Safe mine operations and optimal economical decision making in the context of lignite resources require an adequate level of knowledge about the spatial distribution of critical attributes in terms of geometry and quality in the deposit. Therefore, ore body models are generated using different approaches in geostatistics, depending on the problem to be solved. In this article the analysis of geostatistical methods used for deposits modeling has been presented. Based on exploration data concerning caloric value Q, models of one exemplary lignite deposit has been made. Two models of deposit were prepared using two different methods: ordinary kriging (OK) and sequential Gaussian conditional simulation (SGSIM). Different models of the same deposit were analyzed and compared with source data using criterion of fidelity to statistical attributes like: mean value, variance, statistical distribution. Models, which have been created based on exploration data, were compared with in-situ data gained from survey activities in the exploitation process. As a result of comparison correlation factor and measures of deviations were computed: average relative error, absolute relative error. Models were compared with in situ data, considering statistical features and local variability as well. In conclusion, the study gives valuable information into the benefits of using certain geostatistical approaches for variable tasks and problems in the lignite deposits design process. For the assessment of average values of deposit parameters ordinary kriging provides appropriate effects. Geostatisical simulation (e.g. sequential Gaussian simulation - SGSIM) provides much more relevant information for tasks connected to probability (or risk) of defined threshold exceedences than ordinary kriging. Models made with simulation method are characterized by high fidelity of spatial distribution in comparison to source data.
Anisotropy of variations of Polish mineral deposit parameters is rarely the subject of interest of geologists who carry on the assessment projects . However, if the anisotropy is strong its description and mathematical modeling are rational and justified as it may affect the accuracy of many calculations suitably for mining geology and mining engineering, e.g. estimation of resources and grade of particular raw-material, interpolation of deposit parameters values and construction of their contour maps, designing of optimum grade mining operations or densification of sampling grid. In geostatistics anisotropy is described with directional semivariograms which represent average variability of values of particular deposit parameter in various directions, depending on the distance between sampling sites. Convenient graphic presentation of anisotropy is map of directional semivariograms and good mathematical presentation are functions describing the anisotropy models.
The paper presents the results of geostatistical descriptions of various anisotropy types in selected examples of Polish mineral deposits. Taking into account the spherical variability model, the influence of anisotropy on the results of deposit parameters estimations has been theorized for both the interpolation point and calculation block (area). It was found that anisotropy is effective for parameters estimation if three mutually interrelated factors are considered: power of directional diversification of parameters variation, contribution of random component to total, observed variation of parameters and the range of semivariograms (autocorrelation) of parameter referred to the average sampling grid density.
The results demonstrate that anisotropy influences much more the estimations of parameters value in interpolation points than those of average values of parameters calculated for particular parts of deposit (calculation blocks). Moreover, anisotropy is unimportant when the random component of variability dominates the overall variability of analyzed parameter. Therefore, the simpler, isotropic variability model can be applied to geostatistical estimations of deposit parameters.
The purpose of the paper is to analyze the spatial variability of precipitation in Poland in the years 1981–2010. The av-erage annual rainfall was 607 mm. Precipitation in Poland is characterized by high spatial and temporal variability. The lowest annual precipitation was recorded in the central part of the country, where they equaled 500 mm. The highest annual precipitation totals were determined in the south, equaling 970 mm. The average precipitation in the summer half-year is 382 mm (63% of the annual total). On the basis of data from 53 climate stations, maps were made of the spatial distribution of precipitation for the period of the year and winter and summer half-year. The kriging method was used to map rainfall distribution in Poland. In the case study, cross-validation was used to compare the prediction performances of three periods. Kriging, with exponential type of semivariogram, gave the best performance in the statistical sense. Their application is justices especially in areas where landform is very complex. In accordance with the assumptions, the mean prediction error (ME), mean standardized prediction error (MSE), and root mean-square standardized prediction error (RMSSE) values are approximately zero, and root-mean-square prediction error (RMSE) and average standard error (ASE) reach values well below 100.
The paper discusses some of the recent advances in kriging based worst-case design optimisation and proposes a new two-stage approach to solve practical problems. The efficiency of the infill points allocation is improved significantly by adding an extra layer of optimisation enhanced by a validation process.
Material parameters identification by inverse analysis using finite element computations leads to the resolution of complex and time-consuming optimization problems. One way to deal with these complex problems is to use meta-models to limit the number of objective function computations. In this paper, the Efficient Global Optimization (EGO) algorithm is used. The EGO algorithm is applied to specific objective functions, which are representative of material parameters identification issues. Isotropic and anisotropic correlation functions are tested. For anisotropic correlation functions, it leads to a significant reduction of the computation time. Besides, they appear to be a good way to deal with the weak sensitivity of the parameters. In order to decrease the computation time, a parallel strategy is defined. It relies on a virtual enrichment of the meta-model, in order to compute q new objective functions in a parallel environment. Different methods of choosing the qnew objective functions are presented and compared. Speed-up tests show that Kriging Believer (KB) and minimum Constant Liar (CLmin) enrichments are suitable methods for this parallel EGO (EGO-p) algorithm. However, it must be noted that the most interesting speed-ups are observed for a small number of objective functions computed in parallel. Finally, the algorithm is successfully tested on a real parameters identification problem.