A Multi-Scale Localization Approach to an Ensemble Kalman filter

Miyoshi

Tanaka

2013

SOLA

Self Cite

In the ensemble Kalman filter, covariance localization plays an essential role in treating sampling errors in the ensemble-based error covariance between distant locations. We may limit the influence of observations excessively, particularly when the model resolution is very high, since larger-scale structures than the localization scale are removed due to tight localization for the high-resolution model. To retain the larger-scale structures with a limited ensemble size, the dual-localization approach, which considers two separate localization scales simultaneously, has been proposed. The dual-localization method analyzes smallscale and large-scale analysis increments separately using spatial smoothing and two localization scales. These are the control parameters of the dual-localization method, and this study aims to investigate the parameter sensitivities by performing a number of observing system simulation experiments using an intermediate AGCM known as the SPEEDY model. Two smoothing functions, the spherical harmonics spectral truncation and the Lanczos filter, are tested, and the results indicate no significant difference. Also, sensitivity to the two localization parameters is investigated, and the results show that the dual-localization approach outperforms traditional single localization with relatively wide choices of the two localization scales by about 400-km ranges. This suggests that we could avoid fine tuning of the two localization parameters.

Section: Resultsmentioning

confidence: 99%

Section: Experimental Settingsmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Parameter Sensitivities of the Dual-Localization Approach in the Local Ensemble Transform Kalman Filter

Miyoshi

Tanaka

2013

SOLA

Self Cite

The 10,240‐member ensemble Kalman filtering with an intermediate AGCM

Miyoshi

Geophysical Research Letters

Imamura

2014

Self Cite

The local ensemble transform Kalman filter (LETKF) with an intermediate atmospheric general circulation model (AGCM) is implemented with the Japanese 10 petaflops (floating point operations per second) "K computer" for large-ensemble simulations of 10,240 members, 2 orders of magnitude greater than the typical ensemble size of about 100. The computational challenge includes the eigenvalue decomposition of 10,240 × 10,240 dense covariance matrices at each grid point. Using the efficient eigenvalue solver for the K computer, the LETKF computations are accelerated by a factor of 8, allowing a 3 week experiment of 10,240-member LETKF with an intermediate AGCM for the first time. The flow-dependent 10,240-member ensemble revealed meaningful long-range error correlations at continental scales. The surface pressure error correlation shows teleconnection patterns like the Pacific North American pattern. Specific humidity error correlation shows continental scale wave trains. Investigations with different ensemble sizes suggest that at least several hundred members be necessary to capture these continental scale error correlations.

Multilocalization data assimilation for predicting heavy precipitation associated with a multiscale weather system

Yang

Chen

et al. 2017

J Adv Model Earth Syst

Self Cite

High‐resolution numerical simulations are regularly used for severe weather forecasts. To improve model initial conditions, a single short localization is commonly applied in the ensemble Kalman filter when assimilating observations. This approach prevents large‐scale corrections from appearing in a high‐resolution analysis. To improve heavy rainfall forecasts associated with a multiscale weather system, analyses must be accurate across a range of spatial scales, a task that is difficult to accomplish using a single localization. This study is the first to apply a dual‐localization (DL) method to improve high‐resolution analyses used to forecast a real‐case heavy rainfall event associated with a Meiyu front on 16 June 2008 in Taiwan. A Meiyu front is a multiscale weather system characterized by storm‐scale convection, a mesoscale front, and large‐scale southwesterly monsoonal flow. The use of the DL method to produce the analyses was able to correct both the synoptic‐scale moisture flux transported by southwesterly monsoonal flow and the mesoscale low‐level convergence offshore of southwestern Taiwan. As a result, the forecasted amount, pattern, and temporal evolution of the heavy rainfall event were improved.