Updated: July 22, 9:30 AM PDT

Departures from normal could apply to a lot of things recently, but I’m speaking here of daily average temperature departures from 30-year normals, as calculated by the National Oceanic and Atmospheric Administration (NOAA) and the National Weather Service (NWS) for Seattle-Tacoma International Airport.

Seattle and the Pacific Northwest is currently undergoing a historic heat wave. The first time-series chart is a bar chart representing every day since January 1, 2000. That is 7850 days up to June 28, 2021, so the bars appear as thin lines. Each bar plots the departure from 30-year daily normals as reported by the NWS. Red bars or lines represent positive numbers (warmer than normal days). Blue bars or lines represent negative numbers (cooler than normal). A 365-day smoothing curve has been applied to the chart.

On Sunday, June 27, I added a series of red data points where the average daily temperature for a given day matched or exceeded 20˚F. I added a series of blue data points where the average daily temperature for a given day matched or fell below -20˚F. The 20˚F limits were arbitrary and chosen to:

• limit data clutter

• show only really hot or cold days (for a given day), and

• because yesterday was extraordinarily hot and was 22˚F warmer than normal for June 26 (and later in the day, added data for June 27, an even hotter and more extreme outlier).

I’ve since added similar data for Monday, June 28

What I found…

• Since roughly 2013-14, the trend has been towards slightly warmer days (e.g. 1-3 degrees F).

• The number of days 20˚F or cooler than normal were four (4). All occurred before 2011.

• The number of days 20˚F or warmer was ten (10). All but one of the really warm days (for a given date) occurred for this data set after 2016. I added data for June 28 to the chart this morning. The NWS has released preliminary data that it reach 108F, a record, at 5:13 PM. This was 34˚F greater than the normal high for the date, and the daily average temperature was 23 degrees warmer than the 30Y norm for this date. That 23 degrees was actually down from the day before even though the daily high was higher than Sunday’s.

• Note: there is an extra data element shown here ‘unrelated’ to temperature, For a period from roughly 2014 until mid-2018, SeaTac received roughly 42 inches more of rain than normal. This is more than one year of rain beyond what we normally see. I just had a roof replaced. It got quite mossy over the past 6-7 years.

Figure 1B simply shows a cropped image of Figure 1 to give a clearer picture of what the detail data underlying Figure 1 appears as.

Figure 1C is new and focuses in on just data since the start of this year.

Table 1 shows the specific dates and temperature departures displayed on the timeline of Figure 1. Table 1 is sorted in chronological order, oldest to most recent. Shaded rows highlight successive days.

Table 1 highlights several things:

1. Since 2000, very cool days (-20˚F or below normal) for a given date have occurred over a four year span from 2006-10.

2. These much cooler than normal days typically occurred in late November or December, out darkest months in Seattle.

3. Since 2000, much warmer than normal days for a given date have occurred over a nearly twelve year span. There are ten (10) of these days. All but one of those days have occurred in the past five years.

4. The warmer days occur over a range of seasons including mid-Spring to mid-Summer.

5. About half of the extreme days occur on consecutive days. This makes sense. Weather systems are very large and cam take several days to break patterns.

Histogram

Figure 2 displays a histogram plot of the total population of daily temperature departures from 30Y climatic normals and it shows a normal distribution. I suppose this is to be expected. Climate does change over time, but relatively slowly. A 20 year time span of day-to-day data is unlikely to show highly skewed distributions. This distribution does skew slightly towards higher temperatures.

Box Plot Distributions

A box plot is a statistical tool which can assist with seeing data distributions. It divides data sets into quartiles. It shows the median of the data. If you add whiskers to the plot it can help identify potential outliers in data. A primer on box plots can be found here.

I plotted the same data set as above using a box plot. It includes (as of 6/28/21) 7850 data points. The box plot is shown in Figure 3. You can click on it to expand it.

what I found…

• The NWS reports daily temperature departure in whole numbers or integers. So the data points shown line up directly over discreet ticks along the X-axis.

• The median temperature departure for all 7850 data points was 1˚F.

• The mean temperature departure for this 21+ year period was 0.9022 ˚F, up .001˚F after Monday’s record high temperature.

• The first quartile for all data points occurred at -2˚F.

• The third quartile occurred at 4˚F.

• The IQR was 6˚F.

• The coolest departure was -23˚F.

• The warmest departure for any given day was 25˚F (6/27/21).

• The range of data spanned 48˚F

• The software application [1] that I use for graphics development draws outliers when the data points exceed 1.5X the Inner Quartile Range (IQR) from the median for the data set. The whiskers (or lower and upper adjacent values) are drawn only to the smallest or largest non-outlier. Extreme outliers that are 3X the IQR are drawn as open circles. With 7850 data points to consider with a widely variable data point like daily temperature fluctuations, there are bound to be some outliers to the data. Figure 3 shows three of the most extreme outliers occurred June 26-28, 2021.

Finally, I’ll add a few more box plots showing the same data in slightly different formats. The first of these additional box plots, Figure 4, adds the data points in between the lower and upper adjacent “whiskers” and inside the Interquartile range.

Often, these individual data points are of little concern to the greater picture. But I wanted to show those unfamiliar to box plot diagrams that there are data points between the whiskers on the chart. Essentially, this diagram design is now a strip plot with the IQR and whisker elements of a box plot added as a top layer.

I did two other things:

• I also added a little ‘noise’ to the data to jitter the values slightly. If I did not do this, all data points would fall on discreet integer tick mark values due to the NWS dataset assigning whole numbers to Temperature Depart data. In essence, most of these 7000+ data points would be stacked on top of each other and be hidden if I did not add ‘noise’ to the data.

• Second, I turned the color transparency value of the data dots to 15% opacity to help see overlapping data points. There are many, many more data points in the IQR region of the graph and due to the overlapping created by the noise these appear as ovals instead of circles.

I left the outlier layer turned on in Figure 4.

In Figure 5, I basically plotted the same chart as in Figure 4. However, in Figure 5 I’ve turned off the outlier layer. There are no solid black or unfilled outlier circles shown. But because these points are beyond the left and right chart whiskers, we know them to be outliers.

And I left the outlier date labels for June 26-28. In some ways, this visually enhances the depiction of the outliers, showing more slightly jittered points than the solid or outlined circles do.

The past several days, temperatures have risen into the triple digits. These values have been 20-30 degrees higher than normal. I believe it’s been worse in Portland. Places east of the Cascade Range have been hit very hard as well, as has Northern California.

But I live in Seattle, on Capitol Hill, and that is where my focus will be. I plot all sorts of data as a hobby, much like others do crossword puzzles daily. I wanted to compare the elevated daily temperatures we’ve seen recent and plot these along a timeline.

My data is limited. I only have data going back to January 1, 2000, or about 21 years. Of course data exists for SeaTac Airport well before 2000. I even have it downloaded but it requires a lot of post processing to fit my data model. The NOAA/NWS CSV format for that earlier data is significantly different than more recent years. So I’ll plot what I have for now.

Full disclosure: I am not a statistician, applied statistician, meteorologist, nor a climate scientist. I did have a 30+ year engineering career in civil aviation so I am familiar with numbers. And I enjoy looking at data and applying basic statistics to the numbers. But a skilled statistician or professional in the weather field might quibble with some of what I’ve shown here, and they would likely be correct.

I prefer seeing data as it presents itself. I prefer keeping my politics out of what I find and publish. The truth is, I don’t know what significance any thing I might find in the data may have on the larger picture of things. Climate science is not my specialty. I really doubt anything I present on this little blog is not already well known by scientists in the fields of interest, whether climate science, meteorology or any related field. But if it helps push people who have a direct interest in this data to dig deeper, more power to them. I work on these diagrams and charts like others work on a daily crossword puzzle. It just interests me.

Having said that, I think it’s fair to say the past few days have been historic and extraordinary in the Pacific Northwest, or at least at Seattle-Tacoma International.

Updates: 6/28, 8:10 AM:

1. Corrected t-Depart values for 6/27 based on updated official data from NWS.

2. Added Figure 1B for more detail and clarity of underlying data

Updates: 6/28, 3:15 PM:

1. Added a histogram to the blog post.

2. Renumbered Figure Nos. to reflect new chart.

Updates: 6/29, 7:00 AM:

1. Updated all Figures (except for Figure 1B) with data from 6/28, Seattle’s hottest day in recorded history.

2. Added Figure 1C.

Updates: 7/22, 9:30 AM

1. Updated Figure 1 for data through July 21, 2021.

2. Updated Figure 1C for data through July 21, 2021

Data sources:

NOAA / NWS Forecast Office: https://w2.weather.gov/climate/index.php?wfo=sew

Software tools:

1. Datagraph 4.7.1 by Visual Data Tools, Inc. for macOS X for statistical charts.

2. Microsoft Excel for macOS X for data aggregation, sorting, filtering and Pivot Table functions.

COPYRIGHTS

All datagraphics on this site © David Blackwell, Seattle, 2021. All rights reserved.
Please contact me using the form linked at left for permission to use.