Geo-intelligence remote sensing platforms situated over spatially diverse areas are often tasked with geo-intelligence surveillance and adversarial monitoring for military organizations. Limited resources disallow continuous sampling of local areas at the same time, necessitating a need for smart sensing of diverse environments according to a rational evidence-based rule. Such algorithms should not only provide insight into which local region should be focused on, but should also facilitate decisions as to which environmental features should be measured over time once a local site has been selected. Multicomponent optimal learning observational arrays are demonstrated using numerically simulated data of turbulent flow to show not only the feasibility of how individual observational platforms should be chosen in a Bayesian sense, but also how goal state directed sampling of complex systems or turbulent processes over local regions can be accomplished. A Bayesian amalgamation algorithm guides which observational arrays perform knowledge gradient policy based optimal learning to smartly sample observations in local regions. Machine learning and operations research algorithms function as data agnostic, Bayesian processors demonstrating how geo-intelligence information can be efficiently captured to help solve data-driven problems.
KEYWORDS: Data modeling, Matrices, Statistical analysis, Statistical modeling, In situ metrology, Data acquisition, Sensors, Machine learning, Data processing, Systems modeling
Data characterization of an eight-site nitrate-level time-series array using a suite of intra- and inter-site dimensional reduction and analysis algorithms was performed as the preliminary stage of a full Bayesian state-estimation approach for understanding the Illinois section of the Mississippi watershed. Preliminary analysis shows high mean nitrate levels in the northern, western, and southern parts of the Illinois watershed with significant correlations of nitrate levels appearing not only in the southern region, but also across a north-south transect. Intra-site dimensional reduction of the eight-site array, based on empirical orthogonal function analysis and nonnegative matrix factorization, demonstrates that specific time series, lower in number than the eight-site dimension, are responsible for both global and local variability. Inter-site dimensional reduction based on Gaussian mixture modeling applied to sets of dual-site time series in the north and south shows multimodal clusters characterized by mean and covariance information. Competitive-leaky-learning-based intersite data group modeling depicts nonlinearly generated data clusters possessing labels also based on distinct mean and covariance structure. Hidden Markov model parameter estimation applied to dual time-series sets across northern and southern regions, and over two different seasonal time scales, provides emission matrix tables with maximum probability trends consistent with the results from Gaussian mixture modeling. All facets of the machine-learning results offer a means for quantitatively describing the Illinois watershed’s nitrate-level dynamics over a fall-winter seasonal time scale.
Environmental engineering remote sensing platforms using hyperspectral imagery are often responsible for monitoring coastal regions in order to safeguard national waters. This objective requires determining subsurface turbulent structure from surface water spatial measurements for flow state assessment and decision-making. The inability of remote sensing platforms to penetrate the water column at depth because of turbulence-induced sediment-concentration modulation necessitates using models that dynamically link surface and subsurface structures. A hidden Markov model is applied to large-eddy simulated three-dimensional turbulent flow for the purpose of exploring the feasibility of constructing a system model possessing diagnostic/prognostic statistical power for turbulent state evolution. The data-driven model is based on machine-learning techniques that rely on data statistical covariance structure. Initial results suggest strong nonlinear coupling between the mean flow directed vorticity, cross mean flow velocity, and sediment concentration. In addition, a Bayesian-based state-action estimation algorithm is employed that demonstrates which turbulent feature variables should be focused on at specific times, given the desire to reach a known goal state, and given only a limited number of observations. Such a model gives experimentalists time- and resource-saving guidance for determining what turbulent variables to measure at different times in order to reach a known turbulent goal state.
Nonnegative matrix factorization-based feature selection analysis performed on land based hyperspectral imagery of the Mississippi river identifies ten spectral bands in the visible and near infrared portion of the electromagnetic spectrum that are significant contributors to the resulting structural image clustering of sediment-laden water. Different distance metrics provide clear evidence of the potency of these spectral bands for class separation of turbid, sediment-laden water from clear water, provided that the data contains low noise. In addition, feature ranking of spectral band subsets of the identified characteristic spectral bands allows insight into the relative importance of smaller spectral band subsets for water-sediment characterization. Results support present day multispectral satellite design methods for land-water imagery where payload power resources are relegated to certain spectral bands at the expense of others.
Direct numerically simulated data can serve as a proxy for understanding many issues concerning multidimensional remotely sensed data. As a step towards performing operational Bayesian belief network modeling for rivers, which is of practical utility to naval intelligence, direct numerically simulated sediment-laden oscillatory flow is used to estimate statistical surface layer spatial eddy scales. This is done using spatial realizations of the sediment concentration, vertical velocity, and pressure fields along with feature extraction algorithms which utilize self-organizing mapping, independent component analysis, and two-dimensional omnidirectional Morlet wavelet analysis. Stress versus scale distributions exhibit distinct phase modulation over the three ambient forcing phases of maximum negative velocity, zero velocity, and maximum positive velocity. The stress versus sediment concentration scale distribution, which is of great pertinence to riverine remote sensing, exhibits a significant amount of large eddy scales suggesting coherent large-scale sediment structure formation possibly due to particle interstitial forces. The estimated statistical results can serve as feature parameters for naïve Bayesian belief network prediction of bottom boundary layer stress from surface eddy scale observations.
Naïve Bayesian belief network modeling is applied to direct numerically simulated imagery of oscillatory sedimentladen flow to illustrate the feasibility of creating a system model which captures the statistical interrelationship of the surface layer sediment concentration, pressure, and vertical velocity eddy scales with the sub-surface Reynolds stress. From a prognostic reasoning viewpoint, preliminary model results suggest that large sediment concentration eddy scales may result from the application of large positive Reynolds stress. However, from a diagnostic reasoning viewpoint, initial results suggest that robustly inferring sub-surface boundary layer stress from surface sediment concentration eddy scales may be a difficult task. The model formulism used allows for the ability to statistically characterize flow structure at depth from observations taken across a surface boundary layer, making the results relevant to image analysis at the airsea interfacial boundary layer in large-scale coastal and riverine systems.
A nonlinear cluster analysis algorithm is used to characterize the spatial structure of a wind-sheared turbulent flow obtained from the direct numerical simulation (DNS) of the three-dimensional temperature and momentum fields. The application of self-organizing mapping to DNS data for data reduction is utilized because of the dimensional similitude in structure between DNS data and remotely sensed hyperspectral and multispectral data where the technique has been used extensively. For the three Reynolds numbers of 150, 180, and 220 used in the DNS, self-organized mapping is successful in the extraction of boundary layer streaky structures from the turbulent temperature and momentum fields. In addition, it preserves the cross-wind scale structure of the streaks exhibited in both fields which loosely scale with the inverse of the Reynolds number. Self-organizing mapping of the along wind component of the helicity density shows a layer of the turbulence field which is spotty suggesting significant direct coupling between the large and small-scale turbulent structures. The spatial correlation of the temperature and momentum fields allows for the possibility of the remote extrapolation of the momentum structure from thermal structure.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.