Entry Date:
April 30, 2012

Data Assimilation

Principal Investigator Dennis McLaughlin

Co-investigators F Morgan , William L Rodi

Data Assimilation: also known as Linear and Non-Linear Regression, Data and Models, Optimization, Parameter Estimation, Inference from Data and Models, Inverse Problems, etc. is an area of teaching and research that permeates virtually all our studies in ERL.

In Geophysics and related fields such as Reservoir Science, we are generally taking inadequate, insufficient and inaccurate data often only on the surface of earth or in a limited number of boreholes. The Inverse Problem is to obtain the best possible model based on these constraints in the data. The results are usually non-unique solutions with large associated errors because of the poor quality of the assumed model(s) and the lack of complete and sensitive data. The art of the subject is to attempt to merge this difficult situation with other information such as from geology, or jointly with other data (eg. seismic and EM) to obtain a more unique solution with improved accuracy.

There are two fundamental courses taught in EAPS in this subject area, namely: Data and Models (formerly taught as Inverse Problems: is the oldest course of its kind at MIT), Inference from Data and Models. Because of the breath of these subjects they draw graduate students from virtually all departments at MIT. Furthermore, there are other related classes in this broad subject taught in engineering and economics/management.

Note that the Data and Models has also been taught successfully for over 15 years as a one-week MIT Professional Institute Summer course to participants primarily from varied industries. The course can be taught at industrial or national lab locations at great cost savings.