Mid-term calibration and performance of low-cost multi-sensor systems for urban air quality monitoring

Abstract This study evaluates a mid-term calibration strategy for MONICA, a compact low-cost multi-sensor system designed for urban air quality monitoring. The goal is to optimize the trade-off between calibration duration and monitoring performance, addressing the operational constraints of real-world deployments. A three-and-a-half-month co-location campaign was conducted with reference-grade instruments to calibrate and assess MONICA's performance for PM\((2.5)\), PM\((₁₀)\), and NO\((₂)\) using three models: Multiple Linear Regression (MLR), Random Forest (RF), and Generalized Additive Models (GAM).MLR showed the most stable performance across the entire monitoring period, while RF and GAM provided high short-term accuracy but exhibited reduced robustness outside the calibration window. A two-week calibration period was sufficient for PM sensors, while one week sufficed for NO\((_2)\). Despite gradual performance degradation—especially for NO\((_2)\)—MONICA successfully captured pollutant trends and regulatory exceedances defined by EU Directive 2024/2881. The results provide practical guidance for implementing scalable LCS networks in time-limited urban monitoring campaigns.