Edge Analytics based Acoustic Rain Gauge

Non-Mechanical Rain gauge

The Acoustic Rain Gauge leverages sound waves to detect raindrops and measure rainfall intensity with high precision and efficiency. Designed to operate without moving parts, this system minimizes maintenance while providing real-time rainfall data. The non-contact approach enhances durability and reliability, making it suitable for various environmental conditions.

By analyzing sound variations produced by raindrop impacts, the acoustic rain gauge offers accurate rainfall measurements, supporting applications in agriculture, water resource management, disaster preparedness, and climate resilience. This innovative technology underscores the importance of integrating IoT devices and machine learning for sustainable development.

Technologies Used

Hardware:

• Professional USB microphone with Type-C connector
• Raspberry Pi 4 Model B with wireless LAN, Bluetooth, and USB ports
• Tipping bucket rain gauge integrated with Arduino Pro Mini
• 50Ah Li-ion battery pack with battery management system
• 100W photovoltaic cell for solar charging

Software:

• Deep learning algorithms with LSTM blocks in Keras
• LoRaWAN for data transmission
• Chirpstack server for receiving tipping bucket data
• Python for audio processing and machine learning

Working Principle

The acoustic rain gauge operates by recording sound waves generated by raindrop impacts, with the key steps as follows:

1. Sound Data Collection: A professional USB microphone captures raindrop frequencies.
2. Processing and Storage: The Raspberry Pi processes the audio data and stores it locally.
3. Comparison with Tipping Bucket Data: Acoustic data is compared with rainfall measurements from a tipping bucket rain gauge for validation.
4. Rainfall Prediction: A Deep Neural Network (DNN) model using LSTM blocks predicts rainfall intensity based on the recorded sound data.

The tipping bucket rain gauge collects rainfall data every three minutes, transmitting it via LoRaWAN to a Chirpstack server. This data serves as a reference for training the acoustic rain gauge's machine learning model.

Experimental Setup

• Power System: Both the acoustic and tipping bucket setups are powered by solar energy. The acoustic system relies on a 50Ah Li-ion battery with solar charging, while the tipping bucket uses a 3.8V battery and solar charger.
• Data Collection: The tipping bucket rain gauge records rainfall data at three-minute intervals, while the acoustic system continuously captures audio data.
• Data Modeling: The acoustic data, structured as a time series, is split into training (75%) and testing (25%) datasets. A deep learning model developed in Keras evaluates rainfall predictions, reporting accuracy using Mean Absolute Percentage Error (MAPE).

Applications

• Meteorological research
• Disaster management and flood early warning systems
• Precision agriculture and water resource management
• Environmental monitoring and climate resilience

Conclusion

The Acoustic Rain Gauge combines IoT edge devices, deep learning, and renewable energy to create a robust and sustainable solution for rainfall measurement. By integrating acoustic data with traditional tipping bucket measurements, it ensures accurate and reliable predictions. This system represents a significant step forward in meteorological research and disaster management, enabling informed decision-making and supporting sustainable practices.