Early cherry blossom blooming is not evidence of ‘climate change’
This spring, cherry blossoms saw the record earliest first-blooming in various locations in Japan. There are many news reports, including one by The Washington Post and BBC
These articles tell concerns about the trend of first-blooming getting earlier and earlier, which they say is a sign of ‘climate change’.
The purpose of this project is to understand the statistical trend of first-blooming dates of cherry blossoms in Japan, as well as to explore some theories regarding first-blooming dates.
Exploratory data analysis questions
The following are three exploratory data analysis questions I’m going to explore in this project:
- Are cherry blossoms’ first-blooming date really getting earlier over time? Cherry blossoms may have bloomed the earliest ever, but it could have happened by random chance. To determine what is really happening, I’m going to investigate the general trend regarding first-blooming date over time.
For this question, I’m first going to compare the average first blooming dates over the decades, followed by geographical analysis to show in which locations cherry blossoms were blooming on April 1st over time.
I’m then comparing each year’s actual first-blooming dates to 30-year-average in various locations in Japan.
- What is the geographical distribution about whether 400- and 600-degree theories hold true? There are several theories to determine cherry blossom’s first-blooming date. One is the 400-degree theory, (the first-blooming happens when the cumulative daily average temperature since February 1st reaches 400 degrees). Another is the 600-degree theory (first-blooming occurs when the cumulative daily high temperature since February 1st reaches 600 degrees).
In the previous project, I examined these theories in six major locations in Japan and concluded these theories would hold true in some locations but not others. In this section, I’m going to conduct the same analysis for all observation locations and see if there is a geographical pattern of where these theories would be true.
Moreover, for places where these theories are found not to expect the actual first-blooming date, I’m going to calculate how many days before or after the expected date (based on the theories) first-blooming is likely to happen.
- Is there a significant correlation between difference in days and latitude? Based on the analysis on Question 2, I’d like to determine if there is a significant correlation between latitude and difference in days (between actual and expected first-blooming date based on 400- and 600-theories).
Data
The main data set I’m using is about cherry blossom’s first-bloom dates from 1953 to 2020 in more than 100 locations in Japan, published by Japan Meteorological Agency(JMA). I downloaded a version posted on Kaggle.com.
This data set includes cherry blossom’s first-blooming dates in more than 100 locations in Japan where the first-blooming were ever observed by JMA. For this project, I’m going to focus on 58 locations where observations are currently conducted.
I’m also using geographical data about JMA’s observation locations, available on the agency’s list of observatories.
Finally, to analyze the second and third questions, I’m using data about daily high and average temperature in more than 50 locations in Japan. Those data sets were directly downloaded from JMA, by using the agency’s downloading system.
For all tests in this project, I’m going to apply the alpha value of 0.05.
First, I’m going to load the data set I’m going to mainly analyze on this project.
Load first-blooming date data
The cleaned data set has 3944 rows and 4 columns. Each row shows an observation, and columns are as follows:
- location: the location the cherry blossom’s first-blooming was observed
- year: the year of observation
- date_x: date of first-blooming (Year-Month-Day)
- doy_x: date of first-blooming, converted as the number of days after January 1st
I’m computing the dates as number of days after January 1st. Negative numbers mean the first-blooming was before January 1st (late December). I count those cases as first-blooming in the coming spring (thus negative number of days after January 1st).
Load observation location data and conduct an initial analysis
First, I’m going to plot the data to see overall distribution.
As the visualization above shows, there are a group of observations that are plotted far from the majority of observations. In those locations, first-blooming dates of the year are below 50 days after January 1st. Where are those locations?
When we look at the data set, arranged by doy_x, we can see those extremely early first-blooming were observed in five sothern locations in Japan – Naze, Ishigaki Island, Miyakojima, Naha, and Minamidaitojima.
To confirm this, I’m going to visualize the relationship between latitude and average first-blooming dates per locations.
As shown in the visualization, as I suspected, those five locations are far from majority of data, thus can be considered as outliers.
Therefore, for this project, I’m going to filter out those five locations from all analyses.
So for this project, I’m going to conduct analyses over 53 locations in total.
This is taken from a long document with a lot more images. See it here yuriko-schumacher.github
Header image: Japan Guide
Please Donate Below To Support Our Ongoing Work To Defend The Scientific Method
PRINCIPIA SCIENTIFIC INTERNATIONAL, legally registered in the UK as a company 
incorporated for charitable purposes. Head Office: 27 Old Gloucester Street, London WC1N 3AX.
Trackback from your site.
Tom
| #
Fake experts don’t need any proof whatsoever. All they require is a passel of brain dead idiots to believe everything they say.
Reply
Joseph Olson
| #
The Japanese were introduced to the Chinese logographic system of writing in 206 BC and adopted the Chinese calendar in the 12th century, after adjusting to their own astronomical observations. They kept accurate cherry blossom dates that show earlier blooms during the Medieval Warming and later blooming during the Little Ice Age Cooling periods. They made 800 years of observations without the benefit of a Mann hockey stick….
Reply