The operational global four‐dimensional variational data assimilation system at the China Meteorological Administration

Abstract

Since 1 July 2018, the GRAPES (Global/Regional Assimilation and PrEdiction System) global 4‐dimensional variational (4D‐Var) data assimilation system has been in operation at the China Meteorological Administration (CMA). In this study, the GRAPES global 4D‐Var data assimilation system is comprehensively introduced. This system applies the non‐hydrostatic global tangent‐linear model (TLM) and the adjoint model (ADM) for the first time. The use of a digital filter as a weak constraint is achieved. A series of linear physical processes is developed, including vertical diffusion, subgrid‐scale orographic parametrization, large‐scale condensation, and cumulus convection parametrization. The vertical diffusion and subgrid‐scale orographic schemes are used in the operational suite and the linear convection parametrization and large‐scale condensation scheme remain under assessment. The Lanczos and conjugate gradient (Lanczos‐CG) algorithm and the limited‐memory Broyden‐Fletcher‐Goldfarb‐Shanno (L‐BFGS) algorithm are also developed. In terms of computational optimization, the total computational time of the GRAPES global TLM and ADM is approximately threefold that of the GRAPES global nonlinear model (NLM). Before it became operational, a one‐year retrospective trial was performed on the GRAPES global 4D‐Var data assimilation system. The entire system was stable, and the analysis and forecasting performances were significantly better than those of the 3D‐Var data assimilation system, especially in the Southern Hemisphere.

Studies of FY-3 Observations over the Past 10 Years: A Review

“…Since 2004, China began to independently develop the Global/Regional Assimilation PrEdiction System (GRAPES). The GRAPES 3D-Var [14] and 4D-Var [15] versions became operational in 2016 and 2018, respectively.…”

Challenges of Increased Resolution for the Local Ensemble Tangent Linear Model

“…Tangent linear models (TLMs) and their adjoints play a key role in generating optimal initial conditions in four-dimensional variational (4DVar) data assimilation (DA) systems used for numerical weather prediction (NWP) (e.g., Rabier et al 2000;Rosmond and Xu 2006;Gauthier and Thépaut 2001;Gauthier et al 2007;Rawlins et al 2007;JMA 2019;Zhang et al 2019). Initial conditions are obtained by minimizing a cost function that quantifies the combined error-weighted differences between the forecast and observations and between the forecast and a control background forecast.…”

“…Ideally, TLMs incorporate all features of the nonlinear forecast model. In practice, development and maintenance of TLMs of physical parametizations can be difficult; therefore, operationally they only approximate the ideal linear model (Janisková and Lopez 2012;Zhang et al 2019). In some circumstances, this limitation is not serious [e.g., when neglected processes are too slow to noticeably impact the state on the DA time scale (6-12 h)].…”

Multi‐resolution incremental 4D‐Var for WRF: Implementation and application at convective scale

“…The European Centre for Medium‐range Weather Forecasts (ECMWF) implemented the first operational 4D‐Var in November 1997 (Klinker et al ., 2000; Mahfouf and Rabier, 2000; Rabier et al ., 2000). Since then, 4D‐Var has become the mainstream operational DA algorithm at global numerical weather prediction (NWP) centres, including France (Janiskova et al ., 1999; Gauthier and Thépaut, 2001), the United Kingdom (Rawlins et al ., 2007), Canada (Gauthier et al ., 2007), Japan (Kadowaki, 2005), the US Navy (Xu et al ., 2005; Rosmond and Xu, 2006), and China (Zhang et al ., 2019). Some of the NWP centres have also developed 4D‐Var for their regional models (Zupanski, 1993; Honda et al ., 2005; Kawabata et al ., 2007; Gustafsson et al ., 2012; Tanguay et al ., 2012; Ballard et al ., 2016), even though these regional 4D‐Var systems were not always implemented into operation.…”