In this tutorial, you have the option to explore up to 3 different visualization libraries in Python, with the task of generating an interactive choropleth map of temperature data.
This activity pairs nicely with time series analysis of temperature data, since this kind of data can easily be plotted on a map (assuming you can aggregate data on a country name level).
This repository contains step-by-step, fill-in-the-blank instructions for reading, manipulating, and generating geographic data in GeoPandas, and visualizing this data with the library of your choice:
- Folium
- Plotly
- Bokeh
Of course, you can try all 3 of them. The goal of these tutorials is not so much to learn the syntax of these -- quite complex -- libraries (that is why a lot of boilerplate code is mostly already set up for you), but rather to learn some concepts behind interactive, browser-based visualization libraries in Python (as opposed to "static" libraries like matplotlib or seaborn) and to get a taste of what is possible with these libraries. Plus, knowing more advanced data visualization tools generally makes you a better Data Scientist.
Here is a quick, high-level overview of the 3 libraries found in this repo that you can use for plotting nice interactive choropleth maps:
| Library | Difficulty | Has Web Server / Back-End? | Front-End Framework | Features / Comments |
|---|---|---|---|---|
| Folium | Easy | No | Leaflet.js | Easy-to-use. Maps only. |
| Plotly | Medium | No | Plotly.js | Comprehensive vizualization library; includes many types of graphs and charts |
| Plotly/Dash | Hard | Yes (Flask) | React.js | Powerful framework for building entire web applications (back- and front-end). Built on top of Plotly. |
| Bokeh | Hard | Yes (Tornado) | BokehJS | Powerful interactive visualization library (also back- and front-end). Older / more established. |
And what about GeoPandas?
GeoPandas is not an interactive visualization library (though it can be used to plot simple, static maps); rather, it's a powerful library used for working with geospatial data in Python.
Each of the 3 tutorials begin with the exact same steps: using GeoPandas to read in a shapefile and convert it to a so-called "GeoDataFrame" (the library is built on top of Pandas) so that we can perform all the convenient data wrangling operations we're already used to, such as filtering, grouping, and merging.
The fundamental difference between a regular Pandas DataFrame and a specialized GeoDataFrame, however, is that the the latter understands geometric data types such as Points, Lines, and Polygons, and can easily perform geometric / spatial operations on them.
And what's really nice for us, is once the GeoDataFrame is in the shape and format we want, we can easily export it to a GeoJSON file -- a data format that most interactive, JavaScript-based visualization libraries (like the 3 above) require for making maps.
Using GeoPandas (for pre-processing) and one of the 3 visualization libraries (for making maps) is a powerful combination that provides a nice introduction into the field of GIS, without having to invest a lot of time in using full-blown solutions like QGIS or any of the proprietary ESRI products.
Clone this repository:
git clone https://github.com/pawlodkowski/interactive_climate_map.gitPick one (or more) of the libraries you'd like to try out, and then navigate into its respective sub-directory:
Note that the the content for Plotly/Dash is also included in plotly/ directory. However, the tutorial for it is not yet written; though I've included some helpful resources on how to get started.
Once you've navigated to the appropriate directory, open up the corresponding Jupyter Notebook
(.ipynb) file. Follow the step-by-step tutorial with the notebook, which includes:
- Installation Instructions
- Data Wrangling with GeoPandas (Steps 1-4)
- This is done as an in-class exercise. Also, these steps are identical for each of the tutorials.
- Making the Interactive Choropleth Map (Steps 5-end)
- This is done as a self-guided activity after the lesson (though a teacher will of course be available in the classroom for debugging support!)
The data for this particular lesson is synthetic / fake! It was originally scraped from Berkeley Earth, but then regenerated based a simple polynomial trend and monthly seasonality modeled from the original data source & injected with random noise. Hence, the data in this repo is somewhat realistic but not authentic.
- This was done in the interest of protecting the original data source / preventing its redistribution.
Most of the code here was either written or adapted (from outside sources, e.g. library documentation, published tutorials) by Paul Wlodkowski, Data Science Instructor @ Spiced Academy in Berlin.
Some parts of the tutorials were based off code snippets made by fellow colleagues at Spiced Academy. More detailed credits can be found in the respective Jupyter Notebooks.
Work in this repository is published under the conditions of the MIT license.