Isoprene is the dominant non-methane organic compound emitted to the atmosphere, where it drives ozone and aerosol production, modulates atmospheric oxidation, and interacts with the global nitrogen cycle. Isoprene emissions are highly variable and uncertain, as is the non-linear chemistry coupling isoprene and its primary sink, the hydroxyl radical (OH). Space-based isoprene measurements can help close the gap on these uncertainties, and when combined with concurrent formaldehyde data provide a new constraint on atmospheric oxidation regimes. Here we present a next-generation machine-learning isoprene retrieval for the Cross-track Infrared Sounder (CrIS) that provides improved sensitivity, lower noise, and thus higher space-time resolution than earlier approaches. The Retrieval of Organics with CrIS Radiances (ROCR) isoprene measurements compare well with previous space-based retrievals as well as with the first-ever ground-based isoprene column measurements, with 20-50% discrepancies that reflect differing sources of systematic uncertainty. An ensemble of sensitivity tests points to the spectral background and isoprene profile specification as the most relevant uncertainty sources in the ROCR framework. We apply the ROCR isoprene algorithm to the full CrIS record from 2012-2020, showing that it can resolve fine-scale spatial gradients at daily resolution over the world’s isoprene hotspots. Results over North America and Amazonia highlight emergent connections between isoprene abundance and daily-to-interannual variations in temperature, nitrogen oxides, and drought stress.
