What Happened
Summer weather in 2023 was a hot topic. Pun intended. And leading the wave of weather-related headlines was Phoenix, Arizona. Articles titled “’Hell on earth’: Phoenix’s extreme heatwave tests the limits of survival” (The Guardian, July 14, 2023), “Phoenix braves relentless wave of extreme heat in US Southwest” (Reuters, July 16, 2023), and “Phoenix Heat Becomes a Brutal Test of Endurance” (The New York Times, July 13, 2023) gave the nation the impression that Phoenix was experiencing heat on a biblical scale. Records were being “shattered”, heat-related deaths were “climbing”, and the outlook was bleak. I personally fielded a few bizarre heatwave-related questions from those I work with who live outside Arizona like “Are car bumpers melting?” and “Can you get 3rd degree burns when walking on asphalt without shoes?” To the outside world, Phoenix and its inhabitants were melting away.
As a resident of Phoenix since 1986, all this national attention forced me to pause and question the claims being made about this summer’s “scorching” temperatures in the Valley of the Sun. Was the summer of 2023 as hot as the media portrayed? I personally didn’t notice a difference from past summers. But then again, my wife has claimed I can be oblivious from time to time. So, I decided to investigate myself and analyze the data because as W. Edwards Deming put it “In God we trust. All others must bring data.” Here is what I found.
A few notes before we begin. The temperature data for this analysis was pulled from the U.S. National Weather Service (www.weather.gov). While this article will focus on the summer months for the past 37 years, additional insight will be made using data as far back as 1900. Finally, summer in the northern hemisphere technically begins on the solstice, which typically occurs around June 21st. It spans 3 months and ends on the autumnal equinox, which typically occurs around September 22nd. However, this analysis will consider summer months to be May through September to capture months that regularly see temperatures above 100F.
Getting to Know the Data
The best place to start when analyzing data is to understand its general behavior. Table 1 provides statistics of the daily high temperatures by summer month from 1986 through 2023. Figure 1 is a modified scatter plot of this same data and provides an excellent visual of the behavior.
Table 1: Descriptive statistics of the daily high temperatures for the summer months for the years 1986 through 2023. Note, the sample size is larger for May, July, and August because these months have 31 days whereas June and September have only 30 days.
Figure 1: Scatter plot of the daily high temperatures for the summer months for the years 1986 through 2023. The blue line is a smoothed mean line that crosses the average high temperature of each month. Note, months with taller, skinnier groupings have higher variation or less temperature consistency. The month of May is a good example of this shape.
We can see from Table 1 that July has the hottest average high temperature of the summer months despite not having the highest recorded temperature during this time, which occurred in June. August has the next hottest average temperature, followed by June, then September, and finally May. However, we do find there is no statistically significant difference between the average high temperatures of June and August. Therefore, we cannot say with confidence that the average high temperature of either month is hotter than the other. This can be concluded by the “Prob > F” statistic seen in Figure 2, which is greater than 0.05, indicating the groups are not significantly different. The results of this analysis are consistent when examining the temperature data as far back as 1900.
Figure 2: Analysis of Variance for the Oneway Fit comparing June and August high temperatures for years 1986 through 2023. The chart in the figure shows the daily high temperatures for the months of June and August and the mean diamonds (green) for each group. Essentially, if the mean diamonds overlap enough then the populations are not significantly different. The “Prob > F” statistic, or p-value, in the table below the chart assigns a value to the amount of overlap that exists. If it is greater than 0.05 (95% confidence) then there is no significant difference between the averages, and therefore no differences between the groups.
Table 1 also provides statistics for the dispersion or variation in high temperatures for each month. In other words, how consistent are the temperatures. July and August have the least amount of variation of the summer months and therefore have more consistent temperatures. The variation in June and September are slightly higher than July and August, and the variation in May is the highest of the summer months.
We can also look at how the data behaves over time. Analysis shows there is a statistically significant upward trend in high temperatures for the months of June through September from 1986 through 2023. The trend for July can be seen in Figure 3 as an example. We see the “Prob > F” statistic is less than 0.05, indicating there is a statistically significant relationship between the variable on the x-axis (year) and the variable on the y-axis (high temperature). While the trend for May was not significant over this same period of time, all 12 months show a statistically significant upward trend when the sample size is increased to include data from 1900 to 2023. This tells us temperatures are increasing over time, and so we should expect 2023 to have record-breaking temperatures. But let’s dive into the data.
Figure 3: Analysis of Variance for the Linear Fit Model of high temperatures vs year for years 1986 through 2023. The chart in the figure shows a scatter plot of the data, the line that best fits the data (red), confidence interval shading around the best fit line (light red), and the mean line (green). The mean line is a horizontal line, and therefore no relationship between the variable on the x-axis and the variable on the y-axis exists. This means that any change in the x-axis variable does not affect the value of the variable on the y-axis. The confidence interval shading indicates where the actual best fit line could fall for the population with 95% confidence. The “Prob>F” statistic, or p-value, in the table below the chart assigns a value to how much of the mean line resides within the confidence intervals. So, if the mean line falls within the confidence interval shading around the fit line, then there is a possibility that there is no relationship between x-axis and the y-axis variables and the “Prob>F” statistic will be greater than 0.05.
Temperature Talking Points
Three Phoenix weather-related statistics received a lot of publicity in 2023. The first was the number of successive days the high temperature reached 110F or greater, which was 31 days. This deserved the attention as it was almost twice the number of the previous record of 18 successive days set in 1974. More noteworthy however, is the fact that between 1900 and 2022, there were only 3 years whose total number of days with a high temperature of 110F or greater was more than 31. Said another way, 2023 had more days in a row at or above 110F than the total number of days at or above 110F for 97.6% of the years between 1900 and 2022.
While we are on the topic, the total number of days Phoenix had a high temperature of 110 or higher was also notable. 2023 set a record with 55 total days, surpassing the previous record of 53 days set in 2020. This may not seem significant since the previous record was set only 3 years prior and was only 2 days less. However, the record set prior to 2020 was set in 2011 and had 20 less days. This equates to a 60.6% increase in the number of days at or above 110F from 2011 to 2020 and a 66.7% increase from 2011 to 2023. Figure 4 provides a visual of this data for each year since 1900.
Figure 4: The bar chart of the number of consecutive days with a high temperature of 110F or higher (blue) and the total number of days with a high temperature of 110F or higher (red) for years 1900 through 2023. Note, there is a statistically significant positive trend for both the number of successive days and the total number of days with a high temperature at or above 110F vs time. This can be concluded from this bar chart as both the blue and red lines tend to grow in magnitude as the years progress.
Another statistic frequently reported in the news was July’s average temperature, which set a record at 102.7F. The previous record of 98.9F was set in 2020. We can start to comprehend the significance of the 3.8F increase when looking at how this metric is calculated. The average temperature is calculated by averaging the average high and average low temperatures for the month. Since this statistic is dependent only on these two values, both must be abnormally high, or one must be extremely high for it to exceed the previous record. In July 2023, both were extremely high. July’s average high temperature was 114.7F, more than 8F greater than the 106.5F mean for years 1986 through 2022. This was the first time since we started recording temperatures that the average high temperature for the month of July was above 110F. July’s average low temperature was 90.8F, almost 7F greater than the 84.0F mean for the years 1986 through 2022. This was the first time the average low temperature has been above 89F for any month in any year. Another way to provide prospective on the magnitude of the 3.8F increase is to note that July would have tied the previous record of 98.9F if the average low remained at 90.8F but the average high was 7.7F less, at 107.0F. Alternatively, the previous record would have been tied if the average high remained at 114.7F but the average low was 7.7F less at 83.1F.
The final statistic repeatedly seen in weather-related news for Phoenix was the number of daily high temperature records that were set or tied over the summer. This statistic is also deserving of the publicity it was given. The summer of 2023 had 13 days that set a record for the daily high temperature and 8 days that tied a record. This is a total of 21 days that currently hold (alone or shared) a daily high temperature record, or 13.7% of the 153 days from May through September. Put differently, one year out of 123 years holds more than 1/8th of the summer daily high records. The year with the next highest number of days that hold a daily high record is 2020, with 19 days. However, outside of 2023 and 2020, no other year currently has more than 9 days with a record daily high temperature (alone or shared).
Figure 5: The bar chart of the number of daily high temperature records currently held alone (blue), the number of daily high temperature records currently shared/tied with at least one other year (red), and the total number of daily high temperature records currently held alone or shared/tied (green) for years 1986 through 2023.
What Can Be Concluded
Based on the statistics above, it appears that the summer of 2023 deserved the attention it was given. The number of consecutive days at or above 110F and the number of total days at or above 110F were the highest in recorded history. The average temperature for July was significantly higher than in years past. And the 2023 summer had the greatest number of days with a daily high temperature record since 1900. But we do find that each of these statistics is a result of the abnormally high temperatures seen in the month of July, not the summer as a whole. Case in point, 30 of the 31 consecutive days with a temperature at or above 110F occurred in July and this makes up over half of the 55 total number at or above 110F. July alone set 7 daily high records, or 53.8% of the records set this summer, and tied 5 daily high records, or 62.5% of the records tied this summer.
We also find it was an average start to the summer. May and June were not statistically different than previous years between 1986 and 2022 as seen in the comparison found in Figure 6. Additionally, the first day of the year to reach 110F or higher was June 25, which is later than 84.2% of the years within the same period.
Figure 6: Analysis of Variance for the Oneway Fit for May and June comparing high temperatures for 2023 with the high temperatures for years 1986 through 2022. The average high temperature for May 2023 was higher than the average for the years 1986 through 2022 but not enough to be statistically significant. The average high temperature for June 2023 was lower than the average for the years 1986 through 2022 but not enough to be statistically significant.
But in the end, my wife was right. I can be oblivious from time to time. Although the 2023 summer in Phoenix started out like any other, it ended up hot. The months of July, August, and September were statistically hotter than previous years. In addition to July, August and September also set and/or tied multiple daily high records. August alone had 18 total days at or above 110F, second only to August in 2020. Even more astonishing was the fact that 2023 had 22 days with a high temperature of 115F or greater. This is 8 more days than the year with the next highest, and 15 more days than the year with the third highest. And I personally didn’t notice any of this.
One final thought. As I pointed out in the beginning of the article, temperatures in the Valley of the Sun are statistically rising over time. This upward trend was seen in data set after data set throughout my analysis and therefore it is safe to say, this is only the beginning.
Copyright © 2024 sheetsfreaks - All Rights Reserved.
Powered by GoDaddy
We use cookies to analyze website traffic and optimize your website experience. By accepting our use of cookies, your data will be aggregated with all other user data.