An Optimized Gravity Terrain Correction Algorithm and Test Results from the Rio Tinto Valley,

The computer calculation of gravity terrain corrections from digital elevation models (DEMs) is a relatively routine task. The conventional approach has been to calculate the gravitational attraction at an observed location by summing the contribution from all DEM nodes. However, even with current high-powered computers, this task can be very time consuming, to the point where very accurate corrections are often not practical. To address this problem, currently available terrain correction algorithms use some sort of optimization strategy at the cost of accuracy. This paper discusses an optimization method that provides a maximum of accuracy and performance for even very large data sets. The approach combines gravity terrain correction methods described by Kane (1962), Nagy (1966), with new grid-mesh interpolation, zoning and de-sampling techniques. Test results from an area of the Rio Tinto valley in Spain are used to illustrate accuracy and performance against a standard terrain correction algorithm from Geosoft. Terrain Correction Algorithm The terrain correction algorithm described here can be run as either a one-step or two-step process. The choice of process depends on the accuracy required, data volume and operational efficiency (specifically, the need to add more stations and dynamically calculate terrain corrections as a survey progresses). Generally, the one-step process can be used to generate highly accurate results with high performance whereas the two-step process two-step process is optimized for large regional gravity surveys (for example, 100,000 or more stations) or ongoing surveys. With the one-step process, the algorithm generates the terrain (regional plus local) corrections based on a usersupplied correction distance. With the two-step method (default), the process consists of calculating a one-time regional correction grid, and then calculating local corrections at each station with the same local correction distance and summing the regional and local correction at each station to obtain a total correction. Because the regional correction grid is only calculated once, this method is designed to save significant computational time for large data sets or surveys in which stations are being added dynamically