Why temperature readings in decimals is important

  • 4
  • Idea
  • Updated 3 months ago

By now it appears that the admin of the new WUnderground is
either unwilling or unable to accommodate their patrons in regards to the
display of decimal values for temperature recordings. No surprise there. Perhaps whomever it was within their team that was responsible for this compromise approach to the interface felt somehow justified that whole numbers alone are sufficient when displaying Celsius or Fahrenheit readings - even though this represents a compromise of the integrity of contributors’ weather data from their PWS.

The basis of this post is not to persuade the admin team that this approach is wrong because this is most likely a futile attempt but rather to remind the viewing public why this is so.

Part of the problem perhaps is the insistence of the US to resist the adoption of SI units (the metric system) like every other country in the World. Since the US clings to the archaic default method of measuring temperature in whole degrees Fahrenheit (this is still the accepted method of The National Weather Service) certain companies not mindful of the Global standard think that this is okay and even suppose that conversion to even cruder whole degrees Celsius readings are likewise acceptable. However, WMO guidelines proclaim that the resolution of automatic weather stations should be in 0.1 K (0.1 C) with an uncertainty of 0.2K. (0.1 deg Celsius ~ 0.2 deg Fahrenheit). The meteorological community worldwide recognizes that modern instrumentation is sufficient to display it as such which is why practically every National Meteorological Service in the World records official temperature in a resolution of 0.1 deg Celsius. Likewise, American users have become accustomed to their own weather stations and MesoNets displaying in tenths of degrees Fahrenheit since the manufacturers of even the most basic temperature-recording devices have set the precision of their instruments to this level.

To disregard the virtue of precision in the reporting of scientific data when it is conveniently available does the end-user a disservice and is an embarrassment to the supplier of this data.
Photo of Patrick

Patrick

  • 1 Post
  • 0 Reply Likes

Posted 4 months ago

  • 4
Photo of Roger Lewis

Roger Lewis

  • 14 Posts
  • 4 Reply Likes
I agree with you. and to add to the discussion I have never understood the logic of "rounding up" when it comes to reading and reporting ambient temperature as with 64.4° is read and reported as 64°yet 64.5° is read and reported as 65° when there is ONLY 0.1-degree variance between the two. same as 64.5 and 65.4° are both read and reported as 65°when clearly there is 0.9degree variance. it doesn't seem to be an accurate method. I am VERY disappointed with the so-called "NEW and IMPROVED" Wunderground website. At least there should be a resolution option in the settings to use 0.1 ...OR... the whole number "rounded up" method.
Photo of jachen

jachen

  • 9 Posts
  • 4 Reply Likes
I think this application is being done by a machine (AI). Cannot talk to humans, has no empathy. Eventually improves things as it did on the underlying google map on a MacBook Pro running Chrome.
Its probably very difficult to migrate all WU services one by one to IBM's Watson. It might take years to finish. And we might get back the decimals on temperature. Be patient or leave.
Photo of Salonika

Salonika

  • 3 Posts
  • 6 Reply Likes
It is not about being patient or not. It is about doing things properly and right. You cannot be (probably) the most important PWS website providing lots of services and meteorological stuff, and suddenly, throw all of it off the boat without the security of keeping your standards and not bothering all of the people that are, for free, giving you their data to support and feed your business.

A lot of people, me included, rely on this portal to see and check our data since a most of current PWS upload data here without the option of keeping it at our PC's. And all of a sudden, you change it all, lose most of your identity and leave your data providers without such a crucial thing as seeing the data just as it is, with decimals (not to talk about downloading your OWN information). It seems unbelievable that such a big company allows itself to do such a poor work. The fact that we are here discussing, 4 months after the change, of such an obvious thing in meteorology as the resolution of the data is pretty amazing. Updates in websites should always be towards a more clean, detailed, trustworthy and rich content, and this is just non-sense...

I hope we get everything fixed soon.
Regards from a spanish fellow.
Photo of Roger Lewis

Roger Lewis

  • 14 Posts
  • 4 Reply Likes
I think they should've made all the changes and so-called updates to the website in BETA and ONLY released when EVERYTHING had been tested and completed... thus leaving us with at least a half functioning website.. in MY opinion, Wunderground has never been the same since they "updated" the website back in 2009-2010 or whenever it was.
Photo of Ul KE

Ul KE

  • 13 Posts
  • 18 Reply Likes
@Salonika: Well explained and nothing to add on this. I have also written a lot of E-Mails in the meanwhile to address this problem, with no response! If there will not be a change within the next few days, I will discontinue providing my (very accurate data) to this weather service.

This is not that, what I expect from a professional weather site!

I am sorry to do that, but I am really, really pissed off right now about the current situation!
(Edited)
Photo of Paul Grace

Paul Grace

  • 23 Posts
  • 5 Reply Likes
I'm surprised that you guys seem to think you have thermometers that accurately report the actual temperature to a 10th of a degree, even under the best of circumstances.  Read the specs...if is says "± 0.1F" it doesn't.  Precision equipment is expensive to buy and maintain.
Then add in all the rooftops, back decks, fence posts, above lawns, next to the house, close to a tree, near irrigated plants- siting errors and the lack of proper solar shielding and regular calibration ($$$) to a known reference standard and you're off by several degrees.

Folks, you're forgetting the difference between resolution and precision.  Just because your display is printing a number in 10ths doesn't mean that's an actual temperature.  You're in the ballpark, but you can't tell if it's left field or a hotdog cart in the stands behind home plate.

Relax.

(Edited)
Photo of Doc_Bill

Doc_Bill

  • 3 Posts
  • 2 Reply Likes
Ill call BS on that 1/10 calibration. That would cost you a fortune to get and keep your WS that close to a NIST standard. Standard ASTM calibration thermometers are usually good to only +/- 0.2F.
Photo of Paul Grace

Paul Grace

  • 23 Posts
  • 5 Reply Likes
It doesn't do that either, James. 
But now I see a bit more about how you folks think, and why they won't take you seriously.
(Edited)
Photo of Ul KE

Ul KE

  • 13 Posts
  • 18 Reply Likes
@Paul Grace: Thats not the point. When you look at digits, you can have 5,6 or 6,4 °C. Both is rounded 6 °C, but in practice, there is a difference of 1 °C. Even worse is it, if you look at the average of a day, month or a year. 
And yes, For us, who bought a high sophisticated weather station, who provides this kind of data, it is just a hugh step backwards to see such inaccurate Data on such a sophisticated weather page, which wunderground would like to be (or was in the past).
Photo of Paul Grace

Paul Grace

  • 23 Posts
  • 5 Reply Likes
"For the rest of us 1/10 degree reading are normal and obtainable"

<chuckle>

No, they aren't obtainable from consumer-grade weather boxes.
Suppose you have a top-of-the-line Davis that you spent more than $1000 on.
The Davis sensor has a non-linear error of ±0.3°C (±0.5°F) so if the temp is exactly 6°C, Davis might report that anywhere at random between 5.7°C, to 6.3°C, a 0.6°C range.  Rounding that at 0.5°C does not affect the data, which is really only good to 0.6°C.  Then the fan-aspirated box adds another ±0.3°C (±0.6F) error on top of the sensor error, assuming it is sited perfectly.  If you don't use the fan-aspirated sensor, it just gets worse, with errors of up to 4°F added in.

You guys are fixated on what we call "false precision"  The sensor errors are at least  1°C, so measuring it to 0.1°C is pointless, and worse, you all think you are getting better data than you really are.  

James suggests using a 4° rounding, but of course when your error is a random ~ 1.2°C spread, you should round to a 1°C with a digital display, since the 0.1°C is imaginary and the 1°C is pretty close to correct.  That's exactly why WU doesn't bother to report that 0.1°C - because it's wrong.  They understand the data quality issue, which is why they drop the false precision...there is no point in using it because it isn't real.

And in Fahrenheit, consumer temps are only correct to the nearest whole degree.  Fractions are just figments of your imagination, numbers displayed at random on your box.

My instrument suite is not the issue (although my anemometer alone costs more than 3 davis weather stations-the mast it is attached to costs more than a Davis weather station)
Understanding the difference between precision and resolution is the issue.
Understanding the sources of error are the issue.
(Edited)
Photo of jachen

jachen

  • 9 Posts
  • 4 Reply Likes
You are right as far as sensor precision and physics o f the system ist concerned and also taking the error correction calculation into account one degree is about enough.

On the other hand there is also a question of aesteics of the temperature plot. A stairs plot like the one we get right now hurts ones eyes. This might be the reason most weather companies in Europe and Asia and many other more plot their temperature graphs considering decimals. 

And my question now: How come IBM ist plotting the pressure graph even using 2 decimals? I am stuck.

Sample Aestetic Plot (Source: Meteo Switzerland):
Photo of Paul Grace

Paul Grace

  • 23 Posts
  • 5 Reply Likes
Thank you- Aesthetics is a valid reason... but since the data isn't good enough to warrant displaying the noise, it creates a false sense of accuracy... which does harm the understanding people have of their own data.  Look at this thread and that it's been running for months - many because they erroneously believe their "data" has been lost or some people might say "taken away by the greedy IBM Corporation."

My opinion is that the resource would be better used teaching people why 1 degree is all they can have under the best conditions, because it is all they can measure given the accuracy of their equipment.

I don't mind if the fractions are printed, since it's only an aesthetic decision - but I don't think people should become so exercised about the aesthetic choices of an overworked, under-appreciated programmer.

Whoever you are, I thank you for providing what we do have, even though it is not be perfect for everybody.
(Edited)
Photo of Ul KE

Ul KE

  • 13 Posts
  • 18 Reply Likes
Sorry, the accuracy of the Davis temperature sensors is at 0,1 °C. But also thats not the point. When it comes to the calculation of a daily, monthly, yearly average, you need the digits to get an accurate result and this is now not possible looking to the current resolution of only 1 °C.
Photo of Paul Grace

Paul Grace

  • 23 Posts
  • 5 Reply Likes
Not according to Davis.  There are 3 different sensors used (SHT31, SHT11 and the SHT 15) and none of them are that accurate.  The 31 is the most accurate, is not suitable for all VP stations, and is not that accurate.  But don't take my word for it.

https://www.manula.com/manuals/pws/davis-kb/1/en/topic/temperature-humidity-general

Photo of Paul Grace

Paul Grace

  • 23 Posts
  • 5 Reply Likes
"you need the digits to get an accurate result"

Since the decimal digits are not accurate, using them in your average pollutes your average.  It doesn't make your average accurate.
(Edited)
Photo of William J Schmidt

William J Schmidt

  • 31 Posts
  • 14 Reply Likes
The best Davis sensor is the 7346.070 SHT31 Upgraded Digital Temperature/Humidity sensor's temperature accuracy is +/-0.5°F (+/-0.3°C)... That means you could be off a whole degree.  And only available after Jan 2016.
Photo of Paul Grace

Paul Grace

  • 23 Posts
  • 5 Reply Likes
Accuracy: The difference between the reported value and the actual value.
Precision: The repeatability of a steady value
Resolution: The smallest change that can be measured.

As noted above, the best Davis sensor's best-case accuracy is only ±0.5°F, ±0.3°C
It is bad to report that with a false precision of 0.1° because it is misleading all of you.

Everyone that believes your consumer equipment is giving you .1° accuracy is mistaken, it isn't - you are being misled.  There are also additional errors caused by mistakes in design and in your installation that can add several degrees of error on top of the basic sensor errors, introduced by humidity, wind, sun, and yes, even by the temperature... above 90°F the accuracy gets even worse.

Just because it's printing numbers on the console doesn't make them real.
Photo of Roger Lewis

Roger Lewis

  • 14 Posts
  • 4 Reply Likes
Paul, given your continuous rant on this subject I take it you do not now, or have ever, nor will ever own a PWS due to the inaccuracies you have pointed out. as It would be a complete waste of your money. WHY  on earth would you invest in a product that is incapable of providing TRUE100% accuracy in ALL values of measurements?
 Based on your "knowledge" of this subject I have NO doubt you have been sought out by NOAA and The NWS to assist in calibrating their sensors... which I might add my PWS seems to be just as accurate as my local AWOS sites which I compare my data with regularly Thank you very much.
Photo of William J Schmidt

William J Schmidt

  • 31 Posts
  • 14 Reply Likes
One point of clarification from a guy with an advanced math degree...  One never rounds data before averaging it.  You average raw data (one or two or ten or 50 decimal places... valid or not) and then round the final result to the number of significant digits.  If you don't do this, you add additional bias by skewing the solution higher or lower.  If you have a good reference, at least as accurate as the device you are using, It gives you additional data to determine the statistical precision and accuracy within the specified numbers for both sensors.  As I said previously, I have no problem showing additional decimals as an aesthetically pleasing, feel better about my investment, etc. thing.  I just wouldn't get too serious about those decimals as they are probably meaningless... and even worse given the shear differences in suboptimal weather station installations (locations, heights, etc.).
Photo of Charles Modica

Charles Modica

  • 1 Post
  • 2 Reply Likes
I have a very hard time understanding why they won't adjust the site to show decimal information, because it's still being logger/captured. If you use their widget code to show your weather station data, it will still show temperatures down to the decimal.

The new web interface has been a significant step backwards in terms of how granularly we can access our data (i.e. decimals, specific date ranges, etc)
(Edited)
Photo of Victoria Gardner

Victoria Gardner, Official Rep

  • 644 Posts
  • 92 Reply Likes
AHEM!!

It's completely possible, I hope, to disagree without falling into ad hominem attacks (calling into question someone else's intelligence, for instance).

The decimal question is quite important for metric users.  We know that.  This is purely a question of allocation of scarce resources (programmers) to problems. 

Fahrenheit users may not need the added level of precision.  Or they may want it anyway.  The beauty of a community is that we do not all need to be exactly the same. 

--Victoria
Photo of DJTH

DJTH

  • 10 Posts
  • 2 Reply Likes
Or you can bring back the old page/system untill this is sorted and programmed into the new page/systems.
Photo of Tim Roche

Tim Roche, Official Rep

  • 337 Posts
  • 44 Reply Likes
Folks, this isn't a conspiracy, we're bringing decimals back. 
We haven't released it yet because we're doing thorough testing of the newer API, we've discovered a particularly tricky issue and are working to get it resolved.  Testing takes time. 
Please be civil to one another, one person has already been banned from this board for their comments towards other users, others are close.  
Photo of Ul KE

Ul KE

  • 13 Posts
  • 18 Reply Likes
Many thanks for that info - hope it will change back to normal soon :-)
Photo of Roger Lewis

Roger Lewis

  • 14 Posts
  • 4 Reply Likes
Speaking on behalf of ALL that is waiting for the decibel format to be reimplemented this is good news. however, please excuse MY skepticism....BUT....it probably will not be any time soon perhaps a year or 2 down the road..... OR longer.
(Edited)
Photo of Doc_Bill

Doc_Bill

  • 3 Posts
  • 2 Reply Likes
I dont have any problem showing decimals of you feel better about it or if it looks better. But like Paul said... its all statistical randomness in the decimal places/ not real.
Photo of Tim Roche

Tim Roche, Official Rep

  • 337 Posts
  • 44 Reply Likes
The folks who view the data in Celcius do have a point, 1 degree is a pretty big jump, in F I agree its pretty noisy
Photo of Ul KE

Ul KE

  • 13 Posts
  • 18 Reply Likes
Yes, as I said before, between a 5,6 and 6,4 it ́s a difference of 1°C. Rounded you have only the 6 °C in both cases. When you analyse data in that way, a lot of information is going elsewhere. 
(Edited)
Photo of Andrea Giglietti

Andrea Giglietti

  • 3 Posts
  • 0 Reply Likes
Another big problem is with 5.4°C and 6.6°C. The difference is about 1°C, but for WU that shows 5°C and 7°C are 2°C of difference.
Photo of Paul Grace

Paul Grace

  • 23 Posts
  • 5 Reply Likes
You're misquoting that error figure, Andrea. 
It's not 5.4 and 6.6, it's 5.6 and 6.4.  Both round to 6, and represent all the accuracy your system is giving you anyway.
Photo of Roger Lewis

Roger Lewis

  • 14 Posts
  • 4 Reply Likes
Paul, Andrea's comment was not referencing Ul KE's comment. She was adding another example using the words: "Another big problem is with"...  5.4°C and 6.6°C. which is ACTUALLY a difference of 1.2°C

5.4°C  is rounded down to the whole degree of 5°C whereas 6.6°C is rounded up to the whole degree of 7°C creating a difference on WU of  2°C which actually makes the reading off by 0.8°C  (almost a FULL degree)

Using the same numbers in a different scenario and rounding to the nearest whole number value...  If you loaned me $5.40 -or whatever your currency is- and I only paid you back $5.00 ( rounding down ) Would you be ok with losing .40 cents? and.. if you bought an item for $6.60 -in your currency- but were charged a flat $7.00..( rounded up) would you be ok with paying MORE than the item actually cost?
Photo of Roger Lewis

Roger Lewis

  • 14 Posts
  • 4 Reply Likes
... and IF we are going to use the WHOLE number value format... why NOT use them CORRECTLY... as in this example, 7.0° to 7.9° would read as 7° instead of SPLITTING at. .5°  making 7.0° to 7.4° rounded down to 7°... and 7.5° to 7.9° rounded up to 8°... ANYvalue within 7.0 to 7.9 is STILL 7°  that's why the decibel format is SO important. 
(Edited)
Photo of Paul Grace

Paul Grace

  • 23 Posts
  • 5 Reply Likes
It's amazing to read this thread.  Yes, I'm fine rounding things to the nearest dollar.  Half the time I gain $0.5, half the time I lose $0.5.  This is noise, and doesn't matter.  My taxes are paid every year to the nearest dollar, all my deductions are rounded to the nearest dollar.

My point is that reporting false resolution doesn't improve accuracy.  Just because you see the decimals doesn't mean they are right.  Commercial themometers just are not correct to 10ths of degrees, so there's not much sense it getting so agitated about not reporting them.  They are noise, not signal.  Recording the noise is not really "SO" important.

Accuracy
Precision
Resolution

These all mean different things.  If anyone is concerned about these topics, they should spend some time researching them. 
Since we're all interested in Accuracy, we should all understand what it is, how to mitigate sources of error, and how much resolution is useful, considering the available resolution and accuracy limitations.

"5.4°C and 6.6°C. which is ACTUALLY a difference of 1.2°C"  

No it isn't.  Reported temperatures are not necessarily the "ACTUAL" temperatures.  We don't know what the actual differece is between reported temps of 5.4°C and 6.6°C.  That's the real problem.  Saying it is "actually" 1.2 degrees is a mistake.
"It's a degree or two" is about all you can honestly say.
(Edited)
Photo of Roger Lewis

Roger Lewis

  • 14 Posts
  • 4 Reply Likes
using the "resolution" of a whole number value format is less "accurate" than the decibel system provided the "precision" is calibrated correctly. 

At any rate... those of us who want to use the decibel system value SHOULD have that as an option on WU... just as it is used on MOST of the other personal weather station websites ie: PWSweather.com, APRS/CWOP.com, WeatherBug.com, Ambientweather.net, and WeatherLink.com just to name a few....and just as it's an option on MOST weather PWS consoles.

Those who think the decibel format is useless are more than welcome to use the whole number (round down or up ) format.. if they chose.
(Edited)
Photo of Roger Lewis

Roger Lewis

  • 14 Posts
  • 4 Reply Likes
I do calibrate my PWS ISS sensor once every 3 months with a CERTIFIED NIST calibrated digital thermometer with a 0.1° resolution both places in room with no airflow for 24 hrs. which takes my weather station offline for the calibration process period.
(Edited)
Photo of William J Schmidt

William J Schmidt

  • 31 Posts
  • 14 Reply Likes
Holy crap... the BS in this thread.  ISO80000-2 defines the rules for rounding.  0.1-0.499 round down.  0.500-0.999 round up.  Bankers rounding.  Also.  The engineering firm that I OWN, as part of our services, witness the calibration of devices for custody transfer.  Temperature devices in particular.  While your NIST device MAY have a RESOLUTION of 0.01 degrees (or ever 0.00000001 degrees), it's accuracy, unless it is a platinum thermometer costing thousands of dollars is not more accurate than 0.1F.  It just is not possible.
Photo of Paul Grace

Paul Grace

  • 23 Posts
  • 5 Reply Likes
And Roger, it's obviously not a typo, so it's "decimal", not "decibel", which is a  logarithmic scale and not used in temperature measurements in commercial weather stations.
(Edited)
Photo of Roger Lewis

Roger Lewis

  • 14 Posts
  • 4 Reply Likes
William, Perhaps with your knowledge of Banking rounding you might explain the logic behind it as applied to measuring ambient temperature. How is it assumed that 1.4 remains as "1" yet 1.5 becomes "2" rather than accepting their original values and maintaining both as "1" when using the whole number format.? What is the logic in rounding down and up? Why is maintaining the value of 1.0 -1.9  as "1", 2.0 - 2.9 as "2", 3.0 - 3.9 as "3" and so on... NOT the acceptable standard practice?

As to the accuracy of the NIST digital thermometer I use to calibrate my sensor. I can not speak to the accuracy of this particular device and the type of metals used in its construction however, you seem to be a bit skeptical. Do you not trust certified NIST thermometers? I do not dive into the particulars of it as you seem to do. I am assuming the accuracy of it based on the NIST standards  Is it truly 100% accurate? Who REALLY knows... BUT.... I'm satisfied.
Thank you for your input.  : )

Oh, and BTW, I also calibrate my watch and ALL clocks in my home twice a month (on Sundays) to avoid the common drift using the official NIST US time at www.time.gov  for calibration.  Are you going to tell me that is not an accurate means of timekeeping?  lol


Paul, Thank you for pointing out my error. Have a Nice Day.  : )
(Edited)
Photo of Paul Grace

Paul Grace

  • 23 Posts
  • 5 Reply Likes
Roger, rounding is used wherever it is appropriate to discard useless information, to make the useful information easier to work with.  Rounding gets people closer than truncating, because errors are often ± and we don't know which way they are wrong from the reported value.  Like a thermometer, for example.

As to your "NIST digital thermometer" - What model do you have?  When was it last certified?  You must be aware that the device needs to be regularly recalibrated, that just because it was calibrated a year ago doesn't mean it's still accurate, right?

Looking at the specifications of the "Sper 800043 Ultimate Thermometer", an NIST tracable digital themometer, it lists the accuracy with its calibrated PT100 thermocouple as (copying from my manual) "± 0.2% + 0.5°C or 1°F"

That means for an outdoor say, 80F reading, there is a 0.16°F + 1°F = ±1.16°F degree error value, which is a 2.32°F spread.
That's for a top-of-the-line NIST certified, digital themometer, fresh out of the box, in perfect working order, before you drop it in your radiation shield and place it near your house, or a tree, or  fence, or near your driveway.

Please don't worry about the 10ths.  Your NIST thermometer fresh from the calibrator is not that accurate, and your weather station is further off.  If you're within a few degrees, you're doing well.
(Edited)
Photo of Paul Grace

Paul Grace

  • 23 Posts
  • 5 Reply Likes
Roger, don't get me started on time.  I run two Stratum-1 PTP time servers in my home, based on oven-controlled oscillators calbrated by GPS in real time and guaranteed to run within a 3 microsecond error (3 millionths of a second) as long as it maintains a 3D GPS fix (so it can calculate the propagation and relativity corrections).  It is guaranted to maintain a 9 microsecond error figure for up to 30 hours after the time of last GPS fix, whereupon it drops out of the time server stack and sends me a failure email (which it has never done)  I am considering a rubidium clock, but they are very expensive.
I don't bother with NTP because without timestamps, you can't know when your NTP packets were sent, so you have unknowable errors in your clocks.

Your time server (based on time.nist.gov) is about 4 orders of magnitude worse than my time server.  You have unknowable latency errors that mean you are lucky if your time is accurate to 100 milliseconds or so, and windows is not a real-time operating system (unlike my time servers).  My time server stack is the same type used by Wall Street traders and banks.  If my first server drops out, and my second server loses GPS lock, I'll still have a time error < 250ms after a year without any lock.

I also wear a Seiko Astron 8X53 OAA0-2 GPS Solar wristwatch (mine is #2022 out of 3000) that syncs itself to the GPS constellation every day.

I'm a data geek.
(Edited)
Photo of Roger Lewis

Roger Lewis

  • 14 Posts
  • 4 Reply Likes
Paul, thanx for your quick reply. 

I am using the Fluke 1523-156  for my calibrations with the 1523-P1 which is NIST certified. I actually send it back to the manufacturer once a year for re-calibration which usually takes about 4 weeks to turn around. I find this to be quite acceptable for my needs. and once again as I mentioned in above comment.. it may NOT be 100% TRUELY accurate  BUT... it DOES proform to MY satisfaction as a means of calibrating my ISS sensor once every 3 months You may beg to differ.. but at this point, it matters not.

As to the Time calibration I use the NIST US time as a means of calibration due to the fact that is what the servers at my job are calibrated with, thus keeping my home clocks and work servers in sync. again you may beg to differ with the TRUE accuracy of the www.time.gov. which is certainly your right...BUT.... it matters not. 

 As far as you being a " Data Geek" I respect that and would love to sit down with you over a few drinks and have a GOOD conversation in that regard..
as I 'm sure I could learn from your knowledge.  Again, Thank you for your input.
(Edited)
Photo of William J Schmidt

William J Schmidt

  • 31 Posts
  • 14 Reply Likes
With respect to the ISO80000-2 rounding rules.... It's pretty obvious how it works... it you are less than half way to the next decade digit, your round down.  If you are half way or more to the next decade, you round up.  I can put that section of ISO80000 on my FTP site if people are interested.


The device we use for testing customers devices is the Fluke 1523-P2 with a Platinum Resistance probe.  The whole bundle cost my company about $8K with annual calibration each (I think we have four now).  I have a friend that owns the calibration lab so get a bit of a break.  The device is certified to +/1- 0.1F accuracy even though the fluke reads to three digits of precision past the decimal place.  This does not matter because in my world, better resolution would not change the answer.  Those extra digits are good for repeating and relative readings only.

Now... as I've stated before... I don't have any problem at all showing decimal places on weather stations.  In fact, the rules for averaging in mathematics say that you average at the highest resolution available and then round the resultant to the number of decimals that matches the precision error of the device.  Extra decimal places makes for pretty graphs and other aesthetics.  HOWEVER don't get pounding the table that those digits actually mean something.  They don't.
Photo of Roger Lewis

Roger Lewis

  • 14 Posts
  • 4 Reply Likes
William, Thanx for your reply... I guess I am a "stickler' for the decimal format... as I'm not a fan of the "round down / up" standard of measurement. be it accurate or not.

 I use the  Fluke 1523-156  with the 1523-P1 which is NIST certified. for my calibration process and yes, it IS a bit pricy.. cost me a little over 3K not to mention the annual cost of re-calibration.  I also use it to calibrate my HVAC system and indoor temperature sensors which I have in every room of my home.
Photo of Paul Grace

Paul Grace

  • 23 Posts
  • 5 Reply Likes
What sensor do you use with your 1523?  It supports many types for many temp ranges.

Yes, I'm sure your devices meet your needs, because that is an aesthetic decision that you get to make for yourself.  That isn't about accuracy, that is only about aesthetics.

"You may beg to differ"
It's not my opinion, it's metrology, it's science.  I am pasting the accuracy of the Davis sensor directly from Davis' sheets.  The point is your ISS is not reliably reporting the actual temperature to within even a degree of the actual value, because the sensor and application are **not capable** of that level of accuracy even if it is **perfectly calibrated**.  It is reporting the value to 0.1°F with some random noise of  ±0.5°F on top of that, before you even get to the siting issues.
You can't know the actual temp with any accuracy beyond the range of a whole degree and probably worse.  More like 3-4 F degrees if it is outdoors, if the sun is shinging, clouds, wind, the yard was watered recently...and quite possibly even worse if the siting is bad.  Worrying about the 10ths (and what IBM does with them) is hard to understand when the whole degrees are possibly wrong.  People are getting agitated about the aesthetics, not the data, but think they are missing data.  Some people don't even understand why rounding numbers works, and why the NIST requires rounding off numbers.

The 10ths are not data.  They are noise.  The sensor does not have the basic accuracy to measure temperature to ±0.3°C, much less ±0.1°F , regardless of what is printed on the display.  It's false accuracy - not data.

FYI, "NIST certifiable" doesn't mean much in this discussion.  "±0.5°F" is an NIST certifiable statistic.  My refridgerator and freezer thermometers are NIST certifiable, all it means is they perform within a given error metric.  You just shouldn't think you have 0.1°F accuracy from a ±0.5°F device.
(Edited)
Photo of William J Schmidt

William J Schmidt

  • 31 Posts
  • 14 Reply Likes
We have several ranges of probes... thermocouples, and RTDs in our shop... and take whatever we need for the job.  It's important that you select a sensor such that the temperature range you want to measure in in the center of the probes range.  That's where the best linearity is.  

BTW... NIST stopped calibrating consumer devices about three or four years ago.  Calibration shops now are the only people to use NIST tracible devices as the calibration reference.  It just got too expensive to continue doing it.
Photo of William J Schmidt

William J Schmidt

  • 31 Posts
  • 14 Reply Likes
Update.  While looking through my ISO documents... I realized that rounding is actually now in Annex B of ISO80000-1 published 2009 (for the first time).  I'm in the process of up-loading Annex B to my FTP site.
Photo of Roger Lewis

Roger Lewis

  • 14 Posts
  • 4 Reply Likes
Paul, IF I understand you correctly.. you are saying there is ABSOLUTELY NO PWS on the market that is CAPABLE of measuring the ACTUAL (TRUE)  ambient outdoor temperature at any given time. EVEN the AWOS stations? and the "official" reading from the AWOS at the Airports?  which I might add..I am less than 1/8 of a nautical mile from... and at ANY GIVEN TIME my station is within ± 0.5° or 1° ( whole number value ) of that station's readings...Which would suggest even the AWOS stations are INACCURATE... Is THIS what you are suggesting if so... that is as a BOLD statement!!
Photo of Roger Lewis

Roger Lewis

  • 14 Posts
  • 4 Reply Likes
William, I am aware of NIST no longer performing the calibration process.. this is why I have been sending my unit to the manufacturer for calibration.
Photo of Paul Grace

Paul Grace

  • 23 Posts
  • 5 Reply Likes
Roger, no, you don't understand me correctly.  I am referring to consumer weather stations, not all possible temperature measuring instruments.

All you have to do is read the specifications of the equipment you are using.  You will find they have about 1°F error figures or worse, depending on various errors in siting and conditional non-linearity (meaning you can't offset for them)

"at ANY GIVEN TIME my station is within ± 0.5° or 1° ( whole number value ) of that station's readings"

I simply don't believe that.  Even 760 feet away the actual temperature would not always be within 1 degree, so it points to both of you having severe errors, like stuck sensors (which I also doubt).  AWOS provides a 5 minute moving-average temp, once a minute, sited near the touchdown area of a primary runway..  Is that what you are comparing your instantaneous temps to?   Is your station sited on a runway apron?  Are you using the same radiation shield?  AWOS uses a platinum wire RTD, which your ISS does not.  

Finally, the spec for the AWOS temperture, here I am pasting from the ASOS guide:

Parameter         Range              RMS Error       Max Error       Resolution
 Ambient     -58°F to +122°F         0.9°F               ±1.8°F              0.1°F
   Temp

(FYI the errors double outsdide of that range)

The average (RMS) error is 0.9°F  Worst case is ±1.8°F, and (this bit made me laugh) "These 5-minute averages are rounded to the nearest degree Fahrenheit"

And you think you are getting 0.1°F accuracy?  The airport doesn't get close, and neither do you.

When you make wild claims like this it makes me wonder what you are trying to accomplish.  What exactly are you comparing to?

I'm glad you're happy with your readings, but something is going seriously wrong.

Here's the NOAA spec for you to read yourself, straight from the guys who wrote the book on the subject...They don't bother with 10ths either.  See page 12
https://www.weather.gov/media/asos/aum-toc.pdf

Is that bold enough for you? 

Cheers, friend!
(Edited)
Photo of William J Schmidt

William J Schmidt

  • 31 Posts
  • 14 Reply Likes
Hey as a cheerful side note...

Once in a while I will look at the WUG map that shows the PWS's around me and will just take note of the difference in readings (like temperature, wind speed, direction, heading, etc.).  I notice that for the most part, my station is at the median (reports the same number as the majority of PWSs).  That's about all I care about.  It's really irrelevant if we all report the exact right number.  I just don't want to see my station reporting a number that is 20% higher or lower than my neighbors.  The collective is what is important for forecasting.  The more (good) data points the better.
Photo of Roger Lewis

Roger Lewis

  • 14 Posts
  • 4 Reply Likes
Ok, Paul, you got me! I must admit  OVER EXAGGERATING!...lol... not "at ANY given time"... BUT... there are many times when I call the automated ASOS minute observations number of this particular ASOS station which is updated every minute I will have to convert the readings from C to F for the ambient temp. then compare to my readings for that specific time reference on my data logger which is set for1-minute intervals and it's fairly close... for the most part... but NOT ± 0.5° at any given time... I also look at the 5-minute tabular display of this ASOS station that is uploaded to MesoWest.utah.edu and compare with my station's 5-minute tabular readings.. again, fairly close under NORMAL weather conditions... but NOT "at any given time".  thanx for calling me out on this...lol 

 I used to have a subscription to anyawos.com to retrieve the ASOS 1-minute observations either by a phone call or online. but now I just periodically call for the 1-minute obs..

I appreciate you providing the info on the ASOS users guide.  I actually have on file...and have read it before.. always good to refresh my memory.  THANX.
(Edited)
Photo of Paul Grace

Paul Grace

  • 23 Posts
  • 5 Reply Likes
Yes, I suspected you were exaggerating.

"which is updated every minute" - don't forget, ASOS "1-minute observations" are moving 5-minute averages.  The ASOS specification is to not publish instantaneous observations, so you are comparing AWOS average temps to your instantaneous temps.  Not exactly apples and oranges, but let's say grapefruit and oranges...  And they are rounded to the nearest whole degree F, so you can't even tell if you are within 0.5F.  Add in the error from your preference for truncating decimals instead of rounding them, and you can be sure that you are off by at least 0.5F about half the time.

You also have very different siting particulars, so you'll get more delta from that.

You are also going to be way off whenever the temps are changing rapidly (like whenever a fast front moves in) because your temps will lead the airport averages.

When all is said and done, I'm sure you're not within 0.5F from the airport temps.  It just isn't possible when we consider the facts that everything that can be different, is different, between your observations and the airport observations 760 feet away.

Don't sweat it though.  It seems like you are running a well-calibrated PWS, and the tenths don't matter.

Cheers!

Photo of juan parra

juan parra

  • 2 Posts
  • 1 Reply Like
Por favor devolver los decimales a la temperatura. Es una lástima que se actualice la página y en el momento de mejorar lo que se hace es empeorar. Creo que no está tratando bien con los usuarios y no creo que sea muy costoso el añadir los decimales. Son millones de estaciones y con ello millones de datos que se utilizan sin que los usuarios reciban nada a cambio, al menos añadir los decimales
Photo of Doc_Bill

Doc_Bill

  • 3 Posts
  • 2 Reply Likes
Lo es más eficaz dirigirse cuestiones en un idioma eso no es nativo al anfitrión?