By Anthony Watts
The headline of a recent story on CNBC claimed, “Scientists Are Using Twitter to Measure the Impact of Climate Change.”
I did a double-take and checked the calendar to make sure this was not April Fools’ Day, thinking this had to be some sort of a joke.
Sadly, it is not.
Incredibly, scientists are basing claims of a climate crisis on the number of people tweeting about climate events—a very bad sign for science, indeed.
The CNBC story featured a newly published study titled, “Using Remarkability to Define Coastal Flooding Thresholds.” (“Remarkability” is a fancy, sciencey-sounding name for Twitter volume.) A pair of scientists from the University of California at Davis and the Max Plank Institute for Human Development examined Twitter messages to measure how often people complained about flooding nuisances—typically caused by backed-up stormwater drains—along coastal counties, including Boston, Miami, and New York.
“Coastal floods and inundation are projected to produce some of the primary social impacts of climate change, imposing significant costs on communities around the world,” the study claims.
“Flooding due to high tides, storm surges, or a combination of the two is increasingly common in many coastal areas and is projected to become more frequent and severe as sea-levels rise globally.”
However, the study ignored hard, objective data like rainfall rates, choosing instead to build a scientific case for worsening coastal flooding by noting that people are tweeting about it more often. The researchers defined a “remarkable threshold” for coastal flooding when the number of Twitter posts in a particular county complaining about flooding rose by 25 percent. Then, they compared the Twitter data with official flood records.
The kinds of Tweets that would qualify as scientific evidence of increasing, climate-driven flooding would include, “Hey neighbors! The street is flooded again because the city didn’t clear the storm drain of junk and leaves. Don’t park out front.”
The study reveals trends of social media commentary, but certainly not objective, factual data about climate. It also reflects trends of social media volume in general, as well as people reflecting the inundation of climate propaganda coming from media sources. None of these are scientific evidence of climate change or climate change impacts.
Here is another interesting tidbit: For some strange reason, the researchers limited the scope of their study to a relatively short period, ranging from March 2014 to November 2016. I’m always suspicious of any scientific study that doesn’t use the entire available dataset. Why not from 2014 to 2018? In many cases, analysts limit their choice of data because when they analyze data for a study and the full dataset does not provide the answer they were hoping to find, they report misleading results from a partial dataset instead.
To their credit, the researchers noted that Twitter data might be misleading. They mentioned earlier research demonstrated that the more people experience things, the less remarkable they become. In other words, when storms and floods occur less often, they are more likely to be exciting and deserving of a Twitter post when they finally do occur.
Here is the biggest flaw in the study: Nowhere in the study did the authors look at the increase of Twitter users or tweets during the same period, and that’s a shocking oversight on their part. According to data for the United States compiled by Statista, Twitter’s audience grew massively from the first quarter of 2013 to the fourth quarter of 2014, from 48 million to 63 million monthly active users. This 31.25 percent increase in the number of Twitter users overlapped the period studied in the previously mentioned (and dubious) flooding study.
Gosh, do you think there might have been an increase in tweets about street flooding because more people were using Twitter during the months at the end of the study period than were using Twitter at the beginning of the study period?
I weep for science, and I especially weep for climate science.
Anthony Watts (AWatts@heartland.org) is senior fellow at The Heartland Institute. He is a former broadcast meteorologist and operates the world’s most-viewed climate website, WattsUpWithThat.com.
Discover more from Watts Up With That?
Subscribe to get the latest posts sent to your email.
How do the Climate Change Bots figure into all of this?
Classic feedback loop.
Mark Steyn said about social media (from memory, paraphrasing Churchill):
Or, to paraphrase Dr. King:
Indeed.
In today’s world MLK is a racist
Yes, Martin Luther King wanted everyone to get along with each other, and judge each other only on the content of a person’s character.
Today’s modern Radical Democrats do not want everyone to get along, they want to divide groups of people in the hopes that will allow the Democrats to gain political power. There is no “Martin Luther King thought process” going on among these people. Martin Luther King was *not* full of hate. The Radical Democrats *are* full of hate. It’s what energizes them.
Postmodern science.
We’ll be a laughing stock in the future. The dark ages of science.
Flood tweets and tide height
Reminds me of the old one about increases in radio licences correlating with increases in cases of insanity.
Why are we surprised? That is how social ‘sciences’ work. Feelings not facts.
“facts don’t care about your feelings” – Ben Shapiro
Twitter is the greatest testament to the human intellect in cultural history. It is an extraordinary marketplace of ideas, where fine minds come together to exchange carefully considered opinions and to uplift the hearts of people everywhere. Twitter is humanity at its best, a wonderful carnival of brilliance, brains and wit.
[/sarc off]
No one used the word “twit” in this discussion yet.
This general approach is great -in some cases, we can learn a lot more than we can, otherwise.
I was at some research training many years ago – now approaching 20 years. This was a two-week, competitive deal, with expert to trainee ratio of about 1:3. A very nice deal. All of us trainees got to know each other, and our topics. One trainee said she was with a health department for a major city, and they were using 9-11 calls to quickly identify quick-onset health problems such as flu outbreak. SARS had set them on this path.
This was before the hype of “Big Data.” I think this was the advent of Big Data. Big Data has a few definitions. One is our ability to take big data sets that exist for one purpose, such as twitter, and apply them to some other worthwhile purpose, such as estimating where and how much flooding is going on, at the moment, in some locale. This view of Big Data is the view that these massive data sets exist, and we now have computational capacity to discern patterns and relationships, so let’s forge ahead and see where it works.
You could look at a local spike in food delivery to map power outage – and so on. There are grand ideas that we may be able to use a variety of data sources to get at predictors of cancer. Our longitudinal cohort studies have weaknesses / limits, and Big Data studies can be complementary.
I helped a team examine such out-of-the-box ideas for cancer prediction. A predictor of cancer incidence was: growing up with a grandparent as head of your household. a single item out of many population-based questionnaires on various topics, wedded regionally to cancer incidences. Cancer incidences were greater where general-population surveys had greater portions of people saying they were raised by grandparents versus parents. What does this mean? We don’t know. Maybe these are lower SES people? Were our methods OK for avoiding chance findings? Did we winnow down variables the right way? SEM? Discriminant function analysis? Backwards-stepping? Frontwards? Or other methods, altogether? This effort was really provoking thought versus genuinely being intended to cure cancer incidence. Compelling, and earnest. A far reach. Yes. Confirmatory? No.
Of course, all of the well-known principles of measurement apply. We want measures that are specific and sensitive. that have convergent and discriminant validity. Candidate measures should be tested by replication to deal with capitalization on chance, regression to the mean, and biases such as experimenter effects.
With some Big Data applications, generalizability or robustness concepts do not apply; we will only see the advent of Coronavirus, or Ebola in the United States, once in our lives. After that, the context will change, and increases in tweets about Coronavirus will not have the same meaning.
The Climate Fear Mongers have no shame about co-opting recognized research methods, and even the mantel of “science” to advance their politics. We should only expect more of this misguided, politically-motivated “science.”
Consensus and conjecture are the new “Science”. How dare you disagree.
May God help us!
If a creek flooded in a remote forest and no one tweeted about it, did it still flood? Uh, yes. But it wouldn’t show up in a study on Twitter.
Of course, on the average dry day with no storms in sight, nobody will be tweeting about the weather, but they might tweet about what they or their friends did that day. So the Twitter researchers will conclude that there’s never any dry weather in this country.
Trying to calculate flood frequency from tweets about floods is similar to calculating trends in hurricane frequency since before the 1970’s. Back before the age of weather satellites, people only knew about a hurricane if it hit land in a populated area, or if a large ship crossed its path without sinking. Now, with weather satellites keeping an eye on all the oceans, we give names to all storms with winds stronger than 35 mph, regardless of whether they ever hit land, or fizzle out in the middle of the ocean. This leads to more reported storms, but not an increase in frequency over time, since we didn’t know about many storms before the advent of weather satellites.
This is how we get our data. We don’t have to go far – take the 97% claim. Cook pulled this one out of thin air. He simply looked at press articles and how many times Climate Change was mentioned and discarded those that did not as not relevant. Automatically blowing the share that suggests that human-made Climate Change is real out of proportion. After redoing the numbers seriously, the real number was low single digits. Food for thought?