Mille Cymru - Welsh 1000 - 23-26/07/2010

[mattc]

I look at this thread and wonder if I'm taking my attempt to do the Gt Eastern seriously enough (note: nice'n'flat). 
At the moment I'm considering riding through each night and napping on the verge once the day warms up again.  This is probably a silly idea 3 nights on the trot.

Well, it _is_ a silly idea - but these things are relative Many have ridden PBP like that, but mainly in hot years, I think. Your ride is a bit shorter and flatter, I believe?
(I think if I was you I'd book a B&B for maybe the 2nd night, depending on how the times work out etc ... )

However, the Welsh ride has a few other factors:
- It's in wales, so some bad weather at night doesn't seem unlikely,
- We've all entered with the lure of (some sort) of sleep stop every night, so it will be rather demoralising to not really get much use from this,
- For most of us, it's a bloody hard ride, so taking away the creature comforts is unappealing,
- I'm expecting some nice scenery, so minimising day-time sleeping seems attractive

[arabella]

shorter and flatter indeed. 
I used an average of 16.67 kph (including food stops) and came up with 60 hours.  This gives me a 12 hour stop and still a few hours spare.  I'm looking at the Thorne-Knaresborough leg if anyone has any suggestions, possibly Wetherby (just over 1/2 way).

I didn't enter Mille Cymru on account of said hills.  Too hard for me in an I know my limits kind of way. (but then I once thought that about a 300)

[vistaed]

Or if you can go fast enough to do this and get enough sleep: start day 2 at 5 am and day 3 at 4 am (just at/before dawn)  -  and ride day 3 as 349k.

umm, this seems like a plan.

[JayP]

A bit more on the day 1 climbing saga.  
I heard about this site
GPS Visualizer: Convert GPS files to plain text or GPX
This  takes you to a form where you can browse your hard drive to your gpx-file-with-dodgy-elevation-data and have it overwritten with data looked up from Digtal Elevation Model (DEM) data held in NASA’s SRTM3 database (Shuttle Radar Topography Mission). You can also have estimated fields like distance gradient and speed added if you like. You can save the transformed file in a variety of formats including Tab Delimited or csv text. The database is described as having a 90 m horizontal resolution. I’m not clear what this means but common sense suggests a 90m square grid with accurate elevation recorded at the vertices and smooth interpolations elsewhere. The data supplied is quoted to one decimal place, that’s to the nearest 10 cm. There would be no point in such precision if the data was unreliable.

So, back to the spreadsheet with my shiny new data. This time it totted up 5972m of climbing. HHmm  well better than last time but still over a 1000m above Mr Dulates figure. I then got the spreadsheet to flag up just the peaks and troughs in the elevation data. I stripped out all the rest. This reduced the file from 4366 trackpoints to a mere 1193 with the total climb remaining at 5972.   Then I got the sheet to flag up all those peaks preceded by at least 9m of continuous climb and tot up a new climbing figure based on these only. It found 164 peaks over 9m and totted up 4706m of climb (close agreement with John’s figure). The ignored less-than-9m peaks are the minor undulations that can occur unnoticed between 10m contour lines.
This is the climbing you miss in a contour count.

So result? Day 1 has around 4700-4800 m of climbing ignoring ‘minor undulations’ totalling a further 1000m.(ish).  
+ I have a way of contour counting without actually counting any contours.  It’s a bit like cycling to Holyhead really, or sex, ISMA the travelling is more fun than the arriving.
Here are some more interesting facts.  
There are 72 peaks preceded by between 10 and 20 m of climbing.
27 between 20 and 30
8 between 30 and 40
5 between 40 and 50
3 between 50 and 60
10 between 60 and 70
4 between 70 and 80
5 between  80 and 90
2 between 90 and 100
2 between 100 and 110
1 between110 and 120
1 between130 and 140
1 between 160 and 170

[kcass]

That explains why my Garmin always comes up with a higher (than the published) climbing figure.

Good work btw.

[JohnHamilton]

Indeed. Calculating ascent depends entirely on the resolution being used. The harder you look the more you find. FWIW my official figures are based on contour counting. I also have a more or less full set of GPS files for the route which I will be sending to the AAA man for comparison. Steve's analysis of GPS files uses a frequency interval deliberately designed to give a result comparable with contour counts (which remain the established benchmark for AAA purposes).

The contour count derived figures and the GPS figure for the previous route were in agreement so I don't expect a significant change as a result.

I've always found any kind of DEM data to give very inaccurate results in Welsh terrain. I've ridden this route and there's no way it has nearly 6000m of climbing in that first day.

[simonp]

Quite. Tracklogs said the bcm was 12500m.

The interpolation is the problem

[Nuncio]

there's no way it has nearly 6000m of climbing in that first day.

Shame. I was beginning to look forward to the 'nearly 6,000m' of descending.

[JohnHamilton]

Quite. Tracklogs said the bcm was 12500m.

The interpolation is the problem

A figure of 1.4-1.5 times the contour counted figure is usually what I've found too. And TrackLogs uses OS elevation data which ISTR incorporates SRTM derived data.

What is most important in any altitude calculation is that it's comparable relative to other events.

[MattH]

Personally, I like the GPS claims for climb and calorie burn.

I can show them to MrsH as justification for eating triple sized portions in the days after the event.

[Greenbank]

Quite. Tracklogs said the bcm was 12500m.

The interpolation is the problem

A figure of 1.4-1.5 times the contour counted figure is usually what I've found too. And TrackLogs uses OS elevation data which ISTR incorporates SRTM derived data.

Exactly, and if you look how the SRTM data are derived you'll realise how inaccurate it will be for roads that run along the sides of valleys (rather common in Wales), or even anything undulating.

"
Contours are now derived from a finer height grid. The 3 arcsecond * 3 arcsecond SRTM grid (about 60*90m at our Latitude) has been subdivided into 2 wide by 3 high (thus about 30m square) points, which are interpolated using a 36-point spline (i.e. taking account of the nearest 6*6 grid of SRTM datapoints). The resulting finer grid was then contoured as before (using 3DEM and DEM2TOPO). This gives 'rounder' contours, which take account of local gradients and gradient changes .. however small features will still be missing, since you can only 'round' what was seen to be there, not 'invent' things which were too small to get measured in the first place.
"

So the individual height point is based on the height of a 6*6 grid of SRTM datapoints. A 6x6 grid of those points represents an area of 300x450m. And you can be as far as ~55m from a real measured point in a 60*90m grid.

Climbing figures from a GPS will be inaccurate for a different reason, jitter. Leave a GPS on a window sill for an afternoon and see how far you've traveled and how much climbing you've done.

I 'climbed' 163m playing 5-a-side football last night. On a perfectly flat pitch. Elevation Plot from Garmin Forerunner 405 GPS here

I 'climbed' 2066m in my sleep one night too (I forgot to disable the GPS when using the Forerunner to log my HR during a nights' sleep). Going from 26m (which is about right), even going as far down as 62m below sea level and as high as 120m ASL. Magic.

So, back to the spreadsheet with my shiny new data. This time it totted up 5972m of climbing. HHmm  well better than last time but still over a 1000m above Mr Dulates figure. I then got the spreadsheet to flag up just the peaks and troughs in the elevation data. I stripped out all the rest. This reduced the file from 4366 trackpoints to a mere 1193 with the total climb remaining at 5972.   Then I got the sheet to flag up all those peaks preceded by at least 9m of continuous climb and tot up a new climbing figure based on these only. It found 164 peaks over 9m and totted up 4706m of climb (close agreement with John’s figure). The ignored less-than-9m peaks are the minor undulations that can occur unnoticed between 10m contour lines.
This is the climbing you miss in a contour count.

I've got a very similar thing done as a perl script working off GPX tracklogs (or GPX files from routes created on Bikely and the like).

GPX/GPS utilities/scripts

It works by removing intermediate points (b where: a > b > c and a < b < c). You're then left with a set of alternating points of peaks/troughs. You can then apply an algorithm to get rid of peaks that climb below a certain amount (i.e. 10m) and then apply the first algorithm again. Repeat until you don't remove any points.

I use it for picking out the major peaks (over 50m climbing) of a planned ride (based on the bikely GPX) to add as extra waypoints so I know where the tops of the hills are; this allows me to pace myself up the hill and also helps avoid the soul destroying 'I must be almost at the top by now' as you round a corner to see another several km of climbing...

With a lower limit for climbing (c.f. 10m) I've had climbing figures that agreed with the organisers figures, although the climbing figures for some rides seem to be based on the raw figures from a GPS, so it seems like a hillier ride than it actually is.

I hoping that the release of the Ordnance Survey data (including spot height database) will help as the OS spot heights are often placed directly on the road at the top (and bottom) of roads (and therefore climbs). That data will allow for a far more accurate idea of climbing figures.

I'm also hoping that this can help in providing a consistent figure for AUK climbing. I'm sure Mr Snook and Mr Coates will be looking at this, and I hope I can do something to help too.

Anyway, probably a better idea to continue one of the other threads on this subject:

GPX/GPS utilities/scripts
Altitude measurement - who to believe!
etc..

rather than clog up the MC1000 thread.

[Euan Uzami]

all very mathematical 

[Weirdy Biker]

Audax math must have pie.  And a confidence analysis by tea-distribution.  Possibly also cake-ulus.

[MattH]

Fixed point arithmetic seems to feature for some riders too.

[Weirdy Biker]

AUK = tea to the power of pie

[cyclone]

Or, possibly to the power of cake 

[JayP]

Heh dont take the mikey were only having a training work-out

[Shinna]

Hi All, I've been off-line for a while busy helping set up a new business since Jan . Just started reading this thread again, it's taken me a while to catch up with all of the posts. I can honestly say that the expectations for this really seems to be building.

I've not ridden any Audax's yet this year, due to the above commitment, but have been managing to get 150 - 250 miles a week in, although relatively flat. I have also entered the Snowdon and Coast 400. so hope to be in a fit state to "sort of" enjoy this, in a self harm sort of way. All of the talk re GPS files, and the recent demise of my cateye bike comp has got me itching to buy a GPS. If only to make it feel harder than it already is.

Really looking forward to this now guys. 12 weeks and counting.

[Weirdy Biker]

12 weeks and counting.

F**king hell.  That sounds so soon 

[Shinna]

Yep, thank God the website doesn't have a countdown timer on it.

[Weirdy Biker]

Yep, thank God the website doesn't have a countdown timer on it.

It'd look like this:

[Shinna]

I was just fiddling around with the website posts and noticed that this thread is fifth in the all time number of views. Tells you something eh.

[Hummers]

Quite. Tracklogs said the bcm was 12500m.

The interpolation is the problem

A figure of 1.4-1.5 times the contour counted figure is usually what I've found too. And TrackLogs uses OS elevation data which ISTR incorporates SRTM derived data.

Exactly, and if you look how the SRTM data is derived you'll realise how inaccurate it will be for roads that run along the sides of valleys (rather common in Wales), or even anything undulating.

"
Contours are now derived from a finer height grid. The 3 arcsecond * 3 arcsecond SRTM grid (about 60*90m at our Latitude) has been subdivided into 2 wide by 3 high (thus about 30m square) points, which are interpolated using a 36-point spline (i.e. taking account of the nearest 6*6 grid of SRTM datapoints). The resulting finer grid was then contoured as before (using 3DEM and DEM2TOPO). This gives 'rounder' contours, which take account of local gradients and gradient changes .. however small features will still be missing, since you can only 'round' what was seen to be there, not 'invent' things which were too small to get measured in the first place.
"

So the individual height point is based on the height of a 6*6 grid of SRTM datapoints. A 6x6 grid of those points represents an area of 300x450m. And you can be as far as ~55m from a real measured point in a 60*90m grid.

Climbing figures from a GPS will be inaccurate for a different reason, jitter. Leave a GPS on a window sill for an afternoon and see how far you've traveled and how much climbing you've done.

I 'climbed' 163m playing 5-a-side football last night. On a perfectly flat pitch. Elevation Plot from Garmin Forerunner 405 GPS here

I 'climbed' 2066m in my sleep one night too (I forgot to disable the GPS when using the Forerunner to log my HR during a nights' sleep). Going from 26m (which is about right), even going as far down as 62m below sea level and as high as 120m ASL. Magic.

So, back to the spreadsheet with my shiny new data. This time it totted up 5972m of climbing. HHmm  well better than last time but still over a 1000m above Mr Dulates figure. I then got the spreadsheet to flag up just the peaks and troughs in the elevation data. I stripped out all the rest. This reduced the file from 4366 trackpoints to a mere 1193 with the total climb remaining at 5972.   Then I got the sheet to flag up all those peaks preceded by at least 9m of continuous climb and tot up a new climbing figure based on these only. It found 164 peaks over 9m and totted up 4706m of climb (close agreement with John’s figure). The ignored less-than-9m peaks are the minor undulations that can occur unnoticed between 10m contour lines.
This is the climbing you miss in a contour count.

I've got a very similar thing done as a perl script working off GPX tracklogs (or GPX files from routes created on Bikely and the like).

GPX/GPS utilities/scripts

It works by removing intermediate points (b where: a > b > c and a < b < c). You're then left with a set of alternating points of peaks/troughs. You can then apply an algorithm to get rid of peaks that climb below a certain amount (i.e. 10m) and then apply the first algorithm again. Repeat until you don't remove any points.

I use it for picking out the major peaks (over 50m climbing) of a planned ride (based on the bikely GPX) to add as extra waypoints so I know where the tops of the hills are; this allows me to pace myself up the hill and also helps avoid the soul destroying 'I must be almost at the top by now' as you round a corner to see another several km of climbing...

With a lower limit for climbing (c.f. 10m) I've had climbing figures that agreed with the organisers figures, although the climbing figures for some rides seem to be based on the raw figures from a GPS, so it seems like a hillier ride than it actually is.

I hoping that the release of the Ordnance Survey data (including spot height database) will help as the OS spot heights are often placed directly on the road at the top (and bottom) of roads (and therefore climbs). That data will allow for a far more accurate idea of climbing figures.

I'm also hoping that this can help in providing a consistent figure for AUK climbing. I'm sure Mr Snook and Mr Coates will be looking at this, and I hope I can do something to help too.

Anyway, probably a better idea to continue one of the other threads on this subject:

GPX/GPS utilities/scripts
Altitude measurement - who to believe!
etc..

rather than clog up the MC1000 thread.

 

Holy shit - and you're not even riding it Alex!

Were you in any way involved in the Mars Beagle Lander?

Hmmmmmmmm?

H

[simonp]

 

Holy shit - and you're not even riding it Alex!

Were you in any way involved in the Mars Beagle Lander?

Hmmmmmmmm?

H

I think this is Hummers' way of saying we're all doomed. 

[simonp]

12 weeks and counting.

F**king hell.  That sounds so soon 

3 months to gain 50 Watts and lose 2-3kg.  Easy...