Free archaeology

[Doug]

Yep, yep and yep. All good points.

Though can't help prodding a bit more.
What constitutes significant archaeology can vary from situation to situation, site to site, region to region.
And yes topsoil, or to be more precise, what is in it can be significant archaeological remains requiring recovery and recording in some areas.

See the EH Research Strategy for Prehistory.......though not sure how widely it has been spread as yet (we got a consultation draft).

Some jobs require mitigation of the loss of worked flint from the topsoil prior to the development going ahead.

As many will know, mathematical modelling is dependent on the basic assumptions of the model (what is a site, what are the factors affecting its location, what errors exist in the data, what skewing factors exist). The proportion of the total dataset entered into any model will dictate its usefulness in predicting......just look at the climatic model for example.

Now I would argue that for archaeology we are still only looking at a small proportion of the total as a start, archaeology is only a proportion of past human activity.
Surviving archaeology represents a small proportion of the former total.
Known archaeology is only a tiny fragment of the surviving archaeology.

You just have to be involved in large projects in areas where DBA's used to come up with nothing, like the Holderness plain in East Yorkshire, South Yorkshire and the 'missing' Iron Age of Durham to get an idea of just how little we know about even simple distributions of most 'sites'.

This is not meant to be insulting Doug, as if you've been studying this for 40 years you must know this. I am fascinated by your claim and want more elucidation, if that is possible.

In my experience of site prediction when for instance being involved in preliminary archaeological works on pipelines, there are rules of thumb that are useful. But people are, and were, contradictory, arsey and stubborn. Sites always seem to crop up where unexpected.

I am interested in how you test your model.

Not insulting in the least bit. PM even after 40 years still has much to be improved upon. Though I have not personally worked on it for 40 years.

Yes- what is important differs from place to place- in southwest US it is 10 lithics within 100m or 3 anything that is datable and all over 50 years of age (not still in use). Here it is what the council arc or EH/HS/Cadw/ etc. says is important with some guidelines.

If you are interested in more I would recommend this publication- https://openaccess.leidenuniv.nl/handle/1887/11863 stroll to the bottom to get the full pdf


Chapter one will give you a good overview and the other chapters cover a lot of your questions.

I would recommend reading the beginning of chapter 7. It illustrates my point made ages ago in this thread- even with predictive modelling being very good you always, always test. The other chapters are fairly good at explaining the different ways PM works in detail.


For interesting work on heritage management and PM-
http://www.assemblage.group.shef.ac.uk/i...inger.html


For the most recent work on the subject this is a good read

http://www.academia.edu/433078/P._Verhag...and_Theory

Overall predictive modelling is basically a deskbased assessment. The only difference is that instead of going to the HR and seeing if anything has been found before in an area and then making a judgement call you follow a set methodology. Not saying what people do with deskdased assessment is flawed just that with PM there are ways to account for problems with the HR that people may not employ on their own deskbased assessment.

Can't iterate this enough- PM can be very very good but it is never 100%. As the Dutch have learned (see that chapter 7) it can not replace actual investigations. It can however sever as a good consulting tool for those in the business of consulting clients about what sort of archaeology to expect. Well, what sort of archaeology they will be required to deal with. As Kevin and Jack point out archaeology is everywhere but not everyone agrees on what it is or what is important.

Jack- some of those resources are some pretty heavy jargon laced reading that I even have trouble getting through at times. Also, it doesn't capture all of the possible ways of doing things. Your question about testing models actually has about 4-5 different ways of doing it. Actually- two main ways but lots of sub-catagories. One is you go out and test your models (can be expensive) e.g. I say you will find site type X here but not there so lets excavate here AND there using a sampling strategy that can then be applied to eerything (people often make the mistake of only looking where they predict things to be- you can see the flaw in that. Very relevant to the geophys talk as well.) or two you look at past data- which involves all sorts of extra steps to eliminate response bias (which are not perfect). If you want I can point you in the direction of good examples. I am afraid that in only a couple of hundred words I am leaving out all the nuances and problems people have found with PM (and that PP finds entertaining) so it is probably not the best explanation of how PM works.

Hope those readings help. If you have anymore questions please fire away.