Ensemble Data Assimilation of GSMaP precipitation into the nonhydrostatic global atmospheric model NICAM

Shunji Kotsuki* and Koji Terasaki, Guo-Yuan Lien, Takemasa Moyoshi, and Eugenia Kalnay
RIKEN Advanced Institute for Computational Science, Japan
Department of Atmospheric and Oceanic Science, University of Marylan

It is generally difficult to assimilate precipitation data into numerical models mainly because of non-Gaussianity of precipitation variables and nonlinear precipitation processes. Lien et al. (2013, 2015) proposed to use an ensemble Kalman filter approach to avoid explicit linearization of models, and a Gaussian transformation (GT) method to deal with the non-Gaussianity of precipitation variables. Lien et al. pioneered to show that using an EnKF and GT helps improve the forecasts by assimilating global precipitation data both in a simulated study using the SPEEDY model, and in a real-world study using the NCEP GFS and TRMM Multi-satellite Precipitation Analysis (TMPA) data.

This study extends the previous study by Lien et al. and assimilates the JAXA’s Global Satellite Mapping of Precipitation (GSMaP) data into the Nonhydrostatic Icosahedral Atmospheric Model (NICAM) at 112-km horizontal resolution. This study newly develops a method to construct the two GTs (forward and inverse GTs) for observed and forecasted precipitation using the previous 30-day precipitation data. Using this new forward GT, precipitation variables are transformed to be the Gaussian variables, and we succeeded in improving the forecasts by assimilating the GSMaP precipitation. We also found the use of the inverse GT, so that we can obtain observation-like precipitation fields transformed by observation-based inverse GT from the model forecasts. Moreover, we also explore online estimation of model parameters related to precipitation processes using precipitation data. This presentation will include the most recent progress up to the time of the conference.

*email: shunji.kotsuki@riken.jp
*Preference: Oral