This study evaluates the performance of dynamical downscaling of global prediction generated from the NOAA Climate Forecast System (CFSv2) at subseasonal time-scale against dense in-situ observational data in Korea. The Weather Research and Forecasting (WRF) double-nested modeling system customized over Korea is adopted to produce very high resolution simulation that presumably better resolves geographically diverse climate features. Two ensemble members of CFSv2 starting with different initial conditions are downscaled for the summer season (June-July-August) during past 10-year (2011-2020). The comparison of simulations from the nested domain (5km resolution) of WRF and driving CFSv2 (0.5°) clearly demonstrates the manner in which dynamical downscaling can drastically improve daily mean temperature (Tmean) and maximum temperature (Tmax) in both quantitative and qualitative aspects. The downscaled temperature not only better resolves the regional variability strongly tied with topographical elevation, but also substantially lowers the systematic cold bias seen in CFSv2. The added value from the nested domain over CFSv2 is far more evident in Tmax than in Tmean, which indicates a skillful performance in capturing the extreme events. Accordingly, downscaled results show the reasonable performance in simulating the plant heat stress index that count the number of days with Tmax above 30 °C and extreme degree days that accumulate temperature exceeding 30 °C using hourly temperature. As the likelihood of extreme hot temperatures is projected to increase in Korea, the modeling skill to predict the ago-meteorological indices measuring the effect of extreme heat on crop could have significant implications for agriculture management practice.

中文翻译：

---

对全球农业应用的全球气候预报系统（CFSv2）进行动态降级的绩效评估

这项研究评估了NOAA气候预测系统（CFSv2）在亚季节时标下针对韩国密集的实地观测数据进行的全球预测动态降尺度的性能。采用在韩国定制的天气研究和预报（WRF）双嵌套建模系统来生成非常高分辨率的模拟，该模拟可能更好地解决了地理上不同的气候特征。在过去的10年（2011-2020年）内，夏季（6月至7月至8月）的初始条件不同的CFSv2的两个合奏成员将被缩减规模。来自WRF的嵌套域（分辨率为5 km）和驱动CFSv2的仿真比较（0。5°）清楚地说明了动态降尺度可以从数量和质量两方面大幅改善每日平均温度（Tmean）和最高温度（Tmax）的方式。降温后的温度不仅可以更好地解决与地形海拔高度密切相关的区域变异性，而且还可以大大降低CFSv2中出现的系统性冷偏差。在Tmax中，嵌套域在CFSv2上的增加值比在Tmean中要明显得多，这表明在捕获极端事件方面具有熟练的性能。因此，按比例缩小的结果显示出在模拟植物热胁迫指数时的合理性能，该指数可计算Tmax高于30°C的天数和使用小时温度累积温度超过30°C的极端度天数。