Where a process of Nature is not completely understood or can only be partially measured with today's technology, then one approach within a larger mathematical model is to substitute a few variables in lieu of a true representation of the 'missing' process, and then to 'tune' these variables as best possible, based on the 'skill' of the larger model to achieve a measurable macro-level performance. This method of approximation within a model is commonly known as 'parameterization' - and is subject to continuous improvement as more knowledge of the 'missing processes' is gained.