Can the Temperature of Summer be Predicted?
I recently read an article written by Julien Emery titled; “Data Science for weather forecast: how to prove a funny theory.” Below is a link the article:
In this article, Julien described a data analysis project where he and his partner; Aouss Sbai analyzed weather patterns in five cities in California. They investigate a previous claim made by their mentor; Palmen Nedeltchev, that based off three days in May (19th, 20th, 21st) he was able to predict if the summer was going to be hot or not. To put it simply, they were trying to predict the weather for three months off of three days. It was never stated why the 19th-21st of May was chosen, but the claim is explained in a way that it could be any three days, they just chose the same three as their mentor. They scraped all of their temperature data from “the old Farmer’s Almanac” which provided them with all of the measurements that would be needed for the analysis. They also claimed that this was the only source of data that they found that would be able to provide historical data. This is the first claim that they made that I want to address because initially I thought that that was hard to believe, and after research I discovered that there are plenty of other sites that provide data such as this, for example; Weather Underground, Weather.gov, and Climate.gov just to name a few. After scraping the “Farmer’s Almanac” site, they began their analysis on the data. They chose to average the three days in May because the measurements were “very similar.” This is the second part of the article that I would like to address because I feel that by doing this they almost discredited their entire project. I say this because it is important when conducting data analysis that you are as accurate as possible, and without the other temperatures, we are left with the averages that they claim are “similar.” They then run the analysis and create visuals that show a scatterplot with the Summer temperature and the May averaged temperature. (see below for example):
They found that there is a positive relationship between the two variables, and that means as the temperature of the three days increases, so does the temperature of the Summer. From this point, they summarize their final points, and state that the only city that they found with a significant relationship was San Jose.
In order to asses the claim that the Summer will be hot or not based on three days in May, I believe that it is important to first look at the strength of the effect. In this case, there is a lot of weight on this effect because you are essentially taking three days and using them to predict around ninety-plus days. As far as the effect being replicated, this can be observed, but the days in May do not have a direct effect on the whole summer. Another issue with this relationship is that there is no exact reasoning that can be pinpointed for the days in May predicting future temperatures. Something that does support this claim is that the alleged cause comes before the effect. An analysis on the different days in May would be interesting to see because we would have a stronger understanding of if there is a relationship with May predicting Summer temperature. Another value that could be added to this analysis would be to see if the temperatures in May directly affected the Summer, for example if the three days were higher than a pervious year was the summer hotter? As far as plausibility, I do think there is something that could be said about May temperature helping predict summer temperature, but the question is how accurate it would be. There are also many other factors on any given day that can affect weather and temperature, so it would be possible to explore other ways to predict or analyze the summer temperature. If the project is to remain specific, I do not believe there is another way to interpret this data besides temperature history. This project highlights predicting future temperature which is something that is done everyday by weather reporters, so it would be interesting to see what that community thinks of this and any other opinions about temperature and its relationship with future temperature.
