Monitoring and assimilation tests with TROPOMI data in the CAMS system: near-real-time total column ozone

Antje Inness, Johannes Flemming, Klaus-Peter Heue, Christophe Lerot, Diego Loyola, Roberto Ribas, Pieter Valks, Michel Van Roozendaël, Jian Xu, Walter ZimmerAtmospheric 2019 · DOI

Abstract

The TROPOspheric Monitoring Instrument (TROPOMI) on the Sentinel-5 Precursor satellite provides high-resolution near-real-time (NRT) total column ozone (TCO3) data. This study evaluates the integration of TROPOMI data into the Copernicus Atmosphere Monitoring Service (CAMS) from November 2017 to November 2018. While TROPOMI data generally agree within 2% of existing sensors between 60°N and 60°S, significant anomalies appear at high latitudes due to surface albedo climatology limitations. Despite these challenges, TROPOMI enhances system resilience and improves tropical tropospheric ozone estimates.

The Challenge of Global Ozone Monitoring

Monitoring global ozone is critical for tracking stratospheric health and understanding tropospheric air quality. CAMS produces daily global forecasts by assimilating satellite observations into numerical models. With aging instruments like OMI and GOME-2, TROPOMI represents a major advancement, offering a horizontal resolution of 3.5 km × 7 km—roughly ten times the density of previous sensors. This study validates whether this high-volume data stream is accurate enough for operational use and assesses its impact on an already well-constrained analysis system.

Methodology: 4D-Var and Super-obbing

The researchers utilized the ECMWF Integrated Forecast System (IFS) with a 4D-Var data assimilation scheme. Because TROPOMI’s resolution is significantly finer than the model’s ~40 km grid, a "super-obbing" technique was employed. This process averages high-quality TROPOMI pixels within a model grid box to create a representative observation, reducing computational load and random noise.

Comparison of high-resolution TROPOMI TCO3 data (a) and the super-obbed data (b) used for assimilation into the CAMS model grid.

The study involved two phases: passive monitoring to identify biases without influencing the model, and active assimilation to update the model state and compare results against ground-truth data from Brewer spectrometers, ozone sondes, and IAGOS aircraft profiles.

Key Findings: Accuracy and Anomalies

Globally, TROPOMI TCO3 shows excellent agreement (within 2%) with CAMS analysis and other sensors between 60°N and 60°S. However, substantial errors—up to 40 Dobson Units—were identified at high latitudes during winter and spring. These anomalies are primarily attributed to the coarse surface albedo climatology used in the NRT retrieval, which struggles to distinguish between snow-covered and dark land pixels at TROPOMI's high resolution.

Latitude-time distribution of the differences between TROPOMI and GOME-2B TCO3, illustrating seasonal biases at high latitudes.

Additionally, a slight dependency on the satellite's scan position (cross-track bias) was noted, a common characteristic of wide-swath sensors that requires ongoing calibration.

Impact on CAMS Analysis

Active assimilation of TROPOMI data led to a small but consistent improvement in the fit to ground-based Brewer data. The most significant vertical impact was observed in the tropical troposphere, where TROPOMI helped reduce model biases compared to aircraft profiles over West Africa. While the absolute change in accuracy was modest (less than 2%) due to the redundancy of existing satellite data, TROPOMI significantly increases the system's resilience against the potential failure of older instruments.

Future Directions

To address current limitations, future work will focus on integrating the "Offline" TROPOMI product, which fits albedo during the retrieval process and avoids high-latitude anomalies. Updates to the NRT algorithm will include higher-resolution surface albedo maps and the implementation of averaging kernels to better account for vertical sensitivity during assimilation.

Reference: Inness, A., et al. (2019). Monitoring and assimilation tests with TROPOMI data in the CAMS system: near-real-time total column ozone. Atmos. Chem. Phys., 19, 3939–3962.

Summary updated

Abstract

The TROPOspheric Monitoring Instrument (TROPOMI) on board the Sentinel-5 Precursor (S5P) satellite launched in October 2017 yields a wealth of atmospheric composition data, including retrievals of total column ozone (TCO3) that are provided in near-real-time (NRT) and off-line. The NRT TCO3 retrievals (v1.0.0–v1.1.2) have been included in the data assimilation system of the Copernicus Atmosphere Monitoring Service (CAMS), and tests to monitor the data and to carry out first assimilation experiments with them have been performed for the period 26 November 2017 to 30 November 2018. The TROPOMI TCO3 data agree to within 2 % with the CAMS analysis over large parts of the globe between 60∘ N and 60∘ S and also with TCO3 retrievals from the Ozone Monitoring Instrument (OMI) and the Global Ozone Monitoring Experiment-2 (GOME-2) that are routinely assimilated by CAMS. However, the TCO3 NRT data from TROPOMI show some retrieval anomalies at high latitudes, at low solar elevations and over snow/ice (e.g. Antarctica and snow-covered land areas in the Northern Hemisphere), where the differences with the CAMS analysis and the other data sets are larger. These differences are particularly pronounced over land in the NH during winter and spring (when they can reach up to 40 DU) and come mainly from the surface albedo climatology that is used in the NRT TROPOMI TCO3 retrieval. This climatology has a coarser horizontal resolution than the TROPOMI TCO3 data, which leads to problems in areas where there are large changes in reflectivity from pixel to pixel, e.g. pixels covered by snow/ice or not. The differences between TROPOMI and the CAMS analysis also show some dependency on scan position. The assimilation of TROPOMI TCO3 has been tested in the CAMS system for data between 60∘ N and 60∘ S and for solar elevations greater than 10∘ and is found to have a small positive impact on the ozone analysis compared to Brewer TCO3 data and an improved fit to ozone sondes in the tropical troposphere and to IAGOS aircraft profiles at West African airports. The impact of the TROPOMI data is relatively small because the CAMS analysis is already well constrained by several other ozone retrievals that are routinely assimilated. When averaged over the periods February–April and September–October 2018, differences between experiments with and without assimilation of TROPOMI data are less than 2 % for TCO3 and less than 3 % in the vertical for seasonal mean zonal mean O3 mixing ratios, with the largest relative differences found in the troposphere.
BibTeX
@inproceedings{inness2019monitoring,
  author = {Antje Inness and Johannes Flemming and Klaus-Peter Heue and Christophe Lerot and Diego Loyola and Roberto Ribas and Pieter Valks and Michel Van Roozendaël and Jian Xu and Walter Zimmer},
  title = {{Monitoring and assimilation tests with TROPOMI data in the CAMS system: near-real-time total column ozone}},
  booktitle = {Atmospheric},
  year = {2019},
  doi = {10.5194/acp-19-3939-2019},
}
Figure