Page 147 - 《高原气象》2025年第6期
P. 147

6 期        谭淇昌等:基于深度学习方法改进CMIP6模式对中国东部沿海复合极端风雨事件的模拟能力                                      1555










































                     图7 2001 -2014年DL-MRM(a)、 DL-MSE(b)、 多模式集合平均(c)和多模式集合中位数(d)对中国东部沿海
                                 年平均复合极端风雨事件(CWPE)模拟相对于观测的相对偏差分布(单位: %)
                                                    黑色打点为超过95%显著性检验
                Fig. 7 The spatial distribution of relative biases in the simulation of annual mean Compound Wind and Precipitation Extremes
                  (CWPE) along the eastern coastal region of China by DL-MRM (a), DL-MSE (b), Multi-Model Ensemble Mean (c),
                            and Multi-Model Ensemble Median (d) during 2001 -2014, relative to observations. Units: %.
                                          The black points indicate exceed the 95% significance test


















                    图8 2001 -2014年DL-MRM模型与DL-MSE模型(a)、 多模式集合平均(MME-Mean)(b)和多模式集合中位数
                           (MME-Median)(c)对中国东部沿海年平均复合极端风雨事件(CWPE)模拟的相对偏差对比
                           红色区域表示DL-MRM相对偏差较低, 表现较好; 蓝色区域表示较高, 表现较差; 白色区域表示不确定,
                                                   即两个对比结果的相对偏差不一致
                Fig. 8 Comparison of relative biases in the simulation of annual average Compound Wind and Precipitation Extremes (CWPE)
                over the eastern coastal region of China from 2001 to 2014 between the DL-MRM model and DL-MSE model (a), Multi-Model
                   Ensemble Mean (MME-Mean) (b), and Multi-Model Ensemble Median (MME-Median) (c). Red areas indicate lower
                      relative biases for DL-MRM, performing better; blue areas indicate higher relative biases, performing worse;
                          white areas indicate uncertainty, i. e. , the relative biases of the two compared results are inconsistent
   142   143   144   145   146   147   148   149   150   151   152