Study on Cross-Instrumentation Gas Concentration Detection Based on Domain Adaptation

The ultraviolet differential optical absorption spectroscopy (UV-DOAS) technique is widely used for detecting the concentration of atmospheric pollutants. However, its concentration inversion model usually relies on specific measuring instrumentation such as spectrometers and light sources. When the instrumentation is replaced, differences in instrumentation characteristics, environmental conditions, and measurement parameters mean that the model trained with data from the original instrumentation cannot be directly applied to the new instrumentation. To address this issue, this article proposes a cross-instrumentation gas concentration detection method based on domain adaptation (DA). The research objects are nitric oxide (NO) and sulfur dioxide (SO2). The goal is to achieve the reuse of the source domain model on new instrumentation. This method only requires a small amount of target domain data to achieve DA through Gaussian process regression (GPR). The results show that the method performs excellently under different spectrometers and light source conditions. The coefficient of determination ( ${R} ^{2}$ ) for NO and SO2 exceeds 0.999. The linear error (LE) is lower than 2.00%. This method is expected to be widely applied in cross-instrumentation gas concentration detection.