This is not a website that explains fundamental geostatistical principles in a series of lessons. There are many classes, short courses, academic papers, and books on geostatistics that provide resources for learning and reviewing the details of specific aspects of geostatistics. The aim of Geostatistics Lessons is instead to explain how to formulate a solution methodology and how to make different implementation decisions.
The concept of Geostatistics Lessons was initiated at the Centre for Computational Geostatistics. The early attempt at a Guide to Best Practices failed because of a too-large scope; these Lessons provide the needed guidance in manageable units.
Geostatistics provides a probabilistic view to spatial prediction. The true spatial distribution of geologic properties is the result of a complex succession of physical, chemical and biological processes. A limited number of data are collected with error. Uncertainty arises because of this limited sampling and geological variability at all scales. We fall back on probabilistic tools to quantify uncertainty and generate numerical models of what the spatial distribution of geological properties might be like.
Geostatistics provides many different tools for specific problems in spatial prediction. These tools come from different fields of application, different amounts of data, different world views and schools of thought, and different levels of software and mathematical sophistication. There are 10s of books, 10s of different software packages, 1000s of papers and many more ways to make mistakes in the application of geostatistics. The responsible practitioner of geostatistics wishes to apply the appropriate subset of geostatistical tools in the best manner possible. They wish to apply the correct amount of professional and computer resources to meet the project objectives. They want to be open and transparent in their modeling decisions. The resulting numerical models and decisions are to be repeatable, thoroughly checked and auditable. There are limitations and the results may not be applicable for all purposes, but the numerical model is fit for the purpose it was intended and the limitations and future work are well documented. Although other experts may have approached certain aspects differently, there is an understanding and general agreement that the approach, software tools and implementation decisions meet with current best practice. Specific software tools are discussed where reasonable, but the focus of Lessons is on the required choices and decisions in a geostatistical workflow.
There are few places for the practitioner to find specific advice on best practice. There are proprietary best practice documents in certain companies. Senior technical staff members often approve best practice and mentor junior staff. The career chain from apprentice to journeyman to master is well established and has proven effective. This system has great value, but is inefficient given a relative lack of masters, the fast pace of technological advancement and the mobility of professionals. Geostatistics Lessons is an open transparent disclosure of what constitutes best practice in geostatistical modeling reviewed by a team of editors.
It is not possible to consider all geological situations, problem settings, project goals and data. There are alternative schools of thought with equally valid approaches to problems. There are many ways to solve a particular problem and there are site-specific factors to consider including historical, social and cultural considerations. Nevertheless, there is value in experienced geostatisticians disclosing their belief of what constitutes best practice. Those beliefs will change as experience is gained; Geostatistics Lessons will be updated as new lessons are authored, revised and reviewed.
Jared Deutsch (Canada) http://jareddeutsch.com/