[A-CG43] Science in the Arctic Region

[ACG43-P02] Validation results of the GCOM-C/SGLI Cryosphere product third version upgrade

*Rigen Shimada^1,2, Masahiro Hori³, Teruo Aoki⁴, Tomonori Tanikawa², Sumito Matoba⁵, Masashi Niwano², Knut Stamnes⁶, Wei Li⁶, Nan Chen⁷ (1.Japan Aerospace Exploration Agency, 2.Meteorological Research Institute, 3.University of Toyama, 4.National Institute of Polar Research, 5.Institute of Low Temperature Science, Hokkaido University, 6.Stevens Institute of Technology, 7.National Center for Atmospheric Research)

Keywords:GCOM-C, SGLI, Cryosphere

Global Change Observation Mission for Climate (GCOM-C) which carries Second-generation Global Imager (SGLI) has been launched on 23 December 2017. GCOM-C/SGLI observes various geophysical variables in the Atmosphere, Ocean, Land and Cryosphere. The version 1 and 2 products were released in public in 2018 and 2020. Through these upgrades the accuracies of the most products were improved. In 2021, the version 3 products were released in public through blush-up of algorithms and further validation works. In this study, we show the third version upgrade validation results of the cryosphere products. GCOM-C/SGLI regularly creates cryosphere product including classification product (C1) and snow properties product (C2). C1 product has focused on cloud mask and surface classification for polar and high altitude region. C2 product has focused on snow and ice physical parameters: snow grain size and snow and ice surface temperature. C1 classification algorithm and C2 snow grain size retrieval algorithm were based on the neural network machine learning method^{1, 2}. All training data were simulated by the DISORT radiative transfer model. In this upgrade, the training data sets were revised in classification and snow grain size retrieval algorithms. C1 product was validated by comparing Terra/MODIS snow and sea ice extent product (snow area: MOD10C2 Snow Cover Extent Product, sea ice area: MOD29E1D Sea Ice Product). In the comparison of snow and ice extent, its accuracy showed 6.5% as relative error (defined as quotient of root mean square error and average of validation data). This accuracy was improved from version 1 and 2 accuracy (9.4% and 8.5%). C2 product was validated by comparing with in-situ observation results obtained at the E-GRIP site on the North-eastern Greenland Ice Sheet in July 2018, the Nakasatsunai site on the Northern Japan in March 2019 and February 2020, and around Dome Fuji site on the Antarctic ice Sheet in November 2018 to January 2019. The in-situ snow grain size were measured with IceCube3 and Handheld Integrating Sphere Snow Grain Sizer (HISSGraS). In the comparison of shallow layer snow grain size, its accuracy showed 50.2 % as relative error. This accuracy was got worse just focused on the number from version 2 accuracy (34.1%, Arctic region only). However the evaluation range of snow grain size was expanded from around 200 micro-meter to over 1000 micro-meter. The version 3 accuracy focused on the Arctic region same as version 2 showed 34.8% as relative error, that was about same as version 2. These results mean that the upgraded products become robust. From these validation results, JAXA decided to release the GCOM-C/SGLI version 3 product including C1 and C2. And we are planning to continue the validation and algorithm improvement.

References
1. Chen, N., W. Li, C. Gatebe, T. Tanikawa, M. Hori, R. Shimada, T. Aoki, K. Stamnes, New neural network cloud mask algorithm based on radiative transfer simulations, Remote Sens. Environ., 219, 62-71, 2018.
2. Chen, N., W. Li, Y. Fan, Y. Zhou, T. Aoki, T. Tanikawa, M. Niwano, M. Hori, R. Shimada, S. Matoba, K. Stamnes, Snow parameter retrieval (SPR) algorithm for GCOM-C/SGLI, Remote Sens. Environ., submitted.
3. Gallet J.C, F. Domine, C. S. Zender, and G. Picard, Measurement of the specific surface area of snow using infrared reflectance in an integrating sphere at 1310 and 1550 nm, The Cryosphere, 3, 167-182, 2009.