Thursday, August 15, 2013 at 08:01:06
If you haven’t clocked it already, I am involved in organising a workshop on the mapping of glacial landforms this September. Titled GMapping, it is concerned with:
“….the “GMapping Workshop” will present and discuss results from glacial gemorphological mapping by different interpreters for statistically representative synthetic drumlins within a real landscape. This can then inform both the differences/similarities in mapping and quantify the impacts upon the calculation of derived metrics. The key outcomes of this workshop will be the initial development of a set of objective criteria for geomorphological mapping.”
This is a highly topical area at the moment as manual, interpretive, mapping of visually complex landscapes is used extensively throughout the geosciences. And whilst it is preferable to have objective, repeatable, automated techniques, these approaches are still some way off sufficient levels of accuracy. So manual mapping remains the tried and tested approach….. yet comparability of results remains unquantified. Part of the problem is that with a real landscape we cannot a priori know what actually exists meaning it is hard to test the efficacy of individual maps.
One solution to this problem I was involved in with a my colleague John Hillier over in Loughborough….. here we used a real landscape but removed drumlins from it and then inserted our own back in to the landscape. We now have a real landscape with known landforms which means we can test how well individuals are able to identify them and, more importantly, variations in this identification and impacts upon the subsequent calculation of landform metrics.
So the workshop fast approaches and we now have a large set of mapped data (and shortly a preliminary report!) from which discussion can follow - perhaps the most important outcomes of this workshop are twofold:
1. gain a better understanding of mapping error and its impact upon landform metrics
2. development of a protocol for manual mapping to maximise accuracy