Pilot production of new Sentinel-3 LAND Thematic products
ESA and the Sentinel-3 Mission Performance Cluster (MPC) have recently developed dedicated delay-Doppler and Level-2 processing chains for the generation of new Sentinel-3 STM LAND level-2 products over (1) Inland Waters, (2) Sea-Ice, and (3) Land Ice areas. The objectives of these so-called Thematic Instrument Processing Facilities (T-IPF) are to improve the performances of the current Sentinel-3 STM LAND level-2 products, and address the specific needs of the user communities for these three continental surfaces.
As a first step, Test Data Set (TDS) were generated with the newly developed Thematic IPFs and delivered to the users for a first evaluation. Information related to these TDS is available here:
Since the preliminary evaluation of these TDS was positive (more information below), the Sentinel-3 STM LAND Thematic products are now routinely generated and made available to the users as “pilot products”, in parallel to the operational STM LAND level-2 products (Processing Baseline version of the current operational products is 2.72/1.49, respectively for S3A/S3B). These pilot products are freely available through the ESA Copernicus Data Hub. They can be found by selecting the product types “SR_2_LAN_HY”, “SR_2_LAN_SI”, “SR_2_LAN_LI” (respectively for Hydrology, Sea Ice and Land Ice) on SCI-HUB.
Compared to the current Sentinel-3 STM level-2 products, the format of the new T-IPFs level-2 products have significantly evolved. The document describing the new Sentinel-3 level-2 Product Data Format Specifications (PDFS) is available at this link.
The Product Notice summarising the recent evolutions of the Sentinel-3 STM LAND Processing Baseline is available at this link.
Users are invited to download and test these new Sentinel-3 LAND Thematic products. As presented below, the first algorithms implemented in the Thematic IPFs already provide a clear data quality improvement for the three surfaces. For any feedbacks or questions regarding this pilot products, please contact: email@example.com
By the end of 2022, the full Sentinel-3A and Sentinel-3B missions will start being reprocessed with the Thematic IPFs. Once the reprocessing will be finished, the Thematic products will be produced operationally in substitute to the current LAND level-2 products.
First major evolutions implemented in the Sentinel-3 Thematic IPFs
Sea Ice and Hydrology
The T-IPFs first versions already include major processing evolutions. Over sea-ice and hydrological surfaces, these evolutions are namely the zero-padding and hamming processing.
The objective of the zero-padding is to provide a finer sampling of the altimetry waveforms (Jensen et al., 1999; Dinardo et al., 2018). As implemented in the T-IPF, the sampling is improved by a 2 factor. The range resolution therefore decreases from ~47 cm to ~23.5 cm, and now the waveforms contain 256 samples, compared to 128 before. The zero-padding is particularly important over specular surfaces, such as sea-ice leads and hydrological targets, where waveform leading edges can have a very limited number of samples. By better sampling sharp leading edges, the accuracy and the precision of the surface topography derived from the altimetry waveforms is enhanced.
Hamming window processing
The azimuth impulse response of the Sentinel-3 unfocused SAR processing has side lobes with relatively high energy levels, -13 dB for the first side-lobe. For that reason, the unfocused SAR waveforms available in the current Sentinel-3 LAND Level-2 products can contain spurious energy peaks, originating from specular along-track targets outside the Doppler band footprint. By applying a hamming window to the burst of pulses, prior to the delay-Doppler processing, the energy of the azimuth impulse response secondary lobes decreases down to ~ -42 dB, removing almost entirely any spurious energy from the waveforms. The added-value of the hamming window is therefore mainly over specular surfaces, such as sea-ice and hydrological targets.
The Land Ice T-IPF first version includes the “extended-window processing” to optimize the generation of SAR mode waveforms over rough and/or steeply sloping surfaces. Over such surfaces, large tracker range variations generally occur within the ~2 seconds of integration time in SAR mode. Hence, after range migration operations, useful energy sampled in the Doppler looks can be moved out the 128-samples window used to stack the data. To compensate for this effect, the window analysis of the delay-Doppler stack is artificially extended, from 128 to 512 samples. This way, all potential useful energy remains kept in the delay-Doppler stack. After multi-looking operation, the SAR mode waveform is re-centred within a conventional 128-sample window analysis. This processing is already implemented in the CryoSat-2 Payload Data Ground Segment, and also documented in Aublanc et al. (2018).
In addition, the slope models used by the Land Ice T-IPF to relocate the measurement at Point Of Closest Approach (POCA) are improved compared to the versions used by the current LAND IPF. These surface slope models are precomputed from digital elevation models published by Helm et al. (2014), for both Antarctica and Greenland. Thanks to this update, a better accuracy is expected in the measurement relocation at POCA.
Preliminary evaluations of Sentinel-3 LAND Thematic products
Test Data Set have been released to the users and assessed by the Expert Support Laboratories of the Mission Performance Cluster. A first summary of the main results is presented below.
To assess the performances of the Sentinel-3 STM LAND Thematic products over sea ice, comparisons were made to the CryoSat-2 mission, also operating in SAR Mode over this surface. The delay-Doppler processing and freeboard calculation is supposed to be now equivalent in the two processing chains (respectively the Sentinel-3 Sea Ice Thematic IPF and the CryoSat-2 ICE Baseline-D processor). In particular, the zero-padding and hamming window are also implemented in the CryoSat-2 ESA processor. Therefore, the two missions shall provide consistent freeboard estimations. This was confirmed by the assessment of the Sea Ice Thematic products TDS, with data measured from December 19th 2019, to February 11st 2020. The following figure shows map of radar freeboard over the Arctic Ocean, followed by the corresponding histograms. The freeboard from the Sentinel-3 Thematic products (in blue) is now in much better agreement with CryoSat-2 (in red), compared to the current Sentinel-3 LAND products (in green).
Figure 1 The radar sea ice freeboard maps on the top for cycle 54, and the corresponding histograms on the bottom show the good agreements between the Sentinel-3 new Sea Ice thematic products and CryoSat-2. Their SAR processing are both endowed with zero-padding and Hamming, unlike the current S3A LAND products.
An assessment of the “extended window processing” was performed at Antarctica global scale. In this assessment, the measurement failure was determined based on the absence of clear waveform leading edge(s) in the window analysis. Using this diagnosis, the map below displays the percentage of valid waveforms recovered over the Antarctica continent in the Land Ice Sentinel-3 Thematic products, in comparison to the current operational LAND level-2 products. Results show data the coverage is clearly improved over the ice sheet margins as expected. In addition, an analysis performed over the Filchner-Ronne Ice Shelf demonstrates that previous data gap close to the grounding line are now filled with the Land Ice Thematic products.
Figure 2 The percentage of additional records retrieved by the Thematic IPF over Antarctica, relative to the existing S3A Land product, showing the improvement in coverage due to the extended window processing, especially around the coastal margin where topography is more complex. (b) and (c) Examples of data coverage of repeated cycles from the existing product (b) and the TDS (c) close to the grounding line of the Filchner-Ronne Ice Shelf, illustrating the additional data retrieved (yellow arrow) within this glaciologically important region.
A global assessment of the Water Surface Heigh (WSH) provided in the Hydrology Thematic products was performed over a large amount of worldwide hydrological targets (823 lakes with area between 1 and 5 km²). The noise level of the WSH was estimated by computing the topography variation over consecutive 20 Hz measurements. Results show that the noise level decreases of about 50 % in the new Hydrology Thematic products (7 cm vs 13 cm) as presented in the figure below.
Figure 3 Noise levels of the Water Surface Height represented in histograms, and computed over 1 full cycle acquired between June and July 2020, and over 823 worldwide lakes with area between 1 and 5 km². Noise level from the current operational Sentinel-3 LAND products (yellow) is compared to the new Sentinel-3 Hydrology Thematic products (blue)
This noise reduction is mainly achieved thanks to the implementation of the zero-padding processing, improving the range resolution of the measurements. Additionally, other preliminary analyses indicate that the application of the hamming window processing could allow to estimate the water surface level over a larger number of hydrological targets.
Aublanc, J., Moreau, T., Thibaut, P., Boy, F., Rémy, F., & Picot, N. (2018). Evaluation of SAR altimetry over the Antarctic ice sheet from CryoSat-2 acquisitions. Advances in Space Research, 62(6), 1307–1323. https://doi.org/10.1016/j.asr.2018.06.043
Dinardo, S., Fenoglio-Marc, L., Buchhaupt, C., Becker, M., Scharroo, R., Joana Fernandes, M. and Benveniste, J.: Coastal SAR and PLRM altimetry in German Bight and West Baltic Sea, Advances in Space Research, 62(6), 1371–1404, doi:10.1016/j.asr.2017.12.018, 2018.
Helm, V., Humbert, A., & Miller, H. (2014). Elevation and elevation change of Greenland and Antarctica derived from CryoSat-2. The Cryosphere, 8(4), 1539–1559. https://doi.org/https://doi.org/10.5194/tc-8-1539-2014
Jensen, J. R.: Radar altimeter gate tracking: theory and extension, IEEE Transactions on Geoscience and Remote Sensing, 37(2), 651–658, doi:10.1109/36.752182, 1999.