Diffus part calculation in PVsyst

[ckoessler]

I have a question about the diffuse part calculation using the Liu-Jordan correlation.

By analyzing the figure attached to this post, I noticed that I do not understand totally the procedures of PVsyst's algorithms when it comes to the diffuse correlation.

The figure shows the scatter plot between the diffuse fraction (kd=DHI/GHI) and the clearness index (kt=GHI/GHIo) of two samples.

these two samples are:

a) Hourly data of global horizontal radiance (generic) (mean weather station hourly data over 10 years)

b) Monthly data of global horizontal radiance based on a) . --> Hourly data is generated synthetically in PVsyst

As we only provide global horizontal radiance, PVsyst calculates automatically the diffuse part.

The question is now how this is actually done (step by step)?

a) hourly generic data is used together with the clearness index ( hourly or daily) to calculate the diffuse radiance thanks to the Liu-Jordan hourly correlation. --> blue points in my figure

b) monthly data is transformed into hourly data by PVsyst. then the Liu-Jordan hourly correlation is applied to the synthetic data. --> red points in my figure

If this is true, so why is there a difference in the dispersion of the red and blue scatter plot?? It seams to me, that data based on monthly values (red) is some how normalized every month to reach some kind of target value. But this would imply that the diffuse part is calculated for the monthly values before generating hourly values. Is my assumption correct?

Thank you for your explication.