Every claim in the report is arithmetically true. Every one of them misleads.
"Carbon dioxide emissions down 15.5% since 2016." The number is right: from 2016 to 2020, emissions fell from 1,031.5 to 872.0 million tonnes. The sentence does not say where the period ends, that emissions have risen 6.2% since 2020, or that from 2015 to 2022 the fall was 8.2%. A green number arrives inside a sentence built to be believed, and the choices behind it (which years, what was counted, whose average, compared with what) are rarely printed. When the data is published, anyone who can compute the number again can ask the questions the sentence left out.
Green on Paper teaches both halves of that work. It teaches Python from the first line, with pandas for tables and matplotlib for charts, on environmental data: greenhouse gas emissions and CO₂-equivalent, electricity capacity and generation, water, waste and air quality. And it teaches how to judge an environmental number, chart it and state it fairly. It assumes no programming at all.
Inside:
- Python from zero: values, names, lists, dictionaries, loops, functions, and how to read an error message
- pandas on real-shaped tables: reading CSV files, filtering and sorting, totals, averages and rates, grouping and reshaping, joins, dates and units, with each move shown again in a spreadsheet and in SQL
- Where environmental numbers come from: national greenhouse gas inventories, global warming potentials, capacity against generation, water withdrawn against water consumed, recycling rates, air quality stations and the WHO guidelines, company reports and claims of avoided emissions
- Checks before you believe: duplicate rows, a unit slip, numbers stored as text, missing rows, flags that contradict the data, stations that write dates their own way; then cleaning in code, with a record of every decision
- A scorecard that sets six claims of a green progress report beside what the checked data says, and ten checks to run before publishing an environmental number
- Charts that tell the truth with matplotlib, ten ways a true number makes a misleading green claim, the rules against misleading green claims in the EU, the UK and the US with the date of each source, and reporting frameworks in plain language: national inventories, the GHG Protocol scopes, corporate sustainability reporting
The examples use twelve tables about the Verdant Union, a fictional union of five countries with ten monitored cities. The Union, its cities, companies and projects do not exist, and no number in the tables describes a real place. The data was built to behave like real environmental statistics, faults included: 25 of them, planted on purpose for the reader to find. The book runs 528 blocks of Python on these tables, in one session from the first chapter to the last, each printed with what it returned, and 65 figures show what the code does and where the claims bend.
The quickest way to run the examples is the LucidPlexus Workbench, a notebook in your browser with the twelve tables already loaded and one task per chapter with an automatic check: nothing to install. Data: lucidplexus.com/community/dataset.php?slug=greenery, lucidplexus.com/community/dataset.php?slug=carboncoxial, lucidplexus.com/community/dataset.php?slug=renewflow. The same code runs in JupyterLab, VS Code or a script on your own computer, with Python 3.11 or later and pandas 3.
For sustainability officers, policy analysts, journalists, consultants and NGO staff who work with environmental numbers and want to stop taking them on trust; for students; and for analysts who want to learn Python on data that matters. No programming is needed to start, and no background in environmental science: you need to have used a spreadsheet and to be comfortable with percentages. Nothing in the book is legal, regulatory or investment advice.
Every purchase includes the companion download: the two-page cheat sheet, the workbook, the 20 Jupyter notebooks with a script that checks every block against the book, and the twelve data tables.
A LucidPlexus Field Book. First edition, 2026.