|The areas in red mark the fossils sites which the ANN unit |
predicted to be present in Wyoming's Great Divide Basin
Long, hard study of geological maps and a hefty chunk of luck was required to find fossils. However scientists from Western Michigan University believe that they have developed a way of locating new localities that will revolutionise Palaeontology.
The team, led by palaeoanthropologist Dr Glenn Conroy, have developed a software model that combines satellite images and maps, both geographical and geological, to accurately pinpoint productive sites within a pre-programmed area. The computer model uses information such as elevation, slope, terrain and other topographical data to predict the locations based upon the same sets of conditions at known fossil bearing formations in the region.
The programme, called ANN (artificial neural network - a series of computer networks that bear similarities to the neuron pathways in a human brain), was taken for a test run to the Great Divide Basin in Wyoming, USA. The ANN unit correctly identified 79% of the area's known fossil formations, and out of all the tagged areas, 99% were fossiliferous. A second trial at the Bison Basin also gave a relatively positive result. 'That gave us encouragement that a blind test based on a neural network for a different basin still gave us pretty good predictive results" says Conroy.
Conroy led the 1991 team which discovered the only known, pre-hominid ape fossil, in South Africa. He hopes to return to the area with the ANN unit to try and find more fossils of such creatures. If the technology proves to be a success and is refined in the future, then many new discoveries may be made with far greater ease. It is possible that the programme may shed light upon the missing links between many animal groups, both modern and extinct.