I have written a C++ code to do the following: For a given curve,
the code will filter out the seasonal factor.  It then fits the deseasoned
curve to a smooth function of the form  f(t)=a+A/(t+B)c, with 0<c<.5. ?The curve fitting is achieved by using Conjugate Gradient method. ??As Tanya pointed out, this form of function ensures that  ?f(t)2*t - f(t-1)2*(t-1) >0, which is useful in calculating forward-forward ?vol. ?The outputs are season patterns (2 different patterns are allowed)?and the smoothed curve.  One can then calculate forward-forward vol?from the smoothed curve and superimpose the seasonal factors back?onto the curve.??This procedure is not useful for VaR calcuation. However, I believe it ?is useful for other situations which require the knowledge of  seasonal ?forward-forward vols.  For example, in power plant valuation and?credit exposure simulations, it would reflect the reality more closely?if we add seasonality to our forward-forward vol in simulating the forward?prices. ??Attached are the spreadsheet and XLL file. I will forward the C code if?any of you is interested. ??Alex??