[p1] Applications in various physical and biological systems


[p1-9]

Local Ensemble Kalman Filter data assimilation for SWOT mission: Example of the Amazon river basin

Jean-Francois Vuillaume (JAMSTEC), D.Yamazaki (JAMSTEC), D.Ikeshima (Tokyo Institute of Technology), S.Kanae, (Tokyo Institute of Technology)

 
Abstract

The Surface Water Observation (SWOT) mission projected to be launched in 2020 is expected to provide high spatial altimetry resolution data (from 50 to 100m) for both ocean and river surface. The prime advantage of this mission is the use of an interferometric radar known as KaRIN (Ka-band Interferometer) which will provide a 120 km swath. Several studies have been conducted to evaluate the potential of such observation on hydro model quality improvement in the US, Ohio river (Andreadis., 2007), Niger (Pedinotti., 2014, Munier 2015), Amazon (Paiva., 2013) and Tennessee River Basin (Yoon., 2013) but with different assimilation data, methods and results even if all of them show discharge forecast measurement.

In this study, we used the global flood model CaMAFlood (Yamazaki et al., 2014) and the global data assimilation model developed by Ikeshima (2016). The model used a Local Ensemble Transform Kalman Filter (LEnKTF) scheme with 20 ensemble members. In our experiments, we assimilate the anomaly of river elevation that has a width higher than 50 meters. We evaluated the global performance of the model and in particular the Amazon river basin for the period 2008-2010 to improve discharge computation.

The framework of the study is based on a twin experiment to simulate (1) the "true" system state and then (2) an ensemble of corrupted model states. The virtual SWOT observations of river equivalent to the CaMAFlood global hydrological model level high were assimilated into the model with a repeat cycle of 21 days to assimilate the elevation anomaly of the pseudo SWOT observations over the Amazon basin for the period 2008-2010.

Finally, the performance of the assimilation is discussed in term of seasonality, upstream-downstream and observation assimilation. In addition, we compare (1) non-assimilated, assimilated and the "true" CaMaFlood forecast.(2) the performance of the assimilation of both the absolute and the anomalies of water elevation.