The Terrain Data Fusion System (TDFS) is an application employed to generate a Digital Terrain Data Base (TDB) that will be used as input to various Synthetic Natural Environments (SNE) used in constructive warfare simulations, for the US ARMY and DoD Joint Simulation environments. The TDFS is a collection of Arc/Info Open Development Environment (ODE) tools which, imports a variety of standard National Imagery and Mapping Agency (NIMA) data types, performs operations to fix common problems, co-registers the data layers, and then exports the integrated data into a form that can be used by the SNE generators. The system makes extensive use of reusable ActiveX com objects created with Arc/Info’s ODE

This document describes an application developed by LMIS-ASC to aid in the production of digital Terrain Data Bases (TDB) for the Synthetic Natural Environment (SNE) Development Agent (DA) of the Warfighter Simulation 2000 (WARSIM), and the Joint Simulation System (JSIMS) warfare training simulation systems. The impetus behind this project was to provide an automated method of TDB generation that could be quick and repeatable. The outcome of the project was a more universal solution than the past practice of ad hoc TDB creation for individual simulations by the creation of a geospatial data model and an application to populate the model appropriately.

The process of creating a terrain database is dependent on three key elements: First, a data model is required which meets the requirements of the user community. Second, for the sake of both speed and consistency, an automated system is required to process input data into the aforementioned data model. Finally, source data are required which are sufficiently robust in terms of geospatial detail for the end-use implementation.

In this instance, the data model is the Terrain Common Data Model (TCDM) designed specifically for the Military simulation community. The application is entitled the Terrain Data Fusion System (TDFS) and was developed in Microsoft Visual Basic (VB) as an extension of the Arc/Info Open Development Environment (ODE) in Microsoft Windows NT 4.0. Though highly extensible, the TDFS presently supports the import and conversion of standard NIMA data products, and exports shape files (figure 1.) that are ready for input into the SNE compiler used by Warfighters Simulation 2000 (WARSIM), and the Joint Simulation System (JSIMS).

As the development of the entire TDFS is dependent on the structure and characteristics of the TCDM, the following section begins with a brief discussion of the rationale and role of the TCDM in the design and development of the TDFS. Following the discussion of the TCDM, the TDFS will be detailed with respect to the application itself, the internal data management regime, and various aspects of TDFS functionality.

In the past, TDB creation was slow, usually done by hand on an ad hoc basis, and in a format unique to the simulation being used. Furthermore, because of the complexity of both the input data and the production process, a highly trained staff was required. For future production simulations to be useful to the U. S. Army, however, TDB’s should ideally be made quickly, as warfare simulations may be needed at any time for any part of the globe. The TDFS provides a semi-automated method of TDB construction that may be used by relatively inexperienced personnel on a real-time basis.

In the past, the creation of the digital Terrain Data Base (TDB) was often left up to the developers of the simulation. The organization and content of the TDB was driven by the needs of the hardware and software created for the simulation. Many programs saw only their individual problems and issues, and resolved them internally.

A Process Model was generated to develop the process for the population of a general geospatial data model for describing objects the TDB needed to provide the SNE. Following the development of the process model the Terrain Common Data Model (TCDM) was developed with inputs from the Army user community and software behavior developers to describe those feature objects and their attributes (Kruck et. al. 1998).

A geospatial data model, such as the TDFS, is a description of the data elements, including their attributes as well as their logical relationships to other data members. Each major data type with important or explicit relationships is captured to show its logical relationship to other data types. The strength of the data model is its capability to completely and unambiguously define these relationships (Kruck et. al. 1998).

The Terrain Data Fusion System is the application employed to generate JSIMS and WARSIM Terrain Data Bases (figure 2.). It is a collection of tools which, when used together, produce a digital description of a geographic region. It blends multiple, and sometimes diverse, source data into a cohesive representation. This terrain data base is intended to serve as the baseline data base for the derivation of various Synthetic Natural Environments (SNEs) employed throughout JSIMS/WARSIM. At a minimum, it is anticipated that the data base will serve as an input to the 3-D polygon generation application and the Computer Generated Forces (CGF) real-time compiler.

The TDFS provides several operational capabilities including automated and interactive database production and interactive edit capabilities. The system core is a rules-based system built on top of the Esri product Arc/Info. Programmed primarily in Visual Basic (VB), the system provides a graphical user interface (GUI) and query services in addition to graphical map/data displays.

TDFS is able to process large quantities of geospatial data through numerous intermediate steps. Both vector and raster data types are manipulated, so both coverages and grids are managed. A more efficient method of organization is used to keep track of coverages and grids that are generated, without the overhead of Arc/Info Librarian, or ArcStorm databases.

Keeping track of all of this data is accomplished by organizing all the necessary information about data in a set of Microsoft Access databases (Miller, et. al. 2000). These databases store such information as: coverage locations, names, dataset availability, processing history, and mapping rules. Essentially, the data contained within the various Access databases embody an operational version of the TCDM alongside supporting project status, structural and historical information.

A Vector Library (Miller, et. al. 2000) was developed for vector data. TDFS Vector Libraries share similar naming and tiling structures with Arc\Info Librarian libraries, but function very differently. Vector Libraries are VB based data structures that store information about the libraries in Access databases that are updated and maintained by the TDFS.

TDFS raster data structures are based on the Arc\Info Raster Catalog (Miller, et. al. 2000)), and are called Raster Catalogs. Both images and grids may be stored in Raster Catalogs. This data structure is based closely on the Arc\Info Raster Catalog, but Raster Catalog information is stored in an Access database.

Both of these data management objects are hierarchical collections (figure 3.). Navigating through the spatial data hierarchy provides a simple mechanism for doing repetitive manipulations on large quantities of spatial data. For instance, projecting every coverage in a Vector Library is a simple nested loop that traverses each member of the specified library.

Currently the TDFS imports the National Imagery and Mapping Agency (NIMA) Vector Product Format (VPF) level 0 (VMAP0) and level 1 (VMAP1) vector data (National Imagery Mapping Agency 1990), and Digital Terrain Elevation Data (DTED®) Level 1 (National Imagery Mapping Agency 1988) Digital Elevation Models. In addition, Controlled Image Base (CIB), Arc Digitized Raster Graphics (ADRG) (National Imagery Mapping Agency 1990) and Compressed Arc Digitized Raster Graphics (CADRG) (National Imagery Mapping Agency 1990) image formats are supported.

Native VPF libraries store their spatial data in complex coverages. The VPFIMPORT command in Arc\Info creates complex coverages as output. The TDFS Vector import process creates a Vector Library of complex coverages. These complex coverages are converted into simple coverages to simplify processing.

The first step in processing is creating simple coverages from the VPFIMPORT output. All region and route subclasses were converted into polygon or line coverages. The simple coverage names were the same as the subclass names. The coverages were stored in a workspace with the same name as original complex coverage name. New workspace names correspond to the GeoTile Reference System (GTRS) Specification (Birkel 1999).

Coverages created by the import process are in Geographic coordinates. Mapping features in these coverages into the TCDM data model requires area and length measurements. In addition, all polygon coverages must have an area attribute calculated in m². The TDFS will project entire Vector Libraries into a number of Arc\Info supported projections. Parameters for custom projections, such as Lambert Conformal Conic, are derived at runtime from the geographic extents of the Vector Library being projected.

Controlling the projection process through the TDFS allows precise tracking and management of any projection activity. As multiple vector libraries may be stored for any given database, any number of projections may be utilized with consistent results. Manual errors such as miss-specified projection parameters are eliminated.

Numerous operations must be performed to convert the simplified VPF data directly to the TCDM data model. Features must be selected from the VPF Simple coverages, put into the appropriate TCDM simple coverage, then attribute fields must be created and calculated. Rules for which VPF features become what TCDM features are stored in an Access database. Future changes to the TCDM can be implemented by changing the rules in the database, and don’t require rewriting TDFS.

Some VPF feature types do not map directly to TCDM feature types. In these instances, features are "typecast" to a new feature geometry that is TCDM compliant. An example would be changing a line to a point, or a point to an area. VPF attribute values in conjunction with TCDM specifications are used to determine the shapes and sizes of the new feature types. An example would be clusters of point features corresponding to oil wells. These points are typecast into Oil Field Polygons.

The term deconflict covers numerous operations that ensure that the spatial data has no internal conflicts with other spatial data. Removing these conflicts is necessary to facilitate realistic movement for the simulated troops and equipment. Some of these processes are common GIS operations such as snapping points to lines and ensuring linear continuity (eg. Snapping dams to hydrology, or making sure road networks connect). Other deconflict processes involve more detailed work. Examples of more detailed operations are the creation of surface areals, bridges, and areal feature centerlines.

The surface areal coverage created by the TDFS is similar to a land use/cover map, and provide the core mobility information required to calculate vehicle movement. It is used to define a seamless 3D surface used by the simulations, and no gaps or overlaps in the polygons may exist. The surface areal coverage is created by a three-step process. First, each of 31 possible polygon coverages from the Vmap1 data are checked for overlap with the other 30 coverages. The overlap is then removed based on various logical and spatial criteria. Second, the 31 coverages are combined into one polygon coverage. Third, any gaps, seams or spaces not filled with an area from the 31 coverages are filled with a worldwide soil type coverage.

Bridge creation is also and involved process. Any intersection of hydrology and transportation must have a bridge or a ford in order for simulated forces to cross it correctly. Many bridges are missing in the Vmap1 source data. Also, bridges over larger waterways, lakes, or bays need to broken into individual spans for targeting purposes. A complex series of processes finds each intersection of hydrography and transportation, develops a complex bridge feature and set of spans, and derives attribute information from the associated transportation feature. Derived attributes include such characteristics as lane width, road or rail type, etc.

Simulated force movement may be disrupted by breaks in a linear network. It is common in GIS data for a linear stream coverage to stop and start at a lake, double line river, bay, ocean, etc. To move along these areal water features centerlines are created and connected to the linear stream features ensuring that no gaps occur in the linear features.

Lastly, a number of operations are built into the system to facilitate the reporting of conflicts between the various thematic data groups. As an example, if a building polygon fell within a stream, the TDFS will automatically log the potential inaccuracy and allow the user to interactively correct it.

In addition to the vector Vmap based data that is converted to the TCDM data model, Digital Elevation Models are processed to create 3D terrain. The TDB’s are required to have a three dimensional terrain "skin" to allow other features to be draped over to mimic the real world as close as possible.

Currently the best commonly available unclassified data set with world wide coverage is the DTED level 1 data. Unfortunately, the DTED data can contain numerous artifacts and anomolies which can effect synthetic environments. The TDFS provides users with 3 operations to process the DTED data to remove or mitigate some of the most commonly found errors.

Elevation errors called "cornrows" can occur in DTED Level 1 data. These artifacts are a product of the stereographic method by which DTED’s data are created. The TDFS gives the user a filtering routine based on an algorithm developed by Bardeen (1997) to decrease the amount and severity of the cornrow effect. The filter is useful in mitigating the effect of severe striping seen in large areas of relatively little topographic relief (figure 5.).

Terrain features are generally present in adjacent DTED’s, but often base elevations between the two do not edgematch. Terrain models created in which this anomaly occur, show "cliffs" or "tears" that are not present on the ground. A local filter is run along the edge of each adjacent pair of DTED’s (figure 6.). The filter only changes the values of an area of predetermined width.

DTED has inadequate bathymetry data for the needs of JSIMS/WARSIM. To supplement the otherwise insufficient bathymetric data, areas of Ocean or Sea are replaced with bathymetry (elevation) from ETOPO5 data furnished by the National Oceanographic and Atmospheric Agency (NOAA). A water mask is created with the Ocean coverage. All elevations within the water are taken from resampled ETOPO5. Any other elevation samples come from the DTED. This process is called clamping, and insures that coastlines conform with appropriate bathymetric data.

After the vector data has been converted to the TCDM, and the digital elevations smoothed, the data must be exported in a form that most TDB compilers can digest. Currently the TDFS exports vector data in the Esri Shapefile format, future enhancements will include the ability to export in the VPF format.. Digital elevations can be exported in any of the Arc/Info supported raster conversion formats.

Quality assurance is part of the export process in the TDFS. Coverage attribution is checked thoroughly to make sure values are valid. An Access database contains a set of rules for each TCDM attribute. Any attribute values that are not valid are then calculated based on the best available information. Attribute fields are also checked for compliance with the TCDM.