Here's an important notice from The National Geodetic Survey (NGS) concerning the new GEOID12 model:
The National Geodetic Survey has determined that the GEOID12 model was developed using erroneous input control points, primarily in the Gulf Coast region. NGS immediately corrected these errors and developed a replacement model: GEOID12A. GEOID12A is now available for public analysis and comment on the NGS Beta website for GEOID12A and will soon be released for production.
Localized errors were found in parts of Louisiana, Mississippi, Texas, Alabama, Oklahoma, and Canada. Errors in Enid, Oklahoma, and Canada were identified shortly after the August 8, 2012 error notice was posted. After receiving critical input from surveyors, the control data used to develop GEOID12 were immediately re-evaluated, and problematic control stations were identified. Once the problem control stations were corrected, a new model, GEOID12A, was developed. The differences between GEOID12A and GEOID12 are displayed in the map, below. (Click on the map to open a full-size PDF version.) The map shows all areas where GEOID12A differs from GEOID12; in all other areas, the two models are identical (except for the localized errors in Canada).
In most of the areas where GEOID12A differs from GEOID12, the error is less than 10 centimeters, although larger errors were found in four regions containing problem bench marks. The four regions are impacted by bench marks with the following NGS Permanent Identifiers (PIDs): Meridian, MS (PID: CO1081); Alexandria, LA (PID: BX0916); Enid, OK (PID: AF9213); and midway between Lubbock and Fort Worth, TX (PID: DO0454). Corrections applied to heights on all bench marks resulted in the above map.
Although the points were involved in the determination of Vertical Time-Dependent Positioning (VTDP) rates, these points lay outside the area where VTDP was used. Consequently, the originally published leveled values should have been used, rather than the VTDP values. The greatest impact was found at the two bench marks in Meridian and Alexandria (noted above), with sub-decimeter effects for five other points in Louisiana (PIDs: BJ0021, BJ0026, BJ0055, BW0051, and BX0822). An additional 37 VTDP points in Louisiana were dropped (red Xs in the map above), with minimal impact (in the 1 to 2 centimeter range).
The problem point in Texas (PID: DO0454) was due to an inaccurate published ellipsoid height (rather than an incorrect orthometric height). A recent Online Positioning User Service (OPUS) occupation of this point yielded an ellipsoid height that differed by approximately 50 centimeters from its published value. In such cases, preference is normally given to the published value. However, when the GPS dataset used to determine the published ellipsoid height was reprocessed, the resulting value was more consistent with the OPUS solution. So in this one location, for GEOID12A, the OPUS solution was used rather than the published value.
The point in Enid (PID: AF9213) was initially used in GEOID12 to improve coverage in this area. It was removed from GEOID12A because its leveled orthometric height, upon inspection, was deemed unreliable.
Finally, three points in Canada, identified for rejection, were accidentally used in GEOID12 and were removed from GEOID12A. Because these points are localized to Canada—an invalid area for NGS hybrid geoid models—the impact of the errors is negligible, but worth noting in the interest of full disclosure.
While the maximum impact of these error corrections is as much as 40 centimeters, many regions will see similar valid changes when compared to GEOID09. In part, the changes are due to the expansion of the VTDP region west into Texas, and to the east into Alabama and Florida, where VTDP had not previously been implemented. Other changes resulted from inclusion of OPUS-Database points, as well as inclusion of greater terrain signal in the underlying USGG2012 model. These errors went unnoticed, because such changes had been expected in many locations, and NGS will strive to do better in the future. We have placed GEOID12A on the NGS Beta website to ensure both internal and external vetting of the control data and GEOID12A.
The way forward is to allow a few weeks of vetting on the NGS Beta website, where feedback is welcomed. The data will be available to OPUS during this period but will not be used for GPS projects submitted to NGS, until a final GEOID12A model has been vetted and posted to the public NGS website (www.ngs.noaa.gov). The superseded GEOID12 model is available through the interactive tool on the GEOID12 webpage, however the GEOID12 grid files are no longer available for download.
NGS apologizes for the inconvenience caused by these errors. We could not successfully provide our products and services without the support of our customers. We understand the public relies on the integrity of our products, and we thank you for your support. NGS particularly appreciates the input of those users who provided feedback on GEOID12, as their input helped us identify the errors. We will adopt a similar method of using customer input to validate future hybrid geoid models prior to official release.