Once upon a time, I used to model real-time systems to engineer software. I know a thing or two about modeling and I can tell you without a doubt some very intuitive things. 1.) All models lack detail from what they are modeling. They are an abstraction. If they had the detail of the thing they are modeling, it would be the real thing. 2.) You need more than one model to come close to modeling with any accuracy. The more than one model has to be a different type of model. Not the same basic model with different variables or a tweek here or there. 3.) Complex systems always have unknowns, and only multiple iterations of prototyping/developing of the solution do you discover the unknowns that could be used to correct the model. 4.) As for modeling the ChiCom Flu, weather, the Climate and other highly complex things, the complexity is so great that models never come close in long range predicting, e.g., weather more than a week. See, number 3, as you are closer to the real thing in time, you can adjust the model to be correct, but you won't be close to the real thing six months from now, until six months from now. By that time, you no longer need the model, because you have the real thing.
Is modeling useful? Yes, it is a tool and as a tool it needs to be used properly and its results need to be interrupted correctly.
Such nonsense!
I had to model a system that was required to track geosynchronous satellites which actually describe a small pattern in the sky that nearly repeats every sidereal day. Rain fades are a problem. "Nearly" is the key because you don't actually know whether the diminution of signal is due to atmospheric problems or satellite drift.
The software to control the tracking had to be subjected to more difficult conditions more often than might occur in the real world.
Sometimes models work. Sometimes they don't. The problems arise when someone is trying to model a system with too many variables and/or too many unknowns. The existence or planetariums proves that a model and the real thing can be pretty damn close.
ML/NJ