LessRad SWDR_Terrain 青藏高原考虑地形效应的日均短波下行辐射数据集（2016-2018）

由于地形复杂、卫星观测有限，准确估算山区日均短波下行辐射（SWDR）仍然是一项重大挑战。本研究提出了一种新方法，将物理地形校正的辐射传输模型与机器学习相结合，利用Himawari 8卫星数据有效地估算山区日均 SWDR。在青藏高原崎岖地区，对逐景H8数据（10 min/景）地形校正结果进行日平均得到山区日均SWDR参考数据，与未校正的 SWDR 相比，本研究数据集将多年平均均方根误差（RMSE）从 23.62 W/m² 降低到 5.54 W/m²，平均偏差误差（MBE）从 20.04 W/m² 降低到 4.33 W/m²，并将判定系数（R²）从 0.833 提高到 0.985。

Accurate estimation of daily mean shortwave downward radiation (SWDR) in mountainous areas remains a significant challenge due to complex terrain and limited satellite observations. This study proposes a novel method that combines physically based topographically corrected radiative transfer models with machine learning to efficiently estimate daily mean SWDR in mountainous areas using Himawari 8 satellite data. In the rugged terrain of the Qinghai-Tibet Plateau, reference data of daily mean SWDR in mountainous areas was obtained by daily averaging the topographically corrected results of per-scene Himawari 8 data (10 minutes per scene). Compared with uncorrected SWDR, the dataset developed in this study reduced the multi-year average root mean square error (RMSE) from 23.62 W/m² to 5.54 W/m², the mean bias error (MBE) from 20.04 W/m² to 4.33 W/m², and improved the coefficient of determination (R²) from 0.833 to 0.985.

搜集汇总

背景与挑战

背景概述

该数据集是2016-2018年青藏高原地区考虑地形效应的日均短波下行辐射数据，通过融合物理地形校正模型和机器学习方法，基于Himawari 8卫星观测生成，显著提高了山区辐射估算精度（如均方根误差降低至5.54 W/m²）。数据以GeoTIFF格式提供，时间分辨率为日，空间分辨率为1km-10km，适用于大气科学和遥感研究。