Adjoint-Derived Observation Impact Using WRF in the Western North Pacific

Abstract

An increasing number of observations have contributed to the performance of numerical weather prediction systems. Accordingly, it is important to evaluate the impact of these observations on forecast accuracy. While the observing system experiment (OSE) requires considerable computational resources, the adjoint-derived method can evaluate the impact of all observational components at a lower cost. In this study, the effect of observations on forecasts is evaluated by the adjoint-derived method using the Weather Research and Forecasting Model, its adjoint model, and a corresponding three-dimensional variational data assimilation system in East Asia and the western North Pacific for the 2008 typhoon season. Radiance observations had the greatest total impact on forecasts, but conventional wind observations had the greatest impact per observation. For each observation type, the total impact was greatest for radiosonde and each Advanced Microwave Sounding Unit (AMSU)-A satellite, followed by surface synoptic observation from a land station (SYNOP), Quick Scatterometer (QuikSCAT), atmospheric motion vector (AMV) wind from a geostationary satellite (GEOAMV), and aviation routine weather reports (METARs). The fraction of beneficial observations was approximately 60%–70%, which is higher than that reported in previous studies. For several analyses of Typhoons Sinlaku (200813) and Jangmi (200815), dropsonde soundings taken near the typhoon had similar or greater observation impacts than routine radiosonde soundings. The sensitivity to the error covariance parameter indicates that reducing (increasing) observation (background) error covariance helps to reduce forecast error in the current analysis framework. The observation impact from OSEs is qualitatively similar to that from the adjoint method for major observation types. This study confirms that radiosonde observations provide primary information on the atmospheric state as in situ observations and that satellite radiances are an essential component of atmospheric observation systems.

Adjoint-Based Observation Impact of Advanced Microwave Sounding Unit-A (AMSU-A) on the Short-Range Forecast in East Asia

“…OSEs의 한계를 극복하기 위하여 비교적 최근에 제 시된 수반모델에 기반한 관측민감도(Forecast sensitivity to observations: FSO) 진단법 또는 관측영향 진단법을 이용하면, 단기예보에 대한 관측영향을 단 시간 내에 독립적으로 산출할 수 있다 Jung et al, 2013;Kim andKim, 2014, Kim et al, 2017). 이 방법은 수십만 개의 위성 자료의 관측영향을 준 현업 과 비슷한 속도로 계산하기 때문에, 어떤 종류 또는 채널의 위성 자료가 수치예측 성능 향상에 도움이 되 는지를 효율적으로 판단하게 해준다 (Cardinali, 2009;Gelaro and Zhu, 2009;Gelaro et al, 2010;Joo et al, 2013;Jung et al, 2013;Lorenc and Marriott, 2014;Kim and Kim, 2014;Kim, 2016).…”

“…이 방법은 수십만 개의 위성 자료의 관측영향을 준 현업 과 비슷한 속도로 계산하기 때문에, 어떤 종류 또는 채널의 위성 자료가 수치예측 성능 향상에 도움이 되 는지를 효율적으로 판단하게 해준다 (Cardinali, 2009;Gelaro and Zhu, 2009;Gelaro et al, 2010;Joo et al, 2013;Jung et al, 2013;Lorenc and Marriott, 2014;Kim and Kim, 2014;Kim, 2016). 수반모델에 기반한 관측민감도 진단법이 가진 장점에도 불구하고, 이 방 법에서 사용하는 선형가정 때문에 중, 장기 예측에 대 한 관측영향을 산출하는 데는 어느 정도 한계가 있다.…”

“…관측민감도 진단법을 이용하여 관측영향을 살펴 본 최근 연구에서는 AMSU-A 복사관측 자료가 전 지구 예측에 대한 기여도가 가장 크다고 알려져 있다 (Cardinali, 2009;Gelaro and Zhu, 2009;Gelaro et al, 2010;Joo et al, 2013;Lorenc and Marriott, 2014;Kim and Kim, 2014;Kim, 2016). 또한 AMSU-A 복 사관측 자료는 2008년 태풍 시즌의 동아시아와 북서 태평양 지역의 날씨 예측에 있어서도 관측 자료들 중 가장 큰 영향을 보인 바 있다 (Jung et al, 2013 Joo et al (2013), Kim and Kim (2014) …”

The importance of appropriate verification metrics for the assessment of observation impact in a convection‐permitting modelling system

“…Since then, the FSOI method has been applied to assess observation impact in several different forecasting systems (e.g. Langland, ; Gelaro et al ; Weissmann et al ; Jung et al ; Lorenc and Marriott, ; Cardinali and Healy, ; Holdaway et al ; Privé et al ; Lupu et al ; Janisková and Cardinali, ; Horányi et al ; Mallick et al ).…”

Assessment of errors in Precipitable Water data derived from Global Navigation Satellite System observations

“…Now, we can deliver the water vapor observations from GNSS to the meteorological agencies with low latency (60 min) (Bosy et al, 2012;Karabatic et al, 2011) or using most advanced real-time International GNSS Service orbits and clocks (Hadas and Bosy, 2014) with no latency at all. These observations later on can be used in the assimilation step (Hurter and Maier, 2013) of nowcasting (Karabatic et al, 2011) and forecasting (Bennitt and Jupp, 2012) models as a valuable piece of water vapor information (ranked first in per observation impact of NWP data source, Jung et al, 2013).…”