Virtual Holiday Day Ten - Bullas#
I really like this ride. Although it is 105 km long, in my mind it is basically all over by the halfway café stop in Bullas. From a simulation point of view, the interesting bits are:
| Start (km) | End (km) | Description |
|---|---|---|
| 12 | 18 | I’m pretty sure this is a local time-trial course |
| 18.5 | 18.6 | A river crossing that could easily trash wheels |
| 35 | 45 | Quiet, twisty wooded section |
| 58 | 73 | Wonderful 15 km ascent / terrifying 15 km descent |
Last year#
Strava 2025🔗:
Changes for this ride#
I don’t know if we will have time to implement this today.
I want to test the new COP30 terrain. Apart from being more accurate we are actively embracing the lo-poly aesthetic and not de-noising it. I am simultaneously hoping for smoother gradients and more jagged mountains. A good testing strategy would be:
- Pick a section: the SW climb out of Pliego has plenty of interesting terrain to the left and right.
- Record a video of de-noised SRTM terrain (I’ve done this already).
- Record a video of raw (i.e., not de-noised) SRTM terrain.
- Record a video of raw COP30 terrain.
That sounds worthy, but tedious - so I think we’ll just do a block and I’ll see what it looks like as I whiz down it today.
This year#
Strava 2026🔗:
Post-ride analysis#
We did manage to reprocess a block of AS30s with the Copernicus data.
Reprocessed AS30s:
And this is the view from the cliff in the top-right AS30:
This is better than I hoped. There is more detail and no noise visible in this admittedly small sample. Of course - it makes the 64,351 other AS30s (derived from SRTM) immediately rubbish!
In other news - populating the instanced trees array starting with the closest AS30s definitely made a difference on the few occasions when the tree array maxed out.
On the final 24 km descent there was some really bad spiking. My working theory is that it is due to the thousands (30,000 at one point) of “they’re behind you” trees having their transformation matrices recalculated when the AS30 changes.