Consistency in terrain representations between run-time databases is a prerequisite for interoperability in Distributed Interactive Simulation (DIS). It has been suggested in previous research that one hundred percent alignment of databases will never occur in a simulation that utilizes distributed geometric databases. However, statistical certification of terrain database elevations offers a means of ensuring the degree of consistency necessary for interoperability. In this paper we define a statistical metric for terrain database certification. Starting with a review of the existing work on quantitative terrain database metrics, we examine a basis for specification and statistical certification of terrain elevation data. Using classical acceptance sampling, hypothesis testing will be introduced as a method by which a terrain database (TDB) is certified. A method for determining the critical error value for the desired accuracy proportion and consumers risk (Type II error) will be discussed. From these results the producers risk associated with the test is evaluated for several different accuracy proportions. Using data collected at the 1992 I/ITSEC as a basis for comparison, the utility of acceptance sampling is demonstrated using data collected at the 1994 I/ITSEC. A distinction is drawn between tests designed for TDB certification and tests with inherent diagnostic capability. As an example of the latter, the use of the cross-correlation metric is introduced for the purpose of detecting linear shifts between the terrain skins of a baseline database and a trial database. Using a portion of the Hunter-Liggett high definition area, an example of linear shift detection is provided for the case of a shift by an integer number of samples.