Abstract:
The area of the northern Adriatic Sea has a high productivity rate regarding the fishing activities and it is recognised as one of the most exploited areas of the Mediterranean Sea. In order to make fishing activities sustainable and to guarantee a productive and healthy ecosystem, there is a strong need to develop effective fishery management plans for constant monitoring and for prediction. For this reason, it is of fundamental importance to analyse the data relating to the movements of fishing vessels and their catch.
Starting from the AIS (Automatic Identification System) data, we reconstruct and enrich the trajectories by assigning to each segment the activity carried out by the boat (in port, entering/exiting from the port, navigation and fishing). In this way, considering only the portions of the trajectory in which the vessel is fishing, we compute the fishing effort, an essential indicator for monitoring the fishing pressure on an area of interest over time. By enriching our initial dataset with daily environmental factors such as sea temperature, waves height, wind and salinity, a variety of prediction methods are used in order to assess their prediction ability related to the fishing effort.
Concerning the Northern Adriatic Sea exploitation, another fundamental aspect is the underwater noise generated by vessels, which has a significant short and long term impact on animal species. Using AIS data and the characteristics of the boats, we build a model for the propagation of underwater noise based not only on the technical characteristics of the fishing boats' engine but also on environmental factors that vary in each season. In this way we get a map of the underwater noise useful for identifying areas where underwater noise can damage the marine environment, even permanently.
To accomplish this project we use MobilityDB, an extension of PostgreSQL and PostGIS that allows the storage and analysis of space-time objects. We also use Python and in particular Scikit-learn library to perform the experiments using machine learning models for regression. Furthermore, using the QGIS open source software, we create maps to display, in a simple and clear way, the areas in the Adriatic Sea characterised by the most intense noise pollution and to view the fishing effort forecasts.
