There are many types of models such as analytical (usually quite simple in nature), computational (simple or complex, predictive in nature) and physical (measurements such as in a wind tunnel or using water as the medium). A model is a representation of reality that is often used to support an idea or design and then improve its performance.
The competence of the ideal modeller extends beyond the delivery of acceptable results, within the tolerance of the chosen method, to providing an appropriate interpretation of the results communicated at different levels to different audiences. A review process internally or externally, if deemed necessary, should be considered essential to providing confidence in any results and their interpretation. If any results are called into question, transparency and rigour will usually provide the required answers. An environment with good QA procedures will provide added assurances.
The choice of type of model appropriate to delivering the desire objectives is a key element of the decision making process. A good understanding of what is commonly used for said objective else what could be further developed is usually based upon the experience of the modeller and/or the positioning of the specific industry.
For example, hydraulic and hydrology models can sometimes be quite simple, in terms of their algorithms, with many assumptions, simplifications, boundary conditions and limitations not declared by the model author. Therefore, interpretations can sometimes exceed the abilities of the model and the results are sometimes not put into context. The advantages and disadvantages (pros and cons) of all methods and models under consideration should be understood - see this paper looking into the Strengths and Weakness of the Modelling Techniques for 3D hydraulic models (similar in many ways to 1D or 2D).
Everyone knows that 2 + 2 = 4. Models are sometimes sold as giving definitive answers whereas this may be far from the truth. Often the seller of the interpretations has little or no knowledge of the modelling process, limitations etc. Sometimes the project director would not even know how to open up the software. Here lies a huge dichotomy, particularly when so heavily relied upon for a project of national interest, and absolutely supports the need for an independent, transparent review process.
Modelling is the outcome of scientific application. Artistic license is sometimes drawn by those trying to sell the results for their own means. If any results are sold without reference to tolerances, sensitivities, simplilfications, limitations and, possibly, need for further work then a competent reviewer might ask 'what are they trying to hide'?