Ground-Based Validation of Sentinel-5P TROPOMI Atmospheric Products using Calibration-Informed Low-Cost Multi-Spectral Sensors

Abstract. Ground-based validation of satellite atmospheric products is essential for ensuring data quality and algorithm performance. We present a validation approach for Sentinel-5P TROPOspheric Monitoring Instrument (TROPOMI) cloud fraction products using a multi-spectral ground station (DG2MCM-15) located in Kempten, Bavaria, Germany. The ground observatory combines professional metrological experience from ISO/IEC 17025 accredited laboratory environments with low-cost commercial sensors, creating a citizen science validation capability.

Our validation dataset comprises 276 temporally matched observations between Sentinel-5P overpasses and ground measurements over a four-week period (January 11 – February 8, 2026). Ground-based cloud detection using an MLX90614 infrared pyrometer achieves strong agreement with Sentinel-5P cloud fraction retrievals (Pearson R = 0.879, N = 27 after quality filtering). The root mean square error of 29.1 % cloud fraction reflects a systematic positive bias from spatial scale mismatch between the ground sensor field of view and satellite pixel dimensions. The method reliably distinguishes between clear, partially cloudy, and overcast conditions, though the derived cloud fraction values exhibit clustering due to the temperature-ratio approach used. Exploratory comparison with TROPOMI aerosol index products yielded negligible correlation due to the absence of UV spectral coverage in the ground sensor, identifying a clear instrumentation requirement for future aerosol validation work.

Temporal matching between satellite overpasses and ground observations achieved a mean time difference of 2.7 minutes, with 95 % of matches within 8 minutes of satellite observation time. Spatial co-location analysis confirms all validation points fall within the nominal TROPOMI pixel footprint (3.5 km × 5.5 km at nadir), though the spatial scale mismatch between the ground sensor field of view and satellite pixel dimensions remains the primary source of validation uncertainty.

Our results demonstrate that low-cost infrared sensors, when operated with calibration-informed measurement protocols, can provide scientifically useful satellite cloud product screening data, reliably distinguishing between clear, partially cloudy, and overcast conditions. The quasi-discrete nature of the derived cloud fraction highlights the need for improved cloud detection algorithms in future work. This approach offers a scalable pathway for expanding ground-based validation networks in regions lacking dedicated atmospheric monitoring infrastructure.