Impact of altimeter-buoy data pairing methods on the validation of Sentinel-3A coastal significant wave heights

Sea state information is critical for a broad range of human activities (e.g. shipping, marine energy, marine engineering) most of them being concentrated along the coastal zone. Satellite altimeter records of significant wave heights represent the largest source of sea state observations available to date. However, the quality of altimeter observations is downgraded in the coastal zone due to surface heterogeneity within the radar signal footprint. In the last decades, increasing efforts have been devoted to exploit altimeter observations closer to the coast, using new sensor technologies (e.g. Ka-band or Delay-Doppler radar altimeters), dedicated waveform retracking algorithms or post processing filtering techniques. One major difficulty that remains to assess the performance of coastal altimetry in the coastal zone is the reduced number of valid altimeter records and the increased sea state variability, which requires the development of new methods to pair and compare nearby altimeter and buoy data. In this study, we use a high-resolution numerical wave model implemented over the European coastal waters in order to characterize the spatial variability of sea states in the proximity of coastal in situ buoys. Areas of sea state similarity are defined from the computation of systematic and random errors between the time series simulated at the station and those of neighboring nodes. These areas are then used to compute buoy-altimeter matchup statistics and estimate altimeter errors with respect to the buoy data. Additional methods, based on the dynamic comparison between model results at the buoy location and at the altimeter ground measurement are also investigated. These different methods are used to assess the quality of Sentinel-3A 20Hz significant wave height acquisitions in both offshore and coastal waters. Three Sentinel-3A data flavours are compared: the pseudo low resolution mode acquisitions,the SAR mode acquisitions and the Low Resolution with Range Migration Correction (LR-RMC) acquisitions. Comparisons against in situ and high resolution model clearly demonstrate the improved performance of the SAR and LR-RMC data relative to PLRM data. In particular, the LR-RMC processing is shown to provide consistent SWH records within the 0-20km coastal strip (and as close as 1km from the coast), with average normalized bias = 2.4%, scatter index = 18.9% and correlation coefficient = 0.95.