Possible Sources of Forecast Errors Generated by the Global/Regional Assimilation and Prediction System for Landfalling Tropical Cyclones.PartⅠ:Initial Uncertainties

This paper investigates the possible sources of errors associated with tropical cyclone （TC） tracks forecasted using the Global/Regional Assimilation and Prediction System （GRAPES）. The GRAPES forecasts were made for 16 landfaIling TCs in the western North Pacific basin during the 2008 and 2009 seasons, with a forecast length of 72 hours, and using the default initial conditions （＂initials＂, hereafter）, which are from the NCEP-FNL dataset, as well as ECMWF initials. The forecasts are compared with ECMWF forecasts. The results show that in most TCs, the GRAPES forecasts are improved when using the ECMWF initials compared with the default initials. Compared with the ECMWF initials, the default initials produce lower intensity TCs and a lower intensity subtropical high, but a higher intensity South Asia high and monsoon trough, as well as a higher temperature but lower specific humidity at the TC center. Replacement of the geopotential height and wind fields with the ECMWF initials in and around the TC center at the initial time was found to be the most efficient way to improve the forecasts. In addition, TCs that showed the greatest improvement in forecast accuracy usually had the largest initial uncertainties in TC intensity and were usually in the intensifying phase. The results demonstrate the importance of the initial intensity for TC track forecasts made using GRAPES, and indicate the model is better in describing the intensifying phase than the decaying phase of TCs. Finally, the limit of the improvement indicates that the model error associated with GRAPES forecasts may be the main cause of poor forecasts of landfalling TCs. Thus, further examinations of the model errors are required.

Helmholtz Solving and Performance Optimization in Global/Regional Assimilation and Prediction System

Despite efficient parallelism in the solution of physical parameterization in the Global/Regional Assimilation and Prediction System(GRAPES),the Helmholtz equation in the dynamic core,with the increase of resolution,can hardly achieve sufficient parallelism in the solving process due to a large amount of communication and irregular access.In this paper,optimizing the Helmholtz equation solution for better performance and higher efficiency has been an urgent task.An optimization scheme for the parallel solution of the Helmholtz equation is proposed in this paper.Specifically,the geometrical multigrid optimization strategy is designed by taking advantage of the data anisotropy of grid points near the pole and the isotropy of those near memory equator in the Helmholtz equation,and the Incomplete LU(ILU)decomposition preconditioner is adopted to speed up the convergence of the improved Generalized Conjugate Residual(GCR),which effectively reduces the number of iterations and the computation time.The overall solving performance of the Helmholtz equation is improved by thread-level parallelism,vectorization,and reuse of data in the cache.The experimental results show that the proposed optimization scheme can effectively eliminate the bottleneck of the Helmholtz equation as regards the solving speed.Considering the test results on a 10-node two-way server,the solution of the Helmholtz equation,compared with the original serial version,is accelerated by 100,with one-third of iterations reduced.

Conjugate Gradient Algorithm in the Four-Dimensional Variational Data Assimilation System in GRAPES

Minimization algorithms are singular components in four-dimensional variational data assimilation(4DVar).In this paper,the convergence and application of the conjugate gradient algorithm(CGA),which is based on the Lanczos iterative algorithm and the Hessian matrix derived from tangent linear and adjoint models using a non-hydrostatic framework,are investigated in the 4DVar minimization.First,the influence of the Gram-Schmidt orthogonalization of the Lanczos vector on the convergence of the Lanczos algorithm is studied.The results show that the Lanczos algorithm without orthogonalization fails to converge after the ninth iteration in the 4DVar minimization,while the orthogonalized Lanczos algorithm converges stably.Second,the convergence and computational efficiency of the CGA and quasi-Newton method in batch cycling assimilation experiments are compared on the 4DVar platform of the Global/Regional Assimilation and Prediction System(GRAPES).The CGA is 40%more computationally efficient than the quasi-Newton method,although the equivalent analysis results can be obtained by using either the CGA or the quasi-Newton method.Thus,the CGA based on Lanczos iterations is better for solving the optimization problems in the GRAPES 4DVar system.

Developments of the Three-Dimensional Variational Data Assimilation System for the Nonhydrostatic GRAPES

Based on the original GRAPES（Global/Regional Assimilation and PrEdiction System）3DVAR（p3DAR）, which is defined on isobaric surface,a new three-dimensional variational data assimilation system（m3DVAR） is constructed and used exclusively with the nonhydrostatic GRAPES model in order to reduce the errors caused by spatial interpolation and variable transformation,and to improve the quality of the initial value for operational weather forecasts.Analytical variables of the m3DVAR are fully consistent with predictands of the GRADES model in terms of spatial staggering and physical definition.A different vertical coordinate and the nonhydrostatic condition are taken into account,and a new scheme for solving the dynamical constraint equations is designed for the m3DVAR.To deal with the diffculties in solving the nonlinear balance equation atσlevels,dynamical balance constraints between mass and wind fields are reformulated,and an effective mathematical scheme is implemented under the terrain-following coordinate.Meanwhile,new observation operators are developed for routine observational data,and the background error covariance is also obtained.Currently,the m3DVAR system can assimilate all routine observational data. Multi-variable idealized experiments with single point observations are performed to validate the m3DVAR system.The results show that the system can describe correctly the multi-variable analysis and the relationship of the physical constraints.The difference of innovation and the analysis residual forπalso show that the analysis error of the m3DVAR is smaller than that of the p3DVAR.The T s scores of precipitation forecasts in August 2006 indicate that the m3DVAR system provides reduced errors in the model initial value than the p3DVAR system.Therefore,the m3DVAR system can improve the analysis quality and initial value for numerical weather predictions.

Effects of Different Initial Fields on GRAPES Numerical Prediction

In this paper, a heavy rainfall process occurring in the Huaihe River Basin during 9-10 July 2005 is studied by the new generation numerical weather prediction model system-GRAPES, from the view of different initial field effects on the prediction of the model. Several numerical experiments are conducted with the initial conditions and lateral boundary fields provided by T213 L31 and NCEP final analyses, respectively. The sensitivity of prediction products generated by GRAPES to different initial conditions, including effects of three-dimensional variational assimilation on the results, is discussed. After analyzing the differences between the two initial fields and the four simulated results, the memonic ability of the model to initial fields and their influences on precipitation forecast are investigated. Analyses show the obvious differences of sub-synoptic scale between T213 and NCEP initial fields, which result in the corresponding different simulation results, and the differences do not disappear with the integration running. It also shows that for the same initial field whether it has data assimilation or not, it only obviously influences the GRAPES model results in the initial 24 h. Then the differences reduce. In addition, both the location and intensity of heavy rain forecasted by GRAPES model Further is very close to the fact, but the forecasting area of strong torrential rain has some differences from the fact. For the same initial field when it has assimilation, the 9-12-, 12-24-, and 0-24-h precipitation forecasts of the model are better than those without assimilation. All these suggest that the ability of GRAPES numerical prediction depends on the different initial fields and lateral boundary conditions to some extent, and the differences of initial fields will determine the differences of GRAPES simulated results.

New Version of the CMA-GFS Dynamical Core Based on the Predictor–Corrector Time Integration Scheme

The operational numerical weather prediction system established by the China Meteorological Administration(CMA),based on the Global/Regional Assimilation and Prediction System(GRAPES)model,adopts the classical semi-implicit semi-Lagrangian(SISL)time integration algorithm.This paper describes a major upgrade to the dynamical core of the CMA global forecast system(CMA-GFS),which was successfully incorporated into operation in 2020.In the upgrade,the classical SISL is further developed into a predictor–corrector scheme,a three-dimensional(3D)reference profile instead of the original isothermal reference profile is applied when implementing the semi-implicit algorithm,and a hybrid terrain-following vertical coordinate system is also applied.The new version of the dynamical core greatly improves the model performance,the time integration reaches second-order accuracy,the time step can be extended by 50%,and the efficiency is greatly improved(by approximately 30%).Atmospheric circulation simulation is systematically improved,and deviations in temperature,wind,and humidity are reduced.The new version of the dynamical core provides a solid foundation for further development of the entire operational system of the CMA.

Applying the WRF Double-Moment Six-Class Microphysics Scheme in the GRAPES_Meso Model： A Case Study

This study incorporated the Weather Research and Forecasting（WRF） model double-moment 6-class（WDM6） microphysics scheme into the mesoscale version of the Global/Regional Assimilation and Pr Ediction System（GRAPES_Meso）. A rainfall event that occurred during 3–5 June 2015 around Beijing was simulated by using the WDM6, the WRF single-moment 6-class scheme（WSM6）, and the NCEP 5-class scheme, respectively. The results show that both the distribution and magnitude of the rainfall simulated with WDM6 were more consistent with the observation. Compared with WDM6, WSM6 simulated larger cloud liquid water content, which provided more water vapor for graupel growth, leading to increased precipitation in the cold-rain processes. For areas with the warmrain processes, the sensitivity experiments using WDM6 showed that an increase in cloud condensation nuclei（CCN）number concentration led to enhanced CCN activation ratio and larger cloud droplet number concentration（Nc） but decreased cloud droplet effective diameter. The formation of more small-size cloud droplets resulted in a decrease in raindrop number concentration（Nr）, inhibiting the warm-rain processes, thus gradually decreasing the amount of precipitation. For areas mainly with the cold-rain processes, the overall amount of precipitation increased; however, it gradually decreased when the CCN number concentration reached a certain magnitude. Hence, the effect of CCN number concentration on precipitation exhibits significant differences in different rainfall areas of the same precipitation event.

An MPI+OpenACC-Based PRM Scalar Advection Scheme in the GRAPES Model over a Cluster with Multiple CPUs and GPUs

A moisture advection scheme is an essential module of a numerical weather/climate model representing the horizontal transport of water vapor.The Piecewise Rational Method(PRM) scalar advection scheme in the Global/Regional Assimilation and Prediction System(GRAPES) solves the moisture flux advection equation based on PRM.Computation of the scalar advection involves boundary exchange,and computation of higher bandwidth requirements is complicated and time-consuming in GRAPES.Recently,Graphics Processing Units(GPUs) have been widely used to solve scientific and engineering computing problems owing to advancements in GPU hardware and related programming models such as CUDA/OpenCL and Open Accelerator(OpenACC).Herein,we present an accelerated PRM scalar advection scheme with Message Passing Interface(MPI) and OpenACC to fully exploit GPUs’ power over a cluster with multiple Central Processing Units(CPUs) and GPUs,together with optimization of various parameters such as minimizing data transfer,memory coalescing,exposing more parallelism,and overlapping computation with data transfers.Results show that about 3.5 times speedup is obtained for the entire model running at medium resolution with double precision when comparing the scheme’s elapsed time on a node with two GPUs(NVIDIA P100) and two 16-core CPUs(Intel Gold 6142).Further,results obtained from experiments of a higher resolution model with multiple GPUs show excellent scalability.