Toward optimized temperature sum parameterizations for forecasting the start of the pollen season
Toward optimized temperature sum parameterizations for forecasting the start of the pollen season
复制标题
优化温度总和参数化以预测花粉季节的开始
作者：
A. Pauling;R. Gehrig;B. Clot
过敏性物种开花的开始日期在过敏的背景下是非常感兴趣的。这样的预测可以帮助规划治疗和设计医学治疗。为此，通常采用温度总和模型。设计温度总和模型需要选择不同的参数，例如基础温度和温度总和的开始日期。然而，最佳参数化往往是未知的，并因地点和物种而异。本研究的目的是系统地测试参数化的温度总和模型的基础上，12个瑞士花粉站。所研究的分类群包括榛属、桤木属、白蜡属、桦属和禾本科。我们测试了简单的热模式类型（强迫只有模式），它只依赖于强迫温度和顺序模式类型，也包括冷却温度。平均绝对误差用于评估模型的性能。我们的研究表明，序贯模型不能实现一个明显的减少（统计）的错误相比，强迫只模型的所有类群。平均绝对误差大致在2至4天之间，桦木的值最低。在2012年花粉季节期间，使用优化参数化结合温度预测和气候平均值，以提供所有五个分类群开花开始的每日更新预测。通过测试额外的温度参数以及降水或辐射等其他气象因素，有可能实现模型的改进。确定一个面向过程的模型，具有高的统计性能，所有站将有利于在数值花粉扩散模型的实施。
The start date of flowering of allergenic species is of great interest in the context of allergy. Such forecasts can help to plan the therapy and design the medical treatment. For this purpose, temperature sum models are usually employed. Designing temperature sum models requires the selection of different parameters such as the base temperature and the start date for the temperature sum. However, the optimal parameterization is often unknown and varies depending on location and species. The purpose of this study was to systematically test parameterizations of temperature sum models based on 12 Swiss pollen stations. The examined taxa include Corylus, Alnus, Fraxinus, Betula, and Poaceae. We tested the simple thermal model type (forcing-only model hereafter) which relies solely on forcing temperatures and the sequential model type that includes also chilling temperatures. The mean absolute error was used to assess the performance of the models. Our study shows that the sequential model could not achieve a discernible reduction of the (statistical) error compared to the forcing-only model for all taxa. The mean absolute error lies roughly between 2 and 4 days with the lowest values for Betula. The optimized parameterizations in combination with temperature forecast and the climatological mean were used during the 2012 pollen season to provide daily updated forecasts of the start of flowering for all five taxa. Improvements of the models could possibly be achieved by testing additional temperature parameters as well as other meteorological factors such as precipitation or irradiation. Identification of a process-oriented model with high statistical performance for all stations would facilitate the implementation in numerical pollen dispersion models.