We present an empirical C-band Cross-Polarization Ocean (C-2PO) model for wind retrievals from synthetic aperture radar (SAR) data collected by the RADARSAT-2 satellite. The C-2PO model relates normalized radar cross section (NRCS) in cross polarization to wind speed at 10-m height. This wind retrieval model has the characteristic that it is independent of wind direction and radar incidence angle but is quite linear with respect to wind speed. To evaluate the accuracy of the proposed model, winds with a resolution on the scale of 1 km were retrieved from a dual-polarization SAR image of Hurricane Earl on 2 September 2010, using the C-2PO model and compared with CMOD5.N, the newest available C-band geophysical model function (GMF), and validated with collocated airborne stepped-frequency microwave radiometer measurements and National Data Buoy Center data. Results suggest that for winds up to 38 m s−1, C-2PO has a bias of −0.89 m s−1 and a root-meansquare error of 3.23 m s−1 compared to CMOD5.N, which has a bias of −4.14 m s−1 and an rms difference of 6.24 m s−1. Similar results are obtained from Hurricane Ike, comparing wind retrievals from C-2PO and CMOD5.N with H*Wind data. The advantage of C-2PO over CMOD5.N and other GMFs is that it does not need any external wind direction and radar incidence angle inputs. Moreover, in the presently available quad-polarization dataset, C-2PO has the feature that the cross-polarized NRCS linearly increases even for wind speeds up to 26 m s−1 and reproduces the hurricane eye structure well, thereby providing a potential technique for hurricane observations from space.

A large collaborative program has studied the coupled air‐ice‐ocean‐wave processes occurring in the Arctic during the autumn ice advance. The program included a field campaign in the western Arctic during the autumn of 2015, with in situ data collection and both aerial and satellite remote sensing. Many of the analyses have focused on using and improving forecast models. Summarizing and synthesizing the results from a series of separate papers, the overall view is of an Arctic shifting to a more seasonal system. The dramatic increase in open water extent and duration in the autumn means that large surface waves and significant surface heat fluxes are now common. When refreezing finally does occur, it is a highly variable process in space and time. Wind and wave events drive episodic advances and retreats of the ice edge, with associated variations in sea ice formation types (e.g., pancakes, nilas). This variability becomes imprinted on the winter ice cover, which in turn affects the melt season the following year.

A new geophysical model function (GMF), called C_SARMOD2, has been developed to relate high-resolution C-band normalized radar cross section (NRCS), acquired in VV polarization over the ocean, to the 10-m height wind speed. A total of 3078 RADARSAT-2 and Sentinel-1A VV-polarized synthetic aperture radar (SAR) images, acquired under different wind speed conditions, were collocated with in situ buoy measurements. The paired dataset was used to derive transfer functions and coefficients of C_SARMOD2, and then to validate the wind speed retrievals. The comparison between SAR-retrieved wind speeds and buoy measurements show almost no bias and a root mean square error of 1.84 m/s. Two representative quad- and dual-polarization SAR images, acquired from coastal regions, are used as case studies to examine C_SARMOD2 performances. The case study and statistical validation results suggest that the proposed C_SARMOD2 has the potential to measure coastal wind speeds at subkilometer resolutions. Although derived from low resolution NRCS measurements, this study also confirms the great robustness of CMOD5.N and recent CMOD7 when applied to SAR data. In addition, it shows that with the new generation of SAR satellite-borne sensors, it is no longer mandatory to rely on scatterometers in order to build a GMF, which will be used for SAR applications. Such an approach is particularly important in view of the upcoming RADARSAT Constellation Mission with new polarization configurations. Moreover, it also opens new perspectives on the derivation of GMFs in HH-polarization. However, these results also suggest that for coastal areas, the increase of the resolution to define the GMF is less important than adding other geophysical parameters to improve wind retrieval performance. This advocates for the necessity of revisiting the methodologies for ocean surface wind speed measurements in coastal areas.

The Data Interpolating Empirical Orthogonal Functions (DINEOF) method has demonstrated usability and accuracy for filling spatial gaps in remote sensing datasets. In this study, we conducted the reconstruction of the chlorophyll-a concentration (Chl-a) data using a convolutional neural networks model called Data-Interpolating Convolutional Auto-Encoder (DINCAE), and we compared its performance with that of DINEOF. Furthermore, the cloud-free sea surface temperature (SST) was used as a phytoplankton dynamics predictor for the Chl-a reconstruction. Finally, four reconstruction schemes were implemented: DINCAE (Chl-a only), DINCAE (Chl-a and SST), DINEOF (Chl-a only), and DINEOF (Chl-a and SST), denoted rec1, rec2, rec3, and rec4 respectively. To quantitatively evaluate the accuracy of these reconstruction schemes, both the cross-validation and in situ data were used. The study domain was chosen to be the Northern South China Sea (SCS) and West Philippine Sea (WPS), bounded by 115–125°E and 16–24°N to test the model performance for the reconstruction of Chl-a under different Chl-a controlling mechanisms. The in situ validation showed that rec1 performs best among the four reconstruction schemes, and that adding SST into the Chl-a reconstruction cannot improve the reconstruction results. However, for cross validation, adding SST can slightly improve spatial distributions of the root mean square error (RMSE) between the reconstructed data and the original data, especially over the SCS continental shelf. Furthermore, the potential of DINCAE prediction is confirmed in this paper; thus, the trained DINCAE model can be re-applied to reconstruct other missing data, and more importantly, it can also be re-trained using the reconstructed data, thereby further improving reconstruction results. Another consideration is efficiency; with similar reconstruction conditions, DINCAE is 5–10 times faster than DINEOF.