Recently, Associate Prof. LU Hui (correspondent author), Dr. YAO Panpan (first author) from the Department of Earth System Science, Tsinghua University, together with other researchers at the Aerospace Research Institute, Chinese Academy of Sciences, and Massachusetts Institute of Technology, have published an article entitled "A long term global daily soil moisture dataset derived from AMSR-E and AMSR2 (2002-2019)" in Scientific Data, describing the dataset of 18 years (2002-2020) global spatio-temporal consistent surface soil moisture. The resolution is 36 km at daily scale. This dataset adopts the soil moisture neural network retrieval algorithm developed by Yao et al. (2017). It transfers the merits of SMAP to AMSR-E/2 through using an Artificial Neural Network (ANN) in which SMAP standard SSM products serve as training targets with AMSR-E/2 brightness temperature (TB) as input. Finally, long term soil moisture data are output. The accuracy is about 5% volumetric water content (evaluation accuracy of 14 dense ground network globally).

Figure 1. 3-day composite distribution map of global surface soil moisture

Surface soil moisture is one of the important variables in the Earth's environment and climate system, which is considered as the basic climate variable. Stable and consistent long time series soil moisture is very important for monitoring global environment and climate change. Microwave remote sensing has a unique ability of soil moisture observation, which can provide high-precision global soil moisture data products at all time. The current multiple soil moisture satellite products suffer from inconsistent accuracy and spatial resolution, discontinuous product time span range, etc., which cannot meet the application requirements.

The recently launched SMAP satellite (Soil Moisture Active Passive) carries an L-band radiometer that provides the highest accuracy soil moisture observations available, while the AMSR-E (Advanced Microwave Scanning Radiometer) and AMSR2 series sensors provide long time series multi-band radiometer observations (C, X and K bands). Combining the respective advantages of the two satellite data, an artificial neural network approach is used to transfer the advantages of SMAP to AMSR-E/2 using the standard soil moisture products of SMAP as training targets and the bright temperature of AMSR-E and AMSR2 as input to obtain a long time series soil moisture dataset (NNsm) with approximate accuracy to the SMAP soil moisture products from 2002-2019, and the accuracy of the product was evaluated using a global network of 14 dense observatories, and compared with two products from JAXA and LPRM of AMSR-E/AMSR2, and with soil moisture products of SMOS and CCI.

The results show that (1) NNsm can reproduce the accuracy of soil moisture products from SMAP well in both the training and simulation phases; (2) NNsm can represent interannual variability well compared with soil moisture observations from the ground station network, and the accuracy is better than the two AMSR-E/2 soil moisture products from JAXA and LPRM; (3) NNsm can seamlessly integrate well with SMAP products on the daily scale, providing wider spatial coverage and more frequent observations within days (Figure 2), and meanwhile, it is a product to fill the gap of SMAP soil moisture products when SMAP data are temporarily missing; (4) Combined with the continuous observations of AMSR2 and the ongoing AMSR3 satellite mission, this continuous and consistent dataset of high-precision soil moisture products can last for more than 20 years and can provide valuable information for climate change studies, especially in trend analysis and extreme event studies.

Figure 2. NNsm and SMAPsm daily data fusion results

The completion of this dataset was supported by the National Key R & D Program of China (2017YFA0603703), the Second Tibetan Plateau Scientific Expedition (STEP) program (2019QZKK0206), the Strategic Priority Research Program of the Chinese Academy (XDA20100103), and the China Postdoctoral Science Foundation (2019M660606).

Data available at:

https://data.tpdc.ac.cn/en/data/c26201fc-526c-465d-bae7-5f02fa49d738/

Article:

Yao, P.P., Lu, H., Shi, J.C., Zhao, T.J., Yang K., Cosh, M.H., Gianotti, D.J.S., & Entekhabi, D. (2021). A long term global daily soil moisture dataset derived from AMSR-E and AMSR2 (2002-2019). Scientific Data. (Accepted)