Login for PhD students/staff at UCPH      Login for others
Introduction to the theory and practice of Multiple-Point Statistics
Provider: Faculty of Science

Activity no.: 5422-17-05-31
Enrollment deadline: 10/08/2017
PlaceDepartment of Geoscience and Natural Resource Management
Date and time15.08.2017, at: 09:00 - 18.08.2017, at: 12:00
Regular seats20
ECTS credits4.00
Contact personKarsten Høgh Jensen    E-mail address: khj@ign.ku.dk
Enrolment Handling/Course OrganiserKarsten Høgh Jensen    E-mail address: khj@ign.ku.dk
Written languageEnglish
Teaching languageEnglish
Course workload
Course workload categoryHours
Preparation / Self-Study60.00
Course hours45.00
Evaluation/reporting20.00

Sum125.00


Content


Aim and content

Over the last 20 years, multiple-point statistics (MPS) became a very active field of research for stochastic reservoir modeling (see for example the special issues on MPS published in Mathematical Geoscienecs in 2015).

The aim of this course is to enable users to understand and use the full power of Multiple Point Statistics method.

The main strength of MPS techniques is that they allow the user to incorporate rather precisely a conceptual model within a stochastic simulation framework. The conceptual model can be based on very general geological knowledge, or it can also be based on pre- existing large datasets typically from analog situations allowing to infer high order statistics. MPS methods are particularly well suited for geological modeling when connectivity patterns are expected to play a key role. MPS can simulate both facies or continuous properties such as permeability or porosity.

The course will include the basic theoretical knowledge required to understand how the methods work and practical exercises based essentially on the use of the Direct Sampling algorithm (available as an SGEMS plugin). As compared to other MPS methods, the Direct Sampling allows simulating multivariate data sets, continuous and categorical variables, and account for non stationarity both in the training data as well as in the simulation.

Audience

Participants are expected to have some background knowledge on statistics and its application in geology or hydrology. No pre-requisite is expected in geostatistics or reservoir modeling.

Students will be awarded 4 ECTS for participation in the course.

Lecturers

Philippe Renard and Grégoire Mariethoz have been working on the development of MPS techniques during the last 10 years. They were involved in the development of several different pixel based or patch algorithms including IMPALA, DEESSE, RPA, etc. They applied those methods to a variety of field of applications including geological modeling, simulation of rainfall, climatic variables, hyperspectral imaging, etc.

Location

Department of Geosciences and Natural Resource Management, University of Copenhagen, Øster Voldgade 10, 1350 Copenhagen K.

Schedule

The course will start every morning at 9 am and end around 5 pm every day.

Day 1.

Morning. General introduction:

  • Overview of the course
  • Basic geostatistical concepts (covariance, simulation, estimation, indicator variables)
  • Overview of different types of models for geological heterogeneity
  • Why using MPS and how it works
  • Overview of some MPS algorithms
  • How to use and parametrize DeeSse

Afternoon. Practical session using SGeMS:

  • Introduction to SGeMS
  • Sequential Indicator Simulations on a discrete case
  • Understanding DeeSse parameters in a categorical case and parameter sensitivity
  • Comparing DeeSse and Sequential Indicator Simulations on a discrete case

Day 2.

Morning. Non stationarity:

  • Requirements of a training image / how to get it and what should be its properties
  • Simulation of categories with trends / rotation / affinity / invariances
  • Accounting for local proportion maps / comparison with tau model
  • Regions with different training images / or parameters

Afternoon. Practical session using SGeMS:

  • How to create a training image
  • Generating trend maps
  • Simulation with trends / rotation / affinity
  • Global and local proportion maps

Day 3.

Morning. Advanced possibilities and applications offered by DeeSse

  • Direct Sampling in the continuous case

  • Extension to multivariate problems

  • Handling non stationarity in the training image (e.g. analog data, outcrop scan, or previous model)

  • Conditioning continuous MPS simulations with averaged block data (ex. tomography)

Afternoon. Practical session using SGeMS:

  • Time series simulation and filtering

  • Using a non stationary training image

  • Reconstruction of incomplete data sets

  • Multivariate data sets and alternative applications (remote sensing, climate data)

  • A short introduction to Conditional Image Quilting

Day 4.

Morning. Synthesis of the first 3 days: the overall MPS workflow

  • Practical recommendations and discussion
  • From TI selection and design to quality checks

Late morning + early afternoon. Practical session using SGeMS:

  • A geological case study

Afternoon. Closing:

  • An overview of current research topics and new algorithms
  • Final open discussion and course evaluation

Course fee

The course is free of charge for national and international PhD students.

Participants are requested to secure funding for their own travel, accommodation and meals.

Registration

Please use this link http://www.hobecenter.dk/index.php/registration/meetings/introduction-to-the-theory-and-practice-of-multiple-point-statistics


Search
Click the search button to search Courses.


Course calendar
See which courses you can attend and when
JanFebMarApr
MayJunJulAug
SepOctNovDec



Publication of new courses
All planned PhD courses at the PhD School are visible in the course catalogue. Courses are published regularly.