Four-dimensional Data Assimilation Research Articles

The Computational Complexity and Parallel Scalability of Atmospheric Data Assimilation Algorithms

Abstract This paper quantifies the computational complexity and parallel scalability of two algorithms for four-dimensional data assimilation (4DDA) at NASA's Global Modeling and Assimilation Office (GMAO). The first, the Goddard Earth Observing System Data Assimilation System (GEOS DAS), uses an atmospheric general circulation model (GCM) and an observation-space-based analysis system, the Physical-Space Statistical Analysis System (PSAS). GEOS DAS is very similar to global meteorological weather forecasting data assimilation systems but is used at NASA for climate research. The second, the Kalman filter, uses a more consistent algorithm to determine the forecast error covariance matrix than does GEOS DAS. For atmospheric assimilation, the gridded dynamical fields typically have more than 106 variables; therefore, the full error covariance matrix may be in excess of a teraword. For the Kalman filter this problem will require petaflop s−1 computing to achieve effective throughput for scientific research.

Evaluation of Regional Climate Simulations of the 1998 and 1999 East Asian Summer Monsoon using the GAME/HUBEX Observational Data

A regional climate model based on the Penn State/NCAR Mesoscale Model (MM5) was used to simulate the 1998 and 1999 East Asian summer monsoon conditions. Simulations were performed for 1 April-31 August of each year, with initial and lateral boundary conditions provided by the ECMWF analysis. Observations from the 1998 and 1999 GAME/HUBEX experiments were used to evaluate the regional climate simulations. Based on observations, large differences can be found between the 1998 and 1999 meteorological conditions and surface energy budgets at the Shouxian station during the IOPs, with much higher rain intensity but only slightly higher rain frequency in 1998 than 1999. For 1998, although the regional climate model was able to reproduce the general spatial distribution of monthly mean rainfall quite well during the summer monsoon season, large discrepancies can be found in comparing the observed and simulated surface climate and energy fluxes in the HUBEX region. By using Four Dimensional Data Assimilation (FDDA) technique, which constrains the simulated large-scale circulation with observations from 21 soundings in the HUBEX α-scale region, both the root mean square error and mean bias in rainfall were greatly reduced. The improvements in simulating rainfall were related to both reduction in errors of precipitation amount and timing. In the control simulation, a mean bias of −63 W/m² (−36%) was found in the simulated surface net radiation at Shouxian, which suggest large errors in simulating clouds in the region. With FDDA, the bias was significantly reduced to −23 W/m² (−13%), with corresponding reduction of bias in the latent heat flux. This suggests that at least part of the model bias in simulating net radiation is related to errors in simulating the large-scale circulation, which can affect cloud amount and vertical distribution.Comparing the 1998 and 1999 simulations, both without FDDA, smaller biases were found in the surface fluxes during 1999. Percentage biases in the net radiation and latent heat flux were −18% and −33% in 1999 and −36% and −50% in 1998 respectively. Based on observations, large differences in the net surface radiation, and small differences in cloud fraction between the two years suggest that cloud optical depth and/or vertical distribution were very different, with more cloudy conditions observed during 1999. Although the 1999 simulations were sensitive to the cumulus convective parameterizations (Grell scheme versus Kain-Fritsch scheme) as shown by the sensitivity experiments, the large differences in simulation skill between the 1998 and 1999 cases, regardless of the convection schemes used, suggest possible dependence of model errors on cloud properties that deserve further investigations.

Seasonal Contrasting Features of Heat and Moisture Budgets between the Eastern and Western Tibetan Plateau during the GAME IOP

Abstract Using the Global Energy and Water Cycle Experiment (GEWEX) Asian Monsoon Experiment (GAME) four-dimensional data assimilation (4DDA) upper-air data, the large-scale heat source (Q1) and moisture sink (Q2) over the western and eastern Tibetan Plateau are examined for a 4-month period from 1 May to 31 August 1998. The computations were performed on the sigma–pressure hybrid coordinate, named η-coordinate, since the analysis area includes high-elevation mountains. Over the western Tibetan Plateau in May, there is a deep layer of heating occupying the whole troposphere with the maximum value exceeding 3 K day−1 around 400–600 hPa. The smaller magnitude of the apparent moisture sink is confined in the lower troposphere 1 km above the ground surface. Vertically integrated heat sources of 103 W m−2 over the western Tibetan Plateau are accompanied by a moisture sink of about half that (60 W m−2). These results indicate that the latent heat release associated with condensation plays an important role in the total heating besides the sensible heat supply from the land surface. Later in July, the moisture sink over the eastern Tibetan Plateau nearly equals the heat source indicating the dominance of moist processes associated with summertime monsoon rains. The contrasting features of the heat source and moisture sink are closely related to the circulation fields. Throughout May and June, we observe strong upward motion along the western and southwestern slopes of the western Tibetan Plateau, while there is salient subsidence motion over the eastern Tibetan Plateau. The analyses of static stability and lifting condensation level indicate that the release of latent heat relevant to moist convection is a dominant factor for tropospheric heating after the monsoon onset, while the premonsoon period (May) is composed of both convective rainfall and dry thermal convection. Thus, the heating mechanism prior to the onset of the monsoon, especially over the western Tibetan Plateau, can be characterized by the hybrid nature of “wet” processes due to condensation heating and “dry” processes associated with the sensible heat flux from the elevated mountain surface.

Methods and examples for remote sensing data assimilation in land surface process modeling

Land surface process models describe the energy, water, carbon, and nutrient fluxes on a local to regional scale using a set of environmental land surface parameters and variables. They need time series of spatially distributed inputs to account for the large spatial and temporal variability of land surface processes. In principle many of these inputs can be derived through remote sensing using both optical and microwave sensors. New approaches in four-dimensional data-assimilation (4DDA) form the basis to combine remote sensing data and spatially explicit land surface process models more effectively. This paper describes basic techniques for 4DDA in land surface process modeling. Two case studies were carried out to demonstrate different successful approaches of remote sensing data assimilation into land surface process models. The assimilation of surface soil moisture estimates from European Remote Sensing (ERS) synthetic aperture radar data in a flood forecasting scheme is presented, as well as the combination of a land surface process model and a radiative transfer model to improve the accuracy of land surface parameter retrieval from optical data [Landsat Thematic Mapper (TM)].

Application of Cost Function Method for Combining Measured Data and CFD to Unsteady Flow Field

In order to reduce measurement errors and fill missing data points, we have already proposed the cost function method (CFM) in which governing equations are applied to the correction and interpolation of measured data. In this paper, in order to study the applicability to the unsteady flow fields, we compare the properties of CFM with those of the four-dimensional data assimilation (FDDA), which is a method that is conventionally applied to unsteady flow fields in the field of numerical weather prediction. It was found that CFM produces a more suitable solution that does FDDA, according to our evaluation criteria. Then we applied PIV to a Karman vortex flow behind a plate and applied CFM to the data obtained. Furthermore, we examined the effect of the number of measured data points on the CFM results. It was found that the CFM is effective in correcting errors in the data, and also in filling missing data points smoothly both spatially and temporally. It is difficult to model a complicated flow field correctly using PIV or CFD separately. However, we can conclude from the results of this study that for the estimation of flow fields, CFM may be an effective technique also for unsteady flow.

MASINGAR, a global tropospheric aerosol chemical transport model coupled with MRI/JMA98 GCM: Model description

This paper presents the first description of a new three-dimensional aerosol chemical transport model, called the Model of Aerosol Species IN the Global AtmospheRe (MASINGAR), which has been developed to study the distributions of atmospheric aerosols and related trace species. MASINGAR is an on-line chemical transport model (CTM) coupled with the MRI/JMA98 GCM. MASINGAR includes nss-sulfate, carbonaceous, mineral dust, and sea-salt aerosols, and accounts for advective transport, subgrid-scale eddy diffusive and convective transport, surface emission, and dry/wet depositions, as well as chemical reactions. The advective transport is calculated using the semi-Lagrangian transport scheme. Parameterization of convective transport is based on the convective mass flux derived by the Arakawa-Schubert scheme. The space and time resolutions of the model are variable, with a standard spatial resolution of T42(2.8°×2.8°) and 30 vertical layers (up to 0.8hPa) with a 20-minute time step. In addition, the model has a built-in, four-dimensional data assimilation (FDDA) system with a nudging scheme incorporating an assimilated meteorological field, which enables the model to realistically simulate a specific period and a short-period forecast of aerosols. The model simulation of mineral dust aerosol in April 2002 suggests that MASINGAR simulates the synoptic scale aerosol events.

Application of MM5 model in the northwest area of Greece using a four-nest procedure

MM5 is a meteorological modelling system characterized by multiple nesting capability, nonhydrostatic dynamics, four-dimensional data assimilation capability as well as enhanced physical parameterisation options combined with portability to a wide range of computing platforms. MM5 has been applied in many cases across the world. The modelling system, using four nests, was applied for the first time in the Northwest area of Greece, focusing the area of Epirus, a region characterized by highly complex topography as well as intensely variable land-water and land-use distribution. The application reveals regional, as well as, mesoscale circulation patterns and air flow phenomena that could not be detected by the poorly organized meteorological network of the area.

GFDL-Type Typhoon Initialization in MM5

Abstract The Geophysical Fluid Dynamics Laboratory (GFDL) hurricane initialization algorithm is implemented in the community fifth-generation Pennsylvania State University–National Center for Atmospheric Research Mesoscale Model (MM5). This work is applied to the MM5-based Regional Data Assimilation and Prediction System model (RDAPS), the Korea Meteorological Administration's regional forecast model. The bogus procedure starts by initializing the winds within the bogus area. The main difficulty lies in the generation of other variables, such as humidity, temperature, geopotential, etc., which are dynamically consistent with the prescribed wind. However, it was found that there is a simple and practical way of tropical cyclone (TC) initialization. It is achieved by the use of the built-in function of MM5, the four-dimensional data assimilation (FDDA). In order to do so, a miniature RDAPS is constructed. After the initialization of wind within the filter region, all other variables are generated by the mod...

On the Smoothness Constraints for Four-Dimensional Data Assimilation

An algorithm for determination of the weights of smoothness penalty constraints for the four-dimensional variational data assimilation technique is proposed and evaluated. To study the nature of smoothness penalty constraints, a simple nonlinear harmonic oscillator problem is first considered. Penalizing smoothness constraints is found to make the modified Hessian matrix of the cost function more positive definite, akin to the idea behind the modified line search Newton's methods. However, the use of the derivative smoothness constraints with a fixed coefficient does not warrant uniform imposition of these constraints at every iteration. A remedy is to control the ratio of the smoothness penalty function over the cost function, which can dramatically increase the positive definite area. On the other hand, the large smoothness coefficients obtained from this approach can deteriorate the convergence property of the minimization problem. Based on these observations, an algorithm for tuning the weights of smoothness constraints is proposed to overcome the aforementioned problems. The algorithm is first applied to a simple dynamic problem. It is then tested on the retrieval of microscale turbulent structures in a simulated convective boundary layer. This method is further evaluated on the retrieval of a strong meso-scale thunderstorm outflow from Doppler radar data. The results show that the algorithm yields efficient retrieval.

Numerical simulation of a super cyclonic storm, Orissa 1999: impact of initial conditions

AbstractNumerical simulations are performed using the Penn State University/ National Center for Atmospheric Research Mesoscale Model (MM5) to study the impact of initial conditions on the super cyclone which hit the coast of Orissa in 1999. Because analysis of the cyclone's circulation was inadequate in the initial fields owing to the coarse resolution of the operational analysis systems and sparse oceanic data coverage, synthetic vortex data were generated using empirical relations and used in the analysis. Four‐dimensional data assimilation is performed in order to assimilate the synthetic vortex in the initial stage to the model.Considerable improvement in the track is obtained by using the synthetic vortex. With better specification of the initial vortex structure, the model successfully simulated the typical tropical cyclone characteristics, such as asymmetries in the wind field: the strongest winds occurred in the east and close to the cyclone's centre, strong wind gradients were found between the centre and the maximum wind region, and there was a slow decrease in wind speed up to the middle troposphere. Despite failing to produce the intense pressure drop observed for this cyclone, the model shows much better cyclone development with enhanced initial condition than the analysis. Copyright © 2002 Royal Meteorological Society.

The Sensitivity of Precipitation to Circulation Details. Part II: Mesoscale Modeling

Abstract A base case and the best time-coherent, regional analog to that case, identified in Part I, were analyzed to determine the precise mechanisms responsible for the differences in precipitation amounts. This analysis was conducted using two modeling approaches. A model dataset that provides time continuity and dynamic coupling among the various fields was generated using a continuous four-dimensional data assimilation procedure for 60-h integrations within the Pennsylvania State University–National Center for Atmospheric Research fifth generation Mesoscale Model (PSU–NCAR MM5). A nonhydrostatic, axisymmetric cloud model with large-scale convergent forcing was also used to perform selected experiments to assess the triggering of convection in a controlled setting. The findings from these analyses are the following. Differences in moisture source origins of parcels arriving within the precipitation region, observationally linked in Part I to variations in the structure and intensity of the subtropical...

Increasing vulnerability to extreme floods and societal needs of hydrological forecasting

Recent climatic variability seems to be intensifying the hydrological extremes resulting in devastating hydrological hazards, especially floods all over the world, including the arid regions. Disasters are always more tragic in poor countries than in rich ones, as more poor people seem to suffer and die, and the economic losses have a greater impact. On the other hand, in advanced countries, although flood control can reduce the number of deaths, it cannot always reduce economic losses despite the large public investments that have long been applied. This is due to the ever increasing damage potential in a basin where reliance on safety has been created by flood control measures. It is proven that proper prediction and preparedness are the only practical means that can be relied upon to reduce the death toll and economic losses. In a time of climatic variation, physically-based distributed models are to be pursued to obtain reasonable predictions, since the climate is non-stationary and both basin land use and water use are constantly changing. The proper integration of a distributed watershed model with atmospheric circulation models, four-dimensional data assimilation and mesoscale hydrometeorological nested models is the direction for hydrologists to be heading.

The 1998 Ice Storm—Analysis of a Planetary-Scale Event

Abstract The ice storm of 5–9 January 1998, affecting the northeastern United States and the eastern Canadian provinces, was characterized by freezing rain amounts greater than 100 mm in some areas. The event was associated with a 1000–500-hPa positive (warm) thickness anomaly, whose 5-day mean exceeded +30 dam (+15°C) over much of New York and Pennsylvania. The region of maximum precipitation occurred in a deformation zone between an anomalously cold surface anticyclone to the north and a surface trough axis extending from the Gulf of Mexico into the Great Lakes. The thermodynamic impact of this unprecedented event was studied with the use of a four-dimensional data assimilation spanning an 18-day period ending at 0000 UTC 9 January 1998. A moisture budget for the precipitation region reveals the bulk of the precipitation to be associated with the convergence of water vapor transport throughout the precipitation period. The ice storm consisted of two primary synoptic-scale cyclonic events. The first even...

Emission Strength Validation Using Four-Dimensional Data Assimilation: Application to Primary Aerosol and Precursors to Ozone and Secondary Aerosol

ABSTRACT Three-dimensional air quality models (AQMs) represent the most powerful tool to follow the dynamics of air pollutants at urban and regional scales. Current AQMs can account for the complex interactions between gas-phase chemistry, aerosol growth, cloud and scavenging processes, and transport. However, errors in model applications still exist due in part to limitations in the models themselves and in part to uncertainties in model inputs. Four-dimensional data assimilation (FDDA) can be used as a top-down tool to validate several of the model inputs, including emissions inventories, based on ambient measurements. Previously, this FDDA technique was used to estimate adjustments in the strength and composition of emissions of gas-phase primary species and O3 precursors. In this paper, we present an extension to the FDDA technique to incorporate the analysis of particulate matter (PM) and its precursors. The FDDA approach consists of an iterative optimization procedure in which an AQM is coupled to an inverse model, and adjusting the emissions minimizes the difference between ambient measurements

The Impact of NORPEX Targeted Dropsondes on the Analysis and 2–3-Day Forecasts of a Landfalling Pacific Winter Storm Using NCEP 3DVAR and 4DVAR Systems

Abstract During the North Pacific Experiment (NORPEX), both the Navy Operational Global Atmospheric Prediction System and the National Centers for Environmental Prediction (NCEP) operational forecast systems found a 48-h forecast degradation over the NORPEX forecast verification region due to the inclusion of a set of NORPEX targeted dropsondes deployed north of Hawaii during 29–30 January 1998. The NCEP three- and four-dimensional varitional data assimilation (3DVAR and 4DVAR) systems are used here to reassess the impact of these dropsonde observations on model predictions. The assimilation of these targeted dropsondes excluding the conventional observations improved the 48-h forecast over the NORPEX forecast verification region. However, the addition of the dropsonde data to an analysis that already contained various conventional observations degraded the 48-h forecast over the NORPEX forecast verification region. In the later case, the dropsonde data still improved and had its largest impact on the for...

A review on the use of the adjoint method in four-dimensional atmospheric-chemistry data assimilation

A review on the use of the adjoint method in four-dimensional atmospheric-chemistry data assimilation

Estimation of emission adjustments from the application of four-dimensional data assimilation to photochemical air quality modeling

Four-dimensional data assimilation applied to photochemical air quality modeling is used to suggest adjustments to the emissions inventory of the Atlanta, Georgia metropolitan area. In this approach, a three-dimensional air quality model, coupled with direct sensitivity analysis, develops spatially and temporally varying concentration and sensitivity fields that account for chemical and physical processing, and receptor analysis is used to adjust source strengths. Proposed changes to domain-wide NO x , volatile organic compounds (VOCs) and CO emissions from anthropogenic sources and for VOC emissions from biogenic sources were estimated, as well as modifications to sources based on their spatial location (urban vs. rural areas). In general, domain-wide anthropogenic VOC emissions were increased approximately two times their base case level to best match observations, domain-wide anthropogenic NO x and biogenic VOC emissions (BEIS2 estimates) remained close to their base case value and domain-wide CO emissions were decreased. Adjustments for anthropogenic NO x emissions increased their level of uncertainty when adjustments were computed for mobile and area sources (or urban and rural sources) separately, due in part to the poor spatial resolution of the observation field of nitrogen-containing species. Estimated changes to CO emissions also suffer from poor spatial resolution of the measurements. Results suggest that rural anthropogenic VOC emissions appear to be severely underpredicted. The FDDA approach was also used to investigate the speciation profiles of VOC emissions, and results warrant revision of these profiles. In general, the results obtained here are consistent with what are viewed as the current deficiencies in emissions inventories as derived by other top-down techniques, such as tunnel studies and analysis of ambient measurements.

Numerical Investigation of Boundary-Layer Evolution and Nocturnal Low-Level Jets: Local versus Non-Local PBL Schemes

Numerical simulations of the evolution of the planetary boundary layer (PBL) and nocturnal low-level jets (LLJ) have been carried out using MM5 (version 3.3) with four-dimensional data assimilation (FDDA) for a high pollution episode in the northeastern United States during July 15–20, 1999. In this paper, we assess the impact of different parameterizations on the PBL evolution with two schemes: the Blackadar PBL, a hybrid local (stable regime) and non-local (convective regime) mixing scheme; and the Gayno–Seaman PBL, a turbulent kinetic energy (TKE)-based eddy diffusion scheme. No FDDA was applied within the PBL to evaluate the ability of the two schemes to reproduce the PBL structure and its temporal variation. The restriction of the application of FDDA to the atmosphere above the PBL or the lowest 8 model levels, whichever is higher, has significantly improved the predicted strength and timing of the LLJ during the night. A systematic analysis of the PBL evolution has been performed for the primary meteorological fields (temperature, specific humidity, horizontal winds) and for the derived parameters such as the PBL height, virtual potential temperature, relative humidity, and cloud cover fraction. There are substantial differences between the PBL structures and evolutions simulated by these two different schemes. The model results were compared with independent observations (that were not used in FDDA) measured by aircraft, RASS and wind profiler, lidar, and tethered balloon platforms during the summer of 1999 as part of the NorthEast Oxidant and Particle Study (NE-OPS). The observations tend to support the non-local mixing mechanism better than the layer-to-layer eddy diffusion in the convective PBL.

Mesoscale Moisture Analysis of the North American Monsoon

Abstract The regional circulations that contribute moisture to the large precipitation over northwestern Mexico, the core region of the North American monsoon, are investigated using three summer seasons (July–September 1995–97) of Eta Model mesoscale analyses and forecasts. Analyses are produced by the Eta Model’s own four-dimensional data assimilation system that includes a diverse mix of observations. Comparison of the forecast precipitation with satellite estimates and previous observational studies shows similarity in location, shape, and scale of the patterns over northwestern Mexico; the magnitude of the precipitation over the slopes of the Sierra Madre Occidental is also similar to that from climatologies based on rain gauge observations. Examination of the morning and evening forecast precipitation also reveals agreement with equivalent estimates from high-resolution satellites. Excessive model forecast precipitation is found over the Isthmus of Tehuantepec in eastern Mexico, which seems related,...

Iterative Inverse Modeling and Direct Sensitivity Analysis of a Photochemical Air Quality Model

A four-dimensional data assimilation approach that combines emission-based modeling and inverse modeling techniques has been developed to provide a method to help identify emissions inventory improvements. The method is based on linking formal direct sensitivity analysis of three-dimensional air quality models with an inverse modeling technique such that observational data of multiple species can be used to suggest improvements in emission strengths, patterns, and compositions of various source categories simultaneously. Information regarding the characteristics of the data and the emissions is incorporated into the model by means of weighting factors. A penalty function allows determining if the method is fully observation-driven, emission-driven, or mixed. The assimilation of the data requires an iterative process since the sensitivity coefficients change as emissions are adjusted, though results suggest that only three or four iterations are necessary. The method has been applied to the Atlanta Metro Area and tested with pseudo-observations. Perturbed emissions were adjusted close to their known original value in these test scenarios. When noisy pseudo-observations were used, the final simulated concentration errors were of the order of the noise applied to the pseudo-observations.