Data di Pubblicazione:
2021
Abstract:
The Venice Lagoon (45.15°-45.6° N and 12.1°-12.8° E) ,
located in the northern part of the Adriatic Sea (Fig. 1), northeastern
part of Italy, extends ~50 km by 10 km along a North-
East axis. The lagoon is subjected to the main wind systems
of the northern Adriatic Sea, i.e. the south-easterly Sirocco
and the north-easterly Bora, the former blowing roughly
parallel to the Adriatic Sea major axis producing the well
known storm surges over the city, the latter blowing roughly
parallel to the lagoon major axis producing local storm surges
inside the lagoon.
SAR is the only instrument at present available to retrieve the
wind field over such small areas at a spatial resolution below
1 km. This is possible once robust methodologies to extract
the wind direction from SAR images, without external
informations, are available. In this context, the use of wind
directions from atmospheric models should be avoided
because of their well know difficulty to account for the small
spatial scale variations of the wind in coastal and semienclosed
areas as the the Venice lagoon. A technique based
on deep residual network (ResNet), a variant of Convolution
Neural Network (Zanchetta and Zecchetto, 2021) is, at present, the best approach to the determination of the
wind direction necessary to compute the wind speed through the available Geophysical Model Functions.
The results of the wind fields retrieval by our ResNet technique on 48 Sentinel-1 SAR images are shown in
terms of statistics from comparisons with in-situ data and description of the wind spatial structure typical of
the above mentioned winds (Fig. 2). Comparisons with the ESA OCN wind products are also provided, to
stress the importance of our results which do not depend on external direction and provide a better spatial
coverage of the Venice Lagoon and of the coastal areas.
Tipologia CRIS:
04.03 Poster in Atti di convegno
Keywords:
Deep residual network; Synthetic Aperture Radar (SAR); Wind field; Venice lagoon
Elenco autori:
Zecchetto, Stefano
Link alla scheda completa: