Surface Interpolations

 

This module was quite challenging and involved the creation of 14 different "maps" (for grad students) that are combined into a sort of poster for demonstrating the power of surface interpolation.  I found these exercises to be extremely practical as I am personally focusing on landscape archaeology and these sorts of statistical exercises and surface interpolations seem certain to factor into my future work.

The first set of maps demonstrates different density and surface interpolation examples on the same set of shovel test data (total artifacts) created by Dr. Palumbo in Panama.  I preferred the result from Kernel Density (with a radius of 50) as it seemed the most straightforward result.  The interpolation surfaces (Kriging, IDW, Spline, and Natural Neighbor) use various mathematical models to create predictive surfaces.

The second poster focuses on data converted from AutoCAD for the Machalilla region.  This was a useful exercise in understanding how to convert common data types to ArcGIS and then manipulate it by using tables and joins.  One complicating factor was that the AutoCAD data has no coordinate system.  The data appears to be in kilometers (probably former 17S UTM data set to a local datum). This allowed me to create a scale bar in kilometers.  We also used the largest polygon in the conversion as the study area boundary and this was used for clipping the rasters.  For reasons inexplicable, I had one clip exhibit "the donut effect" (kernel density) but the other clips do not.

Finally, the grads continued with the Machalilla data to create a series of Inverse Distance Weighting interpolations set to differing powers in an attempt to replicate the Peterson & Drennan (2005) article. The general idea is that social interaction occurs within communities (in this case, in the Regional Development Period) and that clusters occur across the landscape with a (subjective) distance (perhaps of < 2 km) acting as the limiting factor of daily communication between these community clusters. By applying an IDW with various weightings, we can find the smoothing point where community clusters reach this interaction limit and gain some understanding of how communities were manifest on the landscape.  The power of 1 seemed to provide a good view that didn't have too much "steepness" or intensity while also smoothing out to about 2 km clusters.  This seems to show intense community interaction in the southeast corner of the region that seems to sweep in a curve west and north to and up the coast.  These seems to be an isolated region in the middle of this area of interconnectedness.

Finally, the grads worked some stats problems using the t-test.  I liked the Drennan (1996) reading so much, I bought his book.