In last decade great efforts have been made to develop models to predict the Dst index from solar wind data. The aim of these models is double: to know the physical mechanisms of RC dynamics and to study the influence of solar wind in the terrestrial environment. These models can be summarized into three groups: those based in a first order differential equation, those with linear filters [21] and those developed from neural networks [22]. In the first group outstands the work of [23], where the time evolution of Dst is modeled as a difference between an injection function, Q(t),and a recovery term of the RC with a characteristic time . The injection function is associated to the energy coming by reconnection. Although the expression for Q(t) has been discussed in many papers (see as an example [24]), it is accepted that
magnetospheric injection is directly related to dawn-dusk component of solar wind electric field [25, 9, 26]. On the other hand, the recovery term is associated to the decay due to any loss process in the RC and it is proportional to the own Dst index.