30 Year "Normals" changing - How does this affect historical data?

[easternsnowman]

No it is not.

It's making the 1981-2010 normals irrelevant because we're always above them.

[tamarack]

It's making the 1981-2010 normals irrelevant because we're always above them.

For a lot of places, the 1980s - early '80s especially - were comparatively cold, with the '90s and "Oughts" significantly warmer. Temps 2011-on resemble the 90s-00s, and thus will be warmer than a 30-yr avg that includes the 1980s.

I'm on year 15 at my current location and keep a running avg of data for that period. Wishing to see how that compared to the 30-yr "normals" which are lacking for MBY, I took 1998-2011 data from the nearest COOP with complete data (6.5 miles to my WNW and similar elevation) and compared those 14 years to their 1981-2010 avg. Their temps 1998-2011 have averaged 1.14F milder than for 1981-2010, with individual months ranging from +0.40 (July) to +1.94 (Sept).

[snowman21]

I can't say for sure if there has been a increase in the amount of days where the low doesn't drop below 50°F, but that is excessively warm for here in the winter time for a low. That's close to our normal high during the coldest part of winter which is 56°F, used to be 54°F.

My whole point was that the entire climate record is a better depiction of climatology than just 30 years. The climate has not changed significantly enough to be that different and the entire record period shows a bit better data with more accurate means in my opinion. Specifically, the last 30 years definitly skews things toward the warmer and I'm in not in the camp that believes in "Global Warming." To me that's the same as only averaging the most recent 30 grades out of a series of 100 projects to determine someone's grade. It's arbitrary.

Normals are generally not arithmetic means. Each normals period uses different methods for computation. For the 1981-2010 set, the daily normal temperature is based on a two week window centered around the day being calculated. This means for any given day of the year, there are roughly 15 * 30 usable data points instead of just taking the average of the 30 data points recorded on that calendar day. Note that there is some filtering that takes place on the available data points, so it's still not just a simple average. The last set of normals used a spline interpolation for the daily values fit through the monthly averages. Adjustments are made to the climate record to account for station siting changes, instrumentation changes, observation times, and so on (the climate record homogenization process). A lot more goes into it than most people who are far removed from the process realize - and this is just for temperatures; never mind precipitation, degree days, and the other elements which are computed using different methods.

Your shift of coldest normal high could be entirely due to the methodology without implying any real warming (hard to know without researching a bit more). Looking back at past normal sets, DFW had a lowest normal daily high temperature of 53 in the 1951-1980, 1961-1990, and the 1971-2000 sets (DFW normals were not produced for the 1921-1950, 1931-1960, and 1941-1970 sets). Now, the lowest as you've noted is 56 although technically it's 55.5. Remember about half of the 1981-2010 normals incorporate ASOS data which will impact the climate record.

[JBG]

The way the climate has warmed recently it is making the 30 year norms meaningless!

Not really.

At least in New York, if you go back to the period from around 1949 to very early 1956, you have a slew of almost snow-less winters and some very hot summers. In fact, KNYC went over 100 during 1952, 1953, 1954, 1955 (four years in a row) and 1957. Most of those winters were boringly mild, until a March 1956 snowstorm (until 2010-11 the only KU event during a moderate or strong La Niña). The summer heat of the early 1950's was very similar to the current run of 2010, 2011 and 2012.

Another issue that use of a longer base period would address is the over-weighting of snow events to March for periods such as the 1950's. Both before and after, February is the snowiest month. In recent years we've done better in December and January than during the 1950's and 1960's. Use of a longer base period would smooth out these anomalies. The period from roughly 1993 to the present has also had a relatively snowy cast, much like the late 1950's through about January 1, 1971. These periods, as well as the "snow droughts" such as the ones of the period from 1949 through February 1956, and March 1983 through February 1993 would become less distortive than they are during a 30 year base period.

So it's not warming that militates against using a 30 year base but in fact its opposite, cyclicality. The base should be about 60 years, now that measurements are available for at least that long a period at many stations.

[Mallow]

Not really.

At least in New York, if you go back to the period from around 1949 to very early 1956, you have a slew of almost snow-less winters and some very hot summers. In fact, KNYC went over 100 during 1952, 1953, 1954, 1955 (four years in a row) and 1957. Most of those winters were boringly mild, until a March 1956 snowstorm (until 2010-11 the only KU event during a moderate or strong La Niña). The summer heat of the early 1950's was very similar to the current run of 2010, 2011 and 2012.

Another issue that use of a longer base period would address is the over-weighting of snow events to March for periods such as the 1950's. Both before and after, February is the snowiest month. In recent years we've done better in December and January than during the 1950's and 1960's. Use of a longer base period would smooth out these anomalies. The period from roughly 1993 to the present has also had a relatively snowy cast, much like the late 1950's through about January 1, 1971. These periods, as well as the "snow droughts" such as the ones of the period from 1949 through February 1956, and March 1983 through February 1993 would become less distortive than they are during a 30 year base period.

So it's not warming that militates against using a 30 year base but in fact its opposite, cyclicality. The base should be about 60 years, now that measurements are available for at least that long a period at many stations.

For "sparse" data like snowfall, I agree that a longer base period makes some sense.

[JBG]

For "sparse" data like snowfall, I agree that a longer base period makes some sense.

But not for temperature data?

[Mallow]

But not for temperature data?

There's a big enough (nonzero) sample size in thirty years of temperature data that I think it's a reasonable representation of the "current" state of the climate.