Debasmita Misra

A hybrid ensemble model of kriging and neural network for ore grade estimation

This paper presents a new hybrid methodology involving kriging and artificial neural network for ore grade estimation of two variables namely, Al2O3% and SiO2%, in a bauxite deposit. The dataset was divided into three statistically similar subsets: training, calibration and validation sets using a genetic algorithm. The proposed hybrid ensemble model was formed using a kriging model and several neural network models. The outputs of these component models were combined using two methods to produce a unified prediction. The results indicated that the hybrid model was not a better estimator than the kriging model for the variable Al2O3%. However, it provides slightly better performance in comparison to any of the other component models in the ensemble for the variable SiO2%.

Using the one-dimensional S-transform as a discrimination tool in classification of hyperspectral images

A standard part of processing remote sensing data is image classification, in which we assume each pixel belongs to a class or theme with a unique spectral signature. Discrimination may be defined as the phenomenon where multiple themes exhibit very similar spectral patterns within a wavelength range of interest and is a common challenge in remote sensing. As a result, researchers may not achieve the desired classification accuracy. A robust discrimination technique must be capable of detecting very minor spectral differences between classes with similar spectral signatures. Using the one-dimensional S-transform, a spectral localization technique to discriminate similar lithologic classes on a hyperspectral satellite image, we investigated the S-amplitude spectra efficiency in enhancing the spectral information of each pixel of a known class. We compared the overall accuracy of classified themes using a support vector classification (SVC) scheme, with and without using the enhanced spectral information. We found that SVC aided by spectral enhancement from the S-transform provided better classification accuracy. Thus, this method may prove very useful in scenarios where pixels of a known class are sparse and not easily separable.

Historical changes in trace metals and hydrocarbons in nearshore sediments, Alaskan Beaufort Sea, prior and subsequent to petroleum-related industrial development: Part II. Hydrocarbons

Concentrations of Fe, As, Ba, Cd, Cu, Cr, Pb, Mn, Ni, Sn, V and Zn in mud (<63μm size), and total and methyl Hg in gross sediment are reported for Arctic Alaska nearshore. Multivariate-PCA analysis discriminated seven station clusters defined by differences in metal concentrations, attributed to regional variations in granulometry and, as in Elson Lagoon, to focused atmospheric fluxes of contaminants from Eurasia. In Colville Delta-Prudhoe Bay, V increase was noted in 1985 and 1997 compared to 1977, and Ba increase from 1985 to 1997. Presumably the source of increased V is the local gas flaring plant, and the elevated Ba is due to barite accumulation from oil drilling effluents. In Prudhoe Bay, concentration spikes of metals in ∼1988 presumably reflect enhanced metals deposition following maximum oil drilling in 1980s. In summary, the Alaskan Arctic nearshore has remained generally free of metal contamination despite petroleum-related activities in past 40 years.

Machine Learning Algorithms and Their Application to Ore Reserve Estimation of Sparse and Imprecise Data

Traditional geostatistical estimation techniques have been used predominantly by the mining industry for ore reserve estimation. Determination of mineral reserve has posed considerable challenge to mining engineers due to the geological complexities of ore body formation. Extensive research over the years has resulted in the development of several state-of-the-art methods for predictive spatial mapping, which could be used for ore reserve estimation; and recent advances in the use of machine learning algorithms (MLA) have provided a new approach for solving the prob-lem of ore reserve estimation. The focus of the present study was on the use of two MLA for estimating ore reserve: namely, neural networks (NN) and support vector machines (SVM). Application of MLA and the various issues involved with using them for reserve estimation have been elaborated with the help of a complex drill-hole dataset that exhibits the typical properties of sparseness and impreciseness that might be associated with a mining dataset. To investigate the accuracy and applicability of MLA for ore reserve estimation, the generalization ability of NN and SVM was compared with the geostatistical ordinary kriging (OK) method.

Effect of concentration gradients on biodegradation in bench-scale sand columns with HYDRUS modeling of hydrocarbon transport and degradation

The present research investigated to what extent results obtained in small microcosm experiments can be extrapolated to larger settings with non-uniform concentrations. Microbial hydrocarbon degradation in sandy sediments was compared for column experiments versus homogenized microcosms with varying concentrations of diesel, Syntroleum, and fish biodiesel as contaminants. Syntroleum and fish biodiesel had higher degradation rates than diesel fuel. Microcosms showed significantly higher overall hydrocarbon mineralization percentages (p < 0.006) than columns. Oxygen levels and moisture content were likely not responsible for that difference, which could, however, be explained by a strong gradient of fuel and nutrient concentrations through the column. The mineralization percentage in the columns was similar to small-scale microcosms at high fuel concentrations. While absolute hydrocarbon degradation increased, mineralization percentages decreased with increasing fuel concentration which was corroborated by saturation kinetics; the absolute CO2 production reached a steady plateau value at high substrate concentrations. Numerical modeling using HYDRUS 2D/3D simulated the transport and degradation of the investigated fuels in vadose zone conditions similar to those in laboratory column experiments. The numerical model was used to evaluate the impact of different degradation rate constants from microcosm versus column experiments.

Modeling Water Table Mounding and Contaminant Transport beneath Storm-Water Infiltration Basins

The objectives of this study were to link an unsaturated and saturated flow model for the purpose of evaluating mounding and contaminant transport beneath an infiltration basin, to calibrate and test the combined water table flow model using experimental data collected from an infiltration basin, and to evaluate the potential for contaminant transport with a numerical fate and transport model. Mound formation may reduce the thickness of the soil available to retard pollutant movement, reduce the infiltration rate of the basin if the mound intersects the basin bottom, and facilitate contaminant movement away from the basin. A 0.10-ha infiltration basin serving a 9.4-ha residential subdivision in Oconomowoc, Wisconsin, was instrumented. Two storm events were modeled using the three-dimensional saturated numerical model MODFLOW. Recharge used in MODFLOW was taken from the seepage flux of the unsaturated one-dimensional model HYDRUS. A good fit was achieved between modeled and measured timing and magnitude of...

Evaluation of artificial neural networks and kriging for the prediction of arsenic in Alaskan bedrock-derived stream sediments using gold concentration data

The detection of arsenic in sediments of placer gold mining areas is critical for planning future controls on migration and mitigation, or tapping uncontaminated groundwater resources for public water use. Arsenic (As) is often found to be collocated and correlated with gold in sediments. However, due to biogeochemical processes, arsenic can partition between the solid and the dissolved fractions in sediments and their interstitial waters. Such partitioning can mobilize arsenic into areas away from the co-located gold distribution in the sediments. In such cases, it is critical to detect the dispersed arsenic concentration. In this paper, neural network (NN) and kriging techniques were used to predict the presence of arsenic in the sediments of Circle City, Alaska using the gold concentration distribution within the sediments. The results obtained using kriging were more promising than those using NNs, albeit a statistically low correlation existed between the observed and the predicted arsenic concentrations. However, irrespective of the method used, the prediction of arsenic value without using gold concentration data was extremely poor.

Analysis and Application of Support Vector Machine Based Simulation for Runoff and Sediment Yield

Physics-based models for simulation of runoff and sediment yield from watersheds are quite complex and involved due to tremendous spatial variability of watershed characteristics and precipitation patterns. Recently, pattern-learning algorithms such as the artificial neural networks (ANN) have gained popularity in simulating the rainfall-runoff-sediment yield processes producing comparable accuracy. We have simulated daily, weekly, and monthly runoff and sediment yield from an Indian watershed (area= 7820 Sq.Km), with data from the monsoon period, using support vector machines (SVM), a statistical learning theory based pattern-learning algorithm. The performance of the model was evaluated using correlation coefficient (r) and coefficient of efficiency (E). The time series data was split into a training set for the learning process and a prediction set for comparison of the model’s forecasting ability. The results of SVM were compared to those of ANN. We concluded that SVM provided significant improvement in both training and prediction abilities as compared to those of ANN. ANN being a computationally intensive method, SVM could be used as an efficient alternative for runoff and sediment yield predictions providing at least comparable accuracy.

GIS Based Marine Platinum Exploration, Goodnews Bay Region, Southwest Alaska

Goodnews Bay, southwest Alaska, is known for platinum (Pt) reserve that extends offshore in the Bering Sea. To assess the nearshore placer potential we first collated marine Pt concentrations available since 1960 in a geographic information system (GIS) database. Subsequently, in 2005, we collected 23 pipe dredge sediment samples and 26 vibracores from unexplored sites and analyzed them for Pt. This sampling was supplemented by magnetic (Sea Spy) and seismic (side scan, geoacoustic and datasonic bubble pulser) surveys. Integrating results of geospatial analysis of Pt concentrations with geophysical analysis using GIS techniques led to delineate four locations encouraging for further Pt exploration. Of these, two locations fall close to paleochannels and drowned ultramafic source, while the other two coincide with high energy environments in the Goodnews Bay and close to the Carter Bay.

Assessment of Mixed Finite Element Method Applied to One-Dimensional Transient Unsaturated Flow

Accurate simulation of Darcy flux is essential to simulate contaminant transport in the unsaturated zone accurately. The mixed finite element method has been used to obtain highly accurate flux distribution in unsaturated zone flow applications. We have demonstrated that application of mixed finite element method to steady state unsaturated flow yields accurate moisture content and flux profiles (Misra and Nieber, 2003, ASAE, Paper No. 032112). The objective of this paper is to present a few case studies of application of mixed finite element method to transient unsaturated flow problems, compare those to the solution obtained from the conventional finite element methods and to analyze any special properties of the solutions obtained from the mixed finite element methods. It has been shown in this paper that the solutions obtained from the mixed finite element method are highly accurate in rapidly changing flux distributions and heterogeneous flow distributions. The effect of approximating the hydraulic conductivity or the flux over the element has been discussed.