3D Morphometrics of Ceramic Data (Update)

3D scans of Brazilian ceramics from the Calderon Collection (housed at the Museu de Arqueologia e Etnologia in Salvador, Bahia, Brazil) are being used to refine our approach to morphometrics.


Our revised approach includes an analysis of characteristic ceramic variables using (2D) Elliptic Fourier Analysis (EFA) (using screenshots of the 3D models) and an update to our method used to apply landmarks. We are abandoning the landmarks used in our pilot study with the exception of CB (central base); the single homologous landmark that will be used across the entirety of the assemblage, and eight semi-landmark curves that are draped down the profile of each vessel to bridge the area between what was RI (rim) and CB – creating four complete profiles of each vessel.

Step 1

In order to avoid as much investigator bias as possible, we use Design X to calculate a rotational vector (assumption) to identify the central axis of each vessel. Every subsequent step is built upon this.

Ceramic vessel with vector.

Step 2

After adding the vector, we identify region groups. These are used to create the first plane on either the base or the rim of the vessel (if a rounded base, we use the rim).

Region Groups_Blog
Ceramic vessel with region groups and plane at base.

Step 3

We create a mesh sketch using the vertical silhouette of the vessel (defined by the angle of the base) to ensure that we capture the widest point away from the vector. That mesh sketch is used to extrude a cylinder around the vessel from which deviations (from the vector) can be calculated, ultimately helping us to identify the single point on the vessel farthest from the vector.

Mesh sketch of vertical silhouette.

 Extruded cylinder with high deviations in red and low deviations in blue.

Step 4

Using the vector and the now-defined widest point of the vessel, we apply a single plane (coplanar to the vector) to capture the widest profile. This is used to calculate the positions for four splines (locked to the mesh) across the profile of the vessel.

Ceramic vessel with vector and plane.

Step 5

The plane from Step 4 is then used as the base plane for a mesh sketch. That sketch is extended so that we end up with four equally spaced profiles (starting with the widest profile – all at perfect 45-degree angles), all of which intersect at the vector on the base (CB).


Base of ceramic vessel with vector, plane, and mesh sketch.

Step 6

The first, and only, (quasi) landmark to be applied is always the central base (CB). Subsequently, each spline (profile) is populated with 40 equally-spaced landmarks – beginning with the profile that sits along the previously-defined plane.

Populating Points_Blog
Ceramic vessel with vector, plane, mesh sketch, landmark, and semi-landmarks applied.


For our study of ceramics from Bahia, Brazil, 161 data points were collected from each vessel for the preliminary analysis. Data were exported from Design X, formatted appropriately to include qualitative measures (temper, firing, etc.), then imported to Morphologika for a generalized Procrustes analysis (GPA) and principal components analysis (PCA). Additionally, PCA1 and Centroid Size were used to demonstrate variation in vessel size within the sample.


Results of GPA in Morphologika.

Preliminary morphometric results for ceramic vessels from the Calderon Collection.

Results of the cluster analysis (again, preliminary) point to four possible shape variants in our sample from Brazil. In this initial run, ceramics from the Aratu tradition (at top – click here to learn more) were found to group together, while vessels from the Tupi-Gurarani tradition appear to be more variable in shape.


Written by zselden

My research is focused at the confluence of archaeology, engineering, computer science, and the humanities. I am particularly interested in the application of 3D technologies to archaeological problems, geometric morphometrics, network analyses, archaeological theory, and archaeological science.