When is low the Low and high the High?

I have been in a discussion with myself for the past month about Low and High values. The question I can’t agree on is when is the low value the Low and the high value the High?

Is it the earliest time each is recorded? Is it the latest time each is recorded? Is it value between 6am and 6pm?

I can make the case for each of these periods. Each has its merits and each is flawed.

Tell me your thoughts as why you think one is preferred over the other.

@dsj @WFstaff I would like to know your thoughts on this too.

1 Like


I don’t an answer… but rather a concession. I’m leaning toward the High/Low values since midnight…


1 Like

I agree. High/low since midnight seems the best choice.

  • Jim

…but which midnight – UTC or local? :wink:

I believe that weewx reports the latest time of a peak measurement in the event of duplicate measurements at different times.

If you want to add to your headache, there was an interesting thread on WXforum about decimal rounding in high/low temperature figures. Should a low of 32.4 be rounded down to 32, or up to 33? Ditto for a high of 75.6, and should it be rounded to 75 or 76???

1 Like

I prefer High/Low since midnight local… That’s usually what the avg consumer is looking for.

As for the rounding… DON’T round… just use the decimal! :wink: That’s what I do.



In our apps the high is the absolute high and the low is the absolute low for the local day (12:00:00 am to 11:59:59 pm in the local timezone). But we don’t report the time it happened so there’s no ambiguity if there are duplicates. But I don’t think there’s any correct answer to your question - as long as you your users know what you’re doing.

This reminds me of Brian Norcross and his argument that meteorologists should consider “the art of numbers” in their reporting.


We all know it’s generally colder at night than at day. We expect the low to be at night.

Let us say the low is 0 at midnight. Then it starts to warm up and reaches 40 at 10 am. Then it drops to 0 at 2 pm.

Which time is more relevant? It can be argued that the 2 pm time is of more relevance since it is a better indicator of what is happening.

@dsj, That’s a nice article but to me, as a data analyst (where numbers actually matter), he talked himself into a circle and left where he started.

I’m still in my circle and no closer to a time that gives any meaning.


How about another thought, just to muddy the waters… daytime H/L and nighttime H/L?

For determining the local conditions, we are using basically ‘sunrise’ to ‘sunset’ for daytime values… and ‘sunset’ to ‘sunrise’ for nighttime values. Could do the same… forget the midnight to midnight concept and go with a “full day” being ‘sunrise’ (today) to ‘sunrise’ (tomorrow)… more like those TV forecasters do (geez, am I really agreeing with THEM???). And since most of us (humanity) consider “today/tonight” being from when I got up this morning until when I get up tomorrow morning…



The daytime high and low is the value, I think, that most people can relate with.

I’m not most people. I deal in real numbers with real meanings. I know night is from midnight to 6 am, morning is from 6 am to noon, afternoon is from noon to 6 pm and evening is from 6 pm to midnight.

When I mentioned 6 am to 6 pm, it was a generalization for sunup to sundown. I haven’t reached the point where I round numbers. And for that article David linked F is a lot more accurate than C. I’d much rather now it’s 72 F than 22 C. Or is it 73 F and 22 C.

1 Like

You got that right!!!

Then again… most of us here aren’t most people! :slight_smile:

A agree with:

1 Like

if they’re the low/high since midnight in localtime

1 Like

9am to 9am is the accepted standard by the Met Office. I wish I had started my Davis data logging from that because I have missed out on dozens of station records due to doing it midnight to midnight. One in particular is, the daily ‘Highest Minimum Temperature’

For example, back in the Summer, I recorded 18°C as the highest minimum temp at around 5am, but it was wiped out by a recording of 17.8°C shortly before midnight on the same day. If I had originally set my Davis rollover at 9am, then I would have broken the record. It has happened so many times… There’s no point in changing it now, as all my data would be skewed.

Midnight to midnight is there just to keep it simple for most folk.

1 Like

for reporting the values in a data base I would say use the 24 hour period starting at midnight.
I thought about reporting it for the last 24 hours on the weather station (but still report from midnight to midnight for databases, or historic data). These were my two scenario’s:
When a user gets up in the morning and looks at the weather station, he probably want to see how cold it was during the night. The coldest period mostly happens after midnight, so that is fine. When a user looks at 15.00 at the station, he or she is probably curious of how hot it is getting, not of how hot it was the day before at 16.00
So reporting it for the last 24 for hours is, I think, not a good idea.

Does that answer your question? not really, just that I think that midnight to midnight is fine. If you wanted, you could have max today and max yesterday,

1 Like

It gives a good perspective from another view. I’m looking for ideas to better display the data. As far as the database, every record is time stamped so it does not matter. The data can be grouped by any timeframe. I’m thinking I will add an option to let the user decide on when the 24 hour period starts.

Personally, I’m more interested in the low reading AFTER the timestamp of the High…As the temperature reaches a high, what did it drop to and when? In Colorado, it can climb to 50 then drop down to 20 during the day. That is of more interest to me than it being 20 at night


Are you perhaps more interested in the trend?
Or perhaps you are more interested in a tiny graph of the last 24 hours.

1 Like

I have NULL interest in graphs.

I’m interested in non-linear movement.

Like ‘how much did the temperature drop during the total eclipse last summer’ (answer, about 8 degF just SW of Portland OR - it got noticeably colder where I was at for the event).

I think a lot of software has the idea of a rolling 24 hour window, as well as things like “what’s the highest low temperature for a day this month” and the like, so there are a lot of fun ways to slice+dice the data.

FWIW, everything I can find from NWS and NOAA online seems to say they consider a day to be midnight-to-midnight local time for their reporting, but that might just be for consistency as it’s a pretty unambiguous definition when you look back at old data.