Abstract's details
Wet Tropospheric Correction for Sentinel-3: a better tuned retrieval algorithm for open-ocean
CoAuthors
Event: 2020 Ocean Surface Topography Science Team Meeting (virtual)
Session: Instrument Processing: Propagation, Wind Speed and Sea State Bias
Presentation type: Type Forum only
Contribution: PDF file
Abstract:
Sentinel-3 (S3) mission is currently composed of two operational satellites (S3A and S3B). To determine the Wet Tropospheric Correction (WTC) of the S3 altimeter observations, each satellite possesses a two-band (23.8 and 36.5 GHz) Microwave Radiometer (MWR). Making use of algorithms firstly designed for EnviSat mission, S3 products provide two different MWR-derived WTC: computed from 3 and 5 inputs. Since the MWR on board Sentinel-3 does not possess a third band near 18 GHz to account for the surface contribution in the MWR measurements, these algorithms allow to overcome this instrumental limitation. They use additional inputs able to include this contribution of the sea surface, such as the altimeter σ0 (describing the changes in the sea surface emissivity due to wind-induced sea surface roughness) and sea surface temperature.
Based on these current algorithms adopted in the Sentinel-3 data records, this study describes an improved algorithm for the WTC retrieval over open ocean, better tuned for Sentinel-3. This is a purely empirical algorithm, based on a simple neural network only with 4 inputs: brightness temperatures at 23.8 and 36.5 GHz, altimeter backscatter coefficient and sea surface temperature (SST) interpolated from ERA5. The learning database has been established using one year of valid S3A measurements and WTC computed from ERA5 fields.
Comparisons with independent and stable WTC sources show that the WTC derived from the proposed algorithm instead of those available in the S3 products leads to a decrease in the RMS values of the WTC differences with respect to the reference WTC by about 1 mm globally, while this decrease can reach almost 1 cm over some regions. These results are more pronounced for distances from coast between 30 and 250 km, where the improvement of this algorithm over those adopted in Sentinel-3 products is globally almost 3 mm. These improvements are mainly attributed to two different factors. First, this is an algorithm originally tuned for Sentinel-3 mission, considering a suitable learning. Second, the contribution of the surface in the MWR measurements is better accounted for by means of SST interpolated from ERA5 (instead of seasonal tables as adopted in S3 products), which introduces additional information on the surface contribution in the MWR measurements, becoming the fifth input (the atmospheric temperature lapse rate) unnecessary.
Originally designed for Sentinel-3, this study proposes a better tuned WTC retrieval algorithm, than those currently used in the S3 products (firstly developed for EnviSat mission).
Based on these current algorithms adopted in the Sentinel-3 data records, this study describes an improved algorithm for the WTC retrieval over open ocean, better tuned for Sentinel-3. This is a purely empirical algorithm, based on a simple neural network only with 4 inputs: brightness temperatures at 23.8 and 36.5 GHz, altimeter backscatter coefficient and sea surface temperature (SST) interpolated from ERA5. The learning database has been established using one year of valid S3A measurements and WTC computed from ERA5 fields.
Comparisons with independent and stable WTC sources show that the WTC derived from the proposed algorithm instead of those available in the S3 products leads to a decrease in the RMS values of the WTC differences with respect to the reference WTC by about 1 mm globally, while this decrease can reach almost 1 cm over some regions. These results are more pronounced for distances from coast between 30 and 250 km, where the improvement of this algorithm over those adopted in Sentinel-3 products is globally almost 3 mm. These improvements are mainly attributed to two different factors. First, this is an algorithm originally tuned for Sentinel-3 mission, considering a suitable learning. Second, the contribution of the surface in the MWR measurements is better accounted for by means of SST interpolated from ERA5 (instead of seasonal tables as adopted in S3 products), which introduces additional information on the surface contribution in the MWR measurements, becoming the fifth input (the atmospheric temperature lapse rate) unnecessary.
Originally designed for Sentinel-3, this study proposes a better tuned WTC retrieval algorithm, than those currently used in the S3 products (firstly developed for EnviSat mission).