Simulation of regional geochemical survey maps at variable sample density

The production of geochemical maps of the world as proposed by the International Geochemical Mapping project (IGM) will require sampling of new areas at low densities and amalgamation of new data with existing geochemical data to ensure map production within reasonable time scales and budgets. Assuming compatibility in sampling and analytical techniques, two important considerations regarding sample density are: (i) how far can sample densities be reduced before meaningful geochemical patterns are lost, and (ii) how can datasets of different sample density be amalgamated? These considerations are examined in the present study by applying two methods of computational data reduction simulating low density sampling to the British Geological Survey's high precision, high resolution ( 1 sample per 1.5 km 2) stream sediment datasets for Northern Britain, exemplified by Ni and B. A series of grids with grid squares of 25 km 2, 100 km 2, 500 km 2 and 160000 km 2 corresponding to sample densities recommended elsewhere for IGM (Table 1 ) are superimposed on the data. For each grid square the data are reduced by (i) selecting a single sample at random, and (ii) calculating the median value. Results are presented as a suite of image processed maps contoured with similar percentile levels enabling comparisons of element distributions to be made. The maps demonstrate that geochemical patterns become distorted at sample densities lower than 1 per 25 km 2 using the random selection method. Random sub-sampling of existing datasets with high sample densities is therefore unlikely to be successful. Employing the median value method, geochemical patterns are maintained with a reasonable degree of accuracy to densities as low as 1 sample per 500 km 2. The optimum reduced sample density for the Northern Britain datasets for Ni and B is 1 per 25 km 2. The size of the geological feature (s) and the magnitude of the geochemical variation are the principle factors controlling the resolution of geochemical patterns at low sample densities. Hence it is unrealistic to recommend an optimum sample density suitable for geochemical mapping throughout the world. Additional factors which influence the choice of sample density include the objectives of the survey (eg. regional reconnaissance, mineral reconnaissance, environmental monitoring), logistical controls on sampling (access, vegetation, climate etc.) and funding constraints. A sample density structure based on grid sizes of 25600 km% 6400 km 2, 1600 km% 400 km 2, 100 km 2 (Garrett, pers. commun., 1992) and 25 km 2 which takes account of these various factors and allows surveys of different sample densities to be related to each other, is therefore proposed for the International Geochemical Mapping project.