Deep learning downscaled CMIP6 high-resolution (0.1°) daily near surface meteorological datasets over East Asia (ensemble mean)

Downscaling is a method to obtain high-resolution meteorological data. UNet, a great potential deep learning technology, is gaining popularity in downscaling. Based on 19 models from the Coupled Model Intercomparison Project Phase 6 (CMIP6) and the Multi-Source Weather (MSWX) dataset, we use bias correction and UNet downscaling to develop an East Asian dataset, called Climate Change for East Asia with Bias Corrected UNet Dataset (CLIMEA-BCUD). CLIMEA-BCUD provides nine meteorological variables including 2-m air temperature, 2-m daily maximum air temperature, 2-m daily minimum air temperature, precipitation, 10-m wind speed, 2-m relative humidity, 2-m specific humidity, downward shortwave radiation and downward longwave radiation with 0.1° horizontal resolution at daily intervals over the historical period of 1950-2014 and three future scenarios (SSP126, SSP245 and SSP585) of 2015-2100. The validations against MSWX shows that CLIMEA-BCUD shows good performance in terms of climatological and can simulates seasonal cycle and future change well. CLIMEA-BCUD can prompt the development of deep learning in climate research and be used for climate change, hydrology, agriculture, energy, etc.Here we provide the multi-model ensemble of CLIMEA-BCUD.

降尺度（downscaling）是一种获取高分辨率气象数据的方法。U-Net（UNet）作为极具潜力的深度学习技术，在降尺度任务中愈发受到关注。本研究基于耦合模式比较计划第六阶段（Coupled Model Intercomparison Project Phase 6, CMIP6）的19个模式以及多源气象数据集（Multi-Source Weather, MSWX），通过偏差校正（bias correction）与U-Net降尺度方法，构建了一套东亚区域数据集，命名为偏差校正U-Net东亚气候变化数据集（Climate Change for East Asia with Bias Corrected UNet Dataset, CLIMEA-BCUD）。CLIMEA-BCUD包含9类气象变量，分别为2米气温、2米日最高气温、2米日最低气温、降水量、10米风速、2米相对湿度、2米比湿、向下短波辐射与向下长波辐射；该数据集水平分辨率为0.1°，时间分辨率为逐日，覆盖1950-2014年的历史时段，以及2015-2100年的三种共享社会经济路径（Shared Socioeconomic Pathways, SSP）情景（SSP126、SSP245与SSP585）。通过与MSWX数据集开展比对验证，结果表明CLIMEA-BCUD在气候态特征模拟方面表现优异，能够较好地还原季节循环过程并精准模拟未来气候变化趋势。CLIMEA-BCUD可推动深度学习技术在气候研究领域的发展，且可应用于气候变化、水文学、农业、能源等多个研究方向。本研究公开提供CLIMEA-BCUD的多模式集合数据集。

搜集汇总

背景与挑战

背景概述

该数据集是通过深度学习技术对CMIP6和MSWX数据进行降尺度处理生成的东亚地区高分辨率气象数据集，包含9个气象变量，覆盖历史时期和未来情景，具有0.1°的水平分辨率和日间隔，适用于气候变化、水文、农业和能源等领域的研究。