Simpler models in environmental studies and predictions

Authors: HONG, EUNMI - Orise Fellow; Pachepsky, Yakov; WHELAN, GENE - Us Environmental Protection Agency (EPA)

Submitted to: Critical Reviews in Environmental Science Technology
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 10/17/2017
Publication Date: 12/13/2017
Citation: Hong, E., Pachepsky, Y.A., Whelan, G. 2017. Simpler models in environmental studies and predictions. Critical Reviews in Environmental Science Technology. 47(18):1669-1712.

Interpretive Summary: Models became the staple of environmental research and management. Different conceptualization and modeling tools that vary in scale and detail can describe the same phenomena. Both simplicity and complexity in models have advantages and disadvantages. There are similarities in development and application of simpler models across environmental science and technology. The objective of this work was to present the summary of model simplification methods and results. We focused on approaches that addressed the basic questions of (a) why and when to use simpler models? (b) how to build simpler models? and (c) what are the relationships between simple and complex models? We reviewed the major directions and techniques of simple model development both based and not based on the original complex model, and provided recommendations for such development. We expect this work to be useful for wide group of environment-minded professionals in that it summarizes the experience and provides guidance in the burgeoning field of research and management activity.

Technical Abstract: Current computational opportunities have not eliminated basic questions of (a) why and when to use simpler models? (b) how to build simpler models? and (c) what are the relationships between simple and complex models? Answers continue to evolve over time, and this work aims to provide a synopsis of this evolution in the environmental modeling. The major directions of simple model development include metamodeling, statistical regression-based and related empirical models, and mechanistic models with reduced structures. Applications of process-based models encounter a range of problems stemming from limitations of model equations relative to a heterogeneous reality, lack of a theory of subgrid-scale integration, practical constraints on solution methodologies, and dimensionality in parameter calibration Simpler models may also be favoured because of limited observational data, error propagation, large uncertainty in many environmental modelling projects, intent to use the model as a component of a multimedia or multi-compartmental model. Differences between complex and simple models often become clearer when one distinguishes between their accuracy and reliability. Decision-making often relies on simple models. Model simplification can be useful in understanding the behavior of complex models. A relatively simple suite of models can capture big issue problems. This is best accomplished when discipline-specific models are seamlessly linked to facilitate communication, and the modeling framework is specifically designed to allow users to construct workflows. Simpler model development follows one of three approaches: 1) independent, 2) simplification of an existing model (model abstraction), or 3) approximation of a base model output to construct a metamodel. Understanding the existence of models of different complexity, model comparisons and simplifications are an important part of the modern epistemology of environmental science and technology.