2011-2012 IAMG Distinguished Lecturer Series

Amilcar Soares, the Distinguished Lecturer for 2011-2012, is a Professor at the Instituto Superior Técnico in Portugal and is also head of the Centro de Modelização de Reservatórios Petrolíferos at IST.  He has been extremely active in promoting mathematical geology, worldwide, particularly geostatistics, since the 1980’s. Soares is one of the world leaders in applying geostatistics in environmental engineering with recent work on characterizing desertification and he is making considerable impact on applications to practical problems, not just in theoretical developments. Amilcar has also organized one of the Geostat Congresses and two of the geoENV conferences and has been an IAMG member for many years.
More information can be found on the web pages of Centro de Modelização de Reservatórios Petrolíferos:  http://cmrp.ist.utl.pt/

Anyone interested in hosting Dr. Soares at their institution, please contact the Chairman of the DL Committee, Sean McKenna, at: samcken@sandia.gov

The monitoring and management of environmental projects of soil contamination and soil degradation, air pollution in urban and industrial sites, surface water and groundwater contamination usually share common problems and challenges related with the complexity of natural phenomena involved, as well as a general scarcity of available information. Geostatistical methodologies have been widely employed in attempting to tackle these problems, namely with regard to natural resource evaluation, development of sampling strategies, characterizing hot-spots or time periods with high pollutant concentrations, and management of the uncertainties and risks of different phases of the project.

This presentation will provide the non-specialists with the basic geostatistical concepts behind the use of stochastic simulations in the assessment of uncertainty and risk.

Desertification is a phenomenon that is affecting two-thirds of the planet. Most of the affected areas belong, geographically and politically, to developing countries with almost no resources to combat its effects. It is extremely important to identify local or regional desertification indicators in order to observe, control and manage the critical situations. However, because local resources are generally not available for their observation and analysis these indicators become worthless as a tool to combat desertification.

Hence Earth Observation data (primarily no-cost satellite images) are a fundamental tool for monitoring the dynamics of desertification phenomena in order to manage critical situations in those countries.

In this presentation, Geostatistical and image analysis classification methods used to characterize the spatial and temporal behavior of individual biophysical factors of desertification (climate, vegetation, water and soil) are discussed. Based on the dynamics of extreme climatic factors (droughts and floods) and the behavior of vegetation and soil through time for the crucial land use classes, joint geostatistical interpolation methods produce regional indicators of desertification. Real case studies of selected semi-arid regions in Portugal, Mozambique and Brasil are presented in order to illustrate those methodologies. This is the kernel of DESERTWATCH (Extended), a project supported by the European Space Agency, which is a tool to characterize desertification indicators at the global scale.

Geostatistics has been used over the last decades as a tool for pollutant characterization and uncertainty assessment in a wide range of environmental applications, like polluted soil sites, air quality, surface water and groundwater contamination, (geoENV conferences proceedings). Geostatistical models basically share an identical framework, i.e., assuming the variables of interest as spatial, or space-time, stationary random functions. Measures of uncertainty and mean behavior of those variables are determined through local or global conditional distribution functions (cdfs) by using estimation (indicator or multiGaussian kriging) or stochastic simulation algorithms. Cost functions and risk analysis are then derived from these local or global cdfs.

However, more complex phenomena cannot be satisfactorily represented with those common approaches, mostly when they present a determinant dynamic component that is hard to directly account for in geostatistical models as in, for example, the flow of contaminants at the surface, in different soil layers or in deep aquifers, or the dynamics of contaminant plumes in air in industrial or urban areas. Usually, in order to integrate the influence of such dynamic factors, deterministic dynamic simulators are used separately in inverse modeling approaches, such as in groundwater problems, to “calibrate” and update parameters that have been characterized by geostatistical techniques.

This presentation addresses some new proposals to approach such problems. The integration of main dynamic characteristics through hybrid models is presented. This presentation is guided by two real case studies: i) The main vectors of a fluid flow (given by a dynamic simulator), are used to characterize local anisotropy behavior of the phenomenon in the stochastic simulation of the main sediment pollutants in a coastal lagoon (Horta et al, 2010); ii) The results of a Gaussian plume simulation of air pollution emissions of an industrial area are downscaled with multi-scale simulation (Liu and Journel, 2008) by using local point data of monitoring stations and block data from the deterministic model (Pereira, et al., 2010 ). This algorithm shows to be a nice solution for mitigating the non-exact and coarse scale results of deterministic Gaussian plume simulations, which is a very common practice in air quality characterization.

Geostatistics has been commonly used in forward modeling and in inverse modeling to integrate seismic information in stochastic fine grid models. The quality of seismic data, the downscaling of seismic attributes to the fine grid of the well measurements are still valuable challenges to which existing geostatistical methods only give partial answers.

Seismic inversion is an established geophysical technique whereby the rock property of acoustic impedance is estimated directly from seismic amplitude data. Seismic inversion is based on an iterative approach to minimize an objective function – the difference between the forward convolution of the reflectivity of an impedance model and the real seismic amplitudes.

In this presentation a new seismic inversion methodology – global stochastic inversion GSI - is proposed based on i) direct sequential simulation and co-simulation approaches for “transforming” 3D images of acoustic impedances in an iterative process, and ii) following the sequential procedure of the genetic algorithms optimization to converge the transformed images towards an objective function.

This is presented in two different situations of seismic inversion: acoustic inversion, where the GSI is applied to a post stack seismic signal; and elastic inversion where the stochastic inversion is applied to a pre-stack seismic. In this last framework, a new method for reproducing the bi-distribution of acoustic impedances Ip and Is is presented. In both acoustic and elastic inversion approaches of GSI, maps of final porosity and corresponding uncertainty are obtained.

Case studies of Brasilian, African and Middle East reservoirs are presented.

last updated 2011-01-08