Abstract. Reliable precipitation data are highly necessary for geoscience research in the Third Pole (TP) region but still lacking, due to the complex terrain and high spatial variability of precipitation here. Accordingly, this study produces a long-term (1979–2020) high-resolution (1/30°) precipitation dataset (TPHiPr) for the TP by merging the atmospheric simulation-based ERA5_CNN with gauge observations from more than 9000 rain gauges, using the Climatology Aided Interpolation and Random Forest methods. Validation shows that the TPHiPr is generally unbiased and has a root mean square error of 4.5 mm day^-1, a correlation of 0.84 and a critical success index of 0.67 with respect to all independent rain gauges in the TP, demonstrating that this dataset is remarkably better than the widely-used global/quasi- global datasets, including the fifth-generation atmospheric reanalysis of the European Centre for Medium-Range Weather Forecasts (ERA5), the final run version 6 of the Integrated Multi-satellitE Retrievals for Global Precipitation Measurement (IMERG) and the Multi-Source Weighted-Ensemble Precipitation version 2 (MSWEP V2). Moreover, the TPHiPr can better detect precipitation extremes compared with the three widely-used datasets. Overall, this study provides a new precipitation dataset with high accuracy for the TP, which may have broad applications in meteorological, hydrological and ecological studies. The produced dataset can be accessed via https://doi.org/10.11888/Atmos.tpdc.272763 (Yang and Jiang, 2022).

中文翻译：

---

TPHiPr：基于高分辨率大气建模和密集观测的第三极地区长期高精度降水数据集

摘要。可靠的降水数据对于第三极（TP）地区的地球科学研究非常必要，但由于该地区地形复杂，降水空间变异性大，目前仍缺乏可靠的降水数据。因此，本研究通过将基于大气模拟的 ERA5_CNN 与来自 9000 多个雨量计的观测数据相结合，为 TP 生成了一个长期（1979-2020 年）高分辨率（1/30°）降水数据集（TPHiPr），使用气候辅助插值法和随机森林法。验证表明 TPHiPr 通常是无偏的，并且均方根误差为 4.5 mm day ^-1，与青藏高原所有独立雨量计的相关性为 0.84，关键成功指数为 0.67，表明该数据集明显优于广泛使用的全球/准全球数据集，包括第五代大气再分析欧洲中期天气预报中心 (ERA5)、全球降水测量综合多卫星反演 (IMERG) 的最终运行版本 6 和多源加权集合降水版本 2 (MSWEP V2)。此外，与三个广泛使用的数据集相比，TPHiPr 可以更好地检测极端降水。总体而言，本研究为青藏高原提供了一个新的高精度降水数据集，可能在气象、水文和生态研究中具有广泛的应用价值。