A field’s nitrous oxide emissions can rise after fertilisation, tillage or rain following dry weather. USDA measurement guidance calls for sampling that captures these episodes, with attention to where chambers sit relative to plants and fertiliser. When and where researchers measure helps determine the emissions picture they obtain.[2]
The guidance also asks for soil temperature and moisture alongside gas measurements, so an emissions number can be read with its field conditions.[2]
Researchers used more than 12,000 chamber observations from 17 US sites. Five ecosystem simulators supplied daily soil and crop conditions, including moisture, mineral nitrogen and biomass, to four machine-learning models.[1]
The simulators’ averaged estimates filled gaps in recorded conditions. Their nitrous oxide predictions were excluded.[1]
Training used 13 sites; four were held out. Across them, the combined model achieved R² of 0.84, versus 0.26 without simulated inputs. R² measures how predictions track observed variation. For November–February test observations, it was 0.50.[1]
Sources & further reading
- Coupled machine learning–ecosystem ensemble models substantially improve predictions of nitrous oxide (N₂O) fluxes from US croplands PNAS · 4 Mar 2026 · Accessed 20 Oct 2026
- USDA LTAR Common Experiment Measurement: Chamber-based fluxes of nitrous oxide (N2O) and methane (CH4) from soil USDA Agricultural Research Service · 23 Sep 2024 · Accessed 20 Oct 2026