This page describes an experimental method and tool produced by Climate Policy Radar. We’re aware of several areas for improvement, and we intend to continue its development. Feedback is always appreciated! Please get in touch.
Finding targets in climate policy documents is critical step in identifying gaps, trends, and opportunities to accelerate climate action.
Identifying these targets typically involves a researcher sifting through thousands of documents by hand. As a result, the scale of this research is rarely comprehensive.
We’ve built a text-classifier to automatically extract targets from our database of national laws, policies, and UN submissions. Using our tool, we can now find and compare 10,000s of global climate targets in minutes, rather than weeks.
Based on a new dataset of automatically extracted targets, you can explore different target types, deadlines, sources, and geographies, against the year in which the targets were set.
We hope that this work enables users to explore the breadth and depth of climate-related national targets in greater detail, and to identify gaps, trends, and opportunities at scale.
The core of our method is the creation of a dataset, high-quality, expert-annotated sample of national climate change laws and policies and UNFCCC submissions from climate-laws.org, published by public institutions only (e.g. government agencies and departments). This data is then used to train a classifier, which aims to accurately predict whether a passage of text contains a target or not.
The class definitions for hand-labelling are based on those in the **ClimateBERT-NetZero study methodology,** itself based on the Net Zero Tracker codebook. This builds on existing work by Net Zero Tracker and ClimateBERT to identify ‘Net-zero’ and ‘Emissions Reduction’ targets, and extended to identify ‘Other’ quantified targets made by national governments in climate policy documents.
We also expanded the Net-Zero and Emissions Reduction targets definitions to include different greenhouse gases (rather than just general greenhouse gas targets) and to include sector-specific targets (such as emissions reduction targets for the transport sector).
The complete definitions are given in full in the following section.
An aim to achieve something, rather than stating something concrete about the future. **Often this means that the phrase indicates a level of uncertainty.
Quantifiable; it contains a reference to a measurable quantitative value. This may be numeric or non-numeric. For example, words like all, every, double, halve, eradicate, no, none and independent of refer to measurable quantities. IMPORTANT: Quantifiable targets differ from verifiable statements of intent (see negative cases below). Classifying verifiable statements as targets would risk creating a model that classified any statement about a future action by a policymaking body as a target. For more information, see CCLW methodology and methodology section of GRI’s 2018 policy brief