Introduction
I awoke this morning with an idea. I wanted to visualize global warming using a gridded approach to show the data. I was going to grid the data every 1 degree of latitude and 1 degree of longitude. I instinctively knew that this would create some interesting imagery.
Sometimes the Gods of analytics just shine on me. Today was one of those days.
Background
It turns out that there are gridded data sets of temperature anomaly data already prepared by some high-powered groups. The GHCND gridded datasets (HadGHCND) are produced through a joint effort between the United States National Oceanic and Atmospheric Administration (National Climatic Data Center) and the United Kingdom’s Hadley Centre. This work extends the Caesar_et_al-2006, as shown in the following reference (just click the link to read the paper for all the details of the methods used).
Reference:
Caesar, J., L. Alexander, and R. Vose (2006), Large-scale changes in observed daily maximum and minimum temperatures: Creation and analysis of a new gridded i data set, J. Geophys. Res., 111, D05101, doi:10.1029/2005JD006280.
The availability of this data quickly made my idea of displaying gridded temperature anomaly data a possibility. All this took was a little work in Alteryx followed by a little work in Tableau to produce the results shown in this article. I have focused on the anomaly data produced from daily maximum temperature data, but I could also do the same using the daily minimum temperature data set.
Alteryx Processing
I wrote an Alteryx workflow to process the 69 years of data included in both the Tmax (max daily temperature) and Tmin (min daily temperature) anomaly datasets. For both data sets, daily temperature anomalies in each grid cell are computed using average daily temperatures from the base period of 1961-1990.
Although I don’t want to go into all the technical details of how the data was calculated and interpolated by the organizations who prepared it (click here), I do want to give you knowledge on how to understand the anomaly data. This is a high-level overview of the methods used.
For any grid cell, find all temperature data present in the cell over the 30 year base period of 1961-1990. Compute the average daily temperature for Jan 1 (1961-1990), Jan 2, Jan 3, etc. There is a lot of detail in the paper on how this was done. Use these averages to compute the temperature anomaly for all Jan 1 dates between 1950 – 2018, and for all other dates during the year.
As an example, the Jan 1, 1950 anomaly would be Jan 1, 1950 – Jan 1 (average of 1961-1990). For any given cell, if the Jan 1, 1950 is 30 deg F and the baseline average is 40 deg F, the anomaly would be 30 – 40 = -10 deg F. This value of -10 deg F would mean that Jan 1, 1950, was 10 degrees colder than the average conditions measured during the baseline period. Therefore, negative anomalies indicate colder days and positive anomalies indicate warmer days with respect to the 30-year baseline period. Blue coloring is used for colder days, red for warmer days in the graphics included in this report.
Once those anomalies were computed for each day, the authors used an interpolation algorithm (see section 3.3 of the paper) to distribute the data across the gridded domain.
For the Tmax based data set, there are over 63.8 million daily anomalies present between 1950 and 2018. For the Tmin based data set, there are over 58.3 million daily anomalies. Alteryx computed the data sets and wrote Tableau hyper files in 8 minutes for each data set. Each data file required about 1Gb to store the 69 years of daily data in the Alteryx *.yxdb format.
I shifted the location of the measurement from the lower left corner of each grid cell by adding 1.25 degrees of Latitude and 1.875 degrees of Longitude for each cell. This means that in the graphics I show in Tableau, I am plotting the data at the centroid of each grid cell. There are 7,002 grid cells in these data sets.
Figure 1 shows the workflow I wrote to process the data.
Figure 1 – The Alteryx workflow used to process the 69 years of Tmax anomaly data.
Figures 2 and 3 show samples of the input and output data created by the workflow.
Figure 2 – The first few records of the 1950 Tmax dataset.
Figure 3 – The first few records of the Tmax output file, which is over 63.8 million lines long.
Analysis of Data Using Tableau and Mapbox Visualizations
My buddy AllanWalker, who works for Mapbox, is a visionary when it comes to mapping and Tableau. We have known each other for years, and every once in a while, Allan will impart some of his wisdom to me. Today was such a day.
I wanted to render this data on an outstanding world map. I didn’t have a background map that would suffice, so Allan showed me how to obtain one using Mapbox Studio.
In under five minutes, the background map I show in this article was created and it allowed me to perfectly display the data in the way I wanted to. Thank you Allan, as you are always so kind and generous with your knowledge. Just as you have told me in the past, Mapbox is Awesome!
Figure 4 shows time-series plots for each month of the year, using the Tmax anomaly data. The linear models were processed (see red dashed trend lines) and tabulated using the methods I previously described. For all months of record, it is easy to see that the anomalies based on daily maximum temperatures are increasing over time. This means that the earth is getting hotter!
Figure 4 – Monthly time-series plots showing the trend of Tmax anomalies over time. This is the world-wide representation of the data. Click the image for a full-scale version.
Figure 5 was produced to show the monthly total temp change over the past 69 years. As I have previously shown, March is the month that has been impacted the most, with a 3.9 deg F change over 69 years.
Figure 5 – Total maximum temperature change by month over 69 years (1950 to 2018).
As shown in Figure 6, the time rate of change in Maximum Daily temp varies by month. As expected, March leads the pack with nearly a 0.06 degree F increase per year in Max temperature. This rate is about twice that of July, which is 0.03 deg F per year. To understand why that is the case, you should read my theory that is explained in this article.
Figure 6 – The time rate of change of maximum temperature (deg F per year) by month over the period of 1950 to 2018.
With this data, it is possible to zoom into any region of the world where data has been collected to visualize the time rate of change of maximum temperature. Figure 7 shows an example of one grid cell selected and it’s monthly March history being shown in the time series plot.
Figure 7 – The March monthly time series shown for one selected cell in the northern latitudes, east of Greenland (the orange highlighted cell).
For anyone interested, I have produced pdf files that show the max temp distributions for every month, for every year of record. Each pdf file has 69 or 68 pages included. Figure 8 shows one example of one month/year combination from April 2018. You can easily notice the preponderance of red cells, which indicates hotter conditions than the baseline period.
Figure 8 – Temperature anomalies from April 2018. The following pdf files have every month/year combination possible with this data set.
Retrieve PDF Files
You can retrieve the pdf files with the links below. Each file is between 4 and 5 Mb so they don’t take long to download.
- Jan_Tmax_temp_anomaly
- Feb_Tmax_temp_anomaly
- Mar_Tmax_temp_anomaly
- Apr_Tmax_temp_anomaly
- May_Tmax_temp_anomaly
- Jun_Tmax_temp_anomaly
- Jul_Tmax_temp_anomaly
- Aug_Tmax_temp_anomaly
- Sep_Tmax_temp_anomaly
- Oct_Tmax_temp_anomaly
- Nov_Tmax_temp_anomaly
- Dec_Tmax_temp_anomaly
Watch Silent Animations By Month
If you want to see animations that scroll through the pdf files, click the links shown below. They are about 50 seconds long, they show 68 or 69 years of data, and they have no sound.
- January
- February
- March
- April
- May
- June
- July
- August
- September
- October
- November
- December
- All Months Shown in One Video (with narration)
Daily Temperature Anomalies
It is also possible to examine daily temperature anomalies from any gridded cell as shown in Figure 9 for March 30, 1950, and in Figure 10 for March 30, 2018.
Figure 9 – Temperature Anomalies on March 30, 1950.
Figure 10 – Temperature Anomalies on March 30, 2018, which is 69 years later than Figure 8.
Animation of March Anomalies Based on Daily Max Temperatures
In the animation shown below, I examine the change in March temperature anomalies over time from 1950 to 2018. This 69-year history shows how increasing temperatures have occurred, especially in the northern latitudes.
Figures 11 and 12 put all the pieces together for the year 1950 and 2017. You can easily see how the distribution of blues/oranges has shifted to more orange. The complete animation is shown below these figures.
Figure 11 – The anomalies for each month of 1950.
Figure 12 – The anomalies for each month of 2017.
I can also produce animations of daily temperature anomalies. Those are very interesting because they show the movement of temperatures across space and time. You can see cold and hot zones that move across the earth. For this reason, that level of detail is very interesting. If anyone is interested in performing this type of work, let me know and I’ll provide you the data.
Final Thoughts
Global warming continues to progress and I continue to work on visualizing the data. I don’t think I’ll ever be satisfied with my work, so expect more of this in the future.
If you are interested in understanding more, click this link to review my previous work. Thanks for reading.
