Coastal PenWP

This study presents a refinement of the assessment of QuikSCAT normalized radar cross sections (σ0 s) noise, together with an assessment of the intra-egg biases induced by the incidence angle variation and their impact on noise esti- mates. In addition, it describes a new implementation of Land Contribution Radio (LCR). Furthermore, it shows the implementation of an LCR-based σ0 correction scheme to reduce the impact of land contaminations. Finally, cor- rected σ0 s are integrated in the four pol-view σ0 s to be used in the retrieval step. Four different averaging methodologies are considered, which differen- tiate according to the weight assigned to each slice: homogeneous (boxcar), Kp dependent, lcr dependent, or Gaussian dependent on the retrieval error. The results confirm that the noise levels provided in the full resolution files have some significant differences with the estimates. In particular, exter- nal H-Pol acquisitions are noisier than expected, especially those closer to the antenna for medium-high wind regimes, while inner acquisitions are less noisy than reported in the files, especially those V-Pol. The impact of the intra-egg biases on the noise estimates is less than 7%, therefore all previous conclusions are valid. The correction of σ0 s shows that the new LCR method (LCRf ull ) seems to provide more suitable estimates of LCR. The correction method is effective in reducing the coastal σ0 values, even if many negative values occur. This trend is expected and the main reason is due to the high σ0 -dependent noise level that affects QuikSCAT measurements. A new formulation of the cor- rection scheme that considers an additional noise-dependent regularization parameter is recommended. Finally, integration methodologies show some differences, but all show some residual contamination near the shore. These preliminary results are not sufficient to select the most suitable one. It is rec- ommended to make this choice after the retrieval step is performed. The soft- ware structure is robust, but the LCR computation is very time-consuming. An optimization of the code is required after it is implemented in a suitable low-level programming language.