Posted by James Watson on 21 Mar, 2017
The full version of this post is available for download here.
A key VOACAP input parameter is the Sun Spot Number (SSN), traditionally derived from data published by the National Geophysical Data Center (NGDC) / National Oceanic and Atmospheric Administration (NOAA). However, at the end of 2016, NOAA discontinued providing predicted data, instead directing users to the SIDC website. The SIDC SSN values were previously very similar to those of the NGDC / NOAA and employed a common smoothing algorithm, designed to reduce the effect of short term perturbations without obscuring the behaviour of the underlying 11-year solar cycle. Unfortunately, this no longer the case and in July 2015 the SIDC dataset underwent a major revision, the most significant component of which was the removal a 0.6 scaling factor. "This scale change, when combined with the recalibration, leads to a net increase of about 45% (correction variable with time) of the most recent part of the series, after 1947” (SIDC, 2017).
I've spent a little time looking into the effect of using the revised numbers using ITU-R P.1148-1 as the basis for comparison. Three candidate sets were evaluated; - NGDC The original NGDC / NOAA values. - SIDC The revised SIDC values (i.e. post July 2015). - SIDC_ADJ A modified form of the SIDC values that applies a 0.7 scaling factor to the published SIDC values.
The three SSN sets are compared in the figure below
The survey looked at the effects of the revised SSN values on VOACAP running with both the Normal and IONCAP absorption models and the application of some proposed modifications by Alex Shovkoplyas. These are are defined at run time by the version number suffix in the file itshfbc\database\Version.w32.
The results indicated that VOACAP performs best when using the original NGDC / NOAA SSN values. This was teh expected outcome as these are the values used during the application's development. In the absence of NGDC / NOAA SSN values going forward, the 'adjusted' SIDC values make a good substitute. More surprising was that the IONCAP model consistently outperformed the Normal model.
The results further indicate that the modifications proposed by Shovkoplyas do indeed improve the accuracy of the model when using Standard Deviation as a metric (as prescribed by ITU-R P.1148-1).
This leads to the recommendation that for best performance, VOACAP should be run with adjusted SSN values and the model suffix set to 'a' in the file itshfbc\database\Version.w32.
The full survey is available for download here and I'd welcome any follow-up correspondence by email address jimwatson -at- mac -dot- com.