Mid-term Performance and Calibration of Multi-sensor Low-Cost Systems for Air Quality Monitoring in Urban Environments

This study evaluates mid-term calibration strategies for MONICA, a compact low-cost multi-sensor device for urban air quality monitoring, in the context of the upcoming EU Air Quality Directive 2024/2881. A key objective is to optimize the trade-off between calibration duration and long-term monitoring performance, balancing statistical robustness with practical deployment constraints. This question is particularly relevant in mid-latitude regions, where seasonal variability may introduce biases if calibration and observation periods are misaligned. Over a three-and-a-half-month winter co-location with reference-grade instruments, we assessed three models—Multiple Linear Regression (MLR), Random Forest (RF), and Generalized Additive Models (GAM)—across three pollutants: PM \(2.5\) , PM \(₁₀\) , and NO \(_2\) . MLR emerged as the most stable model for temporal extrapolation, while RF and GAM, although accurate short-term, showed performance degradation outside the training range. A two-week calibration period was sufficient for PM, whereas NO \(_2\) required only one week. Although sensor accuracy declines over time-especially for NO \(_2\) —the MONICA system remains effective in tracking temporal trends and identifying regulatory exceedances. These results support the development of efficient and scalable low-cost sensor networks, offering practical insights for planning reliable air quality monitoring campaigns.