1. Fail to accurately predict or explain observations: This is the most fundamental reason. If a model consistently produces predictions that don't match real-world data, it needs revision or replacement. This could be due to:

* Incomplete understanding: The model might be based on incomplete information about the system it's trying to describe.

* Oversimplification: The model might make assumptions that are too simplistic and don't capture the complexities of the real world.

* New discoveries: New observations and data may reveal aspects of the system that were previously unknown, requiring the model to be updated.

2. Are contradicted by new evidence: As science progresses, new evidence can emerge that challenges existing models. This evidence might come from:

* New experiments: Experiments may reveal unexpected results that force a re-evaluation of the model.

* New technologies: Advancements in technology can allow for more precise measurements and observations, leading to discrepancies with previous models.

* New theories: The development of new, more comprehensive theories may render older models obsolete.

3. Become too complex or unwieldy: While complexity is sometimes necessary, models can become so convoluted that they are difficult to understand, interpret, or apply. This can hinder progress and lead to the need for simplification or a new approach.

4. Are no longer useful for their intended purpose: As scientific understanding evolves, the goals and applications of models may change. A model that was once useful for a specific purpose might become outdated or inadequate for new applications.

In summary:

Scientific models are constantly evolving. They are not considered absolute truths, but rather tools that help us understand the world around us. They are refined and adjusted as new information emerges, ultimately leading to a deeper and more accurate understanding of the natural world.