The increasing demand for cleaner and more efficient refining processes has driven the development of advanced upgrading technologies for heavy crude residues. This study investigates a novel Visbreaking-Supercritical Fluid Extraction (SFE) approach to upgrade the Merey vacuum residue (VR), integrating experimental analysis with molecular dynamics (MD) simulations for atomic-level mechanism exploration. The Visbreaking process is optimized at 400 °C for 40 min, achieving a viscosity reduction of 89.0 % while minimizing coke formation. The SFE process fractionates the visbroken VR, with total extraction yields ranging from 70.1 wt% to 70.7 wt%, demonstrating remarkable efficiency. Higher extraction pressures enhance deasphalted oil (DAO) yield but compromise its quality with higher metal and sulfur contents, while lower temperatures improve extraction selectivity. The integrated process effectively removes Fe, Ni, V, and Na, with demetalization efficiencies exceeding 62 %, 75 %, and 95 %, respectively. Molecular dynamics simulations provide atomic-scale insights into solubility mechanisms, revealing that higher pressures and lower temperatures enhance solvent compatibility with lighter visbroken VR fractions. The extracted DAO meets marine fuel oil blending specifications, while raffinates show potential for bitumen production and modification. These findings highlight the Visbreaking-SFE combination as a promising and sustainable upgrading strategy for heavy crude residues.

This paper reports on consolidated ground-based validation results of the atmospheric NO2 data produced operationally since April 2018 by the TROPOspheric Monitoring Instrument (TROPOMI) on board of the ESA/EU Copernicus Sentinel-5 Precursor (S5P) satellite. Tropospheric, stratospheric, and total NO2 column data from S5P are compared to correlative measurements collected from, respectively, 19 Multi-Axis Differential Optical Absorption Spectroscopy (MAX-DOAS), 26 Network for the Detection of Atmospheric Composition Change (NDACC) Zenith-Scattered-Light DOAS (ZSL-DOAS), and 25 Pandonia Global Network (PGN)/Pandora instruments distributed globally. The validation methodology gives special care to minimizing mismatch errors due to imperfect spatiotemporal co-location of the satellite and correlative data, e.g. by using tailored observation operators to account for differences in smoothing and in sampling of atmospheric structures and variability and photochemical modelling to reduce diurnal cycle effects. Compared to the ground-based measurements, S5P data show, on average, (i) a negative bias for the tropospheric column data, of typically-23 % to-37 % in clean to slightly polluted conditions but reaching values as high as-51 % over highly polluted areas; (ii) a slight negative median difference for the stratospheric column data, of about-0:2 Pmolec cm-2, i.e. approx.-2 % in summer to-15 % in winter; and (iii) a bias ranging from zero to-50 % for the total column data, found to depend on the amplitude of the total NO2 column, with small to slightly positive bias values for columns below 6 Pmolec cm-2 and negative values above. The dispersion between S5P and correlative measurements contains mostly random components, which remain within mission requirements for the stratospheric column data (0.5 Pmolec cm-2) but exceed those for the tropospheric column data (0.7 Pmolec cm-2). While a part of the biases and dispersion may be due to representativeness differences such as different area averaging and measurement times, it is known that errors in the S5P tropospheric columns exist due to shortcomings in the (horizontally coarse) a priori profile representation in the TM5-MP chemical transport model used in the S5P retrieval and, to a lesser extent, to the treatment of cloud effects and aerosols. Although considerable differences (up to 2 Pmolec cm-2 and more) are observed at single ground-pixel level, the near-real-time (NRTI) and offline (OFFL) versions of the S5P NO2 operational data processor provide similar NO2 column values and validation results when globally averaged, with the NRTI values being on average 0.79 % larger than the OFFL values.