Jagath J. Kaluarachchi

Modular neural networks to predict the nitrate distribution in ground water using the on-ground nitrogen loading and recharge data

Abstract Artificial neural networks have proven to be an attractive mathematical tool to represent complex relationships in many branches of hydrology. Due to this attractive feature, neural networks are increasingly being applied in subsurface modeling where intricate physical processes and lack of detailed field data prevail. In this paper, a methodology using modular neural networks (MNN) is proposed to simulate the nitrate concentrations in an agriculture-dominated aquifer. The methodology relies on geographic information system (GIS) tools in the preparation and processing of the MNN input–output data. The basic premise followed in developing the MNN input–output response patterns is to designate the optimal radius of a specified circular-buffered zone centered by the nitrate receptor so that the input parameters at the upgradient areas correlate with nitrate concentrations in ground water. A three-step approach that integrates the on-ground nitrogen loadings, soil nitrogen dynamics, and fate and transport in ground water is described and the critical parameters to predict nitrate concentration using MNN are selected. The sensitivity of MNN performance to different MNN architecture is assessed. The applicability of MNN is considered for the Sumas-Blaine aquifer of Washington State using two scenarios corresponding to current land use practices and a proposed protection alternative. The results of MNN are further analyzed and compared to those obtained from a physically-based fate and transport model to evaluate the overall applicability of MNN.

Applicability of statistical learning algorithms in groundwater quality modeling

[1] Four algorithms are outlined, each of which has interesting features for predicting contaminant levels in groundwater. Artificial neural networks (ANN), support vector machines (SVM), locally weighted projection regression (LWPR), and relevance vector machines (RVM) are utilized as surrogates for a relatively complex and time-consuming mathematical model to simulate nitrate concentration in groundwater at specified receptors. Nitrates in the application reported in this paper are due to on-ground nitrogen loadings from fertilizers and manures. The practicability of the four learning machines in this work is demonstrated for an agriculture-dominated watershed where nitrate contamination of groundwater resources exceeds the maximum allowable contaminant level at many locations. Cross-validation and bootstrapping techniques are used for both training and performance evaluation. Prediction results of the four learning machines are rigorously assessed using different efficiency measures to ensure their generalization ability. Prediction results show the ability of learning machines to build accurate models with strong predictive capabilities and hence constitute a valuable means for saving effort in groundwater contamination modeling and improving model performance.

Parameter estimation using artificial neural network and genetic algorithm for free-product migration and recovery

Artificial neural network (ANN) is considered to be a universal function approximator, and genetic algorithm (GA) is considered to be a robust optimization technique. As such, ANN regression analysis and ANN-GA optimization techniques can be used to perform inverse groundwater modeling for parameter estimation. In this manuscript the applicability of these two techniques in solving an inverse problem related to a light-hydrocarbon-contaminated site is assessed. The critical parameters to be evaluated are grain-size distribution index α and saturated hydraulic conductivity of water Ksw, since these parameters control free-product volume predictions and flow. A set of published data corresponding to a light-hydrocarbon-contaminated unconfined aquifer was used as the base case to determine the applicability of these methods under a variety of scenarios. Using limited monitoring- and recovery-well data under homogeneous and heterogeneous conditions, the critical parameters were evaluated. The results were used to determine the relative effectiveness of each method and corresponding limitations. The results of the work suggested that ANN regression analysis has limited utility, especially with heterogeneous soils, whereas the ANN-GA optimization can provide superior results with better computational efficiency. Finally, a general guideline for solving inverse problems using the two techniques is outlined.

Impacts of the 2004 tsunami on groundwater resources in Sri Lanka

The 26 December 2004 tsunami caused widespread destruction and contamination of coastal aquifers across southern Asia. Seawater filled domestic open dug wells and also entered the aquifers via direct infiltration during the first flooding waves and later as ponded seawater infiltrated through the permeable sands that are typical of coastal aquifers. In Sri Lanka alone, it is estimated that over 40,000 drinking water wells were either destroyed or contaminated. From February through September 2005, a team of United States, Sri Lankan, and Danish water resource scientists and engineers surveyed the coastal groundwater resources of Sri Lanka to develop an understanding of the impacts of the tsunami and to provide recommendations for the future of coastal water resources in south Asia. In the tsunami-affected areas, seawater was found to have infiltrated and mixed with fresh groundwater lenses as indicated by the elevated groundwater salinity levels. Seawater infiltrated through the shallow vadose zone as well as entered aquifers directly through flooded open wells. Our preliminary transport analysis demonstrates that the intruded seawater has vertically mixed in the aquifers because of both forced and free convection. Widespread pumping of wells to remove seawater was effective in some areas, but overpumping has led to upconing of the saltwater interface and rising salinity. We estimate that groundwater recharge from several monsoon seasons will reduce salinity of many sandy Sri Lankan coastal aquifers. However, the continued sustainability of these small and fragile aquifers for potable water will be difficult because of the rapid growth of human activities that results in more intensive groundwater pumping and increased pollution. Long-term sustainability of coastal aquifers is also impacted by the decrease in sand replenishment of the beaches due to sand mining and erosion.

Application of artificial neural network and genetic algorithm in flow and transport simulations

Abstract Artificial neural network (ANN) is considered to be a powerful tool for solving groundwater problems which require a large number of flow and contaminant transport (GFCT) simulations. Often, GFCT models are nonlinear, and they are difficult to solve using traditional numerical methods to simulate specific input–output responses. In order to avoid these difficulties, ANN may be used to simulate the GFCT responses explicitly. In this manuscript, recent research related to the application of ANN in simulating GFCT responses is critically reviewed, and six research areas are identified. In order to study these areas, a one-dimensional unsaturated flow and transport scenario was developed, and ANN was used to simulate the effects of specific GFCT parameters on overall results. Using these results, ANN concepts related to architecture, sampling, training, and multiple function approximations are studied, and ANN training using back-propagation algorithm (BPA) and genetic algorithm (GA) are compared. These results are summarized, and appropriate conclusions are made.

Multi-criteria decision analysis with probabilistic risk assessment for the management of contaminated ground water

Abstract Traditionally, environmental decision analysis in subsurface contamination scenarios is performed using cost–benefit analysis. In this paper, we discuss some of the limitations associated with cost–benefit analysis, especially its definition of risk, its definition of cost of risk, and its poor ability to communicate risk-related information. This paper presents an integrated approach for management of contaminated ground water resources using health risk assessment and economic analysis through a multi-criteria decision analysis framework. The methodology introduces several important concepts and definitions in decision analysis related to subsurface contamination. These are the trade-off between population risk and individual risk, the trade-off between the residual risk and the cost of risk reduction, and cost-effectiveness as a justification for remediation. The proposed decision analysis framework integrates probabilistic health risk assessment into a comprehensive, yet simple, cost-based multi-criteria decision analysis framework. The methodology focuses on developing decision criteria that provide insight into the common questions of the decision-maker that involve a number of remedial alternatives. The paper then explores three potential approaches for alternative ranking, a structured explicit decision analysis, a heuristic approach of importance of the order of criteria, and a fuzzy logic approach based on fuzzy dominance and similarity analysis. Using formal alternative ranking procedures, the methodology seeks to present a structured decision analysis framework that can be applied consistently across many different and complex remediation settings. A simple numerical example is presented to demonstrate the proposed methodology. The results showed the importance of using an integrated approach for decision-making considering both costs and risks. Future work should focus on the application of the methodology to a variety of complex field conditions to better evaluate the proposed methodology.

Critical assessment of the operator-splitting technique in solving the advection-dispersion-reaction equation: 1. First-order reaction

Abstract Operator-splitting technique (OST) is a common mathematical approach used in the solution of the advection-dispersion-reaction equation (ADRE), especially in the presence of biological decay, where the scales of transport and biological decay are far apart. The OST introduces a time-lag between the advection-dispersion and reaction stages by splitting the ADRE causing a breakdown of the physics of the problem, thus limiting its applicability. In this work, the applicability of the operator splitting technique is studied in parts. This first manuscript addresses the critical limitations of the operator-splitting technique as related to first-order decay, and the second manuscript extends the work to include coupled transport between hydrocarbon and oxygen with Monod kinetics describing biological decay. The critical assessment is performed to address the errors associated with the time-lag due to splitting. The assessment is based on mass balance errors, deviations of concentration prediction and sensitivities of concentration deviations. The results show that the splitting introduces an inherent error independent of the discretization errors. This inherent error increases with reaction rate and time-lag, and the overall errors can be reduced by using the alternate operator-splitting technique suggested by previous researchers.

Optimizing separate phase light hydrocarbon recovery from contaminated unconfined aquifers

Abstract A modeling approach is presented that optimizes separate phase recovery of light non-aqueous phase liquids (LNAPL) for a single dual-extraction well in a homogeneous, isotropic unconfined aquifer. A simulation/regression/optimization (S/R/O) model is developed to predict, analyze, and optimize the oil recovery process. The approach combines detailed simulation, nonlinear regression, and optimization. The S/R/O model utilizes nonlinear regression equations describing system response to time-varying water pumping and oil skimming. Regression equations are developed for residual oil volume and free oil volume. The S/R/O model determines optimized time-varying (stepwise) pumping rates which minimize residual oil volume and maximize free oil recovery while causing free oil volume to decrease a specified amount. This S/R/O modeling approach implicitly immobilizes the free product plume by reversing the water table gradient while achieving containment. Application to a simple representative problem illustrates the S/R/O model utility for problem analysis and remediation design. When compared with the best steady pumping strategies, the optimal stepwise pumping strategy improves free oil recovery by 11.5% and reduces the amount of residual oil left in the system due to pumping by 15%. The S/R/O model approach offers promise for enhancing the design of free phase LNAPL recovery systems and to help in making cost-effective operation and management decisions for hydrogeologists, engineers, and regulators.

Stochastic analysis of oxygen- and nitrate-based biodegradation of hydrocarbons in aquifers

A Lagrangian stochastic framework was used to analyze field-scale aerobic biodegradation in a heterogeneous aquifer, using Monod-kinetics based reactions between the contaminant, oxygen and microbes. Subsurface heterogeneity was represented by closed-form travel time distributions, derived from a spatially correlated random hydraulic conductivity field with a log-normal distribution. The solution to the coupled and nonlinear, one-dimensional Lagrangian transport equations was obtained using the operator-splitting technique. The presence of nitrate, and considering nitrate as a second electron acceptor, produced significantly different results under intrinsic conditions for different scales of heterogeneity and sorption. In general, nitrate as a second electron acceptor can substantially lower the peak contaminant concentration and increase the maximum remediation under various conditions of heterogeneity and sorption. There exists a critical value for retardation coefficients of both contaminant and microbes that produce complete degradation of mass, and this value depends on the availability of the electron acceptor(s) and is independent of the heterogeneity. Maximum remediation and peak contaminant concentration were sensitive to half-saturation constants. Enhanced remediation using oxygen and nitrate showed that maximum remediation can be increased by approximately 15% when oxygen or nitrate concentration was increased by 50%, but a further increase may be obtained if injection occurred at a more effective location. The proposed stochastic methodology is capable of analyzing field-scale biodegradation using multiple electron acceptors in a simple and computationally attractive manner, producing useful results on design parameters. The key contributions arising from the Lagrangian stochastic framework in field-scale analysis, its limitations and potential approaches for overcoming these limitations are also discussed.

Critical assessment of the operator-splitting technique in solving the advection-dispersion-reaction equation: 2. Monod kinetics and coupled transport

Abstract In the first manuscript of this two-part series, the behavior of the inherent time-lag error in the operator-splitting technique has been discussed. However, the discussion has been limited to the advection-dispersion-reaction equation with first-order kinetics. In this manuscript, the discussion is extended to address the applicability of the operator-splitting technique in solving the advection-dispersion-reaction equation with Monod kinetics. The discussion also considers the two-species coupled transport problem. The results of the analysis show that the behavior of the time-lag error causing a mass balance error is rather similar to those observed with first-order kinetics. The time-lag error directly depends on the reaction rate, and hence, affects the oxygen equation due to the high rate of oxygen consumption. Also, the alternate operator-splitting scheme shows a greater accuracy compared to the normal operator-splitting technique. Upon observing such similarities, a concept of equivalent first-order reaction rate is derived. The equivalent reaction rate has been found to be extremely useful for predicting the time-lag error of the single- and two-species mass transport with Monod kinetics using the result of the first-order reaction problems obtained in the first manuscript.