This update is based on a dataset change that shifts the observation period by 15 years. Buried within a footnote that nobody will ever read:
> The 2012 USDA hardiness zones were calculated using the average lowest winter temperature for the observation period of 1976-2005. The new zones are calculated using the years 1991-2020. These two observation windows overlap. Colors show the difference between the two 30-year averages for each place on the map.
So while climate change may or may not be the explanation for broader trends in the map, you cannot rule out noise for any particular location, especially given that the methodology uses the lowest winter temperatures in that window as the basis for the zone. That's a fundamentally noisy metric. For example, this drops the blizzard of 1978 -- one of the largest ever in US history -- from the dataset. It's hard to know how that affects the zones without careful inspection, but major weather events like that can easily affect simple outlier analysis. You can see this in the dot plots for Raleigh (the default), which had two outliers in the winters around 1995, and then an otherwise flat series of data. What was the prior dataset? We don't know. Just look at the two default examples:
It would only take a small number of cold or warm winters to shift those means in either direction.
I don't want to overemphasize a silly little article about plants, but IMO, the fact that NPR buried this information while hyper-focusing on specific locations is just another failure of basic journalism. Telling people that an arbitrarily defined metric has been changed is meaningless, unless you tell people the specific change that was made. It should be the first or second thing they tell you, not buried in a hidden footnote.
And by "buried" you mean "dedicated a whole screen to it at legible size"? I don't wanna go back and count, but as I was scrolling through the sequence on mobile, it's definitely in there, up front and clear. Maybe you had a different experience of the website?
I appreciate data skepticism, but your point about the 1978 blizzard doesn't really help, as blizzards and snow events in general are not generally associated with extreme cold events.
Also, the site tells you about the changes in whatever place you plugged in. I'm not clear on why you're calling this "hyper-focusing on specific locations". It's literally telling you what's going on in your neighborhood and gives good reasons to even then discount the new number.
By "buried", I mean: "hidden in a collapsed 'about this data' accordion, below the 'change in lowest winter temperature' color bar, in the fourth text box that appears on downward scrolling."
Unless you're seeing a wildly different website than I am, this is a strange definition of "up front and clear".
But just to belabor the point a bit, if I were a journalist covering this, my lede -- the very first paragraph of the story -- would be something like this:
"Recently, the USDA updated its plant hardiness map for the first time in 11 years. Among other changes, it based the definition of winter temperatures on data from 1991-2020, dropping data from 1976-1991 as part of the update. This resulted in a calculated increase in lowest winter temperatures in many locations across the United States."
Simple. Factual. Unbiased. Provides necessary context. It isn't that hard to be a good journalist here. I hesitate to speculate, but my suspicion is that this was buried because it weakens a narrative that the writers and editors were trying to achieve, by adding nuance.
I don't think you understand how much gardeners, the nursery industry, and farmers rely on this data. We know things aren't the way they used to be. That's why you're starting to see so many "green thumb" people getting more vocal about climate change. They have first-hand experience with what's going on that isn't easy to dismiss with a hand-wave.
I know HN loves pedantry, but this criticism is overblown to the point of being misleading.
Look at the Raleigh graph - there are a few outliers, but the average (the one number being reported) is right in that cluster of flat years. And that's what's apparently gone up 5°
There is always going to be information lost when you condense 30 years into 1 number, but you are "pushing a narrative" far more than they are
The point was that you don't know what data they dropped by shifting the data 15 years forward, and that it only takes a few such "outliers" to shift the mean in an otherwise flat dataset. Moreover, you wouldn't know how flat the trend in the data actually is from the article itself.
For Raleigh (new zone high 8a, almost 8b) there are three winters in the low end of the "previous" zone of 7b in the last 10 years. For St Louis (new zone 7a) there were two winters in zone 6b and one in 5b (!) in the time since 2013. There were an equal number of winters in the "old zone" of 6b and the "new zone" of 6a in the dataset shown. Even calling that an outlier is stretching the definition of the term! And I didn't cherry-pick these examples...these were the ones NPR chose to highlight.
This isn't "pedantic". It's a basic understanding of statistics and the methods being applied here.
> The 2012 USDA hardiness zones were calculated using the average lowest winter temperature for the observation period of 1976-2005. The new zones are calculated using the years 1991-2020. These two observation windows overlap. Colors show the difference between the two 30-year averages for each place on the map.
So while climate change may or may not be the explanation for broader trends in the map, you cannot rule out noise for any particular location, especially given that the methodology uses the lowest winter temperatures in that window as the basis for the zone. That's a fundamentally noisy metric. For example, this drops the blizzard of 1978 -- one of the largest ever in US history -- from the dataset. It's hard to know how that affects the zones without careful inspection, but major weather events like that can easily affect simple outlier analysis. You can see this in the dot plots for Raleigh (the default), which had two outliers in the winters around 1995, and then an otherwise flat series of data. What was the prior dataset? We don't know. Just look at the two default examples:
https://imgur.com/a/oZjQ36Y
It would only take a small number of cold or warm winters to shift those means in either direction.
I don't want to overemphasize a silly little article about plants, but IMO, the fact that NPR buried this information while hyper-focusing on specific locations is just another failure of basic journalism. Telling people that an arbitrarily defined metric has been changed is meaningless, unless you tell people the specific change that was made. It should be the first or second thing they tell you, not buried in a hidden footnote.