Prediction Model for Annual Variation in Total Pollen by Allergenic Trees in Korean Cities

Area of Science:

• Aerobiology
• Environmental Science
• Allergy Research

Background:

• Pollen forecasting aids respiratory allergy management by predicting concentrations.
• Current systems (e.g., KMA) offer short-term (2-day) forecasts.
• Long-term tree pollen analysis reveals significant annual variations not captured by daily models.

Purpose of the Study:

• To develop annual prediction models for allergenic tree pollens.
• To utilize long-term, multi-site pollen and meteorological data.
• To enhance the accuracy of pollen forecasting systems.

Main Methods:

• Collected daily pollen data (1998-2021) from nine Korean sites using Hirst-type traps.
• Used daily weather data from Korea Meteorological Administration (KMA) stations.
• Developed prediction models using three variable selection methods: TM (t-test/stepwise), FM (full model/stepwise), and LM (LASSO).

Main Results:

• Models were developed using 1998-2017 data and validated with 2018-2021 data.
• The TM models demonstrated superior suitability for actual forecasting compared to FM and LM.
• While TM had lower R² than FM/LM, its performance in error metrics (RMSE, MAE, ME) indicated better forecasting capability.

Conclusions:

• The developed annual variation models can be integrated into existing daily pollen forecast systems.
• Incorporating annual pollen potential control can significantly improve daily forecast accuracy.
• This approach offers enhanced tools for managing seasonal respiratory allergies.