7
Conclusions and Recommendations

The panel has reached four major conclusions. These conclusions are stated below, followed by a discussion setting forth the basis for each conclusion.

CONCLUSIONS

  1. Four-dimensional (space and time) data assimilation as a subdiscipline of geophysical sciences is fundamental for the synthesis of diverse, temporally inconsistent, and spatially incomplete observations into a coherent representation of an evolving geophysical system.

Within the past decade, data assimilation has emerged as a robust strategy whereby prior information and current information are combined to describe a geophysical system. Temporal and spatial consistency enters in the description through a numerical assimilating model that possesses predictive, quality control, and validation capabilities, based on the governing equations for the given geophysical systems. Data assimilation produces data sets that, by virtue of their integrity and consistency, provide information and value that significantly exceed those of the incoming observations. The method provides for systematic validation of observations, synthesized data fields, and the prediction model itself.



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 61
Four-Dimensional Model Assimilation of Data: A Strategy for the Earth System Sciences 7 Conclusions and Recommendations The panel has reached four major conclusions. These conclusions are stated below, followed by a discussion setting forth the basis for each conclusion. CONCLUSIONS Four-dimensional (space and time) data assimilation as a subdiscipline of geophysical sciences is fundamental for the synthesis of diverse, temporally inconsistent, and spatially incomplete observations into a coherent representation of an evolving geophysical system. Within the past decade, data assimilation has emerged as a robust strategy whereby prior information and current information are combined to describe a geophysical system. Temporal and spatial consistency enters in the description through a numerical assimilating model that possesses predictive, quality control, and validation capabilities, based on the governing equations for the given geophysical systems. Data assimilation produces data sets that, by virtue of their integrity and consistency, provide information and value that significantly exceed those of the incoming observations. The method provides for systematic validation of observations, synthesized data fields, and the prediction model itself.

OCR for page 61
Four-Dimensional Model Assimilation of Data: A Strategy for the Earth System Sciences Viewed as an integral element of the scientific process, four-dimensional model assimilation of geophysical data is a systematic, quantitative, objective, iterative means of inference and testing aimed at advancing understanding and prediction of nonlinear dynamical geophysical systems where interactions occur continually among relevant physical, chemical, and biogeochemical processes. Use of data assimilation will advance understanding of geophysical phenomena governed by nonlinear, time-dependent dynamics as opposed to time-invariant conditions. Through systematic confrontation between observations and a priori understanding, as expressed by nonlinear governing equations and conceptual models, data assimilation provides a unique learning tool for promoting scientific understanding of the earth system. The learning process involved is inherently open ended and iterative. Current developments in data assimilation methodology will permit the use, continuously in time, of data that are nonlinearly related to model variables. Data assimilation is the only objective way to synthesize four-dimensional, multifaceted data that describe atmospheric, oceanic, and land surface processes. Systematic and iterative confrontation of model-predicted states on all time scales with observations (the learning process) has exposed and will continue to expose defects in both the data and the models. This confrontation leads to model improvement by identifying missing processes and critical observation gaps, by providing estimates of the magnitude of individual processes, and by providing estimates of the feedbacks between processes. The learning process leads to improvements in observational systems and sampling strategies. Within this iterative process, the evaluation of assimilated data by research users, including insertion of limitations, shortcomings, and critical comments into the archive record of the data sets, is important. Such action should become a routine step in studies in order to stem the propagation of error in subsequent analysis should error be established. Since the physical and dynamical consistency of model-assimilated data sets results in a level of information and added value that significantly exceeds that of the incoming observations, assimilation data sets have been and will be used even more extensively by the scientific community for diagnostic, predictive, and process studies, supplemented by original observations when needed. A major part of the progress achieved in operational weather prediction in recent years must be credited to improvements in data assimilation and the increasing use of model-assimilated and simulated data sets in research.

OCR for page 61
Four-Dimensional Model Assimilation of Data: A Strategy for the Earth System Sciences For example, the data sets produced by the National Oceanic and Atmospheric Administration's (NOAA) Geophysical Fluid Dynamics Laboratory (GFDL) and the European Centre for Medium Range Weather Forecasts (ECMWF), along with the operational global, hemispheric, and regional data sets produced by the National Meteorological Center (NMC), have been the principal basis for improvements in forecast skill, weather prediction research, studies of climatic processes, diagnostic studies of the planetary circulation, and interacting, coupled atmospheric-oceanic models. Studies of mesoscale circulation systems in the atmosphere and oceans are increasingly using both assimilated and simulated data sets produced by high-resolution models. Model-assimilated data sets are internally consistent with the dynamics and physical processes of the atmosphere as represented by the models employed at the operational and research centers. These data sets use all available observations to provide the best possible description of the atmosphere in a compact, orderly, gridded, or spectral format. As such, they form an exceedingly valuable resource that has become the data set of choice for the great majority of atmospheric researchers. Model-assimilated data sets for the ocean, land surface, and other components of climate change and global change processes should prove to be equally valuable for research once they have been developed. Atmospheric model data assimilation is a proven strategy in an advanced state of development. Although still in an early stage of development for the earth sciences as a whole, model data assimilation is strategically situated to address the pressing national need to observe and understand global change as it occurs in the coming decades. The immediate need is to develop and provide a nationally focused capability to synthesize, test, and utilize for understanding and prediction the information from the greatly enhanced new ground-based and satellite observing systems already scheduled for the next two decades. Implementation of advanced observation technology for the earth sciences is rapidly surpassing the capability of the scientific community to manage and utilize effectively the potentially overwhelming output of global data in the coming decades. NOAA, for example, is rapidly implementing, as part of the National Weather Service (NWS) Modernization Program, new networks of WSR-88D Doppler radars, atmospheric profilers, and advanced satellites that together will increase the flow of meteorological and oceanic data by up to 100 times. The National Aeronautics and Space Administration's (NASA) Earth Observing System (EOS), together with the Upper Atmosphere Research Satellite (UARS), will provide obser-

OCR for page 61
Four-Dimensional Model Assimilation of Data: A Strategy for the Earth System Sciences vations at a rate and volume never before seen. These data streams must be managed through a nationally focused program that produces model-assimilated data sets that provide a high-quality, valuable, coherent, and integrated understanding of the earth system as a whole. RECOMMENDATIONS The panel recommends the following strategy for a nationally focused program for four-dimensional model assimilation of data for the earth system. The strategy for the nationally focused program will:s Develop and expand applications of model-based data assimilation efforts to interdisciplinary areas that are necessary for integrated earth-ocean-atmosphere-biogeochemical models. Implement an integrated, multicenter, nationally focused archive system for model-assimilated and-tested data sets and provide for ready access to these sets by the scientific community over the long term. Provide for continuing scientific exchange and collaboration among the various groups and individuals engaged in geophysical data assimilation, particularly scientists involved with the Earth Observing System Data and Information System (EOSDIS); the national meteorological, oceanographic, and climate centers; and the High Performance Computer and Communications Initiative. Provide for the generation of routine research-quality, model-assimilated and-tested geophysical data sets to serve a broad range of national endeavors, including climate and global change research and predictions. Establish a working group to develop an implementation plan for this nationally focused program. Given the growing national focus on societal problems stemming from climate and global change, and the greatly enhanced observational capabilities already planned and scheduled for the 1990s and beyond, the need exists nationally to utilize global geophysical information to address these problems. This need to detect and monitor climate and global change as it occurs places unusual demands on both observing and data assimilation capabilities. Development and application of global and regionally assimilated data sets prepared by state-of-the-art prediction models for understanding the earth-ocean-atmosphere system and the physical processes that determine its evolution require a nationally focused program. To date, the

OCR for page 61
Four-Dimensional Model Assimilation of Data: A Strategy for the Earth System Sciences production of gridded analyses using data assimilation has largely been driven by numerical weather prediction activities located within operational weather centers. The production of optimum research-quality assimilated sets requires a delayed mode of analysis whereby all available relevant geophysical data are utilized. Within this larger effort, use of the quality control and validation information gained from operational assimilation is essential as a two-step process for production of optimum global data sets. Such efforts must be embodied in a nationally focused program to meet the varied needs of academic, government, and public sectors concerned with climate and global change, in making important economic and political policy decisions to cope with the rate of change predicted. Model data assimilation applied to these issues would provide the capability to synthesize these heterogeneous data consistently and objectively. A nationally focused program for data assimilation would ensure an archiving system that would make these coherent data sets readily available for the broad range of national needs. The strategy and implementation plan should also include the following specific actions. To validate and maintain quality control of new types of remotely sensed and experimental in situ geophysical data, including research data from field programs, the application of operational data assimilation models is essential. Funding agencies should routinely provide sufficient funds and computer capacity for this purpose, including provision for timely communication of such new and experimental data sets to designated operational assimilation centers. Over the past two decades, operational data assimilation for global weather forecasting has been developed to the point where it is invaluable for quality control and validation of remotely sensed and experimental data. The method ensures an internally consistent synthesis of all available data and includes the means for validating observational data streams against other data types. These capabilities have been systematically exploited to identify errors and biases in many types of in situ and remotely sensed data. Quality control and validation information from operational assimilation are important for producing research-quality data sets. Validations, both global and regional, performed in this way consider a broader range of atmospheric conditions than traditional field-validation campaigns. The ready provision of new research data to operational centers can provide for the timely preparation of assimilated data sets, including immediate quality control and consistency checks, the timely nature of which is important for the conduct of experiments. As such, this new approach to validation and quality monitoring is a valuable supplement to field cam-

OCR for page 61
Four-Dimensional Model Assimilation of Data: A Strategy for the Earth System Sciences paigns. These benefits can only be realized, however, if the research data are transmitted to suitable operational centers in a timely fashion and assimilated data sets are likewise returned. A coordinated national program should be implemented and funded to develop consistent, long-term assimilated data sets (extended back to about 1950) for the study of climate and global change. This effort will reanalyze with a state-of-the-art global-and regional-scale data assimilation model all atmospheric and oceanic data available since about 1950 in order to produce the best possible, validated, temporally and spatially consistent data sets for the study of climate and global change. Model biases should be identified and eliminated as far as possible at this time. This reanalysis should be repeated as advances in the state of the art require. Existing global model-assimilated data sets for climate studies are compromised by changes in assimilating models, in methods of dealing with raw data, in the observing networks and systems, and in assimilation procedures. These factors introduce discontinuities and inconsistencies in long time series of model-assimilated geophysical data, originally produced for weather prediction purposes, that make them only marginally useful for climate and global change analysis and predictions. An analysis of the entire useful global climate record from about 1950, using a single state-of-the-art data assimilation system, is the optimum means for producing temporally and spatially consistent and continuous global model-assimilated data sets useful for climate and global change purposes. Since data sets of this quality do not currently exist, a comprehensive reanalysis of the existing observed and analyzed data is needed now. During this effort, changes in observing capabilities (such as the inclusion of global satellite data beginning in the 1970s) can be more accurately assimilated and the results reflected in statistics of climate and global change. The lack of satellite data globally in the 1950s and early 1960s, as well as other changes in observing capabilities, introduces some degree of uncertainty in the existing data sets. As assimilation strategies and computers are developed in the future, further reanalyses with more advanced data assimilation models will be more useful. Future reanalyses will provide an orderly way to maintain consistency and eliminate model biases, thus increasing the reliability of the climatic record. The model used in any extended analysis or reanalysis of the record should be the best available at the time of analysis. Currently, such an analysis could probably be conducted using an operational or advanced research global atmospheric general circulation model (GCM) with sea surface temperature (SST) specified by the observed values. Future analyses

OCR for page 61
Four-Dimensional Model Assimilation of Data: A Strategy for the Earth System Sciences will require increasingly sophisticated models with more processes included, eventually culminating in a fully coupled atmosphere-ocean-land-biosphere-cryosphere model that assimilates all forms of data (National Research Council, 1990). To make these long climatic analyses possible, the observed data must be collected and archived in uniform and accessible formats, with comprehensive documentation of instrumental calibration and other characteristics that affect data quality. Because each successive analysis would be of the entire climate record to that date, special efforts to archive observed data in a convenient, compact electronic format will be needed. Data assimilation models for the mesoscale should be developed in concert with regional prediction models. The systems should integrate data from the enhanced observational capabilities of the 1990s that will be provided through the NWS Modernization Program, advanced satellite systems, and the EOS. These models should be capable of integrating specially observed data with the conventional data stream and also provide for realistic responses to various types of system forcing for the time and space scales emphasized. The mesoscale assimilation systems should also ensure effective nesting with larger-scale analysis systems. Use of model-based assimilation systems for the generation of mesoscale data sets is especially appropriate because many new mesoscale data sources can resolve smaller spatial and temporal scales than current synoptic-scale operational models are designed to do. High-resolution atmospheric assimilation systems have the valuable ability to generate mesoscale structures that are forced by processes such as differential surface heat, moisture and momentum fluxes, topographic variations, and latent and radiative heating. The assimilating model should include complete physics in order to simulate mesoscale structures and processes that are not directly observed. Since the lateral boundaries of an analyzed mesoscale domain must frequently be shifted in time to follow the evolution of mesoscale disturbances, the mesoscale assimilation system must be effectively and accurately linked with a coarser-scale (e.g., global and synoptic) model assimilation system. In order to provide ready access by the scientific community, a multicenter geophysical data archive system, electronically linked for maximum effectiveness for assimilated dataset management, transfer, and usage, should be created. The data management activity for the multicenter archiving system will ensure the routine compilation and availability of

OCR for page 61
Four-Dimensional Model Assimilation of Data: A Strategy for the Earth System Sciences observed, model-assimilated, and model-predicted data sets in a structural format jointly adopted and coordinated with the EOSDIS. There is a vital need to archive together with the data sets the modeling codes, information on the input data and how they were processed, and so forth for the scientist to be able to evaluate and understand the archived data sets. Accessibility to archived model-assimilated data sets should be part of a preplanned, service-oriented archive system, including on-line electronic links, low-cost publication media for use on individual workstations, routine inclusion of metadata and software for unpacking and manipulating data sets, and high-quality imaging capabilities. The assembly of a properly formatted and sequenced interdisciplinary assimilated data archive that will span approximately the last 40 years and be extended into the future is a major and crucial task for studies of climate and global change and will require a designated, funded management activity. The archive should consist of data used for operational purposes (e.g., numerical weather prediction) as well as data received in a delayed mode from a variety of national and international sources. The archive should also include model-simulated data that are to be used in scientific and validation studies involving intercomparisons with observed and assimilated data. Assembly of such an archive from a multitude of different data collection points is relatively difficult and time consuming. Therefore, insofar as feasible with respect to efficiency and detail, most data should be collected through the operational system in real time. Delayed data also need to be gathered systematically before they are scattered among too many sub-archives. The goal should be to prepare comprehensive global data sets, with each observational component in a good format and in the sequential order required for long-term analyses. This activity needs to accommodate the diversity of data types, which include rawinsonde data, ship marine data, aircraft reports, cloud drift winds, drifting buoys, expendable bathythermographs (XBTs), satellite retrieval information on atmospheric and sea surface conditions, sea ice coverage, river discharge, and land surface conditions. Data sets should be readily available to users on request from archives. Advanced methods are needed to lower the cost of obtaining data and to facilitate the user's access to the data. Considerable portions of the archived data sets should be ''published'' on media devices such as data acquisition tapes (DATS) or compact disks that can be readily reproduced at low cost. Software routines that access data should accompany the device, and selected browse-and-display routines also should be available. A display option is a converter that costs less than $10,000 and has the ability to prepare video cassette recorder tapes.

OCR for page 61
Four-Dimensional Model Assimilation of Data: A Strategy for the Earth System Sciences Interdisciplinary and discipline graduate education and research opportunities in four-dimensional geophysical data assimilation should be created. To provide essential expertise to cope with the 100-fold increases in observational data from greatly enhanced observing systems in the 1990s and beyond, the development of graduate education, including advanced degree and research opportunities in the fundamental subdiscipline of four-dimensional data assimilation, is needed. Immediate support should be provided for graduate courses and research fellowships in university departments of atmospheric, oceanic, global change, and related sciences. These programs should be explicitly coordinated with related system development, such as EOSDIS, the High Performance Computer and Communications program, and national computer-linked networks. A serious shortage of research scientists with the background, interest, and ability to develop advanced four-dimensional data assimilation methodology and use it in producing high-quality model-assimilated data sets exists at present. The 102 to 103 increase in global data in the 1990s will require at least a 10- to 100-fold increase in data assimilation modeling, the development and utilization of which will require an equal increase in the number of scientists engaged in these efforts. Graduate students and junior researchers need to be made aware of the problems and challenges of four-dimensional data assimilation for the earth science system and be provided with opportunities to develop their abilities in this area. The central importance of data assimilation in advancing understanding in the atmospheric, oceanic, and related sciences must be stated and endorsed through support for graduate courses and research fellowships at universities in order to attract students and young scientists to this crucial effort that seeks to understand climate and global change.