TY - JOUR
T1 - Using the Linearized Observation Operator to Calculate Observation Space Ensemble Perturbations in Ensemble Filters
AU - Shlyaeva, Anna
AU - Whitaker, Jeffrey S.
N1 - Publisher Copyright:
©2018. The Authors.
PY - 2018/7
Y1 - 2018/7
N2 - Within the National Oceanic and Atmospheric Administration National Weather Service, the hybrid ensemble-variational system (Gridpoint Statistical Interpolation, GSI) is run together with the 80-member ensemble square root filter (EnSRF) operationally for the global forecast data assimilation system. EnSRF uses observation operator from GSI: current operational configuration requires 81 runs of GSI in the observation operator mode to run EnSRF (for each of the 80 ensemble members and for the ensemble mean). To reduce data assimilation cycle computation time, a GSI-EnSRF configuration that requires a single run of the GSI system in the observation operator mode was developed. In this configuration EnSRF uses full observation operator for the ensemble mean and linearized observation operator for the ensemble perturbations. Comparison of the two approaches shows that using linearized observation operator for ensemble perturbations compared to using full observation operator does not change the analysis results significantly and allows to reduce overall data assimilation cycle computation time.
AB - Within the National Oceanic and Atmospheric Administration National Weather Service, the hybrid ensemble-variational system (Gridpoint Statistical Interpolation, GSI) is run together with the 80-member ensemble square root filter (EnSRF) operationally for the global forecast data assimilation system. EnSRF uses observation operator from GSI: current operational configuration requires 81 runs of GSI in the observation operator mode to run EnSRF (for each of the 80 ensemble members and for the ensemble mean). To reduce data assimilation cycle computation time, a GSI-EnSRF configuration that requires a single run of the GSI system in the observation operator mode was developed. In this configuration EnSRF uses full observation operator for the ensemble mean and linearized observation operator for the ensemble perturbations. Comparison of the two approaches shows that using linearized observation operator for ensemble perturbations compared to using full observation operator does not change the analysis results significantly and allows to reduce overall data assimilation cycle computation time.
KW - data assimilation
KW - ensemble Kalman filters
KW - observation operator
UR - https://www.scopus.com/pages/publications/85050595686
U2 - 10.1029/2018MS001309
DO - 10.1029/2018MS001309
M3 - Article
AN - SCOPUS:85050595686
SN - 1942-2466
VL - 10
SP - 1414
EP - 1420
JO - Journal of Advances in Modeling Earth Systems
JF - Journal of Advances in Modeling Earth Systems
IS - 7
ER -