<p>Environmental monitoring is vital for understanding, pollution, climate change, and biological health. Historical monitoring systems, usually comprised of expensive stations, provide limited monitoring and inefficiencies. IoT sensors and Machine Learning (ML) algorithms are changing things, allowing us to scale resources and provide significant environmental sensing at lower cost, ideally in real-time. This research describes an AI-powered, IoT-based air quality monitoring system that employs inexpensive electrochemical sensors to monitor Carbon monoxide (CO) and Nitrogen dioxide (NO₂), and infrared sensors when measuring particulate matter present in the atmosphere. The system uses solar-rechargeable batteries and long-range wireless communication using Zigbee which allows it to be energy sustainable. To improve data accuracy, the Adaptive Prairie Dog Optimized Efficient Isolation Forest (APDO-EIF) method is employed for sensor calibration, temperature-humidity compensation, and noise reduction. The system is deployed to collect real-time data. The collected data undergoes preprocessing, then feature selection and optimization using APDO. A T-test was performed to statistically validate the performance improvements of the proposed model compared to traditional methods. Comparative analyses with traditional methods demonstrate significant improvements in accuracy and precision, with the proposed model achieving 98.95% accuracy, and 98.2% recall: 97.8% F-1 score, and 96.8% precision. The novelty of this approach is the implementation of a majority voting scheme that improves decision-making through the combination of outputs of several sensors, making results more reliable. This work adds to the evidence of the utility of ML-driven IoT systems for biodiversity tracking and, thus, demonstrates the potential of ML-driven IoT in environmental monitoring, and makes strides toward better decision-making for sustainability and public health efforts.</p>