Probabilistic Freeze-Thaw Record for the Northern Hemisphere, 2016-2020

This dataset provides a probabilistic freeze/thaw (FT) data record from 2016 to 2020 for the Northern Hemisphere derived using a deep learning model (U-Net). The model was informed by satellite multi-frequency microwave brightness temperature retrievals from the NASA SMAP (Soil Moisture Active Passive) and JAXA AMSR2 (Advanced Microwave Scanning Radiometer 2) radiometers, and trained using daily soil temperature observations from Northern Hemisphere weather stations and global reanalysis data (ERA-5). Unlike other available FT data records that provide only a binary classification of frozen or non-frozen conditions, this product includes both binary FT and continuous variable estimates of the probability of thawed conditions. This product is designed to complement other established binary FT data records, including the NASA FT Earth System Data Record and SMAP Level 3 FT operational products, by providing a probabilistic FT variable with enhanced accuracy and sensitivity to near-surface (<=5 cm depth) soil FT condition. The data are provided in cloud optimized GeoTIFF (COG) format.

本数据集提供了一种基于深度学习模型（U-Net）的北半球概率性冻结/解冻（FT）数据记录，时间跨度为2016年至2020年。该模型受NASA SMAP（土壤水分主动被动）和JAXA AMSR2（高级微波扫描辐射计2）辐射计的多频微波亮度温度反演结果所启发，并利用北半球气象站每日土壤温度观测数据和全球再分析数据（ERA-5）进行训练。与仅提供冻结或非冻结状态二分类的其他可用FT数据记录不同，本产品同时包含二元FT和解冻条件概率的连续变量估计。本产品旨在补充其他已建立的二元FT数据记录，包括NASA FT地球系统数据记录和SMAP第3级FT运行产品，通过提供具有增强精度和对近地表（<=5厘米深度）土壤FT条件敏感性的概率性FT变量，以实现对现有数据记录的完善。数据以云优化GeoTIFF（COG）格式提供。