1. The algorithm SSE was designed for prediction of relatively large earthquakes following a strong earthquake. A subsequent strong earthquake can be an aftershock or a main shock of larger magnitude.

The SSE algorithm resulted from the analysis of the 21 case histories in California and Nevada. Then it was retrospectively tested in 8 seismic regions of the world.

Given a strong earthquake, i.e. an earthquake with magnitude M M₀, that occurred at time t, we aim at predicting the subsequent strong earthquake, which magnitude, time, and location depend on the parameters of the strong earthquake:

Consider a strong earthquake with magnitude M and occurrence time t. The problem is to predict whether a subsequent strong earthquake will occur.

To solve this problem we analyse the aftershocks of the first earthquake during the first 40 days in the magnitude range between M and M - 3, and the earthquakes which occurred during 5 years before it in the magnitude range between M and M - 1. The aftershocks are counted within the distance R = 0.03 10^0.5M [km] of the strong earthquake under consideration; the preceding earthquakes are counted within a larger distance R₁ =1.5R. The formulation of the Problem is illustrated in figure 1.

Hypothesis: High activity and irregularity in aftershock sequence precede the reoccurrence of a strong event in a vicinity of the first shock.

Design of the algorithm. The prediction algorithm for the reoccurrence of a large earthquake was found by retrospective analysis of 21 large California earthquakes with M 6.4 [1].

Premonitory phenomena are seven characteristics of the sequence of aftershocks reflecting its following features: the number of the aftershocks, the total area of their sources, the largest distance from the main shock, and the irregularity of this sequence. One more characteristic is the number of earthquakes preceding the first large earthquake.

Prediction is made in two steps.

1. If the number of the aftershocks is less than 10, the next large earthquake is not expected within the time and distance ranges given above, whatever the other characteristics may be.
2. If this number is 10 or more, eight characteristics reflecting premonitory phenomena are determined and a pattern recognition technique known as the Hamming distance is used [2].

4. An alarm is declared if at least five functions out of eight votes for SSE.

Nine regions for prediction of SSE are shown in the map (Figure 2) and listed in Table, as well as magnitudes of strong earthquakes after which we will predict SSE.

• The Schedule of Predictions

The prediction by SSE algorithm has no regular schedule being connected with times of strong earthquakes.

The prediction is announced during two months after each strong earthquake occurred in the regions. The delay of two months is necessary to receive and process aftershock data.