It isn't easy to navigate. Especially in ecology, the time series and data we have are often very sparse, though we are starting to see good examples of long-term time series available for us to work with and are trying to incorporate them into the models. Often those time series and those observations don't come with a good understanding of exactly what the error is in them, so there is some struggle there putting, for example, a probability or a precise likelihood to, say, a management strategy that management wishes to evaluate. There are ongoing challenges with that and I think it's just something that may always be there.
The other thing I'd like to say about it is almost a caveat. The simulation modelling and computer modelling does play an outsized role in marine ecology, partially because you can't isolate variables. You can't do controlled experiments as easily. They are just models at the end of the day.