^{1}

^{2}

^{1}

^{2}

^{3}

^{4}

^{1}

^{2}

Earthquake prediction is a main topic in Seismology. Here, the goal is to know the correlation between the seismicity at a certain place at a given time with the seismicity at the same place, but at a following interval of time. There are no ways for exact predictions, but one can wonder about the causality relations between the seismic characteristics at a given time interval and another in a region. In this paper, a new approach to this kind of studies is presented. Tools which include cellular automata theory and Shannon's entropy are used. First, the catalogue is divided into time intervals, and the region into cells. The activity or inactivity of each cell at a certain time is described using an energy criterion; thus a pattern which evolves over time is given. The aim is to find the rules of the stochastic cellular automaton which best fits the evolution of the pattern. The neighborhood utilized is the cross template (CT). A grid search is made to choose the best model, being the mutual information between the different times the function to be maximized. This function depends on the size of the cells β on and the interval of time τ which is considered for studying the activity of a cell. With these β and τ, a set of probabilities which characterizes the evolution rules is calculated, giving a probabilistic approach to the spatiotemporal evolution of the region. The sample catalogue for the Iberian Peninsula covers since 1970 till 2001. The results point out that the seismic activity must be deduced not only from the past activity at the same region but also from its surrounding activity. The time and spatial highest interaction for the catalogue used are of around 3.3 years and 290x165 km<sup>2</sup>, respectively; if a cell is inactive, it will continue inactive with a high probability; an active cell has around the 60% probability of continuing active in the future. The Probabilistic Seismic Hazard Map obtained marks the main seismic active areas (northwestern Africa) were the real seismicity has been occurred after the date of the data set studied. Also, the Hurst exponent has been studied. The value calculated is 0.48±0.02, which means that the process is inherently unpredictable. This result can be related to the incapacity of the cellular automaton obtained of predicting sudden changes.