You examine the current mathematical climate models and determine their limitations in terms of available CPU power, RAM, disk on even the largest supercomputers.

Then you take into account the lack of any accurate model at all being available for the operation of the seas and oceans - we don't have an accurate mathematical model for even the Gulf Stream, for instance; much less the rest of the world.

Then you realize that all prediction models are so error-prone due to the above that they cannot be relied upon with a reasonable degree of certainty.