The TROPOspheric Monitoring Instrument (TROPOMI) provides high-resolution, global measurements of carbon monoxide (CO) for environmental pollution monitoring. The Automated Plume Detection and Emission Estimation algorithm (APE), developed by SRON, identifies pollution plumes and estimates emissions based on the satellite data. This study implemented four machine learning algorithms to enhance APE and applied them to 180 steel plant locations for 6 years to estimate the average emissions from detected plumes using the divergence method. The models identified up to 136.1% more plumes than APE. Comparing the estimated emissions with the European Pollutant Release and Transfer Register (E-PRTR) dataset shows the ResNet-44 model achieves the lowest bias (1.20 kg/s) and standard deviation (2.36 kg/s) compared to APE, which had a bias of 3.41 kg/s and a standard deviation of 2.40 kg/s. This demonstrates the potential of machine learning to improve plume detection and emission estimation for remote sensing of pollution from space.

This thesis investigates TROPOMI NO₂ observations over low clouds and fog. Such scenes present unique opportunities due to the increased reflectivity of clouds in the lower troposphere, which enhances sensor sensitivity. However, in order to accurately estimate vertical NO₂density in these scenarios, a better understanding of the retrieval process in such conditions is imperative. Two NO₂ retrieval case studies, both exhibiting a spurious retrieval in the presence of low clouds and fog above the North Sea, are examined: February 14𝑡ℎ, 2023, and April 9𝑡ℎ, 2023. The main focus was on the influence of the air-mass factor on the retrieval. This factor converts measured slant column data along the satellite path into vertical tropospheric columns. Spurious spatial patterns are strongly amplified after the application of the air-mass factor. Key dependencies influencing the conversion from slant to
vertical column are identified, being the the a-priori NO₂ profile, air mass factor calculation, and cloud characterization. Notably, the low spatial resolution of the a-priori model and its pronounced peak near the surface contribute to inflated NO₂ values over clouds. The majority of the a-priori NO₂ profile is simulated underneath the cloud field, leading the retrieval algorithm to assume it did not capture a significant portion of the concentration, thus strongly increasing retrieved NO₂  densities over clouds. Furthermore, spatial reflectivity patterns in the North Sea are not always accounted for in the surface reflectivity climatology. This results in misinterpreted clouds with a low cloud fraction. Misinterpreted clouds over open sea areas, combined with low cloud pressures, result in downward corrections over clear-sky seas. These effects culminate in an exacerbated cloudy-clear contrast in NO₂ densities, aligning with the outlines of low cloud fields. Additionally, the heterogeneous cloud field challenges the assumption of a fixed-height, fixed-albedo Lambertian reflector. This inflation effect is modulated by the cloud height. The uncertain height of the Lambertian cloud in the lowest troposphere, combined with an assumed NO₂ profile at the same height, dominates the air-mass factor calculation . Four recommendations to enhance NO₂ retrieval accuracy in the presence of low clouds and fog over the North Sea are presented, which include a Cloud-as-Layers characterization, high-resolution simulations and in-situ measurements of vertical NO₂ profiles in the presence of low clouds, an a-priori correction of the NO₂ profile in the presence of low clouds and implementing auxiliary data relating to spatial patterns in sea-colour of the North Sea.

Anthropogenic nitrogen dioxide (NO₂) in the troposphere is mainly produced by combustion engines in traffic, industry and energy production, and continues to affect air quality, human health and environment. Daily global measurements of tropospheric NO₂ columns are obtained by satellites with increasing spatial resolution and signal-to-noise levels, to improve monitoring of emission sources and air quality forecasting. The recently launched TROPOMI instrument on-board ESA’s Sentinel-5 Precursor satellite measures tropospheric NO₂ with a spatial resolution of 7.1 km by 3.6 km. During its commissioning phase, the instrument was temporarily operated in ’zoom-mode’ to measure at a resolution of 2.4 km by 1.8 km. This research presents the processed results from this unique data-set, which allows mapping NO₂ pollution sources from space with unprecedented detail. Comparison to measurements at operational resolution shows the improvement in spatial information content, at the cost of increased noise and uncertainty. The benefits and possibilities of measuring tropospheric NO₂ at high resolution are explored with several case studies. Comparisons with chemical transfer model forecasts show the improved ability of these measurements to capture local NO₂ enhancements and possibly improve emission inventories. The found correlations with co-located space-borne CO₂ column observations and the performance of a plume detection algorithm applied to the data-set provide additional support for simultaneous high resolution measurements of co-emitted CO₂ and NO₂, planned for future satellites to improve CO₂ emission monitoring. Finally, test retrievals with the zoom-mode data, using experimental high resolution surface reflectivity and NO₂ profile shape input, demonstrate the potential impact of high resolution a priori databases on the retrieval performance.