Image - Press Release - Sentinel-3 - 19/02/2016

Altimeter 1D-VAR Tropospheric Correction (AMTROC)

Image - Press Release - Sentinel-3 - 19/02/2016
Image - Press Release - Sentinel-3 - 19/02/2016

This study aimed to implement a novel method to derive Total Column Water Vapour (TCWV) and a Wet Tropospheric Correction (WTC) from observations from the Microwave Radiometer (MWR) instrument on-board the Sentinel-3 series of satellites.

Last Updated

14, November 2020

Long-term observations of trends in Sea Surface Height (SSH), as well as Total Column Water Vapour (TCWV), are critical for understanding the impacts and risks of climate change. In particular, changes in SSH are of major societal importance globally, as well as regionally. The Global Climate Observing System (GCOS) has identified both SSH and TCWV as essential climate variables (ECVs).

Atmospheric water vapour increases the refraction index of air, leading to a reduction in the speed of light as compared to dry air. The accuracy of SSH estimates from radar altimetry, therefore, depends strongly on the so-called 'wet tropospheric correction' (WTC), aimed at eliminating the temporally and spatially varying impact of the atmospheric water vapour on SSH retrievals. In fact, the spatial and temporal variability of water vapour is such that an instantaneous estimation of its impact is needed to meet the SSH accuracy requirements. Consequently, the primary role of the nadir-looking Microwave Radiometer (MWR), of the Sentinel-3 altimetry missions, is to provide the necessary water vapour observations.

In addition to WTC, TCWV is a highly important climate variable in its own right. The atmospheric water vapour feedback is believed to be the strongest feedback mechanism in climate change, approximately doubling the direct warming impact of increased CO2 forcing [Cess et al., 1990; Forster et al., 2007].

In the frame of the AMTROC study, a novel method has been implemented to derive TCWV and WTC, together with their respective uncertainties, from observations of the MWR instrument above the global ice-free oceans. The method was then applied to 10 months of continuous Sentinel-3A observations between 15 June 2016 and 15 April 2017.

Altimeter 1D-VAR Tropospheric Correction (AMTROC) - TCWV climatology
Figure 1: TCWV climatology (monthly means for 3°×3° areas), covering the period 06/2016 to 04/2017, derived from the MWR instruments.


The overall objective of this study was to contribute to enhancing the scientific quality of the Level 2 TCWV and WTC products to be derived from observations by the Microwave Radiometer (MWR) instruments onboard the Sentinel-3 series of satellites.


Altimeter 1D-VAR Tropospheric Correction (AMTROC) - differences
Figure 2: Difference between 1D-VAR and ERA-Interim reanalysis (red) as well as ANN and ERA-Interim reanalysis (blue) as a function of ERA TCWV (left) and wind speed (right) for a subset of 8,000 randomly selected MWR data points.

The primary study goal was to implement a 1D-VAR method to provide accurate and robust TCWV and WTC retrievals from Sentinel-3A MWR observations, building on recent achievements such as the EMiR project, funded by the European Space Agency. The results with this new method have been compared with the Sentinel-3 operational products obtained with a semi-empirical method based on neural networks (ANN). The following results were obtained:

  1. First estimation of the absolute brightness temperature bias in the two MWR channels at 23.8 and 36.5 GHz as provided by the reprocessed Sentinel-3A operational L2 product, by comparison against the corresponding channels of the better calibrated Advanced Technology Microwave Sounder (ATMS) instrument on the Suomi NPP satellite mission.
  2. Implementation of the 1D-VAR algorithm for TCWV and WTC from MWR observations of the Sentinel-3A satellite for the period 15 June 2016 to 15 April 2017, together with the corresponding per-footprint uncertainties.
  3. Provision of flags to allow for masking unreliable (e.g. rain-affected) retrievals.
  4. Validation of bias corrected brightness temperatures as well as TCWV and WTC retrievals by comparison against independent observations, as well as through an analysis of sea surface height variances at cross-over points. The results of 1D-VAR and ANN are very close to each other with the 1D-VAR performing slightly better when compared against reanalysis. In addition, 1D-VAR shows a reduced dependency on wind speed compared to ANN.
  5. Computation of retrieval errors. This is one of the main advantages of the 1D-VAR compared to neural networks method.
  6. Provision of the retrieval software code (in python) together with the accompanying documentation.

Further details of these results can be found in the Study Documents listed below.

Other Copernicus algorithm studies

Other Copernicus algorithm...