The next step in uncertainty analysis is to make different samples from the existing distribution. Choosing the most appropriate sampling method and size of the samples has an important role in uncertainty based analysis, as it should be large enough to provide accurate results \cite{wang2014monte}. Indeed, the input covers the entire range of space, in order to guarantee that the samples are reliable representatives of the entire uncertain space \cite{Saltelli_2007}.
The Latin Hypercube Sampling method is selected for this step. Regarding that, the number of 20 random samples
from each year of climatic data between 2012-2016 is considered. This resulted in an
overall sum of 100 annual simulations (8760 hours) of incident solar
radiation on the surface of a vertical wall facing south. The simulations were handled with JePlus simulation manager. As
mentioned before, this study is only focused on the cooling season,
which for Milan city starts from the first of July and endson the
9th of September \cite{Dongellini_2015} .