Here's the recommended solution:

[ Diagram of slightly more complex Geography dimension. ]

We say "recommended solution" because there are many possible ways to implement a multidimensional data model. Usually one way emerges as the best when flexibility, ease of maintenance, and robustness are considered. (For example, if you plan to analyze many other facts with respect to climate, you would be justified in creating an entirely new dimension for the MICRO CLIMATE attribute.)

Notice that our model allows cities to be grouped by microclimate or by county. Microclimates are not determined by counties in any way, only by cities.

What if some cities extend across microclimates? How would we have to change our model? If you start to lose sleep at night, or are interested in hearing our answer, contact us!

Back to "What is multidimensional modeling?"...