Terrain Awareness Warning System Databases for the Civil Aviation Industry

Lisa Wynn

Jeppesen, the leading supplier of aeronautical navigation data and charts, now produces elevation databases for the aviation industry. Recent FAA regulations mandate that all aircraft carrying six or more passengers be equipped with Terrain Awareness Warning Systems (TAWS) by 2005. TAWS terrain databases must meet critical data quality standards due to their role in accident avoidance. Jeppesen produces high-quality TAWS databases by employing Workstation ArcInfo with GRID and TIN, ArcGIS, and ArcView with Spatial Analyst to generate, validate, document, and distribute these databases. An overview of TAWS and the Jeppesen database production process will be presented in depth.

Introduction

Jeppesen, headquartered in Englewood, Colorado, has been producing aeronautical information including navigational charts, for civil aviation for more than 60 years. Captain Elrey B. Jeppesen, the company's founder, flew airmail delivery before the days of published aeronautical charts. With no way to know the terrain, flying and landing in poor visibility was risky. Capt. Jeppesen drew his own charts, and when other pilots learned of them, business rapidly expanded. Today, Jeppesen produces visual and instrument charts and other types of navigational and planning information, used by major airlines to private pilots worldwide.

Figure 1. Jeppesen Navigational Chart

Over the years, computerization has made it possible to produce, and for on-board computers to use, new types of aeronautical information. Jeppesen has long utilized both CAD and GIS systems to produce electronic charts and navigation data. Now, in response to flight safety concerns about terrain-related accidents, Jeppesen produces terrain databases for newly mandated Terrain Awareness and Warning Systems (TAWS).

History of CFIT Safety Issues

In the early 1970s, studies began to look into incidents involving controlled flight into terrain (CFIT). CFIT refers to a situation in which pilots in full control of the aircraft fly into terrain, water, or obstacles with no apparent awareness of the danger. These studies found that in many cases installation of Ground Proximity Warning Systems (GPWS) could have helped avoid such incidents. In 1974 the Federal Aviation Administration (FAA) required all large turbine-powered aircraft to be fitted with GPWS equipment. By 1992 this mandate had expanded to include all turbine-powered airplanes with 10 or more passenger seats.

In the last decade, advances in terrain mapping technology have made possible the kind of databases that could give earlier warning of impending flight into terrain. National Transportation and Safety Board (NTSB) investigations into several additional CFIT accidents resulted in recommendations and a final FAA ruling on March 22, 2000. The ruling stated that all turbine-powered aircraft registered in the United States and having 6 or more passenger seats, be equipped with Terrain Awareness Warning Systems (more modern and enhanced versions of GPWS) within 5 years. In fact, all such aircraft rolling off assembly lines after March 29, 2002 must be TAWS-equipped at the time of production.

Figure 2. TAWS Terrain Display

TAWS are forward-looking systems that compare geographical location of the aircraft with the terrain elevation and vertical obstacles within a given search volume of the flight line. They are required to issue various aural and visual alerts and warnings based on conditions such as excessive rate of descent or altitude loss and excessive rate of closure into terrain. They depend upon terrain elevation databases that are loaded into the avionics computer on-board the aircraft, as well as other detection equipment. These databases must be produced under strict quality control standards.

The Jeppesen Terrain Database

Jeppesen has been developing terrain databases for use in TAWS systems over the last several years. The effort involves researching and acquiring elevation databases with the highest quality possible. The spatial resolution, vertical and horizontal accuracy need to be as high as possible, and the processing history must be traced to the source. Detailed metadata are extremely important. These quality issues are raised since the data used in TAWS must be compliant with DO-200A standards, which specify "Standards for Processing Aeronautical Data." That means that TAWS terrain database quality must be traceable all the way back to the original source data.

The Jeppesen Terrain Database is developed specifically to meet requirements of individual avionics manufacturers who build TAWS systems. They in turn must meet the requirements outlined in the FAA Technical Standard Order (TSO-C151a) on "Terrain Awareness and Warning Systems." That TSO recommends certain gridding resolution of terrain data, and requires that the data must have been processed to meet DO-200A standards. The Jeppesen Terrain Database has undergone rigorous development and scrutiny in order to meet the TSO and DO-200A standards.

Jeppesen's Database consists of a worldwide set of gridded elevation values standardized to 30-arcsecond (30as, or approximately 1 kilometer) spatial resolution, also known as post spacing. As will be described below, data have been resampled from the original source data to capture the maximum elevation value at each post. Data are registered to geographic coordinates and to the WGS84 datum, and coordinates are stored in units of decimal seconds. Elevation values are stored in either meters or feet.

The database is developed using two main sources of elevation data: the USGS National Elevation Dataset (NED) (1as, or 30 meters), and the NOAA Global Land One-Kilometer Base Elevation (GLOBE) dataset (30 as). The GLOBE dataset was developed from 18 separate regional elevation databases, and it contains extensive metadata useful in providing traceability. Other data sources which may be used as needed include NIMA DTED Level 0 (30as), USGS GTOPO30 (30as), and other localized higher-resolution data as available.

Jeppesen's Terrain Database Production Methodology

Jeppesen's GIS processing starts with analysis of requirements of TAWS systems, and incorporates analysis of available source data and ArcInfo processing methods. The processing is grouped into five major steps, called "procedures" in DO-200A terminology. These five procedures are discussed below.

Figure 3. Jeppesen Terrain Database GIS Processing

Procedure 1: Procure Raw Terrain Data from Source Data Providers.

Specifications for source data are issued to private data vendors, or are used to procure data from government providers. Once the data are delivered to Jeppesen, the files are loaded onto the system. During loading the files are checked to ensure they meet Jeppesen specifications, and that the data were loaded without corruption. Data providers are notified of any datasets not meeting specifications, so that the correct files may be obtained.

Procedure 2: Convert Data into Usable Format.

The source data files are now processed into 1° x 1° ArcInfo GRID tiles, named according to the lower-left coordinate of the tile. The map projection, coordinate system, datums, and grid cell size (i.e., post spacing) are checked. If the post spacing does not meet original Jeppesen specifications, the originator is notified so that the proper dataset is delivered. If the post spacing is correct, the data are transformed if necessary into the required geographic coordinate system, registered to the WGS84 datum, with coordinates stored in decimal seconds.

Elevation values are stored in floating-point meters. If sea level values are coded as "0" from the source, they must be converted to -500m to distinguish them from legitimate 0-meter elevations on land. This involves constructing binary "land vs. ocean" grids, then comparing the elevation grids with them to determine true ocean-surface grid cells and assign the proper sea-level value.

Visual checks and descriptive statistics are then run on the grids to look for spikes, pits, or other anomalous elevation values. Since DO-200A standards prohibit aeronautical data processors like Jeppesen from changing original source data values, anomalies must be reported to the source provider, who then must make the corrections and deliver new versions to Jeppesen. Once the files are checked and transformed they are moved into a production directory for further processing.

Procedure 3: Process Terrain Data.

The ArcInfo grids are further processed according to avionics customer requirements for output data resolution and tiling schemes. The grids are interpolated from true-surface values into maximum elevation values at required post spacing. In order to construct a dependable maximum elevation grid, the original grids must first be sub-sampled into much smaller cell sizes. ArcInfo's GRID BLOCKMAX algorithm is then applied to the subsampled grids to generate new grids containing maximum elevation values at the specified post spacing.

The results of this process are NED and GLOBE grids resampled and interpolated into maximum elevation grids at 3, 6, 15 and 30-arc-second post spacing, organized into 1° x 1° tiles. These maximum elevation grids are now merged together into 5° x 5° tiles, where GLOBE-derived and NED-derived data are kept separate.

The final step is to merge the GLOBE-derived with the NED-derived data into a continuous worldwide database. For this process, the NED grids are considered the primary grids, and the GLOBE grids are used to fill in the remainder of any 5° x 5° tiles not covered by NED. This usually involves using GLOBE data to fill in open ocean areas not covered by NED. To accomplish this, the MERGE command is constructed such that the GLOBE-based dataset is the last file input, because the last file has the least priority in the MERGE algorithm. The result is that wherever NED-based data are missing in a tile, GLOBE-based data are filled in. Output grids include merged 3, 6, 15 and 30-as grids, which are further processed according to customer specifications.

Procedure 4: Convert to Customer Required Format.

The grids must now be formatted into the customer's specifications for their particular TAWS system. Typical options for elevation data are shown in the following table.

Table 1. Customer Specifications for Jeppesen Terrain Data Specification

Specification	Typical Options
post-spacing	15as or 30as
tile size	1° x 1° or 5° x 5°
elevation value format	integer or floating-point
elevation units	feet or meters
sea-level value	-500 or -9999
NODATA value (there should be none)	-9999 or -32767
output file format	USGS-DEM, USGS-SDTS, IEEE-binary or ASCII, and BIL image format

Customer-formatted data are stored in a client-specific directory on the system, in preparation for writing to the delivery medium. Metadata files are produced and delivered along with the terrain data, in XML format.

Procedure 5: Write to Customer Output Medium and Distribute.

At this stage the data must be compressed and grouped for maximum efficiency of distribution. If the customer works in unix, they may want a .tar file that has been compressed using "gzip" or unix "compress." If they work in Windows NT or 2000, they may want a WinZIP or a "gzip" file. In either case, a 32-bit cyclic redundancy check (CRC) must be run so that the customer knows whether the data have been corrupted in transit. Once the data have been packed using the above methods, the files are written onto CD-ROM for distribution.

Quality Control, Verification and Validation

Quality control of the Jeppesen Terrain Database is an ongoing process throughout the production of the data. There are many steps where the data are either visually checked, or values are compared with known surveyed elevations, or statistical checks are run. Errors that are easy to detect are shown in the figure, below.

Figure 4. Visual Inspection of Jeppesen Terrain Data

The three errors shown in this figure include incorrect values existing in the source data, missing the processing of an individual 5° x 5° tile, and incorrect coding of elevations on land. When these errors are found, the GIS team reviews the procedures and examines individual AMLs to find the source of the error. Any source-data errors are reported back to the originating agency for future correction. Such anomalies, if they cannot be rectified by the source agency before delivery to the customer, are delivered as-is with full documentation for the customer. AML processing steps causing any errors are corrected, and again reviewed by the GIS team. Data are re-processed, and inspected once again before delivery.

The GIS team can run more rigorous testing of the data when high-quality ground survey data can be obtained. This is the case for the United States proper, where National Geodetic Survey (NGS) control benchmarks are obtained. These independent sources of elevation control are correlated with source data such as NED or GLOBE. The correlations are run against true-surface data, and results show how accurate the source data are, how much of the time. Figure 5 shows, for example, that NED terrain data match the NGS data values within 20 meters, close to 95% of the time.

Figure 5. Correlation Between NED Elevations and NGS Benchmark Elevations

Similar correlations can be run against elevations of airport runway endpoints. These are elevations surveyed at the endpoints, or thresholds, of runways at the major airports around the world. This, however, still leaves much of the world's terrain surface to analyze against ground control surveys.

Future Directions

The Jeppesen Terrain Database is designed to be a continually improving database. There is an ongoing need for higher-resolution data meeting rigorous quality standards. There is a great need for ground-control benchmarks for the rest of the world, similar to the NGS benchmarks. Better statistical correlation methods need to be developed to analyze the maximum-elevation values contained in the data output from the Jeppesen Terrain Database. More efficient data storage solutions are being developed to manage this rapidly growing database, which is already nearing one terabyte. The GIS procedures already in place will benefit from more efficient automation, taking advantage of ArcObjects, Java, and XML. The goal will always be to deliver TAWS databases of the highest resolution, accuracy, and quality possible.

References

Special Committee 193. User Requirements for Terrain and Obstacle Data (DO-276). Washington, D.C.: RTCA, Inc., March, 2002.

Special Committee 181. Standards for Processing Aeronautical Data (DO-200A). Washington, D.C.: RTCA, Inc., September, 1998.

Federal Aviation Administration. 14 CFR Parts 19, 121, 135: Terrain Awareness and Warning System. Washington, D.C.: U.S. Department of Transportation, March 29, 2001.

Federal Aviation Administration Aircraft Certification Service. Terrain Awareness and Warning System (Technical Standard Order TSO-C-151a). Washington, D.C.: U.S. Department of Transportation, November, 1999.

Lisa Wynn
GIS Data Analyst
Jeppesen Sanderson
55 Inverness Drive East
Englewood, CO 80112