Technical Note: DACNO₂ – A Multi-Constraint Deep Learning Framework for High-Resolution 3D NO₂ Field Estimation

Abstract. Accurate, high-resolution 3D fields of nitrogen dioxide (NO₂) are critical for air quality management and satellite retrievals, yet traditional chemistry-transport models (CTMs) face challenges in fine-scale modeling. Machine learning (ML) alternatives often struggle with generalization and transferability, inheriting biases from CTMs or being limited by sparse surface measurements. We present the Deep Atmospheric Chemistry NO₂ model (DACNO₂), a deep learning model that generates daily 2 km 3D NO₂ fields over Western Europe. The model's three-phase and multi-constraint training strategy begins by pre-training on European Copernicus Atmosphere Monitoring Service (CAMS) reanalysis data to learn large-scale atmospheric patterns, then fine-tunes with both CAMS and in-situ European Environmental Agency (EEA) surface data to correct biases and refine local detail, and completes with an adaptive fine-tuning to capture evolving trends. An evaluation for 2023 shows that DACNO₂ reproduces broad-scale 3D CAMS fields (R² = 0.90) while improving agreement with independent EEA stations over the CAMS reanalysis (R² enhanced from 0.61 to 0.66; bias reduced from -1.15 to -0.38 µg/m³). The model resolves more spatial detail and learns physically interpretable relationships. This hybrid training approach fuses the physical consistency of a process-based model with the real-world accuracy of surface measurements, overcoming the limitations of using either constraint data alone. Applying DACNO₂ a-priori profiles to TROPOMI retrievals increases tropospheric NO₂ columns by 3 % on average over those using European CAMS profiles, with larger enhancements over emission hotspots. These results demonstrate the framework's potential to advance air quality monitoring and satellite remote sensing.