Power Efficient Non-Mechanical Weather Station: Harnessing TinyML for Rainfall Classification and Environmental Sensing

In response to the pressing need for accurate and cost-effective weather monitoring in agricultural settings, this paper presents a next-generation Tiny Machine Learning (TinyML) smart weather station. Leveraging TinyML technology, this work offers real-time classification of rainfall intensity using audio signals. Through meticulous data collection, preprocessing, and feature extraction, coupled with a well-designed 1D Convolutional Neural Network (CNN) architecture, the TinyML model achieves an accuracy of 98.967% while maintaining low computational and memory requirements. Deployed on the Arduino Nano 33 BLE Sense, the weather station provides farmers with essential weather data for informed decision-making, contributing to enhanced agricultural productivity and resilience. With a focus on power optimization, this work ensures prolonged operation without frequent battery replacements, making it suitable for deployment in remote or resource-limited environments. This paper details the development process, model architecture, deployment strategy, and evaluation results, highlighting the potential impact on weather monitoring practices in agricultural contexts.