The re-interpretations are something like this:
- There are approximately 6240 positions within the system, and each of these positions have a bit of different property / behavior which ultimately translate to variable inputs into the simulation.
In terms of specifying the inputs into UQ Lab such that they are randomly varied, the easiest way for me to achieve it was by “randomly selecting” a unique historical data entry among the 6240 positions, rather than directly and randomly varying the inputs associated with the positions. This is because the inputs are tied in relation to the 6240 positions, and the trend / shape of the input in relation to the positions are not so easy to describe. This is because behavior-to-position relation is influenced by human decisions. This could be done using a shape function to describe the overall trend of the behavior-to-position without significantly increasing the number of input parameters, but I chose not to do it.
So basically, among the 19 inputs into UQ Lab, the last 2 inputs describe the “position” among 6240 and “random point in history” for that position. The re-interpreted inputs are the behavior /properties associated with that random selection that is printed as part of the 25 outputs.