Stochastic forecasting of mine water inrushes

An event-based stochastic forecasting approach is used to model water inrushes into underground works under karstic water hazard. The stochastic properties of inrushes are related to the statistical properties of fissures in the karstic rock. The probability distributions (DF) of five random variables of interest in design are estimated; namely, inrush yield q, number N of inrushes per unit area, distance L between inrushes, maximum q(max) in N events and total yield Q. The phenomenological hypotheses of log normal DF of q(max) and Poisson DF of N are reinforced by observation data. On the basis of these DF, a Monte Carlo simulation of a spatial Poisson process of inrushes is run to estimate the DF of q(max) and Q. The derivation of Bayesian DF to account for parameter uncertainty is discussed. The stochastic model is used for design and operation of minewater control facilities in the Transdanubian karstic region of Hungary.