a fixed-threshold rule for finding sudden storm drain anomalies?
• What is the effect of adding temperature correction on the false alarm rate of a conductivity logger?
• Does sampling interval change the detector's ability to identify short discharge events?
• To what extent does baseline variability in normal storm drain data affect CUSUM detection accuracy?
• Which combination of conductivity and temperature features gives the earliest reliable alert for a discharge event?
• How does sensor drift over days or weeks affect anomaly detection performance?

Basic Materials

• ESP32 microcontroller board.
• Conductivity sensor probe compatible with the ESP32.
• Waterproof temperature sensor or thermistor probe.
• MicroSD card module for data storage.
• Real-time clock module for time stamps.
• Breadboard and jumper wires.
• 5V or USB power supply.
• Waterproof enclosure for outdoor deployment.
• Digital multimeter for circuit checks.
• Distilled water for baseline testing.
• Tap water and simple salt solutions for calibration.
• Laptop for data download and analysis.

Advanced Materials

• Lab-grade conductivity meter for calibration checks.
• Reference conductivity standards.
• Thermometer or temperature bath for sensor validation.
• Peristaltic pump for controlled flow tests.
• Flow channel or flume for repeatable stormwater simulation.
• Data-logging weather station for rainfall and ambient context.
• Field deployment housing with cable glands and strain relief.
• Replacement probes for drift comparison.
• GPS-enabled handheld device for site mapping.
• Particle filter setup for pretesting sensor fouling effects.

Software & Tools

• Arduino IDE: Programs the ESP32 and records sensor data in a simple format.
• Python: Cleans time-series data, runs CUSUM, and makes detection plots.
• pandas: Organizes sensor logs and helps you align conductivity with temperature and time.
• NumPy: Supports numerical calculations for baseline statistics and detector tuning.
• ImageJ: Measures visual change in calibration images if you document sensor fouling or test setups.

Experiment Steps

1. Define the exact anomaly you want to detect, such as a sudden conductivity shift, and decide how you will label it in your data.
2. Build a clean baseline dataset so you know what normal storm drain behavior looks like before any change-point testing.
3. Choose the detector inputs, then decide whether you will use conductivity alone or conductivity plus temperature correction.
4. Set up a calibration plan that converts raw sensor readings into a consistent signal you can compare across days.
5. Design a comparison between your detector and at least one simpler method, such as a fixed threshold or moving average rule.
6. Plan how you will score performance, including detection delay, false alarms, and missed events.

Common Pitfalls

• Treating every conductivity spike as pollution, which confuses road salt, rainfall dilution, and true illicit discharge.
• Ignoring temperature swings, which can shift conductivity enough to create false change points.
• Logging data at an interval that is too slow, which can miss short discharge events.
• Calibrating with only one standard, which leaves you unable to check whether the sensor response stays linear across your full range.
• Testing only on clean tap water, which hides fouling, drift, and the messy background noise that makes field detection hard.

What Makes This Competitive

A stronger project does more than say, “My sensor found a spike.” You would compare several detection rules, quantify false alarms, and test whether temperature correction or baseline normalization improves performance. You can also strengthen the project by using real field data, then checking your detector against known weather events, maintenance logs, or site observations. That kind of careful validation makes the project feel like environmental engineering, not just data logging.

Project Variations

• Test the same detection idea in roadside catch basins instead of storm drains to see whether traffic runoff changes the signal.
• Compare conductivity-based anomaly detection with turbidity-based detection to see which sensor catches discharge events earlier.
• Run the detector on simulated runoff mixtures with salt, soap, or fertilizer to separate different pollution signatures.

Learn More

• USGS Water Science School: Explains conductivity, water quality, and stormwater concepts, and you can find it by searching USGS Water Science School conductivity and stormwater.
• EPA Stormwater Program: Covers urban runoff, illicit discharge, and water protection basics, and you can find it by searching EPA illicit discharge stormwater.
• NOAA National Water Prediction Service: Offers rain and runoff context for field sites, and you can find it by searching NOAA stormwater rainfall data.
• PubMed: Search for review articles on conductivity sensing, water quality monitoring, and anomaly detection in environmental data.
• MIT OpenCourseWare: Look for courses on environmental engineering, sensors, and data analysis to build background on field measurement and signal processing.