北半球季节冻深数据集（1850-2100年）

基于北半球1220个长时间序列的最大冻结深度站点数据作为训练数据，融合最大冻结深度的站点数据和气温、降水、冻结指数、融化指数、积雪深度、太阳辐射、叶面积指数、土壤质地等预测因子，采用XGBoost算法构建北半球土壤冻结深度模型。在模型训练过程中，采用十折交叉验证，运行300次，输出最...

Utilizing 1220 long-term time-series in-situ maximum frost depth observations from Northern Hemisphere stations as training data, this work combined the in-situ maximum frost depth measurements with a suite of predictor variables including air temperature, precipitation, freezing degree days, thawing degree days, snow depth, solar radiation, leaf area index and soil texture, and adopted the XGBoost algorithm to build a Northern Hemisphere soil frost depth model. During the model training process, ten-fold cross-validation was implemented, and 300 runs were conducted to output the optimal...

搜集汇总

背景与挑战

背景概述

北半球季节冻深数据集（1850-2100年）是一个基于多源数据和XGBoost算法构建的高精度冻深模型，覆盖历史及未来四种气候情景，空间分辨率0.5°，时间分辨率逐年，适用于冻土变化研究。