News IBM Weather Report.

[Notebook]

CES 2019: IBM's hourly weather reports will cover entire Earth

But besides this crowdsourced data from members of the public, Graf will also analyse information from thousands of commercial flights.

All based around atmospheric pressure?

https://www.bbc.co.uk/news/technology-46790221

 

[Urwumpe]

All based around atmospheric pressure?

https://www.bbc.co.uk/news/technology-46790221

No, but also. Just pressure wouldn't be enough to describe air. The big difference to classic forecasting seems to be, that they plan to include smartphone pressure sensors into the calculations. But this would then depend on the density of smartphones in a small region to get somewhat accurate data.

 

[Marijn]

That's interesting. I didn't know there are many smartphones with atmospheric pressure sensors. I believe such a system must correct for the elevation of the terrain where the smartphone is, ̶s̶o̶ ̶t̶h̶e̶y̶ ̶n̶e̶e̶d̶ ̶G̶P̶S̶ ̶d̶a̶t̶a̶ ̶a̶s̶ ̶w̶e̶l̶l̶.̶ ̶D̶o̶ ̶p̶e̶o̶p̶l̶e̶ ̶h̶a̶v̶e̶ ̶t̶h̶e̶i̶r̶ ̶G̶P̶S̶ ̶r̶u̶n̶n̶i̶n̶g̶ ̶a̶l̶l̶ ̶t̶h̶e̶ ̶t̶i̶m̶e̶?̶ ̶I̶ ̶o̶n̶l̶y̶ ̶t̶u̶r̶n̶ ̶i̶t̶ ̶o̶n̶ ̶w̶h̶e̶n̶ ̶I̶ ̶n̶e̶e̶d̶ ̶i̶t̶ ̶t̶o̶ ̶s̶a̶v̶e̶ ̶t̶h̶e̶ ̶b̶a̶t̶t̶e̶r̶y̶. - Tringulation would do - Also, I wonder how much the measurement is affected by temperature. How do these sensors work? Not with a membrane box I think.

 

[Notebook]

I didn't know there were any smartphones with pressure sensors! But that's my age.
If its using all that phone data in real-time, that's a lot of data crunching?

N/

 

[Urwumpe]

How do these sensors work? Not with a membrane box I think.

Well, according to my smartphone (Motorola Nexus 6), I have a TDK Invensense sensor suite, with MEMS barometer (measuring hPa).


So, practically a microscale membrane box.

 

[Marijn]

So, practically a microscale membrane box

Interesting. I've been researching the question whether altitude measured with a barometer is more accurate than GPS data some years ago. My conclusion then was that GPS data is more accurate, mainly because of the effect of external influences such as temperature and moisture on the sensors.

So I wonder what happens if someone takes the stairs to the roof of a high building. Increased body heat and lower pressure because of the elevation would distort the data I would think.

---------- Post added 01-09-19 at 12:06 AM ---------- Previous post was 01-08-19 at 11:52 PM ----------

But IBM's new tool provides reports down to more specific, 3km-wide areas.
The company says it can even predict individual thunderstorms.

That's also very interesting. To predict thunderstorms with this accuracy, you need to know the vertical temperature distribution of the atmosphere at the location from the ground up.

Many years ago when I was working on a glider airfield in Austrialia, every morning before breakfast, I hopped in the backseat of a Belanca Scout plane for a 'temp trace'. That meant holding a thermometer outside the window in the free air and writing down the temperature every 300 feet. After landing, I averaged the values of the way up and down and gave the results to the meteorologist. He then was able to predict very accurately the start time of the thermals, their vertical speeds and the time at which they would over-develop into thunderstorms.

This makes me wonder if IBM does have access to the data from ESA's Aeolus project. That makes more sense to me than the usage of smartphones to predict thunderstorms.

 

[Urwumpe]

Interesting. I've been researching the question whether altitude measured with a barometer is more accurate than GPS data some years ago. My conclusion then was that GPS data is more accurate, mainly because of the effect of external influences such as temperature and moisture on the sensors.

Depends on the GPS... I remember that I saw some huge errors in altitude with GPS on some days during a past employment. We had been 9 meters off in our measurements including WAAS/SBAS corrections using a fixed antenna on the roof with a $20K LRU GPS receiver as used in aircraft. Without WAAS, the best you can get in altitude is about 10 meters (30ft), but it can be up to 30 meters off.



An aircraft barometric altimeter can be accurate to single feet at low altitude and with favorable conditions, but it should at least get 80 ft accuracy in 10000ft altitude in the worst case to comply to US standards. That is still more accurate than GPS can be in the worst case (about 150 ft off). You would require DGPS to beat it.

 

[Marijn]

Depends on the GPS...

Actually, the logic was that GPS is not affected by external influences at all and therefore, overall, more accurate than barometric height. Especially membrane boxes are subject to changes in temperature.

I found that many GPS devices and apps display their data incorrectly or at least suboptimal. My Samsung phone always showed an offset of about 150 feet. When I created an app for it, I discovered that I had to correct for geodetic height myself. Once I did that, the values were very accurate. I am quite sure the 150ft offset you mention is caused by this as well. The geodetic offset in Western Europe is quite big.

---------- Post added at 01:22 AM ---------- Previous post was at 12:43 AM ----------

I think it's perfectly normal for a GPS which has a good signal to ouput +45 meters anywhere in Europe. See the image below. The developers of apps usually output the raw data without bothering about this. But if you would substract the geodetic offset of the area you are, the data becomes quite accurate.

Barometric height is based on the standard atmosphere. But the real atmosphere on a particular day is always different. It's favourable for GPS data when these differences tend to be bigger.

In aviation, all altitudes such as the ceiling of restricted areas are referenced to the standard atmosphere. That means everybody is making the same error and therefore the data can be interchanged. Not per se that the error is small.

If you want to measure the amount of meters between an airplane and the hard ground (which unlike artificial ceilings do not care about mathematical models), I doubt whether altitudes based on barometric heights will give better results than GPS altitudes which are properly corrected for geodetic height.

WSG-84 Geoid Height: https://www.google.nl/imgres?imgurl...9_fAhX_BGMBHaEFC0wQMwhCKAIwAg&iact=mrc&uact=8

WGS84 defines the Earth as an ellipsoid: a squashed ball. This ellipsoid is a pretty good approximation to the mean sea level around the planet, but is recognised as having errors of between -100 m and +70 m with respect to the geoid, depending where you are on the planet.

 

[Urwumpe]

Its is not geodetic offset - it is ionospheric! Accuracy of GPS depends highly on solar activity. And remember: We used an antenna on the roof of a building, we knew very precisely where the antenna was relative to Earth. And we recorded GPS signals 24h a day.

 

[Marijn]

Its is not geodetic offset - it is ionospheric! Accuracy of GPS depends highly on solar activity.

We're talking about different things now. Space weather does indeed effect GPS. With +- 15meters on the Z-axis, it's the biggest contributor to inaccuracies. So like the standard atmosphere, there is a standard space 'atmosphere'. GPS receivers assume there are no special space weather related events going on.

A GPS receiver does know it's position relative to the satellites. Without adding data, it doesn't even know Earth exists, let alone what your position on it's surface would be. Therefore a model is required. The most common model is WGS84. WGS84 defines the Earth as an ellipsoid: a squashed ball. GPS receivers hold the model in their memory. Sometimes other models can be selected.

Like any model, it is not 100% equal to the real thing.

The picture I posted previously shows where the errors lie. In Europe the ellipsoid is about 50m above the geoid (note thats quite like 150ft). In south India it is nearly 100m below. Some GPS systems using WSG84 add the geoid correction to give geometric height and some don’t.

Have a look at this picture: https://nl.wikipedia.org/wiki/Geoïde#/media/File:Geoida.svg

..we knew very precisely where the antenna was relative to Earth.

Can you tell me now how you determined the precise Z-position of the antenna on the roof of your building?

 

[Urwumpe]

Actually quite simple. We used the very accurate geodetic reference data for Germany (DHHN85) and converted it to WGS84:

https://www.lgln.niedersachsen.de/g...etze/geodaetische-grundlagennetze-143648.html

Since we had no mines underneath or floating pogs, this data should be accurate to about 10 mm, also the reference altitude did not change, contrary to the eastern states of Germany after the reunification (Also, the new DHHN2016 shows less than 1 cm difference to DHHN85 in this spot) . And right next to us was the airport of Braunschweig, another reason to be sure that the geodesy is reliable.

Finally - the building was constructed around 1935 with thick stone walls, as common in the Nazi era - only little concrete used. Height from foundation to antenna about 6 meters, maximum known range of temperature extremes is +/- 30K - even assuming the worst coefficients, its just 0.45 mm thermal expansion between record low temperature and record high temperature. So error in altitude to WGS84 is limited to about 12mm.

At least we can claim the following: We knew the location of our antenna far better than all smartphone users and aircraft.

 

[Marijn]

Ok. That sounds like you did pay attention to these datails. Setting the best geodetic offset is the same as I did. Then I cannot exaplain why the GPS altitude you mentioned was so inaccurate. Unless you were employed at a time when there was indeed more than normal solar activity. Could that be the case?

I think all GPS systems, including the cheap Chinese boards, should be able to give an altitude accuracy well below 150 ft.

 

[Urwumpe]

Ok. That sounds like you did pay attention to these details. Setting the best geodetic offset is the same as I did. Then I cannot explain why the GPS altitude you mentioned was so inaccurate. Unless you were employed at a time when there was indeed more than normal solar activity. Could that be the case?

Not really, it was around 2006-2008, at solar minimum. We had a few X-flares in the space weather newsletter every other week, but rarely strong activity.

Still, those few events had a strong effect on the measurements. "Should be able" is also not really the problem. My smartphone can be about 2 meters accurate on a good day when geocaching, it did not fare too bad against better receivers. The problem is, can you rely on it? On a good day GPS is better than barometric. on a bad day, it is far worse. On the average, a good and properly calibrated barometric altimeter is only homeopathically better than GPS with WAAS and both are worse than a radar altimeter.

GPS with WAAS/SBAS and LAAS/GBAS should be able to compete well with CAT I ILS (<4m vertically), but still fails to compete with CAT III ILS, which a future standard should achieve.