This paper presents models of dependences between seismic activity (of the N number and seismic energy of tremors E) induced by conducted longwall exploitation and observed quantities of ground surface deformations. In the models, the observed quantities of ground surface deformations have been described with subsidence through cross-section areas Pw along observation line parallel to the advancing longwall face. Linear models of regression with a single independent variable have been adopted. Dependences between accrued quantities of the total number (N), and total seismic energy of tremors (E) and increasing quantities Pw were determined. The article presents verification of models for longwall exploitation conducted in the area of Upper Silesia Coal Basin by the hard-coal mine KWK 'Halemba'. The surface Pw has been determined basing on the results of geodetic measurements (levelling and angle-linear measurements) carried out on an observation network implemented in the area of exploitation. Measurements carried out at the observation line located approximately within the symmetry axis of the longwall field perpendicular to the face of the longwall were applied. The data concerning the induced seismic activity was obtained from the Mining Geophysics Station. Verification of the models was carried out for increasing quantities of E, N and Pw in time periods of realized cycles of geodetic measurements. For a determined regressional dependence of E in relation to Pw at the significance level of p < 0.05, it has been found that there is significant influence of the variable Pw on E. The value of the determination coefficient R2 equalled 0.9772. For the determined regressional dependence of N in relation to Pw, the value of the determination coefficient R2 equalled 0.9905 The regression coefficient b and the correlation coefficient are significantly different from 0 (zero) at the significance level of p < 0.05. The analysis of remainder variables in the time function, assorted values and an independent variable excluded any reservations regarding adequacy of the considered models. The models proposed in this paper may be used for development of a method for seismic risk forecasting.