the wind tunnel at A&M isn't far =)

can't think of any locations in austin that are even level

let alone windless

---

---

Kat Hunter reports on the San Dimas Stage Race from inside the GC winning team
Aeroweenie.com -Compendium of Aero Data and Knowledge
Freelance sports & outdoors writer Kathryn Hunter

Don't need level now. Won't need windless soon.

Yeah, had your method in mind when asking. But I do need a good loop that doesn't require any braking and doesn't involve stop lights or signs.

http://anonymous.coward.free.fr/.../recommendations.pdf

Seems like the Veloway would be a perfect place to use Robert's VE method, at least if you could find a time when the wind is calm and no one else is using it.

Also, Zilker Park south of the river might work...although I can't recall if you can make a complete loop, or have to exit the park and re-enter it.


West of Zilker on the north side of the river is pretty flat in places, so might be good for regression testing...again, if you can find a time when wind and/or traffic didn't interfere.

Thought of the Veloway, but discounted it due to having to brake around several of the corners. Then again, I don't need a 3-mile loop do I?

mopac loop can probably provide what you need

1st street just south of slaughter maybe too

---

---

Kat Hunter reports on the San Dimas Stage Race from inside the GC winning team
Aeroweenie.com -Compendium of Aero Data and Knowledge
Freelance sports & outdoors writer Kathryn Hunter

'

you can do the short loop and skip the two hairpins on the veloway, it is relatively well shielded from the wind too

---

---

Kat Hunter reports on the San Dimas Stage Race from inside the GC winning team
Aeroweenie.com -Compendium of Aero Data and Knowledge
Freelance sports & outdoors writer Kathryn Hunter

Great Northern just south of Farwest/Mopac could be a good run. It is roughly 1.4 miles and going South to North is a elevation gain. I don't think that run will make a good "loop". You will have to go out there early in the morning to make sure there is no traffic or cyclists.

The back run:
http://www.gmap-pedometer.com/?r=3935887

Start:
http://maps.google.com/...=0,0.020878&z=16

Middle:
http://maps.google.com/...cbp=12,25.58,,0,9.62

End:
http://maps.google.com/...;cbp=12,17.4,,0,5.46

I am actually very curious to find out if this would work well for testing. Also if you see a dude on a Black/Grey P3 say Hi. Thats me.

Great Northern actually sounds like a great idea.

1.4 mi is quite a bit more than you really need, so you could use just the part with the most consistent grade if you want to employ the regression approach.

Again, though, "wind is thine enemy", so don't expect great results (from any method*) unless you test when the winds are quite light (<0.5 m/s) and there is absolutely no traffic.

*With the exception of when you actually measure the wind speed, as we did in our wind tunnel validation study.

Run Far Does a free TT every other Tuesday evening as long as you have your own chip. It's an out and back. Might be best "worst case".

I've did some Google searching for specifics. Is this the method that you're referring to with "regression testing"? http://www.sportsci.org/2006/index.html

If so, a few more questions:
- Would multiple runs with a Garmin 705 be sufficient for getting accurate elevation profiles? Or would even multiple runs not be good enough? Is getting the elevation specs from the city the only reliable option?
- Is a handheld weather station a necessity or would local weather measurements be sufficient for temp, pressure, and humidity? The MesoWest site linked in the above testing method shows 3 different reporting stations with a few miles of Great Northern Blvd. It has data for temp, pressure, humidity, and wind. Updates seem to be every 10 minutes.
- I would think that the wind measurements from the above would not be reliable enough and that a local measurement of speed and direction would be needed. I could see spending the $20-30 for one of those devices, but most of them don't seem to have pressure and humidity readings built in.

I'm a smart guy but don't have the statistical or engineering background to even come close to following the discussions you and Robert have regarding the merits of the different test protocols. When it comes to this stuff, I'm more of a "do this, measure this, plug them into this equation, here's your answer" kind of guy, just like the above method provides.

Yeah, but it's 8 miles over varying terrain that doesn't end in the same spot it begins (the finish elevation is several feet lower than the start). Since the roadways are separated significantly, you're also going to get very different wind effects on the out and back, even if you do measure the wind speed.

That's an example of the what he's talking about, and it has a spreadsheet with the air density calcs built-in.

The regression method doesn't actually need an elevation profile. It needs to have either zero slope or, if it's not zero, known constant slope. Of course, whether you have true zero slope or known constant slope, or measure with a Garmin 705 depends on how much error you're willing to accept. The best precision Andy has ever been able to get with the regression approach is 1.5% so that's a target to shoot for.

Local weather measurements to get air density are probably fine.

Then just plug your data file into Aerolab. Nothing simpler.

Actually, it has been as low as 1.3% on multiple occasions...but more to the point:

1) you yourself have emphasized that this is likely an overestimate of the true imprecision, and

2) no one has ever done any better across multiple trials regardless of the approach they have used.




In fact, local weather measurements are not always "fine" - it depends on the sites in question, and in particular the extent to which local heating impacts air density (wind data from weather station reports is only useful if all you care about is making some crude approximations).

Pressure and relative humidity, generally yes, but temperature, it depends.


Your thinking is correct.


You can now get a hand-held weather station that measures everything for around $100, but I wouldn't spend the money on one just for wind speed measurements, as few, if any, can accurately measure very low wind speeds. For that you need one that relies on a hot wire anemometer or perhaps a pressure differential flow meter rather than a mechanical device that has inertia.

Case-in-point:



The Brunton measurements are obviously made on-site, whereas the airport in question is ~2 km away as the crow flies. When I first started using my "natural wind tunnel", I did not own the Brunton, and so could only cross-check the airport data againt the thermometer of my car and also my SRM. I did not observe any significant differences, but on most occasions it either wasn't very warm, or it was quite warm overnight. Now that I have the Brunton and can record data at high frequency throughout each testing session, it has become obvious that on "in between" days, the airport temperature starts increasing sooner/more rapidly than the local temperature. This is because the airport is in an almost-treeless flood plain surrounded by lots of commercial development, whereas the road I use for testing runs through dense woods. (And it was because I knew that this could result in important temperature gradients that I collected data using my car's thermometer and the SRM in the first place.)

Thanks for the replies, Robert.

I phrased the second question poorly. I know regression testing requires slope, not elevation profile. However, an elevation profile will obviously allow you to determine the slope of a particular section. What I'm trying to figure out is if multiple runs with my Garmin would produce an accurate enough picture of that slope.

Looking at Andy's site, it looks like Aerolab is not a standalone app, but rather only a patch on GC. Is that right?

AC, thanks to you for the replies as well.

The reporting site appears to be an individual's house who reports as part of the network. The foliage coverage, pavement coverage, elevation, etc are pretty consistent between that area and Great Northern.

Given what you've said about the portable anemometers, do you have any recommendation for ensuring as accurate a test as possible without significant expense? Some of the sub-$100 models from have specs that state minimum wind speeds of 0.3 m/s and accuracy of +/- 3%. I would think there are two options: test when the wind is as non-existent as possible or get one of the portable stations and be sure the wind is blowing at least above the device's stated minimum.

That sounds okay, but an NWS weather station might be better as presumably it will be accurate and won't go away any time soon.



Unless you can accurately measure both wind speed and direction on site, your only real option if you want highly reproducible data is to test when winds are as light as possible.

Seems like wind direction would be the easiest part of all. There are lots of ways to create a simple weather vane (if not just buy one) add a compass and you're in business.

Right - then you just need someone to stand there by the side of the road while you're testing and write down the reading every second or so. ;-)

(Actually, just once per run is sufficient if the wind direction isn't flucuating very much...if it is, though, then multiple readings would probably be necessary).

Hmmm. Well, first, thanks for that correction of 1.5% to 1.3%. Second, I'm pretty surprised that I've ever used the term "true imprecision" -- I don't recall using it, it's not something I can easily imagine myself saying, and I'm not sure there is such a thing so if I did I must've misspoken. I have talked about OLS being an inefficient estimator when the error term has nonconstant variance but that isn't really relevant to this particular discussion.

Tom A. has collected data across trials that are quite a bit better than 1.5% -- or 1.3% for that matter.

Sigh. That's why I said "probably." The handy thing about Aerolab is that you can quickly see exactly how much a change in air density is going to affect the estimated CdA.