[jeffchien]

Not to mention the dissolved solids as well, so it's not as simple as using tap water.

IMHO Celsius is the SI unit least deserving of praise. It's just bolted onto the system through the Boltzmann constant and isn't intricately linked like other SI/CGS units. It's also a little too coarse for modern use. Sure, I could use decicelsius myself, but that doesn't stop every other weather station, AC/thermostat manufacturer, etc. from still using whole degrees. You can buy $2000 Italian espresso machines with elaborate PID controllers that can only be tuned to whole degrees, which ironically means that using Fahrenheit is more precise!

[pocketarc]

I think there’s some confusion here - the Kelvin is the SI unit of temperature, not Celsius.

[triactual]

The isotopic content of the water matters too. Boiling point are not used to calibrate thermometers though. Freeze points and triple points are used for temperature calibration. And outside of a primary standards lab, it’s usually just a transfer standard (another, more accurate thermometer).

[eqvinox]

> doesn't stop every other weather station, AC/thermostat manufacturer, etc. from still using whole degrees

I am European and while I agree on weather stations, I have never seen an AC or thermostat that doesn't at least use 0.5°C increments. (And I'd argue a whole °C is perfectly appropriate for weather.)

[arlort]

Besides Celsius not being the SI unit.

I'd be shocked if weather forecasting is ever precise to less than 1-2 °C

[defrost]

> I'd be shocked if weather forecasting is ever precise to less than 1-2 °C

That's not a forecast .. actual forecasts for a region are often more accurate than that wrt "mean tempreture" for that region.

Where I live the BoM (Bureau of Meteorology) cannot "accurately" predict the upcoming daily tempreture and rainfall for the town that I'm in.

They do very accurately predict the mean temp. and rainfall for the wider region.

Doppler RADAR gives a very accurate measure of the amount of moisture held by rain bearing clouds, stable wrt broad area winds give a good path for that moisture's travel in the next 12-24 hours, experience + data give solid predictions for the amount of water to be shed, along with bulk cooling from shade and water fall + evaporation.

The difficulty rarely lies in the bulk parameters for a broad area, the devil is in the small details which are inherently turbulent.

To measure the accuracy of forecasting it first has to be ring fenced by what area you intend to predict the mean attributes for, the timescale you intend to project for, etc.

[arlort]

Yes, but all of that is unrelated to the original point

The commenter I replied to was complaining about weather stations not using decimals.

As you point out we can predict averages with a better accuracy, but services which have that accuracy do use decimals, where decimals aren't used is in consumer weather forecasts where an accuracy of less than 1 degree is anyway useless because

> the devil is in the small details which are inherently turbulent

[Ekaros]

Also temperature is variable. It fluctuates through the day and then things like clouds or rains can drop it quite a lot or keep it higher during winters in parts of the world.

You can get pretty good hour level forecast in some scenarios. But how do you effectively communicate those outside showing some type of graph...

[The_Colonel]

The only use case I encountered where Celsius is too coarse is measuring body temperature.