Evaluation of an ensemble based data assimilation system over Southern South America: Sensitivity to the use of perturbed boundary conditions

MARIA EUGENIA DILLON; YANIAN GARCIA SKABAR; EUGENIA KALNAY; JUAN RUIZ; MARCOS SAUCEDO; ESTELA COLLINI

Santiago de Chile

Congreso; 11th ICSHMO International Conference on Southern Hemisphere Meteorology and Oceanography; 2015

American Meteorological Society

One of the big challenges in numerical weather prediction is to reduce the uncertainty in the initial conditions. At the National Meteorological Service (SMN) of Argentina, many efforts have been carried out to address this issue: a regional data assimilation system is being developed using the Local Ensemble Transform Kalman Filter (LETKF) coupled with the Weather Research and Forecasting Modeling System (WRF). The selection of this method is motivated mainly because it generates an ensemble of initial conditions that can be used for ensemble forecasting and also by the favorable results obtained by many authors and its computational efficiency. In this manner, two important objectives are achieved: the improvement of the initial conditions and the explicit estimation of its uncertainty that can be propagated forward in time using an ensemble of forecasts. In this work, the LETKF-WRF data assimilation system is evaluated over Southern South America during a two month period including November and December 2012. A horizontal resolution of 40 km and 40 ensemble members are used, and the analyses are obtained with a 3 hour frequency, assimilating the observations available in the PREPBUFR files from the Global Data Assimilation System. Encouraging results about the performance of this regional data assimilation system have been achieved with different configurations by the authors, for example with a multischeme structure or with the inclusion of the vertical profiles retrieved from the Atmospheric Infrared Sounder (AIRS) in the assimilation cycles. In this work we explore the sensitivity of the analysis accuracy to the use of perturbed boundary conditions. Previous studies demonstrated that this can be done in order to explicitly represent the uncertainty in the boundary conditions in the data assimilation system, thus improving its performance. We compare the results of the analysis cycles obtained without using perturbations at the boundaries with an experiment where the boundary conditions are perturbed using random balanced perturbations. These perturbations are generated as scaled differences of randomly selected atmospheric states as described by the Global Forecasting System (GFS) forecasts. The impact of lateral boundary perturbations upon the assimilation system is evaluated through the verification of the 6 hour forecasts. In addition, an intense precipitation case study is selected in order to evaluate the performance of a 48 hour forecast. The results are encouraging and indicate that boundary conditions perturbations should be considered for the operational implementation of this system. Future work would focus on the implementation of a complete configuration taking advantage of a multischeme ensemble, the assimilation of the AIRS retrievals and the perturbed boundary conditions altogether.