Operational forecasting of climate-sensitive ragweed pollen exposure events under delayed and simulated real-time monitoring scenarios in Central Europe.

Ambrosia artemisiifolia is a climate-sensitive invasive species whose airborne pollen constitutes a major allergenic burden across Central Europe. Operational forecasting is constrained by monitoring system design: conventional Hirst-type samplers involve processing delays that preclude real-time data availability, whereas automated systems provide near-instantaneous data. This study evaluates exceedance-based forecasting under these two paradigms across eight urban sites in Slovakia and the Czech Republic, spanning a gradient from high to minimal ragweed abundance. The real-time scenario was simulated by assuming immediate availability of Hirst-type data rather than using measurements from automated pollen monitors. Four elastic net-penalised logistic regression models incorporating seasonal, meteorological, and Ambrosia pollen history predictors were evaluated across forecast horizons of 1-7 days using an expanding window cross-validation. Exceedance thresholds were 20 pollen/m3 (Slovak sites) and 5 pollen/m3 (Czech sites). Forecast performance was strongly governed by exceedance prevalence. At Slovak sites (prevalence 8-17%), the best real-time models generally achieved higher F1 scores at a 1-day horizon (0.67-0.82) than the best delayed-reporting models (0.59-0.72). At Czech sites (prevalence below 3%), F1 scores rarely exceeded 0.30, and sensitivity was a more informative metric under severe class imbalance. The real-time advantage diminished progressively with forecast horizon, converging towards delayed-reporting performance at day 7. Sunshine duration and maximum temperature were the most consistently positive predictors, relative humidity was predominantly positive across most sites, whereas precipitation and wind speed showed weaker, less consistent associations. These findings provide a transferable framework for ragweed pollen early warning systems adaptable to heterogeneous monitoring infrastructure.