Data-driven predictive modeling of pollen concentration for allergy prevention in Islamabad and Rawalpindi

Airborne pollen, notably from highly allergenic taxa such as pollen from trees and grasses, especially Broussonetia papyrifera, significantly impacts the quality of life for individuals prone to allergic reactions in Islamabad and Rawalpindi, Pakistan. Forecasting elevated pollen concentrations and disseminating this information is crucial for taking proactive measures to protect public health. This study employs a data-oriented method to use operational forecasting models for seasonal cycles, phenological stages forecasts, and daily average airborne pollen concentrations (1–7 days ahead). All datasets spanning 2004–2024 for different modeling approaches were received from the Pakistan Meteorological Department, Islamabad. The accumulated heat unit principles are used in a phenological model to forecast different growth stages of pollen flowering.

Additionally, patient data collected through a survey was utilized to develop an allergy index, from which a weekly probabilistic forecast model was generated using computational intelligence. The results show that the phenological model accurately predicted pollen characteristics, with a variation of ± 2.8 days for onset and ± 4.2 days against the flowering period. However, the long-range pollen forecast model exhibited overestimation, so the forecast was multiplied by a bias factor of 0.658. With an index of agreement ranging from 0.85 to 0.93, the overall performance indicates the promising operational potential of the models. The probabilistic forecast model gives the predicted hazard index a 73% hit alarm rate. The end users of this research are the Pakistan Meteorological Department health agencies, the National Disaster Management Authority, the general public, the media, and district administration. Consequently, proactive public health measures can be implemented effectively.