VOACAP reliability, the end of myth ? part 4

Following the publication of the series of articles on VOACAP reliability, Jari, OH6BG (voacap.com) sent me the following email (reproduced here with his authorization). I have just underlined the key-elements.
Hi Pat,
Kari OH2BP wrote to me today pointing out your monitoring results vs. VOACAP predictions blog at http://f6irf.blogspot.com/. A very interesting reading indeed, and a constant reminder for all of us that predicting HF propagation is indeed more of an art form than an exact science, hi!
I have made similar monitoring sessions with Faros on July 2007, and, on 14 MHz, VOACAP miserably failed to predict circuits from Finland to the United States in the middle of the night. I reasoned then that the twilight hours that prevail at that time of the year here in the North generate many hours of signal-enhancing "sunset" effect at this end, which VOACAP simply cannot see.
Below I gathered my personal observations, which can perhaps help you in your endeavour of tackling the issue in hand...
1. A quick run reveals that the use of the older IONCAP absorption model can, in your case, give up to 15 dB better SNR/SDBW values over the more conservative "Normal" absorption model, used in your calculations.
2. You must not only look at the predicted SNR values, please note also the predicted median Signal Power values, S DBW.
3. You must not only look at the predicted median values of Signal/Noise and Signal Power. Calculate also the predicted upper decile (10%) values that tells you the probability values for SNR and S DBW on 3 days (10%) in a month. The SDBW10 values specifically suggest an extremely poor, but probably distinguishable (if the real Noise Power is not too overwhelming), signal at the time periods that have been monitored, although the values themselves are extremely conservative, but to me, a right tendency can be seen. Formulas: SNR10 = SNR + SNR UP SDBW10 = SDBW + SIG UP
4. I have understood that extremely long circuits (such as this, far far beyond 10,000 km) can be troublesome for VOACAP and probably the predicted values are of a more conservative note. Therefore, I am pretty amazed at VOACAP being able to figure out the correct tendency anyhow.
5. In addition, I can only speculate that perhaps the greatest influence for the "better-than-predicted" monitoring values lies in the fact that the time periods of monitored best reception on 7 MHz seems to occur approximately at ionospheric sunset/sunrise on either side of, or along, the circuit. As far as I know, VOACAP is not able to take into account the signal-enhancing sunrise/sunset effect, which I can also confirm by my own monitoring on 14 MHz.
73 Jari OH6BG/OG6G

Attached to Jari's email was the run output table which I have plotted here vs "the observed average" over near to 3 weeks period (plot published in article part 2)This plot has been made using OH6BG run prediction table. It uses the old IONCAP absorbtion model, assumes Isotropic antennas at each end of the path, and a very quiet man-made noise level of -155dBW/Hz at 3Mhz. The purple curve is the "upper decile" (3 days per month) while the blue one is the SNR (as shown earlier equivalent to SNR50, so 15 days per month). The red curve is an averaged measured value over a 3 weeks period.

Following are the parameters that Jari used:

SHORT PATH: F6IRF to ZL2TLD, using IONCAP absorption model
CCIR Coefficients ~METHOD 30 VOACAP 08.1023I PAGE 1 Oct 2008 SSN = 7. Minimum Angle= 3.000 degrees f6irf zl2tld AZIMUTHS N. MI. KM 45.80 N 6.00 E - 40.96 S 175.59 E 61.06 306.11 10267.9 19014.6 XMTR 2-30 + 0.0 dBi[samples\SAMPLE.00 ] Az= 61.1 OFFaz=360.0 0.001kW RCVR 2-30 + 0.0 dBi[samples\SAMPLE.00 ] Az=306.1 OFFaz= 0.0 3 MHz NOISE = -155.0 dBW REQ. REL = 90% REQ. SNR = 4.0 dB
LONG PATH: F6IRF to ZL2TLD, using IONCAP absorption model
CCIR Coefficients ~METHOD 30 VOACAP 08.1023I PAGE 1
Oct 2008 SSN = 7. Minimum Angle= 3.000 degrees
f6irf zl2tld AZIMUTHS N. MI. KM
45.80 N 6.00 E - 40.96 S 175.59 E 241.06 126.11 11345.0 21009.3
XMTR 2-30 + 0.0 dBi[samples\SAMPLE.00 ] Az= 61.1 OFFaz=180.0 0.001kW
RCVR 2-30 + 0.0 dBi[samples\SAMPLE.00 ] Az=306.1 OFFaz=180.0
3 MHz NOISE = -155.0 dBW REQ. REL = 90% REQ. SNR = 4.0 dB

In a later email Jari wrote:
Although I am by no means a specialist of low-band propagation, my little experience tells me that enhanced signal levels and openings on lower bands can be observed 1 to 2 hours before sunset and similarly 1 to 2 after (if not even longer) hours after sunrise. And both the stations need not be on the terminator zone, "greyline"... This particular case with ZL2 could fit into this framework. When it comes to using VOACAP for predicting propagation on lower bands, I firmly believe that predictions on 80M, not to talk about 160M, are simply doomed to fail for any longer path... George Lane writes about this at http://www.voacap.com/lowband.html as follows, "... George Haydon said there was very little data below 4 MHz but there was some for short paths that did go down to 2 MHz. So they modeled a fit to those cases. Risky, but it has proven to give good results for Near Vertical Incidence Skywave (NVIS) situations. ..."
So, trying to use VOACAP to predict low-band DX propagation is asking for trouble. Maybe you can get better results just by simply looking at sunrise/sunset maps! hi.

Many thanks Jari for your contribution... I think that the message is clear !