Although originally designed to track cloud patterns, weather satellites observe the Earth's surface through gaps in the cloud cover on a daily basis. Receiving stations around the world regularly process such data to give information on the strength of vegetation or the temperature of the ocean. This project has combined data from several of these satellite systems to construct a photorealistic view of the Earth's surface which is readily understood. Areas of forest, mountain, desert and ocean are clearly recognised in their natural colours, offering an alternative to traditional topographic and political maps. The satellite systems used offer frequent global coverage, but do not acquire images in true colour. To simulate natural colours over the land, red and near-infrared wavelengths are used to differentiate land cover types on the basis of their spectral reflectance characteristics. Ground-level, aerial and orbital photography are then used to assign red, green and blue colour values to known feature types, giving realistic land surface colour for anywhere in the world. Ocean colour is derived from a combination of ocean depth and ocean chlorophyll observations. Polar areas covered by sea ice are derived from individual optical images. On land, the courses of rivers are merged in from a digital map database and the appearence of physical features has been enhanced by hill-shading using a digital terrain model. To minimise cloud-cover over the land, it was necessary to process a full year of satellite observations - over 30 gigabytes of input data. Software was written to select the most cloud-free parts from many overlapping images. So far, two man-years of effort have been spent on the Satellite Imagemap and it is continually being improved as new data becomes available. The Satellite Imagemap is held in a geographical coordinate system based on a latitude-longitude grid for easy combination with ancillary data layers, eg, digital terrain, water and political masks, nighttime imagery. Linework such as coastlines, borders, rivers, roads and railways may be directly overlaid without registration errors. The full-resolution Satellite Imagemap, with a pixel size of 1km, is 43,200 by 21,600 pixels - 2.3 gigabytes of data in 24bit colour. Printed as a poster at 150 pixels-per-inch, it would measure 24 feet wide by 12 feet high. The Satellite Imagemap has been used in numerous educational and public information projects for television, exhibitions, print and electronic publication. For details of terms and pricing, please visit the Satellite Imagery website