Central-government climate targets play a directive role in China, cascading through the economy via the country’s target-responsibility system. To support evidence-based analysis of this evolving target landscape, the China Climate Target Tracker compiles national-level quantified climate and climate-relevant targets since 2004, systematically extracted from official policy documents and government website entries. The dataset adopts a broad definition of “climate targets”: quantified commitments that directly reduce GHG emissions, contribute indirectly to emissions reduction, or significantly influence emissions pathways.

The tracker is updated on a quarterly basis.Latest data update: June 2026.

Chinese policy documents vary in how firmly they formulate climate targets. More than 85 percent of targets in the dataset are expressed in strong, definitive language, using direct formulations such as “will reduce … by X%,” “will reach X” or “will achieve X”. A smaller share uses more tentative or aspirational language, with formulations such as “should reach” or “strive to”. The full downloadable dataset links each target to the original source sentence, allowing users to review the precise wording of each formulation.

To extract quantified targets at scale, we applied a text-mining pipeline to a corpus of national-level, climate-related strategic policy documents issued by China’s central government since 2004. The corpus comprised PDF and HTML documents in both Chinese and English. Each document was converted to plain text and cleaned to standardise spacing and sentence boundaries. The text was then segmented into sentences, and rule-based pattern matching was used to identify sentences containing target-relevant temporal expressions (e.g., “到2025年”, “与2010年相比”, “2020-2030年”), leveraging the standardised structure and recurring target phrasing typical of Chinese policy documents. The code used for target extraction has been uploaded to GitHub.

Extracted sentences were stored with metadata linking them to their source document, translated into English where required, and further processed using rule-based filters and NLP methods to disaggregate target elements.

Should you have any questions about the data and the underlying methodology, please get in touch with 705404@soas.ac.uk.