With all the heat waves and wildfires around, many people are asking whether today’s extremely hot temperatures are an anomaly or something that used to be the same in the past.
In this article I carry out a brief and simple analysis to shed some light on this question.
Let me start out by defining what I mean by extreme temperatures. Extreme temperatures in this article will refer to temperatures that are unusually hot or cold compared to the general temperature trends in a given region. The frequency of such temperatures is arguably very well measured by how recent hot and cold temperature records are in a given region.
I decided to look at extreme temperature data all over the world. Using Wikipedia’s “List of weather records” article, I calculate how many days ago the most recent temperature records were recorded in each country for which the data is available. Then I aggregate the data by calculating its mean and median.
The idea behind this is that if on average the current cold records were recorded say 15,000 days ago, but the current hot records were recorded 5,000 days ago, then it seems that indeed extremely hot weather is becoming more common. In other words, if there were no trends in temperature in either direction, one would expect that we see new records in both directions with roughly the same frequency.
I took Wikipedia’s data, got rid of all countries for which there were no dates or there were multiple dates. The resulting sample had cold records for 69 countries and hot records for 86 countries.
On average, cold records were observed 19,640 days ago, that is November 1, 1959. The median is pretty much the same (19,885). On the other hand, hot records were observed on average 9,547 days ago, that is June 20, 1987. The median here is actually even more alarming, it is only 3,653 days, that is August 10, 2003.
Visually we have this:
This graph shows when exactly each hot and cold record in the sample was recorded. One can clearly see that while cold records are more or less uniformly distributed until about 30,000 days ago, hot records are much more concentrated in the less than 10,000 days ago range.
The standard deviations are big, however; above 11,000 days in both cases. This means that strictly speaking, statistically we cannot really be sure that the mean cold record age (19,640 days) is significantly different from the mean hot record age (9,547 days). It could all just be a statistical coincidence.
The data may be biased because hot records are available for more countries than cold records. This could perhaps be because certain developing countries don’t keep records of extreme cold temperatures because they don’t really have too cold temperatures (speculation). Whatever the reason, these countries can potentially drive down the average hot record age while leaving the average cold record age unchanged. As a robustness check, I remove all countries that are not present in both samples from my list (I call this a ‘restricted sample’).
The resulting sample has 57 cold and hot records in the very same countries. These countries appear to measure both extreme ends of temperature officially. The results are largely unchanged. The cold mean is 20,773 days (median 21,011), the hot mean is 13,023 days (median: 7,302). As expected, the difference decreases but it is still large. Standard deviations do not change considerably.
Since the standard deviations are quite high (around 11,000-12,000), one may expect that there are continent- or region-specific effects. Some regions may experience more new records than others. How to divide up the world into regions from this perspective is hard to tell. A very easy and convenient division is by continents. And indeed, there are considerable differences in average record age among continents:
Also note that this table sheds light on another problem with the original analysis. Namely, that some regions (e.g. Europe) have a considerably higher number of countries. Consequently, there are more observations from those regions, so their temperature trends get weighed more heavily in our aggregate analysis. The table above presents an unweighted average, which is simply the mean of continent means with no regard to how many countries each continent contains.
One consequence is that since Europe appears to experience more severe differences between extreme hot and cold temperatures, they drive up the average (weighted) difference (which is 19,640-9,547=10,093 days). With the unweighted average, we get that the difference is only 7,085 days, as seen in the table above.
Looking at the table above we can see that it is only in North America and in Antarctica that the hot-cold record age difference is considerably lower and/or inexistent. Furthermore, South America also has a low mean difference but their median is high, close to the overall median. All this suggests that perhaps warm extremes are not more frequent than cold extremes on these continents.
Using the restricted sample to carry out the same analysis, we get similar results. North America and Antarctica have no issues with extreme hot temperatures, neither does Asia with this data. South America is still contradictory (with a low difference in means but a high-ish one in medians), the same goes now for Africa. It is really only Europe that clearly has considerably more recent hot records than cold records.
Let us not forget though that one reason for grouping countries into continents was to lower the standard deviations, which were quite high with the aggregated sample. Unfortunately, as can be seen in the table above, this cannot be achieved by continental grouping. Standard deviations remain high suggesting considerable variation within continents. The grouping would thus have to follow another pattern. To try to shed some light on this, let us look at data at a more micro level: U.S. states and European countries.
Using some more data from Wikipedia, I checked out the age of the records for U.S. states. The average hot record age is 23,290 days (median: 28,156), the average cold record age is 23,930 days (median: 26,143), thus the difference is 640 days (median: -2,013). To me this reinforces the fact that North America (or at least the U.S.) doesn’t seem to have an extreme hot temperature problem. The aggregate results for Europe can be found in the table above and as already discussed, as opposed to the U.S. data, it seems to indicate that Europe has been experiencing extreme hot weather more frequently than normal.
To visually check whether there is a regional trend, I tried mapping differences. In the following two pictures the colors represent the difference in the age of cold and hot records in each country/state, measured in thousands of days. Its cold minus hot, therefore a larger number indicates a more recent hot record compared to the respective country’s or state’s cold record.
There are no very clear trends, but one can see that in the U.S. the Midwest seems to experience more recent cold records. So inhabitants of those states see more extreme cold records nowadays. Most of the Southwest, the Southeast and the Northeast seem to experience somewhat more hot extremes than cold extremes.
The same color scheme is used for Europe, so one can compare the two maps directly. First, the map also indicates that Europe clearly has more of a warming problem than the U.S. does. Especially the Central and Eastern European belt going from Austria to Ukraine seem to be experiencing more hot extremes. Secondly, cold countries are almost nonexistent.
Clearly though we can conclude that extremely hot temperatures are becoming more common in the world on average, but this doesn’t apply equally to every region of the world. This may be partially due to random effects but the fact that new hot records are not common in North America for instance cannot be disputed. On the other hand, quite a lot of those states in the Midwest have recent cold records indicating perhaps that temperatures are becoming more extreme on both sides of the Atlantic, it’s just that the U.S. is experiencing more extreme cold temperatures on average.