Environmental CH 4 detection is critical in several applications from different sectors, from accident prevention to cost reduction in production and avoiding equipment and infrastructure damage. This paper’s demonstrates the feasibility of using low-power (low-cost) IoT end devices that incorporate metal-oxide-semiconductor (MOS) sensors with machine learning models for measuring CH 4 concentrations. MOS sensors are characterized by high sensitivity and low power consumption, making them capable of detecting small concentrations of certain gases and a good choice for systems with a low energy (and low cost) budget. However, these sensors present strong nonlinearities. One way to get around this problem is by implementing tinyML models. This paper aims to analyze the limitations of conventional MOS sensors in detecting CH 4 and to propose potential solutions. Various strategies will be detailed, including sensor heating at different temperatures and data fusion from multiple sensors to enhance the accuracy of CH 4 detection using low power IoT end devices.