Assimilation Window Research Articles

Mesoscale Data Assimilation of Myanmar Cyclone Nargis

A mesoscale data assimilation (DA) system was developed for low latitudes, and DA experiments for the tropical cyclone Nargis were conducted. A tropical cyclone bogus (TCB) procedure was developed for the Bay of Bengal, and the impact was investigated.The Meso 4D-Var system of the Japan Meteorological Agency (JMA), which was designed for operational mesoscale DA in the mid-latitudes, was modified for DA application in the tropics. Since the relationship of the geostrophic wind was not available near the equator, a weighting function for the regression coefficient matrix for the unbalanced wind was determined based on statistics. Six DA experiments were performed to produce the initial field at 1200 UTC on April 30, 2008, in order to assess the impact of the DA periods (12 h and 24 h) and two different TCBs. Analyzed initial fields were compared with the JMA global analysis (GANAL), the best track data of RSMC, New Delhi, and the cyclone track estimation of the Indian National Centre for Ocean Information Services (INCOIS).Precipitable water vapor (PWV) around the cyclone was less than 65 mm in GANAL; in DA analyses, areas with more than 60-mm PWV were more widespread, especially to the east of the cyclone, and the maximum values exceeded 70 mm near the cyclone center. The cyclone center position was also improved by DA and the implementation of TCB; however, the cyclonic circulation was underestimated in the 24-h DA without TCB (MA24). Lack of satellite observations of the Bay of Bengal in the first 12-h assimilation windows was determined to be the reason for this underestimation.Mesoscale numerical predictions using initial fields produced by Meso 4D-Var were conducted, and the results were compared with the prediction using GANAL (control run; GA). Overestimation of the cyclone speed and the resultant positional lag in the control run were ameliorated by modifying the initial field by DA. Differences between the upper flow in the GA analysis and that in DA analyses might account for the difference in cyclone speed. Cyclone development was well predicted except in the MA24 experiment, in which underestimation of cyclonic circulation at the initial time caused subsequent insufficient development. Northward biases were observed in several predicted cyclone tracks. Differences between precipitation in the MA24 and that in other experiments might be one of the reasons for the differences in cyclone track.

Four‐dimensional variational assimilation in the unstable subspace and the optimal subspace dimension

AbstractKey apriori information used in 4D‐Var is the knowledge of the system's evolution equations. In this article we propose a method for taking full advantage of the knowledge of the system's dynamical instabilities in order to improve the quality of the analysis. We present an algorithm for four‐dimensional variational assimilation in the unstable subspace (4D‐Var – AUS), which consists of confining in this subspace the increment of the control variable. The existence of an optimal subspace dimension for this confinement is hypothesized. Theoretical arguments in favour of the present approach are supported by numerical experiments in a simple perfect nonlinear model scenario. It is found that the RMS analysis error is a function of the dimension N of the subspace where the analysis is confined and is a minimum for N approximately equal to the dimension of the unstable and neutral manifold. For all assimilation windows, from 1 to 5 d, 4D‐Var – AUS performs better than standard 4D‐Var. In the presence of observational noise, the 4D‐Var solution, while being closer to the observations, is farther away from the truth. The implementation of 4D‐Var – AUS does not require the adjoint integration. Copyright © 2010 Royal Meteorological Society

Towards an integrated observation and modeling system in the New York Bight using variational methods. Part I: 4DVAR data assimilation

Four-dimensional variational data assimilation (4DVAR) in the Regional Ocean Modeling System (ROMS) is used to produce a best-estimate analysis of ocean circulation in the New York Bight during spring 2006 by assimilating observations collected by a variety of instruments during an intensive field program. An incremental approach is applied in an overlapped cycling system with 3-day data assimilation window to adjust model initial conditions. The model-observation mismatch for all observed variables is reduced substantially. Comparisons between model forecast and independent observations show data assimilation improves forecast skill for about 15 days for temperature and salinity, and 2–3 days for velocity when the model is forced by a concatenation of successive 24-h meteorological forecasts. These time scales for forecast improvement due to data assimilation may be less in practice with real-time multi-day forecast meteorology. Tests that limit the data used to certain subsets show that assimilating satellite sea surface temperature data improves the forecast of surface and subsurface temperature, assimilating in situ temperature and salinity data from gliders improves the subsurface temperature and salinity forecast, and assimilating HF-radar surface current data improves the velocity forecast yet degrades the subsurface temperature forecast – an effect that is attributed to the lack of cross-variable covariance in the univariate background error covariance used here. During some time periods the convergence for velocity is poor as a result of the data assimilation system being unable to adjust for errors in the applied winds because surface forcing is not among the control variables. The capability of a 4DVAR data assimilation system to reduce model-observation mismatch and improve forecasts in the coastal ocean is demonstrated, and the value of accurate meteorological forcing is highlighted.

Implementation of a reduced rank square-root smoother for high resolution ocean data assimilation

Optimal smoothers enable the use of future observations to estimate the state of a dynamical system. In this paper, a square-root smoother algorithm is presented, extended from the Singular Evolutive Extended Kalman (SEEK) filter, a square-root Kalman filter routinely used for ocean data assimilation. With this filter algorithm, the smoother extension appears almost cost-free. A modified algorithm implementing a particular parameterization of model error is also described. The smoother is applied with an ocean circulation model in a double-gyre, 1/4° configuration, able to represent mid-latitude mesoscale dynamics. Twin experiments are performed: the true fields are drawn from a simulation at a 1/6° resolution, and noised. Then, altimetric satellite tracks and sparse vertical profiles of temperature are extracted to form the observations. The smoother is efficient in reducing errors, particularly in the regions poorly covered by the observations at the filter analysis time. It results in a significant reduction of the global error: the Root Mean Square Error in Sea Surface Height from the filter is further reduced by 20% by the smoother. The actual smoothing of the global error through time is also verified. Three essential issues are then investigated: (i) the time distance within which observations may be favourably used to correct the state estimates is found to be 8 days with our system. (ii) The impact of the model error parameterization is stressed. When this parameterization is spuriously neglected, the smoother can deteriorate the state estimates. (iii) Iterations of the smoother over a fixed time interval are tested. Although this procedure improves the state estimates over the assimilation window, it also makes the subsequent forecast worse than the filter in our experiment.

Short-Term Forecasting of a Midlatitude Convective Storm by the Assimilation of Single–Doppler Radar Observations

Abstract The McGill University radar data assimilation system is used to initialize a convective storm at high resolution (1 km) from single–Doppler radar observations. In this study, the background term in the assimilation system is improved. Specifically, by assuming the correlation of the errors of the control variables to be isotropic and homogeneous, the background error covariance matrix is modeled by a recursive filter. In addition, a 3-h-prior high-resolution model forecast is used as the background field. The analysis fields from the assimilation system successfully trigger the convective storms in the radar-observed regions from a single assimilation window. Without data assimilation, the modeled storms did not occur at the right time and place. To account for the rapid evolution of the convective storms and to correct the forecast errors with time, a cycling process is applied for a very short-term forecast. It is found that the first assimilation window can maintain the prediction of the storms for less than 1 h. The cycling process helps to maintain the intensity of the storm cells for a longer period of time. However, a comparison of radar observations with the 90-min simulation indicates an error in the position of the convective cells. The error of the radial component of the wind field between the observation and the simulation is larger at the upper levels. A wavelet analysis between the observation and simulated reflectivities indicates that the forecast is able to adequately predict the convective scale (∼10–20 km) during the first 20 min, whereas the simulation has more predictability at the longer scale (>30 km) beyond 20 min.

Characterization of the impact of geostationary clear‐sky radiances on wind analyses in a 4D‐Var context

AbstractThe impact of geostationary clear‐sky radiances (CSRs) on 4D‐Var wind analyses has been investigated by running a set of observing system experiments. Analysis scores have been calculated to measure the ability of individual satellite datasets to improve the wind analysis, starting from a no‐satellite baseline. In this context, the assimilation of CSRs from the two water‐vapour channels on Meteosat‐9 has been found to improve the wind analysis throughout the troposphere, with the strongest signal at 300 and 500 hPa. Indeed, for the Northern Hemisphere and the Tropics, the CSR impact at these levels exceeds that of the Meteosat‐9 atmospheric motion vectors (AMVs), the sampling of the latter in the assimilation being rather limited. Conversely, the impact of AMVs exceeds that of CSRs in the lower troposphere, where the latter provide very little direct information. This demonstrates the complementarity of the two datasets in the operational 4D‐Var wind analysis. The mechanisms through which the assimilation of CSRs can impact wind analyses have been isolated. The dominant effect is that of humidity‐tracer advection, by which the wind field is adjusted in order to fit observed humidity features via the minimization of the 4D‐Var cost function. Other mechanisms, such as balance constraints and the cycling of the forecast model that links the humidity and wind variables, have been found to play a minor role. The benefit of having frequent CSR images within the assimilation window has also been demonstrated. Copyright © 2009 Royal Meteorological Society

On the extraction of wind information from the assimilation of ozone profiles in Météo–France 4-D-Var operational NWP suite

Abstract. By applying four-dimensional variational data-assimilation (4-D-Var) to a combined ozone and dynamics Numerical Weather Prediction model (NWP), ozone observations generate wind increments through the ozone-dynamics coupling. The dynamical impact of Aura/MLS satellite ozone profiles is investigated using Météo-France operational ARPEGE NWP 4-D-Var assimilation system for a period of 3 months. A data-assimilation procedure has been designed and run on 6-h windows. The procedure includes: (1) 4-D-Var assimilating both ozone and operational NWP standard observations, (2) ARPEGE transporting ozone as a passive-tracer, (3) MOCAGE, the Météo–France chemistry and transport model re-initializing the ARPEGE ozone background at the beginning time of the assimilation window. Using observation minus forecast statistics, it is found that the ozone assimilation reduces the wind bias in the lower stratosphere. Moreover, the Degrees of Freedom for Signal diagnostics show that the MLS data covering the 68.1–31.6 hPa vertical pressure range are the most informative and their information content is nearly of the same order as tropospheric humidity-sensitive radiances. Furthermore, with the help of error variance reduction diagnostics, the ozone contribution to the reduction of the horizontal divergence background-error variance is shown to be better than tropospheric humidity-sensitive radiances.

A new methodology for the extension of the impact of data assimilation on ocean wave prediction

It is a common fact that the majority of today's wave assimilation platforms have a limited, in time, ability of affecting the final wave prediction, especially that of long-period forecasting systems. This is mainly due to the fact that after “closing” the assimilation window, i.e., the time that the available observations are assimilated into the wave model, the latter continues to run without any external information. Therefore, if a systematic divergence from the observations occurs, only a limited portion of the forecasting period will be improved. A way of dealing with this drawback is proposed in this study: A combination of two different statistical tools—Kolmogorov–Zurbenko and Kalman filters—is employed so as to eliminate any systematic error of (a first run of) the wave model results. Then, the obtained forecasts are used as artificial observations that can be assimilated to a follow-up model simulation inside the forecasting period. The method was successfully applied to an open sea area (Pacific Ocean) for significant wave height forecasts using the wave model WAM and six different buoys as observational stations. The results were encouraging and led to the extension of the assimilation impact to the entire forecasting period as well as to a significant reduction of the forecast bias.

Comparison of Local Ensemble Transform Kalman Filter, 3DVAR, and 4DVAR in a Quasigeostrophic Model

Abstract Local ensemble transform Kalman filter (LETKF) data assimilation, three-dimensional variational data assimilation (3DVAR), and four-dimensional variational data assimilation (4DVAR) schemes are implemented in a quasigeostrophic channel model. Their advantages and disadvantages are compared to assess their use in practical applications. LETKF and 4DVAR, which take into account the flow-dependent errors, outperform 3DVAR under a perfect model scenario. Given the same observations, LETKF produces more accurate analyses than 4DVAR with a 12-h window by effectively correcting the fast-growing errors with the flow-dependent background error covariance. Even though 4DVAR performance benefits substantially from using a longer assimilation window, LETKF is also able to achieve a satisfactory accuracy compared to the 24-h 4DVAR analyses. It is shown that the advantage of the LETKF over 3DVAR is a result of both the ensemble averaging and the information about the “errors of the day” provided by the ensemble. The analysis corrections at the end of the 12-h assimilation window are similar for LETKF and the 12-h window 4DVAR, and they both resemble bred vectors. At the beginning of the assimilation window, LETKF analysis corrections obtained using a no-cost smoother also resemble the corresponding bred vectors, whereas the 4DVAR corrections are significantly different with much larger horizontal scales.

A Sequential Variational Algorithm for Data Assimilation in Oceanography and Meteorology

Abstract This study theoretically establishes a sequential variational (SVAR) method for the data assimilation in oceanography and meteorology defined on the model space. Requiring a significantly smaller amount of computer memory, theoretically SVAR gives the same optimal model state estimate as the four-dimensional variational data assimilation method. Its computational cost is similar to that of the four-dimensional variational data assimilation and representer methods. In addition to the optimal state estimates, SVAR computes error covariances at the end of the assimilation window. These advantageous properties of the new algorithm are obtained by combining the sequential methodology with suitable definitions of several new l2 norms, which implicitly provide required estimates.

Coupling ensemble Kalman filter with four-dimensional variational data assimilation

This study examines the performance of coupling the deterministic four-dimensional variational assimilation system (4DVAR) with an ensemble Kalman filter (EnKF) to produce a superior hybrid approach for data assimilation. The coupled assimilation scheme (E4DVAR) benefits from using the state-dependent uncertainty provided by EnKF while taking advantage of 4DVAR in preventing filter divergence: the 4DVAR analysis produces posterior maximum likelihood solutions through minimization of a cost function about which the ensemble perturbations are transformed, and the resulting ensemble analysis can be propagated forward both for the next assimilation cycle and as a basis for ensemble forecasting. The feasibility and effectiveness of this coupled approach are demonstrated in an idealized model with simulated observations. It is found that the E4DVAR is capable of outperforming both 4DVAR and the EnKF under both perfect- and imperfect-model scenarios. The performance of the coupled scheme is also less sensitive to either the ensemble size or the assimilation window length than those for standard EnKF or 4DVAR implementations.

Identification of optimal topography of the barotropic ocean model in the North Atlantic by variational data assimilation

The use of the data assimilation technique to identify optimal topography is discussed in frames of time-dependent motion governed by nonlinear barotropic ocean model. As-similation of arti cially generated data allows to measure the in uence of various error sources and to classify the impact of noise that is present in observational data and model parameters. The choice of length of the assimilation window in 4DVar is discussed. It is shown that using longer window lengths would provide more accurate ocean topography. The topography de ned using this technique can be further used in other model runs that start from other initial conditions and are situated in other parts of the model's attractor.

Four-Dimensional Variational Data Analysis of Water Vapor Raman Lidar Data and Their Impact on Mesoscale Forecasts

Abstract The impact of water vapor observations on mesoscale initial fields provided by a triangle of Raman lidar systems covering an area of about 200 km × 200 km is investigated. A test case during the Lindenberg Campaign for Assessment of Humidity and Cloud Profiling Systems and its Impact on High-Resolution Modeling (LAUNCH-2005) was chosen. Evaluation of initial water vapor fields derived from ECMWF analysis revealed that in the model the highly variable vertical structure of water vapor profiles was not recovered and vertical gradients were smoothed out. Using a 3-h data assimilation window and a resolution of 10–30 min, continuous water vapor data from these observations were assimilated in the fifth-generation Pennsylvania State University–NCAR Mesoscale Model (MM5) by means of a four-dimensional variational data analysis (4DVAR). A strong correction of the vertical structure and the absolute values of the initial water vapor field of the order of 1 g kg−1 was found. This occurred mainly upstream of the lidar systems within an area, which was comparable with the domain covered by the lidar systems. The correction of the water vapor field was validated using independent global positioning system (GPS) sensors. Much better agreement to GPS zenith wet delay was achieved with the initial water vapor field after 4DVAR. The impact region was transported with the mean wind and was still visible after 4 h of free forecast time.

Toward a weak constraint operational 4D-Var system: application to the Burgers' equation

Strong constraint (perfect model assumption) 4D-Var algorithms are increasingly used for synoptic and global scale atmospheric data assimilation at operational numerical weather prediction centers around the world. The data assimilation windows currently range from 6 hours to 12 hours at different centers. It is preferable to have as many independent observations as possible in each data assimilation window under the variational framework. Longer data assimilation windows generally increase the information content from the observations, but also make the perfect model assumption more improper. It is clear that weak constraint (imperfect model assumption) 4D-Var algorithms will be required to properly combine the background forecast with high resolution observations in longer data assimilation windows in the not too distant future. A weak constraint observation space atmospheric 4D-Var system, NAVDAS-AR (NRL Atmospheric Variational Data Assimilation System - Accelerated Representer), is currently under development at the Naval Research Laboratory (NRL) in Monterey. The impact of the model errors is presented in a residual term resulting from a four-dimensional matrix/vector multiplication of the model error covariance with the adjoint field in the right hand side of the tangent linear model. The calculation of the residual term is generally very computationally intensive. However, it can be obtained accurately and efficiently when some special choices are made for the model error covariances. In this short contribution we present an approach to include the impact of the model errors in the NAVDAS-AR. The steps involved in this approach are demonstrated through an application to the Burgers' equation.

Model‐error estimation in 4D‐Var

AbstractCurrent operational implementations of 4D‐Var rely on the assumption that the numerical model representing the evolution of the atmospheric flow is perfect, or at least that model errors are small enough (relative to other errors in the system) to be neglected. This paper describes a formulation of weak‐constraint 4D‐Var that removes that assumption by explicitly representing model error as part of the 4D‐Var control variable.The consequences of this choice of control variable for the implementation of incremental 4D‐Var are discussed. It is shown that the background‐error covariance matrix cannot be used as an approximate model‐error covariance matrix. Another model‐error covariance matrix, based on statistics of model tendencies, is proposed.Experimental results are presented. They show that this approach to accounting for and estimating model error captures some known model errors and improves the fit to observations, both in the analysis and in the background, but that it also captures part of the observation bias. We show that the model error estimated in this approach varies rapidly, and cannot be used to correct medium‐ or long‐range forecasts. We also show that, because it relies on the tangent linear assumption for the entire assimilation window, the incremental formulation of weak‐constraint 4D‐Var is not the most suitable formulation for long assimilation windows. Copyright © 2007 Royal Meteorological Society

Initialization and simulation of a landfalling typhoon using a variational bogus mapped data assimilation (BMDA)

Recently, a new data assimilation method called “3-dimensional variational data assimilation of mapped observation (3DVM)” has been developed by the authors. We have shown that the new method is very efficient and inexpensive compared with its counterpart 4-dimensional variational data assimilation (4DVar). The new method has been implemented into the Penn State/NCAR mesoscale model MM5V1 (MM5_3DVM). In this study, we apply the new method to the bogus data assimilation (BDA) available in the original MM5 with the 4DVar. By the new approach, a specified sea-level pressure (SLP) field (bogus data) is incorporated into MM5 through the 3DVM (for convenient, we call it variational bogus mapped data assimilation – BMDA) instead of the original 4DVar data assimilation. To demonstrate the effectiveness of the new 3DVM method, initialization and simulation of a landfalling typhoon – typhoon Dan (1999) over the western North Pacific with the new method are compared with that with its counterpart 4DVar in MM5. Results show that the initial structure and the simulated intensity and track are improved more significantly using 3DVM than 4DVar. Sensitivity experiments also show that the simulated typhoon track and intensity are more sensitive to the size of the assimilation window in the 4DVar than that in the 3DVM. Meanwhile, 3DVM takes much less computing cost than its counterpart 4DVar for a given time window.

Impact of the Different Components of 4DVAR on the Global Forecast System of the Meteorological Service of Canada

Abstract A four-dimensional variational data assimilation (4DVAR) scheme has recently been implemented in the medium-range weather forecast system of the Meteorological Service of Canada (MSC). The new scheme is now composed of several additional and improved features as compared with the three-dimensional variational data assimilation (3DVAR): the first guess at the appropriate time from the full-resolution model trajectory is used to calculate the misfit to the observations; the tangent linear of the forecast model and its adjoint are employed to propagate the analysis increment and the gradient of the cost function over the 6-h assimilation window; a comprehensive set of simplified physical parameterizations is used during the final minimization process; and the number of frequently reported data, in particular satellite data, has substantially increased. The impact of these 4DVAR components on the forecast skill is reported in this article. This is achieved by comparing data assimilation configurations that range in complexity from the former 3DVAR with the implemented 4DVAR over a 1-month period. It is shown that the implementation of the tangent-linear model and its adjoint as well as the increased number of observations are the two features of the new 4DVAR that contribute the most to the forecast improvement. All the other components provide marginal though positive impact. 4DVAR does not improve the medium-range forecast of tropical storms in general and tends to amplify the existing, too early extratropical transition often observed in the MSC global forecast system with 3DVAR. It is shown that this recurrent problem is, however, more sensitive to the forecast model than the data assimilation scheme employed in this system. Finally, the impact of using a shorter cutoff time for the reception of observations, as the one used in the operational context for the 0000 and 1200 UTC forecasts, is more detrimental with 4DVAR. This result indicates that 4DVAR is more sensitive to observations at the end of the assimilation window than 3DVAR.

Variational assimilation of GPS precipitable water vapor and hourly rainfall observations for a meso-β scale heavy precipitation event during the 2002 mei-yu season

Recent advances in Global Positioning System (GPS) remote sensing technology allow for a direct estimation of the precipitable water vapor (PWV) from delayed signals transmitted by GPS satellites, which can be assimilated into numerical models with four-dimensional variational (4DVAR) data assimilation. A mesoscale model and its 4DVAR system are used to access the impacts of assimilating GPS-PWV and hourly rainfall observations on the short-range prediction of a heavy rainfall event on 20 June 2002. The heavy precipitation was induced by a sequence of meso-β-scale convective systems (MCS) along the mei-yu front in China. The experiments with GPS-PWV assimilation successfully simulated the evolution of the observed MCS cluster and also eliminated the erroneous rainfall systems found in the experiment without 4DVAR assimilation. Experiments with hourly rainfall assimilation performed similarly both on the prediction of MCS initiation and the elimination of erroneous systems, however the MCS dissipated much sooner than it did in observations. It is found that the assimilation-induced moisture perturbation and mesoscale low-level jet are helpful for the MCS generation and development. It is also discovered that spurious gravity waves may post serious limitations for the current 4DVAR algorithm, which would degrade the assimilation efficiency, especially for rainfall data. Sensitivity experiments with different observations, assimilation windows and observation weightings suggest that assimilating GPS-PWV can be quite effective, even with the assimilation window as short as 1 h. On the other hand, assimilating rainfall observations requires extreme cautions on the selection of observation weightings and the control of spurious gravity waves.

Impact of 4DVAR assimilation of rainfall data on the simulation of mesoscale precipitation systems in a mei-yu heavy rainfall event

The multi-scale weather systems associated with a mei-yu front and the corresponding heavy precipitation during a particular heavy rainfall event that occurred on 4{5 July 2003 in east China were successfully simulated through rainfall assimilation using the PSU/NCAR non-hydrostatic, mesoscale, numerical model (MM5) and its four-dimensional, variational, data assimilation (4DVAR) system. For this case, the improvement of the process via the 4DVAR rainfall assimilation into the simulation of mesoscale precipitation systems is investigated. With the rainfall assimilation, the convection is triggered at the right location and time, and the evolution and spatial distribution of the mesoscale convective systems (MCSs) are also more correctly simulated. Through the interactions between MCSs and the weather systems at difierent scales, including the low-level jet and mei-yu front, the simulation of the entire mei-yu weather system is significantly improved, both during the data assimilation window and the subsequent 12-h period. The results suggest that the rainfall assimilation flrst provides positive impact at the convective scale and the in∞uences are then propagated upscale to the meso- and sub-synoptic scales. Through a set of sensitive experiments designed to evaluate the impact of difierent initial variables on the simulation of mei-yu heavy rainfall, it was found that the moisture fleld and meridional wind had the strongest efiect during the convection initialization stage, however, after the convection was fully triggered, all of the variables at the initial condition seemed to have comparable importance.

Impact of Configurations of Rapid Intermittent Assimilation of WSR-88D Radar Data for the 8 May 2003 Oklahoma City Tornadic Thunderstorm Case

Abstract Various configurations of the intermittent data assimilation procedure for Level-II Weather Surveillance Radar-1988 Doppler radar data are examined for the analysis and prediction of a tornadic thunderstorm that occurred on 8 May 2003 near Oklahoma City, Oklahoma. Several tornadoes were produced by this thunderstorm, causing extensive damages in the south Oklahoma City area. Within the rapidly cycled assimilation system, the Advanced Regional Prediction System three-dimensional variational data assimilation (ARPS 3DVAR) is employed to analyze conventional and radar radial velocity data, while the ARPS complex cloud analysis procedure is used to analyze cloud and hydrometeor fields and adjust in-cloud temperature and moisture fields based on reflectivity observations and the preliminary analysis of the atmosphere. Forecasts for up to 2.5 h are made from the assimilated initial conditions. Two one-way nested grids at 9- and 3-km grid spacings are employed although the assimilation configuration experiments are conducted for the 3-km grid only while keeping the 9-km grid configuration the same. Data from the Oklahoma City radar are used. Different combinations of the assimilation frequency, in-cloud temperature adjustment schemes, and the length and coverage of the assimilation window are tested, and the results are discussed with respect to the length and evolution stage of the thunderstorm life cycle. It is found that even though the general assimilation method remains the same, the assimilation settings can significantly impact the results of assimilation and the subsequent forecast. For this case, a 1-h-long assimilation window covering the entire initial stage of the storm together with a 10-min spinup period before storm initiation works best. Assimilation frequency and in-cloud temperature adjustment scheme should be set carefully to add suitable amounts of potential energy during assimilation. High assimilation frequency does not necessarily lead to a better result because of the significant adjustment during the initial forecast period. When a short assimilation window is used, covering the later part of the initial stage of storm and using a high assimilation frequency and a temperature adjustment scheme based on latent heat release can quickly build up the storm and produce a reasonable analysis and forecast. The results also show that when the data from a single Doppler radar are assimilated with properly chosen assimilation configurations, the model is able to predict the evolution of the 8 May 2003 Oklahoma City tornadic thunderstorm well for up to 2.5 h. The implications of the choices of assimilation settings for real-time applications are discussed.