Global warming is something I hear about every day and I have been interested in this topic for the past couple of decades. For this reason, I wanted to take a dive into climate data to see what temperature changes have happened during my lifetime.
I want to cut through the political agendas and understand what is going on in our atmosphere over time, with a focus given to changes that have occurred over the past 50 years.
In part 1 of this series, I explained where the data originated and how the data was processed for this work.
In this article, I wanted to answer a very specific question. I wanted to understand how the spatial distribution of maximum air temperatures have changed over time. In other words, I want to know how much warming has occurred during my lifetime and where the warming has occurred. This seemed like a fun topic to pursue, and much to my surprise, the answers “blew me out of the water”.
My instincts and all the information I had been reading indicated to me that global warming is pervasive and ubiquitous. I was expecting to find rampant warming over the past 50 years, from the decade of the 1960’s to the 2010’s.
The results of this work were so different than this, however, that I had to check, double check and triple check my work. I was stunned. I have never seen climate data presented in this way.
Three Alteryx workflows and three Tableau dashboards later, I finally determined that what I was seeing is real, is based on solid data, and is very interesting. I hope you enjoy seeing what this analysis has uncovered.
I am not a climatologist. I will not be attempting to explain what factors lead to the results I am going to show. I am not politically motivated.
I know that predicting air temperature changes over time is notoriously difficult for many reasons. I spent over 20 years performing numerical simulations of natural systems (groundwater and surface water) and I learned how hard it is to accurately simulate non-linear and chaotic systems such as our atmosphere.
For these reasons, I am only going to present the results without worrying about what caused the changes that I have visualized. In other words, I am not promoting any agenda in this work – I am simply using Alteryx and Tableau to visualize real-world data.
When I was a little boy growing up in Chicago in the mid-1960’s, the cold was biting. There were days when the cold hurt my bones as I walked to and from school. Now when I visit Chicago, the cold that once existed doesn’t seem to occur as often and it doesn’t seem as intense. Last week I was there and the wind was whipping, but the temperatures were not representative of the Decembers that I remembered.
When I moved to Tennessee in the late 1980’s, the winters were much harsher than they are now. In fact, my grass is still growing and it is now January! It is clear to me that air temperature changes are taking place during our lifetime. These changes are not imaginary.
However, since I have only lived in a few places, I didn’t want to focus my analysis on those areas. I wanted to take a look across the globe to see how maximum air temperatures have changed from my childhood (1960’s) to my adulthood (2010’s). Here is how I did it.
In a decade, there are 10 years. This means that there are 310 days in January (10 years * 31 days per month), as well as 310 days in March and 300 days in April, etc. Since the data set I have is based on daily observations of maximum air temperature (as well as min temps), what I did was calculate the average monthly maximum air temperature by decade by month. I did this for every decade of data that existed in the data set.
Before calculating the averages, I removed all data that had any data quality issues. What was left appears to be a very clean, robust data set for the 1,788 monitoring stations I showed in part 1 of this series. Of these stations, over 1400 of them are based in the United States. Some of these stations did not have data in the 1960’s, so these stations dropped out of the analysis.
Part 1 – Visualizing the Daily Data
In the first part of the analysis, I needed to visualize the daily data to make sure that the Alteryx workflow that was used to parse the data was working correctly. An example of the dashboard is shown in Figure 1. The data looked steady and clean from a large number of monitoring stations that I interrogated.
The value of Dashboard 1 is that it is fascinating to pop around the world and see how the maximum temperatures vary by day, by month, by decade, as well as by location.
This fascinating dashboard is driven by over 64 million rows of daily data and used the powerful Tableau union capability to read nearly 1800 csv files at one time. Of course, I could have used Alteryx to do this union operation but I wanted to give Tableau a test and it did just fine. Maybe I’ll throw this data at Power BI and watch it buckle under the pressure of the job.
Part 2 – Visualizing the Max Temperature Changes From the 1960’s to the 2010’s
Although it is tempting to play with dashboard #1 all day long, I needed to move onto another form of analysis that would let me quickly visualize the differences between the monthly temperatures across multiple decades. In other words, I wanted to quickly compare the max temps from the 1960’s to the 2010’s, so dashboard number 2 was developed as shown in Figure 2.
Although I was tempted to play with dashboard #2 all day, I still wasn’t able to see what my curious mind was longing to see. If you look at the difference in air temperature for June in this example, there has been a 10 degree warming from the 1960’s to the 2010’s.
What I wanted to see was a map of this type of maximum air temperature change for all monitoring stations in the database, for any month I was interested in. This lead me to create the third Alteryx workflow and the third Tableau dashboard, as well as playing around with some very powerful LOD calculations in Tableau.
Part 3 – Visualizing the Spatial Distribution of the Maximum Temperature Changes
Before I looked at the results I am about to show you, I formed a mental picture of what I expected to see. I expected to see a shotgun blast across the map with some stations showing more warming than others, possibly with a relationship to the monitoring station elevation.
What I immediately saw, blew me away and I instantly knew I had uncovered something very interesting. This is a perfect example of the pure power of Tableau visualizations. Figure 3 shows the change in maximum air temperatures from the 1960’s to the 2010’s for the month of January. Sometimes point data can be very instructive compared to spatially interpolated data that we typically see for this type of analysis.
Figure 4 shows the same type of results for February. Notice the difference? Yes, the differences are huge! A large swath of the country has experienced lower max temperatures in the month of February, which indicates cooling conditions. The clustering of this data is not a figment of my imagination.
Figure 5 shows the same type of results for March. Once again, do you notice the difference? Yes, the differences are very prounced with much of the country having undergone a warming in March. There is a particularly noticeable warming in the central region of the US, especially over Iowa and Nebraska. This type of pattern is what I expected to see for every month!
To understand how interesting this analysis is, I created two pdf files. The first file is focused on the US-based monitoring stations like those shown in Figures 3-5. The second file is a view across the entire globe, as demonstrated in Figure 6 for the month of April. Notice the warming that has occurred in the other parts of the world, although there is both warming and cooling in the US. This data continues to amaze me in both the observed patterns (clusters) as well as the spatial consistency of the trends across the globe.
As for the remaining 8 months of the year, there are interesting patterns that developed. I discuss the different patterns for each month for both the US-based view and the entire world view in the following video.
Part 4 – Visualizing the Relationship Between Maximum and Minimum Temperature Changes from 1960’s to 2010’s
Just as the Tmax data was processed for these stations, so was the Tmin data. I wanted to see what the relationship looked like between Tmax and Tmin changes over the 50 years. I hoped that there were would be a strong correlation between the changes such that if Tmax showed signs of warming, so would Tmin. I also wondered whether Tmin or Tmax would show more perturbations over the 50 year period of analysis.
- Maximum daily air temperatures are changing over time but not in the way that I expected.
- Minimum air temperatures are also changing, and for many months the changes are greater than the maximum air temperature changes.
- Although I expected widespread increases in the maximum air temperatures over the past 50 years, there are many examples of large regions where maximum air temperatures have dropped, which indicates cooler air being present in the 2010 decade compared to the 1960’s. I definitely did not expect this result.
- The magnitude of the air temperature changes is significant, is asymmetric and appears to range most commonly between -5 to +10 degrees F (cooling to warming). The distribution indicates more warming has occurred overall, but in certain months there has been definitive cooling that has occurred.
- The patterns of maximum air temperature changes must be being controlled by big-scale forces, such as El Niño & La Niña events, changing ocean currents, and other things that I probably don’t even know about. I am not even going to attempt to guess how these changes have occurred, but I am fascinated to have discovered these changes in min and max air temperatures!
If I had unlimited time, a fast computer, and a lot of storage space, I’d let Alteryx rip into the 80,000 other weather monitoring stations and I’d process the full-monty of this data set. I would also compute results for other decade sets to see if these types of patterns emerged multiple times. I now wonder if the results I have seen are just a result of chaos reigning within our atmosphere or if there are underlying causes that could be determined.
I think this work is awesome and has already changed the way I am thinking about the changing nature of our climate. I’ve got snowfall, snowfall depth and precipitation to play with too, so stay tuned if you like this stuff. I sure do love the power of Alteryx and Tableau!
Speaking of Weather Data
I’m adding this section because I want to remember the work of this person. Nathalie Miebach takes weather data and turns it into sculptures and music. She obviously is very talented, so I recommend that you check out her work! The hurricane-inspired musical score attached below was downloaded from her site.