Across the United States, Summer 2012 so far has seen some of the hottest temperatures in history. Record-breaking highs in the Midwest, the South, and particularly on the East Coast in June and July have shattered documented averages, and many have begun to cite these temperatures as evidence for global climate change.
At Topsy, we wondered: to what extent could such increased temperature be reflected in the way people discuss weather on Twitter?
We searched historical Twitter records for the consecutive terms “hot outside” in tweets from the United States throughout June and July 2012. Below, we can observe distinct spikes in the number of mentions during the most severe periods of the heat wave, on June 20 and June 28-30:
To delve deeper into these trends, we compared mentions of “hot outside” in tweets from the United States to deviations from historically-normal temperatures. Since the heat wave was particularly severe in the mid-Atlantic region, we gathered weather data from Washington, DC. Using publicly available temperature records by city (here), we subtracted the average maximum temperature in previous years from the temperature observed in 2012. This deviation in temperature (in degrees Fahrenheit) is displayed on the x-axis below, and mentions of the heat on Twitter are displayed on the y-axis. A significant positive correlation between the variables was observed (Pearson’s R = .65, p < .001).
Note: Extreme outliers during severe heatwaves disrupted the flow of normal weather discussion, and were thus removed from the correlation analysis.
So, discussion on Twitter can indeed reflect aspects of the real world. Notably, however, this analysis presents how tweets can even exhibit a rather external measurement such as temperature. The fact that deviations from average temperature are positively correlated with Twitter mentions implies that perhaps the Web can also present human interpretation of these external occurrences: many people simultaneously observing that the temperature is rather different than it is normally could be considered a type of collective intelligence, and this too can be quantified. All of the above results indicate that the public conversation held daily on the social Web is a powerful tool for prediction and explanation of the world we live in.