Quantifying the present and predicting the past : theory, method, and application of archaeological predictive modeling

post3 head

The volume entitled, “Quantifying the present and predicting the past: theory, method, and application of archaeological predictive modeling” published in 1989 by the U.S. Bureau for Land Management (BLM) is the foundation for all approaches to Archaeological Predictive Modeling (APM).  Past or present, no single volume goes into greater depth on the true underlying issues associated with APM or attempts to synthesize the variety of basic approaches to the data.  I received my first copy from the back of  a BLM warehouse 1999 to 2000; but I have recently run across these scanned copies online.  A PDF is available at this link from the BLM Quantifying the Present and Predicting the Past [PDF] and also an  Alternative Version on Archive.org that allows for download in a variety of formats and is searchable. (the Archive.org link is part of the FEDLINK library and BLM collection that have a ton of great hard to find publications from days gone by)

This most ambitious undertaking was complied and edited by a large team of archaeological All-Stars (Sebastian, Kvamme, Kohler, Altschul, et al.) and sought to identify the fundamental questions the APM should address, the issues that are specific to our data, and methods to implement these models.  This volume is a synthesis of the previous decades worth model based archaeology that had developed out of “New Archaeology” and was directed by the needs of large-are land surveys for the newly formed Cultural Resources Management (CRM) industry. Editorially, I will suggest that this was at the peak of APM study as a central theme that such leaders in the field would devote themselves to.  In the years following, easy access to computer core cycles and push-button GIS made glossing over the details of these methods all to easy.  (I am a guilty contributor to this cycle, so no finger pointing here.)

From the methodological side, the technical details in this volume are not particularly sexy or exciting.  Logistic regression and PCA (Kvamme, Chapter 8) are not nearly as cool to talk about as Machine Learning and Deep Convolutional Neutral Nets, but they are incredibly important to understanding how we attempt to model site locations and are still appropriate models in many situations.  That is not to say that modern statistical learning algorithms are not important of have good performance, but just the notion that you cannot ignore the past (bad pun intended). I has to be recognized that this book was released when personal computers were relatively available andModel Output in ASCII desktop GIS was just getting off the ground. As such, model output that use ASCII characters as indicators of raster quantities was very cool for its day.  Still pretty cool in a retro sort of way.  While the modeling methods may seem outdated, the content in every chapter is well worth the read.  If not only for learning the information the first time, but also for understanding the basis for most of the APM literature from the decades to follow.

For example, many people recite Kvamme’s (pg. 327-328) basic archaeological assumptions for location based models, but wrap them in odd alternative meanings that are not likely intended.  Read this and you will get the original context and better understanding. Speaking of Kvamme, you also get to read his introduction the “Kvamme Gain”; the most widely applied (and mis-applied) measure of an archaeological models predictive ability. (Having studied many of the modern day metrics for classification models I’ll say that for as simple as it is, it is quite a good measure and holds up very well over time)

Kvamme Gain

I recommend reading this whole damn thing cover to cover.  But, say you don’t have a few extra weeks of spare time and are looking for a place to start, I recommend Sebastian’s discussion of correlative and explanatory models (I have MUCH more to say about this, but will hold my tongue for now); dip into Kholer’s contextual discussion in Chapter 2; the intro and model types of Altschul (Chapter 3); hit some high points in Ebert and Kholer; become very cozy with Chapter 5 with Rose and Altschul; read every word of Chapter 7 by Kvamme; and give some serious thought to what Judge and Martin say in Chapter 12.  Chapters 8 through 11 are very interesting, but not as critical as the others.

Advertisements
Quantifying the present and predicting the past : theory, method, and application of archaeological predictive modeling

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s