Stacy E. Howington

Soil Moisture and Thermal Behavior in the Vicinity of Buried Objects Affecting Remote Sensing Detection: Experimental and Modeling Investigation

Improvements in buried mine detection using remote sensing technology rest on understanding the effects on sensor response of spatial and temporal variability created by soil and environmental conditions. However, research efforts on mine detection have generally emphasized sensor development, while less effort has been made to evaluate the effects of the environmental conditions in which the mines are placed. If the processes governing moisture and temperature distribution near the ground surface can be captured, sensor development and deployment can be more realistically tailored to particular operational scenarios and technologies. The objective of this study is to investigate the effects of the soil environment on landmine detection by studying the influence of the thermal boundary conditions at the land-atmosphere interface and the buried objects themselves on the spatial and temporal distribution of soil moisture around shallow-buried objects. Two separate large tank experiments were performed with buried objects with different thermal properties. Experimental results were compared to results from a fully coupled heat and mass transfer numerical model. Comparison of experimental and numerical results suggests that the vapor enhancement factor used to adjust the vapor diffusive flux described based on Ficks law is not necessary under dry soil conditions. Data and simulations from this study show that the thermal signature of a buried object depends on the complex interaction among a soils water content and its thermal and hydraulic properties. Simulated thermal and saturation contrasts were generally very different for a buried landmine than for other buried objects.

A Suite of Models for Producing Synthetic, Small-scale Thermal Imagery of Vegetated Soil Surfaces

A high-resolution, computational suite has been constructed to produce synthetic thermal imagery of vegetated soil surfaces. Because a soil’s moisture affects its thermal response, the model suite must include both moisture and energy movement within the soil and plants. Thus, the suite consists of a soil model, a vegetation model, and a ray-casting model. The models run simultaneously on a single, parallel or serial computer and communicate using sockets. The soil model is a three-dimensional, spatially adaptive, continuous Galerkin, finite element model that simulates partially-saturated flow and heat transport, coupled to two-dimensional surface water flow. The vegetation model simulates infrared absorption, reflection, and transmission by discretized plant leaves and stems. Ray casting provides boundary conditions for the soil and vegetation thermal models, and produces multi-spectral images of energy reflected and emitted from the synthetic scene. Subsurface phase change, distributed root zone moisture uptake and transpiration, and flow through macropores and cracks are processes under construction. Example calculations to be presented include a multi-million-element simulation for an arid test site that is only a few meters in its longest dimension. The models are driven with meteorological data and are built using material property data collected at the field site. Synthetic images produced are compared against those from thermal cameras. A long-term goal of this work is to help build inversion software to estimate ground state information (soil moisture and physical property distributions) from airborne imagery.

Binary fuzzy measures and Choquet integration for multi-source fusion

Countless challenges in engineering require the intelligent combining (aka fusion) of data or information from multiple sources. The Choquet integral (ChI), a parametric aggregation function, is a well-known tool for multi-source fusion, where source refers to sensors, humans and/or algorithms. In particular, a selling point of the ChI is its ability to model and subsequently exploit rich interactions between inputs. For a task with N inputs, the ChI has 2N interaction variables. Therefore, the ChI becomes intractable quickly in terms of storage and its data-driven learning. Herein, we study the properties of an efficient to store, compute, and ultimately optimize version of the ChI based on a binary fuzzy measure (BFM). The BFM is further motivated by empirical observations in the areas of multi-sensor fusion and hyperspectral image processing. Herein, we provide a deeper understanding of the inner workings, capabilities and underlying philosophy of a BFM ChI (BChI). We also prove that two fuzzy integrals, the ChI and the Sugeno integral, are equivalent for a BFM. Furthermore, only a small subset of BFM variables need be stored, which reduces the BChI to a relatively simple look up operation.

Application of the pseudo-transient technique to a real-world unsaturated flow groundwater problem

Modeling unsaturated flow using numerical techniques such as the finite element method can be especially difficult because of the highly nonlinear nature of the governing equations. This problem is even more challenging when a steady-state solution is needed. This paper describes the implementation of a pseudo-transient technique to drive the solution to steady-state and gives results for a real-world problem. The application discussed in this paper does not converge using a traditional Picard nonlinear iteration type finite element solution. Therefore, an alternate technique needed to be developed and tested.

Genetic programming based Choquet integral for multi-source fusion

While the Choquet integral (Chi) is a powerful parametric nonlinear aggregation function, it has limited scope and is not a universal function generator. Herein, we focus on a class of problems that are outside the scope of a single Chi. Namely, we are interested in tasks where different subsets of inputs require different Chls. Herein, a genetic program (GF) is used to extend the Chi, referred to as GpChI hereafter, specifically in terms of compositions of Chls and/or arithmetic combinations of Chls. An algorithm is put forth to leam the different GP Chls via genetic algorithm (GA) optimization. Synthetic experiments demonstrate GpChI in a controlled fashion, i.e., we know the answer and can compare what is learned to the truth. Real-world experiments are also provided for the mult-sensor fusion of electromagnetic induction (EMI) and ground penetrating radar (GPR) for explosive hazard detection. Our mutli-sensor fusion experiments show that there is utility in changing aggregation strategy per different subsets of inputs (sensors or algorithms) and fusing those results.

Improved Well Modeling Tools for Unsaturated Flow Pump-and-Treat Remediation Studies

Many Department of Defense military sites and Environmental Protection Agency superfund sites benefit from pump-and-treat systems for their remediation. One of the ways used to determine the effectiveness of a pump-and-treat system is by using a groundwater finite element method program to compute flow patterns from proposed well placement. As the governing Richards’ equation for unsaturated flow is highly nonlinear, the finite element discretization often has trouble converging. This is especially true when (1) a significant part of the flow region has unsaturated flow, (2) soils such as sand have relative hydraulic conductivities that go from almost vertical to almost horizontal, (3) water in very dry soil is being applied at the top of the flow region, and (4) the wells become partially above the free surface when the pumping is turned on. This paper focuses on an improved well-modeling technique that allowed the finite element solution to go from nonconvergent to converging in 15 nonlinear iterations for the Higgins Farm Superfund site The Higgins Farm Superfund site, which is approximately 75 acres in size and currently operated as a cattle farm, is located in a rural area along Route 518 in Franklin Township, Somerset County, New Jersey. The site is primarily pastureland where a drum burial dump was discovered. Extraction wells were placed around the perimeter with an onsite treatment plant. The finite element run not only did not converge but also the modeling of wells was initially very tedious. A given well has a specified flow, and the user would manually try to distribute this flow among the given set of nodes modeling the well. During the iteration toward convergence, if a node of a well became above the free surface (pressure head less than zero), the code stopped. The user then manually redistributed the flow for that well and started the process again. However, combining all of the nodes for a well into one well supernode was determined to be the best solution. The nodes representing a partially penetrating well are combined into a single degree-of-freedom well supernode. Now, one total value of flow is used as input to the well supernode, and one single value of total head is computed as output. Further improvement was added to the well model by implementing a nonreversible elimination of well nodes above the water table. This paper will give the details of this computational experience using the high performance computers at the ERDC and illustrate the general applicability of this approach to other studies.

Imbedding velocity autocorrelation into simulators for constituent transport through porous media

Simulation based on the traditional advection-dispersion equation often is inadequate for making fate-transport predictions. The dominant process in transport through porous media is differential advection, which can be characterized statistically through spatial and temporal correlations in the velocity field. A particle tracking scheme that contains appropriate velocity statistics predicts realistic contaminant plumes. A particular generating scheme for random correlated velocities is introduced that can be mathematically transformed to a system of transport equations for parallel media whereby each equation transports a fraction of the contaminant mass, each with its characteristic advection velocity. The rate of transport among the parallel media determines the spectrum of length scales represented. The formulation of a continuum, multi-scale transport model thus follows naturally from the particle tracking scheme. This resulting computational scheme is one familiar to some fracture flow modelers—the multiple-interacting continuum approach. This discussion ties together the correlated random walk, non-local, and multiple-interacting continuum approaches and provides a means for model calibration.