Articles | Volume 17, issue 6
Nonlin. Processes Geophys., 17, 809–815, 2010
Nonlin. Processes Geophys., 17, 809–815, 2010

Research article 23 Dec 2010

Research article | 23 Dec 2010

Energy-based predictions in Lorenz system by a unified formalism and neural network modelling

A. Pasini1, R. Langone2, F. Maimone3, and V. Pelino3 A. Pasini et al.
  • 1CNR, Institute of Atmospheric Pollution Research, Rome, Italy
  • 2Katholieke Universiteit Leuven, Department ESAT/SISTA, Leuven, Belgium
  • 3Italian Air Force, CNMCA, Pratica di Mare (Rome), Italy

Abstract. In the framework of a unified formalism for Kolmogorov-Lorenz systems, predictions of times of regime transitions in the classical Lorenz model can be successfully achieved by considering orbits characterised by energy or Casimir maxima. However, little uncertainties in the starting energy usually lead to high uncertainties in the return energy, so precluding the chance of accurate multi-step forecasts. In this paper, the problem of obtaining good forecasts of maximum return energy is faced by means of a neural network model. The results of its application show promising results.