Modelowanie procesu osiadania terenu górniczego Kopalni Węgla Brunatnego Bełchatów – nowe podejście

Abstract

Land subsidence modeling in mining area of open pit Brown Coal Mine Bełchatów – a new approach.A b s t r a c t. From the beginning of open-pit mining works (i.e. ground massive dewatering, access excavation, cover dumping) in 1976, which were strictly connected with an exposure a brown coal beds on Bełchatów field it was noticed, that a land surface subsided in the vicinity of Brown Coal Mine Bełchatów. Quantitative land subsidence assessments, which are based on deterministic models (elastic ground model, consolidation model), are not efficient enough to simulate the process – adjusted coefficient of determination amounts R2kor = 0.15–0.28. The alternative approach a quantitative description of the land subsidence, which is presented in the doctoral thesis, is the application of statistical models (multiple regression, neural networks), which are built on ground of collected sampling data in 1976–2005. Among three model types, which were divided according to available variables sets, the most efficient were autocorrelation models (R2kor = 0.96–0.97), which included as a variable the subsidence depth from previous years. To the variables, which have a significant impact on land subsidence progress, belong also the history of ground massive dewatering and the depth of Cenozoic cover. The most significant effects of the land subsidence appear within a one or two years delay towards the dewatering process, which is a relevant  indication in the assessment of the consolidation delay. Recorded differences in land subsidence and dewatering semivariogram distances prove the spatial anisotropy of ground massive – on average the distance is twice longer in parallel of latitude direction, which is caused by the structure and geographical location of Kleszczów Graben. As the results show, the optimal frequency of land subsidence and decline of water table measurement is about every 2–3 years and the measurement points should be located not farther than 1200 m from one another. The minimum number of observations, which are needed to build an efficient land subsidence model, in a data set is about 200–300. Presented coherent statistical methods enable to compare different types of linear and non-linear land subsidence models and allow applying proposed methodology for modeling subsidence process in another mining areas – included contiguous Szczerców Brown Coal Field.