<p>Atmospheric events are driven by surface sea physical parameters, including the exchanges of water vapor with the overlying atmosphere. Oceans cover around 70 percent of the Earth's surface and influence the atmospheric circulation, causing some of the main weather events. The lack of surface observations over the vast ocean areas is a critical problem to be addressed for improving the performance of weather forecasting.</p><p>Even if weather observations over sea from ships have been collected for over 200 years and used for meteorological research and climate applications, only recently the availability of different telecommunication solutions make real time access to measurements possible, even from remote areas. This is consequently opening new opportunities to use data from marine areas in operational weather applications.</p><p>Ground based GNSS receivers has been used for many years to determine a quantity that is of major interest for meteorologists and climatologists, the water vapor content, derived from the Zenith Path Delay. GNSS meteorology has been also tested over ships during some measurement campaigns in the past.</p><p>This work presents the implementation of the first GNSS meteo infrastructure on ships operating on the northwestern Mediterranean Sea, involving 9 commercial vessels, real-time collecting a list of GNSS meteo parameters: the signals from Galileo, GPS, GLONASS and Beidou constellations, measurements of pressure, temperature, humidity, wind and precipitation. These 9 moving platforms are complemented by a number of fixed ground platforms, used as a reference.</p><p>The difficulties in ship based GNSS meteorology, with respect to the classical approaches from fixed stations, lie both in the exposure of the hardware instruments to challenging environmental conditions as in the open sea and in the computation algorithms, which must be applied to kinematic conditions and continuously solve the receiver position with very high accuracy.</p><p>Two different processing schemes have been applied to the dataset (i.e. few months): the first one is based on differential GNSS using the TRACK suite of GAMIT software, and the second one is based on precise point positioning using the GLAB software. As it is well known, if network solutions are adopted (as in the first case), the satellites and receivers clock errors can be eliminated with very high accuracy, while PPP-based methods (as in the second case) require ultrafast precise satellite ephemeris products, but they give the possibility to implement standalone instruments, so not to send large amounts of full RINEX files to a ground processing centre.</p><p>The ZPD quantities retrieved from the first period of observations aboard ships are shown, using both the techniques. The comparison shows some discrepancies both in the absolute quantity and in the short-term trends. Even if preliminary, the comprehension of the quality of such an unprecedent source of information is of great interest, because the perspectives of this infrastructure are both scientific and operational, thinking for example to the data assimilation into numerical weather prediction models.</p>
DSNet: an efficient CNN for road scene segmentation
