Traditional versus structure-based model development strategies

Abstract Text:

An important challenge in agriculture modeling is deciding how to mathematically represent biological phenomena. The objective of this paper is to compare more traditional model development methods (e.g., empirical models) with structure-based modeling (SBM) such as system dynamics (SD). Substantial overlap exists between traditional and SBM approaches, but there are important differences. The overall steps of the modeling process and scientific rigor are quite similar, but their focus and implementation can differ substantially. The steps of both modeling approaches often comprise the (1) identification of a problem (research objective), (2) formulation of the mathematical (and/or statistical) statements, (3) data collection (experimentation), (4) model evaluation and quantitative analysis relevant to the modeling objectives. SBM often differs from traditional approaches in each of these phases such as defining the problem as the replication of observed dynamic behavioral modes (e.g., s-shaped growth or oscillations) rather than situational point prediction or statistical estimation of parameters (step 1), giving more attention to system structure based on cause-effect relationships in terms of the stock-flow (i.e., level variables and rate variables) and feedback processes that generate observed behavior and visualizing these relationships in causal loop diagrams (CLD) and stock and flow diagrams (SFD) (step 2), and data collection that encompasses a broader range of sources (experimental, secondary, expert opinion, participatory exercises) and may include concepts hypothesized to be important but for which limited data are available (step 3). Model evaluation criteria can also differ due to the intrinsic nature of SBM as greater focuses are given to behavioral mode replication and feedback loop dominance analysis (step 4). In general, traditional modeling approaches focus on defining analytical functions and their statistical consistency with observed biological responses, whereas SBM focus on the mechanistic explanations for system behaviors and the feedback relationships that led to them. For example, a traditional modeling approach could use a saturating function to describe movement of a substrate across a membrane, whereas SBM would focus on feedback processes that represent decreasing affinity of the membrane for that substrate as concentration increases. Although they can differ substantially in their implementation, these two mathematical modeling strategies should be viewed as complementary rather than competing tools.

Keywords:

Modeling, Simulation, Methodology