Dynamical Framework of IAP Nine-Level Atmospheric General Circulation Model

The dynamical framework of the nine-level version of the IAP AGCM is presented in this paper. The emphasis of the model's description is put on the following two aspects:(1) A model's standard atmosphere, which is a satisfactory approximation to the observed troposphere and lower stratosphere standard atmosphere, is introduced into the equations of the model to permit a more accurate calculation of the vertical transport terms, especially near the tropopause; (2) The vertical levels of the model are carefully selected to guarantee a smooth dependence of layer thickness upon pressure in order to reduce the truncation error involved in the unequal interval vertical finite-differencing. For testing the model, two kinds of linear baroclinic Rossby-Haurwitz waves, one of which has a dynamically stable vertical structure and the other has a relatively unstable one, are constructed to provide initial conditions for numerical experiments. The two waves have been integrated for more than 300 days and 100 days respectively by using the model and both of them are propagating westward with almost identical phase-speed during the time period of the integrations. No obvious change of the wave patterns is found at the levels in the model's troposphere. The amplitudes of both two waves at the uppermost level, however, exhibit rather significant oscillation with time, of which the periods are exactly 20 days and 25 days espectively.The explanation of this interesting phenomena is still under investigation.

The African Climate as Predicted by the IAP Grid-Point Nine-Layer Atmospheric General Circulation Model (IAP-9L-AGCM)

ＴｈｅＡｆｒｉｃａｎＣｌｉｍａｔｅａｓＰｒｅｄｉｃｔｅｄｂｙｔｈｅＩＡＰＧｒｉｄ－ＰｏｉｎｔＮｉｎｅ－ＬａｙｅｒＡｔｍｏｓｐｈｅｒｉｃＧｅｎｅｒａｌＣｉｒｃｕｌａｔｉｏｎＭｏｄｅｌ（ＩＡＰ－９Ｌ－ＡＧＣＭ）ＣｈｉｎｅｋｅＴｈｅｏＣｈｉｄｉｅｚｉｅ①，...

Simulations of the Tropical Intraseasonal Oscillation by the Atmospheric General Circulation Model of the Beijing Climate Center

The performance of BCC （Beijing Climate Center） AGCM 2.0.1 （Atmospheric General Circulation Model version 2.0.1） in simulating the tropical intraseasonal oscillation （TIO） is examined in this paper.The simulations are validated against observation and compared with the NCAR CAM3 （Community Atmosphere Model version 3） results.The BCC AGCM2.0.1 is developed based on the original BCC AGCM （version 1） and NCAR CAM3.New reference atmosphere and reference pressure are introduced into the model.Therefore,the original prognostic variables of temperature and surface pressure become their departures from the reference atmosphere.A new Zhang-McFarlane convective parameterization scheme is incorporated into the model with a few modifications.Other modifications include those in the boundary layer process and snow cover calculation.All simulations are run for 52 yr from 1949 to 2001 under the lower boundary conditions of observed monthly SST.The TIOs from the model are analyzed.The comparison shows that the NCAR CAM3 has a poor ability in simulating the TIO.The simulated strength of the TIO is very weak.The energy of the eastward moving waves is similar to that of the westward moving waves in CAM3.While in observation the former is much larger than the latter.The seasonal variation and spatial distribution of the TIO produced by CAM3 are also much different from the observation.The ability of the BCC AGCM2.0.1 in simulating the TIO is significantly better.The simulated TIO is evident.The strength of the TIO produced by the BCC AGCM2.0.1 is close to the observation.The energy of eastward moving.waves is much stronger than that of the westward moving waves,which is consistent with the observation.There is no significant difference in the seasonal variation and spatial distribution of the TIO between the BCC model simulation and the observation.In general,the BCC model performs better than CAM3 in simulating the TIO.

A Study on Sulfate Optical Properties and Direct Radiative Forcing Using LASG-IAP General Circulation Model

The direct radiative forcing （DRF） of sulfate aerosols depends highly on the atmospheric sulfate loading and the meteorology, both of which undergo strong regional and seasonal variations. Because the optical properties of sulfate aerosols are also sensitive to atmospheric relative humidity, in this study we first examine the scheme for optical properties that considers hydroscopic growth. Next, we investigate the seasonal and regional distributions of sulfate DRF using the sulfate loading simulated from NCAR CAM-Chem together with the meteorology modeled from a spectral atmospheric general circulation model （AGCM） developed by LASG-IAP. The global annual-mean sulfate loading of 3.44 mg m-2 is calculated to yield the DRF of -1.03 and -0.57 W m-2 for clear-sky and all-sky conditions, respectively. However, much larger values occur on regional bases. For example, the maximum all-sky sulfate DRF over Europe, East Asia, and North America can be up to -4.0 W m-2. The strongest all-sky sulfate DRF occurs in the Northern Hemispheric July, with a hemispheric average of -1.26 W m-2. The study results also indicate that the regional DRF are strongly affected by cloud and relative humidity, which vary considerably among the regions during different seasons. This certainly raises the issue that the biases in model-sinmlated regional meteorology can introduce biases into the sulfate DRF. Hence, the model processes associated with atmospheric humidity and cloud physics should be modified in great depth to improve the simulations of the LASG-IAP AGCM and to reduce the uncertainty of sulfate direct effects on global and regional climate in these simulations.

Prediction of the Asian-Australian Monsoon Interannual Variations with the Grid-Point Atmospheric Model of IAP LASG(GAMIL)

Seasonal prediction of Asian-Australian monsoon （A-AM） precipitation is one of the most important and challenging tasks in climate prediction. In this paper, we evaluate the performance of Grid Atmospheric Model of IAP LASG （GAMIL） on retrospective prediction of the A-AM interannual variation （IAV）, and determine to what extent GAMIL can capture the two major observed modes of A-AM rainfall IAV for the period 1979-2003. The first mode is associated with the turnabout of warming （cooling） in the Nifio 3.4 region, whereas the second mode leads the warming/cooling by about one year, signaling precursory conditions for ENSO. We show that the GAMIL one-month lead prediction of the seasonal precipitation anomalies is primarily able to capture major features of the two observed leading modes of the IAV, with the first mode better predicted than the second. It also depicts the relationship between the first mode and ENSO rather well. On the other hand, the GAMIL has deficiencies in capturing the relationship between the second mode and ENSO. We conclude： （1） successful reproduction of the E1 Nifio-excited monsoon-ocean interaction and E1 Nifio forcing may be critical for the seasonal prediction of the A-AM rainfall IAV with the GAMIL; （2） more efforts are needed to improve the simulation not only in the Nifio 3.4 region but also in the joining area of Asia and the Indian-Pacific Ocean; （3） the selection of a one-tier system may improve the ultimate prediction of the A-AM rainfall IAV. These results offer some references for improvement of the GAMIL and associated seasonal prediction skill.

Effects of Nested Area Size upon Regional Climate Model Simulations

This paper presents a numerical study on the 1998 summer rainfall over the Yangtze River valley in central and eastern China, addressing effect of a nested area size on simulations in terms of the technique of nesting a regional climate model (RCM) upon a general circulation model (GCM). Evidence suggests that the size exerts greater impacts upon regional climate of the country, revealing that a larger nested size is superior to a small one for simulation in mitigating errors of GCM-provided lateral boundary forcing. Also, simulations show that the RCM should incorporate regions of climate systems of great importance into study and a low-resolution GCM yields more pronounced errors as a rule when used in the research of the Tibetan Plateau, and, in contrast, our PσRCM can do a good job in describing the plateau’s role in a more realistic and accurate way. It is for this reason that the tableland should be included in the nested area when the RCM is employed to investigate the regional climate. Our PσRCM nesting upon a GCM reaches more realistic results compared to a single GCM used.

Modelling the January and July Climate of 9000 Years before Present

The January and July climate of 9000 years before present was simulated through a set of perpetual experimentsby means of the newly designed 9 Level Atmospheric General Circulation Model (AGCM) in the institute of Atmospheric Physics (IAP). The results were analysed and compared with previous results simulated by the IAP 2Level AGCM. There exists good agreement between them. It is found that the temperature is higher in July and lower inJanuary in 9000 yBP than that at present. The temperature difference is more obvious in the Northern Hemispherethan in the Southern Hemisphere and greater in July than in January. These results prove the potential abilities of theg-L AGCM in the climate simulation and climate prediction.

Evaluation of Spring Persistent Rainfall over East Asia in CMIP3/CMIP5 AGCM Simulations

The progress made fi＇om Phase 3 to Phase 5 of the Coupled Model Intercomparison Project （CMIP3 to CMIP5） in simulating spring persistent rainfall （SPR） over East Asia was examined from the outputs of nine atmospheric general circulation models （AGCMs）. The majority of the models overestimated the precipitation over the SPR domain, with the mean latitude of the SPR belt shifting to the north. The overestimation was about 1mm d-1 in the CMIP3 ensemble, and the northward displacement was about 3°, while in the CMIP5 ensemble the overestimation was suppressed to 0.7 mm d-i and the northward shift decreased to 2.5°. The SPR features a northeast-southwest extended rain belt with a slope of 0.4°N/°E. The CMIP5 ensemble yielded a smaller slope （0.2°N/°E）, whereas the CMIP3 ensemble featured an unre- alistic zonally-distributed slope. The CMIP5 models also showed better skill in simulating the interannual variability of SPR. Previous studies have suggested that the zonal land-sea thermal contrast and sensible heat flux over the southeastern Tibetan Plateau are important for the existence of SPR. These two ther- mal factors were captured well in the CMIP5 ensemble, but underestimated in the CMIP3 ensemble. The variability of zonal land-sea thermal contrast is positively correlated with the rainfall amount over the main SPR center, but it was found that an overestimated thermal contrast between East Asia and South China Sea is a common problem in most of the CMIP3 and CMIP5 models. Simulation of the meridional thermal contrast is therefore important for the future improvement of current AGCMs.

Severe winter weather as a response to the lowest Arctic sea-ice anomalies

Possible impact of reduced Arctic sea-ice on winter severe weather in China is investigated regarding the snowstorm over southern China in January 2008. The sea-ice conditions in the summer （July-September） and fall （September-November） of 2007 show that the sea-ice is the lowest that year. During the summer and fall of 2007, sea ice displayed a significant decrease in the East Siberian, the northern Chukchi Sea, the western Beaufort Sea, the Barents Sea, and the Kara Sea. A ECHAM5.4 atmospheric general circula- tion model is forced with realistic sea-ice conditions and strong thermal responses with warmer surface air temperature and higher-than-normal heat flux associated with the sea-ice anomalies are found. The model shows remote atmospheric responses over East Asia in January 2008, which result in severe snowstorm over southern China. Strong water-vapor transported from the Bay of Bengal and from the Pacific Ocean related to Arctic sea-ice anomalies in the fall （instead of summer） of 2007 is considered as one of the main causes of the snowstorm formation.

Polar Vortex Response to Pacific Ocean Warming and Its Additive Nonlinearity with the Indian Ocean

A previous modeling study about Pacific Ocean warming derived polar vortex response signals, by subtracting those in the Indian Ocean warming experiments from those in the Indo-Pacific. This approach questions the resemblance of such an indirectly derived response to one directly forced by Pacific Ocean warming. This is relevant to the additive nonlinearity of atmospheric responses to separated Indian and Pacific Ocean forcing. In the present study, an additional set of ensemble experiments are performed by prescribing isolated SST forcing in the tropical Pacific Ocean to address this issue. The results suggest a qualitative resemblance between responses in the derived and additional experiments. Thus, previous findings about the impact of Indian and Pacific Ocean wanning are robust. This study has important implications for future climate change projections, considering the non-unanimous warming rates in tropical oceans in the 21st century. Nevertheless, a comparison of present direct-forced experiments with previous indirect-forced experiments suggests a significant additive nonlinearity between the Indian and Pacific Ocean warmings. Further diagnosis suggests that the nonlinearity may originate from the thermodynamic processes over the tropics.

Quantifying the Response Strength of the Southern Stratospheric Polar Vortex to Indian Ocean Warming in Austral Summer

A previous multiple-AGCM study suggested that Indian Ocean Warming （IOW） tends to warm and weaken the southern polar vortex.Such an impact is robust because of a qualitative consistency among the five AGCMs used.However,a significant difference exists in the modeled strengths,particularly in the stratosphere,with those in three of the AGCMs （CCM3,CAM3,and GFS） being four to five times as strong as those in the two other models （GFDL AM2,ECHAM5）.As to which case reflects reality is an important issue not only for quantifying the role of tropical ocean warming in the recent modest recovery of the ozone hole over the Antarctic,but also for projecting its future trend.This issue is addressed in the present study through comparing the models＇ climatological mean states and intrinsic variability,particularly those influencing tropospheric signals to propagate upward and reach the stratosphere.The results suggest that differences in intrinsic variability of model atmospheres provide implications for the difference.Based on a comparison with observations,it is speculated that the impact in the real world may be closer to the modest one simulated by GFDL AM2 and ECHAM5,rather than the strong one simulated by the three other models （CCM3,CAM3 and GFS）.In particular,IOW during the past 50 years may have dynamically induced a 1.0℃ warming in the polar lower stratosphere （～100 hPa）,which canceled a fraction of radiative cooling due to ozone depletion.

Dependence of the AGCM Climatology on the Method of Prescribing Surface Boundary Conditions and Its Climatological Implication

By using IAP 9L AGCM, two sets of long-term climatological integration have been performed with the two different interpolation procedures for generating the daily surface boundary conditions. One interpolation procedure is the so-called “traditional” scheme, for which the daily surface boundary conditions are obtained by linearly interpolating between the observed monthly mean values, however the observed monthly means cannot be preserved after interpolation. The other one is the “new” scheme, for which the daily surface boundary conditions are obtained by linearly interpolating between the “artificial” monthly mean values which are based on, but are different from the observed ones, after interpolating with this new scheme, not only the observed monthly mean values are preserved, the time series of the new generated daily values is also more consistent with the observation. Comparison of the model results shows that the differences of the globally or zonally averaged fields between these two integrations are quite small, and this is due to the compensating effect between the different regions. However, the differences of the two patterns (the global or regional geographical distributions), are quite significant, for example, the magnitude of the difference in the JJA mean rainfall between these two integrations can exceed 2 mm/ day over Asian monsoon regions, and the difference in DJF mean surface air temperature can also exceed 2?C over this region. The fact that the model climatology depends quite strongly on the method of prescribing the daily surface boundary conditions suggests that in order to validate the climate model or to predict the short-term climate anomalies, either the “new? interpolation scheme or the high frequency surface boundary conditions (e.g., daily or weekly data instead of the monthly data) should be introduced. Meanwhile, as for the coupled model, the daily coupling scheme between the different component climate models ( e.g., atmospheric and oceanic general circulation models) is preferred in order to partly eliminate the “climate drift” problem which may appear during the course of direct coupling.

Numerical Simulation and Comparison Study of the Atmospheric Intraseasonal Oscillation

Daily mean outputs for 12 yr （1978-1989） from two general circulation models （SAMIL-R42L9 and CAM2.0.2） are analyzed and compared with the corresponding NCEP/NCAR reanalysis dataset, and results in two models show clearly that the root-mean square errors （RMSEs） from the simulation of intraseasonal oscillation can take 30-40 percent of the total RMSE, particularly, the distributions of the RMSE in simulating intraseasonal oscillation are almost identical with that of the total RMSE. The maximum RMSE of intraseasonal oscillation height at 500 hPa is shown in the middle latitude regions, but there are also large RMSEs of intraseasonal oscillation wind over the tropical western Pacific and tropical Indian Oceans. The simulated ISO energy in the tropic has very large difference from the result of the NCEP/NCAR reanalysis dataset which means the simulation of tropical atmospheric ISO still possesses serious insufficiency. Therefore, intraseasonal oscillation in the weather and climate numerical simulation is very important, and thus, how to improve the ability of the GCM to simulate the intraseasonal oscillation becomes very significant.

NUMERICAL EXPERIMENTS FOR THE IMPACT OF ANTARCTIC ICE COVER AND SEA SURFACE TEMPERATURE ON CLIMATE VARIABILITY

By using a three-level atmospheric general circulation model (AGCM),we have completed several numerical experiments to study the impacts of sea surface temperature anomaly (SSTA) and antarctic ice cover anomaly (AICA) during 1981—1983 on climate variability.The results show that during the El Nino period of 1982—1983 the impact of SSTA overrides that of AICA.SSTA mainly affects equatorial zonal circulation and produces PNA wave train,and SE-NW wave train in East Asia to influence the weather of China.AICA produces west-east anomalous vortex streets in the middle latitudes of both hemispheres and affects the intensity of the polar vortex of Southern Hemisphere.

A Simulation Study on the Extreme Temperature Events of the 20th Century by Using the BCC_AGCM

Extreme temperature events are simulated by using the Beijing Climate Center Atmospheric General Circulation Model （BCC_AGCM） in this paper. The model has been run for 136 yr with the observed ex- ternal forcing data including solar insolation, greenhouse gases, and monthly sea surface temperature （SST）. The daily maximum and minimum temperatures are simulated by the model, and 16 indices representing various extreme temperature events are calculated based on these two variables. The results show that the maximum of daily maximum temperature （TXX）, maximum of daily minimum （TNX）, minimum of daily maximum （TXN）, minimum of daily minimum （TNN）, warm days （TXg0p）, warm nights （TNg0p）, summer days （SU25）, tropical nights （TR20）, and warm spell duration index （WSDI） have increasing trends during the 20th century in most regions of the world, while the cold days （TX10p）, cold nights （TN10p）, and cold spell duration index （CSDI） have decreasing trends. The probability density function （PDF） of warm/cold days/nights for three periods of 1881-1950, 1951- 1978, and 1979-2003 is examined. It is found that before 1950, the cold day/night has the largest probability, while for the period of 1979-2003, it has the smallest probability. In contrast to the decreasing trend of cold days/nights, the PDF of warm days/nights exhibits an opposite trend. In addition, the frost days （FD） and ice days （ID） have decreasing trends, the growing season has lengthened, and the diurnal temperature range is getting smaller during the 20th century. A comparison of the above extreme temperature indices between the model output and NCEP data （taken as observation） for 1948 2000 indicates that the mean values and the trends of the simulated indices are close to the observations, and overall there is a high correlation between the simulated indices and the observations. But the simulated trends of FD, ID, growing season length, and diurnal temperature range are not consistent with the observations and their correlations are low or even negative. This indicates that the model is incapable to simulate these four indices although it has captured most indices of the extreme temperature events.

Sensitivity of simulated tropical intraseasonal oscillations to cumulus schemes

The sensitivity of simulated tropical intraseasonal oscillations (ISO) to different cumulus parameterization schemes was analyzed using an atmospheric general circulation model (latest version-SAMIL2.2.3) developed at the Laboratory for Numerical Modeling for Atmospheric Sciences and Geophysical Fluid Dynamics (LASG) at the Institute of Atmospheric Physics (IAP) of the Chinese Academy of Sciences.Results show that the basic features of tropical climatological intraseasonal oscillations (CISO) can be captured using all three cumulus schemes.The CISO simulated by the Tiedtke scheme was found to be more realistic than that of the Manabe and Zhang-McFarlane schemes.The results of simulated transient intraseasonal oscillations (TISO) indicate that although the Tiedtke and the Zhang-McFarlane schemes in the new version SAMIL2.2.3 have been adjusted according to different problems,only the latter can simulate the eastward propagation of the 27-50-day TISO mode.It may be associated with the more realistic diabatic heating profile simulated by the Zhang-McFarlane scheme.In addition,the Manabe scheme in SAMIL2.2.3 is the same as that in the prior version SAMIL2.08.However,some aspects of the physical process,such as the radiation scheme and aerosol condition,have been changed.Conversely the eastward propagation from 100°E to the west of the tropical 27-50-day TISO mode only can be simulated using the Manabe scheme of SAMIL 2.08.Consequently,not all the improvements of physical parameterization schemes work well in every respect.The coordinated developments between dynamic frame and physical processes,and among different physical processes,are important methods that may be used to improve the model.

The Observed and Simulated Major Summer Climate Features in Northwest China and Their Sensitivity to Land Surface Processes

Northwest China （NWC） is a typical arid and semi-arid region. In this study, the main summer climate features over NWC are presented and the performance of an atmospheric general circulation model （NCEP GCM/SSiB） over this region is evaluated. Satellite-derived vegetation products are applied in the model. Based on comparison with observational data and Reanalysis II data, the model generally captures major features of the NWC summer energy balance and circulation. These features include： a high surface tem- perature center dominating the planetary boundary layer; widespread descending motion; an anticyclone （cyclone） located in the lower and middle （upper） troposphere, covering most parts of central NWC; and the precipitation located mainly in the high elevation areas surrounding NWC. The sensitivity of the summer energy balance and circulation over NWC and surrounding regions to land surface processes is assessed with specified land cover change. In the sensitivity experiment, the degradation over most parts of NWC, except the Taklimakan desert, decreases the surface-absorbed radiation and leads to weaker surface thermal effects. In northern Xinjiang and surrounding regions, less latent heating causes stronger anomalous lower-level anticyclonic circulation and upper-level cyclonic circulation, leading to less summer precipitation and higher surface temperature. Meanwhile, the dry conditions in the Hexi Corridor produce less change in the latent heat flux. The circulation change to the north of this area plays a domi- nant role in indirectly changing lower-level cyclonic conditions, producing more convergence, weaker vertical descending motion, and thus an increase in the precipitation over this region.

Propagation and Mechanisms of the Quasi-Biweekly Oscillation over the Asian Summer Monsoon Region

The propagation and underlying mechanisms of the boreal summer quasi-biweekly oscillation（QBWO）over the entire Asian monsoon region are investigated,based on ECMWF Interim reanalysis（ERA-Interim）data,GPCP precipitation data,and an atmospheric general circulation model（AGCM）.Statistical analyses indicate that the QBWO over the Asian monsoon region derives its main origin from the equatorial western Pacific and moves northwestward to the Bay of Bengal and northern India,and then northward to the Tibetan Plateau（TP）area,with a baroclinic vertical structure.Northward propagation of the QBWO is promoted by three main mechanisms：barotropic vorticity,boundary moisture advection,and surface sensible heating（SSH）.It is dominated by the barotropic vorticity effect when the QBWO signals are situated to the south of 20°N.During the propagation taking place farther north toward the TP,the boundary moisture advection and SSH are the leading mechanisms.We use an AGCM to verify the importance of SSH on the northward propagation of the QBWO.Numerical simulations confirm the diagnostic conclusion that the equatorial western Pacific is the source of the QBWO.Importantly,the model can accurately simulate the propagation pathway of the QBWO signals over the Asian monsoon region.Simultaneously,sensitivity experiments demonstrate that the SSH over northern India and the southern slope of the TP greatly contributes to the northward propagation of the QBWO as far as the TP area.

Simulation of the Effects of the Preceding SST Anomalies over the Tropical Eastern Pacific on Precipitation to the South of the Yangtze River in June

Numerical experiments are performed to simulate the response of the atmospheric circulation and pre-cipitation over East China in June to the sea surface temperature（SST）anomalies over the tropical eastern Pacific（TEP）from preceding September to June by using an atmospheric general circulation model （AGCM）.We constructed composite positive/negative SST anomalies（P-SSTAs/N-SSTAs）based on the observational SST anomalies over the TEP from September 1997 to June 1998.The results show that：（1） the response of the precipitation in the Yangtze River basin and its southern area（YRBS）to El Nino with different durations varies with the maximum amplitude of the precipitation anomalies appearing when the imposed duration is from November to next June,and the minimum appearing when the SST anomalies is only imposed in June.The anomalies of the precipitation are reduced when the duration of the forcing SST anomalies over the TEP is shortened and the positive SST anomalies in the preceding autumn tend to cause significantly more rainfall in the YRBS.This is in agreement with previous diagnostic analysis results.（2）The simulated precipitation anomalies over the YRBS are always obviously positive under strong or weak positive SST anomalies over the TEP.The intensity of the precipitation anomalies increases with increasing intensity of the SST anomalies in the experiments.The simulation results are consistent with the observations during the warm SST events,suggesting reasonable modeling results.（3）When negative SST anomalies in the TEP are put into the model,the results are different from those of the diagnostic analysis of La Nina events.Negative precipitation anomalies in YRBS could be reproduced only when the negative SST anomalies are strong enough.

IMPACT OF DIABATIC PROCESSES IN AGCM ON 4-DIMENSIONAL VARIATIONAL DATA ASSIMILATION

The impact of diabatic processes on 4-dimensional variational data assimilation (4D-Var) was studied using the 1995 version of NCEP's global spectral model with and without full physics.The adjoint was coded manually.A cost function measuring spectral errors of 6-hour forecasts to 'observation' (the NCEP reanalysis data) was minimized using the L-BFGS (the limited memory quasi-Newton algorithm developed by Broyden,Fletcher,Goldfard and Shanno) for optimizing parameters and initial conditions.Minimization of the cost function constrained by an adiabatic version of the NCEP global model converged to a minimum with a significant amount of decrease in the value of the cost function.Minimization of the cost function using the diabatic model, however,failed after a few iterations due to discontinuities introduced by physical parameterizations.Examination of the convergence of the cost function in different spectral domains reveals that the large-scale flow is adjusted during the first 10 iterations,in which discontinuous diabatic parameterizations play very little role.The adjustment produced by the minimization gradually moves to relatively smaller scales between 10-20th iterations.During this transition period,discontinuities in the cost function produced by 'on-off' switches in the physical parameterizations caused the cost function to stay in a shallow local minimum instead of continuously decreasing toward a deeper minimum. Next,a mixed 4D-Var scheme is tested in which large-scale flows are first adiabatically adjusted to a sufficient level,followed by a diabatic adjustment introduced after 10 to 20 iterations. The mixed 4D-Var produced a closer fit of analysis to observations,with 38% and 41% more decrease in the values of the cost function and the norm of gradient,respectively,than the standard diabatic 4D-Var,while the CPU time is reduced by 21%.The resulting optimal initial conditions improve the short-range forecast skills of 48-hour statistics.The detrimental effect of parameterization discontinuities on minimization was also reduced.