Anomalous temperature events in Portland, Oregon in the PDX era

[IbrChris]

Methodology

1. Gather all the daily highs/lows from PDX during the 1940-2013 period (Project was undertaken in late 2013)

2. Calculate the mean temperature for each calendar day (1940-2013 POR).

3. Calculate the standard deviation for each calendar day (hi, lo, mean)
4. Look at record highs and record lows for each calendar day and determine their degree of anomaly based on standard deviations from normal.

Top events

 

In each event I look for the largest anomaly in terms of standard deviation and compare events based on their standard deviations (sigmas).

Here are the most anomalous events based on sigmas for each temperature category, hi max, hi min, lo max, lo min:

Hi Max
+3.54 (80 deg on Mar 16, 1947)

 

Lo Max
-4.80 (17 deg on Dec 17, 1964), single most anomalous value in PDX history.

Hi Min
+4.00 (74 deg on Jul 17, 1941)

Lo Min
-4.38 (-2 deg on Jan 31, 1950)

In a normal distribution the odds of encountering (on a particular calendar day), in years:

+/- 3 sigma: 1:370
+/- 3.5 sigma: 1:2,149

+/- 4 sigma: 1:15,787
+/- 4.5 sigma: 1:147,160

In other words a +3 or -3 sigma anomaly on May 16th should be encountered roughly every 370 years.

However, temperature data is skewed toward negative deviations...ie larger cold deviations are more common than larger warm deviations, so the return period is considerably less than that of a normal distribution. In other words a -4 sigma event is not a true -4 sigma event on a normal distribution but more like a -3.5 event. A +3 sigma event is more like a +3.5 sigma event on a normal distribution.

Since highs and lows occur daily, for a daily event the odds of a day with anomaly of x sigma, in years:

+/- 3 sigma: Once a year
+/- 3.5 sigma: Every six years
+/- 4 sigma: Every 43 years
+/- 4.5 sigma: Every 403 years
+/- 5 sigma: Every 4,776 years

So the odds of a -4.5 sigma event occurring are once every 403 years in a normal distribution, obviously having multiple -4 sigma events at PDX in the last 74 years again indicates the distribution is not normal but is skewed in the cold direction.

Mean sigma for each calendar day record in the following categories:

Hi Max: +2.45
Lo Max: -2.24
Hi Min: +2.27
Lo Min: -2.56

Basically we see that hi max tend to be greater deviations than lo max, and hi min are less anomalous than lo min.

-4 sigma events at PDX 1940-2013

Lo Max
-4.80 Dec 17, 1964
-4.45 Dec 1, 1985

-4.39 Nov 14, 1955
-4.37 Nov 12, 1955
-4.29 Nov 15, 1955

-4.28 Nov 30, 1985
-4.28 Nov 16, 1955

Interestingly all -4 lo max sigma events are in November and December, with the exception of the +4 sigma hi min in July 1941.

Lo Min

-4.38 Jan 31, 1950

-4.32 Dec 8. 1972

-4.25 Feb 2, 1950
-4.06 Dec 17, 1964
-4.03 Feb 1, 1950

Lo Min -4 sigma events tend to be Dec-Feb.

Most anomalous day in PDX history appears to be Dec 17, 1964 with a high of 17 and low of 7.
 

[IbrChris]

I love seeing this kind of data, now may I ask as many others will, how about Seattle and Vancouver BC?

Seattle wouldn't be tough to do...Vancouver would be fairly time consuming. I like having all the data where I can just copy it into Excel to analyze it using Pivot Tables. I didn't type any of the data into Excel

 

I'll see what I can do with Sea-Tac data though.

[IbrChris]

In all truth though this is the proper method to compare the various anomalous events in a station's history, not by the incomplete brute-force recitation of high/low values in support of an argumentative position as is common on here.

[SilverFallsAndrew]

This is very interesting stuff Chris. Thank you! 

[Front Ranger]

Interesting how all the greatest sigma events were within a 25 year period.

[wx_statman]

No question about December 1964, although that is more a product of a relatively short POR than anything else. If it wasn't for the 6 degrees on 12/16 and 7 degrees on 12/17 that year, the record lows at PDX would be 20 and 19. If it wasn't for the maximums of 17 on both 12/17 and 12/18 in 1964, the records would be 35 and 33. Those are ridiculous margins.

 

A longer POR would have included cold waves like 1924 on those dates, which would serve to lower the sigma values to more reasonable levels.

[Front Ranger]

No question about December 1964, although that is more a product of a relatively short POR than anything else. If it wasn't for the 6 degrees on 12/16 and 7 degrees on 12/17 that year, the record lows at PDX would be 20 and 19. If it wasn't for the maximums of 17 on both 12/17 and 12/18 in 1964, the records would be 35 and 33. Those are ridiculous margins.

 

A longer POR would have included cold waves like 1924 on those dates, which would serve to lower the sigma values to more reasonable levels.

 

Yup, and that goes along with bbkid's point in the other thread - it makes more sense to look at more than just what's occurred that day.

[IbrChris]

No question about December 1964, although that is more a product of a relatively short POR than anything else. If it wasn't for the 6 degrees on 12/16 and 7 degrees on 12/17 that year, the record lows at PDX would be 20 and 19. If it wasn't for the maximums of 17 on both 12/17 and 12/18 in 1964, the records would be 35 and 33. Those are ridiculous margins.

 

A longer POR would have included cold waves like 1924 on those dates, which would serve to lower the sigma values to more reasonable levels.

 

I could build one for downtown...would be more time-consuming however.

[IbrChris]

Great info.  How much time did you put into this?

Couple hours...mainly because I use pivot tables somewhat rarely so I found myself re-learning how to use some features.

[Glacier]

I love seeing this kind of data, now may I ask as many others will, how about Seattle and Vancouver BC?

I've done similar work on BC weather stations in the past, and can't remember where I had saved the work, but the results would be similar to Portland. For example, November 1985, various months in 1955, and January 1950 were some of the coldest up here. December 28th to 31st, 1968 (depending on location) saw the lowest maximum on record, so that would probably be up there as well.