著者
Paula Maldonado Juan Ruiz Celeste Saulo
出版者
Meteorological Society of Japan
雑誌
SOLA (ISSN:13496476)
巻号頁・発行日
vol.17, pp.96-102, 2021 (Released:2021-05-19)
参考文献数
34
被引用文献数
4

This study investigates the impact of applying different types of initial and boundary perturbations for convective-scale ensemble data assimilation systems. Several observing system simulation experiments (OSSEs) were performed with a 2-km horizontal resolution, considering a realistic environment, taking model error into account, and combining different perturbations' types with warm/cold start initialization. Initial perturbations produce a long-lasting impact on the analysis's quality, particularly for variables not directly linked to radar observations. Warm-started experiments provide the most accurate analysis and forecasts and a more consistent ensemble spread across the different spatial scales. Random small-scale perturbations exhibit similar results, although a longer convergence time is required to up-and-downscale the initial perturbations to obtain a similar error reduction. Adding random large-scale perturbations reduce the error in the first assimilation cycles but produce a slightly detrimental effect afterward.
著者
Paula Maldonado Juan Ruiz Celeste Saulo
出版者
Meteorological Society of Japan
雑誌
SOLA (ISSN:13496476)
巻号頁・発行日
pp.2021-015, (Released:2021-04-14)
被引用文献数
4

This study investigates the impact of applying different types of initial and boundary perturbations for convective-scale ensemble data assimilation systems. Several OSSEs were performed with a 2-km horizontal resolution, considering a realistic environment, taking model error into account, and combining different perturbations' types with warm/cold start initialization. Initial perturbations produce a long-lasting impact on the analysis's quality, particularly for variables not directly linked to radar observations. Warm-started experiments provide the most accurate analysis and forecasts and a more consistent ensemble spread across the different spatial scales. Random small-scale perturbations exhibit similar results, although a longer convergence time is required to up-and-downscale the initial perturbations to obtain a similar error reduction. Adding random large-scale perturbations reduce the error in the first assimilation cycles but produce a slightly detrimental effect afterward.