Best Way to Determine Hottest/Coldest Month

[Raissermesser]

Note. No politics or climate debates, please. This strictly concerns how to look at data.

The Why:
It started with a toss-off statement by a TV weather anchor that there had been X number of consecutive days with highs above a certain amount. Except I clearly remembered days where the highs were below that.

That lead to a search for weather information during that period. Finally found it for the TV station's location ( Climate Data Online (CDO) - The National Climatic Data Center's (NCDC) Climate Data Online (CDO) provides free access to NCDC's archive of historical weather and climate data in addition to station history information. | National Climatic Data Center (NCDC) - https://www.ncdc.noaa.gov/cdo-web/ ), downloaded it; imported it into a spreadsheet, and checked. Nope, the weather anchor was wrong.

This had me looking at weather data I keep up with. That's an offshoot from keeping up with it at work. Why? Just Because. September was a hot month; most of us think it was the hottest in our lifetime. But was it?

The Problem:
Straightforward, or so I thought. I keep up with this in a database. Did a search on all Septembers. That was easy. But once I had that, things got confusing. Do I average the highs for the month? Determine the average temperature for each day and average that? Maybe something else?

As it turned out, our September was the hottest in my records regardless of which method I used. But there was some variation in the list of top years. It makes a difference in which year comes up.

Most likely it depends on what constitutes the hottest and coldest. If it's continuous heat/cold, maybe and average of a month's highs and lows. But if it's simply how hot or how cold were the temperatures in a month, maybe averaging those separately?

Comments?

 

[Mr5x5]

There isn't a perfect method. To add another wrinkle... if you are interested in the human scale you could look at heating and cooling degree days. If you are more concerned with vegetation impacts look at growing degree day accumulations for the warmest (coldest you are out of luck) but then you need to decide what base temperature you want to you.

 

[Mr. Scruffy]

Most likely it depends on what constitutes the hottest and coldest. If it's continuous heat/cold, maybe and average of a month's highs and lows. But if it's simply how hot or how cold were the temperatures in a month, maybe averaging those separately?

Comments?

Once you define what is meant by "hottest" or "coldest" it will be easier to measure. You might even need to back up a bit and define what is meant by "hot" and "cold".

 

[Raissermesser]

There isn't a perfect method. To add another wrinkle... if you are interested in the human scale you could look at heating and cooling degree days. If you are more concerned with vegetation impacts look at growing degree day accumulations for the warmest (coldest you are out of luck) but then you need to decide what base temperature you want to you.

About heating and cooling degree days:

Once tried to correlate load to heating and cooling degree days. The assumption is that AC comes on above a certain lever, and heating below another, so it should be a good correlation, or so I thought. Then I came to October in my test data. There my cunning plan fell apart. October that year had increased electricity use, but by the heating and cooling degree days, it should have been low.

Didn't take long to figure out what happened. That October, we had both cool temperatures in the morning and warm temperatures in the afternoon. That meant we had both heating and cooling in use in the same 24 hour period several times that October. The "standard" way to calculate heating and cooling degree days is to take the daily mean temperature and see if it's above or below the base levels. That year, the daily mean averaged out to be between the two.

Obviously, a better method would be to track it to actual temperature change, which roughly follows the pattern of sunrise and sunset, and calculate based on a model of hourly temperature. Yes, that's been done. It didn't work for what we needed, so I just dropped it.

What I ran into by using the daily mean temperature is why I looked at the Hottest September data in two different ways.

Starting to think that a better way of looking at hottest and coldest month would be to sum the hours above and below base levels, like the modified heating and cooling degree day equations.