The spatial characteristics at mesoscale ? of the strongest sea surface winds blowing on the Adriatic Sea and Venice Lagoon, Italy, have been investigated through 500 m ×500 m wind fields extracted, without external information, from the C-band Sentinel-1 synthetic aperture radar (SAR) images interferometric wide (IW) swath ground range detected high resolution (GRDH) level 1 (L1) by a deep learning method based on a residual neural network (ResNet). The derived SAR wind directions have been used to retrieve the wind speed using the C_SARMOD2 geophysical model function (GMF). The resulting wind fields have been compared with European Centre for Medium-range Weather Forecasts (ECMWF) and in situ data: the wind direction bias |??|<= 1° indicates a very good agreement and the value of RMS difference against in situ winds of 16° results smaller than those reported in the literature; the wind speed bias ?w=-1.5ms-1 with respect to in situ data indicates an underestimation of SAR speed. However, the innovation brought by ResNet results is evident in the study of the spatial structure of the wind fields, showing unprecedented details. The spatial gradients of wind direction and speed obtained (? 6°/km and 0.5ms-1/km, respectively) are well consistent with those derived from previous experimental campaigns. The maps of the atmospheric wind stress curl derived from the ResNet wind fields exhibit the presence of contiguous cells with positive and negative values roughly elongated in the wind direction, not reproducible with such profusion of details by any other existing datasets. ResNet wind fields provide exhaustive coverage of the Venice lagoon, allowing to study their spatial structure: under northeasterly winds, the wind speed increases on average from northern to southern lagoon by 30% in agreement with a case study reported in the literature. Local differences of ResNet wind direction with respect to ECMWF (±30° maximum) generate local differences of wind speed as large as ±pm 2ms-1.
Structure of High-Resolution SAR Winds Over the Venice Lagoon Area
Zecchetto S;
2022
Abstract
The spatial characteristics at mesoscale ? of the strongest sea surface winds blowing on the Adriatic Sea and Venice Lagoon, Italy, have been investigated through 500 m ×500 m wind fields extracted, without external information, from the C-band Sentinel-1 synthetic aperture radar (SAR) images interferometric wide (IW) swath ground range detected high resolution (GRDH) level 1 (L1) by a deep learning method based on a residual neural network (ResNet). The derived SAR wind directions have been used to retrieve the wind speed using the C_SARMOD2 geophysical model function (GMF). The resulting wind fields have been compared with European Centre for Medium-range Weather Forecasts (ECMWF) and in situ data: the wind direction bias |??|<= 1° indicates a very good agreement and the value of RMS difference against in situ winds of 16° results smaller than those reported in the literature; the wind speed bias ?w=-1.5ms-1 with respect to in situ data indicates an underestimation of SAR speed. However, the innovation brought by ResNet results is evident in the study of the spatial structure of the wind fields, showing unprecedented details. The spatial gradients of wind direction and speed obtained (? 6°/km and 0.5ms-1/km, respectively) are well consistent with those derived from previous experimental campaigns. The maps of the atmospheric wind stress curl derived from the ResNet wind fields exhibit the presence of contiguous cells with positive and negative values roughly elongated in the wind direction, not reproducible with such profusion of details by any other existing datasets. ResNet wind fields provide exhaustive coverage of the Venice lagoon, allowing to study their spatial structure: under northeasterly winds, the wind speed increases on average from northern to southern lagoon by 30% in agreement with a case study reported in the literature. Local differences of ResNet wind direction with respect to ECMWF (±30° maximum) generate local differences of wind speed as large as ±pm 2ms-1.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
