How to Make a Choropleth Map of Japan: Colouring Prefecture Data and Choosing How to Classify It
A choropleth map shows values for each prefecture, such as population density, average temperature, sales or survey response rates, as shades of colour on a map. It conveys regional trends at a glance far better than a table, but choose the wrong data or classification and it can give an impression that doesn't match the facts. This article explains how to make a choropleth map by pasting data from Excel, compares three classification methods (natural breaks (Jenks), quantiles and equal intervals) using real numbers, and covers how to decide the number of classes, colours and legend.
What is a choropleth map?
A choropleth map divides a value for each area into several classes (steps) and shades each class a different intensity of colour. In Japanese it's called a kaikyu kubunzu ("class division map"). It's often used to present statistics by prefecture, and it's a familiar sight in social studies textbooks, government white papers and news graphics.
There are four main decisions when making one: (1) which data to use, (2) how many classes, (3) where to draw the breaks and (4) which colours. The breaks in (3) in particular can make the same data look very different on the map, so they're compared in detail in the second half of this article.
Steps for making a choropleth map from prefecture data
- 1
Prepare the data
In Excel or a spreadsheet, put a column of prefecture names next to a column of values. Taking statistics from official sources such as e-Stat (the portal site for Japanese government statistics) lets you state the source clearly.
- 2
Open "Color by Value"
Switch the mode of the Visited Prefectures Map to "Color by Value (Choropleth)".
- 3
Paste prefecture names and values
Select and copy the two columns in Excel, and paste them into the "Paste prefecture names and values" box. Use one prefecture per line; tabs, commas or spaces all work as separators. Either "Tokyo" or "Tokyo-to" style names are fine, and it's OK if a prefecture code column comes first.
- 4
Choose the classification, number of classes and colours
Choose the classification method (natural breaks, quantiles or equal intervals), the number of classes (2 to 7) and the colours (red, blue, green, purple, orange or blue ↔ red). The defaults are natural breaks, 5 classes and red.
- 5
Add a title and source, then save
Enter a title and "Source & unit" (for example, "Unit: people (2020 Census)") and save as PNG or SVG. The legend and a table sorted from largest to smallest value are also shown, so check them before saving.
- Thousands separators (1,234,567) and full-width digits are read as they are. Negative values can be written with "-", "▲" or "△".
- If the same prefecture appears twice, the value in the later row is used, and the screen tells you so.
- Prefectures without a value are coloured grey as "No data".
Use rates or "per …" figures, not totals
The most common mistake with choropleth maps is colouring totals such as population or number of cases directly. A choropleth map fills a prefecture's whole area with one colour, so prefectures with large populations end up with large values for everything, and the map simply shows that "populous prefectures are dark".
| What you want to show | Values that mislead if coloured directly | Values suited to colouring |
|---|---|---|
| Where people live | Population | Population density (per km²) |
| How many hospitals there are | Number of hospitals | Hospitals per 100,000 people |
| How likely accidents are | Number of accidents | Accidents per 100,000 people |
| Sales growth | Sales amount | Year-on-year change (%) |
| Response trends | Number who answered "yes" | Share answering "yes" (%) |
When you want to show totals, writing the numbers on the map or using circles of different sizes for each prefecture works better. If you really must colour totals in a choropleth map, make it clear in the title that they're totals, as in "Population (total)".
Natural breaks (Jenks), quantiles and equal intervals compared
| Classification | How breaks are set | Good for | Weak at |
|---|---|---|---|
| Natural breaks (Jenks) | Breaks at the gaps between clusters of values | Clearly showing groups of prefectures that differ | Comparing across years or sources with the same breaks |
| Quantiles | Breaks so that every class has about the same number of prefectures | Showing rankings such as top and bottom | Small differences in value can look like big differences in colour |
| Equal intervals | Divides the range from minimum to maximum into equal widths | Showing the spread of values itself, or getting round-number breaks | Data with an extreme value (outlier) |
Natural breaks (Jenks natural breaks) calculates the breaks so that the variation of values within each class is as small as possible and the differences between classes are as large as possible. It produces breaks close to where a person looking at the data would feel "they split around here". The Visited Prefectures Map solves this calculation exactly using Fisher's dynamic programming method.
Quantiles sorts the values from smallest to largest and divides them evenly by number of prefectures. With 47 prefectures in 4 classes, each colour gets roughly 12 prefectures. Equal intervals, the simplest method, divides the range from minimum to maximum into equal widths.
Comparing classifications on the same data
To get a feel for the differences, let's compare the classifications on two examples. The first is the number of prefectures each prefecture borders on land (0 to 8). The values are integers, and 18 prefectures have exactly 4 neighbours, so the data has many identical values.
| Classification (3 classes requested) | Breaks | Prefectures per class |
|---|---|---|
| Natural breaks | 0–2 / 3–5 / 6–8 | 8 / 33 / 6 |
| Quantiles | 0–3 / 4 / 5–8 | 20 / 18 / 9 |
| Equal intervals | 0–2.67 / 3–5.33 / 6–8 | 8 / 33 / 6 |
Quantiles tries to even out the counts, but because 18 prefectures share the same value of 4, it can't split into exact thirds. And even if you request 4 or 5 classes with quantiles, identical values go into the same class, so in this example it collapses to 3 classes. That's why data with many identical values can end up with fewer classes than requested.
The second is an example with one extreme value. Dividing fictional data into 4 classes, with 24 values scattered between 10 and 120 and a single value of 1,400 (a situation like only Tokyo having sales an order of magnitude higher), gives the following.
| Classification (4 classes) | Breaks | Values per class |
|---|---|---|
| Natural breaks | 10–35 / 38–70 / 76–120 / 1,400 | 10 / 9 / 5 / 1 |
| Quantiles | 10–25 / 28–45 / 48–70 / 76–1,400 | 7 / 6 / 6 / 6 |
| Equal intervals | 10–357.5 / … / 1,400 | 24 / 0 / 0 / 1 |
With equal intervals, 1,400 pulls the width of each class to about 350: all 24 values get the lightest colour, only 1,400 gets the darkest, and not a single prefecture falls into the two middle colours. The map becomes almost one colour, and the difference between 10 and 120 disappears. Natural breaks puts 1,400 in a class of its own and splits the rest into three groups, so differences other than the outlier remain visible. Quantiles evens out the count in each colour, but 76 and 1,400 get the same colour, so the size of the outlier doesn't come across.
Deciding the number of classes and colours
- Aim for 4 or 5 classes: people can only distinguish about 5 to 7 shades of a colour. For 47 prefectures, 4 or 5 classes give each colour enough prefectures and make comparison with the legend easy.
- For larger and smaller values, use shades of one colour: the basic approach is one colour such as red, blue or green, lighter for smaller values and darker for larger ones.
- For increases and decreases, or plus and minus, use two colours: for data whose meaning flips either side of 0, such as year-on-year change, use "Blue ↔ Red (change)", with decreases in blue, increases in red and a near-white colour in the middle.
- To highlight the small end, reverse the colours: turn on "Reverse colors (darker for smaller values)" to emphasise data where the small values are the point of interest, such as low unemployment or short waiting times.
- Colour for prefectures without data: prefectures without a value are grey (No data). Check before saving that it can be told apart from the lightest class colour.
Don't forget the legend and source
Without a legend, nobody can tell what the colours of a choropleth map mean. The Visited Prefectures Map includes each class's range (lower to upper bound) and number of prefectures in the image as a legend. On top of that, writing the following three items in the "Source & unit" box makes it a map people can trust as a document.
- Unit: people, %, people/km², yen and so on. This also shows whether it's a rate or a total.
- Date: when the data is from, such as the 2020 Census or fiscal 2025.
- Source: the name of the statistics or the organisation that published them. For data you collected yourself, write "Survey by ○○".
To colour by groups such as "available / coming soon" rather than values, see How to colour a blank map of Japan by prefecture, and for the list of neighbouring prefecture counts used in the example, see Number of neighbouring prefectures for each prefecture.
FAQ
- Can I make a choropleth map for free?
- Yes. Just paste two columns of prefecture names and values from Excel or similar into "Color by Value" in the Visited Prefectures Map, and you get a choropleth map with the classification, number of classes and colours you choose, which you can save as PNG or SVG with a legend. No sign-up is needed.
- What are natural breaks (Jenks)?
- A method that calculates breaks so that the variation within each class is small and the differences between classes are large. Because it breaks at the gaps between clusters of values, groups of prefectures that differ stand out clearly.
- When should I use equal intervals and when quantiles?
- Equal intervals suits showing the spread of values itself, or when you want round-number breaks; quantiles suits showing rankings such as top and bottom. Be careful when there's an extreme value: with equal intervals, almost every prefecture ends up the same colour.
- I set 5 classes but only 4 appear
- When many prefectures share the same value, identical values go into the same class, so you can end up with fewer classes than requested. For example, classifying the number of neighbouring prefectures (integers 0 to 8) into 5 quantile classes gives 3 classes. Switching to natural breaks or equal intervals may give more classes.
- Can I colour population directly?
- For choropleth maps, population density, figures per 100,000 people or percentages work better than totals such as population or case counts. Colouring totals just makes populous prefectures dark, and the trend you want to show doesn't come across.
Tools in this article
Related articles
How to Colour a Blank Map of Japan by Prefecture: for Lessons, Reports and Service Area Maps
Class handouts, school projects, the service areas in a company report, a travel list of prefectures you want to visit: there are surprisingly many occasions when you want a blank map of Japan coloured by prefecture. But colouring on paper rarely looks neat, and in Excel or PowerPoint even preparing the map shapes takes effort. This article compares ways to colour a blank map by prefecture, walks through making a colour-coded map with a free tool and exporting it as an image, and explains how to choose colours and a legend so the map gets its message across.
Read article →Rice-to-Water Ratio Chart | How Much Water for 1 to 5 Cups (Gō) of Rice in a Rice Cooker, Clay Pot, Saucepan or Pressure Cooker
“How many millilitres of water for 3 cups of rice?” “Does a clay pot need more than a rice cooker?” “Wash-free rice needs a little more and new-crop rice a little less, but how much?” When the lines inside the rice cooker bowl are worn off, when you cook in a pot, or when you switch to a different kind of rice, the right amount of water suddenly becomes hard to judge. For white rice, the guideline is 1.1 times the volume of the washed, drained rice in a rice cooker, 1.2 times in a clay pot (donabe) or saucepan, and 1.0 times in a pressure cooker. This guide explains why those ratios work and gives a chart by amount and cooking method, how to adjust for wash-free rice, new-crop rice, brown rice and glutinous rice, how to make rice firmer or softer, soaking times by season, heat settings and timings for a clay pot, saucepan and pressure cooker, the water for sushi rice, seasoned mixed rice and rice with grains, how much rice to cook for a given number of people, and what went wrong when a batch fails, all in concrete millilitres.
Read article →How Strong Is a 10, 15 or 20 m/s Wind? What Each Wind Speed Does to Umbrellas, Walking, Bikes, Cars, Trains and Typhoons, With km/h and Knot Conversions
A forecast of “maximum wind 15 m/s” does not immediately tell you how strong that wind feels. A 10 m/s wind is 36 km/h and already too strong for an umbrella. At 15 m/s you can no longer walk into the wind, at 20 m/s you cannot stay on your feet without holding on to something, and above 30 m/s trees and utility poles start to come down. This guide explains how to read the wind speed in a forecast (mean wind versus gusts, and the height it is measured at), gives a chart of what happens to people, trees, cars and buildings from 1 to 40 m/s, describes 5, 10, 15, 20 and 30 m/s in detail, and gives guidelines for umbrellas, laundry, cycling, camping, drones, work at height, trains and expressways. It also covers the force on a sign or umbrella in kgf, wind chill, how much stronger the wind is on a high apartment floor, typhoon intensity classes and hurricane categories, converting between m/s, km/h, knots and mph with mental-math shortcuts, and what to do before strong winds or a typhoon, all with concrete numbers.
Read article →