A Tool for Classification and Regression Using Random Forest Methodology: Applications to Landslide Susceptibility Mapping and Soil Thickness Modeling

Classification and regression problems are a central issue in geosciences. In this paper, we present Classification and Regression Treebagger (ClaReT), a tool for classification and regression based on the random forest (RF) technique. ClaReT is developed in Matlab and has a simple graphic user interface (GUI) that simplifies the model implementation process, allows the standardization of the method, and makes the classification and regression process reproducible. This tool performs automatically the feature selection based on a quantitative criterion and allows testing a large number of explanatory variables. First, it ranks and displays the parameter importance; then, it selects the optimal configuration of explanatory variables; finally, it performs the classification or regression for an entire dataset. It can also provide an evaluation of the results in terms of misclassification error or root mean squared error. We tested the applicability of ClaReT in two case studies. In the first one, we used ClaReT in classification mode to identify the better subset of landslide conditioning variables (LCVs) and to obtain a landslide susceptibility map (LSM) of the Arno river basin (Italy). In the second case study, we used ClaReT in regression mode to produce a soil thickness map of the Terzona catchment, a small sub-basin of the Arno river basin. In both cases, we performed a validation of the results and a comparison with other state-of-the-art techniques. We found that ClaReT produced better results, with a more straightforward and easy application and could be used as a valuable tool to assess the importance of the variables involved in the modeling.     

Application of Adjoint Approach to Oil Spill Problems

Adjoint method is applied to various oil spill problems. A three-dimensional model for describing the dispersion of a quasi-passive substance (a pollutant or a nutrient) and its adjoint model are considered in a limited sea region. Direct and adjoint estimates are used to get dual (equivalent) estimates of the mean concentration of the substance in important zones of the region. The role of dual estimates is illustrated with a few examples. They include such oil spill problems as the search of the most dangerous point of the oil tanker route, the oil dispersion with a climatic velocity, and the dependence of the oil concentration estimates on the oil spill rate. One more example is the application of optimal bioremediation strategy for cleaning a few zones polluted by oil. In this case, instead of oil, the model describes the dispersion of a nutrient released to marine environment. Balanced, unconditionally stable second-order finite-difference schemes based on the splitting method for the solution of the dispersion model and its adjoint are suggested. The main and adjoint difference schemes are compatible in the sense that at every fractional step of the splitting algorithm, the one-dimensional split operators of both schemes satisfy a discrete form of Lagrange identity. In the special unforced and non-dissipative case, each scheme has two conservation laws. Every split one-dimensional problem is solved by Thomas’ factorization method.     

Empirical Estimation of Virtual Point Source Height Over a Bank of Coke Ovens

This paper addresses an empirical method of assessment of the height of a virtual point source. In order to achieve such assessment, three planes defining the location of a virtual point source in a space are created and formulas describing the geometry of observer’s sight and plume orientation are used. Data for calculations are obtained on a basis of plume photographs. The final position of a virtual emission point is described by a set of its coordinates in the Cartesian coordinate system. Presented method may be applied as a means to verify and validate the existing formulas describing the rise of plume above the stack exit or as a tool enabling acquisition of information for the sake of creating new formulas of this type. The calculation of the location of a virtual point source ascribed to the point emission from the roof of the bank of coke oven is presented and illustrated with photographs which reveal position of plume, adequate planes, axes and characteristic objects.     

Spatial Validation of Land Use Change Models Using Multiple Assessment Techniques: A Case Study of Transition Potential Models

Using land use and cover change (LUCC) models for the urban growth planning, environmental assessment, and decision-making needs the establishment of an appropriate level of confidence in their performance. The objective of this research is to explore the importance of using multiple assessment techniques in order to fairly evaluate the performance of land use models. An application is conducted by using the Land Change Modeler for Ecological Sustainability (LCM) which is an empirical and transition potential model. LCM is applied to model the agricultural to developed areas transition in Rennes metropolitan area (France). The land demand is estimated using the Markov Chain model; whereas, the transition potential map is implemented using the Multi-Layer Perceptron Neural Network (MLP) method based on historical changes and driving variables. The model performance is assessed based on a variety of the most commonly used validation techniques. At the study area level, the correctness and disagreement analysis shows that LCM performs better at predicting the amount than the allocation of developed areas. Additionally, landscape metrics reveal that LCM tends to predict a fragmented urban form, which seems evident because of the large number of the individual urban patches. At the municipality level, the error budget analysis shows that the model performance, which varies highly between different subareas, needs to be improved. Moreover, the cross-tabulation between the transition potential map and both the observed and the predicted agricultural to developed areas transitions reveals that the order of the transition potential values does not perfectly fit the observed change; whereas, the predicted change is not solely limited to areas with high potential.     

Mercury Air Emission from Coal-Fired Public Power Sector: Uncertainty and Its Monthly Distribution. Case Study from Poland

The paper presents two aspects concerned with the mercury air emission inventory from coal-fired public power and energy plants: an uncertainty analysis, using Monte Carlo simulation (Journal of the American Statistical Association, 44(247), 335–341 1949) and the monthly distributions applying the Denton-Cholette approach (Dagum & Cholette 2006). The analysis determines uncertainty about the estimates mercury air emission from 1990 to 2012 including the development of air pollution control (APC) technologies in the Polish public power and energy sector, also the monthly distributions in comparison with previously obtained results (H?awiczka 2008). The uncertainty of mercury (Hg) content in fuel is 31.6% for hard coal and 42.4% for brown coal. The confidence interval for the estimated emission changed from [kg] (16,082.2; 16,242.2) in 1990 to (10,525.9;10,671.1) in 2012. However, the Denton-Cholette approach overestimates the emissions for the warmer periods of the year, but it could, however, in our view, be applied to attain the monthly distributions.     

Assessing regulatory violations of disinfection by-products in water distribution networks using a non-compliance potential index

Inactivating pathogens is essential to eradicate waterborne diseases. However, disinfection forms undesirable disinfection by-products (DBPs) in the presence of natural organic matter. Many regulations and guidelines exist to limit DBP exposure for eliminating possible health impacts such as bladder cancer, reproductive effects, and child development effects. In this paper, an index named non-compliance potential (NCP) index is proposed to evaluate regulatory violations by DBPs. The index can serve to evaluate water quality in distribution networks using the Bayesian Belief Network (BBN). BBN is a graphical model to represent contributing variables and their probabilistic relationships. Total trihalomethanes (TTHM), haloacetic acids (HAA5), and free residual chlorine (FRC) are selected as the variables to predict the NCP index. A methodology has been proposed to implement the index using either monitored data, empirical model results (e.g., multiple linear regression), and disinfectant kinetics through EPANET simulations. The index’s usefulness is demonstrated through two case studies on municipal distribution systems using both full-scale monitoring and modeled data. The proposed approach can be implemented for data-sparse conditions, making it especially useful for smaller municipal drinking water systems.     

Assessment of physicochemical parameters and prevalence of virulent and multiple-antibiotic-resistant <Emphasis Type="Italic">Escherichia coli</Emphasis> in treated effluent of two wastewater treatment plants and receiving aquatic <Emphasis Type="Italic">milieu</Emphasis> in Durban,South Africa

The poor operational status of some wastewater treatment plants often result in the discharge of inadequately treated effluent into receiving surface waters. This is of significant public health concern as there are many informal settlement dwellers (ISDs) that rely on these surface waters for their domestic use. This study investigated the treatment efficiency of two independent wastewater treatment plants (WWTPs) in Durban, South Africa and determined the impact of treated effluent discharge on the physicochemical and microbial quality of the receiving water bodies over a 6-month period. Presumptive Escherichia coli isolates were identified using biochemical tests and detection of the mdh gene via PCR. Six major virulence genes namely eae, hly, fliC, stx1, stx2, and rfbE were also detected via PCR while antibiotic resistance profiles of the isolates were determined using Kirby-Bauer disc diffusion assay. The physicochemical parameters of the wastewater samples ranged variously between 9 and 313.33 mg/L, 1.52 and 76.43 NTUs, and 6.30 and 7.87 for COD, turbidity, and pH respectively, while the E. coli counts ranged between 0 and 31.2?×?10³ CFU/ml. Of the 200 selected E. coli isolates, the hly gene was found in 28 %, fliC in 20 %, stx2 in 17 %, eae in 14 %, with stx1 and rfbE in only 4 % of the isolates. Notable resistance was observed toward trimethoprim (97 %), tetracycline (56 %), and ampicillin (52.5 %). These results further highlight the poor operational status of these WWTPs and outline the need for improved water quality monitoring and enforcement of stringent guidelines.