Prompt Engineering for Sustainability & Environmental Studies
Climate scientists, policy analysts, and sustainability professionals now work with foundation models trained on decades of Earth-observation and regulatory data. The IPCC, EPA, and corporate ESG reports are thousands of pages each — and increasingly the first readers are LLMs.
Where this is showing up in Sustainability & Environmental Studies
- ClimateBERT (Bingler, Kraus, Leippold, Webersinke — DistilRoBERTa fine-tuned on climate research, news, and corporate reports) ships five task-specific models on Hugging Face, including a TCFD disclosure classifier.
- ClimateGPT (7B/13B/70B parameters, Llama 2-based, trained on 4.2B climate-specific tokens with hierarchical RAG) was designed to synthesize interdisciplinary climate research and was trained using renewable energy.
- IBM and NASA's Prithvi-WxC (2.3B parameters, 40 years of MERRA-2 data, Sept 2024) and Prithvi-EO-2.0 (600M parameters, trained on NASA's Harmonized Landsat and Sentinel dataset) are open foundation models for weather, climate, and geospatial analysis.
- IPCC Assessment Reports, EPA regulatory filings, and corporate ESG/TCFD disclosures are now the standard corpora for RAG systems serving policy teams, sustainability consultancies, and investor-relations groups.
Projects you could build in this course
- A RAG assistant over the latest IPCC AR6 report that cites specific paragraphs and figures
- An ESG-metric extractor that pulls Scope 1/2/3 emissions from 10-K sustainability sections
- An agent that translates a local environmental-impact report into a plain-language community brief