Daily Snow Depth Fusion Dataset for Central Asia

This dataset employs the XGBoost (XGB) machine learning model, adopting a seasonal modeling strategy (winter, spring, and autumn) to integrate the advantages of multiple daily snow depth (SD) products, including ERA5-Land, ERA5, JRA-55, MERRA-2, and GLDAS, based on in-situ SD observations. By coupling multi-dimensional covariates such as topography, meteorological factors, temporal variables, land use, and snow-related parameters, a high-precision daily SD fusion model was developed for Central Asia (CA). The model was then applied to generate a 0.1° daily SD fusion product for CA spanning 1990–2023 (covering winter, spring, and autumn). Evaluation results demonstrate that the dataset achieves an RMSE of 4.1 cm, MAE of 2.3 cm, and R of 0.96 across the CA region, significantly improving accuracy compared to other existing SD products. This dataset provides reliable data support for climate change studies, water resource management, and disaster early warning systems in Central Asia.本数据集采用XGBoost（极限梯度提升树，XGB）机器学习模型，采用季节建模策略（冬季、春季、秋季），基于原位雪深（Snow Depth, SD）观测数据，融合ERA5-Land、ERA5、JRA-55、MERRA-2与GLDAS多款每日雪深产品的优势。通过耦合地形、气象因子、时间变量、土地利用以及积雪相关参数等多维协变量，为中亚（Central Asia, CA）构建了高精度每日雪深融合模型。随后将该模型应用于生成1990–2023年覆盖中亚的0.1°分辨率每日雪深融合产品（涵盖冬季、春季与秋季）。评估结果显示，本数据集在中亚区域内的均方根误差（Root Mean Square Error, RMSE）为4.1 cm，平均绝对误差（Mean Absolute Error, MAE）为2.3 cm，决定系数（Coefficient of Determination, R）为0.96，相较于现有其他雪深产品，精度得到显著提升。本数据集可为中亚地区的气候变化研究、水资源管理以及灾害预警系统提供可靠的数据支撑。