Bayesian Surrogate Analysis and Uncertainty Propagation

The quantification of uncertainties of computer simulations due to input parameter uncertainties is paramount to assess a model’s credibility. For computationally expensive simulations, this is often feasible only via surrogate models that are learned from a small set of simulation samples. The surrogate models are commonly chosen and deemed trustworthy based on heuristic measures, and substituted for the simulation in order to approximately propagate the simulation input uncertainties to the simulation output. In the process, the contribution of the uncertainties of the surrogate itself to the simulation output uncertainties is usually neglected. In this work, we specifically address the case of doubtful surrogate trustworthiness, i.e., non-negligible surrogate uncertainties. We find that Bayesian probability theory yields a natural measure of surrogate trustworthiness, and that surrogate uncertainties can easily be included in simulation output uncertainties. For a Gaussian likelihood for the simulation data, with unknown surrogate variance and given a generalized linear surrogate model, the resulting formulas reduce to simple matrix multiplications. The framework contains Polynomial Chaos Expansions as a special case, and is easily extended to Gaussian Process Regression. Additionally, we show a simple way to implicitly include spatio-temporal correlations. Lastly, we demonstrate a numerical example where surrogate uncertainties are in part negligible and in part non-negligible.

Estimating complex ecological variables at high resolution in heterogeneous terrain using multivariate matching algorithms

1. Simulation models are valuable tools for estimating ecosystem structure and function under various climatic and environmental conditions and disturbance regimes, and are particularly relevant for investigating the potential impacts of climate change on ecosystems. However, because computational requirements can restrict the number of feasible simulations, they are often run at coarse scales or for representative points. These results can be difficult to use in decision-making, particularly in topographically complex regions.2. We present methods for interpolating multivariate and time series simulation output to high resolution maps. First, we developed a method for applying k-means clustering to optimize selection of simulation sites to maximize the area represented for a given number of simulations. Then, we used multivariate matching to interpolate simulation results to high-resolution maps for the represented area. The methods rely on a user-defined set of matching variables that are assigned weights such that matched sites will be within a prescribed range for each variable. We demonstrate the methods with case studies using an individual-based plant simulation model to illustrate site selection and an ecosystem water balance simulation model for interpolation.3. For the site-selection case study, our approach optimized the location of 200 simulation sites and accurately represented 96% of a large study area (1.12 x 106 km2) at a 30-arcsecond resolution. For the interpolation case study, we generated high-resolution (30-arcsecond) maps across 4.38 x 106 km2 of drylands in western North America from simulated sites representing a 10 x 10 km grid. Our estimates of interpolation errors using leave-one-out cross validation were low (<10% of the range of each variable).4. Our point selection and interpolation methods provide a means of generating high-resolution maps of complex simulation output (e.g., multivariate and time-series) at scales relevant for local conservation planning and can help resolve the effects of topography that are lost in simulations at coarse scales or for representative points. These methods are flexible and allow the user to identify relevant matching criteria for an area of interest to balance quality of matching with areal coverage to enhance inference and decision-making in heterogenous terrain.

Quantitative Integration of 4D Seismic with Reservoir Simulation

Summary Standard history-matching workflows use qualitative 4D seismic observations to assist in reservoir modeling and simulation. However, such workflows lack a robust framework for quantitatively integrating 4D seismic interpretations. 4D seismic or time-lapse-seismic interpretations provide valuable interwell saturation and pressure information, and quantitatively integrating this interwell data can help to constrain simulation parameters and improve the reliability of production modeling. In this paper, we outline technologies aimed at leveraging the value of 4D for reducing uncertainty in the range of history-matched models and improving the production forecast. The proposed 4D assisted-history-match (4DAHM) workflows use interpretations of 4D seismic anomalies for improving the reservoir-simulation models. Design of experiments is initially used to generate the probabilistic history-match simulations by varying the range of uncertain parameters (Schmidt and Launsby 1989; Montgomery 2017). Saturation maps are extracted from the production-history-matched (PHM) simulations and then compared with 4D predicted swept anomalies. An automated extraction method was created and is used to reconcile spatial sampling differences between 4D data and simulation output. Interpreted 4D data are compared with simulation output, and the mismatch generated is used as a 4D filter to refine the suite of reservoir-simulation models. The selected models are used to identify reservoir-simulation parameters that are sensitive for generating a good match. The application of 4DAHM workflows has resulted in reduced uncertainty in volumetric predictions of oil fields, probabilistic saturation S-curves at target locations, and fundamental changes to the dynamic model needed to improve the match to production data. Results from adopting this workflow in two different deepwater reservoirs are discussed. They not only resulted in reduced uncertainty, but also provided information on key performance indicators that are critical in obtaining a robust history match. In the first case study presented, the deepwater oilfield 4DAHM resulted in a reduction of uncertainty by 20% of original oil in place (OOIP) and by 25% in estimated ultimate recoverable (EUR) oil in the P90 to P10 range estimates. In the second case study, 4DAHM workflow exploited discrepancies between 4D seismic and simulation data to identify features necessary to be included in the dynamic model. Connectivity was increased through newly interpreted interchannel erosional contacts, as well as subseismic faults. Moreover, the workflow provided an improved drilling location, which has the higher probability of tapping unswept oil and better EUR. The 4D filters constrained the suite of reservoir-simulation models and helped to identify four of 24 simulation parameters critical for success. The updated PHM models honor both the production data and 4D interpretations, resulting in reduced uncertainty across the S-curve and, in this case, an increased P50 OOIP of 24% for a proposed infill drilling location, plus a significant cycle-time savings.

Endogenous Infrastructure Development and Spatial Takeoff in the First Industrial Revolution

This paper develops a model in which the evolution of the transport sector occurs alongside the growth in trade and output of agricultural and manufacturing firms. Simulation output captures aspects of the historical record of England and Wales over 1710–1881. A number of counterfactuals demonstrate the role that the timing and spatial distribution of infrastructure development play in determining the timing of takeoff. There can be a role for policy in accelerating takeoff through improving infrastructure, but the spatial distribution of that improvement matters. (JEL H54, N53, N63, N73, N93, R12, R42)

Mutimodal Transpose Rotation Mobius Transformation Based Cancellable Template Generation Technique

Cancelable biometric is such a template security conspire, that replaces a biometric template when the stored template is taken or lost. It is a feature level area transformation where a misshaped variant of a biometric template is produced and coordinated in the transformed space. The issue persevere with the utilization of unique template can be abstained from utilizing cancellable biometrics. In this work, a nonexclusive structure has been intended for producing irreversible portrayal of templates of multimodal which depends on Mobius transformation on unique picture. So the template security is additionally improved. The simulation output of the proposed framework give better execution in Identification of clients. Another strategy called "MTRMT" is proposed to address the issue of stored Templates. The Proposed epic strategy has been assessed with the ongoing fingerprint got from 50 volunteers of veerapandi village in Coimbatore locale. The exploratory outcome shows the better execution of the proposed framework.