Global tropical cyclone size and intensity reconstruction dataset for 1959–2022 based on IBTrACS and ERA5 data

A global long-term tropical cyclone (TC) size and intensity reconstruction dataset is generated, covering a time period from 1959 to 2022, with a 3-hour temporal resolution. The machine learning model was established by taking ERA5-derived 10 m azimuthal median azimuthal wind profiles in six basins for which TCs were generated as input, while the maximum sustained wind speed and radius to maximum wind speed from the International Best Track Archive for Climate Stewardship (IBTrACS) was used as the learning target. An empirical wind–pressure relationship and six wind profile models were employed to estimate the minimum central pressure and outer sizes (radial distances from the cyclone center to locations where sustained wind speeds of 34, 50 and 64 knots are observed on surface) of the TCs, respectively. Compared to the IBTrACS dataset, the reconsturction dataset contains approximately 3–4 times more data points per characteristic. Over all, this dataset is in terms of both coverage and good accuracy.本研究构建了一套全球长期热带气旋（TC）尺度与强度重建数据集，时间跨度为1959年至2022年，时间分辨率为3小时。 本数据集所采用的机器学习模型，以6个热带气旋生成海盆的ERA5衍生10米方位向中位数风廓线作为输入，以国际热带气旋最佳路径气候管理档案（IBTrACS）中的最大持续风速及最大风速半径作为学习目标。本研究分别采用经验风速-气压关系与6种风廓线模型，估算热带气旋的最低中心气压与外围尺度——即气旋中心至地面观测到34、50、64节持续风速位置的径向距离。相较于IBTrACS数据集，本重建数据集的各特征变量的数据量约为其3至4倍。 总体而言，该数据集在覆盖范围与精度两方面均表现出色。