Prediction of mars meteorological variables using artificial neural networks

Weather forecasting is the task of determining future states of the atmosphere for a given location and time. The techniques to carry out the prediction range from deterministic approaches using complex fluid dynamics models to data-driven approaches using artificial intelligence. While the former is mainly focused on the creation of General Circulation Models, the later are starting to replace them in many situations for Earth’s meteorology and astrophysics. Here, we develop an artificial neural network to perform Mars’ weather forecasting using environmental measurements from the Vikings and Mars Science Laboratory missions. The methodology followed in of this study is a data-driven approach; we make use of computer science expertise which has been long applied to Earth, but not on Mars yet. To do so, we create an artificial neuronal network that predicts the meteorological conditions of the following day using the previous day as input. We show that temperature and pressure are among the most important variables, and that ANN can perform with a 0.5 to 1% accuracy in forecasting diurnal changes in the selected variables. Keywords: ANN, Mars, Weather, Forecast, Curiosity