The quality of paleoecological research depends strongly on site selection. If a researcher is asking a specific question (‘What was the effect of Holocene climate change on the position of treelines in the coastal mountains of BC?’) then they choose sites that are meaningful – sites at or near treeline in the coastal mountains of BC. The current network of paleoecological sample sites (I’m just going to use pollen sites here, but I recognize there are different kinds of sites!) reflects almost 100 years (has it been that long von Post?!) of site selection by researchers for specific research questions.
So what happens when you want to do synthesis work? The sites that have been selected by researchers may not be idea for regional-scale synthesis work:
- Spatial bias is common in the record. My Ph.D thesis was confounded to some degree by the strong coastal bias of pollen cores in BC.
- There is likely to be a bias toward sites where ‘something happens’, either ecologically or climatically, which may or may not be a problem.
- Lots of records have never been added to any sort of large scale database
- Sites have variable taphonomy.
Andrew Gelman has an interesting post up about game theory and incentives to produce high quality research that helps put some of this into a slightly different framework. Sample sites that will best inform large-scale synthesis work are potentially different from the sample sites that will best address individual research questions, but this leads to a conflict in trying to assess which are the best sites. To develop a framework that can help inform future sample site selection, you need to be able to cross-validate the selection procedure, but to cross-validate the procedure you need to be able to assess the contribution of past sample sites so that you have some ability to assign meaning to the sampling statistics
Unfortunately it is difficult to assess the importance of paleo-sites in a meaningful way. A site may be important because it addressed a research question well, and we want to acknowledge that, but it may be unimportant because it is older and may not have adequate dating. As part of PalEON we are struggling to develop framework that will help us site potential new coring locations, while developing metrics with which to evaluate older cores to help us cross-validate the statistical framework.
I suspect that ultimately we may have to settle for an ad hoc method for assigning value to previous cores (something relating 14C dating density and temporal resolution to ecological turnover?) but this isn’t as satisfactory as a uniform metric that can take into account the intrinsic value of a record. But, after all, do individual records have any intrinsic value?
Some music I’m listening to:
Phantom Buffalo – ‘A Hilly Town‘, nice little bass run in the song.
Kashmere Stage Band – ‘Kashmere’, pretty awesome for a bunch of high school kids. Hell, pretty awesome period.