Frontiers in Earth Science (Jan 2024)
Development of a prediction system for precipitation- and wind-causing typhoons affecting the Korean peninsula using observational data
Abstract
Introduction: When the forecasted typhoon track differs from the numerical model’s prediction, the estimated precipitation and wind from the model may not be reliable. Typically, forecasters receive numerical model forecasts with a delay of 4 h or more in calculation time. However, a more timely reference of precipitation and wind forecasts is required in an emergency with an approaching typhoon. Analyses of the observational data of typhoon-related characteristics, such as heavy rainfall and strong winds, from 1997 to 2021 revealed that their distribution areas are considerably affected by typhoon tracks. In this study, we developed a precipitation and wind prediction system based on the observational data of the typhoons that affected the Korean Peninsula.Methods: Typhoon tracks were categorized into west-coast landfalls, southeast landfalls, and those passing the Korea Strait. Each category affects the Korean Peninsula differently in terms of rainfall and wind. We devised a system that predicts these patterns based on incoming typhoon tracks. We can make forecasts by comparing the approaching typhoons to previous instances and analyzing their center, movement direction, and size. Observations from these past typhoons were averaged to produce a forecast grid for each new typhoon.Results: Our system, validated from 2019 to 2022, showed a wind speed root-mean-square error of 3.37 m/s and a precipitation accuracy index of 0.72. For comparison, traditional numerical models yielded 5.04 m/s and 0.75, respectively. This indicates that our system is comparably efficient and computationally less demanding.Discussion: Our system’s strength is its ability to offer real-time typhoon forecasts, often faster than numerical models. However, its dependence on historical data limits its predictive power for atypical weather scenarios. It is essential to consider integrating ensemble models with these observations for enhanced accuracy. Since 2022, this system has been operational at the Korea Meteorological Administration, showing consistent reliability in forecasting.
Keywords
