Temperatures for SE Qld Australia
I have been comparing the temperatures of areas of the SE of Qld including checking my own temperatures manually and the coldest temps are always 4 to 6 degrees too hot for what it actually is, the hottest temps for the day are always 2 degrees to cool. What is up with the inconsistencies? Miles shows it is between 12 to 10 degrees when it is actually 8. Tomorrow it shows it is 7 to 5 degrees when it will actually be 1. This has been constant over the last month and has been for several areas such as Brisbane, Warwick, Wondai, Toowoomba and Caboolture. If I have to check 3 other free weather pages to verify what you have shown whats the use on paying premium?
Windy visualises weather models and is not responsible for their discrepancies with observed temperatures.
Have you used the Observation vs Forecast tool to see which is the best model at predicting daily minimum temps?
Here is the comparison between ECMWF model and the weather station at Miles (Reported temperature)
Seems that ICON and METEOBLUE show less inconsistencies.
@idefix37 Yes I have compared and as of tonight their predictions are wrong and will be wrong for the entire week. Where are they pulling their data from? As i stated before the temperature will be 4 to 6 degrees lower then what it truly is.
You may claim no responsibility for their incorrect data however you are responsible for finding sources that can provide the correct data right?
Here is a source that has proven to be correct 95% (error within 1 degree maybe once a week) of the time compared to windys 5% (and im being generous with that percentage) https://www.weatherzone.com.au/qld/darling-downs/miles
If you are going to charge people for data you need to make sure it is correct first.
vsinceac last edited by vsinceac
Windy displays the best weather forecast models available in the world, provided by WMO centers (NCEP/NOAA, DWD, ECMWF, MeteoFrance, etc. - and maybe one day BOM, too), which use the best technology currently available.
One cannot denigrate these models (excepting if one denigrates science). They are all different, as computed by different algorithms (improved continuously) and approaches, which is an asset in itself, as weather forecasting is not and will never be an exact science.
Btw., when comparing forecast (e.g. ICON) to WMO surface station (e.g. ID 94576, BRISBANE), one can see that computed forecast temperatures match pretty well the measured values (see temperature compared timeseries on screenshot).
Don't forget that forecast models are not computed for a given point in space; they are computed over slices of atmosphere covering big areas (e.g. ~9 km² for ICON, ~27 km² for GFS). Point forecast values are just interpolated from these large area values and they could never compete with real-time values measured by sensors of WMO surface stations, as they cannot cannot get rid of very local influences (large concrete surfaces in cities, deep valleys, etc.).
Over an area of 25 km², if you would have, say, 1 WMO station every square km, these 625 stations will surely measure very different temperatures, too. I guess the mean of the 625 measured values would match pretty well the computed value provided by GFS model for that 25 km² grid cell...
If you would need more precise forecast for your very local area, you should ask your local meteorological service affiliated to WMO, which generally provides more pertinent local forecast than global models; this is the role of meteorologists and forecasters, and still requires human brain and local knowledge in addition to algorithms. Last but not least, even local forecasters and meteorolgists use the same forecast models to generate local forecasts (and even some of them begin using Windy :o)
About Weatherzone web site cited above: this site is not free (6$/month or 75% of the web page covered by stupid ads). It was founded by BOM meteorologists and gets priviledged data from BOM. It is maybe good for local Oz weather forecast, but if one asks forecast for e.g. Paris, France, this site shows very different values compared to local forecast provided by Meteo France.
In addition, some of the more accurate weather forecast models provided by Windy are not that free (e.g. ECMWF and NEMS).
@vsinceac To start with, your screenshot is a useless example as windy isnt showing the correct temperatures which is my primary claim, to make your point valid you would need to show me the data of that point from another source (my main focus is on the lowest temp as the inaccuracy is greater).
If these are the ''best weather forecast models available in the world'' then how come they can constantly get such a thing wrong like temperature? You would think they would have fixed their models by now unless its intentional.
As for being ''computed over slices of atmosphere covering big areas'' this is why I gave several examples over a large area of country including mostly country areas where there is little to know concrete. Most temperatures in the regional areas are taken at airports where there is little outside influence.
Its common sense in my books if your model don't accurately reflect what is happening in something easy as imputing new data from temps to fix your model then how is the rest of the model trustworthy? For farmers where I am close enough is not going to water the crops.
My aim for pointing this out was so someone could hopefully look into the problem but it just looks like everyone wants to excuse it away. So be it.
vsinceac last edited by vsinceac
A closer look to the 2D plot above shows 2 time series for temperatures; as explained above:
- one is for ICON model forecasted values (each 9km²), interpolated for the position of WMO Station ID 94576 (blue dashed line)
- the other (color shading curve) is for values measured by WMO Station ID 94576
For a point on Earth, there are no other trustfull real time temperatures than measured by a WMO surface weather station (here, BRISBANE). As one can see, the two time series are pretty close each other; details for a sample point in time selected on right side show same value on both curves.
Imho., computing temperatures for the next couple of days with such an accuracy, is a great achievement of the science and technology; a couple of dozen of years ago, this was simply impossible.
Btw., ICAO stations on airports generally provide METAR messages, where measured temperatures are conventionally rounded to integer value; thus they are already affected by errors up to +/-0.5°C.
ICON is not "my model", it is provided by DWD; behind such a model are dozens of years of scientific research; computing a global model requires the most powerfull computers in the world. Thus I apologize, I couldn't be able to look closer "into the problem", nor to "fix my model" myself (and I'm sure Windy also couldn't).
But if one would think these models are all so wrong, then one should explain this to the guys who compute them (and provide them for free), and suggest them some good ideas (this should be possible on the web sites of NCEP/NOAA, DWD, ECMWF); I bet they will highly appreciate such constructive help... ("unless its intentional" and they would devilishly introduce some delta values for temperatures just for kidding with Oz people :o))).
Otherwise, when people using weather forecast models want to get the better one for their very local position, generally they try to compare each of them to measures from closest WMO weather stations, then use the one with best match. Generally people don't denigrate these models, people use them as they are; weather forecas modelling never claimed to be an exact science, as you think.
Actually I personally think that the best weather forecast model for Oz should be local models provided by BOM met office; unfortunately (as a developer said two years ago here), Windy cannot provide these models as they are not free.
Anyway... Nowadays everyone can buy a weather station from Amazon for two pence, thus no need for all those wrong, boring scientific models, isn't it? Finally, every aifone and every aiwatch can measure temperature, atmospheric pressure, humidity, heartbeats and even holy koronav presence, can send e-mails and tweets; why spending public money in scientific research - which gets so wrong results - while gadgets vendors take care of the whole stuff so well?