基于AMSR-E和AMSR2数据的全球长时序日尺度土壤水分数据集（2002-至今）

稳定连续的长时序地表土壤水分数据集对于全球环境和气候变化监测等都非常重要。SMAP等卫星搭载的L波段辐射计能提供目前最优精度的全球地表土壤水分观测，但其数据记录的短时间限制了其在长期研究中的应用；而AMSR-E和AMSR2系列传感器能提供长时序多频段辐射计观测（C、X和K波段）。本数据集是一个准实时更新的近20年（2002/07/27~至今）的全球连续一致的地表土壤水分数据集，分辨率为日尺度的36 km，采用EASE-Grid2投影坐标系，数据单位为m3/m3。数据集采用Yao et al.(2017，2021)发展的土壤水分神经网络反演算法，将SMAP的优势传递到AMSR-E/2，以目前卫星最优精度的SMAP标准土壤水分产品为训练目标，以AMSR-E/2的亮温为输入，最终输出长时序土壤水分数据。该数据集能够重现SMAP土壤水分的时空分布，精度与SMAP土壤水分产品相当；同时该数据集精度优于AMSR-E和AMSR2的官方土壤水分产品，通过全球14个密集观测站网的地面观测验证表明，其土壤水分精度为5%左右。该全球长时序数据集目前时间覆盖20年，随着AMSR2的持续在轨观测以及即将发射的后继AMSR3任务，该数据集是可延长的，为气候极端事件、趋势分析和年代际变化的长时序研究提供支持。

Stable and continuous long-term time-series surface soil moisture datasets are critical for global environmental and climate change monitoring and related research. L-band radiometers onboard satellites such as SMAP can provide global surface soil moisture observations with the currently highest precision, but their short temporal record lengths restrict their application in long-term research. In contrast, the AMSR-E and AMSR2 series of sensors provide long-term time-series multi-band radiometer observations (C, X, and K bands).

This dataset is a near-real-time updated, globally consistent and continuous surface soil moisture dataset spanning nearly 20 years (2002/07/27 to present), with a 36 km spatial resolution at daily scale, using the EASE-Grid2 projection coordinate system, with data units in m³/m³. The dataset adopts the soil moisture neural network inversion algorithm developed by Yao et al. (2017, 2021) to transfer the strengths of SMAP to AMSR-E/2: taking the SMAP standard soil moisture product with the currently highest satellite precision as the training target, and using the brightness temperatures from AMSR-E/2 as inputs, it ultimately outputs long-term time-series soil moisture data.

This dataset can reproduce the spatiotemporal distribution of SMAP soil moisture, with accuracy comparable to that of SMAP soil moisture products; meanwhile, its accuracy outperforms the official soil moisture products of AMSR-E and AMSR2. Verified by in-situ observations from 14 dense global observation networks, the retrieval accuracy of this dataset is approximately 5%.

This global long-term time-series dataset currently covers a 20-year time span. With the continued on-orbit observations of AMSR2 and the upcoming follow-on AMSR3 mission, this dataset can be extended, providing support for long-term time-series research on climate extreme events, trend analysis, and interdecadal variability.

搜集汇总

背景与挑战

背景概述

该数据集是基于AMSR-E和AMSR2卫星数据的全球长时序日尺度土壤水分数据集，时间跨度为2002年至今，空间分辨率为36km，采用神经网络反演算法，精度与SMAP土壤水分产品相当。数据集为NC和Tiff格式，总大小为41.58GB，适用于全球环境和气候变化监测等研究。