proof of predict.habhub.org

[Logan Geeves]

Hi. I'm a teacher in NZ with a balloon, tracking and filming equipment ready to go. We have been held up getting a clearance to launch from the airways control center in NZ. We need to provide proof that the website http://predict.habhub.org/ works. Does any one know how to contact the organizes of the site or who has used it before and can provide some proof that it works. 

[Anthony Stirk]

HI Logan,

I don't think there is any "proof" it works as the underlying data it uses is the GFS (Global Forcasting System) 192h from http://www.noaa.gov/. It is like any other weather forcasting data subject to some error margin and should be used in an advisory manner only. That said we've found it generally reliable in predicting the general direction and landing spot of a payload globally 1-2 days in advance. 

Used in combination with the http://habhub.org/calc/ you should get a reasonable level of accuracy on where abouts your balloon is going to go. We can set you an hourly predictor (example http://predict.habhub.org/hourly/cambridge) if needed this gives you every possible landing spot for the next seven days by the hour. Click a point to see the track, this is useful for seeing how the winds are changing. Let me know if this would be of interest.

Regards,

Anthony M0UPU

On Mon, Feb 17, 2014 at 3:30 AM, Logan Geeves <lge...@my.torbay.school.nz> wrote:

Hi. I'm a teacher in NZ with a balloon, tracking and filming equipment ready to go. We have been held up getting a clearance to launch from the airways control center in NZ. We need to provide proof that the website http://predict.habhub.org/ works. Does any one know how to contact the organizes of the site or who has used it before and can provide some proof that it works. 

--
You received this message because you are subscribed to the Google Groups "UKHAS" group.
To unsubscribe from this group and stop receiving emails from it, send an email to ukhas+un...@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.

[Steve Aerospace]

Assuming “works" means "is accurate”  then the proof of the pudding can only be in the eating.   

As long as you run the forecast as close to the lauch time as you can the CUSF predictor is quite accurate if you get all the flight parameters correct :

* measurement of payload weight & gas fill (for correct ascent rate), 

* burst altitude (balloon quality)

* descent rate (parachute sizing).

The trick is getting all the parameters right. 

The other factors - 

* accuracy of GFS model (variation from actual), 

* accuracy of the CUSF predictor models of ascent and descent 

* ballon burst estimator model  

are less of a factor than getting the flight parameters correct in my opinion.

I’ve done 50+ flights using it and I can only recall a couple of times when the results were out more than I expected.  I’m sure other people can vouch for it.

As you would expect accuracy of predicted landing spot is a function of ground track length.  Short Straight line flight paths give the most accuracy - Long ZigZag flight paths much less so.  Deviation from predicted landing spot is much less perpendicular to the flight path than it is parallel to the flight path.

NZ looks as challenging (if not more so) as the UK to launch and land in - quite narrow.  Like the UK you will need the predictor if you don’t want to loose your payload to the sea.

Being cheeky you might ask what the NZ met office use when predicting the landing spot for their Radiosonds?    ;-)

Steve G8KHW

[Felix Plöger]

Hi,

from my experience I can tell that the landing predictor is pretty accurate in most cases. 

As Steve has pointed out already, the correct parameters are important. Also keep in mind that weather changes frequently, and especially under stormy weather the predictions between different GFS Models can differ quite a lot.

If you wanted to use a second source for doing your own predictions, you could also use the Wind-Data produced from local radiosonde launches done by your local MET-Office. As long as the MET-Office reports to the archive of University of Wyoming in a timely fashion you are able to find recent launch data here:

http://weather.uwyo.edu/upperair/sounding.html

From the "Region" dropdown select "New Zealand", then select the one of the two stations which is closest to your launch site.

A new tab will pop up with an url like this and containing a more or less human readable list, see below for an example:

http://weather.uwyo.edu/cgi-bin/sounding?region=nz&TYPE=TEXT%3ALIST&YEAR=2014&MONTH=02&FROM=1621&TO=1621&STNM=93844

This data can be stored as text file and used for analysis to give you a less accurate prediction than the online one might be, you can use the "Ballon Track" software available here:

http://www.eoss.org/wbaltrak/

This program is pretty fiddly, so it will be less comfortable in use than the online predictor.

Another software product which I am using for most of my local predictions and while being mobile is Allen Jordan's "Balloon Prediction" which basically works the same way as the CUSF Predictor but also gives you the ability to use current wind-data from radiosonde launches. It also lets you define altitude depending vertical speeds. You will have to contact Allen by email to obtain a copy, I would assume he would be happy to help you. All details can be found here:

http://www.allenjordan.info/balloonprediction.html

I hope this helps, good luck for your project.

Felix, DL1SGP

2014-02-17 Steve Aerospace <st...@randomaerospace.com>:

[Steve Aerospace]

Also if you want to assess the accuracy of the CUSF processing of the GFS model then the Uni of Wyoming also have a GFS based predictor:  http://weather.uwyo.edu/polar/balloon_traj.html    From what I remember It has several limitations: fixed ascent and descent rate - and no interpolation between GFS models - its also a bit painful to use in comparison - but useful to check CUSF predictor for accuracy.  I spent many an hour comparing predictions when the CUSF predictor was new.

I’ve always assumed that there is no common implementation (other than the GFS itself) between the UoW and CUSF predictors - perhaps someone from the HABHUB team can comment?

Also anyone know how to check which radiosonde wind data is built into the GFS model?

Steve G8KHW

[Daniel Richman]

Hi Logan,

I’m from Cambridge University Spaceflight — we are (primarily)
responsible for building the predictor (more details in the “About”
page top right).

To quote that page “no guarantee is given for the accuracy, precision
or reliability of the data produced by this software, and you use it
entirely at your own risk”.

I must stress that we (CUSF) cannot provide you with any proof or
guarantees, and we must distance ourselves from the endorsements
(flattered as we are) from others on the list lest we become in some
way responsible for any problems with other peoples flights (which we
do not want). Besides, we are yet to conduct a proper “scientific”
study of its accuracy.

For what it’s worth, that’s a fairly unusual requirement. I’m not
aware of anyone else being asked to “prove that it works”. Have they
even given you a reasonable definition of “works”, tolerances etc.? At
risk of stating the obvious, you should be designing your payload to
be as safe as possible when landing in case it does go somewhere you
don’t expect.

I hope you understand,
Daniel

[Nicholas Janzen]

I can also vouch for the accuracy, I've used it for around 15 flights in Canada, and we have almost always found the balloon within 10 KM of the prediction.

I remember times when it was a little off, we had just found the path to be accurate, just stretched out (as in the winds were stronger then we thought or our ascent rate was slower then we punched in).
--

/-------------------------------------------------------------------------\
| Nicholas Janzen Personal Site: http://obsd.com |
| Email: n...@obsd.com Security Site: http://secure6.net |
| VE6OBS VE6TS (Basic+Advanced) Weather Site: http://nicholasjanzen.ca|
| Short URL's: http://clearurl.net/ Balloon Site: http://arawr.ca |
\-------------------------------------------------------------------------/

[chris hillcox]

Hi Logan, I really like to post the balloon trajectory forecast and a map that shows the actual landing site. Here is an example from last year. The CUSF forecast was pretty accurate as you can see:-

http://nsballoonproject.wordpress.com/missions/mission-3/

Cheers,

Chris

[Logan Geeves]

thanks chris

[Kristaps]

Hi, Logan!

Here's another comparison of the predicted and actual flight path from last year:

http://space.people.lv/intra/LAASE-1-prepost/

http://space.people.lv/intra/LAASE-2-prepost/

LAASE-1 balloon (Hwoyee) burst a few kilometers higher than the burst calc assumed, ascent and descent rates were about as expected.

LAASE-2 (Totex) burst prematurely several kilometers lower and descended much faster so not much of a "proof", but it still wasn't far off from the predicted path – helps if it's ZigZag-ish. 

Our CAA asked us for the proof that CUSF Landing Predictor is trustable, too. After long talks and with few restrictions they gave us the permission anyway.

Good luck!

Regards,

Kristaps

[Ed Moore]

I talked about its accuracy for a real use-case for about a minute here: http://youtu.be/sU6j49eCkF4?t=21m2s

[Wolfbl]

Hallo,

because i'm running my own model for trajectorys of meteorologic radiosondes,

i have made many,many benchmarks with predict.habhub.org  against my model and real flights.

My conclusion is :

- predict.habhub.org has well implemented mathematics to transform the wind-predictions in a trajectory.

- the model of ascending and descending speeds can be improved specifically for the behavior of radiosondes.

In detail :

a.) habhub.org assumes a linear ascending speed. In real you find high ascending speed near gound , less speed

in the middle of ascending and increasing ascending speed in the upper half of ascending. 

For instance:

This will produce a significant predication error in case of high stratospheric wind speeds. Habhub move up the

ballon to slow in this hight, thererfor it remains to long in this region and the balloons is moved horizontal to much.

(up to 30 km error with 100 m/s in 10 hPa  and burst above 10 hPa)

b.) habhub asumes the descending speed progression is equivalent to CONST/ sqrt(density)..

    in real i found the descending speed is more equivalent to CONST/sqrt(pressure). (Why ever ,viscosity effects??)

    

I think,

predict.habhub.org is reliable for all flights with unknown specific higher order  ascending or descending parameters.

In case of well known paramters of radiosonde launches ( same stuff every day ! )  improvement are possible.

Best regards

WolfB.

[Ed Moore]

There mechanism for the speed increase about half way up is, I think, caused by the well-known knee in the curve of drag-coefficient vs reynolds number for spheres (modelling the balloon as a sphere, a decent assumption for once). Example of the curve here:

http://www.grc.nasa.gov/WWW/k-12/airplane/Images/dragsphere.jpg

You can see the drag coefficient drops quite suddenly between Re = 10^5 and 10^6 for a smooth sphere, so we'd expect to see velocity increase at the conditions when the Reynolds number is in that range.

Reynolds_number = density * velocity * characteristic_length / dynamic_viscosity

Re arranging for density (so we can calculate altitudes we'd expect to see the velocity increase):

density = reynolds_number * dynamic viscosity / (velocity * characteristic_length)

some rough numbers for a typical hab

dynamic viscosity is a function of temperature and is about 1.3 x 10^-5 kg/ms at -30C

velocity is about 5m/s

characteristic length is a wishy washy concept for a typical size of thing in the flow, diameter of the balloon is a good bet, let's say 3m

reynolds number is, let's say 3 x 10^5 (eyeballing that graph for where the dip occurs)

if we substitute those rough numbers in we get a density of about 0.15kg/m^3, which corresponds to an altitude of about 15km.

So, you'd expect to see the velocity increase at about 15km (plus or minus, given I've just picked some fairly rough values), and last time I looked that was about where it occured.

However, i think this effect is quite small compared to, say, the rather rough and ready estimates people use of their sea-level descent rate, into the predictor. Parachute systems are harder to model than balloons!

Ed

[Steve Aerospace]

An interesting comment - so I take from what you are saying that if:-

balloon size, 

gas type & fill, 

payload weight

parachute size  

are held constant then the variation in:-

actual balloon mass/shape/construction

burst altitude

balloon remnants - weight and potential tangling.

time of day/year atmospheric pressure/density variation.

GFS to actual

produce significantly less variation than the error in the ascent/descent rate model?

I have noticed that the descent rate model used in the habhub tracker (I assume this is the same as the predictor) is

out. The prediction of landing spot takes quite a while to "home in" on the actual landing spot and most often predicts long.  I’ve found (imperially) an e^x model where x is a function of height gives and amazingly good fit to actual descent rate - it starts giving very good landing spot prediction within a few minutes of burst (assuming the payload/parachute/remnants configuration does not change on the way down).

Steve G8KHW

[Steve Aerospace]

Yes Ed you can often see some step changes in rate in the altitude graphs - but also a slow continuous acceleration or sometimes continuous deceleration with altitude.  The amount depends on lifting gas used (Helium often has a straighter altitudeVtime graph than hydrogen).  In my experience the amount is also dependant on balloon size and payload weight.  Hydrogen with a light payload shows a very marked concave curve on the altitude graph. 

I believe that unconstrained by an elastic envelope the balloon+payload would naturally accelerate with altitude.  My feeling is that the the envelope elasticity constrains the acceleration (i.e. has a decelerating effect with increasing altitude) - and can be less or more (or close to balance) than the natural acceleration with altitude.  Also note the elasticity of the latex will vary with the local temperature.

Other factors:

Slower ascent rates near the ground before the balloon fills out to a sphere.

A difference in shape between Totex and Pawan/Hwoyee balloons (because different construction techniques are used)

this is most noticeable at low inflation. 

Steve G8KHW

[Frits PE2G]

Hi,

I've used the predictor for trajectory predictions of many hundreds of
weather balloon flights in the Netherlands, Belgium and Germany.

When the predictor uses a ~6 hour old GFS wind forecast model, landing
site prediction is usually accurate to around 10 km, provided that the
correct burst altitude, ascent and descent rates are entered in the
predictor.

Regards,

Frits PE2G

Op Mon, 17 Feb 2014 04:30:15 +0100 schreef Logan Geeves
<lge...@my.torbay.school.nz>:

--
| PE2G | JO32HI | pe...@xs4all.nl |

http://home.deds.nl/~pe2g