Tuesday, September 15, 2009

Choosing good analogs

The use of analogs is common practice in scientific research.  A good analog can be instrumental in understanding concepts, systems or situations that are difficult or imposible to study or analyze directly.  More importantly, a good analog enables to make predictions away from data or observations. A poorly chosen analog, however, leads to false interpretations or conclusions and bad predictions.
As a geologist, I study present-day Earth processes and use them to understand what happend in the past.  It would be great if I can travel back in time, 50 million years ago for example, and witness first hand what was happening then, but it is impossible, hence the need to "approximate" by using analogs.

This is a picture of a Paleocene conglomerate in the Southern Carpathians of Romania.  The outcrops in this area are not great, but through careful analysis of lateral and vertical relationships, sedimentary structures and internal character, I concluded that these rocks were deposited by alluvial fans.   I can only guess what the landscape was 58 million years ago when these rocks were formed, but I can use analogs to make a prediction. 

I envision the alluvial fans in the Romanian landscape of 58 million years ago being similar to the ones that exist today in the Death Valley, California, shown in this landsat picture. This is a good analog but not a perfect one. I can use it to predict the lateral extent and the overlapping nature of the fans, the size and internal architecture but there are major differences too, like tectonic setting and climate, and these have important implications.

An article published recently in the Scientific American shows the impact that an inapropriate analog has to a subsequent model; it also shows how important it is to test a model or hypothesis once it is proposed. 
More than fifty years ago Nobel laureates Alan Hodgkin and Andrew Huxley came up with a model to calculate the power behind electrochemical currents in neurons.  To come up with this model the two scientists used as an analog the brain of a squid and concluded that the energy split in the brain between action potential propagation and synaptic transmission is about 50-50.  It took more than fifty years for new research to show that the split in the human brain is more like 15-85; the problem with the original prediction was primarily the analog used.  A squid's brain has little in common with the human brain (good to know!); by using it as an analog, instead of a mammalian brain, which is what was used for the recent research, one makes the wrong prediction.

From Romanian conglomerates to alluvial fans in Death Valley, squids and the human brain -- I love where geology takes my imagination.

No comments:

Post a Comment

Related Posts Plugin for WordPress, Blogger...