Humidex - Discuss!

[mcrossley]

I am looking at how Cumulus uses Humidex. At present it is only calculated for the current time, and is not recorded anywhere.
Cumulus also treats Humidex as a temperature, but according to the Canadian meto (who instigated it use) it is a dimensionless index number.

So some questions to start the discussion...

1. If MX treated Humidex as an index, it should stop using the C-->F conversion on it for users who use Fahrenheit?
   (Canada used to use F, but the Humidex scale used then is not the same as the current scale converted using the normal C-->F - they just took the "Fahrenheit" version and multiplied it by 5/9)
   (The "comfort" tables used to interpret Humidex all appear to use the "Celsius-fied" Humidex values)
2. However humidex looks like a temperature, and it appears it is commonly abused as a "feels like" temperature. So we treat it as a temperature?
3. Like Heat Index, Humidex is a scale index used for high temperatures, indeed the formula used breaks down below 7C. So I propose that below 10C that MX stops calculating Humidex and just returns the current temperature (in Celsius? - see Q1!)
4. Like Heat Index, low value records etc are fairly meaningless, we are interested in highest values only?
   (For heat index MX does not produce low records)

Your views please.

[HansR]

I appreciate this discussion and although we already exchanged some views I'll summarise mine here again:

1. It may have originated in Fahrenheit and later 'Celsi-fied', but the calculation says:

   Code: Select all

   humidex = (air temperature in Celsius) + h

   which in my view makes it a raised temperature, indeed kind of Feels Like. the h being a dimensionless part added to the Temperature. But that makes it a temperature because the dimension won't change with that addition. But that would also mean that, if a user uses Fahrenheit iso Celsius, the addition needs to be converted otherwise you get an unexplainable difference.
2. I say yes, see my (1)
3. Agreed
4. Agreed. Depends also a bit on the amount of work involved I guess. And what about the chill factor which has no meaning at high temperatures? Does that have a high record? Maybe just let it be. If its useless, the website does not have to display it.

[mcrossley]

It may have originated in Fahrenheit and later 'Celsi-fied', but the calculation says:

Code: Select all

humidex = (air temperature in Celsius) + h

which in my view makes it a raised temperature, indeed kind of Feels Like. the h being a dimensionless part added to the Temperature. But that makes it a temperature because the dimension won't change with that addition. But that would also mean that, if a user uses Fahrenheit iso Celsius, the addition needs to be converted otherwise you get an unexplainable difference.

Well "h" is not dimensionless - it involves vapour pressure in Pascal and 1/Temperature. But the whole formula adds dimensioned and none dimensioned values so all bets on the normal rules are off!

Depends also a bit on the amount of work involved I guess. And what about the chill factor which has no meaning at high temperatures? Does that have a high record?

Wind chill has a maximum cut-off value of temperature (10C)* above which the function just returns the input temperature. There is no maximum record for it.

* though this can be overridden.

[HansR]

[...] so all bets on the normal rules are off!

Yes, any dimension analysis breaks. Also because it is unknown how those constants in the equation are formed. There are quite some assumptions in there.

Wind chill has a maximum cut-off [...] There is no maximum record for it.

So chill, heat index and humidex are basically treated the same: some cutt off value, no record on the high/low side of the application region. While temperature, dew-point, feels like and apparent work in the full temperature range.

Nice explanatory work for the Wiki for that whole temperature graph and its derivatives

[BCJKiwi]

Take your Pick!!
Environment Canada refer to Humidex as an INDEX with no units.
It seems everyone else treats it as a temperature.

The numbers it creates are on a scale that seems sort of related to heat index.

See the table below where
Humidex reaches 56 with temp at 38 C and RH at 65% OR 53 with temp at 34 C and RH at 85%
This relates to
Heat index which hits 58 C with temp at 38 C and RH at 65% OR 57 C with temp at 34 C and RH at 85%

Right now it is not hot enough here for the Humidex to register ( less than 19 deg C ) BUT
Humidex 19.7
Temp 17.4
App temp 16.2
Feels like 16.5
Heat index 17.4
Wind chill 17.4

So Humidex seems to start a little lower than Heat Index and stays that way (if you consider both to be temperatures).

Humidex.PNG

[HansR]

A link would be nice, so I searched, not sure it is where you got your info from, but I read this one (and look here for the Wind Chill factor). It says:

The humidex combines the temperature and humidity into one number to reflect the perceived temperature.

A perceived temperature, as far as I see it, has a dimension of a temperature. No matter what anybody else says. See also my argument above: it is the current temperature with some addition.

On top of that is the question : how should CMX (and the display layers on top of it) deal with it ? How formal does it need to be with something which is neither measurement nor scientific formal derivative.

1. If we just look ate the text display it is quite easy to display or not the used unit;
2. If the user uses Fahrenheit, then if it is truly dimensionless... how to interpret the value relative to this other scale? As I said before, I think it should be converted. But that means it is interpreted as degrees
3. The mouseover display in the graphs shows degrees Celsius or Fahrenheit as unit. It will be more than troublesome to split those formats. Agree, not a valid or formal argument, but why make life difficult if everybody but the Canadian government (Environment & Climate Change) treats it like a temperature?
4. And if CMX wants to treat this type of derived temperatures - I have no other word for them - it has to remove the degrees units for the HeatIndex and the WindChill factor as well. Also using a low cutoff for the Heat and Humidity indexes and a high cutoff for the WindChill.

My guess is that abiding to the Canadian government (dimensionless usage) in this case is not the best thing to do.
But, having made my point, I abandon the discussion. I can live with any choice.

Further reading: here and here, Interesting because it says:

Environment Canada’s Humidex Comfort Ratings (Celsius)

[mcrossley]

Interesting because it says:

Environment Canada’s Humidex Comfort Ratings (Celsius)

Do you not think that is because the lookup temperatures are in Celsius?

Anyway, anyone else have any opinions on this? I expected more input tbh, it's your chance to influence how this is implemented.

[HansR]

Do you not think that is because the lookup temperatures are in Celsius?

Very probable.

[sfws]

Mark, you prompted me in a PM to reply to this discussion, so I have, but my indoor "chores" are taking priority over this forum.

So some questions to start the discussion...

If MX treated Humidex as an index, it should stop using the C-->F conversion on it for users who use Fahrenheit?
(Canada used to use F, but the Humidex scale used then is not the same as the current scale converted using the normal C-->F - they just took the "Fahrenheit" version and multiplied it by 5/9)
(The "comfort" tables used to interpret Humidex all appear to use the "Celsius-fied" Humidex values)
However humidex looks like a temperature, and it appears it is commonly abused as a "feels like" temperature. So we treat it as a temperature?
Like Heat Index, Humidex is a scale index used for high temperatures, indeed the formula used breaks down below 7C. So I propose that below 10C that MX stops calculating Humidex and just returns the current temperature (in Celsius? - see Q1!)
Like Heat Index, low value records etc are fairly meaningless, we are interested in highest values only?
(For heat index MX does not produce low records)

Your views please.

1. Canadian Humidity Index (Humidex) is an index, you don't quote units. Please see viewtopic.php?f=14&t=150&p=1304&hilit=Humidex#p1304, for original discussion on this, it even mentions the Fahrenheit question where Steve Loft actually says:

Yes - the idea is that it's just an index, unlike the other data, and as it was invented in Canada it uses degrees C to calculate it. It would be wrong to quote it with a degrees C unit, and even more wrong to provide a Fahrenheit equivalent.

2. The only people still using Fahrenheit are in USA, they do use Humidex although they have their own USA Heat Index which also does not have units. FWIW, if someone wants Humidex in oF, then omit "5/9" multiplier, don't apply any 32 offset; i.e. go back to original Canadian formula. That is what my script shared at viewtopic.php?f=18&t=18096&p=142634#p142634 does, and what my aborted log editor did.

3a. As for USA Heat Index, that is only defined above 80 oFahrenheit (approx 27 oC), so it hardly ever applies in UK, I actually coded it to appear above 25 oC on my web pages written a decade ago, as otherwise I never saw it. Just to make that clear, below my cut-off I treat Heat Index as null, i.e. not defined. I can't understand why Cumulus would choose to return air temperature for a dimensionless index; it is a computer programmer's aversion to null, that Steve Loft seemed to apply, just like Cumulus outputs valid values even when a sensor does not exist. I'm sure a meteorologist would prefer null (i.e. report not available) when a statistic is not valid. It puzzles me, as Steve Loft reports "--" for monthly values on first day of month in some web tags.

3b. As a side issue, nobody ever requested Heat Index be added to Cumulus. Given how small a proportion of Cumulus users are in USA (and as far as I know only the USA bother to report Heat Index), it remains a mystery to me why [when Steve Loft added the WX gauges to Cumulus in 1.7.7 (21 Jun 2008)], he also added Heat Index. I suspect the only reason that Heat Index got added to Cumulus 1 was that Steve based a lot of his implementation on weatherlink, which is USA based. Anyway, apparently Steve Loft thought it useful enough that he revised the formula in 1.9.0 (11 Sept 2010) and implemented it as monthly, yearly and all-time derivatives.

4a. In my web pages, I don't display Wind Chill above 10 oCelsius, I don't think replacing it by the air temperature is valid when Wind Chill is not defined. Similarly, if there is no wind (or wind less than 3 mph), there is no wind chill. I always thought it strange that UseWindChillCutOff=1 did not actually cut off the output of wind chill outside its defined range; I can only assume it was because Steve Loft had an aversion to null, as I said for Heat Index. I have to admit, my original JavaScript code on my early web pages actually checked air temperature and wind speed to make the decision whether to include that row in the table, but now my web pages use PHP Hypertext Preprocessor (PHP) script, I simply hide wind chill if it is equal to air temperature making use of the default implemented by Steve Loft.

4b. At this point it is worth noting that Cumulus has only reported the greatest wind chill for daily and longer periods (i.e. highest wind and lowest temperature). That reflects that the least wind chill is obviously when it stops being defined at low wind speeds and high temperatures, so not reporting least wind chill is an extension of not bothering to report lowest wind speed.

5 So wind chill is for cold windy days. What about muggy days? Well my web pages report Absolute Humidity (related to dew point) rather than relative humidity because I want to record how muggy it feels, and Cumulus 1 (and MX) outputs monthly extremes for dew point, so I have a good starting point. But as described later, I also include the Canadian Humidity Index, although Cumulus, so far, makes that hard work.

6a. Because of Heat Index being so rarely quoted, and because Canada does seem to have a climate closer to UK's than the USA, I have been displaying Humidex on my web pages as it gives a better comfort index on those days when Humidity makes the air feel heavy (oppressive). Now obviously I use scripts to enable me to see humidex on my web pages. I have to add humidex to my standard log files, and calculate daily, monthly, and yearly extremes for myself (so I look forward to MX doing it for me).

6b. There was a user request for Humidex and the request came from TNT weather in USA! By the way, this is not the TNT weather channel that started in 2016, but the late Kevin Reed's user name before he died in Jan 2018. It is unclear to me why Steve Loft decided to pursue full implementation of the unrequested Heat Index (he said it was useless to him on Sanday), and not of the requested Humidex. It might interest some people to see viewtopic.php?f=14&t=195&p=1317 where a php function to describe Humidex in text is presented, a bit like Beaufort descriptions. Those with PHP web sites might want to implement that.

6c. As you say Humidex formula breaks down at low temperatures, so following my earlier points, I ignore it at low temperatures. Again, following my heat index points, I think it should be reported as null when it is out of range; a dimensionless index should not suddenly become a temperature.

7. Low extreme records. I notice that the files that were incorrectly released in 3.6.12 (build 3088) do have low humidex value and time in dayfile.txt editor, you copied your feels like temperature code. Humidex is not like feels like temperature which is defined for all temperatures. Humidex is just like wind and rain. There is a fixed lower-value cut-off, so the lowest all-time is known. Yes, sometimes there is interest in lowest humidex value. People may wonder what was the smallest rainfall rate (or smallest wind run) during a period when it rains (or is windy) every day. Put another way, the popular top ten script on your web site has Driest Months, Lowest Max Wind Gust and Min Daily Wind Run. I can imagine Lowest Humidex appearing there in the future.

To sum up:
- I welcome addition of Humidex.
- Humidex and Heat index should continue to be treated as dimensionless
- Ideally, Humidex and Heat index should go to null instead of reporting air temperature when out of range.

[mcrossley]

Thanks @sfws - some interesting background there.

I agree with nulling values when they are not available or not applicable, however MX has not been written to cope with nulls as a concept and adding it now would be both complex and problematic for historic data. I think that will have to wait for the fabled C4! The best we could do in MX would be to use some obviously invalid value as n/a such as -999

[sfws]

The best we could do in MX would be to use some obviously invalid value as n/a such as -999

My log files in my Cumulus 1 days did have some -999s for items like solar and even in some cases when I decided rogue values had crept in for temperature or whatever.

[freddie]

I was going to write something about this, but swfs has given such a good analysis that I don't have anything really to add. The UK Met Office doesn't "do" humidex, and that speaks volumes to me. It is of limited use in the UK climate, so I guess that is not entirely surprising.

[sfws]

swfs has given such a good analysis

Thanks.

It is of limited use in the UK climate,

3 quick responses to that.
1. I'm just supporting the idea that Cumulus offers people the chance to use Humidex.
2. Cumulus is not just for UK.
3. Humidex is at least valid at temperatures seen in UK in many months of year. Your Alcaston Weather web pages, like web pages for the majority of Cumulus users, display Heat Index, which is much more useless in UK.

I am guessing that relatively few Cumulus users actually care about any derived values, despite that being the main functionality Cumulus offers, they just want to see something from their weather station.
I suspect if Mark had initiated a discussion about wind chill, it would have seen as little feedback.

I am not even surprised that nobody from Canada, who invented this measure, has made any comment.

[mcrossley]

OK, given the current restrictions with MX, and to match Heat Index and Wind Chill, here's what I'm going to do...

1. Implement as an index - no units, always based on Canadian values, ie no Fahrenheit conversion
2. Implement a cutoff of 10°C, below that it returns the temperature in Celsius (same a heat index and wind chill)
3. Add to dayfile fields for the max value and time
4. Add records/alarms for max value the same as heat index

Let's see how that goes.

I was giving a bit of thought to null values, and I think that two distinct types of null would be optimum. One for "not implemented/recorded" and one for "out of range" for these derived values, and cases where sensors go out of sensible range. Some thoughts for C4!

[PaulMy]

I am not even surprised that nobody from Canada, who invented this measure, has made any comment.

I don't make it, I just use it

Humidex is generally included in EC forecasts, etc. but TV weathercasts now predominantly use Feels like https://www.canada.ca/en/environment-cl ... .html#toc7 which annoys me as my wife says " WOW 40°" and I have to correct her to say it is only 34°

Code: Select all

dateDetailed Forecast
Today	A mix of sun and cloud. 30 percent chance of showers this morning then 70 percent chance of showers or thunderstorms this afternoon. Risk of a thunderstorm this morning. Wind becoming southwest 30 km/h gusting to 50 this morning. High 32. Humidex 41.
Tonight	Partly cloudy with 30 percent chance of showers early this evening and risk of a thunderstorm. Clearing this evening. Wind southwest 20 km/h becoming light this evening. Low 20.
Mon, 20 Jul	Sunny. Becoming a mix of sun and cloud near noon. Wind becoming northwest 20 km/h near noon. High 28. Humidex 35.
Night	Cloudy. Low 17.

https://www.canada.ca/en/environment-cl ... .html#toc7

Enjoy, and try to stay cool!
Paul