Why can't custom-made TLEs for the DSCOVR launch booster in orbit around Earth work with SDP4?

Question

This fascinating answer to Why has no TLE been published for the DSCOVR satellite and the Falcon 9 R/B? presents links to custom-made TLEs for the Falcon 9 rocket body that put DSCOVR into orbit, and it includes the following:

https://www.projectpluto.com/pluto/mpecs/pseudo.htm

https://github.com/Bill-Gray/tles/

The file '15007b19.tle', for example, contains TLEs covering 2019 for this object (the spent Falcon 9 rocket body). Note that for this particular object, an SDP4 TLE cannot be fitted, even though the object is bound to the earth. The 'ephemeris type' byte is therefore set to be 2 ("use SGP4 only") and the TLEs use only the SGP4 propagator. So the TLEs will work in some software that actually follows the full TLE specification and respects the 'ephemeris type' byte... and not in a lot of software that isn't so carefully written.

Looking at the TLE history in the github link I see that the mean motion is something like 0.06 meaning it's in a very distant 16 or 17 day orbit! That means that it is very loosely bound to Earth and that the effects of the Sun and Moon should be quite significant.

Question: Why can't home-made TLEs for the DSCOVR launch booster in orbit around Earth work with SDP4?

---

For more on SDP4 see

• How do SDP4's "Deep space" corrections to SGP4 account for the Sun's and Moon's gravity?
• SGP4 on Systems Tool Kit (STK); how to check if SDP4 deep space correction is implemented?

Answer 1

SDP4 has a lot of algorithmic nuances to handle objects in 12- and 24-hour orbits. You can get decent (km-level over a week or so) fits to the model. I see no evidence in the SDP4 algorithm that they had higher-orbiting objects in mind. Go much beyond a one-day orbit with SDP4, and you're basically using a hammer as a screwdriver.

If you take a look at the TLEs on that GitHub site, you'll see that I did use SDP4 for some objects in four-day orbits: 1983-020A and D, most Velas, 1977-093A and E. The fits are good to (usually) a dozen or so km over one day, usually better near apogee and worse near perigee. 2011-037A and 2011-037ND (the Russian radiotelescope SPEKTR-R and its booster) are in eight-day orbits, and I can still get a decent fit with SDP4. But in each case, I'd get a better one with SGP4. (I stick with SDP4 where possible, because it's more widely usable. I only switch over to SGP4 when the errors in SDP4 become intolerable.)

Those errors do become intolerable for some higher objects, such as 2015-007B (the DSCOVR booster that started this discussion), the Chang'e 2, 3, and 4 boosters, and a few others. For those, I'd get fits good only to a thousand km or so if I tried using SDP4. So I do a somewhat non-standard thing and fit them to SGP4, setting the ephemeris flag accordingly. A TLE creator's gotta do what a TLE creator's gotta do.