Climate Normals & Daily Temperatures

[CapturedNature]

This is more of a climatology question than a meteorology question but I didn't see a more fitting place to post this. I'll admit that I went to Met school 25 years ago but I've never worked in the field and most of what I learned is long gone but I've always kept and weather record for my own personal use and over the past few years I've been getting back into daily monitoring.

My questions have more to do with climate normals and how they are derived than anything.

1. Why are climate normals computed in 30 year periods and not longer periods? It seems to me that 30 years is not enough time to accommodate decadal patterns and if a longer record is available, why not use the full breath of data that is available?

2. Why are normals only computed every 10 years? Couldn't they be done every year?

3. In terms of daily temperatures, why is the mean used instead of the average? We currently take the mean temp. or the average of high and low for the day but often they may not be reflective of actual temperatures that day. For example, if the high occurs at midnight and drops all day, the mean temp will be higher than what the average (or sum of all hourly temps divided by 24). Doesn't that skew the data?

4. In terms of daily temperatures, we currently use either 0000 to 24:00 or 0700 to 0700. Which is more "correct" in terms of actually depicting a climate? I know that midnight to midnight is a day and determines temperatures on a given day but we also measure seasons in whole months and not 12/21 to 3/20 or 3/21 to 6/20. I kind of like the AM to AM approach as it is usually the min to min, but I'm wondering if there is some official explanation.

I hope that some enlightened folks can help me fill in some knowledge gaps or we could have a discussion on what is best.

Thanks in advance!

[kylemacr]

Here's an answer to the first question.

1. Why are climate normals computed in 30 year periods and not longer periods? It seems to me that 30 years is not enough time to accommodate decadal patterns and if a longer record is available, why not use the full breath of data that is available?

This is true for a few reasons, the main one being historical inertia. This is a quote from a WMO report (remember, the WMO mandates the 30 year normals) that you can read here.

The concept of the 30-year climatological standard normal dates from 1935, when the Warsaw

conference of the International Meteorological Committee recommended that the period 1901-

1930 be used as a world-wide standard for the calculation of normals.

This idea has lots of issues though. Remember that the 30 year choice is really kind of arbitrary. Here's a recent article from BAMS that describes some of the issues in detail. That article makes a lot of really good points.

Bottom line is that there isn't a whole lot that's special about that 30 year cutoff. In many, many applications it would be best to use the whole climatology. When we scientists are in "research mode", we tend to do this. For instance, in many research papers then anomalies are used (whether they be temperature, precipitation, geopotential height, or any other variable) anomalies are created by removing the long-term (whole or most of the climatology) and much of the seasonal cycle.

You really have to know what you're trying to do and figure out the best way to construct your anomalies. Statistics is never a one-size-fits-all problem.

[CapturedNature]

Here's an answer to the first question.

...

Bottom line is that there isn't a whole lot that's special about that 30 year cutoff. In many, many applications it would be best to use the whole climatology. When we scientists are in "research mode", we tend to do this. For instance, in many research papers then anomalies are used (whether they be temperature, precipitation, geopotential height, or any other variable) anomalies are created by removing the long-term (whole or most of the climatology) and much of the seasonal cycle.

You really have to know what you're trying to do and figure out the best way to construct your anomalies. Statistics is never a one-size-fits-all problem.

Thanks for the thoughts and links. I'm well aware of how statistics can be manipulated and it's one of the reasons I asked why the daily mean was used instead of the average. It can really skew the perception of a day.

[famartin]

Of course, there are cases where it is important to use the most recent data as opposed to the whole period of record, such as sites where a lot of urbanization has occurred. You definitely don't want to use the whole period of record for Reno or Las Vegas, where the urban heat island has greatly skewed low temperatures in particular during recent decades.