Background and Methodology for the GED Project

Because of the nature of complex systems, which comprise environment-biosphere interactions, data management, analysis, and synthesis must consider context-dependent factors. In addition to common methods for quantifying and modeling components and processes, we need new ways of capturing the context of whole, functioning systems. This makes data and information management very complicated. It drives technological needs for integrative, relational, and dynamic data structures, comprehensive metadata systems, and new synthesis approaches.

• Features of the GED Project
• Data integration
• Metadata structures
• Characterization and Modeling
• Ecological characterization and synthesis

Features of the Global Ecosystems Database 

• Quality assurance provided through extensive testing and quality control procedures. 
• Thorough documentation and metadata has been prepared for each dataset, in collaboration with the principal investigators. The metadata format precedes but is similiar to current FGDC standards. Full conversion of the metadata format to FGDC and ISO standards is planned for a future release. 
• Scientific Peer Review was conducted involving over 100 scientists for overall design, dataset content, and documentation. The initial peer review comments are summarized with the data base. Each added dataset is individually peer-reviewed.
• The data are completely within the public domain, per agreement by authors of each data set. All "value-added" work performed at NGDC in integrating the data is also in the public domain. Related scientific reprints, reproduced with permission of their independent publishers, may have copyright restrictions.
• Spatial and temporal integration and selectived thematic relationships provide users the ability to compare datasets and combine them in multi-variable processing, modeling, and analysis. 
• Geographic Information System operability is assured through preparation and testing in an operable GIS format created jointly by the GED project and Clark University as part of The Clark University Labs "Idrisi" software. Current versions of Idrisi (4.X, Idrisi for Windows, and Idrisi32) retain backward compatibility with this format. GIS software is not provided with the database. 
• All data are provided in open formats that can be easily accessed for conversion to other formats by varioius means. Format description files are provided with the database.  
• User Feedback. We interact with many of the authors of individual datasets and encourage user feedback to both the Data Center and the original investigators. Comments can be provided by any means, but perhaps most conveniently through the "Comments" link at the top of this page. The feedback that we obtain helps to provide ideas and resources for improving datasets or prioritizing the acquisition or development of future datasets. 

Data Integration

Data integration refers to the process of incorporating data into compatible structures that facilitate their use but also add meaning by defining relationships between data elements. Many levels of integration are possible, from simple database tables to more complicated Geographic Information System structures and Web-enabled technologies. Often intensive quality control, resampling, documentation, and other value-added activities are associated with data integration activities.

Rigorous methods are employed for data integration and quality assessment to produce useful data and information products for the ecosystem sciences. Providing advanced intercomparison and analysis capabilities through the use of scientific Geographic Information Systems is a central focus of these products.

Metadata Structures

Metadata are "data about data" which correspond to the analytical, experimental, or synthetic context of a dataset. Metadata add the semantic content to data. In other words, metadata make data meaningful. Without it, data are nothing but a number series. There are many levels of metadata, from basic information about the units of a variable to documents or reports about the whole dataset or project that produced it. As a practical matter, we generally employ the word "metadata" for that portion of such information that is specifically coded in computer-readable formats, for use in automated query systems (clearinghouse, product interfaces, etc.). As part of our efforts to integrate comprehensive databases for characterization or modeling, we are also involved in defining metadata structures and preparing metadata content. Current efforts aim to adopt and improve the implementation of FGDC/ISO metadata content standards, to develop hierarchical metadata systems that are capable of serving multiple needs, and to prepare metadata in collaboration with the principle investigators for a given dataset and as a routine output of our processing systems. Less structured information about data, such as printed reports and other documents, are considered part of the overall "documentation" of a dataset or product; thus techniques for incorporating both structured and unstructured metadata into documentation are employed.

Characterization and Modeling

The term "characterization" can be used in several contexts. Most common is the reference to characterization of environmental variables in the form of datasets that are generally focused on a specific variable or theme, for example landscape or climate characteristics. We call this environmental characterization to distinguish it from more complex characterizations that attempt to describe an entire ecosystem (see below).

The term "modeling" also has many uses. A fundamental difference in meaning exists between "dynamic" models and "static" models. The former are based primarily on time-dependent equations that are thought to model natural processes (for example, modeling the growth of vegetation under differing CO2 concentrations). The latter are based on observations and associations that are thought to represent important states or coincidences of a system (for example, deriving a drought index from climate data). Most of our work in preparing databases employs static modeling techniques, although databases may contain outputs from both dynamic and static models models along with observational data. The recognition of both static and dynamic modeling as equally important aspects of science, and as fields of study that require different methods and technology, is important to retain balance in scientific methods. Geographic Information Systems technology is appropriate for static modeling, and can be linked to dynamic modeling, or the results of dynamic modeling, in many ways. However, even dynamic models oversimplify complex relationships. A new breed of system model is needed that can deal with mutual and non-linear causation (for example, agent-based models, chaotic models, and perhaps other approaches). Meanwhile, clarifying the role and limitations of static and dynamic models in meeting information needs is extremely important for defining data requirements.

Research is being conducted on Characterization and Modeling methods in collaboration with a number of institutions. Exciting new approaches are being explored for extracting system information from integrated databases using dynamic automated inference techniques. These techniques will support sampling design and the development of environmental and ecological indices through the dynamic generation of predictive maps. These and similar methods will improve our ability to infer system properties from quantitative, observational data.

Ecological Characterization and Synthesis

"Ecological characterization" refers to comprehensive descriptions of ecosystems, which may include environmental characterizations, integrated databases, management information, scenarios and forecasts, models, text descriptions of the ecosystem, and other information. Ecological characterization had its origin in attempts to provide comprehensive and useful descriptions of coastal ecosystems related to management issues, but has since been applied in different forms and in different systems. It is primarily a synthesis of all available information (including operationally proven models) about a system, limited by (a) management issuesvalues and expectations, and (b) present knowledge of ecosystem states and function. Additionally, recent reviews of this approach emphasize the importance of societal factors both in the characterization information and in the preparation process itself. The term "socio-ecological characterization" has been suggested to emphasize this need.

More information is available in: Kineman, J.J., and B.O. Parks, eds. 1996. Ecological characterization: Recommendations of a Science Review Panel. National Oceanic and Atmospheric Administration workshop held March 9-10, 1995. KGRD #32. National Geophysical Data Center, Boulder, CO. 48pp.