地面模式擾動對午後對流系集預報的影響

Abstract

Noah Multi Parameterization（Noah MP）是一個社群地面模式（Land surface model，LSM），針對重要地表物理過程，分別發展多種不同物理參數法，並介接多個天氣或氣候模式，作為研發及作業之用。本研究使用Noah MP設計以地表模式為主的系集擾動方法，並評估其對午後對流系集預報的影響。 本研究首先在離線（Offline）實驗架構下，探討Noah MP在台灣地區的起轉（spin-up）特徵以及不同地表物理過程對系集離散度的影響。研究結果顯示，Noah MP 約需至少三個月的起轉時間以達到平衡狀態（equilibrium state）；在敏感性實驗中共測試了五種地表物理過程，結果顯示對系集離散度的貢獻程度依序為：表面熱交換係數、冠層輻射幾何形狀、冠層氣孔阻力、蒸發表面阻力和氣孔阻力的土壤水分因子。本研究取前四種物理過程的參數法，介接大氣預報模式，建構以地表模式為主的系集擾動方法。 其次，本研究針對五個午後對流個案，共進行兩個系集實驗，以評估LSM擾動對午後對流預報的影響。每個實驗包含24個系集成員，第一個實驗只對大氣初始條件進行擾動，另一個實驗則進行了額外加入LSM擾動。24小時預報的結果顯示，加入LSM擾動後雖然沒有顯著改善模式系集平均的預報能力，但可以有效增加降雨以及近地表大氣變量系集預報的離散度。本研究進一步分析造成系集成員離散度的根因，結果顯示系集離散度的主要源於表面熱交換係數參數法，而這也導致系集成員產生分群的現象。研究亦指出，適當調校地面模式的熱交換係數參數法，是進一步改善系集預報準確度和離散度的關鍵之一。

Noah Multi Parameterization (Noah MP) is a community land surface model (LSM) that incorporates multiple physics parameterization schemes for important land surface processes. This study utilizes Noah MP to design a LSM-based ensemble perturbation scheme and evaluates its impact on ensemble forecasting of afternoon thunderstorms over Taiwan. In this study, we first investigate the spin-up characteristics of Noah MP in the offline experimental framework and examine the influence of different land surface physics processes on the ensemble spread. The results indicate that Noah MP requires a spin-up period of at least three months to reach an equilibrium state. The five tested land surface physics processes that contribute to the ensemble spread are ranked in descending order: surface layer heat exchange coefficient, canopy radiation geometry, canopy stomatal resistance, surface resistance to evaporation, and soil moisture factor for stomatal resistance. The parameterization schemes of the first four land surface processes are selected to construct the LSM-based ensemble perturbation scheme. Next, we conduct two ensemble experiments for five afternoon convection cases to assess the impact of LSM perturbation on afternoon thunderstorms forecasting. Each experiment consists of 24 ensemble members, with the first experiment perturbing only the atmospheric initial conditions and the second experiment incorporating additional LSM perturbations. The 24-hour forecast results show that while the addition of LSM perturbation does not significantly improve the ensemble mean forecast skill, it effectively increases the ensemble spread for rainfall and near-surface atmospheric variables. Further analysis reveals that the main source of ensemble spread lies in the parameterization scheme of the surface layer heat exchange coefficient, leading to the clustering of ensemble members. The study emphasizes the importance of properly calibrating the surface model's exchange coefficient parameterization scheme as a key factor in enhancing the accuracy and spread of ensemble forecasting.