Experimental AI prototype for biogechemical models and hydrodynamic oceanographic centers crop for prediction of variables and environmental impact.

This work develops the interaction of artificial intelligence algorithms with oceanographic models. For this, an experimental pilot was used to predict the time series of different variables that can generate enviromental impacts on the maritime ecosystem. It was necessary to rely on the oceanographic baseline study to obtain data in southern Chile and generate the study in a maritime concession. The processing of the data from Croco-Ocean oceanographic model was carried out through a transformation of the time series from the NetCDF to CSV forma, allowing a better manipulation of the data. The results obtained show a comparison of the performance of different neural networks: dense(DNN), convolutional (CNN) and recurrent (RNN), which allowed determining which method obtains better metrics in time series problems. In order to be able to monitor and predict the variables that affect the cultivation center.