Model-data interaction in groundwater studies: Review of methods, applications and future directions

Abstract

We define model-data interaction (MDI) as a two way process between models and data, in which on one hand data can serve the modeling purpose by supporting model discrimination, parameter refinement, uncertainty analysis, etc., and on the other hand models provide a tool for data fusion, interpretation, interpolation, etc. MDI has many applications in the realm of groundwater and has been the topic of extensive research in the groundwater community for the past several decades. This has led to the development of a multitude of increasingly sophisticated methods. The progress of data acquisition technologies and the evolution of models are continuously changing the landscape of groundwater MDI, creating new challenges and opportunities that must be properly understood and addressed. This paper aims to review, analyze and classify research on MDI in groundwater applications, and discusses several related aspects including: (1) basic theoretical concepts and classification of methods, (2) sources of uncertainty and how they are commonly addressed, (3) specific characteristics of groundwater models and data that affect the choice of methods, (4) how models and data can interact to provide added value in groundwater applications, (5) software and codes for MDI, and (6) key issues that will likely form future research directions. The review shows that there are many tools and techniques for groundwater MDI, and this diversity is needed to support different MDI objectives, assumptions, model and data types and computational constraints. The study identifies eight categories of applications for MDI in the groundwater literature, and highlights the growing gap between MDI practices in the research community and those in consulting, industry and government.