Environmental Effects on NDIR-Based CH4 Monitoring: Characterization and Correction

Abstract

Nondispersive infrared (NDIR) sensors offer high sensitivity, selectivity, and low operational costs, making them particularly well-suited for environmental gas monitoring, where accurate detection of gases such as CH₄ and CO₂ is essential. However, these sensors are highly sensitive to environmental conditions, including temperature and humidity, which can significantly affect detection accuracy. This study characterizes the effects of these conditions and applies machine learning models to correct signal biases caused by multiple environmental factors. Experiments simulating natural environmental conditions for CH₄ monitoring were conducted in the laboratory across a temperature range of 10–40 °C, relative humidity levels of 10–70%, and CO₂ concentrations ranging from 0 to 1000 ppm, revealing significant signal variability under these conditions. The simulations and their results were comprehensively validated at the Ito Natural Analogue Site (INAS), a real-world field-testing location dedicated to investigating environmental impacts. Using machine learning regression algorithms for comprehensive compensation of environmental influences, we successfully mitigated signal biases caused by environmental factors. This offers a cost-effective solution for improving detection accuracy and reliability while reducing system complexity and operational costs.