Answers are grounded in this paper and cite their sources.
Pollen impacts millions of Americans, yet US-based operational models lack its representation. We evaluate real-time, tree, grass, and weed pollen forecasts from the Rapid-Refresh coupled to Chemistry (RAP-Chem) model and the Pollen Emissions for Climate Models version 2 (PECM v2.0) across North America for 2025. We evaluate RAP-Chem using counts from National Allergy Bureau (NAB) stations and a novel observational product from Pollen Sense. Pollen Sense correlated moderately with NAB (Spearman r = 0.50–0.70) and demonstrated moderate skill in reproducing phenology (start, peak, and end-of-season; SOS/POS/EOS), with mean absolute errors (MAE) for tree, grass, and weed pollen of 21.0/13.9/22.4 d; 5.2/18.4/6.6 d, and 19.2/44.2/21.4 d, respectively validating its spatial evaluation utility. At NAB stations, RAP-Chem best simulates weed pollen (r(NAB) = 0.74, r(Pollen Sense) = 0.75), followed by grass (0.53, 0.36) and tree (0.40, 0.29) pollen, with similar skill to Pollen Sense in reproducing NAB phenology (MAE = 19.9/20.7/13.8 d, 24.0/17.6/12.2 d, and 32.8/65.2/12.2 d). Total pollen has correlations r(NAB) = 0.44 and r(Pollen Sense) = 0.42. US-wide, RAP-Chem and Pollen Sense phenology agree best for weed pollen. RAP-Chem underestimates tree and grass pollen at most NAB sites and overestimates Pollen Sense tree pollen in forested areas. RAP-Chem grass and weed pollen magnitudes strongly depend on a post-processing correction for model-omitted gravitational settling, while correlation and phenology are unaffected. This experimental framework marks a critical advancement toward pollen forecasting and public health exposure management in the US.
No takes yet. Share an insight, caveat, or question.
Schnell et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: