High Accuracy Sun Position Calculation

[Kiwi Geoff]

(09-22-2015 06:12 PM)Powersoft Wrote:  Also the next step is calculation Sun-Rise, Sun-Set and Sun-Transit time.

Hello Jan,

Back in March of this year (2015), I tinkered with the SPA code, and found an anomaly with the "Sunrise/Sunset" algorithm.

The SPA header file says the "sunrise and sunset" times are +/- 30 second accuracy, but the following example (on the 20th September 2015) shows a delta of something like 100 seconds (in time).

I just tried using this SPA Calculator.

with this data:

********************************
Start Date 2015 September 20
End Date 2015 September 20

Interval = 10 minutes
latitude = 39.742476;
longitude = -105.1786;
timezone = -7.0;
elevation = 0.0;
pressure = 820;
temperature = 11;
delta_ut1 = 0.0;
delta_t = 67;

azm_rotation = 180;
slope = 0;
atmos_refract = 0.5667;
********************************

The SPA calculator gave the following result:

Date,Time,Local sunrise time,Local sun transit time,Local sunset time
9/20/2015,0:00:00,5.770938,11.902407,18.050313

SPA Calculator (in format HH:MM:SS.ss)
2015-9-20 R 05:46:15.38
2015-9-20 T 11:54:08.67
2015-9-20 S 18:03:01.13

JPL DE430 ephemeris gives the following:
2015-9-20 R 05:46:15.930 Az = 87.884
2015-9-20 T 11:54:08.515 El = 51.252
2015-9-20 S 18:01:21.472 Az = 271.858

There is reasonable agreement between "Rise" and "Transit" times, but the "Sunset" has a delta of something like 100 seconds.

Following is a graph showing SPA.C compared to the JPL DE430 ephemeris for all "sunsets" for 2015 using above configuration.



To help isolate the problem with SPA.C , I tried SPA.C using coordinates near my house here in the Southern Hemisphere. The "sunset" data is now very good but the "sunrise" calculation in SPA has the 100 second error. So this "sunrise/sunset" error is related to which "longitude" hemisphere one is in (East or West).

The following graph , shows the accuracy of SPA over a period of 85 years , showing the Sunset and transit delta for each of the 31,047 days for a location here in Christchurch New Zealand (approx 172.7 East , -43.5 South).



I tried to write to Afshin Andreas (author of SPA.C) at NREL, to advise that I think there is an issue with the SPA.C code regarding Sunrise and Sunset. However I did not hear back from Afshin - maybe my email was rejected from being "overseas" (New Zealand).

So according to my comparison to the JPL DE430 ephemeris Jan, there is an issue with the Sunrise/Sunset algorithm in SPA.

Perhaps a person who has the book by J. Meeus called "Astronomical Algorithms" could check to see where the SPA code has made a possible mistake implementing the method for "Sunrise/Sunset".

Regards, Geoff Hitchcox, Christchurch, New Zealand.