ISRAMAR HaderaANN System - Artificial neural network for wave forecast

Artificial neural network (ANN) is a computing system inspired by the biological neural networks; in other words, it does not have a preset fixed algorithm to give the desired output. Instead, the ANN system must be “trained” to reach a state in which its inner parameters are such that it gives the desired result. Such a “training” is conducted by presenting the system with examples of both the kind of input that will be used for prediction and its matching desired output.

The IOLR HaderaANN system is intended to forecast the wave height at Hadera Meteomarine Monitoring station in the middle of Israel’s Mediterranean coast. The system gives 24 hours forecast every hour. The calculation of the forecast takes less than 30 seconds. The system is innovative in using several types of data as inputs. This includes wind forecast data given by the SKIRON atmospheric modeling system, wave forecast for the station's location given by the SWAN wave model, and observed wave data the station collects. Whereas wave observation and forecast inputs are entered as simple time series, wind input is entered over a grid four-dimensional grid. This grid covers a square with a side length of 3.25° (approximately 325 km).

As the forecast changes hourly with new observations, previous predictions are collected to estimate the prediction range. The observations and SWAN forecast are also presented on the chart as a reference.

 The system uses SKIRON wind fields at 10 m above sea surface with a resolution of 0.05° x 0.05° provided by The University of Athens's Atmospheric Modeling and Weather Forecasting Group (Regional meteorological model SKIRON). The system was developed with the TensorFlow package for Python.

Go back