Mathematical models can be classified according to how their relationships are obtained. Mechanistic models are constructed using an understanding of the processes and mechanisms believed to govern a system. They often use physical laws, conservation principles, biological mechanisms or chemical reaction processes. For example, a mechanical model may be based on Newton's laws of motion. Empirical models are constructed primarily from observed data by identifying statistical or mathematical relationships between measured quantities. A simple regression equation fitted to experimental measurements is an empirical model. These categories are not completely separate. A model may be mechanistic in its main structure but use empirical parameters estimated from observations. Such models are often called semi-empirical or hybrid models. Mechanistic models may provide insight into why a system behaves in a particular way, while empirical models may provide accurate predictions within the range of the data used to develop them. The choice depends on the available scientific knowledge, data quality, modelling purpose and expected conditions of use.
Mechanistic and Empirical Models
Explains the distinction between mechanistic models based on underlying processes and empirical models based primarily on observed data.
An empirical linear model may be written as y = β₀ + β₁x + ε. A mechanistic model may be derived from governing principles such as a conservation equation or a physical law.
Explanation
Variables
| Symbol | Meaning |
|---|---|
y | Observed or modelled response. |
x | Explanatory variable. |
β₀ | Empirical intercept parameter. |
β₁ | Empirical slope parameter. |
ε | Random error or unexplained variation in the empirical model. |
Example or application
A model of falling motion derived from Newton's second law is mechanistic. A regression equation that estimates crop yield from measured rainfall and fertilizer use is empirical if its coefficients are fitted primarily from collected data.
Definition
A mechanistic model represents a system using mathematical descriptions of its underlying processes, whereas an empirical model represents relationships derived primarily from observed or measured data.
Purpose
To understand how models can be constructed from scientific mechanisms, observed data or combinations of both.
Importance
The distinction helps researchers choose appropriate modelling strategies. Mechanistic models can support causal explanations and extrapolation when their mechanisms are understood, while empirical models can be useful when reliable data exist but detailed mechanisms are unknown.
Learning Objectives
Define mechanistic and empirical models.
Explain how scientific principles contribute to mechanistic models.
Explain how observed data contribute to empirical models.
Compare the strengths and limitations of both approaches.
Recognize hybrid and semi-empirical models.
Key Concepts
Mechanistic model
Empirical model
Governing mechanism
Observed data
Regression
Parameter estimation
Hybrid model
Extrapolation
Working Principle
Mechanistic modelling starts from a proposed understanding of how the system works and translates that understanding into equations. Empirical modelling starts from observations and identifies mathematical relationships that describe the data. Hybrid models combine both approaches.
Process
Identify the system and modelling objective.
Determine whether relevant governing mechanisms are known.
Collect observations or experimental data.
Choose a mechanistic, empirical or hybrid approach.
Formulate the mathematical relationships.
Estimate unknown parameters where necessary.
Compare model predictions with independent observations.
Evaluate the model's explanatory ability and domain of validity.
Applications
Physics: mechanistic models derived from conservation laws and motion equations.
Agriculture: empirical crop-yield models based on rainfall and soil data.
Medicine: mechanistic models of physiological processes and empirical risk-prediction models.
Climate science: hybrid models combining physical laws with data-based parameterization.
Engineering: physical system models calibrated using experimental measurements.
Economics: empirical relationships estimated from historical observations.
Worked Example
Problem
A researcher wants to model crop yield as a function of rainfall. Explain how a mechanistic and an empirical approach could differ.
Given
Response variable: crop yield Y.
Explanatory quantity: rainfall R.
Possible additional factors: soil quality, temperature and fertilizer.
Solution
An empirical approach collects rainfall and crop-yield measurements from several growing seasons.
A regression model may then be fitted, such as Y = β₀ + β₁R + ε.
A mechanistic approach investigates how water availability affects soil moisture, plant physiology and crop development.
It may represent those processes using water-balance and plant-growth equations.
A hybrid model could use mechanistic equations for water availability and empirical relationships for uncertain plant-response parameters.
Answer
The empirical model describes a data-based relationship, while the mechanistic model attempts to represent the processes connecting rainfall to crop growth.
Interpretation
An empirical model may be easier to fit when data are available, whereas a mechanistic model may offer more insight into why the relationship occurs. Both require careful validation.
Common Misconceptions
| Misconception | Correction |
|---|---|
Empirical models are simply guesses. | Empirical models are built from observations and use systematic methods such as regression and statistical estimation. |
Mechanistic models never use experimental data. | Mechanistic models often require observations to estimate parameters and validate their predictions. |
A mechanistic model is automatically more accurate than an empirical model. | Accuracy depends on the quality of the mechanisms, assumptions, data and model evaluation. |
Mechanistic and empirical models cannot be combined. | Hybrid models combine theoretical mechanisms with data-based relationships. |
Practice Questions
| Type | Question | Difficulty |
|---|---|---|
classification | Classify a regression model fitted to observed house prices as primarily mechanistic or empirical. | beginner |
conceptual | Explain how Newton's laws can be used to construct a mechanistic model of motion. | intermediate |
comparison | Compare the advantages and limitations of mechanistic and empirical models. | intermediate |
design | Design a hybrid modelling approach for estimating crop growth using physical knowledge and experimental data. | advanced |
Key Takeaway
Mechanistic models represent how a system works, empirical models describe relationships observed in data, and hybrid models combine both. Selecting the right approach depends on scientific knowledge, data and modelling objectives.
Prerequisites
Definition and meaning of mathematical models
Components of a mathematical model
Assumptions and simplifications
Basic statistics and functions
Reference
Giordano, F. R., Fox, W. P., and Horton, S. B. A First Course in Mathematical Modeling. Cengage Learning.
Meerschaert, M. M. Mathematical Modeling. Academic Press.
Box, G. E. P., Hunter, J. S., and Hunter, W. G. Statistics for Experimenters. Wiley.
Related Topics
SCI-MAT-015-001
SCI-MAT-015-013
SCI-MAT-015-014
SCI-MAT-015-521
SCI-MAT-015-581
Sources and verification
Sources are shown as provided in the project data. Not every source has yet been independently verified. Please report errors or missing citations.
- Source verification pending; check an authoritative textbook, university, standards body, or peer-reviewed source.
Suggest a correction or source · More in Mathematical Modelling