Improving Monsoon Prediction With a Coupled EnKF Data Assimilation System

Currently data assimilation for coupled ocean-atmosphere models are typically done in a weakly coupled mode, with 3D-Var used for the ocean DA, 3 or 4D-Var used for the atmospheric DA, the atmospheric model seeing the ocean SST, and the ocean model driven by the surface fluxes. It is also common to perform a separate SST 2-D OI analysis (“Reynolds SSTs”) and use it as a boundary condition for the atmosphere, and as a forcing for the ocean.

Our project, supported by the India Monsoon Mission Directorate, has the ultimate goal of developing a strongly coupled ocean-atmosphere data assimilation based on NCEP's Coupled Forecasting System coupled with the Local Ensemble Transform Kalman Filter (CFS-LETKF). The models will be coupled and vertical localization of the LETKF will be used for the DA coupling. This will allow the near surface atmospheric model variables to be influenced by the ocean observations, and the near surface ocean model variables to be influenced by the atmospheric observations. Student Travis Sluka is developing both a weakly and a strongly coupled DA for a simpler coupled GCM (SPEEDY-NEMO, kindly provided by Fred Kucharski, ICTP). This will allow to efficiently explore different approaches for coupled DA.

As part of this project, Dr. Guo-Yuan Lien developed an advanced GFS-LETKF, documented in http://code.google.com/p/miyoshi/wiki/GfsLetkf With this system, Dr. Lien achieved for the first time successful assimilation of TRMM/TMPA precipitation (Lien, 2014). In addition the implementation of the Ensemble Forecast Sensitivity to Observations (EFSO, Hotta, 2014) allows to select the precipitation observations that improve the short-range forecasts. The experiments that Lien carried out show that the assimilation in a cycle has a positive impact that increases with time.