Uncertainty, Error, and Approximation

Why useful models do not need to be perfect

Before You Start

You should know
That measurements can be imperfect and that real systems are usually more complicated than a simple diagram or formula.

You will learn
How uncertainty, error, and approximation differ, and why a model can still be useful even when it does not match reality exactly.

Why this matters
Readers who expect perfect prediction from every model usually end up either disappointed or mistrustful. Better modelling starts with learning what kinds of imperfection are normal and manageable.

If this gets hard, focus on…
The question “Is this model useful for the job?” That is usually better than asking whether it is perfect.

Models are never perfect.

That sounds like a criticism, but it is actually normal. A model is useful not because it captures every detail of reality, but because it captures enough of the right structure to help us explain, compare, or predict.

1. Error Is Not Always Failure

If a weather model predicts 18 °C and the day reaches 19 °C, the model is not useless. It is close.

If a flood model predicts the wrong neighbourhood entirely, that is much more serious.

So the important questions are not:

  • Was the model perfect?

They are:

  • How wrong was it?
  • In what direction was it wrong?
  • Was that amount of error acceptable for the decision being made?

2. Uncertainty

Uncertainty means that something is not known exactly.

That can happen because:

  • measurements are limited
  • future conditions are unknown
  • natural systems are noisy
  • the model leaves out some processes

Uncertainty is not the same as total ignorance. Often it can be described, bounded, or estimated.

3. Approximation

Approximation means we deliberately replace something complicated with something simpler.

Examples:

  • treating Earth as a sphere
  • treating a slope as uniform
  • treating a city as one centre instead of many

Those are not mistakes by themselves. They are modelling choices.

The real question is whether the simplification is fair for the problem we are trying to solve.

Three Different Imperfections

Do Not Collapse These Into One Idea

The same map can be uncertain, wrong, or simplified for different reasons. Those are not the same failure mode.

Uncertainty

We Do Not Know Exactly

A flood boundary might be shown as a fuzzy zone because the water level or input data is not fully known.

Question to ask:

How wide is the plausible range?

Error

The Result Misses Reality

A model predicts flood impacts on the wrong block or in the wrong direction. The outcome is off, not just imprecise.

Question to ask:

How wrong is it, and does that matter for the decision?

Approximation

We Simplified On Purpose

A city is treated as one center instead of many because that simplification is good enough for the task.

Question to ask:

Is the simplification fair for this job?

Uncertainty describes what is not known exactly, error describes mismatch, and approximation describes an intentional simplification.

4. Why This Matters

Good modelling requires honesty about limits.

Models are:

  • structured simplifications
  • useful despite limits
  • strongest when their assumptions are visible

If readers learn that early, they can read technical work more maturely. They stop looking for impossible certainty and start asking better questions.

5. A Practical Reading Habit

When you meet a model, ask:

  1. What is being simplified?
  2. What is being ignored?
  3. What kind of error could that create?
  4. Is that acceptable for the question at hand?

That habit is more valuable than memorizing formulas without judgment.

If This Gets Hard, Focus On

  • all models simplify
  • approximation is normal
  • uncertainty can still be informative
  • the real question is whether the model is useful for the job

That is a strong and realistic way to read quantitative work.