Sandra E. Thompson

Pacific Northwest GridWise™ Testbed Demonstration Projects; Part I. Olympic Peninsula Project

This report describes the implementation and results of a field demonstration wherein residential electric water heaters and thermostats, commercial building space conditioning, municipal water pump loads, and several distributed generators were coordinated to manage constrained feeder electrical distribution through the two-way communication of load status and electric price signals. The field demonstration took place in Washington and Oregon and was paid for by the U.S. Department of Energy and several northwest utilities. Price is found to be an effective control signal for managing transmission or distribution congestion. Real-time signals at 5-minute intervals are shown to shift controlled load in time. The behaviors of customers and their responses under fixed, time-of-use, and real-time price contracts are compared. Peak loads are effectively reduced on the experimental feeder. A novel application of portfolio theory is applied to the selection of an optimal mix of customer contract types.

Identification of bacterial spores using statistical analysis of Fourier transform infrared photoacoustic spectroscopy data.

Fourier transform infrared photoacoustic spectroscopy (FTIR-PAS) has been applied for the first time to the identification and speciation of bacterial spores. A total of forty specimens representing five strains of Bacillus spores (Bacillus subtilis ATCC 49760, Bacillus atrophaeus ATCC 49337, Bacillus subtilis 6051, Bacillus thuringiensis subsp. kurstaki, and Bacillus globigii Dugway) were analyzed. Spores were deposited, with minimal preparation, into the photoacoustic sample cup and their spectra recorded. Principal component analysis (PCA), classification and regression trees (CART), and Mahalanobis distance calculations were used on this spectral library to develop algorithms for step-wise classification at three levels: (1) bacterial/nonbacterial, (2) membership within the spore library, and (3) bacterial strain. Internal cross-validation studies on library spectra yielded classification success rates of 87% or better at each of these three levels. Analysis of fifteen blind samples, which included five samples of spores already in the spectral library, two samples of closely related Bacillus globigii 01 spores not in the library, and eight samples of nonbacterial materials, yielded 100% accuracy in distinguishing among bacterial/nonbacterial samples, membership in the library, and bacterial strains within the library.

Identification of Sporulated and Vegetative Bacteria Using Statistical Analysis of Fourier Transform Mid-infrared Transmission Data

A combined mid-infrared spectroscopic/statistical modeling approach for the discrimination and identification, at the strain level, of both sporulated and vegetative bacterial samples is presented. Transmission mode spectra of bacteria dried on ZnSe windows were collected using a Fourier transform mid-infrared (FT-IR) spectrometer. Five Bacillus bacterial strains (B. atrophaeus 49337, B. globigii Dugway, B. thuringiensis spp. kurstaki 35866, B. subtilis 49760, and B. subtilis 6051) were used to construct a reference spectral library and to parameterize a four-step statistical model for the systematic identification of bacteria. The statistical methods used in this initial feasibility study included principal component analysis (PCA), classification and regression trees (CART), andMahalanobis distance calculations. Internal cross-validation studies successfully classified 100% of the samples into their correct physiological state (sporulated or vegetative) and identified 67% of the samples correctly as to their bacterial strain. Analysis of thirteen blind samples, which included reference and other bacteria, nonbiological materials, and mixtures of both nonbiological and bacterial samples, yielded comparable accuracy. The primary advantage of this approach is the accurate identification of unknown bacteria, including spores, in a matter of minutes.

Inverse kriging to enhance spatial resolution of imagery

We describe a unique approach to image resolution enhancement, inverse kriging (IK), which takes advantage of the spatial relationship between high- and low-resolution images within an area of overlap. Once established, this mathematical relationship then can be applied across the entire low-resolution image to significantly sharpen the image. The mathematical relationship uses the spatial correlations within the low-resolution image and between the low and high spatial-resolution imagery. Two of the most important requirements of the technique are that the images be co-registered well within the resolution of the larger pixels and that the spatial structure of the training area (where the spatial correlation statistics are computed) is similar to the structure of the remaining image area where it will be applied. Testing was performed using same-sensor and multi-sensor imagery. We show results that indicate that the method does improve the low spatial-resolution imagery. The selection of a training area spatial structure similar to the area being processed is important, as areas with different spatial structure (e.g., vegetation versus buildings and roads) will produce poor results. Comparisons with bilinear interpolation demonstrate that IK could be used as an improved interpolation tool, for example, in the image-registration process.

Graph-Based Analysis of Nuclear Smuggling Data

Much of the data that is collected and analyzed today is structural, consisting not only of entities but also of relationships between the entities. As a result, analysis applications rely on automated structural data mining approaches to find patterns and concepts of interest. This ability to analyze structural data has become a particular challenge in many security-related domains. In these domains, focusing on the relationships between entities in the data is critical to detect important underlying patterns. In this study we apply structural data mining techniques to automate analysis of nuclear smuggling data. In particular, we choose to model the data as a graph and use graph-based relational learning to identify patterns and concepts of interest in the data. In this article, we identify the analysis questions that are of importance to security analysts and describe the knowledge representation and data mining approach that we adopt for this challenge. We analyze the results using the Russian nuclear smuggling event database.

FTIR transmission and photoacoustic spectroscopy for the statistical identification of bacteria

We have previously reported a combined mid-infrared spectroscopic/statistical modeling approach for the discrimination and identification, at the strain level, of both sporulated and vegetative bacteria. This paper reports on the expansion of the reference spectral library: transmissive Fourier-transform mid-infrared (trans-FTIR) spectra were obtained for three Escherichia bacterial strains (E. coli RZ1032, E. coli W3110, and E. coli HB101 ATCC 33694), and two Pseudomonas putida bacterial strains (P. putida 0301 and P. putida ATCC 39169). These were combined with the previous spectral data of five Bacillus bacterial strains (B. atrophaeus ATCC 49337, B. globigii Dugway, B. thuringiensis spp. kurstaki ATCC 35866, B. subtilis ATCC 49760, and B. subtilis 6051) to form an extended library. The previously developed four step statistical model for the identification of bacteria (using the expanded library) was subsequently used on blind samples including other bacteria as well as non-biological materials. The results from the trans-FTIR spectroscopy experiments are discussed and compared to results obtained using photoacoustic Fourier-transform infrared spectroscopy (PA-FTIR). The advantages, disadvantages, and preliminary detection limits for each technique are discussed. Both methods yield promising identification of unknown bacteria, including bacterial spores, in a matter of minutes.

VIM: A Platform for Violent Intent Modeling

Radical and contentious activism may or may not evolve into violent behavior depending on contextual factors related to social, political, cultural and infrastructural conditions. Significant theoretical advances have been made in understanding these contextual factors and the import of their interrelations. However, there has been relatively little progress in the development of processes and capabilities that leverage such theoretical advances to automate the anticipatory analysis of violent intent. In this paper, we describe a framework that implements such processes and capabilities, and discuss the implications of using the resulting system to assess the emergence of radicalization leading to violence.

Hierarchical multi-scale approach to validation and uncertainty quantification of hyper-spectral image modeling

Validating predictive models and quantifying uncertainties inherent in the modeling process is a critical component of the HARD Solids Venture program [1]. Our current research focuses on validating physics-based models predicting the optical properties of solid materials for arbitrary surface morphologies and characterizing the uncertainties in these models. We employ a systematic and hierarchical approach by designing physical experiments and comparing the experimental results with the outputs of computational predictive models. We illustrate this approach through an example comparing a micro-scale forward model to an idealized solid-material system and then propagating the results through a system model to the sensor level. Our efforts should enhance detection reliability of the hyper-spectral imaging technique and the confidence in model utilization and model outputs by users and stakeholders.

MIST Final Report: Multi-sensor Imaging Science and Technology

The Multi-sensor Imaging Science and Technology (MIST) program was undertaken to advance exploitation tools for Long Wavelength Infra Red (LWIR) hyper-spectral imaging (HSI) analysis as applied to the discovery and quantification of nuclear proliferation signatures. The program focused on mitigating LWIR image background clutter to ease the analyst burden and enable a) faster more accurate analysis of large volumes of high clutter data, b) greater detection sensitivity of nuclear proliferation signatures (primarily released gasses) , and c) quantify confidence estimates of the signature materials detected. To this end the program investigated fundamental limits and logical modifications of the more traditional statistical discovery and analysis tools applied to hyperspectral imaging and other disciplines, developed and tested new software incorporating advanced mathematical tools and physics based analysis, and demonstrated the strength and weaknesses of the new codes on relevant hyperspectral data sets from various campaigns. This final report describes the content of the program and the outlines the significant results.

Consequences of Mixed Pixels on Temperature Emissivity Separation

This report investigates the effect that a mixed pixel can have on temperature/emissivity seperation (i.e. temperature/emissivity estimation using long-wave infra-red data). Almost all temperature/emissivity estimation methods are based on a model that assumes both temperature and emissivity within the imaged pixel is homogeneous. A mixed pixel has heterogeneous temperature/emissivity and therefore does not satisfy the assumption. Needless to say, this heterogeneity causes biases in the estimates and this report quantifies the magnitude of the biases.